Infinite Series

An infinite series is the sum of infinitely many terms, $\sum_{n = 1}^{\infty} a_{n}$ , defined as the limit of its partial sums. Key result: a series converges iff its sequence of partial sums has a finite limit; otherwise it diverges. Applies in: calculus, analysis, physics, numerical computation, and probability. Who/when: formalized by Cauchy (1821) and Weierstrass (1850s) after centuries of use by Archimedes, Newton, Euler, and others.

An infinite series is an expression of the form $a_{1} + a_{2} + a_{3} + \dots$ , where the $a_{n}$ are terms of a sequence. Its meaning is not given by "adding infinitely many numbers" — that is meaningless — but by the limit of finite partial sums $s_{N} = \sum_{n = 1}^{N} a_{n}$ . If this limit exists and is finite, the series converges to that value; otherwise it diverges. Series matter because they let us represent functions (power series, Fourier series), compute transcendental constants, and solve differential equations, turning analysis into algebra of coefficients. A surprising fact: rearranging the terms of a conditionally convergent series can make it sum to any real number you like — addition stops being commutative.

The Core Insight

A series is not a sum — it is a sequence in disguise. The object you actually study is the sequence of partial sums $(s_{N})$ . Everything about convergence, value, and behavior of $\sum a_{n}$ is a statement about the limit of $(s_{N})$ . The terms $a_{n}$ are merely the differences $s_{N} - s_{N - 1}$ . This reframing — series-as-sequence — is the single idea that makes the whole subject rigorous.

Formal Statement

Series and Partial Sums

Given a sequence of real (or complex) numbers $(a_{n})_{n \geq 1}$ , the $N$ -th partial sum is

s_{N} = n = 1 \sum N a_{n} = a_{1} + a_{2} + \dots + a_{N} .

⟳This is a finite sum, always well-defined;

N

ranges over positive integers.

The infinite series $\sum_{n = 1}^{\infty} a_{n}$ converges to $S$ if

N \to \infty lim s_{N} = S, S \in R .

⟳This is the standard

ε

–

N

limit applied to the sequence

(s_{N})

;

S

is called the sum.

If no such finite $S$ exists, the series diverges. Here $a_{n}$ is the general term, $n$ the index, and $S$ the sum (a number, not a process).

Absolute and Conditional Convergence

A series $\sum a_{n}$ converges absolutely if $\sum ∣ a_{n} ∣$ converges. It converges conditionally if $\sum a_{n}$ converges but $\sum ∣ a_{n} ∣$ diverges.

Cauchy Criterion for Series

\sum a_{n}

converges if and only if for every

ε > 0

there exists

N

such that for all

m > n \geq N

k = n + 1 \sum m a_{k} < ε .

⟳Apply the Cauchy criterion for sequences to

(s_{N})

in the complete space

R

; the tail sum is

s_{m} - s_{n}

The $n$-th Term Test (Necessary Condition)

If $\sum a_{n}$ converges, then $lim_{n \to \infty} a_{n} = 0$ . Equivalently, if $a_{n} \neq \to 0$ , the series diverges.

n = 1 \sum \infty a_{n} converges ⟹ n \to \infty lim a_{n} = 0.

⟳

a_{n} = s_{n} - s_{n - 1} \to S - S = 0

when both partial sums tend to the same limit

S

The converse of equation [4] is false. $a_{n} \to 0$ does not imply convergence. The harmonic series $\sum 1/ n$ has $a_{n} = 1/ n \to 0$ yet diverges. The $n$ -th term test can only prove divergence, never convergence.

Why It Holds / Derivation

Geometric Series — the prototype

Consider $\sum_{n = 0}^{\infty} r^{n}$ with $∣ r ∣ < 1$ .

Step 1. Write the partial sum and multiply by

r

s_{N} = n = 0 \sum N r^{n}, r s_{N} = n = 0 \sum N r^{n + 1} = n = 1 \sum N + 1 r^{n} .

⟳shift the index by one inside the second sum.

Step 2. Subtract to telescope:

s_{N} - r s_{N} = 1 - r^{N + 1} ⟹ s_{N} = \frac{1 - r ^{N + 1}}{1 - r} .

⟳nearly all middle terms cancel, leaving the first term

1

and the last

- r^{N + 1}

Step 3. Take the limit. Since

∣ r ∣ < 1

r^{N + 1} \to 0

, so

n = 0 \sum \infty r^{n} = N \to \infty lim \frac{1 - r ^{N + 1}}{1 - r} = \frac{1}{1 - r} .

⟳only the

r^{N + 1}

term depends on

N

and it vanishes; the rest is constant.

If $∣ r ∣ \geq 1$ then $r^{n} \neq \to 0$ , so by the $n$ -th term test the series diverges. $■$

Divergence of the Harmonic Series

Step 1. Group terms in blocks of doubling length:

n = 1 \sum \infty \frac{1}{n} = 1 + \frac{1}{2} + (\frac{1}{3} + \frac{1}{4}) + (\frac{1}{5} + \dots + \frac{1}{8}) + \dots

⟳the

k

-th block runs from

n = 2^{k - 1} + 1

n = 2^{k}

and contains

2^{k - 1}

terms.

Step 2. Bound each block below by its smallest term times its count:

n = 2^{k - 1} + 1 \sum 2^{k} \frac{1}{n} \geq 2^{k - 1} \cdot \frac{1}{2 ^{k}} = \frac{1}{2} .

⟳each of the

2^{k - 1}

terms is at least

1/ 2^{k}

, the value of the largest index.

Step 3. Hence

s_{2^{k}} \geq 1 + \frac{k}{2} \to \infty

. The partial sums are unbounded, so the series diverges.

■

Derivation hint (Oresme, ~1350): each parenthesized block contributes at least $1/2$ , and there are infinitely many blocks.

Absolute Convergence Implies Convergence

If $\sum ∣ a_{n} ∣$ converges, then $\sum a_{n}$ converges, and $∣ \sum a_{n} ∣ \leq \sum ∣ a_{n} ∣$ .

⟳by the Cauchy criterion,

\sum_{k = n + 1}^{m} a_{k} \leq \sum_{k = n + 1}^{m} ∣ a_{k} ∣ < ε

; completeness of

R

finishes it.

Augustin-Louis Cauchy | 1789–1857

Cauchy gave the modern, limit-based definition of series convergence in his Cours d'Analyse (1821), ending two centuries of confident but sometimes nonsensical manipulation of divergent series. He introduced the partial-sum definition, the Cauchy criterion, the ratio and root tests, and the Cauchy product of two series. His insistence that a series means its limit of partial sums — and nothing else — turned analysis from a collection of formal tricks into a rigorous deductive science. Where Euler had written $1 - 1 + 1 - 1 + \dots = \frac{1}{2}$ , Cauchy declared the question ill-posed: the partial sums oscillate and have no limit.

Building Intuition

Picture the partial sums as the position of a walker on the number line. Each term $a_{n}$ is a step: positive terms step right, negative terms step left. The series converges if the walker eventually settles arbitrarily close to a fixed point and stays there. It diverges if the walker runs off to infinity (like $\sum 1/ n$ , who keeps inching right forever) or paces back and forth without settling (like $\sum (- 1)^{n}$ ).

Convergence requires the steps to shrink ( $a_{n} \to 0$ ) — but that alone isn't enough. The steps must shrink fast enough that their accumulated total stays bounded. The harmonic walker takes shrinking steps yet covers infinite distance, because $1/ n$ shrinks too slowly. Compare to $\sum 1/ n^{2}$ : steps shrink fast enough, total distance is finite ( $= π^{2} /6$ ).

Partial sums of a convergent (geometric, gold) vs divergent (harmonic, gray) series

Worked Examples

Geometric Series with Numbers

Given: A bouncing ball is dropped from height

h_{0} = 10 m

; after each bounce it returns to

0.6

of its previous height. Find the total vertical distance travelled before coming to rest. Step 1. First drop:

10

m. Each subsequent bounce goes up and comes down, contributing

2 \times

the bounce height. Bounce heights:

10 (0.6), 10 (0.6)^{2}, \dots

D = 10 + 2 n = 1 \sum \infty 10 (0.6)^{n} .

Step 2. Evaluate the geometric series with first term

a = 10 (0.6) = 6

and ratio

r = 0.6

n = 1 \sum \infty 6 (0.6)^{n - 1} = \frac{6}{1 - 0.6} = 15 \Rightarrow n = 1 \sum \infty 10 (0.6)^{n} = 15.

Step 3. Total:

D = 10 + 2 (15) = 40.

Answer:

D = 40 m

. The infinite number of bounces yields a finite total distance because

∣ r ∣ < 1

Telescoping Series with Numbers

Given: Compute

n = 1 \sum \infty \frac{1}{n ( n + 1 )}

. Step 1. Partial fractions:

\frac{1}{n ( n + 1 )} = \frac{1}{n} - \frac{1}{n + 1} .

Step 2. Write the partial sum and cancel:

s_{N} = n = 1 \sum N (\frac{1}{n} - \frac{1}{n + 1}) = (1 - \frac{1}{2}) + (\frac{1}{2} - \frac{1}{3}) + \dots + (\frac{1}{N} - \frac{1}{N + 1}) = 1 - \frac{1}{N + 1} .

Step 3. Take the limit:

N \to \infty lim s_{N} = 1 - 0 = 1.

Answer:

n = 1 \sum \infty \frac{1}{n ( n + 1 )} = 1

(dimensionless). The structure of telescoping makes the exact sum visible.

Common Misconceptions

Five frequent errors:

" $a_{n} \to 0$ means the series converges." False — see the harmonic series.
"You can rearrange terms freely." Only for absolutely convergent series. Conditional convergence is fragile.
"Convergence depends on the first few terms." No — convergence is a tail property; finitely many terms change the sum but never the convergence status.
" $\sum a_{n} =$ the formula $1/ (1 - r)$ always." That formula is valid only for $∣ r ∣ < 1$ .
"Divergent series have no value." Some can be assigned values via summation methods (Cesàro, Abel), but these are not the ordinary sum.

Riemann Rearrangement

The alternating harmonic series converges conditionally:

n = 1 \sum \infty \frac{( - 1 ) ^{n + 1}}{n} = 1 - \frac{1}{2} + \frac{1}{3} - \frac{1}{4} + \dots = ln 2 \approx 0.693.

But by Riemann's rearrangement theorem, its terms can be reordered to sum to any real number

L

— or to

+ \infty

, or

- \infty

. For instance, taking enough positive terms to exceed

5

, then one negative, then more positives, etc., drives the sum to

5

. This is impossible for absolutely convergent series, whose sum is rearrangement-invariant. Commutativity of addition does not survive the limit process for conditional convergence.

Connections

Infinite series sit at the crossroads of analysis. Through Power Series, series become functions and bridge to Taylor Series expansions of $e^{x}$ , $sin x$ , and $ln (1 + x)$ . The Convergence Tests (ratio, root, comparison, integral, alternating) provide the practical machinery for deciding convergence. The Integral Test links series sums to Improper Integrals, so $\sum 1/ n^{p}$ converges exactly when $\int_{1}^{\infty} x^{- p} d x$ does ( $p > 1$ ). Fourier Series generalize the idea to expansions in trigonometric functions and underpin Partial Differential Equations. The notion of Uniform Convergence determines when term-by-term differentiation and integration are valid, connecting to Sequences of Functions. Finally, the Cauchy Criterion for series is the same completeness principle that defines the real numbers via Cauchy Sequences.

Real-World Applications

Compound interest & annuities (finance): the present value of a perpetuity is a geometric series, $\sum P / (1 + r)^{n} = P / r$ .
Signal processing (Fourier series): periodic signals are decomposed into harmonics; JPEG and MP3 compression discard high-frequency series terms.
Numerical computation of constants: $π$ via the Leibniz/Machin series, $e$ via $\sum 1/ n!$ , logarithms via power series.
Quantum mechanics (perturbation theory): energy corrections are computed as a power series in the coupling constant; the famous QED expansion in $α \approx 1/137$ .
Probability (generating functions): $\sum p_{n} x^{n}$ encodes a discrete distribution; expectations and variances fall out as derivatives at $x = 1$ .
Engineering (control systems): the $z$ -transform $\sum x [n] z^{- n}$ is a power series whose region of convergence determines stability.
Special relativity & physics expansions: $γ = (1 - v^{2} / c^{2})^{- 1/2}$ expanded as a binomial series gives the kinetic-energy correction $\frac{1}{2} m v^{2} + \frac{3}{8} m v^{4} / c^{2} + \dots$ .

Historical Development

c. 250 BCE — Archimedes sums a geometric series ( $\sum 4^{- n} = 4/3$ ) to find the area of a parabolic segment via exhaustion.
c. 1350 — Nicole Oresme proves the harmonic series diverges by the doubling-block argument (equations [8]–[9]).
1650 — Pietro Mengoli poses the Basel problem ( $\sum 1/ n^{2} = ?$ ) and studies telescoping series.
1668 — Nicholas Mercator & James Gregory discover power series for $ln (1 + x)$ and $arctan x$ .
1734 — Leonhard Euler solves the Basel problem: $\sum 1/ n^{2} = π^{2} /6$ , and pioneers manipulation of divergent series.
1812 — Carl Friedrich Gauss gives the first fully rigorous convergence study (the hypergeometric series), introducing the Gauss test.
1821 — Augustin-Louis Cauchy defines convergence via partial sums in Cours d'Analyse; states the ratio/root tests and Cauchy criterion.
1837 — Peter Gustav Lejeune Dirichlet distinguishes absolute from conditional convergence and identifies rearrangement dangers.
1854 — Bernhard Riemann proves the rearrangement theorem for conditionally convergent series; advances Fourier theory.
1890s — Cesàro, Abel, Borel develop rigorous summation methods that assign values to certain divergent series.

Subtleties & Limits

Where the theory still bites:

Divergent series with meaning. Expressions like $ζ (- 1) = 1 + 2 + 3 + \dots = - \frac{1}{12}$ (via analytic continuation of the Riemann zeta function) appear in string theory and the Casimir effect. Making such "sums" rigorous requires regularization, not ordinary convergence — the boundary between formal and meaningful remains delicate.
Asymptotic series. Many physical expansions (QED, WKB) are divergent yet give superb numerical results when truncated optimally. The theory of resurgence and Borel summation attempts to recover exact answers from divergent data, an active research area.
Conditional convergence in higher dimensions. Multiple series and lattice sums (e.g., Madelung constants for crystal energies) depend on summation order in subtle ways; their "correct" values require careful regularization, and some convergence questions remain open.
Open arithmetic questions. Whether certain rapidly converging series produce irrational or transcendental constants (e.g., values of $ζ (2 k + 1)$ beyond $ζ (3)$ , proven irrational by Apéry 1978) is largely unresolved.