Absolute vs. Conditional Convergence — Determining Absolute...

SECTION 1

Historical Context & Motivation

The distinction between absolute convergence and conditional convergence arose from deep investigations into the behavior of infinite series during the eighteenth and nineteenth centuries. Early mathematicians such as Euler freely manipulated infinite series—rearranging terms, grouping them, and substituting values—without questioning whether such operations preserved the sum. This uncritical approach occasionally produced spectacular results, but it also led to paradoxes that demanded a more rigorous foundation for the theory of convergence.

1734

Euler's Series Manipulations

Leonhard Euler computed sums of many infinite series by rearranging and regrouping terms freely, sometimes arriving at correct results and sometimes at contradictions, highlighting the need for rigorous convergence criteria.

1827

Cauchy's Convergence Theory

Augustin-Louis Cauchy formalized the definition of convergence using partial sums and introduced the concept of absolute convergence, proving that absolutely convergent series could be rearranged without changing the sum.

1837

Dirichlet's Contributions

Peter Gustav Lejeune Dirichlet clarified the distinction between absolute and conditional convergence and demonstrated that rearrangement theorems require absolute convergence as a hypothesis.

1854

Riemann's Rearrangement Theorem

Bernhard Riemann proved his famous rearrangement theorem: any conditionally convergent series can be rearranged to converge to any desired real number, or even to diverge. This result cemented the importance of distinguishing absolute from conditional convergence.

Riemann's rearrangement theorem revealed a startling truth: a series that converges but does not converge absolutely is, in a precise sense, fragile. Its sum depends on the order in which terms are added, a property that has profound implications for analysis, Fourier series, and numerical computation. The central question this lesson addresses is: given a convergent series, how do we determine whether the convergence is absolute or merely conditional?

SECTION 2

Core Principles & Definitions

Before classifying a series as absolutely or conditionally convergent, one must first establish that the series converges in the ordinary sense. The classification then hinges on whether convergence survives when every term is replaced by its absolute value—a more demanding requirement that eliminates any beneficial cancellation between positive and negative terms.

Convergence of a Series

A series Σaₙ converges if the sequence of partial sums Sₙ = a₁ + a₂ + ⋯ + aₙ approaches a finite limit L as n → ∞. This is the baseline requirement before any classification.

Absolute Convergence

A series Σaₙ converges absolutely if the companion series Σ|aₙ| also converges. Absolute convergence implies ordinary convergence, but not vice versa. Absolutely convergent series are robust: their sums are invariant under any rearrangement of terms.

Conditional Convergence

A series Σaₙ converges conditionally if it converges in the ordinary sense, but Σ|aₙ| diverges. Such series rely on cancellation between positive and negative terms; their sums can change if terms are rearranged.

The Implication Hierarchy

Absolute convergence ⇒ convergence, but convergence ⇏ absolute convergence. Divergence of Σ|aₙ| does not imply divergence of Σaₙ. This asymmetry is the crux of the classification problem.

✦ KEY TAKEAWAY

Think of absolute convergence like a bank account that stays positive even if all withdrawals are converted to deposits—the total is finite regardless of sign changes. Conditional convergence is like a budget that balances only because income and expenses nearly cancel; if you reorder the transactions, you could end up in debt or with a surplus. The question 'absolute or conditional?' is really asking: does the convergence depend on cancellation, or is it robust?

SECTION 3

Visual Explanation

Classification Flowchart for Convergence Type

This flowchart illustrates the two-step classification process. First, determine whether Σaₙ converges at all. If it does not, the series is divergent. If it converges, test whether Σ|aₙ| also converges. A 'yes' yields absolute convergence; a 'no' yields conditional convergence.

The flowchart above captures the essential decision procedure. Notice that the classification is a two-stage test: convergence of Σaₙ is a prerequisite, and only after establishing it do we examine Σ|aₙ|. A common error is to skip Step 1 entirely and jump straight to the absolute value series. While it is true that if Σ|aₙ| converges then Σaₙ converges automatically (so you may proceed directly to Step 2 in that case), if Σ|aₙ| diverges, you must still verify separately that Σaₙ converges before declaring conditional convergence—otherwise the series might simply diverge.

SECTION 4

Mathematical Framework

The formal definitions translate the intuitive ideas of the previous sections into precise mathematical language. Throughout this section, let {aₙ} denote a sequence of real numbers and let Sₙ = Σₖ₌₁ⁿ aₖ denote the nth partial sum of the series.

ABSOLUTE CONVERGENCE

Σaₙ converges absolutely ⟺ Σ|aₙ| = Σₙ₌₁∞ |aₙ| < ∞

A series Σaₙ is absolutely convergent if the series formed by taking the absolute value of each term is itself convergent. Since |aₙ| ≥ 0 for all n, the partial sums of Σ|aₙ| are non-decreasing, so Σ|aₙ| converges if and only if its partial sums are bounded.

CONDITIONAL CONVERGENCE

Σaₙ converges conditionally ⟺ Σaₙ converges AND Σ|aₙ| = ∞

A series is conditionally convergent when the original series converges but the absolute-value series diverges. The convergence depends entirely on cancellation between positive and negative terms.

KEY THEOREM — ABSOLUTE ⇒ CONVERGENCE

If Σ|aₙ| converges, then Σaₙ converges, and |Σaₙ| ≤ Σ|aₙ|

This is proved using the Cauchy criterion: for any ε > 0, the convergence of Σ|aₙ| implies |Σₖ₌ₘⁿ aₖ| ≤ Σₖ₌ₘⁿ |aₖ| < ε for sufficiently large m, n. The inequality |Σaₙ| ≤ Σ|aₙ| is the series analogue of the triangle inequality.

RIEMANN REARRANGEMENT THEOREM

If Σaₙ converges conditionally, then for every L ∈ ℝ ∪ {±∞}, there exists a rearrangement σ such that Σa_{σ(n)} = L

This theorem, due to Riemann, demonstrates that the sum of a conditionally convergent series is not an intrinsic property of the set of terms—it depends on the ordering. In contrast, any rearrangement of an absolutely convergent series converges to the same sum.

These results establish a clear hierarchy: absolute convergence is a stronger condition that guarantees many desirable algebraic properties (rearrangement, multiplication of series via the Cauchy product), while conditional convergence is a weaker, more fragile state. In practice, the determination usually proceeds by first testing Σ|aₙ| for convergence using standard tests (Ratio Test, Root Test, Comparison Test, Integral Test), and if Σ|aₙ| diverges, separately verifying that Σaₙ converges using tests suited to alternating or sign-changing series.

SECTION 5

Classic Examples & Partial-Sum Behavior

The canonical example distinguishing absolute from conditional convergence involves the alternating harmonic series and the p-series. The alternating harmonic series Σ(−1)ⁿ⁺¹/n converges to ln 2 by the Alternating Series Test, yet Σ1/n—the harmonic series—diverges. Thus the alternating harmonic series is conditionally convergent. In contrast, the series Σ(−1)ⁿ⁺¹/n² converges absolutely because Σ1/n² = π²/6 is finite.

The amber curve shows the partial sums of the conditionally convergent alternating harmonic series, which oscillates widely before settling toward ln 2 ≈ 0.693. The cyan curve shows the absolutely convergent series Σ(−1)ⁿ⁺¹/n², whose partial sums converge far more rapidly and with much smaller oscillations to π²/12 ≈ 0.822. The visual difference in oscillation amplitude reflects the role of cancellation.

Classification of common series

Series	Σaₙ Converges?	Σ\|aₙ\| Converges?	Classification
Σ(−1)ⁿ⁺¹/n	Yes (= ln 2)	No (harmonic)	Conditional
Σ(−1)ⁿ⁺¹/n²	Yes (= π²/12)	Yes (= π²/6)	Absolute
Σ(−1)ⁿ⁺¹/√n	Yes (AST)	No (p = ½ ≤ 1)	Conditional
Σ(−1)ⁿ/2ⁿ	Yes (geometric)	Yes (\|r\| = ½ < 1)	Absolute
Σ1/n	No	No	Divergent

SECTION 6

Worked Example

Let us determine whether the series Σₙ₌₁∞ (−1)ⁿ⁺¹ · n/(n² + 1) converges absolutely, converges conditionally, or diverges.

Classify: Σₙ₌₁∞ (−1)ⁿ⁺¹ · n/(n² + 1)

Step 1 — Examine the Absolute-Value SeriesForm the series of absolute values: Σₙ₌₁∞ |aₙ| = Σₙ₌₁∞ n/(n² + 1). For large n, we note that n/(n² + 1) ≈ n/n² = 1/n. This suggests comparison with the harmonic series.

Step 2 — Limit Comparison Test on Σ|aₙ|Apply the Limit Comparison Test with bₙ = 1/n. Compute: lim(n→∞) [n/(n² + 1)] / [1/n] = lim(n→∞) n²/(n² + 1) = lim(n→∞) 1/(1 + 1/n²) = 1. Since this limit is a positive finite number and Σ1/n diverges (harmonic series), the Limit Comparison Test tells us that Σn/(n² + 1) also diverges.

Σ|aₙ| diverges → NOT absolutely convergent

Step 3 — Test Σaₙ for Convergence (Alternating Series Test)Since Σ|aₙ| diverges, we cannot conclude absolute convergence, but the series might still converge conditionally. The series has the form Σ(−1)ⁿ⁺¹ bₙ with bₙ = n/(n² + 1). We apply the Alternating Series Test (Leibniz Test), which requires: (i) bₙ > 0 for all n, (ii) bₙ is eventually decreasing, and (iii) lim(n→∞) bₙ = 0.

Step 4 — Verify the Three AST ConditionsCondition (i): bₙ = n/(n² + 1) > 0 for all n ≥ 1. ✓ Condition (ii): Consider f(x) = x/(x² + 1). Then f′(x) = (x² + 1 − 2x²)/(x² + 1)² = (1 − x²)/(x² + 1)². For x > 1, f′(x) < 0, so f is decreasing for n ≥ 2, which means bₙ is eventually decreasing. ✓ Condition (iii): lim(n→∞) n/(n² + 1) = lim(n→∞) 1/(n + 1/n) = 0. ✓

All three conditions satisfied → Σaₙ converges by the Alternating Series Test

Step 5 — State the ConclusionSince Σaₙ converges (by AST) but Σ|aₙ| diverges (by Limit Comparison), the series Σₙ₌₁∞ (−1)ⁿ⁺¹ · n/(n² + 1) satisfies the definition of conditional convergence.

The series converges conditionally.

💡 Strategy Note

A useful shortcut: if you can show Σ|aₙ| converges, you're done—the series is absolutely convergent and you need not check the original series separately. Only when Σ|aₙ| diverges must you perform a second test on Σaₙ to distinguish conditional convergence from outright divergence.

SECTION 7

Choosing the Right Convergence Test

Determining absolute or conditional convergence often requires selecting the appropriate test for two separate investigations: one for Σ|aₙ| and potentially another for Σaₙ itself. The table below summarizes the most commonly used tests and their applicability in this context.

Convergence tests and their role in absolute/conditional classification

Test	Tests Σ\|aₙ\|?	Tests Σaₙ directly?	Best Used When…
Ratio Test	Yes (inherently uses \|aₙ\|)	Determines absolute conv.	Factorials, exponentials, products
Root Test	Yes (uses \|aₙ\|¹ᐟⁿ)	Determines absolute conv.	nth powers, nested exponents
Comparison / LCT	Yes (applied to \|aₙ\|)	Typically not (needs non-negative)	Rational functions, p-series comparisons
Integral Test	Yes (applied to \|aₙ\|)	Needs positive terms	Continuous, positive, decreasing terms
Alternating Series Test	No	Yes (Σaₙ directly)	Alternating sign, decreasing \|aₙ\| → 0

✦ KEY TAKEAWAY

The Ratio and Root Tests are particularly efficient because they simultaneously test for absolute convergence: if the limit is less than 1, the series is absolutely convergent; if greater than 1, it diverges. Only when these tests are inconclusive (limit = 1) must you resort to comparison-type tests for Σ|aₙ| and the Alternating Series Test for Σaₙ.

SECTION 8

Connections to Rearrangements and Power Series

The distinction between absolute and conditional convergence is not merely a classification exercise—it has far-reaching consequences in analysis, particularly for rearrangements of series, the Cauchy product of two series, and the behavior of power series at the boundary of their interval of convergence.

Absolute vs. conditional: algebraic consequences

Property	Absolutely Convergent	Conditionally Convergent
Rearrangement invariance	Sum unchanged under any rearrangement	Sum can be made any value in ℝ ∪ {±∞} (Riemann)
Cauchy product	Product of two abs. conv. series converges absolutely	Product may diverge unless at least one factor is abs. conv.
Power series at endpoints	Convergence is unambiguous and robust	Endpoint behavior requires careful individual analysis
Grouping of terms	Any grouping preserves convergence and sum	Grouping preserves convergence, but removing parentheses may not

In the context of power series Σcₙ(x − a)ⁿ, the series converges absolutely for |x − a| < R (the radius of convergence) and diverges for |x − a| > R. At the endpoints x = a ± R, the series may converge absolutely, converge conditionally, or diverge—and the classification at each endpoint must be determined individually. For example, the power series for ln(1 + x) = Σ(−1)ⁿ⁺¹xⁿ/n has R = 1: at x = 1 it converges conditionally (alternating harmonic), and at x = −1 it diverges (negative harmonic). These endpoint investigations are precisely the absolute-vs.-conditional analysis studied in this lesson, applied to a concrete function-theoretic setting.

🔭 Looking Ahead

In advanced courses (real analysis, functional analysis), absolute convergence generalizes to unconditional convergence in normed vector spaces. In finite-dimensional spaces, absolute and unconditional convergence coincide, but in infinite-dimensional Banach spaces they can differ—a result known as the Dvoretzky–Rogers theorem.

SECTION 9

Practice Problems

PROBLEM 1 — CONCEPTUAL

Explain why every absolutely convergent series is convergent, but not every convergent series is absolutely convergent. Illustrate your explanation with a specific example of each type.

PROBLEM 2 — BASIC CALCULATION

Determine whether the series Σₙ₌₁∞ (−1)ⁿ/n³ converges absolutely, converges conditionally, or diverges.

PROBLEM 3 — INTERMEDIATE

Determine whether the series Σₙ₌₂∞ (−1)ⁿ/(n · ln n) converges absolutely, converges conditionally, or diverges.

PROBLEM 4 — APPLIED

The Maclaurin series for arctan(x) is Σₙ₌₀∞ (−1)ⁿ x²ⁿ⁺¹/(2n + 1), valid for |x| ≤ 1. Determine whether this series converges absolutely, conditionally, or diverges at x = 1 and at x = −1.

PROBLEM 5 — CRITICAL THINKING

Prove that if Σaₙ converges conditionally, then both Σaₙ⁺ and Σaₙ⁻ diverge, where aₙ⁺ = max(aₙ, 0) and aₙ⁻ = max(−aₙ, 0) are the positive and negative parts of aₙ. (Hint: note that aₙ = aₙ⁺ − aₙ⁻ and |aₙ| = aₙ⁺ + aₙ⁻.)

SUMMARY

Lesson Summary

A series Σaₙ is classified based on the behavior of its companion series Σ|aₙ|. If Σ|aₙ| converges, then Σaₙ is absolutely convergent—a robust form of convergence that is invariant under rearrangement and supports algebraic operations like the Cauchy product. If Σaₙ converges but Σ|aₙ| diverges, then Σaₙ is conditionally convergent, relying on cancellation between positive and negative terms. By the Riemann rearrangement theorem, conditionally convergent series can be rearranged to converge to any prescribed sum.

The determination strategy follows a two-step procedure: first test Σ|aₙ| for convergence using the Ratio Test, Root Test, Comparison Test, or Integral Test. If Σ|aₙ| converges, you are done—the series is absolutely convergent. If Σ|aₙ| diverges, apply the Alternating Series Test (or another appropriate test) to Σaₙ. If Σaₙ converges, the convergence is conditional; if Σaₙ also diverges, the series simply diverges.