For decreasing terms, the series Σφ(n) behaves exactly like the much thinner series Σ2^nφ(2^n), which turns powers into geometric series and logarithms into powers.
The integral test needs a function ϕ(x) of a continuous variable and an integral we can evaluate. §182 gives a test with the same reach that needs neither: only the terms themselves, at the indices 1,2,4,8,16,…. It is the argument Hardy used in §77 to show that ∑1/n diverges, turned into a general theorem. It settles ∑n−s in one line, and in Chapter IX it settles ∑1/(nlogn) and ∑1/(n(logn)2), where every power of n is too coarse.
The idea
Group the terms in blocks of doubling length: the block 2k≤n<2k+1 has 2k terms. If the terms decrease, the largest term of the block is its first, ϕ(2k), and the smallest is its last. So the whole block is at most 2kϕ(2k), and (shifting the blocks by one place) at least about half of the next such quantity. In other words the series behaves like the condensed series
ϕ(1)+2ϕ(2)+4ϕ(4)+8ϕ(8)+⋯=k∑2kϕ(2k),
which keeps only one term per block and multiplies it by the block length.
ExploreBlocks of doubling length
Try this.With ϕ(n)=n−s and s=1, compare each block with its rectangle in both modes, and read the table. Then try s=1.5, and the two logarithmic series.
ϕ(n)=n−1, terms grouped in blocks of length 2k
nφ(n)
terms φ(n), alternate blocks in two colours
rectangle of height ϕ(2k) over the block 2k≤n<2k+1
Compare each block with
k
block 2ᵏ ≤ n < 2ᵏ⁺¹
2ᵏ φ(2ᵏ)
0
1
1
1
0.8333
1
2
0.7595
1
3
0.7254
1
4
0.709
1
5
0.701
1
Each block sum is at most 2kϕ(2k), because the block has 2k terms and the first is the largest. Adding up: 4.728 ≤ 6. So if the condensed series ∑2kϕ(2k) converges, the partial sums of ∑ϕ(n) are bounded. Here 2kϕ(2k)=(21−s)k=1k: a geometric series, convergent exactly when s>1.
What to notice
For s=1 every rectangle has area exactly 1: the block sums neither shrink nor grow, so the series diverges. For s=1.5 the rectangle areas shrink by the factor 2−1/2 from block to block, a geometric series. In each mode the block sums are on the correct side of the rectangle areas: at most the upper ones, at least the lower ones.
Precise statements
TheoremCauchy’s condensation testHardy: §182
If ϕ(n) is positive and decreases as n increases, then the series ∑ϕ(n) and ∑2nϕ(2n) are either both convergent or both divergent.
The test works with any integer base a≥2 in place of 2 (Examples LXXII, 1), but 2 is usually the most convenient.
Corollary∑n−s againHardy: §182
∑n−s converges if s>1 and diverges if s≤1.
Indeed, for s>0 the condensed series is ∑2n2−ns=∑(21−s)n, a geometric series with ratio 21−s, which is less than 1 exactly when s>1. (For s≤0 the terms do not tend to 0.) Condensation turns powers of n into geometric series, and, as Chapter IX will show, powers of logn into powers of n.
Proofs
ProofCauchy’s condensation test
Strategy
Compare each block of terms with its largest term (for an upper bound) and with its smallest term (for a lower bound). Both comparisons produce the condensed series, so the two series have bounded partial sums together.
Upper bound: blocks 2k≤n<2k+1
The block has 2k terms, each at most the first one, since ϕ decreases:
ϕ(2)+ϕ(3)≤2ϕ(2),ϕ(4)+⋯+ϕ(7)≤4ϕ(4),…
and in general ϕ(2k)+⋯+ϕ(2k+1−1)≤2kϕ(2k).
So condensed convergence implies convergence
Adding the blocks k=0,1,…,K: ϕ(1)+⋯+ϕ(2K+1−1)≤∑k=0K2kϕ(2k). If the condensed series converges, the right side is bounded by its sum, so the partial sums of ∑ϕ(n) (which increase) are bounded and the series converges.
Lower bound: blocks 2k<n≤2k+1
Now compare each block with its last, smallest term:
ϕ(3)+ϕ(4)≥2ϕ(4),ϕ(5)+⋯+ϕ(8)≥4ϕ(8),…
and in general ϕ(2k+1)+⋯+ϕ(2k+1)≥2kϕ(2k+1)=21⋅2k+1ϕ(2k+1).
So condensed divergence implies divergence
Adding the blocks k=0,…,K: ϕ(2)+⋯+ϕ(2K+1)≥21∑k=1K+12kϕ(2k). If the condensed series diverges, the right side tends to ∞, and so do the partial sums of ∑ϕ(n).
Conclude
Both series have positive terms, so each either converges or diverges to ∞. The two steps show that they do the same thing.
∎
Notice the role of “decreasing”: it is what lets one term stand for a whole block. For terms that go up and down, a single term at n=2k says nothing about its neighbours.
Worked examples and exercises
Here is the whole laboratory: all five tests on the whole gallery. Use it to check your answers to the exercises, and to see which test is the most economical for each series.
ExploreThe convergence-test laboratory
Try this.Choose 1/(nlogn) and run every test on it, then do the same for 1/(n(logn)2). Finally pick a series from the top of the menu and decide which test is quickest.
∑n=2∞vn,vn=nlogn1
kpartial sums
s2k+1−1, the original series up to n=2k+1−1
∑j≤k2jϕ(2j), the condensed series
Test
At k=11 (that is, n=4095 terms of the original): original 2.9131, condensed 4.3568. 11 condensed terms do the work of about four thousand. Diverges. The condensed series has terms 2kϕ(2k)=2kklog22k=klog21, which diverges; so the original series diverges too. The condensed series is a multiple of the harmonic series.
What to notice
For the two logarithmic series, comparison with powers of n and the root and ratio tests are all silent; the integral and condensation tests decide them, in opposite ways, and about a dozen condensed terms already show the trend that four thousand ordinary terms barely reveal. For n2/2n or 1/n! the ratio test is the fastest; for the interleaved series only the root test or a comparison works.
Exercise●●○Hardy: Examples LXXII, 1
Let a≥2 be an integer. Show that, for positive decreasing ϕ, ∑ϕ(n) converges or diverges with ∑anϕ(an).
Hint
Use blocks ak≤n<ak+1, which contain ak+1−ak=(a−1)ak terms.
Solution
Upper bound. The block ak≤n<ak+1 has (a−1)ak terms, each at most ϕ(ak), so its sum is at most (a−1)akϕ(ak). Summing, ∑n<aK+1ϕ(n)≤(a−1)∑k≤Kakϕ(ak): convergence of the condensed series gives convergence.
Lower bound. The block ak<n≤ak+1 has (a−1)ak terms, each at least ϕ(ak+1), so its sum is at least (a−1)akϕ(ak+1)=aa−1ak+1ϕ(ak+1). Summing shows that divergence of the condensed series gives divergence. The constants a−1 and (a−1)/a do not affect convergence.
Exercise●●○Hardy: Examples LXXII, 2
Show that if ∑2nϕ(2n) converges then 2nϕ(2n)→0. Deduce Abel’s theorem: for positive decreasing ϕ with ∑ϕ(n) convergent, nϕ(n)→0.
Hint
For 2k≤n<2k+1, compare nϕ(n) with 2k+1ϕ(2k).
Solution
The terms of a convergent series tend to 0, so ck=2kϕ(2k)→0.
Now let ∑ϕ(n) converge. By the condensation test ∑ck converges, so ck→0. For 2k≤n<2k+1, since ϕ decreases, 0≤nϕ(n)≤2k+1ϕ(2k)=2ck. As n→∞, the corresponding k→∞, so 2ck→0 and nϕ(n)→0.
ExercisePowers of a logarithm●●○
Assuming the properties of log from Chapter IX, show that ∑n≥2n(logn)p1 converges if p>1 and diverges if p≤1.
Hint
Condense: what is 2kϕ(2k) when ϕ(n)=1/(n(logn)p)?
Solution
For p>0, ϕ is positive and decreasing on n≥2. The condensed terms (from k=1) are
2k2k(klog2)p1=(log2)p1⋅kp1,
a constant multiple of ∑k−p, which converges exactly when p>1. For p≤0 we have (logn)p≤1 once logn≥1 (that is, n≥3), so the terms are at least 1/n and the series diverges by comparison with ∑1/n. Condensation has turned logn into k: one level of the “scale of infinity” has been peeled off.
ExerciseCondensing ∑1/n2 by hand●○○
Use the condensation inequality to show that ∑n≥11/n2≤2, and use the base-2 lower bound to show it is at least 1+21=1.5.
Solution
Upper.∑ϕ(n)≤∑k≥02kϕ(2k)=∑2k4−k=∑2−k=2.
Lower.ϕ(1)=1, and ϕ(2)+ϕ(3)+⋯≥21∑k≥12kϕ(2k)=21∑k≥12−k=21. So the sum is at least 1.5. The true value π2/6≈1.645 lies between. The upper bound is the same as the integral test’s; the lower bound is better than its 1.
Check yourself
Check yourselfChoose one
What is the condensed series of ∑n−s?
Why
2nϕ(2n)=2n(2n)−s=2n(1−s)=(21−s)n, a geometric series with ratio 21−s. Forgetting the factor 2n (the first option) would give a series that converges for every s>0, which is wrong. The last option evaluates ϕ at n instead of at 2n.
Check yourselfEnter a number
For ϕ(n)=1/n2, what is the condensed term 2kϕ(2k) at k=5?
Why
25/(25)2=1/25=1/32=0.03125. It stands for the block 32≤n≤63, whose actual sum is about 0.0159, less than the condensed term, as the upper bound says.
Check yourselfTrue or false?
The condensation test may be applied to ∑n22+(−1)n as it stands.
Why
False: the terms 11,43,91,163,… do not decrease, so a single term ϕ(2k) cannot stand for its block. (Here ϕ(2k)=3/4k happens to pick out the larger values.) The series does converge, since its terms are at most 3/n2; comparison is the right tool.
Check yourselfTrue or false?
The condensed series of ∑1/n is 1+1+1+⋯.
Why
True: 2k⋅2−k=1. This is Hardy’s proof in §77 that the harmonic series diverges: each block 2k+11+⋯+2k+11 is at least 21.
Hardy’s language
Summary
Summary
For positive decreasing ϕ, ∑ϕ(n) and ∑2nϕ(2n) converge or diverge together.
Proof: each block 2k≤n<2k+1 lies between 2kϕ(2k+1) and 2kϕ(2k).
Condensation turns n−s into a geometric series and 1/(n(logn)p) into ∑k−p.
Without the decrease the test means nothing; use comparison instead.