CALCULUS 2 • SEQUENCES & SERIES

Alternating Series Test

A powerful convergence criterion for series whose terms alternate in sign, yielding elegant error bounds.

Historical Context & Motivation

The study of infinite series with alternating signs stretches back to some of the earliest investigations into convergence. Before formal analysis provided rigorous foundations, mathematicians grappled with paradoxical results: certain series whose individual terms do not shrink to zero fast enough for absolute convergence can nonetheless settle to a finite sum when successive terms cancel against one another. The Alternating Series Test (sometimes called the Leibniz criterion) formalized this intuition, providing one of the most elegant and widely used convergence tests in analysis. Its development is intertwined with the broader history of making the calculus of infinite sums rigorous.

1682
Leibniz and the Alternating Harmonic Series
Gottfried Wilhelm Leibniz studied the series 1 − 1/2 + 1/3 − 1/4 + ⋯ and showed it converges to ln 2, establishing the first major example of a conditionally convergent alternating series.
1714
Leibniz's Convergence Criterion
Leibniz articulated the key insight that an alternating series converges if the magnitude of its terms decreases monotonically to zero, though his argument lacked modern rigor.
1821
Cauchy's Cours d'Analyse
Augustin-Louis Cauchy provided a rigorous proof of the alternating series test using the language of limits and sequences of partial sums, cementing it in the foundations of real analysis.
1837
Dirichlet's Refinements
Peter Gustav Lejeune Dirichlet extended the ideas behind alternating series to more general classes of series through his test for convergence, broadening the theory of conditional convergence.

The central question the Alternating Series Test addresses is deceptively simple: when does a series whose terms flip between positive and negative converge to a finite value? Many of the most important series in mathematics — from the expansion of ln(1 + x) to Fourier series representations — are alternating in nature. Having a clean, verifiable criterion for their convergence is indispensable for both theoretical work and practical computation.

Core Principles & Definitions

An alternating series is any series whose terms alternate in sign. The standard forms are ∑(−1)n bn or ∑(−1)n+1 bn, where bn > 0 for all n. The factor (−1)n is responsible for the sign alternation, while bn captures the magnitude of each term. Understanding the test requires distinguishing clearly between the role of the sign-alternation mechanism and the behavior of the magnitudes.

1

Alternating Structure

The series must have terms that strictly alternate between positive and negative. This is encoded by (−1)n or (−1)n+1 multiplied by a positive sequence bn.
2

Monotone Decrease

The magnitudes bn must be eventually non-increasing: bn+1 ≤ bn for all n ≥ N for some integer N.
3

Limit to Zero

The limit lim bn as n → ∞ must equal 0. Without this condition, the series diverges by the Divergence Test.
4

Conditional vs. Absolute Convergence

A series that passes the AST may converge only conditionally — meaning ∑|an| diverges even though ∑an converges.
KEY TAKEAWAY
Think of an alternating series like a person pacing back and forth on a number line, each step shorter than the last. Because each reversal overshoots less than the one before, the person inevitably settles down to a single resting point — the sum of the series. The two conditions of the test (decreasing steps, steps shrinking to zero) are precisely what guarantee this settling behavior.

Visual Explanation

The partial sums of an alternating series exhibit a distinctive oscillatory pattern that is best understood visually. Each odd partial sum overshoots the eventual limit from one side, while each even partial sum undershoots from the other, creating a pair of monotone subsequences that squeeze together toward the series' sum. The following diagram illustrates this convergence behavior for the alternating harmonic series.

The odd partial sums (pink) decrease toward ln 2, while the even partial sums (violet) increase toward ln 2. The dashed cyan line represents the limit. As n grows, the oscillations tighten and both subsequences converge to the same value.

Observe that S2 < S4 < S6 < ⋯ < S < ⋯ < S5 < S3 < S1. The even partial sums form a non-decreasing sequence bounded above by every odd partial sum, while the odd partial sums form a non-increasing sequence bounded below by every even partial sum. By the Monotone Convergence Theorem, both subsequences converge, and because their difference |S2k+1 − S2k| = b2k+1 → 0, they must converge to the same limit.

Mathematical Framework

The Alternating Series Test can be stated with precision and proved using elementary properties of real sequences. The theorem and its built-in error estimate are among the most elegant results in introductory analysis.

ALTERNATING SERIES TEST (LEIBNIZ CRITERION)
If b₁ ≥ b₂ ≥ b₃ ≥ ⋯ > 0 and lim(n→∞) bₙ = 0, then ∑ₙ₌₁∞ (−1)ⁿ⁺¹ bₙ converges.
Here bn denotes the magnitude (positive part) of each term. The two hypotheses — monotone decrease and convergence to zero — are both required. Neither alone suffices.

Proof Sketch

Consider the even partial sums S2k. We have S2k = S2k−2 + (b2k−1 − b2k) ≥ S2k−2 since b2k−1 ≥ b2k. So {S2k} is non-decreasing. Simultaneously, we can write S2k = b1 − (b2 − b3) − (b4 − b5) − ⋯ ≤ b1, showing the sequence is bounded above. By the Monotone Convergence Theorem, {S2k} converges to some limit S. Since S2k+1 = S2k + b2k+1 and b2k+1 → 0, the odd partial sums also converge to S.

ALTERNATING SERIES ESTIMATION THEOREM
|S − Sₙ| ≤ bₙ₊₁
The error in approximating the sum S by the n-th partial sum Sn is bounded by the magnitude of the first omitted term bn+1. This provides a remarkably simple and practical error bound.
Common Pitfall
Students frequently forget that the hypotheses must be checked on the magnitudes bn, not the signed terms an = (−1)n bn. Also, "eventually decreasing" is sufficient — the test applies as long as bn+1 ≤ bn holds for all n beyond some threshold N.

Applying the Test: Decision Flowchart

In practice, applying the Alternating Series Test requires a systematic check of its hypotheses. When one or both conditions fail, different conclusions arise. The following flowchart guides you through the decision process for any series you suspect to be an alternating series.

Start at the top with your alternating series. First check whether bn → 0 (if not, the series diverges outright). Then verify monotone decrease. If both conditions hold, the series converges. The final diamond distinguishes between conditional and absolute convergence.

A subtle but important point: when the AST conditions are not met, this does not automatically mean the series diverges (unless lim bn ≠ 0). If the magnitudes fail to decrease monotonically but do tend to zero, you may need a different convergence test — such as Dirichlet's test or Abel's test — to draw a conclusion.

Worked Example

Let us apply the Alternating Series Test to determine convergence and then use the estimation theorem to bound the error of a partial sum approximation.

Determine whether ∑ₙ₌₁∞ (−1)ⁿ⁺¹ / (2n − 1) converges, and find how many terms are needed for the error to be less than 0.01.
1
Step 1 — Identify the alternating structureThe series is ∑ (−1)n+1 bn where bn = 1/(2n − 1). Since 2n − 1 > 0 for all n ≥ 1, we have bn > 0, and the factor (−1)n+1 provides the alternation.
bn = 1/(2n − 1) > 0 ✓
2
Step 2 — Check the limit conditionWe compute lim(n→∞) bn = lim(n→∞) 1/(2n − 1) = 0. As n grows without bound, the denominator grows without bound, so the fraction tends to zero.
lim(n→∞) bn = 0 ✓
3
Step 3 — Check monotone decreaseWe need bn+1 ≤ bn. Since bn+1 = 1/(2(n+1) − 1) = 1/(2n + 1) and 2n + 1 > 2n − 1 for all n ≥ 1, we have 1/(2n + 1) < 1/(2n − 1), confirming the sequence is strictly decreasing.
bn+1 < bn for all n ≥ 1 ✓
4
Step 4 — Conclude convergenceBoth conditions of the Alternating Series Test are satisfied. Therefore, ∑ₙ₌₁∞ (−1)n+1 / (2n − 1) = 1 − 1/3 + 1/5 − 1/7 + ⋯ converges. (This is the Leibniz formula for π/4.)
The series converges by the AST.
5
Step 5 — Apply the estimation theoremBy the Alternating Series Estimation Theorem, |S − Sn| ≤ bn+1 = 1/(2(n+1) − 1) = 1/(2n + 1). We need 1/(2n + 1) < 0.01, which gives 2n + 1 > 100, so n > 49.5. Thus n = 50 terms suffice.
n = 50 terms guarantee |error| < 0.01

Strengths, Limitations & Comparisons

The Alternating Series Test is a focused tool with clear strengths and well-defined limitations. Understanding where it fits among the broader toolkit of convergence tests helps you choose the right approach for any given series.

Strengths and limitations of the Alternating Series Test
AspectStrengthLimitation
Ease of useOnly two conditions to check — monotone decrease and limit to zero. No comparison series needed.Applies only to series with strictly alternating signs; cannot handle irregular sign patterns.
Error boundsThe estimation theorem gives a free, sharp error bound: |error| ≤ bn+1.The bound may be loose (overly conservative) for series where terms decrease very slowly.
Convergence typeCan detect conditional convergence, which tests like Ratio and Root cannot.Does not determine absolute convergence; a separate test on ∑|an| is needed.
Failure modeWhen the limit condition fails, the Divergence Test immediately gives divergence.When monotone decrease fails but the limit is zero, the AST is inconclusive — not a proof of divergence.
KEY TAKEAWAY
Think of your convergence test toolkit like diagnostic instruments in an engineering lab. The AST is a specialized sensor: extremely effective for the class of signals it's designed to detect (alternating sign patterns with decreasing amplitude), but it won't help you analyze signals outside its design range. When the AST is inconclusive, you pivot to more general instruments — the Ratio Test, Root Test, or comparison methods — just as an engineer switches tools to match the problem at hand.

Connection to Advanced Theory

The Alternating Series Test is a special case of more powerful theorems in analysis, and its implications extend into areas such as rearrangement of series, power series convergence at endpoints, and numerical computation. Understanding these connections deepens your appreciation of why the AST matters beyond a single convergence check.

How the AST connects to more advanced convergence theory
AST ConceptAdvanced Generalization
Alternating signs (−1)nDirichlet's Test: Replaces (−1)n with any sequence whose partial sums are bounded.
Monotone decreasing bn → 0Abel's Test: Pairs a convergent series with a bounded monotone sequence.
Conditional convergenceRiemann Rearrangement Theorem: Any conditionally convergent series can be rearranged to converge to any real number, or to diverge.
Error bound |S − Sn| ≤ bn+1Euler–Maclaurin summation and Richardson extrapolation provide tighter error control for slow-converging alternating series.

A particularly striking consequence of conditional convergence is the Riemann Rearrangement Theorem: if ∑an converges conditionally, then for any target value L ∈ ℝ (or ±∞), there exists a rearrangement of the terms that converges to L. This reveals that the ordering of terms in a conditionally convergent series is not merely cosmetic — it determines the sum entirely. The AST detects exactly the class of alternating series most susceptible to this phenomenon, making it an essential gateway to deeper analytic reasoning. In applications, the estimation theorem plays a key role when computing power series values at endpoints, where alternating series frequently arise (e.g., ln 2, π/4, and many Taylor series evaluated at the boundary of their interval of convergence).

Practice Problems

PROBLEM 1CONCEPTUAL
Explain why the series ∑ₙ₌₁∞ (−1)n · n/(n + 1) diverges, even though its terms alternate in sign.
PROBLEM 2BASIC CALCULATION
Determine whether the series ∑ₙ₌₁∞ (−1)n+1 / n² converges. If it does, state whether the convergence is conditional or absolute.
PROBLEM 3INTERMEDIATE
Does the series ∑ₙ₌₁∞ (−1)n · ln(n)/n converge? Verify all conditions of the AST carefully.
PROBLEM 4APPLIED
The Maclaurin series for arctan(x) evaluated at x = 1 gives π/4 = ∑ₙ₌₀∞ (−1)n / (2n + 1). How many terms of this series must be summed to approximate π/4 with an error less than 0.001?
PROBLEM 5CRITICAL THINKING
Construct an example of a series ∑ (−1)n bn where bn > 0, lim bn = 0, but {bn} is NOT monotone decreasing, and the series still converges. Then construct an example with the same properties where the series diverges. What does this reveal about the necessity of the monotone condition?

Summary & Review

The Alternating Series Test (Leibniz criterion) states that a series ∑(−1)n bn converges whenever two conditions hold: the magnitudes bn are eventually non-increasing, and lim bₙ = 0. The proof relies on showing that even and odd partial sums form monotone bounded subsequences that squeeze together to a common limit. This elegant structure also yields the Alternating Series Estimation Theorem: the truncation error |S − Sn| is bounded by the first omitted term bn+1.

A series passing the AST may exhibit conditional convergence (convergent, but not absolutely convergent), which has profound consequences: the Riemann Rearrangement Theorem shows that reordering the terms of a conditionally convergent series can change its sum to any desired value. The AST generalizes to Dirichlet's and Abel's tests, which handle broader classes of series. When applying the AST, always verify both hypotheses on the magnitudes bn, and remember that failure of the monotone condition makes the test inconclusive rather than proving divergence.

Varsity Tutors • Calculus 2 • Alternating Series Test