LIM INF AND LIM SUP

JOHN QUIGG

Denition. Let (xn ) be a bounded sequence in R. The lim sup of (xn ) is lim sup xn := lim sup{xn : n k },
k

and the lim inf of (xn ) is lim inf xn := lim inf {xn : n k },
k

The properties of lim inf are mirror images of the properties of lim sup. In fact, we have lim inf xn = lim sup(xn ). To justify the denition of lim sup, dene a new sequence (sk ) by sk = sup{xn : n k }. Then (sk ) is decreasing and bounded, hence converges, and lim sup xn = lim sk . Proposition. Let (xn ) be a bounded sequence in R, and let a R. Then: 1. if a > lim sup xn then there exists k N such that for all n N, if n k then xn < a; 2. if a < lim sup xn then for all k N there exists n k such that xn > a. Similarly for lim inf. Proof. By similarly for lim inf, we mean that if a < lim inf xn then there exists k N such that for all n N, if n k then xn > a, while if a > lim inf xn then for all k N there exists n k such that xn < a. We will give the details for the lim sup, then briey indicate how the corresponding result for the lim inf follows. Dene sk = sup{xn : n k } as above, and put s = lim sup xn . 1. Since sk s and a > s, there exists k N such that sk < a. Since sk = sup{xn : n k }, for all n k we have xn < a. 2. Since sk s, a < s, and k N, there exists j k such that sj > a. Since sj = sup{xn : n j }, there exists n j such that xn > a. Since j k we have n k also.
Date : October 24, 2005.
1

JOHN QUIGG

We have veried the result for the lim sup; we now briey indicate how the corresponding result for the lim inf follows: if a < lim inf xn , then a > lim inf xn = lim sup(xn ), so by what we have already proven there exists k N such that for all n N, if n k then xn < a, hence xn > a, and similarly if a > lim inf xn , then a < lim sup(xn ), so for all k N there exists n k such that xn > a, hence xn < a. QED We will show that the lim sup is the largest subsequential limit (and similarly the lim inf is the smallest). For this we need the following characterization of limits of subsequences: Lemma. Let (xn ) be a sequence in R, and let x R. Then there exists a subsequence of (xn ) which converges to x if and only if for all > 0 and k N there exists n k such that |xn x| < . Proof. First let (yj ) be a subsequence of (xn ) which converges to x. Let > 0 and k N. There exists l N such that for all j N, if j l then |yj x| < . Thus the set {n N : |xn x| < } is innite, hence contains an element n such that n k . Conversely, assume the condition regarding and k . Choose n1 N such that |xn1 x| < 1. Inductively, for each k = 2, 3, . . . , choose nk N such that nk > nk1 and 1 |xnk x| < . k Then (xnk ) is a subsequence of (xn ), and it converges to x by the Squeeze Theorem. QED Proposition. Let (xn ) be a bounded sequence in R. Then: 1. There exists a subsequence of (xn ) which converges to lim sup xn . 2. For all t R, if there exists a subsequence of (xn ) which converges to t, then t lim sup xn . Similarly, lim inf xn is the smallest limit of convergent subsequences of (xn ). Proof. Again, we give the details for the lim sup, then indicate how to handle the lim inf. Put s = lim sup xn . 1. Let > 0 and k N. There exists l N such that for all n N, if n l then xn < s + . Then there exists n max{k, l} such that xn > s , hence |xn s| < . Thus by the preceding lemma there exists a subsequence of (xn ) converging to s 2. We argue by contrapositive: suppose t > s. Let = (t s)/2. Since t > s, we can choose k N such that for all n N, if n k then xn < t , hence |xn t| > . Thus by the preceding lemma there is no subsequence of (xn ) converging to t.

LIM INF AND LIM SUP

We have shown that lim sup xn is the largest limit of convergent subsequences of (xn ); we now briey indicate how to deduce the corresponding result for the lim inf: dene a sequence (yn ) by yn = xn . Then by what we have already proved, there is a subsequence (ynk ) of (yn ) converging to lim sup yn , and then (xnk ) is a subsequence of (xn ) which satises

lim xnk = lim ynk = lim sup yn = lim inf xn ,

and similarly if t R and there exists a subsequence of (xn ) converging to t then the corresponding subsequence of (yn ) converges to t, so we must have t lim sup yn , hence t lim inf xn . QED Corollary. Let (xn ) be a bounded sequence in R. Then (xn ) converges if and only if lim sup xn = lim inf xn .

Proof. First assume that (xn ) converges. Then every subsequence of (xn ) converges to lim xn , so we have lim sup xn = lim xn and lim inf xn = lim xn .

Conversely, assume that lim sup xn = lim inf xn , and let s denote this common value. Let > 0. Since s + > lim sup xn and s < lim inf xn , we can choose k1 , k2 N such that for all n N, if n k1 then xn < s + , and if n k2 then xn > s . Put k = max{k1 , k2 }. Then for all n N, if n k then

s < xn < s + ,

hence |xn s| < . Thus xn x.

QED

Denition. A sequence (xn ) in R is Cauchy if for all > 0 there exists k N such that for all n, j N, if n, j k then |xn xj | < .

JOHN QUIGG

Proposition. A sequence of real numbers converges if and only if it is Cauchy. Proof. Let (xn ) be a sequence in R. First assume that (xn ) converges. Put x = lim xn . Let > 0. Choose k N such that for all n N, if n k then |xn x| < /2. Then for all n, j N, if n, j k then |xn xj | |xn x| + |x xj | < so (xn ) is Cauchy. Conversely, assume (xn ) is Cauchy. Claim: (xn ) is bounded. To see this, note that because (xn ) is Cauchy we can choose k N such that for all n k we have |xn xk | < 1. Put M = max{1 + |xk |, |x1 |, |x2 |, . . . , |xk |}. Then for all n N we have |xn | M , proving the claim. Since (xn ) is bounded, to show that it converges it suces to show that lim inf xn = lim sup xn , and for this, because lim inf xn lim sup xn , it suces to show that for all > 0 we have lim inf xn lim sup xn . Since (xn ) is Cauchy, we can nd k N such that for all n, j k we have |xn xj | < /2. Since lim sup xn /2 < lim sup xn , we can nd j k such that xj > lim sup xn /2. Then for all n k we have xn > xj /2 > lim sup xn . Thus no subsequence could have a limit less than lim sup xn . Therefore lim inf xn lim sup xn . QED Unbounded sequences. Denition. Let (xn ) be a sequence in R. The lim sup of (xn ) is lim sup xn := Similarly for the lim inf. The above denition is consistent with the earlier one for bounded sequences. The properties of lim sup and lim inf for arbitrary sequences are similar to the properties for bounded sequences. limk sup{xn : n k } if (xn ) is bounded above if (xn ) is unbounded above. 2 + 2 = ,