You are on page 1of 2

2.

1 The Covergence of Sequences Convergence: > 0 N N |ana| < indices n N. Definition of Convergence: {an} converges to limit L iff >0, NN nN, we have |anL|<. Lemma 2.9 The Comparison Lemma: Let {an} converge to the number a. Then the sequence {bn} converges to the number b if there is a nonnegative number C and an index N1 such that |bnb| C|ana| indices n N1.(2.4) Proof: Let > 0. Need to find index N |bnb|< indices nN.(2.5) However, we have the estimate (2.4). If C = 0, then the estimate (2.4) that the inequality (2.5) holds with N = N1. So suppose C>0. Since {an} converges to a and the number /C is positive, we can choose an index N2 |ana|</C indices n N2.(2.6) Define N=max(N1,N2). Then, from the estimates (2.4) and (2.6), we have |bnb|C|ana|<C/C=. Theorem 2.10 The Sum Property: Suppose that the {an}a and that {bn}b. Then the sequence of the sums {an+bn}(a+b); that is, lim[an+bn]=liman+limbn. Proof: Let > 0. Want to find index N |(an+bn)(a+b)|< indices nN.(2.7) To do so, we observe that for every index n, |(an+bn) (a+b)|=|(ana)+(bnb)|.(2.8) Since {an}a and /2 is positive, we choose index N1 |ana|</2 indices nN1 and since {bn}b and /2 is positive, we choose index N2 |bnb|<</2 indices nN2. Define N=max(N1,N2). From inequality (2.8) and the choice of N1 and N2, it follws that if nN, then |(an+bn) (a+b)||(ana)|+|(bnb)|</2+/2=. Thus, the required inequality (2.7) holds for this choice of N. Lemma 2.11: Suppose that {an}a. Then, , {an} converges to a. Proof: From the Comparison Lemma, observe that for every index n, |ana|=|||ana||||ana|. Thus, the required comparison inequality (2.4) holds with N1=1 and C=||. Lemma 2.12 Suppose that {an} and {bn}0. Then {anbn} also converges to 0. Proof: Let >0. WTS: index N |anbn|< nN1.(2.9) Since >0 and liman=0, index N1 |an| nN1. Similarly, index N2 |bn|< nN2. Define N=max(N1,N2). Then, if nN, |anbn|=|an||bn|<=. Thus, the required inequality (2.9) holds for this choice of index N. Theorem 2.13 The Product Property: Suppose {an}a and that {bn}b. Then the sequence {anbn}ab. Proof: WTS: lim(anbnab)=0.(2.10) Define, for each index n, n=ana and n=bnb. Then limn=0 and limn=0.(2.11) Moreover, for each index n, since an=a+n and bn=b+n, anbnab=(a+n)(b+n)ab=an+bn+nn. From the limits(2.11), the sum property of convergent sequences and the preceding two lemmas, we conclude that lim(anbnab)=lim(an+bn+nn)=lim(an)+lim(bn)+lim(nn)=alim(n)+blim(n)+lim(nn) Thus, the required limit(2.10) holds. Proposition 2.14: Suppose (bn) is a sequence of nonzero numbers that converges to the nonzero number b. Then the sequence {1/bn}b. Proof: Use the Comparison Lemma. Thus, we will find a nonnegative number C and index N1 |1/bn1/b|C|bn b| nN1. (2.12) But observe that for every index n, |1/bn1/b|=|(bbn)/(bbn)|=1/(|bn||b|)|bnb|. Suppose there is an index N1 |bn| > |b|/2 nN1. (2.13) Then |1/bn1/b|=1/(|bn||b|)|bnb|2/|b|2|bnb| nN1.(2.14) Thus, the required comparison inequality (2.12) holds with N1 chosen above and C = 2/|b|2. The convergence of {1/bn} to 1/b follows from the comparison Lemma. To complete the proof, show that N1 (2.13) holds. Indeed, since the number |b|/2 is positive, we can take =|b|/2 and use the definition of the convergence of a sequence to choose an index N1 so that bn (b, b+) nN. Observe if b>0, then =b/2 (b, b+)=(b/2, 3b/2) and hence bn>b/2 nN1. Since b>0, the inequality (2.13) holds for this choice of N1. On the other hand, if b<0 then =-b/2 (b, b+)=(3b/2, b/2) and hence bn<b/2 nN1. Since b < 0, the inequality (2.13) holds for this choice of N1. Theorem 2.15 The Quotient Property: Suppose {an}a and that {bn}b. Also suppose bn0 indices n and that b0. Then the sequence of quotients {an/bn}a/b. Proof: By the preceding proposition and the product property of convergent sequences, we have lim(an/bn)=lim(an1/bn)=a1/b=a/b. Proposition 2.16 The Linearity Property: ,, lim(an+bn)=liman+limbn. 2.2 Sequences and Sets Bounded: {an} is bounded provided M |an| M for every index n. Theorem 2.18: Every convergent sequence is bounded. Proof: Let {an}a. Take =1, it follows from the definition of convergence that we can select an index N such that |ana|<1 nN. Observe that an=(ana)+a, so by the Triangle Inequality, |an|=|(ana)+a||ana|+|a|. So, by the choice of the index N, |an|1+|a| nN. Define M=max{1+|a|,|a1|,,,|aN1|}. Then |an|M for ever index n. Sequential Denseness of the Rationals: For a set S, we say that {xn} is in the set S provided that for each index n, the term xn belongs to S. Proposition 2.19: A set S is dense in R iff x is the limit of a sequence in S. Proof: Assume S is dense in R. Fix a number x. Let n be an index. By the denseness of S in R, there is a member of S in the interval (x, x+1/n). Choose a member of S that belongs to this interval and label it Sn. This defines {sn} that has the property that |snx|<1/n for every index n. Since {1/n}0, it follows from the Comparison Lemma that {sn} converges to x, and, by the above choice, {sn} is a sequence in S. CONVERSE: Suppose that the set S has the property that every number is the limit of a sequence in S. WTS: S is dense in R. Consider interval (a, b). WTS: (a, b) contains a point of S. Consider the midpoint x=(a+b)/2 of the interval. Define =(ba)/2. Then >0. By the definition of a convergent sequence, index N sn (s, s+) for each index nN. However, (s-,s+)=(a,b). The point sN (a,b). Thus S is dense in R. Theorem 2.20 Sequential Density of Rationals: Every number is the limit of a sequence of rational numbers. Proof: Theorem 1.9 asserts that the set of rational numbers is dense in R. By the preceding proposition, every number is the limit of a sequence of rational numbers. Lemma 2.21 Suppose {dn}d and that dn0 nN. Then d0. Proof: Suppose d<0. Let =-d/2. See that >0 and that d+=d/2<0. So the interval (d,d+) consists of negative numbers, so no terms of {dn} belong to this interval. {dn} cannot converge to d. Contradiction! Thus d0. Closed: Theorem 2.22: Let {cn} be a sequence in [a,b]. If {cn}c, then c also belongs to [a,b]. Proof: By the linearity property of convergent sequences, lim[bcn]=bc. However, for each index n, cn [a,b], so bcn0. The preceding lemma implies that bc0. A similar argument shows that ca0. Thus, c[a,b]. 2.3 The Monotone Convergence Theorem Definition: {an} is said to be monotonically increasing if an+1an for every index n. {an} is said to be monotonically decreasing if an+1an for every index n. A sequence is monotone if either situation is satisfied. Theorem 2.25 The Monotone Convergence Theorem: A monotone sequence converges iff it is bounded. Moreover, the bounded monotone sequence {an} converges to i. sup{an|n in N} if it is monotonically , and to ii. inf{an|n in N} if it is monotonically . Proof: WTS: If monotone sequence {an} is bounded, then it converges to its limit. First suppose {an} is monotonically increasing. Then if we define S={an|nN}. By assumption, the set S is bounded above. According to the Completeness Axiom, S has a least upper bound. Define l=sups. Let >0. WTS: l<an<l+(2.15) nN. Since the number l is an upper bound for the set S, we have anl<l+ for every index n.(2.16) On the other hnd, since l is the least upper bound for S, the number l is not an upper bound for S, so there is an index N l<aNan nN.(2.17) From the inequalities (2.16) and (2.17) follows the required inequality (2.15). Thus {an}l Proposition 2.28: Let c R |c|<1. Then limcn=0. Proof: The case c = 0 is clear since we consider the constant sequence whose terms =0. Moreover, since |cn|=|(c)n|, the case c<0 follows from the case c>0. So assume c>0. Since 0<c<1, {cn} is a monotonically decreasing sequence bounded below by 0. According to MCT, the sequence {cn}l R where l=inf{cn|n in N}. We must have l=0, since otherwise, because c>0, we have cn=cn+1/cl/ic for every index n so that l/c is a lower bound for the sequence. Thus l/c is a lower bound for the sequence, and it is larger than l since 0<c<1. Hence l is not the greatest lower bound for the sequence. This contradiction shows that l=0. Theoreom 2.29 The Nested Interval Theorem: For each n N, let an and bn be numbers an<bn and consider the interval In = [an,bn]. Assume that In+1In for every index n.(2.19) Also assume that lim(bnan)=0.(2.20) Then there is exactly one point x that belongs to the interval In for all n, and both of the sequences {an} and {bn} converge to this point. Proof: Assumption (2.19) means precisely that for every index n, anan+1<bn+1bn. In particular, {an} is bounded above by b1. The MCT implies that {an}a and that ana for every index n. A similar argument shows that {bn}bsuch that bbn for every index n. Thus ana and bbn for every index n.(2.21) From assumption (2.20) and the difference property of convergent sequences, we conclude that 0=lim(bnan)=ba. Thus a=b. Setting x=a=b, it follows from (2.21) that the point x belongs to In nN. There can be only one such point since the existence of two points would contradict (2.20) that the length of the intervals converges to 0.

2.4 The Sequential Compactness Theorem Subsequence: Consider {an}. Let {nk} be a sequence of natural numbers that is strictly increasing; that is, n1 < n2 < . Then the sequence {bn} defined by bk = ansubk for every index k. Preposition 2.30: Let {an}a. Then every subsequence of {an} also converges to the same limit a. Proof: Let >0. Since {an}a, we can say NN nN, we have |ana|<. By definition of subsequence {nk} is strictly increasing kN. Thus, nkk kN (SHOW). So |ansubka|< kN Theorem 2.32: Every sequence has a monotone subsequence. Proof: Consider {an}. We call an index m a peak index for {an} provided that anam for all indices nm. Either there are only finitely many peak indices for {an} or there are infinitely many such indices. Case 1: There are only finitely many peak indices. Then we can choose an index N there are no peak indices > N. We will recursively define a monotonically increasing subsequence of {an}. Indeed, define n1=N+1. Now suppose that k is an index such that positive integers n1<n2<<nk have been chosen ansub1<ansub2<<ansubk. Since nk>N, the index nk is not a peak index. Hence there is an index nk+1>nk ansub(k+1)>ansubk. Thus, we recursively define a strictly increasing sequence of positive integers {nk} having the property that the subsequence {ansubk} is strictly increasing. Case 2: There are infinitely many peak indices. For each natural number k, let nk be the kth peak index. Directly from the definition of peak index it follows that the subsequence {ansubk} is monotonically decreasing. Theorem 2.33: Every bounded sequence has a convergent subsequence. Proof: Let {an} be a bounded sequence. According to Thm 2.32, we can choose a monotone subsequence {ansubk}. Since {an} is bounded, so is its subsequence. Hence {ansubk} is a bounded monotone sequence. By MCT, {ansubk} Sequentially Compact: A set of real numbers S is said to be sequentially comact provided that every sequence {an} in S has a subsequence that converges to a point that belongs to S. Theorem 2.36 The Sequential Compactness Theorem: Let a,b R a<b. Then the interval [a,b] is sequentially compact; that is, every sequence in [a,b] has a subsequence that converges to a point {a,b]. Proof: Let {xn} belong to the closed interval [a,b]. Then {xn} is bounded. By Thm 2.33, subsequence that converges. But thesubsequence is in [a,b] and hence, according to Theorem 2.22, its limit is also in [a,b]. Cauchy Sequence: {an} is said to be Cauchy proved that > 0, index N |an-am| < if n N and m N. Preposition 9.2: Every convergent sequence is Cauchy. Proof: Suppose {an}a. Let >0. Need to find index N |anam|< if n>N and mN. But since {an}a, we can choose an index N |aka|</2 for every index kN. Thus, if nN and mN, setting anam=(ana)=(aam), by the Triangle Inequality, |anam|=|(ana)+(aam)||ana|+|ama|</2+/2=. Lemma 9.3: Every Cauchy sequence is bounded. Proof: Suppose {an} is a Cauchy sequence. For =1, we can choose an index N |anam|<1 if nN and mN. In particular, we have |anaN|<1 if n>N. But, setting an=aN+(anaN), by the Triangle Inequality, |an|=|aN+(an aN)||aN|+|anaN|. Consequently, we see that |an|aN|+1 if nN. Define M=max{|aN|+1, |a1|, |a2|,,|aN1|}. Then |an|M for every index n. Theorem 9.4 The Cauchy Convergence Criterion For Sequences: A sequence converges iff it is Cauchy. Proof: According to Proposition 9.2, every convergent sequence is a Cauchy sequence. CONVERSE: Suppose {an} is Cauchy. Lemma 9.3 asserts a Cauchy sequence is bounded. By the Sequential Compactness Theorem, {an} has a subsequence {ansubk}a. WTS: {an}a. Let >0. Since {an} is Cauchy, we can choose an index N |anam|</2 if nN and mN.(9.1) On the other hand, since {ansubk}a, there is an index N |ansubka|</2 if kK.(9.2) Now choose any index k kK and nkN. Using inequalities (9.1) and (9.2) together with the Triangle Inequality, it follows that if n>N, then |ana|=|(anansubk)+(ansubka)||anansubk Triangle Inequalities: |a+b||a|+|b| and |ab||a||b| Theorem 1.5 The Archimedean Property: i. For any positive number c, there is a natural number n n > c. ii. For any positive number , there is a natural number n 1/n < . 0|x|< x=0 Theorem 1.9: The set of rational numbers is dense in R

1. Suppose {xn} is bounded. Prove (xn/nk}0 as n0 kN. xn/nk can be rewritten as xn(1/nk). Look at {1/nk}. We know {1/nk}0. Regardless of xn being divergent or convergent, because {1/nk} converges to 0, the product of 0 with anything will be 0. Therefore, (xn/nk}0. 2. (a) Suppose {xn} and {yn} converge to the same limit. Prove xnyn0 as n. Let >0. Let {xn}a and {yn}a where LR. By the Sum Property, we can write |(xnL)+(ynL)|<. But we are working with the difference of two sequences so we need to rewrite the Sum Property: |(xnL)[-(ynL)]|<. Simplifying |(xnL)[-(ynL)]|, we get |(xnL)(-yn+L)| and by a commutativity, we get |(xnyn)(LL)|. So |(xnyn)0|<. So (xnyn)0. (b) Show the converse to (a) is false. Let {xn}n and {yn}-n. By the sum property, {anbn}0, but neither {xn} nor {yn} converge. 3. Prove the sequence {(-1)n} nN has no limit. 4. Prove the limit {3+1/n} nN is 3. 5. Suppose xR, xn>0 and xnx as n. Prove that xnx. xnx = (xnx)*[(xn+x)/(xn+x)]=(xnx)/(xn+x). Since xn>0, xn+x>x(xnx)/(xn+x)<(xnx)/x=(xnx)*(1/x). We know lim(xnx)0 so by Product Rule, (xnx)*(1/x)0. So we know xnx0. And by the linearity property, we know xnx 6. In the statement of the nested interval property, the interval must be closed. Show that the NIP might not hold if closed is omitted. If closed is omitted, then it is possible for a sequence to converge to a point outside of the interval. Take {1/n} in the interval (0,2). The sequence is within (0,2), but {1/n} converges to 0, which is outside the interval. 7. Show the NIP might not hold if bounded is omitted. If bounded is omitted, there arent necessarily monotone sequences because bnbn+1 (i.e. intervals wont necessarily be nested.). Also, if you are able to find a monotone sequence, because bounded is omitted, you wont be able to infer convergence. 8. Prove that any real sequence {xn} that satisfies |xnxn+1|<1/2n nN is convergent. Let >0. Let N=1/. Let {xn} be a real sequence |xn+xn+1|1/2n. So -1/2n xnxn+11/2n. We know 2n>n so 1/2n<1/n. So |xnxn+1|<1/n<1/N=1/(1/)= So |xnxn+1|< so |xnxn+1| is convergent. 9. Prove every closed bounded interval is sequentially compact. Assume interval [a,b]R a,bR. Let {sn} [a,b] where {sn} is not necessarily convergent. By Theorem 2.32, a monotone subsequence of {sn}, call it {snsubk}. By problem 2.4.1a, {snsubk} is bounded. So {snsubk} is bounded and monotone. So by MCT, {snsubk} is convergent. By Thm 2.2, we know {snsubk} converges to a point in [a,b] A subsequence of a bounded sequence is bounded. True, given a bounded sequence {an}, M R |an| < M indices n. Look at subsequence |an+1| < m. A subsequence of a monotone sequence is monotone. True, every monotone sequence is increasing or decreasing so a subsequence of a monotone sequence cannot have indices doing the opposite of what the sequence was doing (increasing/decreasing) A subsequence of a convergent sequence is convergent. True, by Theorem 2.33 and that a convergent sequence is bounded, a subsequence is convergent. Sequence converges if it has a convergent subsequence. False, {an}= n when odd and 1/n when even. {an} is divergent. But subsequence {ansubk}=1/n The sum of a monotone sequence is monotone False, {an}=n, {bn}={0,0,-2,-2,-4,}, but {an+bn}={1,2,1,2,} A convergent sequence has a unique limit. Proof: Let >0 and {xn} be a convergent sequence that converges to L and M. By the definition of convergence, we know there exists N such that for all nN |xnL|<1 holds. By the definition, we can make the difference between xn and L as small as we need for large enough n. Therefore, 1 is a positive number. Similarly, we know there exists N1 such that for all nN1, |xnM|<2 is true. For both inequalities to be true, we must find the largest number out of N and N1. Let N=max(N1,N2). Then, for all nN, we have 0|LM|=|(Lxn)+(xnM)||Lxn|+|xnM|. We know |Lxn|<1 and |xnM|<2. So |Lxn|+|xnM|<1+2. So we have 0|LM|<1+2. Let 1=/2 and 2=/2. We know 0|LM|< implies LM=0 so L=M. Prove that the sequence {cn} converges to c iff the sequence {cnc} converges to 0. Proof: ()Given {cn}c, that implies |cnc0|<0 which implies |cnc|<. ()|cnc|0 which implies |cnc0|< which implies |cnc|< which implies cnc. Show that the set (,0] is closed. Let SnS and Sn be in the interval (,0]. lim(0-sn)=0s. By the linearity property of convergent sequences, s0 implies s < 0. 2 1 Show + + 3 converges to 1. n n

Scratch Work: 2 1 + + 3 3 n n

So

2 n

We know n n. 1 1 So . n n 2 1 1 1 2 1 2 Add to both sides of to get + + which is n n n n n n n 3 1 2 equivalent to + n n n

1 + 33 = n

2 n

1 n

Back to 2 1 + n n

2 n =

1 : n 2 n + 1 3 3 3 < n n N n

3 Proof: Let > 0. Let N=/3. Choose n N n > N. By the second part of the Archimedean Property, 2 1 2 1 2 1 2 1 3 3 3 3 + + 3 3 = + 0 = + = + < = = 3 n n n n n N n n n n n 2 1 + + 3 3 n n So N should equal

You might also like