When Does a Sequence Converge?
A sequence is an infinite ordered list a(1), a(2), a(3), …, a function on the whole numbers. It converges to a limit L when its terms eventually crowd arbitrarily close to L and stay there: past some index, every term sits within any tolerance you name. Divergence comes in flavors, marching off to ±∞ (like a(n) = n), or oscillating forever without settling (like sin(n), or the alternating (−1)ⁿ).
Sequence limits are the bedrock under everything else in the infinite-process toolkit: a series converges when its partial-sum sequence converges, and the very definition of limits of functions at infinity mirrors this one. Learn to read a sequence's long-run behavior and the rest of the chapter follows.
How to Use This Calculator
Enter the general term in n. The calculator evaluates it at escalating indices, 10, 100, 1,000, up to a million, watches whether the values settle, and declares convergence (with the limit), divergence to infinity, or oscillation. The first forty terms are plotted so you can watch the approach with your own eyes. The chips cover the four canonical behaviors, including the single most important limit in the subject: (1 + 1/n)ⁿ → e, compound interest's road to the exponential.
Worked Example
Find lim (3n² + 1)/(n² + n). Divide top and bottom by n², the highest power:
(3 + 1/n²) / (1 + 1/n) → (3 + 0)/(1 + 0) = 3
The technique, divide by the dominant power, let the leftovers vanish, is identical to limits of functions at infinity and the sampled values (2.99 at n = 100, 2.999 at n = 1000) confirm the approach. For rational terms the shortcut is immediate: equal top and bottom degrees give the leading-coefficient ratio; a bigger bottom gives 0; a bigger top gives ±∞.
Limit Laws and Classic Sequence Limits
| Sequence | Limit | Note |
|---|---|---|
| 1/nᵖ (p > 0) | 0 | polynomial decay |
| rⁿ, |r| < 1 | 0 | geometric decay |
| n^(1/n) | 1 | slow but sure, try the chip |
| (1 + x/n)ⁿ | eˣ | the compound-interest limit |
| c^(1/n) (c > 0) | 1 | roots flatten everything |
| n!/nⁿ | 0 | nⁿ beats n! |
Sums, products, and quotients of convergent sequences converge to the corresponding combinations of their limits (denominators nonzero), the same limit laws as for functions, plus one tool unique to sequences: the monotone convergence theorem, promising a limit to any increasing sequence with a ceiling.
Common Mistakes to Avoid
- Confusing sequence convergence with series convergence. a(n) = 1/n converges (to 0) as a sequence, while Σ1/n diverges as a series. The sequence question is about the terms; the series question about their accumulating sum.
- Assuming bounded means convergent. sin(n) and (−1)ⁿ stay trapped in [−1, 1] forever without settling, bounded oscillation is a genuine third behavior.
- Treating (1 + 1/n)ⁿ as 1^∞ = 1. The base's crawl toward 1 races the exponent's growth, and the photo finish is e ≈ 2.71828, the most consequential indeterminate form in mathematics.
- Trusting early terms. n^(1/n) rises to a peak at n = 3 before its long descent to 1, first impressions mislead, which is why the calculator samples out to a million.
Real-World Applications
Sequence limits describe the long-run destiny of anything evolving in discrete steps. Iterative algorithms are sequences by construction: Newton's method's guesses, gradient descent's parameter updates, and PageRank's successive score vectors each form a sequence whose convergence (and its speed) is the algorithm's entire value proposition, "does it converge?" is a sequence-limit question asked millions of times daily inside software. Fixed-point iteration in numerical computing is the abstract template: a(n+1) = g(a(n)), settling exactly onto solutions of g(x) = x.
In finance, (1 + r/n)ⁿ approaching eʳ is why continuous compounding exists as a concept and a contract type; in population biology, discrete generation models converge to equilibria or cycle according to sequence dynamics; and in statistics, the law of large numbers is a statement that the sequence of sample means converges, the foundational guarantee under every poll, quality-control chart, and Monte Carlo simulation.
Frequently Asked Questions
Can I use L'Hôpital's rule on sequences?
Indirectly: replace n with the continuous variable x, and if the function's limit at infinity exists, the sequence shares it. The route matters for forms like n^(1/n), take logs, resolve ln(x)/x → 0 by L'Hôpital, exponentiate back to 1. The L'Hôpital Calculator handles the continuous middle step.
What is the monotone convergence theorem?
An always-increasing sequence with an upper bound must converge (likewise decreasing with a floor), no need to know the limit in advance. It's the tool for recursively defined sequences like a(n+1) = √(2 + a(n)), where you prove "increasing" and "bounded by 2" separately and conclude convergence, then solve L = √(2 + L) for the limit.
Why doesn't sin(n) have a limit when sin(x)/x-type limits exist?
Because the integers sample the sine wave at points that never repeat and never settle (π being irrational, the samples fill the wave densely), the sequence visits values throughout [−1, 1] forever. Boundedness without settling: the calculator's verdict is "oscillates," and the plot shows the scatter.
How do sequences connect to series tests?
Twice over: a series is the limit of its partial-sum sequence, and the divergence test says a series can't converge unless its term sequence goes to 0. Both connections route through exactly the analysis this calculator performs, see the Partial Sum Calculator for the accumulating side.
What does "convergence rate" mean for iterative methods?
How fast |a(n) − L| shrinks: halving each step (linear), squaring each step (quadratic, like Newton's method), or slower crawls like 1/n. Two convergent sequences can differ enormously in usefulness, rate, not just existence, is what engineering cares about.
What's a subsequence, and why does it matter?
A sequence read along selected indices, every other term, say. If the full sequence converges, every subsequence shares the limit; contrapositively, two subsequences with different limits prove divergence, which is the clean argument for (−1)ⁿ: its even terms sit at 1, odd at −1, so no single limit can exist.
Can a bounded sequence still fail to converge?
Yes. The sequence given by negative one to the power n stays bounded between minus one and one but oscillates forever without settling. Boundedness alone is not enough; a monotonic and bounded sequence, however, always converges by the monotone convergence theorem.