M2PM3 HANDOUT: THE ABEL AND DIRICHLET TESTS FOR
CONVERGENCE
The Abel and Dirichlet tests for convergence of series belong to Real Anal-
ysis rather than, or as much as, to Complex Analysis. We give them here since
they are necessary for handling∑ convergence and absolute convergence of Dirich-
∞
let series – series of the form n=1 an /ns , with s a complex variable. These,
and in particular
∑∞ the most important special case, the Riemann zeta function
ζ(s) := n=1 an /ns , are important in Analytic Number Theory; the zeta func-
tions is crucial for the study of the distribution of prime numbers. In this course,
the main importance of the zeta functions is as an example of the crucially im-
portant concept of analytic continuation (III.8). See Coursework 1 (2010).
Abel’s Lemma below is the discrete analogue of integration by parts, or par-
tial integration. It is accordingly also called partial summation. In what follows,
an , vn are real.
Theorem (Abel’s Lemma, or Partial Summation).
Write sn := a1 + . . . + an . Show that
(i) a1 v1 + . . . + an vn = s1 (v1 − v2 ) + . . . + sn−1 (vn−1 − vn ) + sn vn .
(ii) If m ≤ a1 + . . . + an ≤ M for all n, and vn is positive and decreasing, then
mv1 ≤ a1 v1 + . . . + an vn ≤ M v1 .
(iii) If in (ii) |sn | ≤ M for all n, then |a1 v1 + . . . an vn | ≤ M v1 for all n.
Proof. (i)
a1 v1 + . . . an vn = s1 v1 + (s2 − s1 )v1 + . . . + (sn − sn−1 )vn
= s1 (v1 − v2 ) + s2 (v2 − v3 ) + . . . + sn−1 (vn−1 − vn ) + sn vn .
(ii) As vn ↓, vk − vk+1 ≥ 0. This and m ≤ sk ≤ M give
m(vk −vk+1 ) ≤ sk (vk −vk+1 ) ≤ M (vk −vk+1 ) (k = 1, . . . , n−1), mvn ≤ sn vn ≤ M vn .
Sum over k = 1 to n − 1: the left and right telescope. Using (i) for the middle
gives
mv1 ≤ a1 v1 + . . . + an vn ≤ M v1 .
(iii) If |sn | ≤ M for all n, taking m = −M in (ii) gives
|a1 v1 + . . . + an vn | ≤ M v1 . //
Theorem (Dirichlet’s Text for Convergence).
∑n ∑
If (an ) has bounded partial sums sn = 1 ak and vn ↓ 0, then an vn is con-
vergent.
∑vn ↓ 0: ∀ϵ > 0 ∃N such that for
Proof. As ∑nn
≥ N , 0 ≤ vn < ϵ. As the partial
sums of an are bounded, for some M | 1 ak | ≤ M for all n. So
∑
n ∑
n ∑
m−1
| ak | = | ak − ak | ≤ 2M ∀m, n (m ≤ n).
m 1 1
1
∑n
So by (iii) of Abel’s Lemma, | m ∑ak vk | ≤ 2M ϵ for all m, n ≥ N . By Cauchy’s
General Principle of Convergence, an vn converges (as it is Cauchy).//
Theorem
∑ Abel’s Test for Convergence. ∑
If an converges and vn ↓ ℓ for some ℓ, then an vn converges.
∑ ∑n
Proof. As the series an converges, its sequence sn := 1 ak of partial sums
∑ So (sn ) is bounded. As vn ↓ ℓ, wn := vn − ℓ ↓ 0. So by Dirichlet’s
converges.
Test, an wn converges, to c say:
a1 w1 + . . . an wn → c (n → ∞).
That is
a1 v1 + . . . an vn − ℓ(a1 + . . . + an ) → c (n → ∞).
∑∞
But a1 + . . . an → b := 1 ak . So
a1 v1 + . . . an vn → c + ℓ.b (n → ∞),
∑
i.e. an vn converges. //
We include the following results (used in Coursework 1, 2010, and III.8,
2011) for completeness.
∑
Theorem (Alternating Series Test). If an ↓ 0, (−1)n an converges.
∑n
Proof. Write sn := 1 ak .
s2n = (a1 − a2 ) + . . . + (a2n−1 − a2n ). (1)
Since an ↓, each bracket on RHS is ≥ 0, so s2n ↑. But bracketing differently,
s2n = a1 − (a2 − a3 ) − . . . − (a2n−2 − a2n−1 ) − a2n . (2)
Each bracket on RHS is ≥ 0 as an ↓, and a2n ≥ 0. So s2n ≤ a1 . So s2n is ↑ and
bounded above, so
s2n ↑ s < ∞.
Also s2n+1 = s2n + a2n+1 . But s2n → s, a2n+1 → 0, so
s2n+1 → s + 0 = s.
Combining the odd and even subsequences gives sn → s, as required. //
∑∞ ∫ ∞ Test). If f (x) is decreasing and non-negative on [1, ∞),
Theorem (Integral
1 f (n) and 1 f (x)dx converge or diverge together.
We omit the proof (see a textbook on Analysis); you may quote the result.
NHB, 2011
