WEBVTT

1
00:00:00.000 --> 00:00:04.934
Here is the problem: Σ 1 / n ^ 2, n equals 1 to ∞.

2
00:00:05.384 --> 00:00:11.026
Decide whether Σ 1 / n squared (n from 1 to ∞) converges or diverges.

3
00:00:11.826 --> 00:00:27.534
Write the series. Σ a(n) for n equals 1, 2, 3, … with a(n) equals 1 / n squared. First terms: 1, 0.25, 0.111111, 0.0625, ….

4
00:00:27.984 --> 00:00:37.899
nth-term test. lim a(n) equals 0, so the divergence test is inconclusive; the terms do shrink, so a finer test is needed.

5
00:00:38.349 --> 00:01:03.102
Limit comparison with a p-series. Positive terms: the terms shrink roughly like 1/n squared (estimated from a(2N)/a(N)), so compare with Σ 1/n squared: lim a(n) times n squared equals 1. The limit is a positive finite number, so the series behaves exactly like the p-series Σ 1/n squared, which converges because p greater than 1.

6
00:01:03.552 --> 00:01:30.510
Estimate the sum numerically. Partial sums S N were computed for several N and extrapolated to N → ∞ (Richardson extrapolation; alternating terms are averaged first). Two independent extrapolations agree to better than 10 to the power minus 10, giving S approximately 1.644934067. Any closed form shown is a numerical identification, not a proof.

7
00:01:31.310 --> 00:01:45.672
So the answer is The series converges (limit comparison with Σ 1/n squared); sum approximately 1.644934067 (matches pi squared/6 to 10+ digits).

8
00:01:46.121 --> 00:02:02.375
A series converges when its partial sums approach a finite number. The terms must tend to 0 for that to happen, but that alone is not enough (the harmonic series diverges), so a comparison, ratio, root, alternating or integral test is used to decide.

9
00:02:03.175 --> 00:02:24.026
Partial sums (supporting evidence, not a proof): S₁₀ equals 1.5497677, S₁₀₀ equals 1.6349839, S₁₀₀₀ equals 1.6439346, S₁₀₀₀₀ equals 1.6448341.