What Is the Root Test?
The root test (also called Cauchy's root test) judges a series Σa(n) by extracting each term's effective geometric rate directly, the nth root of its size:
L = limn→∞ |a(n)|^(1/n)
The verdicts mirror the ratio test's exactly, threshold and all: L < 1 means absolute convergence, L > 1 means divergence, L = 1 says nothing. The intuition: if |a(n)| behaves like rⁿ for large n, then its nth root recovers r, the root test reads off the geometric decay rate that the terms are imitating, then applies the geometric series' |r| < 1 criterion. Terms decaying faster than every geometric series give L = 0; terms refusing geometric decay give L = 1 or more. In every case, the single number L summarizes an infinite tail's behavior.
How to Use This Calculator
Enter the general term in n. The calculator computes |a(n)|^(1/n) at n = 10, 100, 1,000, 10,000, 100,000, estimates the limit, and issues the three-way verdict, plotting the sampled roots against the critical height 1 so the settling is visible. The test shines brightest when the whole term is something raised to the nth power: the default chip (n/(2n+1))ⁿ hands its base straight to the limit, no algebra required beyond reading off the base.
Worked Example
Test Σ(n/(2n+1))ⁿ. Taking the nth root strips the outer power in one stroke:
|a(n)|^(1/n) = n/(2n+1) → 1/2
So L = ½ < 1: the series converges absolutely. Contrast the ratio test on the same series: a(n+1)/a(n) involves comparing (n+1)-th powers of one base against nth powers of another, pages of algebra where the root test needed a single limit. Whenever the n lives in the exponent, roots beat ratios for sheer economy.
Root Test vs. Ratio Test
| Root test | Ratio test | |
|---|---|---|
| Quantity examined | |a(n)|^(1/n) | |a(n+1)/a(n)| |
| Best for | Terms with nth powers | Factorials, products building term-to-term |
| Strength | Strictly stronger (works whenever ratios do, and sometimes more) | Usually easier algebra when it applies |
| Blind spot | L = 1 | L = 1 (the same one) |
A classic case where roots win outright: terms that alternate between two decay rates (say 2⁻ⁿ and 3⁻ⁿ by parity). Consecutive ratios oscillate forever, no limit, while the nth roots settle happily below 1, delivering the verdict the ratio test couldn't.
A good habit when using either geometric-rate test: before computing anything, ask what the term's dominant structure is. An nth power in the term means roots will strip it in one line; a factorial or a product built term by term means ratios will telescope it away. Ten seconds of triage routinely saves a page of algebra.
Common Mistakes to Avoid
- Forgetting the standard limit (1 + 1/n)ⁿ → e. Root-test computations constantly produce this shape; treating it as → 1 is the classic slip. The third chip is built to trip exactly this wire: its roots approach e/3 ≈ 0.906, not 1/3.
- Reading L = 1 as convergence. Same blind spot as the ratio test, every p-series lands there. Hand those to the p-Series Calculator.
- Taking the nth root of the ratio, or the ratio of roots. The test is the nth root of the term itself; hybrid concoctions compute nothing meaningful.
- Ignoring absolute values with signed terms. Odd n makes roots of negative numbers treacherous, the test is defined on |a(n)|, and this calculator applies the absolute value automatically.
Real-World Applications
The root test is the natural instrument wherever growth compounds per step and you want the per-step rate back out. In information theory, the capacity of a channel and the growth rate of typical sequences emerge as nth roots of counting quantities; in finance, the annualized (geometric mean) return of an n-period investment is exactly an nth root, and asking whether it clears 1 is a root-test question about long-run growth versus decay. Population biology's long-term growth rate, the dominant eigenvalue extracted as an nth root of n-generation growth, obeys the same threshold logic.
Within mathematics, the root test underlies the Cauchy–Hadamard formula: the radius of convergence of a power series is exactly 1/lim sup |cₙ|^(1/n), the most general such formula, working even when coefficient ratios misbehave. That makes the root test the theoretical backbone of the entire power-series toolkit, including the Radius of Convergence Calculator even in cases where day-to-day practice reaches for ratios first.
Frequently Asked Questions
Why is the root test "stronger" than the ratio test?
Theorem: whenever the ratio limit exists, the root limit exists and equals it, but the roots can settle when ratios oscillate without a limit (the parity-alternating decay example). So every ratio-test success is a root-test success, not conversely. In exchange, ratios often demand less algebra, which is why both survive in the toolkit.
How does |a(n)|^(1/n) recover a geometric rate?
If |a(n)| = C·rⁿ, then its nth root is r·C^(1/n), and C^(1/n) → 1 for any positive constant, polynomial factors wash out the same way, since n^(k/n) → 1. The root operation is blind to constants and polynomial dressing, seeing only the exponential skeleton: exactly the part that decides convergence.
Where does e sneak into root-test problems?
Through (1 + 1/n)ⁿ → e and its variants, the compound-interest limit. Terms like (1+1/n)^(n²)/3ⁿ have nth root (1+1/n)ⁿ/3 → e/3, and missing the e turns a convergent verdict wrong. It's the single most-tested subtlety in root-test exam problems.
What does L = 0 signify?
Super-geometric decay, the terms shrink faster than rⁿ for every r > 0. Factorial-reciprocal series like Σ1/n! live here (via Stirling's formula, (1/n!)^(1/n) ≈ e/n → 0), converging with room to spare. L = 0 is the strongest possible convergence signal the test can emit.
Why does the plot draw a dashed line at height 1?
It's the critical threshold: sampled roots settling below it certify convergence, above it divergence, and hugging it flags the inconclusive case. Watching where the trajectory lands relative to that line is the whole test, visualized.
Why does the calculator cap its sample indices?
Floating-point honesty: a term like (½)ⁿ underflows to exactly zero past n ≈ 1000, and the nth root of a hard zero is uninformative. Sampling at moderate n keeps the arithmetic inside the representable range while the trend toward L is already unmistakable, a small window into how numerical analysis must respect the machine's limits.