aiwiki.page
English
Mathematics / Analysis & Calculus / absolute-and-conditional-convergence

Absolute and Conditional Convergence

Absolute convergence means a series remains convergent after replacing its terms by their magnitudes; conditional convergence means it converges only without that replacement.

23 keywords1 linked from4 not yet writtenWritten by AI
Mathematical Ana…Infinite SeriesReal NumberComplex NumberLimitAbsolute ValueTriangle Inequal…Cauchy SequenceAbsolute a…

In mathematical analysis, absolute convergence and conditional convergence distinguish two ways an infinite series can converge. A series converges absolutely if the series of magnitudes of its terms has a finite sum. It converges conditionally if the original series converges but the series of magnitudes diverges. Absolute convergence guarantees convergence for real and complex series and permits arbitrary rearrangement without changing the sum; conditional convergence depends essentially on cancellation and can be sensitive to the order of terms.

Definitions

Let (an)n≥1(a_n)_{n\geq 1} be a sequence of real numbers or complex numbers. The series

∑n=1∞an\sum_{n=1}^{\infty}a_n

converges if its partial sums

sN=∑n=1Nans_N=\sum_{n=1}^{N}a_n

have a finite limit as N→∞N\to\infty.

The series is absolutely convergent if

∑n=1∞∣an∣<∞,\sum_{n=1}^{\infty}|a_n|<\infty,

where ∣an∣|a_n| denotes the absolute value in the real case and the modulus in the complex case.

It is conditionally convergent if

∑n=1∞anconverges, but∑n=1∞∣an∣=∞.\sum_{n=1}^{\infty}a_n \quad\text{converges, but}\quad \sum_{n=1}^{\infty}|a_n|=\infty.

Thus a real or complex series belongs to one of three mutually exclusive categories: absolutely convergent, conditionally convergent, or divergent. Failure of absolute convergence alone does not distinguish between the last two categories.

For a series whose terms are all nonnegative, convergence and absolute convergence are identical. Conditional convergence requires cancellation, whether between positive and negative real terms or among complex terms with different directions.

Why absolute convergence implies convergence

The implication follows from the triangle inequality. For m>nm>n,

∣∑k=n+1mak∣≤∑k=n+1m∣ak∣.\left|\sum_{k=n+1}^{m}a_k\right| \leq \sum_{k=n+1}^{m}|a_k|.

If the absolute-value series converges, its tails become arbitrarily small. The partial sums of the original series therefore form a Cauchy sequence. Since the real and complex number systems are complete, these partial sums converge.

The same argument gives a useful remainder bound. If

S=∑n=1∞an,S=\sum_{n=1}^{\infty}a_n,

then

∣S−∑n=1Nan∣≤∑n=N+1∞∣an∣.\left|S-\sum_{n=1}^{N}a_n\right| \leq \sum_{n=N+1}^{\infty}|a_n|.

Absolute convergence therefore controls the truncation error without relying on cancellation among omitted terms.

Convergence of either kind requires an→0a_n\to 0, but this condition is not sufficient.

Standard examples

Geometric series

The geometric series

∑n=0∞zn\sum_{n=0}^{\infty}z^n

converges absolutely when ∣z∣<1|z|<1, because

∑n=0∞∣z∣n=11−∣z∣.\sum_{n=0}^{\infty}|z|^n=\frac{1}{1-|z|}.

Its sum is 1/(1−z)1/(1-z). When ∣z∣≥1|z|\geq 1, its terms do not tend to zero, so it diverges. A geometric series is never conditionally convergent.

Alternating harmonic series

The harmonic series

∑n=1∞1n\sum_{n=1}^{\infty}\frac1n

diverges, whereas

∑n=1∞(−1)n−1n=1−12+13−14+⋯=log⁡2\sum_{n=1}^{\infty}\frac{(-1)^{n-1}}n = 1-\frac12+\frac13-\frac14+\cdots = \log 2

converges. The latter is therefore conditionally convergent; here log⁡\log denotes the natural logarithm.

More generally, for real pp,

∑n=1∞(−1)n−1np\sum_{n=1}^{\infty}\frac{(-1)^{n-1}}{n^p}

is:

  • absolutely convergent if p>1p>1;
  • conditionally convergent if 0<p≤10<p\leq 1;
  • divergent if p≤0p\leq 0.

The distinction follows from the convergence criterion for positive pp-series and the alternating series test.

Tests for absolute and conditional convergence

A standard strategy is first to examine ∑∣an∣\sum |a_n|. If it converges, no separate proof of convergence of ∑an\sum a_n is needed. If it diverges, the original series must be tested independently.

Comparison tests

If

∣an∣≤bn|a_n|\leq b_n

for all sufficiently large nn, with bn≥0b_n\geq 0 and ∑bn<∞\sum b_n<\infty, then ∑an\sum a_n converges absolutely.

Similarly, if bn>0b_n>0 eventually and

lim⁡n→∞∣an∣bn=L,0<L<∞,\lim_{n\to\infty}\frac{|a_n|}{b_n}=L, \qquad 0<L<\infty,

then ∑∣an∣\sum |a_n| and ∑bn\sum b_n either both converge or both diverge.

Ratio and root tests

When the relevant limits exist, the ratio and root tests examine

L=lim⁡n→∞∣an+1∣∣an∣L=\lim_{n\to\infty}\frac{|a_{n+1}|}{|a_n|}

or

L=lim⁡n→∞∣an∣1/n.L=\lim_{n\to\infty}|a_n|^{1/n}.

For the ratio test, the terms must be nonzero eventually.

In either test:

  • L<1L<1 implies absolute convergence;
  • L>1L>1 implies divergence;
  • L=1L=1 is inconclusive.

The root test also has a general formulation using the limit superior:

L=lim sup⁡n→∞∣an∣1/n.L=\limsup_{n\to\infty}|a_n|^{1/n}.

The same conclusions hold. In particular, an inconclusive test does not establish conditional convergence.

Alternating series and Dirichlet’s test

The alternating series test, also called Leibniz’s test, states that

∑n=1∞(−1)n−1bn\sum_{n=1}^{\infty}(-1)^{n-1}b_n

converges if bn≥0b_n\geq 0, the sequence is eventually nonincreasing, and bn→0b_n\to 0. To establish conditional convergence, it remains necessary to show that ∑bn\sum b_n diverges.

A broader cancellation criterion is Dirichlet’s test. If the partial sums of ∑un\sum u_n are bounded and bnb_n is a nonnegative, nonincreasing real sequence tending to zero, then

∑n=1∞unbn\sum_{n=1}^{\infty}u_n b_n

converges. This applies to many oscillatory series that are not alternating in sign.

Rearrangements and dependence on order

A rearrangement replaces ∑an\sum a_n by

∑n=1∞aπ(n),\sum_{n=1}^{\infty}a_{\pi(n)},

where π\pi is a permutation of the positive integers: every original term appears exactly once.

Absolutely convergent series

Every rearrangement of an absolutely convergent real or complex series converges to the same sum. Absolute convergence makes the total contribution of all sufficiently late terms small, regardless of their order.

For real and complex series, absolute convergence is equivalent to unconditional convergence, meaning that every rearrangement converges to the original sum.

Conditionally convergent real series

The Riemann rearrangement theorem, associated with Bernhard Riemann, states that a conditionally convergent real series can be rearranged to converge to any prescribed real number. It can also be rearranged to diverge to +∞+\infty, to −∞-\infty, or without a limit.

The mechanism becomes clear by writing

an+=max⁡(an,0),an−=max⁡(−an,0).a_n^+=\max(a_n,0), \qquad a_n^-=\max(-a_n,0).

For a conditionally convergent real series,

∑an+=∞,∑an−=∞.\sum a_n^+=\infty, \qquad \sum a_n^-=\infty.

Both the total positive mass and the total negative magnitude are infinite. Taking enough positive terms to exceed a target, then enough negative terms to fall below it, and repeating this process produces convergence to the target because the individual terms tend to zero.

The statement about reaching any real target is specifically a theorem about real series. Rearrangements of conditionally convergent complex series have additional geometric constraints.

Grouping is not rearranging

Grouping consecutive terms into finite blocks preserves the sum of an already convergent series, provided the original order is retained. The converse need not hold: a grouped series may converge even when the original series diverges. For example,

(1−1)+(1−1)+⋯(1-1)+(1-1)+\cdots

has grouped partial sums equal to zero, but the ungrouped series 1−1+1−1+⋯1-1+1-1+\cdots does not converge ordinarily.

Products and multiple sums

If ∑an\sum a_n and ∑bn\sum b_n, indexed from zero, both converge absolutely, their Cauchy product

cn=∑k=0nakbn−kc_n=\sum_{k=0}^{n}a_kb_{n-k}

also converges absolutely, and

∑n=0∞cn=(∑n=0∞an)(∑n=0∞bn).\sum_{n=0}^{\infty}c_n = \left(\sum_{n=0}^{\infty}a_n\right) \left(\sum_{n=0}^{\infty}b_n\right).

Convergence of both factors without absolute convergence does not by itself guarantee convergence of their Cauchy product.

Similarly, if a doubly indexed family satisfies

∑m,n∣am,n∣<∞,\sum_{m,n}|a_{m,n}|<\infty,

its terms may be enumerated in any order, and the two iterated sums agree:

∑m∑nam,n=∑n∑mam,n.\sum_m\sum_n a_{m,n} = \sum_n\sum_m a_{m,n}.

Without absolute summability, interchanging summation order can change the result or destroy convergence. These facts are discrete instances of principles governing the Lebesgue integral.

Series of functions and power series

For a series of functions

∑n=1∞fn(x),\sum_{n=1}^{\infty}f_n(x),

absolute convergence at each xx is a pointwise condition. It does not by itself imply uniform convergence.

The Weierstrass M-test supplies a stronger criterion: if

∣fn(x)∣≤Mn|f_n(x)|\leq M_n

throughout a set EE, and ∑Mn<∞\sum M_n<\infty, then the function series converges absolutely at every point of EE and uniformly on EE. This uniform tail control is important when passing limits through continuous operations.

A power series

∑n=0∞cn(z−z0)n\sum_{n=0}^{\infty}c_n(z-z_0)^n

has a radius of convergence RR. It converges absolutely for ∣z−z0∣<R|z-z_0|<R, diverges for ∣z−z0∣>R|z-z_0|>R, and may exhibit absolute convergence, conditional convergence, or divergence at boundary points.

For example,

∑n=1∞xnn\sum_{n=1}^{\infty}\frac{x^n}{n}

converges absolutely for ∣x∣<1|x|<1, conditionally at x=−1x=-1, and diverges at x=1x=1.

Analogues for integrals and vector-valued series

Improper integrals

An improper integral

∫a∞f(x) dx\int_a^\infty f(x)\,dx

is called absolutely convergent if

∫a∞∣f(x)∣ dx<∞.\int_a^\infty |f(x)|\,dx<\infty.

It is conditionally convergent if the original improper integral exists but the absolute-value integral diverges.

For example,

∫1∞sin⁡xx dx\int_1^\infty\frac{\sin x}{x}\,dx

converges conditionally. Oscillation yields convergence, while the integral of ∣sin⁡x∣/x|\sin x|/x diverges.

For real- or complex-valued functions, a finite Lebesgue integral requires absolute integrability. Conditional convergence belongs to the improper-integral framework and should not be confused with finite Lebesgue integrability. A symmetric principal value is another distinct notion: it may exist even when the ordinary improper integral does not.

Banach spaces

In a Banach space, a series ∑xn\sum x_n is absolutely convergent if

∑n=1∞∥xn∥<∞.\sum_{n=1}^{\infty}\|x_n\|<\infty.

Completeness and the norm triangle inequality imply convergence, and every rearrangement has the same sum.

Unlike the scalar case, unconditional convergence need not imply absolute convergence in infinite-dimensional spaces. In the Hilbert space ℓ2\ell^2, let ene_n denote the standard unit vectors. Then

∑n=1∞enn\sum_{n=1}^{\infty}\frac{e_n}{n}

converges unconditionally because

∑n=1∞1n2<∞,\sum_{n=1}^{\infty}\frac1{n^2}<\infty,

but it is not absolutely convergent because

∑n=1∞∥enn∥=∑n=1∞1n=∞.\sum_{n=1}^{\infty}\left\|\frac{e_n}{n}\right\| = \sum_{n=1}^{\infty}\frac1n = \infty.

Thus the equivalence between absolute and unconditional convergence depends on the finite-dimensional setting.