We now know that a conditionally convergent series converges only through cancellation, and that comparison tests cannot see cancellation. So we need tests that use the pattern of signs. The simplest pattern is strict alternation: +,−,+,−,…. §195 proves the classical test for such series, often called Leibniz’s test, and it comes with a bonus that is rare in analysis: an explicit, practical bound on the error of every partial sum.
The idea
Write the series as φ0−φ1+φ2−φ3+⋯ with every φn>0, and suppose the sizes φn decrease. Walk along the number line: forward φ0, back φ1, forward φ2, back φ3, … . Since each step is no longer than the one before, every backward step undoes only part of the previous forward step, and every forward step undoes only part of the previous backward step. So the walk zigzags inside a shrinking region:
the positions after forward steps, s0,s2,s4,…, go down;
the positions after backward steps, s1,s3,s5,…, go up;
each “up” position lies below each “down” position.
If the steps shrink to zero, the two sequences squeeze together onto a single point: the sum.
ExploreError bars for free
Try this.For 1−21+31−⋯, step n through 0,1,2,… and compare the true error ∣s−sn∣ in the read-out with the next term. Then try 1−31+51−⋯ and the family ∑(−1)n/(n+1)s for several s.
∑n=0∞(−1)nφn,φn=n+11
nsₙ
sₙ, n even (decreasing)
sₙ, n odd (increasing)
bar from sₙ to sₙ₊₁
the sum s
s5=0.6166667 and s6=0.7595238; the sum s=log2≈0.6931472 lies between them. The error is ∣s−s5∣=0.0765, within the bound φ6=0.143 (the next term).
What to notice
Each green bar joins sn to sn+1, and every bar crosses the dashed line: the sum is always between two consecutive partial sums. The blue points (even n) only go down, the ochre ones (odd n) only go up. The error is always less than the next term, and usually about half of it.
Precise statements
TheoremThe alternating series testHardy: §195
Let φn be positive and tend steadily to zero (that is, φn+1≤φn for all n, and φn→0). Then
φ0−φ1+φ2−φ3+⋯
converges, and its sum s lies between φ0 and φ0−φ1.
Under the same conditions, s lies between sn and sn+1 for every n, where sn=φ0−φ1+⋯+(−1)nφn. Hence
∣s−sn∣≤φn+1:
the error in stopping after a term is at most the size of the first term left out, and it has the sign of that term.
Both conditions on φn are needed. If φn decreases to a positive limit, the terms do not tend to 0 and the series oscillates finitely (Examples LXXVIII, 5). If φn→0 but not steadily, the series may diverge (Examples LXXVIII, 4, below). As usual, the conditions only need to hold from some n on.
Proofs
ProofThe alternating series test
Strategy
Show that the even partial sums decrease, the odd ones increase, every odd one is below every even one, and their difference tends to zero. Then both monotone sequences converge, to the same limit.
Even partial sums decrease
s2n+2−s2n=−φ2n+1+φ2n+2≤0, because φ decreases.
Odd partial sums increase
s2n+3−s2n+1=φ2n+2−φ2n+3≥0.
Odd sums lie below even sums
s2n−s2n+1=φ2n+1>0. So for any m and n, taking k=max(m,n), we get s2m+1≤s2k+1<s2k≤s2n. In particular s1≤s2n+1<s2n≤s0 for every n.
Both sequences converge
s2n is decreasing and bounded below by s1; s2n+1 is increasing and bounded above by s0. Monotone bounded sequences converge (§69): say s2n→L and s2n+1→L′.
To the same limit
L−L′=lim(s2n−s2n+1)=limφ2n+1=0. Since both the even and the odd partial sums tend to L, so does the whole sequence sn. (Given δ, choose n0 that works for both subsequences.)
Where the sum lies
The limit s satisfies s2n+1≤s≤s2n for every n, so s lies between each pair of consecutive partial sums. With n=0 this says φ0−φ1≤s≤φ0, and in general ∣s−sn∣≤∣sn+1−sn∣=φn+1: the corollary.
∎
The steady decrease was used in the first two steps only, but those are what make the even and odd partial sums monotone, and without monotonicity the argument has nothing left. The next figure shows what can go wrong.
ExploreWhen the terms do not decrease steadily
Try this.Choose the series with φn=1/(n+(−1)n) and show 200 terms. Then choose φn=1+1/(n+1).
∑n=2∞(−1)nφn,φn=n+(−1)n1
nsₙ
sₙ, n even
sₙ, n odd
bar from sₙ to sₙ₊₁
Here φ5=0.809, φ6=0.2899: the terms tend to 0 but do not decrease steadily. Each downward step (odd n) is longer than the upward step before it, so the partial sums drift downwards, slowly, like −logn: the series diverges to −∞.
What to notice
In the first, φn→0, but each odd term 1/(n−1) is larger than the even term 1/(n−1+1) before it, so the even and odd partial sums are not monotone and both drift down to −∞. In the second, the bars never shrink below length 1: steady decrease without the limit 0 gives finite oscillation.
Worked examples and exercises
ExerciseA family of alternating series●○○Hardy: Examples LXXVIII, 1–2
Let a>0. Show that ∑(−1)n(n+a)−s is absolutely convergent if s>1, conditionally convergent if 0<s≤1, and oscillatory if s≤0. Deduce that 1−21+31−⋯, 1−21+31−⋯, ∑(−1)n/n+a and ∑(−1)n/(n+a) are conditionally convergent.
Hint
The moduli are (n+a)−s: compare with ∑n−s.
Hint
For 0<s≤1 use the theorem; for s≤0 look at the size of the terms.
Solution
The moduli (n+a)−s behave like n−s (their ratio tends to 1), so ∑∣un∣ converges exactly when s>1 (§181, comparison). So for s>1 the series is absolutely convergent.
For 0<s≤1, φn=(n+a)−s is positive, decreases steadily and tends to 0, so the series converges by the theorem; since ∑∣un∣ diverges, it converges conditionally.
For s=0 the series is 1−1+1−⋯, which oscillates finitely; for s<0 the terms have sizes (n+a)∣s∣→∞ alternating in sign, and the partial sums are unbounded above and below (each even partial sum exceeds the previous odd one by (n+a)∣s∣): infinite oscillation.
The four series: the first two are a=1 with s=1,21 (index shifted). For 1/n+a, take s=21. For 1/(n+a), the terms decrease steadily to 0, and their sizes exceed 1/(2n) for n≥a, so the moduli diverge: again conditional convergence.
ExerciseHow many terms?●○○Hardy: Examples LXXVIII, 3
How many terms of 1−31+51−⋯ (whose sum is π/4) guarantee an error less than 10−3? Less than 10−6?
Hint
After m terms, the first term left out is 1/(2m+1).
Solution
By the corollary the error after m terms is less than 1/(2m+1) (strictly, because the sum lies strictly between consecutive partial sums when φ is strictly decreasing). We need 2m+1≥1000, so m=500 terms suffice; for 10−6, m=500000. The bound is close to the truth: the actual error is roughly half the next term, so about 250 terms give two or three correct decimals of π/4, but not more. This series converges slowly.
ExerciseSteady decrease is essential●●○Hardy: Examples LXXVIII, 4
Show that n≥2∑n+(−1)n(−1)n diverges, although it alternates in sign and its terms tend to 0.
Hint
Write n+(−1)n(−1)n=n(−1)n−χn and find χn.
Hint
Is χn positive? How does nχn behave?
Solution
Since n(−1)n−n+(−1)n(−1)n=n(n+(−1)n)(−1)n(n+(−1)n−n)=n+(−1)nn1, we have
n+(−1)n(−1)n=n(−1)n−χn,χn=n+(−1)nn1>0.
∑(−1)n/n converges by the theorem. ∑χn has positive terms with nχn→1, so it diverges like ∑1/n. The difference of a convergent and a divergent series diverges, here to −∞. The hypothesis that fails is steady decrease: φ2k+1=1/(2k+1−1) is larger than φ2k=1/(2k+1).
If φn decreases steadily to a limit l>0, show that ∑(−1)nφn oscillates finitely.
Hint
Steps 1–4 of the proof did not use φn→0.
Solution
As in the proof, s2n decreases to a limit L and s2n+1 increases to a limit L′, and all partial sums lie between s1 and s0, so they are bounded. But L−L′=limφ2n+1=l>0, so sn has no limit. Bounded and not convergent: finite oscillation.
Let a,b>0 and φn=b(b+1)⋯(b+n)a(a+1)⋯(a+n). Show that ∑(−1)nφn converges if and only if a<b.
Hint
Look at φn+1/φn.
Hint
For a<b, take logarithms and use log(1−x)≤−x.
Solution
We have φn+1/φn=(a+n+1)/(b+n+1).
If a≥b this ratio is at least 1, so φn≥φ0>0 for all n; the terms do not tend to 0, and the series cannot converge.
If a<b the ratio is less than 1, so φn decreases steadily. To see that φn→0, write c=b−a>0 and
logφn=k=0∑nlog(1−b+kc)≤−ck=0∑nb+k1→−∞,
since ∑1/(b+k) diverges like the harmonic series. So φn→0 and the theorem applies. (Hardy also allows negative non-integral a, b: then the factors are positive from some point on, and the same argument applies to the tail.)
Check yourself
Check yourselfEnter a number
What is the smallest number m of terms of 1−31+51−⋯ for which the error bound of the corollary guarantees an error below 0.001?
Why
After m terms the bound is the next term, 1/(2m+1). It is below 0.001 exactly when 2m+1>1000, that is m≥500.
Check yourselfTrue or false?
If the terms of an alternating series tend to zero, the series converges.
Why
False. The terms must also decrease in size steadily (from some point on). ∑n≥2(−1)n/(n+(−1)n) alternates, its terms tend to 0, and it diverges.
Check yourselfChoose one
For 1−21+31−41+51 the partial sum is s4=47/60≈0.783. What does the corollary say about the sum s?
Why
The next term is −61, so s lies between s4 and s5=s4−61, and the error has the sign of the first omitted term. (In fact s=log2≈0.693.) The first option uses the last term included instead of the first term omitted.
Check yourselfTrue or false?
For an alternating series φ0−φ1+⋯ with φn decreasing, the even partial sums s0,s2,s4,… form a decreasing sequence.
Why
True: s2n+2−s2n=φ2n+2−φ2n+1≤0. Similarly the odd partial sums increase.
Hardy’s language
Summary
Summary
If φn>0 decreases steadily to 0, then φ0−φ1+φ2−⋯ converges.
The even partial sums decrease, the odd ones increase, and they squeeze onto the sum.
The sum lies between any two consecutive partial sums: ∣s−sn∣≤φn+1, the first term omitted.
Both hypotheses matter: without φn→0 the series oscillates; without steady decrease it may diverge.