Limit Superior and Limit Inferior

In analysis, sequences don't always converge nicely. They may oscillate, have multiple accumulation points, or behave unpredictably. Limit superior and limit inferior are the essential tools that cut through this complexity, capturing the "worst-case" and "best-case" long-term behavior of a sequence. Mastering these concepts unlocks a deeper understanding of series convergence, provides powerful proof techniques for bounded sequences, and offers precise language for describing asymptotic bounds.

1. Foundational Definitions: Beyond Ordinary Limits

When a sequence $(a_n)$ converges, all its subsequences converge to the same limit. For non-convergent bounded sequences, different subsequences can converge to different values. The limit superior (denoted ${\mathrm{limsup}}_{n\to \infty }{a}_n$ or $\overline{\lim }{a}_n$) is defined as the largest subsequential limit. Conversely, the limit inferior (${\mathrm{liminf}}_{n\to \infty }{a}_n$ or $\underline{\lim }{a}_n$) is the smallest subsequential limit.

A more constructive, set-theoretic definition is often used for calculation and proof. For a bounded sequence $(a_n)$, consider the tail sets $A_N=\{a_n:n\ge N\}$. Let $s_N=\sup A_N$ and $i_N=\inf A_N$. The sequence $(s_N)$ is non-increasing and bounded below, so it converges to its infimum. This limit is the limit superior: $$\underset{n\to \infty }{\mathrm{limsup}}{a}_n=\underset{N\to \infty }{\lim }(\underset{n\ge N}{\sup }{a}_n).$$ Similarly, $(i_N)$ is non-decreasing and bounded above, converging to its supremum, which is the limit inferior: $$\underset{n\to \infty }{\mathrm{liminf}}{a}_n=\underset{N\to \infty }{\lim }(\underset{n\ge N}{\inf }{a}_n).$$ This definition explicitly shows that $\mathrm{liminf}{a}_n\le \mathrm{limsup}{a}_n$, with equality holding if and only if the ordinary limit $\lim a_n$ exists, in which case all three values coincide.

2. Computation and Interpretation

Computing limsup and liminf involves identifying the largest and smallest points that the sequence approaches infinitely often. Consider the oscillating sequence $a_n=(-1{)}^n+\frac{1}{n}$. The subsequence for even $n$ converges to 1, and the subsequence for odd $n$ converges to -1. No subsequence can converge to a value larger than 1 or smaller than -1. Therefore, $\mathrm{limsup}{a}_n=1$ and $\mathrm{liminf}{a}_n=-1$.

For a more complex example, let $a_n=\sin \left(\frac{n\pi }{4}\right)$. This sequence takes values from the set $\{0,\pm 1,\pm \frac{1}{\sqrt{2}}\}$ in a periodic pattern. The supremum of any tail set $A_N$ is always 1, and the infimum is always -1. Thus, $\mathrm{limsup}{a}_n=1$ and $\mathrm{liminf}{a}_n=-1$. The key insight is that these bounds are actually attained by infinitely many terms.

The step-by-step method using the set-theoretic definition is reliable:

1. For each $N$, compute $s_N=\sup \{a_N,a_{N+1},a_{N+2},...\}$.
2. The sequence $(s_N)$ will be non-increasing. Its limit is the $\mathrm{limsup}$.
3. Repeat for $i_N=\inf \{a_N,a_{N+1},...\}$; the limit of this non-decreasing sequence is the $\mathrm{liminf}$.

3. Application: The Root and Ratio Tests for Series

Limit superior and inferior are not just descriptive; they provide the most general forms of powerful convergence tests. For a series $\sum a_n$, consider the Cauchy root test. The standard rule states that if ${\lim }_{n\to \infty }\sqrt[n]{|a_n|}=L<1$, the series converges absolutely. However, this limit may not exist. The definitive version uses the limit superior: Let $L={\mathrm{limsup}}_{n\to \infty }\sqrt[n]{|a_n|}$.

• If $L<1$, the series converges absolutely.
• If $L>1$, the series diverges.
• If $L=1$, the test is inconclusive.

Why does this work? The $\mathrm{limsup}$ captures the "worst-case" growth rate of the terms. If the largest asymptotic growth rate is less than 1, the series is dominated by a convergent geometric series. A similar refinement applies to the ratio test: the conclusive criterion is ${\mathrm{limsup}}_{n\to \infty }\left|\frac{a_{n+1}}{a_n}\right|<1$ for absolute convergence. The ordinary limit is easier to compute when it exists, but the limsup formulation is always valid and strictly more powerful.

4. Role in Theorems and Oscillating Sequences

These concepts are central to several key theorems in real analysis. A fundamental result states: *A bounded sequence $(a_n)$ converges if and only if $\mathrm{limsup}{a}_n=\mathrm{liminf}{a}_n$.* This reframes convergence as the collapse of the "asymptotic envelope" defined by these two values.

For oscillating or irregular sequences, limsup and liminf provide a precise framework for proof. For instance, to prove a sequence is bounded, you might show both its limsup and liminf are finite. In problems involving inequalities, you often encounter statements like "$\mathrm{liminf}{a}_n\ge 0$," which means the sequence does not become permanently negative in the limit; it may dip below zero, but not all its subsequential limits can be negative.

Consider proving that for bounded sequences $(a_n)$ and $(b_n)$, $\mathrm{limsup}(a_n+b_n)\le \mathrm{limsup}{a}_n+\mathrm{limsup}{b}_n$. The proof leverages the set-theoretic definitions: the supremum of a sum of tails is at most the sum of the suprema of the individual tails. This inequality becomes equality if one of the sequences converges, illustrating how limsup interacts with structure.

Common Pitfalls

1. Assuming $\mathrm{limsup}{a}_n$ is the maximum of the sequence. It is not the maximum value the sequence attains, but the largest value that some infinite subsequence approaches. For $a_n=1/n$, the maximum is 1, but $\mathrm{limsup}{a}_n=0$.
2. Believing $\mathrm{limsup}{a}_n=\mathrm{liminf}{a}_n$ implies the sequence is constant. This only implies convergence to a common limit. The sequence $a_n=1/n$ has $\mathrm{limsup}=\mathrm{liminf}=0$, but it is never constant.
3. Forgetting that the definitions require boundedness (or the extended real numbers). For a sequence unbounded above, we define $\mathrm{limsup}{a}_n=+\infty$. For a sequence unbounded below, $\mathrm{liminf}{a}_n=-\infty$. Always check boundedness before applying the standard definitions.
4. Misapplying the inequality $\mathrm{limsup}(a_n+b_n)\le \mathrm{limsup}{a}_n+\mathrm{limsup}{b}_n$ as an equality. This is a frequent error. The equality holds only under additional conditions, such as when one of the sequences converges. A counterexample is $a_n=(-1{)}^n$, $b_n=(-1{)}^{n+1}$. Here, $\mathrm{limsup}{a}_n=\mathrm{limsup}{b}_n=1$, but $\mathrm{limsup}(a_n+b_n)=0$.

Summary

• Limit superior ($\mathrm{limsup}$) and limit inferior ($\mathrm{liminf}$) are defined as the greatest and least subsequential limits of a sequence, respectively. They can be computed as the limits of the sequences of tail suprema and infima: $\mathrm{limsup}{a}_n={\lim }_{N\to \infty }({\sup }_{n\ge N}{a}_n)$.
• A sequence converges if and only if its limit superior and limit inferior are equal. Their values provide tight asymptotic bounds on the sequence's long-term behavior.
• These concepts give the most general form of the root and ratio tests for series convergence, where the decisive criterion is ${\mathrm{limsup}}_{n\to \infty }\sqrt[n]{|a_n|}<1$.
• They are indispensable tools for proving theorems about bounded and oscillating sequences, often simplifying arguments by focusing on extreme asymptotic cases rather than complicated sequence behavior.