Geometric Progression: nth Term & Partial Sum

In an arithmetic progression (AP), consecutive terms differ by a constant additive value called the common difference d. Example: 3, 7, 11, 15 has d = 4 — you always add 4. In a geometric progression (GP), consecutive terms differ by a constant multiplicative factor called the common ratio r. Example: 3, 6, 12, 24 has r = 2 — you always multiply by 2. The practical consequence is enormous: APs grow linearly (doubling n roughly doubles the nth term), while GPs grow exponentially (doubling n squares the nth term). This is why compound interest, population explosions, and viral spread are modeled with GPs, not APs. To test which type a sequence is, check whether consecutive differences are constant (AP) or consecutive ratios are constant (GP). A sequence can be neither if neither condition holds.

When the absolute value of the common ratio is strictly less than 1 (−1 < r < 1, r ≠ 0), each successive term is smaller than the previous one, shrinking toward zero. As a result, the infinite sum converges to a finite value given by the formula S∞ = a₁ / (1 − r). For example, with a₁ = 12 and r = 0.25, S∞ = 12 / 0.75 = 16 — the entire infinite series sums to exactly 16, even though it has infinitely many terms. This convergence property is foundational in physics (total distance of a bouncing ball), finance (present value of a perpetuity where payments shrink over time), and signal processing (total energy of decaying echoes). If |r| ≥ 1, the terms do not shrink and the series diverges — no finite infinite sum exists.

Zero is the only excluded value — r = 0 would collapse every term after the first to zero, making the sequence trivial and the ratio undefined (you'd be dividing by zero when computing aₙ₊₁/aₙ). Every other value is valid. A negative r produces an alternating sequence where terms flip between positive and negative signs each step: e.g., a₁ = 4, r = −2 gives 4, −8, 16, −32, 64. The standard formulas still apply, but you must track signs carefully — partial sums oscillate rather than monotonically increase. A fractional r between 0 and 1 (e.g., r = 0.5) produces a decreasing positive sequence converging toward zero; if it's between −1 and 0, you get a decreasing alternating sequence. Both cases use the same formulas, and if |r| < 1, the infinite-sum convergence formula S∞ = a₁/(1−r) also applies.

Compound interest is the most widely applied real-world instance of a geometric progression. If you invest principal P at a periodic interest rate i (e.g., annual rate expressed as a decimal), the balance at the end of each period forms a GP: P, P(1+i), P(1+i)², P(1+i)³, … Here a₁ = P and r = (1+i). The balance after n periods is aₙ₊₁ = P · (1+i)ⁿ — the standard compound interest formula. The partial sum Sₙ applies when equal periodic deposits are made, yielding the future value of an ordinary annuity. For example, $2,500 invested at 5% annually for 15 years: a₁₆ = 2,500 · 1.05¹⁵ ≈ $5,197. Understanding this GP structure lets you deconstruct any compound-growth financial product — bonds, mortgages, index funds — into its mathematical fundamentals.

If you know term aₘ at position m and term aₙ at position n (with n > m), the common ratio is: r = (aₙ / aₘ)^(1 / (n − m)). For example, if the 3rd term is 18 and the 7th term is 2,916: r = (2,916 / 18)^(1/4) = 162^(0.25) = 3. Once you have r, recover the first term with a₁ = aₘ / r^(m−1) = 18 / 3² = 2. This technique is essential in growth-rate estimation from sparse data — for instance, estimating a company's annual revenue growth rate given only revenues from two specific years separated by several years. It is also used in carbon dating calculations to find decay rates from two measured isotope concentrations.

The geometric mean of n positive numbers is the nth root of their product: G = (a₁ · a₂ · … · aₙ)^(1/n). In a GP, the geometric mean of all n terms equals the middle term (for odd n) or the square root of the product of the two middle terms (for even n). Example: in the 5-term GP 2, 6, 18, 54, 162, the geometric mean = (2 · 6 · 18 · 54 · 162)^(1/5) = (1,679,616)^0.2 = 18 = a₃. This property is directly applied in finance through the Compound Annual Growth Rate (CAGR): if a portfolio grows from $10,000 to $17,000 over 5 years, CAGR = (17,000/10,000)^(1/5) − 1 ≈ 11.2% — the geometric mean of the annual growth ratios. The geometric mean always understates the arithmetic mean for volatile data, making it a more conservative and accurate measure of average investment returns.

GPs appear throughout algorithm analysis. In binary search, each comparison halves the remaining search space: n, n/2, n/4, …, 1 — a GP with r = 0.5 and n = log₂(N) terms, directly explaining the O(log N) time complexity. In merge sort, the number of sub-problems at recursion depth k is 2ᵏ — a GP with r = 2 — and the total work summed across all log₂(N) levels gives the O(N log N) complexity via a GP partial sum. Dynamic arrays (like Python lists or Java ArrayLists) double their capacity each time they run out of space, creating a GP of allocation sizes: 1, 2, 4, 8, 16, … The total memory allocated through k reallocations is S = 2ᵏ⁺¹ − 1, a GP partial sum, which proves amortized O(1) cost per insertion. Recognizing these GP structures is essential for algorithm design and complexity proofs.

Yes. The product of the first n terms of a GP is Pₙ = a₁ⁿ · r^(n(n−1)/2). This follows from multiplying all terms: (a₁)(a₁r)(a₁r²)…(a₁r^(n−1)) = a₁ⁿ · r^(0+1+2+…+(n−1)) = a₁ⁿ · r^(n(n−1)/2). Example: for a₁ = 2, r = 3, n = 5: P₅ = 2⁵ · 3^10 = 32 · 59,049 = 1,889,568. A useful shortcut: for odd n, Pₙ = (middle term)ⁿ; for even n, Pₙ = (√(a₁ · aₙ))ⁿ. The product formula is used in statistics (computing geometric means efficiently), information theory (calculating joint probabilities of independent geometric processes), and number theory.

An infinite geometric series a₁ + a₁r + a₁r² + a₁r³ + … converges to a finite sum only when |r| < 1. The formula is S∞ = a₁ / (1 − r). When |r| ≥ 1, the terms do not approach zero and the sum grows without bound (diverges). Practical examples: a ball dropped from 8 m rebounding to 60% each time has total downward travel S∞ = 8/(1−0.6) = 20 m (plus upward paths). A perpetuity paying $500 annually discounted at 5% has present value S∞ = 500/0.05 = $10,000. The convergence condition |r| < 1 is non-negotiable — applying the infinite sum formula when |r| ≥ 1 produces meaningless results. Always verify the ratio before using S∞.

Divide each term by the preceding term. If the ratio aₙ₊₁/aₙ is identical for every consecutive pair, the sequence is geometric and that constant is r. Example: for the sequence 5, 15, 45, 135: 15/5 = 3, 45/15 = 3, 135/45 = 3 → confirmed GP with r = 3. If the ratios vary, the sequence is not a pure GP. In real-world data, ratios are rarely perfectly constant due to measurement noise. In that case, compute the geometric mean of all consecutive ratios as an estimated r, or use log-linear regression: taking logarithms of the terms converts a GP into an AP (log(aₙ) = log(a₁) + (n−1)·log(r)), and the slope of the best-fit line equals log(r). This log-linearization technique is standard in epidemiology, finance, and physics for fitting exponential growth models to noisy data.

Several errors recur frequently. First, off-by-one in the exponent: the nth term is a₁ · r^(n−1), not a₁ · rⁿ. For n = 1, a₁ = a₁ · r⁰ = a₁ — if you use rⁿ you get a₁ · r, which is actually a₂. Second, misidentifying a₁: ensure a₁ is truly the first term. If a problem gives you a₃ = 12 and r = 2, then a₁ = 12 / 2² = 3, not 12. Third, applying the r ≠ 1 sum formula when r = 1: Sₙ = a₁(rⁿ−1)/(r−1) is undefined when r = 1 (division by zero); the correct formula is Sₙ = n · a₁. Fourth, confusing periodic and annual rates in finance: if interest compounds monthly at 6% APR, r = 1 + 0.06/12 = 1.005 per month, not 1.06. Using the wrong r compounds errors exponentially. Fifth, forgetting to include the initial drop in bouncing-ball problems — the first downward path is not part of the rebound GP.

A geometric progression is the discrete analog of a continuous exponential function. The nth term aₙ = a₁ · r^(n−1) has the same mathematical form as the exponential function f(x) = a · bˣ, where n plays the role of x and r plays the role of b. Taking the logarithm of both sides: log(aₙ) = log(a₁) + (n−1) · log(r), which is a linear equation in n — this is why plotting GP data on a logarithmic scale produces a straight line (log-linear or semi-log plot). This relationship is exploited constantly: epidemiologists plot case counts on log scales to spot exponential growth, engineers use Bode plots (log-log scale) to visualize frequency responses, and financial analysts use log-scale charts to compare percentage growth. Conversely, logarithms solve GP problems involving unknown n: if you need to find when aₙ reaches a target value T, solve n = 1 + log(T/a₁) / log(r).