The root and ratio tests settle ∑n2/2n and ∑1/n! in a line, but they are silent on ∑1/n2, ∑1/n3/2, ∑1/(nlogn): on every series whose terms shrink like a power of n or more slowly. Those are exactly the series that matter most in analysis. §§178–180 explain why the old tests fail there and introduce two new tools: a necessary condition for convergence (Abel’s theorem), and a test that compares a series with an integral (Maclaurin’s integral test). The integral test settles ∑n−s in the next lesson.
The idea
The geometric series is too good a yardstick
Comparison with ∑rn can only prove convergence of series whose terms eventually shrink geometrically. That is extremely fast. Hardy’s example: (2/3)n at first looks much larger than n−12, but far enough out the geometric sequence is smaller by an astronomical factor.
ExploreGeometric decay always wins
Try this.Start with Hardy’s pair r=2/3, k=12 and find where the curves cross. Then push r up towards 0.99 and k up to 30.
rn against n−k, on a logarithmic scale
nlog₁₀ of the term
log10rn
log10n−k
from here on rn<n−k
At n=296: rn≈10−52.12 and n−k≈10−29.66, so rn is smaller by a factor of about 1022.5. From n=148 on, rn<n−k for good: the straight line always wins against the logarithmic curve.
What to notice
On a logarithmic scale rn is a straight line falling at a fixed slope, and n−k is a curve whose slope flattens out like −k/n. However gentle the line and however steep the curve at first, the line wins. So a series like ∑1/n2, whose terms are not eventually below any rn, can never be handled by a test built on the geometric series.
We need a slower yardstick. The series ∑n−s will be it, but first we must decide for which s it converges, and for that the geometric series is useless. The new tests assume something the old ones did not: that the terms decrease steadily. This is true of nearly every series that occurs naturally.
A necessary condition: nun→0
For a convergent series the terms must tend to 0. If the terms also decrease, much more is true: they must tend to 0 faster than 1/n. The idea is that a decreasing sequence cannot hide its size: the n terms un+1,…,u2n are all at least u2n, so their sum is at least nu2n, and that sum is a piece of the tail of a convergent series, which tends to 0.
ExploreDoes nun tend to 0?
Try this.Look at nun for each of the four series, with 300 or 1000 terms. Note in the read-out which ones decrease and which converge.
un=n21
nn·uₙ
nun
Terms decreasing: yes. Series converges: yes. nun→0: yes. (At n=100: nun=0.01, partial sum 1.635.) Convergent and decreasing, so Abel’s theorem applies: n·uₙ = 1/n → 0, as promised.
What to notice
For 1/n2 (decreasing, convergent) nun→0, as the theorem says. For 1/n it does not, so ∑1/n cannot converge. The third series converges although nun=1 infinitely often, because its terms are not decreasing. The fourth decreases and has nun→0, yet diverges: the condition is necessary, not sufficient.
Comparing a sum with an area
If ϕ is a decreasing function, the sum ϕ(1)+ϕ(2)+⋯+ϕ(n) is the area of n rectangles of width 1 standing over the graph, and ∫1n+1ϕ(x)dx is the area under it. The difference consists of thin curved slivers, one per rectangle. Slide them all to the left, into the strip 0≤x≤1: because ϕ decreases, they stack one above the other without overlapping, inside a column of height ϕ(1). So the sum and the integral never differ by more than ϕ(1), and one is bounded exactly when the other is.
ExploreThe integral test as a picture
Try this.With ϕ(x)=1/x, increase n and switch between “in place” and “slide them left”. Then change s to 2, and try 1/(1+x2).
ϕ(x)=x1: rectangles of height ϕ(ν) against the area under the curve
xy
rectangles: total area ϕ(1)+⋯+ϕ(n); shaded: ∫1n+1ϕ
the excess slivers
Slivers
ϕ(1)+⋯+ϕ(8)=2.7179, ∫19ϕ=2.1972; the slivers have total area 0.5206, which is less than ϕ(1)=1 however large n is. So the sums stay bounded exactly when the integrals do: here ∫1ξϕ→∞ as ξ→∞, so the series diverges.
What to notice
The slivers always fit inside the dashed column, so their total area stays below ϕ(1). For 1/x both the sum and the integral grow without bound (like logn); for x−2 both stay bounded. The sum of 1/(1+n2) is trapped between π/4 and 21+π/4.
Precise statements
“Decreasing” here means never increasing: un+1≤un for every n.
If ∑un is a convergent series of positive, decreasing terms, then nun→0.
Two warnings, both shown in the widget. Without “decreasing” the conclusion can fail (Examples LXX, 2). And the converse is false: nun→0 with decreasing terms does not imply convergence (Examples LXX, 3).
TheoremMaclaurin’s (or Cauchy’s) integral testHardy: §180
Let ϕ(x) be positive, continuous and decreasing for x≥1, and let
Φ(ξ)=∫1ξϕ(x)dx.
Then the series ϕ(1)+ϕ(2)+ϕ(3)+⋯ converges if Φ(ξ) tends to a limit l as ξ→∞, and diverges if Φ(ξ)→∞. In the first case its sum is less than ϕ(1)+l.
Since ϕ>0, the function Φ increases, so these are the only two possibilities. The proof gives more than the theorem.
CorollarySum minus integralHardy: §180
Under the same hypotheses,
ν=1∑n−1ϕ(ν)−∫1nϕ(x)dx
increases with n and tends to a limit between 0 and ϕ(1).
For ϕ(x)=1/x this limit is Euler’s constant γ=0.5772… (Examples XC, 1, in Ch. IX).
Proofs
ProofAbel’s theorem
Strategy
Bound nu2n by a block of the tail of the series, which tends to 0. That handles even indices; odd indices are squeezed between even ones because the terms decrease.
A block of the tail tends to 0
The series converges, so its tails un+1+un+2+⋯ tend to 0. The block un+1+⋯+u2n is part of such a tail (positive terms), so it tends to 0 too.
The block is at least nu2n
It has n terms, each at least u2n because the terms decrease. So 0≤nu2n≤un+1+⋯+u2n→0. Hence 2nu2n=2(nu2n)→0: the claim holds along even indices.
Odd indices
Since u2n+1≤u2n,
0≤(2n+1)u2n+1≤2n2n+1⋅2nu2n→1⋅0=0.
Conclude
Both the even and the odd terms of the sequence nun tend to 0, so nun→0.
∎
ProofThe integral test
Strategy
Measure the gap between each term ϕ(ν−1) and the integral over [ν−1,ν]. These gaps are positive and add up to at most ϕ(1) (the slivers in the column), so “sum minus integral” converges. Then the sum and the integral are bounded together.
Squeeze ϕ on each unit interval
Because ϕ decreases, ϕ(ν−1)≥ϕ(x)≥ϕ(ν) for ν−1≤x≤ν.
Define the slivers
Let
vν=ϕ(ν−1)−∫ν−1νϕ(x)dx=∫ν−1ν(ϕ(ν−1)−ϕ(x))dx.
The integrand lies between 0 and ϕ(ν−1)−ϕ(ν), so 0≤vν≤ϕ(ν−1)−ϕ(ν). (This is the area of the sliver over [ν−1,ν], and the bound says it fits in the horizontal strip between heights ϕ(ν) and ϕ(ν−1).)
The slivers add up to at most ϕ(1)
The upper bounds telescope: v2+⋯+vn≤ϕ(1)−ϕ(n)≤ϕ(1). So ∑vν is a series of positive terms with bounded partial sums, and it converges, to a sum between 0 and ϕ(1).
Rewrite the partial sums
Adding the definitions, v2+⋯+vn=∑ν=1n−1ϕ(ν)−Φ(n). This is the Corollary: it increases and tends to a limit in [0,ϕ(1)].
Conclude
Write ∑1n−1ϕ(ν)=Φ(n)+(v2+⋯+vn). The second part converges, so the partial sums of the series converge exactly when Φ(n) does. Since Φ increases, Φ(n) has a limit exactly when Φ(ξ) has one as ξ→∞ through all real values, and otherwise Φ(ξ)→∞. In the convergent case the sum is l+lim(v2+⋯+vn)≤l+ϕ(1). (Hardy notes that the inequality is strict, since the vν cannot all reach their upper bounds.)
∎
Worked examples and exercises
Exercise●○○Hardy: Examples LXX, 1
Use Abel’s theorem to show that ∑1/n and ∑1/(an+b) diverge, where a>0 and an+b>0 for all n considered.
Solution
Both series have positive decreasing terms. For un=1/n, nun=1, which does not tend to 0. For un=1/(an+b), nun=n/(an+b)→1/a=0. If either series converged, Abel’s theorem would force nun→0. So both diverge.
Exercise●●○Hardy: Examples LXX, 2
Let un=1/n when n is a perfect square and un=1/n2 otherwise. Show that ∑un converges but nun does not tend to 0. Why is Abel’s theorem not contradicted?
Hint
Split the series into the terms at squares and the others. Each part can be compared with ∑1/k2.
Solution
The terms at the squares n=k2 are 1/k2, and these form the convergent series ∑1/k2. The remaining terms are some of the terms of ∑1/n2, so their partial sums are bounded by ∑1/n2. Every partial sum of ∑un is at most the sum of the two bounds, so ∑un converges (Theorem B). But nun=1 whenever n is a square, which happens for arbitrarily large n, so nun→0. There is no contradiction because the terms do not decrease: u4=41>u3=91.
Exercise●●●Hardy: Examples LXX, 3
Let u1=1 and un=1/(nv) when 2v−2<n≤2v−1 (v=2,3,…). Show that the terms decrease and nun→0, but ∑un diverges.
Hint
Group the terms into the blocks 2v−2<n≤2v−1. How many terms does a block have, and how small can each be?
Solution
Decreasing. Inside a block 1/(nv) decreases with n; passing to the next block the factor 1/v drops to 1/(v+1) and n increases, so the terms keep decreasing.
nun→0.nun=1/v, and v→∞ as n→∞ (we have v−1≥log2n).
Divergence. The block for v≥2 has 2v−2 terms, each at least 1/(2v−1v). So its sum is at least 2v−2/(2v−1v)=1/(2v). The partial sums up to n=2V−1 are therefore at least 21(21+31+⋯+V1), which tends to ∞ with V because ∑1/v diverges. Hence the converse of Abel’s theorem is false: nun→0 can happen arbitrarily slowly.
Exercise●●○Hardy: Examples LXXI, 8
Show that 2n−2<1+21+⋯+n1<2n−1 for n≥2.
Hint
Use the rectangles of the integral test on both sides of the curve y=1/x.
Solution
Let ϕ(x)=x−1/2, decreasing, with ∫abϕ=2b−2a.
Lower bound. On [ν,ν+1], ϕ(x)≤ϕ(ν), with strict inequality for x>ν; integrating, ∫νν+1ϕ<ϕ(ν). Summing for ν=1,…,n: ∑1nϕ(ν)>∫1n+1ϕ=2n+1−2>2n−2.
Upper bound. On [ν−1,ν], ϕ(x)≥ϕ(ν), strictly for x<ν, so ϕ(ν)<∫ν−1νϕ. Summing for ν=2,…,n and adding ϕ(1)=1: ∑1nϕ(ν)<1+∫1nϕ=2n−1.
Exercise●●○Hardy: Examples XC, 1 (from §180)
Show that 1+21+⋯+n1−logn tends to a limit γ with 0≤γ≤1. (You may use ∫1ndx/x=logn.)
Hint
Apply the Corollary with ϕ(x)=1/x, and account for the last term 1/n.
Solution
By the Corollary with ϕ(x)=1/x, dn=∑ν=1n−1ν1−logn increases and tends to a limit γ with 0≤γ≤ϕ(1)=1. Our sequence is dn+1/n, and 1/n→0, so it has the same limit. (In fact γ=0.5772…; it is Euler’s constant.) This also shows how fast ∑1/n diverges: its partial sums are logn+γ+o(1).
Check yourself
Check yourselfTrue or false?
If ∑un is a convergent series of positive terms, then nun→0.
Why
False without “decreasing”: the series with un=1/n at squares and 1/n2 elsewhere converges, yet nun=1 at every square. Abel’s theorem uses the decrease to say that u2n is the smallest of the block un+1,…,u2n.
Check yourselfTrue or false?
If un decreases and nun→0, then ∑un converges.
Why
False. ∑1/(nlogn) has decreasing terms with nun=1/logn→0, and it diverges (integral test, with ∫dx/(xlogx)=loglogx). Hardy’s Examples LXX, 3 is an example that avoids logarithms.
Check yourselfChoose one
Which series can the integral test be applied to directly, as stated?
Why
The test needs ϕ positive and decreasing. sin2x/x2 is not decreasing (it vanishes at multiples of π); compare it with 1/n2 instead. (−1)n/n is not positive. x2−x increases at first (up to x=1/log2≈1.44); the test could be applied from x=2 on, but the ratio test is far simpler.
Check yourselfEnter a number
For ϕ(x)=1/x2, the integral test bounds ∑n≥11/n2 above by ϕ(1)+l. What is this bound?
Why
ϕ(1)=1 and l=∫1∞x−2dx=1, so the bound is 2. The true sum is π2/6≈1.645, and the lower bound from the same picture is l=1.
Check yourselfTrue or false?
For every r with 0<r<1 and every k>0, we have rn<n−k for all sufficiently large n.
Why
True: nkrn→0 (Examples XXVII, 7), so nkrn<1 eventually. Even r=0.99 and k=100 only postpone the moment (to beyond n=100000).
Hardy’s language
The geometric series … is not only convergent but very rapidly convergent. The tests derived from comparison with it are therefore naturally very crude, and much more delicate tests are often wanted.
— G. H. Hardy, §178
Summary
Summary
Tests built on the geometric series only detect geometric decay; they fail for terms like n−s.
Abel’s theorem: positive, decreasing, convergent implies nun→0. Both hypotheses matter, and the converse is false.
Integral test: for positive decreasing continuous ϕ, ∑ϕ(n) converges exactly when ∫1ξϕ tends to a limit l; the sum then lies between l and ϕ(1)+l.
Sum minus integral converges, to a limit between 0 and ϕ(1): the slivers stack inside one column.