What Is the Alternating Series Test?
An alternating series flips sign every term: b₁ − b₂ + b₃ − b₄ + ⋯, with all b(n) positive. The alternating series test (Leibniz's test) says such a series converges if just two things hold:
- The b(n) are (eventually) decreasing each positive part no larger than the last, and
- b(n) → 0 the parts shrink away entirely.
That's a remarkably light requirement, no geometric decay, no p > 1, because alternation does the heavy lifting. Each partial sum overshoots the target, the next undershoots, and the zigzag corridor narrows as the b(n) shrink, trapping the sum in between. The corridor picture also delivers the test's famous bonus, the error bound: stopping after N terms leaves an error no larger than b(N+1), the first term you omitted.
How to Use This Calculator
Enter the positive part b(n), the calculator supplies the alternating signs. It checks both Leibniz conditions numerically (monotonicity across 200 terms, and the large-n limit), declares the verdict, and, when the test passes, sums 60 terms with the zigzag plotted, converging corridor and all. The failure chip (b(n) → 1) shows the other side: when the parts don't vanish, the partial sums bounce forever between two levels and the series diverges by the divergence test.
Worked Example: the Alternating Harmonic Series
Take b(n) = 1/n: the series 1 − ½ + ⅓ − ¼ + ⋯. Both conditions hold (1/n decreases to 0), so it converges, to ln 2 ≈ 0.6931 one of the most quotable results in the subject. The contrast deserves a pause: the plain harmonic series Σ1/n diverges yet merely alternating its signs tames it. Convergence here is entirely the alternation's doing, the cancellations, not the term sizes, carry the day. Series like this (convergent, but not absolutely) are called conditionally convergent.
The Error Bound in Action
The first-omitted-term bound |S − SN| ≤ b(N+1) is unusually practical, most convergence tests certify convergence without saying how close a partial sum is. For the alternating harmonic series, stopping at N = 100 guarantees the answer within 1/101 < 0.01; wanting three decimals demands b(N+1) ≤ 0.0005, so about 2000 terms. The Leibniz π/4 chip shows the cost of slow decay vividly: computing π this way to six digits would need half a million terms, historically important, computationally hopeless, and a perfect illustration of why convergence rate matters as much as convergence.
Common Mistakes to Avoid
- Checking only b(n) → 0. The decrease condition matters: cunning non-monotonic examples converge or diverge unpredictably without it. Both boxes must tick.
- Applying the error bound to non-alternating series. The first-omitted-term guarantee is a gift of the zigzag geometry, general series' tails can dwarf their next term.
- Confusing conditional with absolute convergence. The alternating harmonic series converges, but the absolute-value version diverges. Absolutely convergent series (like alternating 1/n²) are sturdier, rearrangement-proof, while conditional ones are famously fragile.
- Missing that b(n) must eventually decrease, not always from n = 1. Early misbehavior is harmless; the test cares about the tail. This calculator checks a long stretch and says which condition failed.
Real-World Applications
Alternating structure appears wherever effects push and pull in turns. In signal processing, Fourier series of square and sawtooth waves are alternating series whose partial sums produce the famous Gibbs overshoot, the zigzag corridor made audible in ringing artifacts. Numerical libraries exploit the error bound directly: evaluating sin or cos by their alternating Maclaurin series, the code stops the moment the next term dips below the precision target, with the Leibniz bound as the certificate, one of the few situations in computing where the stopping rule is provably safe and free.
In physics, alternating series compute lattice sums like the Madelung constant of a salt crystal (alternating charges at increasing distances), and perturbation expansions frequently alternate, with the first-omitted-term heuristic guiding where to truncate. The famous conditional-convergence pathology has a physical echo too: rearranging a conditionally convergent lattice sum genuinely changes the answer, which is why crystal-energy calculations must fix a physically meaningful summation order, Riemann's rearrangement theorem enforcing discipline on chemistry.
Frequently Asked Questions
What is Riemann's rearrangement theorem?
For conditionally convergent series, reordering the terms can change the sum, to any target value, or to divergence. The alternating harmonic series can be rearranged to sum to 5, or −π, or ∞. Absolutely convergent series are immune: their sums survive any reordering. It's the sharpest illustration that infinite addition is not commutative by default.
Why does the alternating harmonic series sum to ln 2?
Substitute x = 1 into the Maclaurin series ln(1+x) = x − x²/2 + x³/3 − ⋯, which (by Abel's theorem) is valid at that endpoint. The Maclaurin Series Calculator generates the parent expansion.
What's the Leibniz series for π?
1 − ⅓ + ⅕ − ⅐ + ⋯ = π/4, from the arctangent series at x = 1, the second chip. Beautiful and impractically slow: the error after N terms is about 1/(2N), so each additional digit of π costs ten times the terms.
Does alternating always help convergence?
It never hurts and often rescues: if Σb(n) already converges absolutely, alternation keeps that; if b(n) → 0 monotonically but Σb(n) diverges (like 1/n), alternation saves it. Only when b(n) fails to approach 0 is the case hopeless, no sign pattern can fix terms that won't shrink.
How does the calculator verify the decreasing condition?
It scans 200 consecutive terms for any increase, a numerical stand-in for the analytic check (showing b(n+1) ≤ b(n), or that the continuous version has negative derivative). For a rigorous proof on exam paper, the derivative route via the Derivative Calculator is the standard argument.
What does "absolutely convergent" buy me in practice?
Robustness: absolute convergence permits reordering, grouping, and termwise multiplication without changing the sum, the manipulations casually performed in physics derivations. Conditionally convergent series forbid all three, which is why identifying which kind you have (compare Σ|b(n)| against a p-series) matters beyond bookkeeping.
How accurate is a partial sum of an alternating series?
The error is bounded by the magnitude of the first omitted term, provided the terms decrease monotonically to zero. This remarkably simple bound is unique to alternating series and makes them ideal for computing values to a guaranteed precision.