Geometric Series & Convergence

A geometric series is a sum of terms with a constant ratio between consecutive terms: $a + a r + a r^{2} + \dots$ . Key result: the infinite series converges to $\frac{a}{1 - r}$ if and only if $∣ r ∣ < 1$ ; otherwise it diverges. Applies anywhere repeated proportional growth or decay accumulates — finance, physics, probability, fractals. Studied since antiquity (Zeno, Archimedes); rigorously settled in the 19th century.

A geometric series is the sum of the terms of a geometric sequence, in which each term is obtained from the previous one by multiplying by a fixed number $r$ , called the common ratio. These series matter because they are the simplest infinite sums whose convergence we can fully understand with an explicit formula, and they serve as the benchmark against which more complicated series are tested. They model compound interest, radioactive decay, bouncing balls, and repeating decimals. The surprising fact: an infinite collection of positive numbers — infinitely many things added together — can have a perfectly finite total, a paradox that troubled the Greeks for two thousand years before being resolved by the modern theory of limits.

The Core Insight

The whole subject pivots on one algebraic trick: multiply the partial sum by $r$ and subtract. Almost everything telescopes, leaving only the first and last terms. Whether the "last term" $r^{n}$ shrinks to zero or blows up — governed entirely by whether $∣ r ∣ < 1$ — decides convergence. The behaviour of a geometric series is completely determined by a single number $r$ .

Formal Statement

Geometric Sequence and Series

A geometric sequence is an ordered list $a, a r, a r^{2}, a r^{3}, \dots$ where:

$a$ is the first term (any real number, $a \neq = 0$ ),
$r$ is the common ratio (any real number), found by $r = \frac{any term}{previous term}$ .

The $n$ -th term is:

t_{n} = a r^{n - 1}

⟳Start with

t_{1} = a

and multiply by

r

exactly

(n - 1)

times to reach the

n

-th term.

A finite geometric series is the sum of the first $n$ terms:

S_{n} = k = 1 \sum n a r^{k - 1} = a + a r + a r^{2} + \dots + a r^{n - 1}

⟳Add the first

n

terms of the sequence defined in [1], letting

k

index from

1

n

An infinite geometric series is the limit of $S_{n}$ as $n \to \infty$ , written $S = \sum_{k = 1}^{\infty} a r^{k - 1}$ , when that limit exists.

Geometric Series Formulas

For the finite series with $r \neq = 1$ :

S_{n} = \frac{a ( 1 - r ^{n} )}{1 - r}

⟳Compute

S_{n} - r S_{n}

; all interior terms cancel, leaving

a - a r^{n} = a (1 - r^{n})

For the infinite series, the limit exists if and only if $∣ r ∣ < 1$ , in which case:

S = k = 1 \sum \infty a r^{k - 1} = \frac{a}{1 - r}, ∣ r ∣ < 1

⟳Take

n \to \infty

in [3]; since

∣ r ∣ < 1

, the term

r^{n} \to 0

, so

1 - r^{n} \to 1

If $∣ r ∣ \geq 1$ (and $a \neq = 0$ ), the infinite series diverges.

Why It Holds / Derivation

Deriving the Finite Sum Formula

Step 1 — Write the partial sum:

S_{n} = a + a r + a r^{2} + \dots + a r^{n - 1}

⟳This is just equation [2] written out term by term.

Step 2 — Multiply both sides by $r$ :

r S_{n} = a r + a r^{2} + a r^{3} + \dots + a r^{n}

⟳Distribute

r

across every term of [5]; each exponent increases by one.

Step 3 — Subtract [6] from [5]. Every middle term (

a r, a r^{2}, \dots, a r^{n - 1}

) appears in both lines and cancels:

S_{n} - r S_{n} = a - a r^{n}

⟳Line up [5] and [6] by power; the only survivors are the first term of [5] and the last term of [6].

Step 4 — Factor both sides:

S_{n} (1 - r) = a (1 - r^{n})

⟳Factor

S_{n}

on the left and

a

on the right of [7].

Step 5 — Divide by $(1 - r)$ , valid because

r \neq = 1

S_{n} = \frac{a ( 1 - r ^{n} )}{1 - r}

⟳This is legal only when

1 - r \neq = 0

; the case

r = 1

gives

S_{n} = na

separately.

Step 6 — Take the infinite limit. As

n \to \infty

, the behaviour of

r^{n}

decides everything:

If $∣ r ∣ < 1$ : $r^{n} \to 0$ , so $S_{n} \to \frac{a ( 1 - 0 )}{1 - r} = \frac{a}{1 - r}$ .
If $∣ r ∣ > 1$ : $∣ r^{n} ∣ \to \infty$ , so $S_{n}$ has no finite limit — diverges.
If $r = 1$ : $S_{n} = na \to \pm \infty$ — diverges.
If $r = - 1$ : $S_{n}$ oscillates between $a$ and $0$ , never settling — diverges.

n \to \infty lim S_{n} = \frac{a}{1 - r} ⟺ ∣ r ∣ < 1

⟳The convergence rests entirely on

lim_{n \to \infty} r^{n} = 0

when

∣ r ∣ < 1

, which itself follows from exponential decay of magnitudes below 1.

■

Archimedes of Syracuse | c. 287–212 BCE

Archimedes is the earliest mathematician known to have summed an infinite geometric series rigorously. In Quadrature of the Parabola, he computed the area under a parabolic arc by inscribing triangles whose areas form the series $1 + \frac{1}{4} + \frac{1}{16} + \dots$ , with ratio $r = \frac{1}{4}$ . He proved by his "method of exhaustion" — a precursor of limits — that the total equals exactly $\frac{4}{3}$ of the first triangle. He did not use limit notation, but argued that the sum could not be greater or less than $\frac{4}{3}$ , and so must equal it — a flawless application of what we now write as $\frac{a}{1 - r} = \frac{1}{1 - 1/4} = \frac{4}{3}$ .

Building Intuition

Picture walking toward a wall, each step covering half the remaining distance: $\frac{1}{2}$ metre, then $\frac{1}{4}$ , then $\frac{1}{8}$ , and so on. After infinitely many steps you have covered $\frac{1}{2} + \frac{1}{4} + \frac{1}{8} + \dots = 1$ metre exactly — you reach the wall. The terms shrink fast enough that their total is bounded, even though there are infinitely many of them. The condition $∣ r ∣ < 1$ says each new contribution is a fraction of the last, so the "tail" you still owe collapses geometrically. By contrast, if $r = 2$ each step is bigger than the previous, and the sum races to infinity. The single number $r$ acts like a dial: below 1 the series "falls into" a finite total; at or above 1 it escapes.

Convergence of $1 + \tfrac12 + \tfrac14 + \tfrac18 + \cdots \to 2$: partial sums climb but never pass the line $S=2$.

Worked Examples

Finite Sum — Compound Repayment

Given:

a = 200

, common ratio

r = 1.5

, and we want the sum of the first

n = 6

terms:

200 + 300 + 450 + 675 + 1012.5 + 1518.75

Step 1 — Identify the formula. Since

r = 1.5 \neq = 1

, use [3]:

S_{6} = \frac{a ( 1 - r ^{6} )}{1 - r}

Step 2 — Compute $r^{6}$ :

1. 5^{6} = 11.390625

. Step 3 — Substitute:

S_{6} = \frac{200 ( 1 - 11.390625 )}{1 - 1.5} = \frac{200 \times ( - 10.390625 )}{- 0.5}

Step 4 — Evaluate:

S_{6} = \frac{- 2078.125}{- 0.5} = 4156.25

S_{6} = 4156.25 units

Infinite Sum — Bouncing Ball Distance

Given: A ball is dropped from a height of

h = 10

m. Each bounce returns it to

80%

of the previous height, so

r = 0.8

. Find the total vertical distance travelled before it comes to rest. Step 1 — Set up the distances. The ball falls 10 m, then for each bounce it travels up and down the bounce height. Heights bounced:

8, 6.4, 5.12, \dots

(ratio

0.8

), each counted twice (up + down).

D = 10 + 2 (8 + 6.4 + 5.12 + \dots)

Step 2 — Identify the inner infinite series:

a = 8

r = 0.8

. Since

∣ r ∣ = 0.8 < 1

, it converges. Use [4]:

k = 1 \sum \infty 8 (0.8)^{k - 1} = \frac{8}{1 - 0.8} = \frac{8}{0.2} = 40

Step 3 — Assemble total distance:

D = 10 + 2 (40) = 10 + 80 = 90

D = 90 m

Common Misconceptions

The boundary cases are deadly. The convergence condition is the strict inequality

∣ r ∣ < 1

. Students routinely write

∣ r ∣ \leq 1

— but at

r = 1

the series is

a + a + a + \dots

(diverges to infinity) and at

r = - 1

it is

a - a + a - a + \dots

, whose partial sums oscillate and never settle. Neither endpoint converges. Also beware: formula [4] is only for infinite series with

∣ r ∣ < 1

; never apply

\frac{a}{1 - r}

to a finite sum or a divergent one.

Grandi's Series — When the Formula Lies

Blindly plugging $r = - 1$ and $a = 1$ into [4] gives:

\frac{1}{1 - ( - 1 )} = \frac{1}{2}

suggesting

1 - 1 + 1 - 1 + \dots = \frac{1}{2}

. This is false under the standard definition of a sum. The partial sums are

S_{1} = 1, S_{2} = 0, S_{3} = 1, S_{4} = 0, \dots

— they bounce forever between 0 and 1 and have no limit. The series diverges. The value

\frac{1}{2}

is only meaningful under exotic "summation methods" (Cesàro, Abel) beyond ordinary convergence. Moral: the formula [4] is valid only when $∣ r ∣ < 1$ ; outside that range it produces meaningless numbers.

Connections

Geometric series sit at the crossroads of several topics. They are the prototype for Infinite Series and Convergence Tests, where the ratio test for a general series $\sum b_{n}$ literally compares it to a geometric series via $lim ∣ b_{n + 1} / b_{n} ∣$ . They underpin Repeating Decimals as Fractions, since $0. \overline{3} = \frac{3}{10} + \frac{3}{100} + \dots$ is geometric with $r = \frac{1}{10}$ . The closely related Arithmetic Series uses a constant difference instead of a constant ratio. And the function $\frac{1}{1 - x} = \sum x^{n}$ is the simplest Power Series, the seed from which Taylor expansions grow.

Real-World Applications

Compound interest & annuities: The future value of regular equal deposits earning fixed interest is a finite geometric series; mortgage and loan formulas are derived directly from [3].
Radioactive decay & half-lives: The remaining amount after successive half-lives forms $1, \frac{1}{2}, \frac{1}{4}, \dots$ with $r = \frac{1}{2}$ .
Drug dosing (pharmacokinetics): Repeated doses with constant elimination fraction give a geometric build-up toward a steady-state concentration $\frac{a}{1 - r}$ .
Fractal geometry: The Koch snowflake's perimeter and area, and the total length of self-similar structures, are computed as geometric series.
Digital signal processing: The impulse response of a first-order recursive (IIR) filter is geometric; its gain is $\frac{1}{1 - r}$ .
Economics — the multiplier effect: When money is re-spent at a fixed fraction $r$ , total economic activity is $\frac{1}{1 - r}$ times the initial injection.
Computer science: Analysing the cost of repeatedly halving a problem (binary search, divide-and-conquer) and the geometric capacity growth of dynamic arrays.

Historical Development

c. 450 BCE — Zeno of Elea poses the dichotomy paradox (to cross a room you must first cross half, then half again…), implicitly invoking $\frac{1}{2} + \frac{1}{4} + \dots$ and challenging whether infinitely many steps can be completed.
c. 350 BCE — Aristotle distinguishes potential from actual infinity, an early attempt to dissolve Zeno's puzzle.
c. 250 BCE — Archimedes rigorously sums $\sum (1/4)^{k} = \frac{1}{3}$ in Quadrature of the Parabola via exhaustion.
14th century — Nicole Oresme and the Oxford Calculators study sums of series, including the divergence of the harmonic series, sharpening the contrast with geometric convergence.
1593 — François Viète publishes an explicit closed form for the infinite geometric series in Variorum de rebus mathematicis.
1668 — Nicolaus Mercator and others expand $ln (1 + x)$ using geometric-series integration, linking the series to logarithms.
1715 — Brook Taylor generalises power-series expansions, with the geometric series $\frac{1}{1 - x} = \sum x^{n}$ as the canonical example.
1821 — Augustin-Louis Cauchy, in Cours d'analyse, gives the modern rigorous definition of convergence via limits of partial sums, finally settling exactly when geometric (and all) series converge.
1890s — Ernesto Cesàro formalises generalised summation, assigning $\frac{1}{2}$ to Grandi's divergent series — clarifying that such values lie outside ordinary convergence.

Subtleties & Limits

While the real geometric series is completely solved, its generalisations open deep waters. Replacing the ratio with a complex number $z$ gives $\sum z^{n} = \frac{1}{1 - z}$ , convergent precisely on the open unit disc $∣ z ∣ < 1$ — the boundary circle $∣ z ∣ = 1$ is a frontier of subtle behaviour studied in complex analysis. In $p$ -adic number systems the convergence condition flips entirely, so $1 + 2 + 4 + 8 + \dots = - 1$ becomes legitimate. And the broader question of when a divergent series can be assigned a consistent, useful "sum" (Borel, Cesàro, Abel summation; Ramanujan's $1 + 2 + 3 + \dots = - \frac{1}{12}$ ) remains an active area linking analysis, physics, and number theory — there is no single canonical answer, only methods chosen to fit the problem.