Absolute and conditional convergence

Apply the triangle inequality to a finite block of signed or complex terms. The magnitude of the block sum is bounded by the sum of magnitudes in that block. If the magnitude series converges, its late tails become uniformly small, hence so do all late blocks of the original series. The partial sums are Cauchy and therefore converge in the complete scalar field. This proof shows exactly why completeness enters the usual implication.

Conditional convergence means the original scalar series converges but the magnitude series diverges. It does not mean the original series is only approximately convergent or that its answer is uncertain in the prescribed order. Rather, some operations need stronger hypotheses than convergence in one fixed order. Absolute convergence supplies a finite tail budget independent of signs, while conditional convergence requires preserving the particular cancellation structure used in its proof.

For decreasing nonnegative bₙ tending to zero, pair adjacent terms of the alternating series. Its even partial sums increase, its odd partial sums decrease, and the two bounds differ by a term tending to zero. They therefore enclose one common limit. Stopping after N terms puts the true sum between that partial sum and the next one, giving an error no larger than the next omitted magnitude.

All parts of the hypothesis matter. Alternating signs alone do not suffice, and shrinking magnitudes without a suitable monotonicity or other cancellation estimate do not automatically give the next-term bound. By contrast, reciprocals of n satisfy the alternating criterion while their magnitude series is harmonic and divergent. This furnishes a canonical conditional example. A slowly changing envelope can produce convergent cancellation even when direct positive comparison rules out absolute convergence.

Summation by parts is the discrete counterpart of integration by parts. If the cumulative sums of aₙ stay bounded and bₙ decreases to zero, the series of products aₙbₙ converges by Dirichlet’s test. Bounded cumulative sums encode cancellation; decreasing bₙ controls its changing envelope. This viewpoint covers many trigonometric series whose signs do not simply alternate and explains why phase matters as much as term magnitude.

The spectral lab uses finite sums of sine modes weighted by reciprocal powers. For fixed position away from multiples of a full turn, the unweighted sine partial sums are bounded, and decreasing weights can make the infinite series converge even when their absolute majorant diverges. This is a pointwise argument, not a uniform one on the whole period. Predict how weak coefficient decay changes the profile, while remembering that finitely many modes cannot certify the required infinite cancellation estimate.