Eventually Bounded Sequences

Bounds That Hold After a Threshold

A sequence may have a finite number of unusually large terms and still have all its later terms within a fixed range. This is the idea behind eventual boundedness: we allow exceptions at the start, but require one bound to work for every term from some index onward. For sequences indexed by \(\mathbb{N}_0\), those finitely many early terms cannot make the whole sequence unbounded. This feature distinguishes sequences from settings where the discarded initial portion might itself contain infinitely many values.

The order of the quantifiers matters: there must be one fixed \(M\) that works for all \(n\geq N\). It is not enough that each individual term has some bound depending on its index. Also, the requirement concerns every term on the tail, not merely infinitely many terms.

Eventual Boundedness and Boundedness

The definition only controls a tail. Nevertheless, a sequence indexed by \(\mathbb{N}_0\) has only finitely many terms before any fixed threshold. Each of those terms is a real number and therefore has a finite absolute value. Taking a bound large enough for both the tail and this finite initial portion gives a bound for the entire sequence.

Proof. Suppose first that \((a_n)\) is eventually bounded. Choose \(N\in\mathbb{N}_0\) and \(M\geq0\) such that \(|a_n|\leq M\) for every \(n\geq N\). If \(N=0\), this already bounds every term. If \(N>0\), the finite set \(\{|a_0|,|a_1|,\ldots,|a_{N-1}|\}\) has a largest element, say \(K\). Set \(B=\max\{M,K\}\). For \(n<N\), \(|a_n|\leq K\leq B\); for \(n\geq N\), \(|a_n|\leq M\leq B\). Thus \(|a_n|\leq B\) for every \(n\), so \((a_n)\) is bounded.

Conversely, if \((a_n)\) is bounded, there is an \(M\geq0\) such that \(|a_n|\leq M\) for every \(n\in\mathbb{N}_0\). The same bound works from \(N=0\) onward, so the sequence is eventually bounded. \(\square\)

This equivalence does not make the word “eventually” meaningless. It identifies exactly what the allowance for finitely many exceptions does in this setting: it lets us verify a bound on a tail, while the finite initial segment can be absorbed into a possibly larger bound. For example, a tail bound of \(5\) need not be a bound for the entire sequence if an early term has absolute value \(100\); it still proves boundedness, but a global bound must account for that term too.

Finite Changes Do Not Affect Boundedness

A related fact concerns two sequences that agree from some point onward. They may have different initial terms, but those differences involve only finitely many indices. The tail controls eventual boundedness, and the theorem above then connects that property to boundedness of the whole sequence.

Proof. Suppose \((a_n)\) is bounded. Then it is eventually bounded by the Tail Boundedness Is Equivalent to Boundedness theorem. Since the sequences agree for \(n\geq M\), the same tail bound for \((a_n)\) applies to \((b_n)\) for every \(n\geq M\). Hence \((b_n)\) is eventually bounded, and the theorem implies that \((b_n)\) is bounded. Interchanging the roles of \(a_n\) and \(b_n\) proves the reverse implication. \(\square\)

Comparing Sequences with a Bounded Difference

Sequences need not agree exactly on a tail for their boundedness to be linked. It is enough for their difference to remain bounded there. The triangle inequality shows that adding a bounded amount to the terms of a bounded sequence cannot make the resulting sequence unbounded. Reversing the roles of the sequences gives an equivalence.

Proof. Suppose \((a_n)\) is bounded. By the Tail Boundedness Is Equivalent to Boundedness theorem, there are \(N_a\in\mathbb{N}_0\) and \(A\geq0\) such that \(|a_n|\leq A\) for every \(n\geq N_a\). Let \(N'=\max\{N,N_a\}\). For every \(n\geq N'\), the triangle inequality gives

Thus \((b_n)\) is eventually bounded, and so it is bounded. The same argument with \(a_n\) and \(b_n\) interchanged proves that boundedness of \((b_n)\) implies boundedness of \((a_n)\). Therefore the two sequences are bounded together or unbounded together. \(\square\)

Consequences and Common Pitfalls

A sequence tending to \(+\infty\) or to \(-\infty\) cannot be eventually bounded. For example, if \(a_n\to+\infty\), then for any proposed tail bound \(M\geq0\), the definition of divergence to \(+\infty\), applied with the real bound \(M+1\), gives an index after which \(a_n>M+1\). In particular, \(|a_n|>M\) on that tail, contradicting the proposed eventual bound. If \(a_n\to-\infty\), apply the definition with the bound \(-M-1\); then eventually \(a_n<-M-1\), so again \(|a_n|>M\). This is consistent with the Unbounded Sequences Diverge theorem: an eventually bounded sequence is bounded, and a bounded sequence cannot diverge to either infinity.

The converse is not true: a sequence that does not tend to either infinity need not be bounded. For instance, \(a_n=(-1)^n n\) has arbitrarily large positive terms at even indices and arbitrarily large negative terms at odd indices, so it is unbounded. It does not tend to \(+\infty\), because its odd-indexed terms are negative, and it does not tend to \(-\infty\), because its even-indexed terms are nonnegative. Thus “not tending to infinity” should not be confused with “eventually bounded.”

The equivalence with boundedness also lets us use earlier results about bounded sequences directly. In particular, if \((a_n)\) is eventually bounded and \(b_n\to0\), then \((a_n)\) is bounded by the theorem above, and the earlier theorem Bounded Sequence Times a Null Sequence gives \(a_nb_n\to0\). No additional bound for each individual term is needed; one bound for the sequence suffices.

When checking eventual boundedness, do not choose a new bound for each later index. The required \(M\) must be fixed. Conversely, do not reject a sequence merely because its early terms are large: a finite number of such terms can be included by enlarging a global bound. Finally, when two tail conditions have different thresholds, use their maximum so both conditions hold at once.