last updated: May 9, 2024

Probability with Measure

2.2 Liminf and limsup

The main difficulty with limits is that, in general, limits do not exist. Most sequences do not converge to anything. However there are two closely related concepts that always exist. They are much easier to work with but they take more care to define.

Let (an) be a sequence of real numbers. Note that the sequence bn=supknak is monotone decreasing, because as n gets larger the set {ak;kn} contains less terms. Lemma 2.1.1 implies that this sequence has a limit, with the caveat that the limit might be infinite. The sequence (bn) is monotone decreasing, so its limit is equal to infnNbn. With this is mind we make the following definition:

(2.3)lim supnan=limn(supknak)=infnN(supknak).

Heuristically, lim supnan is the smallest value that the tail of the sequence (an) stays below.

We can do the same construction the other way up, which gives

(2.4)lim infnan=limn(infknak)=supnN(infknak).

Heuristically, lim infnan is the largest value that the tail of the sequence (an) stays above. Note that (2.3) and (2.4) are always well defined, as extended real numbers, and that lim infnanlim supnan.

  • Example 2.2.1 It is helpful to see a picture:

    (A sequence with its liminf and limsup)

    The sequence displayed is a sample of an=cos(n)+10(1)nnUn, where (Un) are i.i.d. uniform random variables on [0,1]. This is chosen to make a clear picture. As n the cos(n) term will oscillate within [1,1] and the second term will tend to zero. Note that the dotted lines converge downwards to lim supnan=1 (in red) and upwards to lim infnan=1 (in green).

We will spend the rest of this section making connections between lim inf, lim sup and lim.

  • Lemma 2.2.2 Let (an) be a sequence of extended reals. Then:

    • 1. The sequence (an) converges if and only if lim infnan=lim supnan.

    • 2. If (an) converges then lim infnan=lim supnan=limnan.

Proof: Suppose first that (an) converges, say ana. We will consider the case aR here; the case a=± is similar and we will omit it, as discussed in Remark 2.1.2. For all ϵ>0 there exists nN such that |aka|ϵ for all kn. Hence aϵaka+ϵ for all kn, which implies that

aϵinfknaksupknaka+ϵ

Without loss of generality we may choose n1ϵ. Letting ϵ0, upon which n, gives that

a=lim infnan=lim supnan.

We have therefore proved both part 2 and the forwards implication of part 1.

We need to prove the reverse implication from part 1. Suppose that lim infnan=lim supnanR (and we don’t yet know that (an) converges). For all nN we have

(2.5)0aninfknaksupknakinfknak.

Note also that

(2.6)limn(supknakinfknak)=lim supnanlim infnan=0.

Combining (2.5) with (2.6) and using the sandwich rule, we have

(2.7)limn(aninfknak)=0.

We can write an=(aninfknak)+infknak, and we know that both of these terms have a limit as n. The first tends to zero by (2.7) and the second converges to lim infnan. From the algebra of limits we thus obtain that (an) converges and limnan=lim infnan. Since lim infnan=lim supnan was our assumption, this completes the proof.   ∎

  • Lemma 2.2.3 Let (an) be a sequence of extended reals. Then lim supn(an)=lim infnan.

Proof: You already know from real analysis that supn(an)=infnan for real an. This equation also holds for extended reals, but we’ll omit checking the extra cases involving infinities here. The result follows from this along with (2.3) for lim sup, and with (2.4) for lim inf.   ∎