Calculus › Section J · Infinite series › Topic J.1
Sequences
A sequence is an infinite list of numbers with a rule behind it, and the central question is where the list is heading. Sequences are the raw material for infinite series and for every algorithm that homes in on an answer step by step.
Infinite lists with a rule
A sequence is an ordered list of numbers \(a_1, a_2, a_3, \dots\), one for each positive integer. Usually the terms come from a formula, like \(a_n = \frac{1}{n}\), which produces \(1, \frac{1}{2}, \frac{1}{3}, \dots\). The subscript \(n\) is the position in the list. You can also define a sequence recursively, by giving the first term and a rule for getting the next one from it, as in \(a_1 = 4\), \(a_{n+1} = \frac{a_n}{2}\), which produces \(4, 2, 1, \frac{1}{2}, \dots\). Formally a sequence is just a function whose inputs are the positive integers, so everything you know about limits at infinity is about to become useful.
Convergence: where is the list heading?
The one question that matters most is whether the terms settle down. A sequence converges to a number \(L\) if \(a_n\) can be made as close to \(L\) as you like by taking \(n\) large enough, written
The sequence \(\frac{1}{n}\) converges to 0, and \(\frac{n+1}{n}\) converges to 1. A sequence diverges if no such \(L\) exists. Divergence has two flavours: \(a_n = n\) grows without bound, while \(a_n = (-1)^n\), which alternates \(-1, 1, -1, 1, \dots\), stays bounded but oscillates and never settles. Bounded alone does not mean convergent.
Computing limits like a pro
Most sequence limits are computed exactly the way you compute limits as \(x \to \infty\). If the same formula \(f(x)\) has \(\lim_{x \to \infty} f(x) = L\), then the sequence \(a_n = f(n)\) also converges to \(L\). So \(\frac{3n + 1}{2n + 5}\) is handled by comparing leading terms: divide everything by \(n\) and get \(\frac{3}{2}\). Powers behave predictably: \(r^n \to 0\) whenever \(|r| < 1\), so \(0.6^n \to 0\), while \(|r| > 1\) explodes. The bridge to functions also means l'Hôpital's rule can be borrowed, which is how you show \(\frac{\ln n}{n} \to 0\): the function version \(\frac{\ln x}{x}\) has the same limit, and one round of l'Hôpital gives \(\frac{1/x}{1} \to 0\). Logarithms grow, but more slowly than any positive power of \(n\).
Worked example: a famous limit
What happens to \(a_n = \left(1 + \frac{1}{n}\right)^n\)? Computing a few terms: for \(n = 1\) it is 2, for \(n = 10\) about 2.5937, for \(n = 100\) about 2.7048, for \(n = 1{,}000\) about 2.7169. The values climb but the climbs get smaller, which suggests convergence. The limit is the number \(e \approx 2.71828\). This is not a coincidence to memorise blindly: it is where \(e\) comes from in the first place. If \(\$1{,}000\) is invested at a 100% annual rate paid once per year you end with \(\$2{,}000\); paid in \(n\) equal instalments, compounding each time, you end with \(\$1{,}000\left(1 + \frac{1}{n}\right)^n\), which approaches \(\$2{,}718.28\) as the payments get more and more frequent. The sequence does not just converge, it converges to the value of continuous compounding.
Two famous divergent patterns
Not every sequence settles, and it is worth knowing the standard offenders. \(a_n = n\) and \(a_n = n^2\) diverge by growing without bound. \(a_n = (-1)^n\) diverges by oscillation: its odd terms are always -1 and its even terms always 1, so it cannot approach a single number. The oscillating type matters later, because a series built from such terms, like \(1 - 1 + 1 - 1 + \cdots\), also fails to settle. A subtler pattern: \(a_n = \frac{(-1)^n}{n}\) does converge, to 0, because the shrinking \(\frac{1}{n}\) drags the alternation down to nothing. Sign flips do not prevent convergence by themselves; only failing to approach one specific number does.
Bounded and monotonic
Two properties guarantee convergence in advance of any computation. A sequence is increasing if \(a_{n+1} \ge a_n\) for every \(n\), and it is bounded above if some ceiling \(M\) satisfies \(a_n \le M\) for all \(n\) (bounded below is defined symmetrically). The monotone convergence theorem says: a sequence that is increasing and bounded above must converge; a sequence that is decreasing and bounded below must converge. The reason is intuitive: each step climbs at least as high as the last, yet an unclimbable ceiling exists, so the terms must crowd together against it. Note the theorem only needs both properties, not a formula for the limit.
Worked example: how a calculator finds √2
Here is a recursive sequence that earns its keep: start with \(a_1 = 2\) and define \(a_{n+1} = \frac{1}{2}\left(a_n + \frac{2}{a_n}\right)\). The terms are \(2, 1.5, 1.4167, 1.41422, 1.41421\), apparently homing in on \(\sqrt{2} \approx 1.41421\). The monotone convergence theorem proves it. Each new term is the average of \(a_n\) and \(\frac{2}{a_n}\). The average of two positive numbers is at least their geometric mean, here \(\sqrt{a_n \cdot \frac{2}{a_n}} = \sqrt{2}\), so every term from \(a_2\) onward is at least \(\sqrt{2}\). And the average of two unequal numbers lies strictly between them, so whenever \(a_n > \sqrt{2}\), the next term satisfies \(\sqrt{2} \le a_{n+1} < a_n\): the sequence is decreasing and bounded below. If the limit is \(L\), then \(L = \frac{1}{2}\left(L + \frac{2}{L}\right)\), which rearranges to \(L^2 = 2\), so \(L = \sqrt{2}\).
Tails, and what does not matter
Only the long-run behaviour of a sequence matters. If you change, delete or add finitely many terms, the limit is untouched: the sequence \(100, -5, 0, \frac{1}{4}, \frac{1}{5}, \frac{1}{6}, \dots\) still converges to 0, because beyond the altered terms it is the same tail as \(\frac{1}{n}\). This is why the starting index is nothing to fuss over: \(\sum\)-style questions later will care about the \(n\)th term's formula for large \(n\), never about the first handful of values. A sequence can even misbehave for a long time, as long as it eventually behaves.
Common mistakes to avoid
The most common slip is thinking bounded means convergent: \((-1)^n\) is bounded by 1 on both sides and never converges, so the monotone part of the theorem is essential. The second is applying l'Hôpital's rule directly to a sequence: the rule is about functions of a real variable, so go through the matching function \(f(x)\) first. The third is assuming a pattern seen in a table proves a limit: the \(\left(1 + \frac{1}{n}\right)^n\) table suggests \(e\), but the proof is the monotone convergence theorem, not the arithmetic. And keep sequences and series straight: a sequence is a list; a series, coming next, is the running sum of such a list.
Key takeaways
A sequence is a function on the positive integers; it converges when its terms approach one specific number and diverges otherwise, either by growing without bound or by oscillating. Compute limits with the same tools as \(x \to \infty\): leading terms, \(r^n \to 0\) for \(|r| < 1\), l'Hôpital through the function version. The landmark limits are \(\left(1 + \frac{1}{n}\right)^n \to e\) and \(\frac{\ln n}{n} \to 0\). The monotone convergence theorem guarantees convergence for sequences that are increasing and bounded above, or decreasing and bounded below, with no formula needed. Finitely many terms never affect the limit.
Flashcards
Flashcard
An ordered list a₁, a₂, a₃, …, one term per positive integer
Test your understanding
Question 1 of 5
What does the sequence aₙ = (4n + 3)/(2n - 1) converge to?
Question 2 of 5
What does the sequence aₙ = (0.5)ⁿ converge to?
Question 3 of 5
Why does the sequence defined by a₁ = 2 and aₙ₊₁ = (aₙ + 2/aₙ)/2 converge?
Question 4 of 5
What does the sequence aₙ = (1 + 1/n)ⁿ converge to?
Question 5 of 5
Which sequence converges?
Standards: 10 (BC)
