The root test of the last lesson needs vn1/n, which is awkward when vn contains factorials or a product of many factors. In §175 Hardy compares successive terms instead. The ratio vn+1/vn of a complicated term is often very simple: for vn=n!/nn it is (n/(n+1))n. The resulting test, d’Alembert’s, is the one most people reach for first. This lesson is about how it works, why it works, and where it fails.
The idea
A geometric series is the series in which every term is the same multiple r of the one before. If the ratio of successive terms of ∑vn is eventually at most r, with r<1, then from some point on the terms shrink at least as fast as a geometric series, and the series converges for the same reason the geometric series does. On a logarithmic scale this is easy to see: a geometric series is a straight line going down, and a series whose ratios tend to r eventually runs parallel to that line.
ExploreTerms that shrink geometrically
Try this.Plot the terms of ∑nkrn with the logarithmic scale on, for k=2 and r=0.7. Then raise k to 5 and show 100 terms.
∑n=1∞nkrn
nuₙ (log scale)
uₙ for nkrn
Plot
After 100 terms s100=44.07407407. The series converges.
What to notice
At first the factor nk makes the terms grow, but the plot always bends over into a straight line of slope log10r. That is because vn+1/vn=(1+1/n)kr→r: eventually every step multiplies the term by nearly r. A large k only delays this.
The ratio test only looks at neighbouring terms, and that is its weakness. A series can converge perfectly well while its neighbouring terms keep jumping up and down.
ExploreWhere the ratio test is blind
Try this.Compare the ratio test and the root test on a+b+a2+b2+a3+⋯ with a=41, b=21. Then try both tests on 1/n!, n!/nn and 1/n2.
∑n=1∞vn,vn=v2m−1=am,v2m=bm
nvₙ₊₁/vₙ
vn+1/vn (below 1)
the critical value 1 (points at or above it)
Test
At n=49: vn+1/vn=3.3554×107. Hollow points are off the top of the plot. Test inconclusive or not applicable. The ratio alternates between (b/a)m→∞ and a(a/b)m→0. It is neither eventually ≤r<1 nor eventually ≥1: d’Alembert’s test fails completely (Ex. LXVIII. 9), although the series converges.
What to notice
For the interleaved series the ratios jump between huge and tiny values (the hollow points are off the scale), so d’Alembert’s test cannot be applied, while every root stays below b≈0.71 and Cauchy’s test proves convergence. For 1/n! and n!/nn the ratio settles quickly and is easy to compute. For 1/n2 both tests sit at 1 and say nothing.
Precise statements
In this lesson all terms are strictly positive, since we divide by them.
TheoremComparison of ratiosHardy: §175, 4 and 5
Let un>0 and vn>0.
If vnvn+1≤unun+1 for all sufficiently large n, and ∑un converges, then ∑vn converges.
If vnvn+1≥unun+1 for all sufficiently large n, and ∑un diverges, then ∑vn diverges.
Taking un=rn in part 1 gives the test itself.
Theoremd’Alembert’s testHardy: §175, 6
If vn+1/vn≤r for all sufficiently large n, where r<1, then ∑vn converges. If vn+1/vn≥1 for all sufficiently large n, then ∑vn diverges.
Most often the ratio, or the root, has a limit, and then the tests take their familiar form.
CorollaryThe limit formsHardy: §175, 7
If vn1/n or vn+1/vn tends to a limit l as n→∞, then ∑vn converges if l<1 and diverges if l>1. If l=1, the test gives no information.
The case l=1 really is undecided: ∑1/n diverges, ∑1/n2 converges, and for both of them the ratio and the root tend to 1. The tests also fail when the ratio or root stays below 1 but comes arbitrarily close to it, and the ratio test fails when the ratio is sometimes above and sometimes below 1, as in the interleaved series above.
Proofs
ProofComparison of ratios
Strategy
Write vn as vn0 times a chain of ratios, compare the chain link by link with the chain for un, and conclude vn≤Kun with a constant K. Then Theorem C finishes the proof.
Telescope the ratios
For n>n0,
vn=vn−1vn⋅vn−2vn−1⋯vn0vn0+1vn0.
Compare factor by factor
Each ratio vν+1/vν with ν≥n0 is at most uν+1/uν, and all factors are positive, so
vn≤un−1un⋯un0un0+1vn0=un0vn0un.
A constant K
K=vn0/un0 does not depend on n. So vn≤Kun for n≥n0, and Theorem C gives convergence of ∑vn.
The divergence half
With the inequalities reversed the same chain gives vn≥Kun, with K>0, and Theorem C gives divergence.
∎
Proofd’Alembert’s test and its limit form
Strategy
Apply the comparison of ratios with the geometric series, then show that a limit below 1 (or above 1) produces the hypothesis needed.
Convergence
Take un=rn, so un+1/un=r. If vn+1/vn≤r for large n and r<1, part 1 of the comparison applies, since ∑rn converges.
Divergence
If vn+1/vn≥1 for n≥n0, then vn≥vn0>0 for all n≥n0. The terms do not tend to 0, so the series diverges.
A limit l<1
Choose r with l<r<1, for example r=21(1+l). Taking δ=r−l in the definition of a limit, vn+1/vn<r for all large n, and the series converges. The root version is the same with Cauchy’s test in place of d’Alembert’s.
A limit l>1
Now vn+1/vn>1 for all large n (take δ=l−1), so the terms eventually increase and cannot tend to 0. For the root, vn1/n>1 for large n means vn>1, with the same conclusion. Hardy leaves this half to the reader.
∎
Hardy remarks that d’Alembert’s test is theoretically less general than Cauchy’s: whenever the ratio test proves convergence, so does the root test. The proof is short.
ProofRatio test succeeds ⟹ root test succeeds
Strategy
The ratio hypothesis gives vn≤Krn. Taking n-th roots turns the constant K into K1/n, which tends to 1 and so does no harm.
From ratios to a bound
If vn+1/vn≤r<1 for n≥n0, the proof above gives vn≤Krn with K=vn0/rn0.
Take roots
Then vn1/n≤K1/nr. Since K1/n→1 (Examples XXVII, 10), for any r′ with r<r′<1 we get vn1/n≤r′ for all large n, which is Cauchy’s hypothesis.
The converse fails
The interleaved series a+b+a2+b2+⋯ satisfies Cauchy’s hypothesis with r′=b, but its ratios are not eventually below any r<1 (Exercise 4 below).
∎
Worked examples and exercises
Exercise●○○Hardy: Examples LXVIII, 1
Let k be a positive integer and r>0. Use d’Alembert’s test to show that ∑nkrn converges if r<1 and diverges if r≥1. Check that Cauchy’s test gives the same.
Hint
Compute vn+1/vn and let n→∞.
Hint
For the root test you need limn1/n=1 (Examples XXVII, 11).
Solution
vnvn+1=(nn+1)kr→r.
So the series converges if r<1 and diverges if r>1. When r=1 the test fails, but then vn=nk≥1, so the terms do not tend to 0 and the series diverges. For the root test, vn1/n=(n1/n)kr→r since n1/n→1, with the same conclusions. In particular nkrn→0 when r<1, however large k is.
Exercise●●○Hardy: Examples LXVIII, 2, 3 and 6
Let P and Q be polynomials of degrees k and l with positive leading coefficients, and r>0. Discuss the convergence of ∑Q(n)P(n)rn (start the sum beyond the largest root of Q).
Hint
P(n)/Q(n)∼(A/α)nk−l, where A and α are the leading coefficients. Use Theorem D.
Hint
When r=1 the ratio test fails. Compare with ∑nk−l and use the last lesson.
Solution
By Theorem D, the series behaves like ∑nk−lrn, since the ratio of the terms tends to A/α=0.
If r=1, d’Alembert’s test applies as in the previous exercise (the exponent k−l may now be negative, which changes nothing): (nn+1)k−lr→r. So the series converges for r<1 and diverges for r>1.
If r=1 the series behaves like ∑n−(l−k), which converges exactly when l−k≥2, that is l>k+1, and diverges when l≤k+1 (Examples LXVIII, 5, and the divergence of ∑1/n; for exponents that are not integers see §181).
Exercise●○○Hardy: Examples LXVIII, 4
Show that Cauchy’s and d’Alembert’s tests both fail for ∑n(n+1)1, although the series converges.
Solution
vnvn+1=n+2n→1, and vn1/n=(n1/n(n+1)1/n)−1→1, since n1/n→1 and (n+1)1/n→1. Both limits equal 1, so both tests are silent, yet the series telescopes to the sum 1. The geometric series is too coarse a yardstick for series whose terms decrease like a power of n.
Exercise●●○Hardy: Examples LXVIII, 9
Let 0<a<b<1. Show that a+b+a2+b2+a3+b3+⋯ converges, that Cauchy’s test proves it, and that d’Alembert’s test fails.
Hint
Write v2m−1=am, v2m=bm, and compute the two kinds of ratio.
Solution
The series is the sum of two convergent geometric series, so it converges (to a/(1−a)+b/(1−b)).
Ratios.v2m/v2m−1=(b/a)m→∞ and v2m+1/v2m=a(a/b)m→0. The ratios are neither eventually ≤r<1 nor eventually ≥1, so d’Alembert’s test says nothing.
Roots.v2m1/(2m)=b, and v2m−11/(2m−1)=am/(2m−1)≤a<b because m/(2m−1)≥21 and a<1. So vn1/n≤b<1 for every n, and Cauchy’s test proves convergence.
Exercise●●○Hardy: Examples LXVIII, 10 and 11
Show that ∑rn/n! and ∑rn/nn converge for every r>0, that ∑n!rn and ∑nnrn diverge for every r>0, and that ∑(n+1nr)n converges if r<1 and diverges if r≥1.
Hint
Use the ratio test for the factorials and the root test for the n-th powers.
Hint
For the last series at r=1, recall (1+1/n)n→e (§73).
Solution
rn/n!rn+1/(n+1)!=n+1r→0, so ∑rn/n! converges; and (rn/nn)1/n=r/n→0, so ∑rn/nn converges.
n!rn(n+1)!rn+1=(n+1)r→∞ and (nnrn)1/n=nr→∞, so the other two diverge (their terms grow without bound).
For the last series, vn1/n=nr/(n+1)→r: convergence for r<1, divergence for r>1. At r=1 the test fails, but vn=(1+1/n)−n→1/e=0, so the series diverges.
Check yourself
Check yourselfTrue or false?
If vn>0 and vn+1/vn<1 for every n, then ∑vn converges.
Why
False. For vn=1/n every ratio n/(n+1) is less than 1, but the series diverges. The ratios must stay below a fixed r<1; here they approach 1. A strictly decreasing sequence of terms is not enough.
Check yourselfChoose one
For a series of positive terms, vn+1/vn→1. What can you conclude?
Why
A limit 1 leaves the question open. The terms may still tend to 0 (as 1/n and 1/n2 do), so the second option is wrong too. The root test does not help: if the ratio tends to 1 then so does the root (Hardy proves this in Ch. IX), so it fails as well. Finer tests come in §§178–183.
Check yourselfEnter a number
For vn=n5/3n, what is limvn+1/vn?
Why
vnvn+1=(1+n1)531→31. The series converges, although its first terms grow (up to about n=5).
Check yourselfTrue or false?
There are series of positive terms whose convergence d’Alembert’s test proves but Cauchy’s test cannot.
Why
False. If vn+1/vn≤r<1 for large n then vn≤Krn, so vn1/n≤K1/nr, which is below any r′∈(r,1) for large n. The root test is the stronger one in theory; the ratio test is often easier in practice.
Check yourselfChoose one
Does ∑n!/10n converge?
Why
vn+1/vn=(n+1)/10→∞. From n=10 on, each term is at least as large as the one before, so the terms do not tend to 0. Factorials eventually beat every geometric growth.
Hardy’s language
Ratio tests are useless for ‘irregular’ series, such as 0 + ½ + 0 + ¼ + 0 + ⅛ + ….
— G. H. Hardy, §175
Summary
Summary
If un,vn>0 and vn+1/vn≤un+1/un for large n, then vn≤Kun: ratios can be compared instead of terms.
d’Alembert’s test: vn+1/vn≤r<1 for large n gives convergence; vn+1/vn≥1 for large n gives divergence.
Limit forms: a ratio or root limit below 1 means convergence, above 1 divergence, equal to 1 means no information.
The root test is more general; the ratio test is easier for factorials and products.
Both tests use the geometric series as a yardstick, which is far too coarse for terms like 1/n2: finer tests follow.