If the absolute-value series converges, must the original follow?
A finite absolute total settles everything. The signed partial sums wobble inside it and land on one point. In reverse, alternating signs can rescue a series whose absolute total diverges.
For example, picture a π§ person stepping forward, then a little back. If all the step lengths add to a finite total, the walker ends up standing somewhere definite.
n=60: since the absolute partial sums Tnβ are Cauchy, so are Snβ.Show proof
Assume ββ£akββ£ converges. Its partial sums Umβ=βk=1mββ£akββ£ converge, hence are Cauchy: for every Ξ΅>0 there is an N such that m>nβ₯N implies β£UmββUnββ£=βk=n+1mββ£akββ£<Ξ΅.
For the original partial sums Smβ, the triangle inequality gives β£SmββSnββ£=ββk=n+1mβakβββ€βk=n+1mββ£akββ£<Ξ΅. Hence (Snβ) is Cauchy too.
R is complete β every Cauchy sequence converges (the completeness axiom from the supremum scene, limits chapter). Hence βakβ converges.
n=60: S60βββ0.68488 heads to βln2ββ0.69315, while T60β=H60ββ4.67987 grows without bound.Show proof
ββk(β1)kββ=βk1β is the harmonic series, hence diverges (the Oresme proof in the harmonic scene). At n=60 already T60β=H60ββ4.67987.
But bkβ=k1ββ0, so the alternating series test gives convergence of βk(β1)kβ (the alternating-series estimate from the Leibniz scene). The sum is βln2ββ0.693147, and S60βββ0.68488.
It converges but not absolutely β conditional convergence, exactly the case where rearrangement can change the sum.