REAL ANALYSIS

PARTIAL SUMMATION (ABEL SUMMATION)

As part of the analogy existing between summation and integration, partial

summation corresponds to integration by parts.
m
X
If u ≤ v and sm = ar then we have the identity
r=u
v
X v
X
am bm = bv+1 sV + sm (bm − bm+1 ) (1)
m=u m=u

Proof
v
X v
X
am b m = (sm − sm−1 )bm
m=u m=u
v
X
= bv+1 sv + sm (bm − bm+1 )
m=u

with the convention that empty sums are zero.

Abel’s lemma With the above notation, suppose that {bm } is a positive
monotonic decreasing sequence, and that |sm | ≤ M for all m.
Then
¯ v ¯
¯X ¯
a m bm ¯ ≤ M b v
¯ ¯
¯
m=u
¯ ¯

Proof
¯ v ¯ ¯ v ¯
¯X ¯ ¯X ¯
¯
¯ am bm ¯¯ = ¯
¯ sm (bm − bm+1 ) + sv bv+1 ¯¯
¯m=u ¯ ¯m=u ¯
v
X
≤ |sm |(bm − bm+1 ) + |sv |bv+1
m=u
v
" #
X
≤ M (bm − bm+1 ) + bv+1
m=u
= M b0

Theorem 6 Dirichlet’s test Suppose that φn is a monotonic decreasing

sequence converging to zero, and that an is a series with bounded
P
∞
X
partial sums. Then an φn is convergent.
n=1

1
 ¯ n ¯
¯X ¯
Proof ¯¯
¯
am ¯¯¯ < K for all n
m=1
¯ v ¯ ¯ v u−1
¯ ¯ v ¯ ¯u−1 ¯
¯X ¯ ¯X X ¯ ¯X ¯ ¯X ¯
a m¯ = ¯ a m− a m¯ ≤ ¯ a m¯ + ¯ a m ¯ < 2K.
¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
¯
¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
m=0 1 1 1 1

ε
Given ε > 0, ∃ a natural number
¯ v
N = N¯ (ε)|phiv < 2K for all u ≥ N .
ε
¯X ¯ µ ¶
By Abel’s Lemma, therefore, ¯¯ am φm ¯¯ ≤ 2K = ε whenever
¯ ¯
m=0 2K
v ≥ u ≥ N ⇒ an φn converges by general principle of convergence.
P

Theorem 7 Abel’s Test Suppose that φn is a monotonic sequence con-

∞
X
verging to a finite limit. Let an be a convergent series. Then an φ n
P
n=1
is convergent.

Proof 1. Suppose φn is monotonic decreasing and z

phin → l as m → ∞ therefore ψn is decreasing and ψn = φn − l →
0 as n → ∞. Therefore by Direchlet’s test an ψn converges.
P

Write

m
X
Ψ = lim
m→∞
an (φn − l)
r=1
Xm ∞
X
= lim an φ n − l an
m→∞
n=1 1
P∞ P∞
Therefore a=1 an φ n = Ψ − l 1 an .
2. Suppose φn is monotonic increasing and φn → L as m → ∞.
Write ψn0 = l − φn ψn0 is increasing and ψn0 → 0. Therefore as
before an φn converges.
P

Theorem 8 Root Test The series un converges or diverges according as

P
1
lim(un ) n is greater than or less than one.
1
Proof 1. Suppose limn→∞ (un ) n = α < 1. Choose β|α < β < 1. Take
ε = β − α > 0. ¿From the property of the upper limit, ∃m =
1
m(β)|(un ) n < β for all n ≥ m so un < β n for all n ≥ m. Therefore
un converges by comparison with β n .
P P
1 1
2. Suppose limn→∞ (un ) n = α > 1 then (un ) n > 1 for an infinity of
n therefore un > 1 for an infinity of n therefore un 6→ 0 as n → ∞
therefore un diverges.
P

2
Theorem 9 ∃ a number R such that the power series an z n converges ab-
P
1
solutely for |z| < R and diverges for |z| > R, and R−1 = limn→∞ |an | n ,
with the appropriate conventions when RHS=0 or +∞.
1 1 1
Proof (i) if |an | n → 0 as n → ∞ |an z n | n = |an | n |z| → 0 as n → ∞ for all
z therefore by Root rest |an z n | converges.
P

1
(ii) If lim|an | n = ∞ the power series does not converge for z 6= 0 since
1 1
lim|an z n | n = lim|an | n R = +∞.
1
(iii) If lim|an | n is finite and non-zero, we write it equal to R1 R >
1
0 lim|an z n | n = |z|
R
. Hence by root test, the series converges or
diverges according as |z| < R or |z| > R. R is called the radius of
convergence.
1
R−1 = lim|an | n with, conventionally, R = 0 if RHS= +∞ and
R = ∞ is RHS=0.