Chapter 1.

2: Properties of Lim inf and Lim sup

The first result we prove shows that if we think of lim inf and lim sup as “squeezing
the sequence as we go to ∞”, then convergence is equivalent to lim inf = lim sup:

Theorem 1.1 Let (xn ) be a real sequence and let l ∈ R. The following are equivalent
(“TFAE”):
(a)
lim xn = l;
n→∞

(b)
lim inf xn = l = lim sup xn .
n→∞ n→∞

Proof: We first show that (a) ⇒ (b): Suppose that limn→∞ xn = l. Since the sequence
(xn ) has a finite limit, it must be bounded both above and below. By Definition 1.1, for
any ε > 0, we can find an integer N = Nε such that

|xn − l| < ε for all n ≥ N.

This is equivalent to the statement that

l − ε < xn < l + ε for all n ≥ N.

Thus, if we let an := inf{xm | m ≥ n} and bn := sup{xm | m ≥ n}, then we have

l − ε ≤ an ≤ bn ≤ l + ε for all n ≥ N.

Since we started with an arbitrary ε > 0, this shows that limn→∞ an = limn→∞ bn = l, so
(b) is true.

To show that (b) ⇒ (a), suppose that lim inf n→∞ xn = lim supn→∞ xn = l. Since l ∈ R
is finite, the sequence (xn ) must be bounded above and below, and by Definitions 1.3 and
1.4 we must have
lim an = lim bn = l
n→∞ n→∞

for (an ) and (bn ) as above. Thus, for any given ε > 0, there exists N = Nε such that

|an − l| < ε and |bn − l| < ε ∀n ≥ N.

In particular, this means l − ε < an and bn < l + ε for all n ≥ N . But by definition we
have an ≤ xn ≤ bn for every n = 1, 2, 3, . . ., and therefore we have l − ε < xn < l + ε
(equivalently, |xn − l| < ε) for all n ≥ N . Again, since ε > 0 was arbitrary, this shows
that limn→∞ xn = l, so (a) is true. (QED)

We have similar results in the case of sequences with infinite limits. First, if the limit
is +∞:

Theorem 1.2 If (xn ) is a real sequence, then TFAE:

(a)
lim xn = ∞;
n→∞

1
 (b)
lim inf xn = ∞.
n→∞

Proof: (a) ⇒ (b): If limn→∞ xn = ∞, it means that for any K ∈ R, there exists an integer
N = NK such that
xn ≥ K for all n ≥ N.
Thus, the sequence (xn ) must be bounded below, and for n ≥ N we have

an := inf{xn , xn+1 , xn+2 , . . .} ≥ K.

Since this can be done for K ∈ R arbitrary, we have limn→∞ an = ∞, which shows that
(b) holds.

(b) ⇒ (a): Conversely, if lim inf n→∞ xn = ∞, then using Definition 1.4 we must have (xn )
bounded below and limn→∞ an = ∞ for an := inf{xm | m ≥ n}. But this means that for
any K ∈ R it is possible to find N = NK such that an ≥ K for all n ≥ N . Since an ≤ xn
by definition of an , this implies that xn ≥ K for all n ≥ N , which means limn→∞ xn = ∞,
i.e. (a) holds. (QED)

Note that since we always have an ≤ bn , for any sequence (xn ),

lim inf xn = ∞ ⇒ lim sup xn = ∞.

n→∞ n→∞

Hence, Theorem 1.2 tells us that if the limit of a sequence is +∞, then its limit superior
is also +∞. The converse to this is not true, in general: there are many sequences with
limit superior +∞ which do not have limit +∞. Spend some time trying to think of some
examples if this is not clear to you. For sequences with limit −∞, we have:

Theorem 1.3 If (xn ) is a real sequence, then TFAE:

(a)
lim xn = −∞;
n→∞

(b)
lim sup xn = −∞.
n→∞

(The proof is very similar to the proof of Theorem 1.2, and is left as an exercise.)

Example 1.2 Find lim inf and lim sup for the following sequences:
(−1)n
(a) xn = 1 + n
;
(b) xn = 2n.

The following theorem gives a kind of “ε − N criteria” for lim inf and lim sup:

Theorem 1.4 Let (xn ) be a real sequence, and let l, L ∈ R be fixed real numbers.

2
(I) TFAE:

(I-a)
lim sup xn = L;
n→∞

(I-b) For any ε > 0, we have: (i) There exists Nε such that xn < L + ε for all n ≥ Nε ;
and (ii) Given any m ≥ 1, there exists nε,m ≥ m such that xnε,m > L − ε.

(II) TFAE:

(II-a)
lim inf xn = l;
n→∞

(II-b) For any ε > 0, we have: (i) There exists Nε such that xn > l − ε for all n ≥ Nε ;
and (ii) Given any m ≥ 1, there exists nε,m ≥ m such that xnε,m < l + ε.

Proof: We give only the proof that (I-a) and (I-b) are equivalent. The proof that (II-a)
is equivalent to (II-b) is very similar, and should be done as an exercise to test your
understanding of the proof given here.

(I-a) ⇒ (I-b): Suppose lim supn→∞ xn = L. Since L is finite, the sequence (xn ) must be
bounded and

lim bn = L (1)
n→∞

for bn := sup{xm | m ≥ n}. Let ε > 0 be given to us. By (1), we can find an integer
N = Nε such that

|bn − L| < ε for all n ≥ N. (2)

Condition (2) means first of all that bn < L+ε for all n ≥ N , which implies that xn < L+ε
for all n ≥ N , as xn ≤ bn . Hence we have shown that condition (i) of (I-b) holds. For
condition (ii) of (I-b), let m ≥ 1 be a given integer. Let k := max{N, m}. Since k ≥ N ,
by (2) we must have bk > L − ε. But since bk := sup{xn | n ≥ k}, this can only be true
if there is some integer nε,m ≥ k such that xnε,m > L − ε (otherwise, the real number
M := L − ε/2 would be an upper bound of the set {xn | n ≥ k} that is less than bk , which
contradicts bk = sup{xn | n ≥ k}).

(I-b) ⇒ (I-a): Suppose (I-b) holds. By condition (i), (xn ) must be bounded above,
and hence lim supn→∞ xn = limn→∞ bn for bn as above. To show that limn→∞ = L, let
ε > 0 be an arbitrary positive real number given to us. By condition (i) of (I-b), we
can find an integer N = Nε such that xn < L + ε for all n ≥ N . By definition of
bn := sup{xm | m ≥ n}, this means that bn ≤ L + ε for all n ≥ N . Now, if n ≥ N is any
integer, let m = n and use condition (ii): this tells us that there exists nε,n ≥ n such that
xnε,n > L − ε. Since nε,n ≥ n, this implies that bn > L − ε (otherwise, bn would not be
an upper bound of the set {xk | k ≥ n}). Thus, we have shown that |bn − L| ≤ ε for all
n ≥ N , so since ε > 0 was arbitrary, limn→∞ bn = L, which shows that (I-a) holds. (QED)

3
Notation Be aware that other books/notes refer to the lim sup as the “upper limit” of a
sequence, and some use the notation

lim xn .
n→∞

Similarly, lim inf is sometimes called the “lower limit” and denoted by

lim xn .
n→∞