Partial Sums: Where Series Meaning Lives
An infinite series Σa(n) is defined through its partial sums: SN = a(1) + a(2) + ⋯ + a(N), the running totals after finitely many terms. The series "equals" a number S exactly when those partial sums converge to S as N grows, infinite addition is a limit of finite additions, not a completed act, there is no moment when the last term goes in, only totals that settle or fail to. That definition makes partial sums the most concrete object in the whole theory: every convergence test, every sum formula, every paradox about infinite series is ultimately a statement about the sequence S₁, S₂, S₃, …
This calculator computes that sequence directly for any general term you can type, sigma notation, evaluated, plotted, and diagnosed.
How to Use This Calculator
Enter the general term as an expression in n the starting index, and how many terms to add. The calculator lists the first few terms (a quick check that your formula means what you intended), reports SN plots the running totals, and adds a tail check, the size of the next omitted term, a rough gauge of whether the sum has stabilized. The chips carry four instructive personalities worth trying in sequence: fast convergence (1/n²), deceptive divergence (1/n), a clean closed form (Σn), and telescoping collapse (1/(n(n+1))).
Worked Example: the Basel Series
Sum Σ1/n². The partial sums crawl upward, S₁₀ ≈ 1.5498, S₅₀ ≈ 1.6251, S₁₀₀ ≈ 1.6350, closing in on a limit Euler identified in 1735 to widespread astonishment:
Σn=1∞ 1/n² = π²/6 ≈ 1.644934
Why π appears in a sum of reciprocal squares is one of mathematics' delicious surprises (Euler connected it to the sine function's infinite product). The convergence is genuine but leisurely, the tail after N terms is roughly 1/N, so even a thousand terms deliver only three decimals, visible in the plot's slow flattening.
Reading a Partial-Sum Plot
| Plot behavior | Diagnosis | Example chip |
|---|---|---|
| Flattens toward a level | Converges to that level | 1/n², telescoping |
| Climbs without leveling | Diverges (perhaps slowly) | harmonic 1/n |
| Grows along a curve | Diverges with a growth law | Σn (parabolic growth) |
| Zigzags with shrinking hops | Alternating convergence | see the Alternating Series Calculator |
The harmonic series deserves its infamy: its terms shrink to zero, yet the partial sums pass any bound, S grows like ln(N), so reaching 20 takes over 270 million terms. "Terms → 0" is necessary for convergence but nowhere near sufficient, and no example teaches that faster.
Common Mistakes to Avoid
- Confusing the sequence with the series. a(n) → 0 is about the terms; ΣA(n) converging is about their accumulating total. The harmonic series splits the two cleanly.
- Trusting a stabilized-looking plot too much. Slowly diverging sums flatten locally, the harmonic partial sums gain only 0.7 per doubling of N. Pair numerical evidence with a test (ratio p-series comparison) for a verdict.
- Off-by-one index errors. Starting at n = 0 versus n = 1 changes every partial sum by the first term, and 1/n² at n = 0 is undefined. The starting-index field exists precisely for this.
- Reading S_N as the infinite sum. It's an approximation whose quality depends on the tail; the tail-check step estimates how much is still missing.
Real-World Applications
Partial sums are how infinite series are actually used no computer ever adds infinitely many terms. Every numerical evaluation of a Taylor expansion, Fourier synthesis of a sound wave, or probability from an infinite discrete distribution is a partial sum cut off when the tail dips below tolerance. Understanding how fast S_N approaches S (the convergence rate) is a practical engineering question: it decides how many terms a calculator's sine routine needs, and how many harmonics an audio synthesizer must sum before the ear can't tell.
Discrete accumulations in daily life are literal partial sums: cumulative interest across compounding periods, total rainfall season-to-date, running project costs, an athlete's accumulating points. And in the mathematics of algorithms, partial sums of series like Σ1/n (the harmonic numbers) appear in average-case analyses of sorting and hashing, the ln(N) growth this calculator displays for the harmonic chip is exactly why those algorithms scale as they do.
Frequently Asked Questions
What is a telescoping series?
One whose terms rewrite as differences, 1/(n(n+1)) = 1/n − 1/(n+1), so that consecutive terms cancel and S_N collapses to 1 − 1/(N+1). The infinite sum is exactly 1, and the chip's plot shows the swift, clean approach. Telescoping and geometric series are the two families with exact partial-sum formulas.
Can this calculator prove convergence?
No, finite evidence never proves an infinite claim. It provides the data (terms, totals, tail size) that make the right conjecture obvious; certainty comes from the tests: ratio root p-series or comparison arguments.
What closed forms exist for common finite sums?
The classics: Σn = N(N+1)/2 (the chip verifies it, try N = 100, giving 5050, Gauss's schoolboy legend), Σn² = N(N+1)(2N+1)/6, Σn³ = [N(N+1)/2]², and geometric sums via a(1−rᴺ)/(1−r). These identities power everything from Riemann-sum derivations to algorithm analysis.
Why does my sum look wrong when the term is undefined at the start index?
Terms like 1/n² or 1/(n(n+1)) blow up at n = 0, the calculator halts with a message rather than silently adding garbage. Shift the start index to the series' natural first term.
How do alternating signs enter the term expression?
Write the sign into a(n) directly, for example (−1)^n or (−1)^(n+1) as a factor... noting that non-integer powers of negative numbers aren't real. Cleaner: use the dedicated Alternating Series Calculator which manages the sign pattern and its error bound for you.
How is a partial sum different from a Riemann sum?
Both are finite approximations of an infinite object, but partial sums approximate a series (adding a sequence's terms) while Riemann sums approximate an integral (adding rectangle areas). The integral test connects them: for decreasing positive terms, the series and the corresponding integral converge together, with partial sums and Riemann sums sandwiching each other.
Why can a partial sum grow without the series converging?
Partial sums of a divergent series may increase without bound, as with the harmonic series whose sums grow like the natural logarithm. Convergence requires the partial sums to approach a finite limit, not merely to keep increasing by ever-smaller amounts.