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

Absolute Convergence

Absolute convergence is convergence of a series of absolute values, ensuring convergence of the original scalar series and invariance of its sum under rearrangement.

25 keywords6 linked from4 not yet writtenWritten by AI
Infinite SeriesReal NumberComplex NumberAbsolute ValueLimitTriangle Inequal…Cauchy SequenceGeometric SeriesAbsolute C…

Absolute convergence is a property of an infinite series that holds when the series formed by taking the absolute value of every term converges. For series of real or complex numbers, it is stronger than ordinary convergence: it guarantees that the original series converges and that its sum does not depend on the order of its terms. The concept also extends to integrals and series in normed vector spaces, where it provides conditions for controlling errors, interchanging operations, and defining infinite sums independently of cancellation.

Definition

Let ∑n=1∞an\sum_{n=1}^{\infty}a_n be a series whose terms are real numbers or complex numbers. It is absolutely convergent if

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

where ∣an∣|a_n| denotes the absolute value, or complex modulus, of ana_n.

Equivalently, the partial sums

AN=∑n=1N∣an∣A_N=\sum_{n=1}^{N}|a_n|

have a finite limit. Because these partial sums are nondecreasing, absolute convergence is also equivalent to their being bounded above.

A series that converges but does not converge absolutely is called conditionally convergent. Thus convergence of a scalar series falls into three mutually exclusive cases:

  • Absolute convergence: both ∑an\sum a_n and ∑∣an∣\sum |a_n| converge.
  • Conditional convergence: ∑an\sum a_n converges, but ∑∣an∣\sum |a_n| diverges.
  • Divergence: ∑an\sum a_n does not have a finite sum.

For a series of nonnegative real terms, convergence and absolute convergence are the same property.

Why absolute convergence implies convergence

The implication follows from the triangle inequality. For integers 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 ∑∣an∣\sum |a_n| converges, the right-hand side becomes arbitrarily small once nn is sufficiently large, uniformly for all m>nm>n. Consequently, the partial sums of ∑an\sum a_n form a Cauchy sequence. Completeness of the real and complex number systems then ensures that these partial sums converge.

Absolute convergence also gives the estimate

∣∑n=1∞an∣≤∑n=1∞∣an∣.\left|\sum_{n=1}^{\infty}a_n\right| \leq \sum_{n=1}^{\infty}|a_n|.

More importantly for approximation, if S=∑n=1∞anS=\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|.

The absolute tail therefore bounds the truncation error without requiring information about cancellations among terms.

For a real series, define

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

Then an=an+−an−a_n=a_n^+-a_n^- and ∣an∣=an++an−|a_n|=a_n^++a_n^-. Absolute convergence is equivalent to finiteness of both ∑an+\sum a_n^+ and ∑an−\sum a_n^-. In a conditionally convergent real series, both of these nonnegative sums are infinite; convergence depends on how positive and negative terms are balanced.

Examples and convergence tests

Geometric and power-decay series

The geometric series

∑n=0∞zn\sum_{n=0}^{\infty}z^n

converges absolutely precisely when ∣z∣<1|z|<1, since its absolute-value series is ∑∣z∣n\sum |z|^n.

For real pp,

∑n=1∞1np\sum_{n=1}^{\infty}\frac{1}{n^p}

converges precisely when p>1p>1. It follows that

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

converges absolutely for p>1p>1, converges conditionally for 0<p≤10<p\leq1, and diverges for p≤0p\leq0.

The familiar case

1−12+13−14+⋯=log⁡21-\frac12+\frac13-\frac14+\cdots=\log 2

is conditionally convergent because the corresponding harmonic series diverges.

Comparison, ratio, and root tests

Tests for series of nonnegative terms can be applied to ∑∣an∣\sum |a_n|.

Comparison test. If ∣an∣≤bn|a_n|\leq b_n for all sufficiently large nn, where bn≥0b_n\geq0 and ∑bn\sum b_n converges, then ∑an\sum a_n converges absolutely.

Ratio test. Suppose an≠0a_n\neq0 eventually and

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

exists. If L<1L<1, the series converges absolutely. If L>1L>1, including L=∞L=\infty, its terms fail to tend to zero and the series diverges. The case L=1L=1 is inconclusive.

Root test. Set

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

If L<1L<1, the series converges absolutely; if L>1L>1, it diverges. Again, L=1L=1 is inconclusive.

Absolute convergence does not by itself indicate rapid convergence. For example, ∑n−p\sum n^{-p} converges absolutely for every p>1p>1, but its tails decay slowly when pp is close to 11.

Rearrangement and order independence

One of the principal consequences of absolute convergence is invariance under rearrangement. If ∑an\sum a_n converges absolutely and π\pi is any permutation of the positive integers, then

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

The rearranged series also converges absolutely.

The reason is that the total absolute size of a sufficiently distant tail is small. Once a rearranged partial sum contains a sufficiently large finite set of the original terms, any remaining discrepancy is controlled by that tail, regardless of the order in which its terms appear.

For real or complex scalar series, absolute convergence is equivalent to unconditional convergence: every rearrangement converges to the same sum.

The contrast with conditional convergence is described by the Riemann rearrangement theorem. A conditionally convergent real series can be rearranged to converge to any prescribed real number, diverge to either positive or negative infinity, or fail to have a limit.

This distinction also matters when grouping terms. Grouping consecutive terms of an already convergent series into finite blocks preserves its sum, but convergence of a grouped series alone does not establish convergence of the original series. Absolute convergence permits arbitrary regrouping and reordering without changing the sum.

Products and multiple sums

Suppose

A=∑n=0∞an,B=∑n=0∞bnA=\sum_{n=0}^{\infty}a_n, \qquad B=\sum_{n=0}^{\infty}b_n

are absolutely convergent. Their Cauchy product has coefficients

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

It is absolutely convergent and satisfies

∑n=0∞cn=AB,\sum_{n=0}^{\infty}c_n=AB,

with the bound

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

This is the sequence version of a basic estimate for convolution.

Similarly, if a doubly indexed family satisfies

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

then the sum can be taken row by row, column by column, or through any enumeration of the index pairs:

∑m(∑nam,n)=∑n(∑mam,n).\sum_m\left(\sum_n a_{m,n}\right) = \sum_n\left(\sum_m a_{m,n}\right).

Without absolute summability or another suitable hypothesis, changing the order of summation can change the result or destroy convergence.

Series of functions

For a series of functions ∑fn(x)\sum f_n(x), absolute convergence at a point xx means

∑n=1∞∣fn(x)∣<∞.\sum_{n=1}^{\infty}|f_n(x)|<\infty.

This is a pointwise property. Absolute convergence at every point of a domain does not, by itself, imply uniform convergence.

A stronger sufficient condition is supplied by the Weierstrass M-test. If constants Mn≥0M_n\geq0 satisfy

∣fn(x)∣≤Mn|f_n(x)|\leq M_n

for every xx in the domain, and ∑Mn<∞\sum M_n<\infty, then the function series converges absolutely at every point and uniformly on the domain. The series ∑∣fn(x)∣\sum |f_n(x)| also converges uniformly.

For a power series

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

there is a radius of convergence RR, possibly 00 or ∞\infty. The series converges absolutely when ∣z−z0∣<R|z-z_0|<R, diverges when ∣z−z0∣>R|z-z_0|>R, and requires separate examination on the boundary. Inside the convergence disk, convergence is uniform on every smaller closed disk. These properties underlie the use of power series in complex analysis.

Absolute convergence of integrals

An improper integral is called absolutely convergent when the corresponding integral of the absolute value is finite. For example,

∫a∞f(x) dx\int_a^\infty f(x)\,dx

is absolutely convergent if

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

Under the usual local integrability assumptions, absolute convergence implies convergence of the improper integral. The converse is false. A standard example is

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

which converges, whereas

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

diverges.

In Lebesgue integration, a real- or complex-valued measurable function is integrable in the finite-valued sense precisely when

∫∣f∣ dμ<∞.\int |f|\,d\mu<\infty.

Thus a conditionally convergent improper integral need not be a finite Lebesgue integral over the same domain.

For a random variable XX, the corresponding condition is

E[∣X∣]<∞.\mathbb{E}[|X|]<\infty.

It ensures that the expectation E[X]\mathbb{E}[X] exists as a finite number. Cancellation in a formal expression or a symmetric limiting procedure does not replace this integrability condition.

Normed spaces and the role of completeness

In a normed vector space XX, a series ∑xn\sum x_n is called absolutely convergent, or absolutely summable in norm, if

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

The triangle inequality makes its partial sums a Cauchy sequence. If XX is a Banach space, meaning that it is complete in its norm, the series therefore converges in XX. Without completeness, the sum may exist only in the completion of XX.

In fact, a normed space is complete if and only if every absolutely summable series in that space converges.

The equivalence between absolute and unconditional convergence has a limitation: it holds in finite-dimensional real or complex normed spaces, but fails in general in infinite dimensions.

For example, let ene_n be the standard unit vectors in the Hilbert space ℓ2\ell^2. The series

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

converges unconditionally, because the squared norm of a tail is controlled by a tail of ∑n−2\sum n^{-2}. Nevertheless,

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

Thus unconditional convergence can be strictly weaker than absolute convergence.

The sequence space ℓ1\ell^1 consists precisely of absolutely summable scalar sequences, with norm

∥(an)∥1=∑n=1∞∣an∣.\|(a_n)\|_1=\sum_{n=1}^{\infty}|a_n|.

Viewed this way, absolute convergence is both a property of individual series and a foundational notion in the structure of spaces of sequences.