1 Definition

The limit inferior, written \(\liminf\), is a way to describe the eventual lower behavior of a sequence or function. It identifies the greatest lower asymptotic value that remains consistent after ignoring finitely many initial terms. When an ordinary limit exists, the limit inferior agrees with it. When no limit exists, it still provides a meaningful summary of long-run behavior.

1.1 Limit inferior of a sequence

For a real sequence \((a_n)\), the limit inferior is defined by looking at the infimum of the “tails” of the sequence and then taking the limit of those tail infimums. This construction captures the lowest values that continue to appear far out in the sequence.

1.1.1 Tail infimums

For each index \(n\), consider the set \(\{a_k : k \ge n\}\). Its infimum is the smallest lower bound of the tail beginning at \(n\). These tail infimums form a nondecreasing sequence, because each new tail omits earlier terms and can only have an infimum that is at least as large as the previous one. The limit inferior is the limit of this monotone sequence, allowing for an extended real value.

1.1.2 Equivalent characterization by subsequences

The limit inferior can also be described as the smallest number that is the limit of some subsequence. More precisely, it is the infimum of all subsequential limits of the original sequence. This viewpoint is especially useful when the sequence oscillates, since it isolates the lowest long-term limiting behavior among all possible subsequences.

1.2 Limit inferior of a function

For a function, the limit inferior is defined by examining values near a point or far out along the domain. It generalizes the sequence version by replacing tails with punctured neighborhoods or regions beyond large thresholds.

1.2.1 As x approaches a point

If \(f(x)\) is defined near a point \(x_0\), then \(\liminf_{x \to x_0} f(x)\) describes the eventual lower edge of the values of \(f\) in arbitrarily small neighborhoods of \(x_0\). It is often defined as the supremum, over all neighborhoods of \(x_0\), of the infimum of \(f\) on those neighborhoods excluding the point itself when needed.

1.2.2 At infinity

The notation \(\liminf_{x \to \infty} f(x)\) measures the lower asymptotic behavior of \(f\) as the variable becomes large. It is obtained by taking infimums over regions of the form \([A,\infty)\) and then letting \(A\) grow. This is useful for describing functions that do not converge but still have stable lower behavior at large arguments.

1.3 Comparison with limit superior

The limit inferior is paired with the limit superior, or lim sup, which captures the eventual upper edge of a sequence or function. Together, these quantities bound the asymptotic behavior from below and above. If the two are equal, then the ordinary limit exists and equals that common value.

2 Basic properties

The limit inferior has several elementary properties that make it a flexible tool in analysis. Many of these follow directly from its definition through tail infimums.

2.1 Existence and extended real values

For any sequence of extended real numbers, the limit inferior always exists as an extended real number, possibly equal to \(-\infty\) or \(+\infty\). For ordinary real sequences, it may still take an infinite value if the sequence is not bounded below or if it diverges upward in a suitable way. This guarantees that \(\liminf\) is defined in broad generality.

2.2 Monotonicity of tail infimums

The sequence of tail infimums is monotone nondecreasing. This monotonicity is the key structural feature behind the existence of the limit inferior. It means that the lower envelope of later and later parts of the sequence can only move upward or stay fixed.

2.3 Behavior under algebraic operations

Limit inferior behaves in a controlled way under basic arithmetic, though it is not fully linear in general. Its algebraic rules are often stated as inequalities rather than exact equalities.

2.3.1 Addition

For two sequences \((a_n)\) and \((b_n)\), one generally has \[ \liminf (a_n + b_n) \ge \liminf a_n + \liminf b_n. \] This inequality reflects the fact that lower asymptotic behavior of sums cannot fall below the sum of the separate lower asymptotic behaviors. Equality may hold in special cases, such as when both sequences converge.

2.3.2 Scalar multiplication

If \(c\) is a real constant, then multiplying a sequence by \(c\) affects the limit inferior in a predictable way. When \(c \ge 0\), one has \[ \liminf (c a_n) = c\, \liminf a_n. \] For negative \(c\), the order reverses, and the limit inferior is related to the limit superior of the original sequence. This distinction comes from the fact that multiplying by a negative number flips inequalities.

2.4 Relation to ordinary limits

If a sequence converges to a finite limit \(L\), then its limit inferior is also \(L\). In fact, the ordinary limit exists if and only if the limit inferior and limit superior coincide. Thus \(\liminf\) can be seen as a weaker but more robust notion than convergence, preserving useful asymptotic information even when no single limit is present.

3 Characterizations

The limit inferior admits several equivalent descriptions. These formulations are useful in different branches of analysis, depending on whether one is working with sequences, subsequences, or neighborhoods.

3.1 Supremum of tail infimums

For a sequence \((a_n)\), the limit inferior can be written as \[ \liminf_{n\to\infty} a_n = \sup_{n \ge 1}\, \inf_{k \ge n} a_k. \] This formula shows that one first finds the worst-case lower bound for each tail and then chooses the largest such value. It expresses \(\liminf\) as the highest level that the tails eventually stay above in an asymptotic sense.

3.2 Infimum of subsequential limits

Another characterization is that the limit inferior is the infimum of all limits of subsequences. Every subsequence converges, if at all, to a value that cannot lie below the limit inferior. Conversely, there is always at least one subsequence whose limit equals the limit inferior whenever that value is finite and the sequence is bounded below in an appropriate sense.

3.3 Neighborhood-based definition for functions

For a function \(f\) near a point \(x_0\), the limit inferior can be expressed as \[ \liminf_{x\to x_0} f(x)

= \sup_{\delta>0} \inf \{f(x): 0<x-x_0<\delta\}.

\] A similar formula applies at infinity by replacing neighborhoods with intervals \([A,\infty)\). These definitions highlight the local or asymptotic lower envelope of the function values.

4 Examples

Examples help distinguish the limit inferior from the ordinary limit and from the limit superior. They also show how \(\liminf\) behaves for convergent, oscillatory, and unbounded data.

4.1 Convergent sequences

If \(a_n = 1/n\), then \(a_n \to 0\), so \(\liminf a_n = 0\). More generally, every convergent sequence has the same limit inferior as its limit. In such cases, the concept adds no new value numerically, but it remains useful as a stable formulation.

4.2 Oscillating sequences

For \(a_n = (-1)^n\), the sequence alternates between \(1\) and \(-1\). It does not converge, but the limit inferior is \(-1\), since values of \(-1\) occur infinitely often and represent the lower asymptotic edge. Similarly, for a sequence such as \(0,1,0,1,\dots\), the limit inferior is \(0\).

4.3 Bounded and unbounded sequences

A bounded sequence can have a finite limit inferior even when it does not converge. For example, \(a_n = \sin n\) stays between \(-1\) and \(1\), and its limit inferior reflects the lowest accumulation behavior of the sequence. For an unbounded sequence like \(a_n = n\), the limit inferior is \(+\infty\), since the sequence eventually exceeds every fixed lower bound.

5 Applications

The limit inferior appears throughout modern analysis because it converts unstable long-term behavior into a precise numerical invariant.

5.1 Analysis and convergence proofs

In analysis, \(\liminf\) is often used to compare sequences that do not converge outright. It helps establish the existence of subsequential limits, prove compactness-related results, and control estimates where only eventual lower bounds matter. It is especially useful in statements involving inequalities, where exact convergence is unnecessary.

5.2 Measure theory and integration

In measure theory, limit inferior is central to results about pointwise convergence and measurable functions. It appears in the study of almost-everywhere behavior and in comparison theorems for integrals. Because it captures the eventual lower edge of a sequence of functions, it supports rigorous limiting arguments when functions vary irregularly.

5.3 Optimization and asymptotic analysis

In optimization, the limit inferior can describe the long-term best achievable lower value of a cost function along a sequence of approximations. In asymptotic analysis, it is used to characterize lower growth rates and to formulate bounds for algorithms or evolving systems. Its robustness makes it valuable when exact stabilization does not occur.

Several nearby notions are closely connected to the limit inferior and often appear alongside it.

6.1 Limit superior

The limit superior, or lim sup, is the corresponding notion for upper asymptotic behavior. While \(\liminf\) tracks the eventual lower edge, \(\limsup\) tracks the eventual upper edge. Together they bracket the set of subsequential limits.

6.2 Accumulation points

An accumulation point of a sequence is a value approached by some subsequence. The limit inferior is the least accumulation point when the sequence has enough boundedness or compactness properties for such points to exist. This link makes subsequences essential to understanding \(\liminf\).

6.3 Lower semicontinuity

Lower semicontinuity is a property of functions that can be described using limit inferior. A function is lower semicontinuous at a point if its value there is no greater than the limit inferior of nearby values. This relation makes \(\liminf\) a natural tool in topology and variational analysis.