Sei sulla pagina 1di 33

Notes on analysis

1 Notes to chapter I: Real numbers


Theorem 1. Bernoulli inequality
For all a > −1, a ∈ R. For all n ∈ N
n
(1 + a) ≥ 1 + na

Proof. Fix a > −1.


Induction on n.

Base: For n = 0:
1≥1

Step:
n+1 n
(1 + a) = (1 + a) (1 + a)
Hypothesis
≥ (1 + na)(1 + a)
= 1 + na + a + na2
See that na2 ≥ 0
≥ 1 + na + a
= 1 + (n + 1)a

Which concludes the proof.


Note that: equality is only possible if n = 1 ∨ a = 0.
Definition 1. Arithmetical and geometrical average
Suppose a1 , . . . , an are non-negative real numbers, then:
Arithmetical: An stand for arithmetical average here.
Pn
an
An := i=1
n

Geometrical: Gn stand for geometrical average here.


v
u n
uY
n
Gn := t an
i=1

Theorem 2. Let a1 , . . . , an stand for non-negative real numbers. Let


An , Gn stand for arithmetical and geometrical averages. Then:

An ≥ Gn

Proof. If ai = 0 for any 1 ≤ i ≤ n then thesis is trivial thus assume all


numbers are non-negative.
Induction on n:

1
Base: A1 = G1 , thesis holds.
Step: Assume thesis holds for n. We are considering case when we
have n + 1 terms.
Without loss of generality assume: a1 ≥ · · · ≥ an+1 .
Then, we have (we are giving an upper bound by taking a maximal
element): Pn
an nan
An ≤ i=1 = = an (1)
n n
Now we try to rewrite the next case to form relating it to Bernoulli
inequality.
nAn + an+1
An+1 = (2)
n+1
an+1 − An
= An + (3)
m+1
 
an+1 − An
= An 1 + (4)
An (m + 1)
Let
an+1 − An
a :=
An (m + 1)
From equation 1 we get a > −1 (for Bernoulli).
We will show that:
An+1 n+1
n+1 ≥ Gn+1

which is equivalent to original statement.


n+1
An+1 n+1
n+1 = An (1 + a) Derivation of An+1
≥ An+1
n (1 + a(n + 1)) Bernoulli inequality
n
= An an+1 Definition of a
≥ Gnn an+1 inductive hypothesis
= Gn+1
n+1

Therefore
An+1 ≥ Gn+1

Theorem 3. Equation for difference of n-th powers


n
X
xn − y n = (x − y) xn−i y i−1
i=1

Proof. If y = 0 then the equality is obvious, so suppose y 6= 0. Then


introduce auxiliary variable
x
q :=
y

2
and see that after dividing both sides of the equation by y n we get:
n
X
q n − 1 = (q − 1) q n−i
i=1

By induction:
Base: q 1 − 1 = q − 1
Step: Suppose thesis holds for n. Then:
n+1
X n
X
(q − 1) q n+1−i = (q − 1) q n−i
i=1 i=0
= (q − 1)q n + q n − 1
= q n+1 − 1

Theorem 4. If n, k ∈ N and k ≥ 2,

k
x := n

then x ∈ N ∨ x ∈ R \ Q.
Proof. First we
√ consider a case k = 2. We will argue by contradiction:
suppose x = n ∈ Q \ N. Then we can define a set:

A := {m ∈ N : mx ∈ N}

and from the properties of Q we know that it’s not empty. We pick the
smallest element:
m0 := min A
and choose:
m1 := m0 (x − bxc) = m0 x − m0 bxc
Clearly m0 > m1 > 0 and m1 ∈ N thus m1 ∈ A.
We get a contradiction, because m0 was suppose to be the minimal
element.


More general case, we fix k and x = k n, x ∈ Q \ N.
Consider a set
B := {s ∈ N : : xs ∈ N}
The set is non-empty, so lets denote its smallest element by s0 (notice
that s0 > 1).
Consider another set:

A := {m ∈ N : ∀ s < s0 : mxs ∈ N}

3
Again the set is non empty, so lets name its smallest element m0 . We
now choose:
m1 := m0 (xs0 −1 − bxs0 −1 c)
As in our toy example: m1 ∈ N and m1 < m0 .
Now we need to show: ∀ s < s0 : m1 xs ∈ N. So fix such s, then:
xs m1 = xs m0 (xs0 −1 − bxs0 −1 c)
xs0 − xs m0 bxs0 −1 c
= xs−1 m0 |{z}
| {z } | {z }
N N N
s
So x m1 ∈ N, therefore m1 ∈ A, but that is contradiction because
m1 < m0 = min A

2 Notes to chapter II: Sequences


Theorem 5. Triangular inequality
For all a, b ∈ R:
|a + b| ≤ |a| + |b|
Proof. By case analysis.


Theorem 6. Lim n
n √
lim n n = 1
n→∞

Proof. Let ε > 0. Let δn := | n n − 1|.
Then:
√ n n
n = n n = (δn + 1)
n  
X n k
= δ
k n
k=0

This means:  
n 2 n(n − 1) 2
n> δ = δn
2 n 2
Equivalently: r
2
δn <
n−1

So to make | n n − 1| less then ε it is enough to find N such that:
r
√ 2
| n − 1| <
n

n−1
But that is easy.

4

Theorem 7. Lim n
x, where x > 0

lim n x = 1
n→∞
√ √
Proof. Use squeeze theorem with n 1, n n, when x ≥ 1. Then see that
the other case reduces to first one:

n
1
x= √
n y

Theorem 8. n-Square root of sum


If 0 ≤ x0 ≤ · · · ≤ xk then
v
u k
uX
n
lim t xni = xk
n→∞
i=0

Proof. Squeeze theorem.


Theorem 9. Every sequence has a monotonic sub-sequence
Proof. Consider An := {am : m ≥ n}. If for all An there exists maximal
element, we pick maximal elements, otherwise we pick the smallest.
Theorem 10. Bolzano–Weierstrass
Every bounded sequence has a converging sub-sequence.
Proof. From previous theorem, choose monotonic sub-sequence, it is
bounded thus it converges.
Theorem 11. If all sub-sequence converge to the same limit, them
sequence converges to that limit
Proof. Suppose the limit of this sub-sequences is g. If g is not a limit
of a sequence, we can find ε0 such that there are infinitely many terms
outside (g − ε0 , g + ε0 ), out of them pick a converging sub-sequence and
get a contradiction.

Theorem 12. Suppose we have a sequence an and some sub-sequences


such that for any element of an it is an element of one of this sub-
sequences and that all have the same limit g. Then limn→∞ an = g
Proof. We fix ε > 0. For each sub-sequence there is Nm such that for
larger n: |an − g| < ε. Choose maximal Nm .
Theorem 13. Stolz-Cesàro theorem
Suppose bn is strictly monotonic and ∀ iinN : bi 6= 0. Suppose an is
other sequence and either:

5
1. limn→∞ an = limn→∞ bn = 0
or
2. limn→∞ bn = ∞

and
an+1 − an
lim =g∈R
n→∞ bn+1 − bn
then
an
lim =g
n→∞ bn

Proof. Without a loss of generality assume bn is strictly increasing (oth-


erwise: take −an , −bn ).
Fix ε > 0. We choose positive number η, but don’t specify its value
yet.
Choose k such that for all s > k:
as+1 − as
g−η < <g+η
bs+1 − bs

Note that bs+1 − bs > 0, thus:

(bs+1 − bs )(g − η) < as+1 − as < (bs+1 − bs )(g + η)

We take two number n, m such that n > m ≥ k and sum over inequal-
ities:
n−1
X n−1
X n−1
X
(g − η) (bi+1 − bi ) < (ai+1 − ai ) < (g + η) (bi+1 − bi )
i=m i=m i=m

this simplifies to:

(g − η)(bn − bm ) < an − am < (g + η)(bn − bm )

and
an − am
g−η < <g+η
bn − bm
Now we consider two cases:
1. limn→∞ an = limn→∞ bn = 0. In previous inequalities, fix m, and
consider:
an − am
g − η ≤ lim ≤g+η
n→∞ bn − bm
an −am am
Because an → 0 as n → ∞ then limn→∞ bn −bm = bm , we get
simply choose η < ε and have:

am
bm − g ≤ η < ε

6
2. limn→∞ bn = ∞. Then we can assume we have bn > 0 and:
(g − η)(bn − bm ) an − am (g + η)(bn − bm )
< <
bn bn bn
Now taking n → ∞ reduces the problem to previous point.
We also can see that assumption that g ∈ R can be dropped, that is
g = ±∞ is also valid.
Theorem 14. If
lim xn = x
n→∞

then sequence of arithmetical averages An of xn is also convergent to


x.
Proof. Using Stolz-Cesàro theorem with sum of elements and how many
there are::
Pn+1 Pn+1
i=0 xi − i=0 xi
lim = lim xn+1 = x
n→∞ (n + 1) − n n→∞

thus:
lim An = x
n→∞

3 Exponential function and logarithm


Theorem 15. Quickly converging to 1
Suppose limn→∞ nan = 0, then:
n
lim (1 + an ) = 1
n→∞

Proof. First, for large enough n:


1
|an | < |nan | < < |1 + an |
2
Bernoulli will come twice, first:
n
(1 + an ) ≥ 1 + nan
1 1
∴ n

(1 + an ) 1 + nan

Observe that:  n
1 an
= 1−
(1 + an )n 1 + an

7
Second Bernoulli:
 n
an nan
1− ≥1−
1 + an 1 + an
Collecting the inequalities:
 n
nan an
1− ≤ 1−
1 + an 1 + an
1
=
(1 + an )n
1

1 + nan
Thesis follow from application of squeeze theorem.

Definition 2. Exponential function exp(x)


We define:  x n
ex = exp(x) := lim 1 +
n→∞ n
As an intermediate step I will denote expn (x) then n − th element of
the sequence.
Theorem 16. expn (x) ≥ 0 for large enough n. Also expn (x) ≤
expn+1 (x).
Proof. Take n > |x|, then:
x
−1 < <1
n
and:
x
0<1+
n
so
0 < expn (x)
Consider
 n+1
x
expn+1 (x) 1 + n+1
= n
expn 1 + nx
x n+1
x  1 + n+1
 
= 1+
n 1 + nx
x+n+1 n+1
!
 x n+1
= 1+ x+n
n n
n+1
nx + n2 + n

 x 
= 1+
n (n + x)(n + 1)
 n+1
 x  x
= 1+ 1−
n (n + x)(n + 1)

8
Using Bernoulli (check conditions)
 
 x x
≥ 1+ 1−
n n+x
  
n+x n
=
n n+x
=1

Theorem 17. For all x ∈ R, exp(x) converges to positive numbers.


Proof. If x = 0 clearly exp(0) = 1. If x < 0 take large enough n, that
is n > |x| and:
x
0<1+ <1
n
thus  x n
0< 1+ <1
n
That means exp(x) is bounded and since its monotonic it converges.

Now take x > 0.


Again choose n > |x|.

x2  x  x
1− 2
= 1− 1+
n n n
Therefore: 2
x 1 − nx2
1+ =
n 1 − nx
Thus:  n
x2
1− n2
exp(x) =
exp(−x)
 2
n
x
The 1− n2 converges by theorem 15. The exp(−x) converges by
previous argument, therefore exp(x) converges. We also have exp(x) >
0 for all x by a previous theorem.
Theorem 18. exp(xy) = exp(x) exp(y)
Proof. We have:
!n !n
exp(x) exp(y) 1 + x+y
n +
xy
n2 xy
= lim = lim 1+ 2
exp(xy) n→∞ 1 + x+y
n
n→∞ n (1 + x+y
n )

Again theorem 15 comes handy.

9
Theorem 19. For all x ∈ R:
exp(x) ≥ 1 + x
Proof. Bernoulli and observation that expn (x) ≥ 1 + x for n > |x|
Theorem 20. Monotonicity of exp
x > y =⇒ exp(x) > exp(y)
Proof.
exp(x) − exp(y) = exp(y) exp(x − y) − exp(y)
= exp(y)(exp(x − y) − 1)
≥ exp(y)(x − y)
Both are non-negative
>0

Theorem 21. For all x < 1:


1
exp(x) ≤
1−x
Proof. If x < 1 then:
exp(x) ≥ 1 − x ≥ 0
Also
exp(x) exp(−x) = 1
So
1 1
exp(x) = ≤
exp(−x) 1−x

1
Theorem 22. If |x| ≤ 2 then:

|exp(x) − 1 − x| ≤ 2|x|2
Proof.
Thm 19 Thm 21
z}|{ z}|{ 1 x2
0 ≤ exp(x) − 1 − x ≤ −1−x= ≤ 2|x|2
1−x 1−x

Theorem 23. Continuity of exp If


lim xn = x
n→∞

then
lim exp(xn ) = exp(x)
n→∞

10
Proof. We should show, using previous theorem, triangle inequality and
squeeze thm that:
lim exp(xn − x) = 1
n→∞

but we skip it.


After that:

0 ≤ |exp(xn ) − exp(x)| = exp x|exp(xn − x) − 1|

and thus exp(xn ) → exp(x).


Theorem 24. Differentiation of exp

if hn → 0 and hn 6= 0 for all n

exp(x + hn ) − exp(x)
lim = exp(x)
n→∞ hn
Proof.

exp(x + hn ) − exp(x) exp(x)(exp(hn ) − 1)


lim = lim
n→∞ hn n→∞ hn
exp(hn ) − 1
= exp(x) lim
n→∞ hn

We know that for large enough n, hn < 21 , thus:



exp(hn ) − 1
0≤ − 1
hn

exp(hn ) − 1 − hn
=
hn
Thm 22
2|hn |2
≤ →0
hn
exp(hn )−1
Thus limn→∞ hn = 1.

Theorem 25. exp is bijection between R and R+


Proof. We already know ∀ x ∈ R : exp(x) > 0
Injectivity
Let exp(x) = exp(y) then

exp(x − y) = 1

then x − y = 0 so x = y.

11
Surjectivity. Suppose y ∈ (0, 1).

Let
A := {t ∈ R : exp(t) ≤ y}
For x ∈ A, we know x < 0, because x > 0 would imply exp(x) ≥ 1+x >
y.
Also, for n ∈ N
1
exp(−n) ≤
1+n
thus there are infinitely many value of this form in A.
So A is non-empty and bounded, thus x := sup(A) is a real number.
We will show x = y.
If x < y then by continuity of exp
1
lim exp(x + ) = exp(x) < y
n→∞ n
That means that for large enough n exp(x + n1 ) < y so x + 1
n ∈ A, but
that is a contradiction.
For x > y things look the same, we take
1
lim exp(x − ) = exp(x) > y
n→∞ n
and arrive at similar contradiction.
So it must be the case that x = y. By monotonicity x is unique.

1 1
If y > 1, then y < 1 and there is x such that exp(x) = y and y =
exp(−x).
Definition 3. Natural logarithm.
We define natural logarithm as inverse of exponential function on R+ :

ln := exp−1

See that for the definition ln(e) = 1, ln(1) = 0.


Theorem 26. For x, y > 0 we have ln(xy) = ln(x) + ln(y)

Proof. Let t, q be such that exp(t) = x, exp(q) = y

ln(xy) = ln(exp(t + q))


=t+q
= exp−1 (x) + exp−1 (y)
= ln(x) + ln(y)

12
Theorem 27. For x, y ∈ R+

x < y =⇒ ln(x) < ln(y)

Proof. Suppose x < y and ln(x) ≥ ln(y), then by monotonicity of exp

x = exp ln(x) ≥ exp ln(y) = y

Theorem 28. For y > 0


1
1− ≤ ln y ≤ y − 1
y
Proof. Let x = ln y. Then:

exp(x) ≥ 1 + x

equivalent
exp(x) − 1 ≥ x
Substituting relevant symbols

y − 1 ≥ ln(y)

Now suppose y < e (x < 1), then


1 1
y = exp(x) ≤ =
1−x 1 − ln y
which is equivalent to
1
ln y ≥ 1 −
y
When ln y ≥ 1 (y ≥ e), then inequality is obvious:
1
1− < 1 ≥ ln y
y

Note that this is equivalent to (for t > −1)


t
≤ ln(1 + t) ≤ t
t+1
Theorem 29. If tn > −1 for all n and tn → 0 then

lim ln(1 + tn ) = ln(1)


n→∞

Proof. Use previous theorem and squeeze.

13
Theorem 30. Continuity of ln.
If yn → y > 0 and yn > 0 for all n, then

lim ln yn = ln y
n→∞

Proof.
  
yn − y
ln yn − ln y = ln y 1 + − ln y
y
 
yn − y
= ln 1 + + ln y − ln y
y
 
yn − y
= ln 1 +
y

Therefore Let
yn − y
tn :=
y
then
yn
−1 < tn = −1 +
y
and tn → 0, thus from theorem 29.

lim ln(1 + tn ) = ln 1 = 0
n→∞

Theorem 31. Differentiation of ln.


Suppose hn → 0 and y > 0, y + hn > 0 for all n ∈ N, then:

ln(y + hn ) − ln y 1
lim =
n→∞ hn y
Proof.
 
hn
ln(y + hn ) − ln(y) ln y + ln 1 + y − ln y
=
hn h
 n
hn
ln 1 + y
=
hn
ln(1 + hyn ) 1
= hn

y
y
hn
ln(1 + 1 y )
= ∗ hn
exp ln(1 + 1 y y )−
  −1
0 hn 1
= exp ln 1 + ∗
y y

14
hn
Clearly ln(1 + y ) → 0, so by taking derivative at exp0 (0)
1
=
y

Definition 4. Power function:


For all a > 0 and x ∈ R we define

ax := exp(x ln a)

Theorem 32. Properties of power function


1. a0 = 1
2. ax+y = ax ay
3. a−x = 1
ax
y
4. (ax ) = axy
x
5. ax bx = (ab)
6. If a > 1 then x < y =⇒ ax < ay and if 0 < a < 1 then
x < y =⇒ ax > ay

7. If xn → x for n → ∞ then

lim axn = ax
n→∞

8. If a < b and x > 0 then ax < bx and if x < 0 then ax > bx .


Proof.
1.
a0 = exp(0 ln a) = exp(0) = 1

2.

ax+y = exp((x + y) ln a) = exp(x ln a) exp(y ln a) = ax ay

3.
1 1
a−x = exp(−x ln a) = = x
exp(x ln a) a
4.
y
(ax ) = exp(y ln exp(x ln a)) = exp(yx ln a) = axy

5.
x
ax bx = exp(x ln a) exp(x ln b) = exp(x ln(ab)) = (ab)

15
6. Let a > 1 and x < y

ax = exp(x ln a) < exp(y ln a) = ay

similar argument for second inequality


7.

lim axn = lim exp(xn ln a)


n→∞ n→∞
 
= exp ln a lim xn
n→∞
= exp(x ln a)
= ax

8.
ax = exp(x ln a) < exp(x ln b) = bx
similar for second case.

Definition 5. Logarithm
If a > 0, a 6= 1 then for all y > 0

loga y = x ⇐⇒ ax = y

Theorem 33. Logarithms are well-defined


Proof. Let a > 0, a 6= 1, then it is enough to check that ax is a bijection
between R and R+ .

Similar argument to the one showing exp is bijection, will work.


Theorem 34. Properties of logarithms.
Let x, y > 0, and a, b > 0, a, b 6= 1
1.
ln y
loga y =
ln a
2.
logb y
loga y =
logb a
3.
loga (x + y) = loga x + loga y

4.
loga (y x ) = x loga y

Proof.

16
1.

aloga y = y
(loga y)(ln a) = ln y
ln y
loga y =
ln a

2.
ln y
logb y ln b
= ln a
logb a ln b
ln y
=
ln a
= loga y

3.
ln(xy)
loga (xy) =
ln a
ln x ln y
= +
ln a ln a
= loga x + logb y

4.
ln(y x )
loga (y x ) =
ln a
Unfolding definition of power in terms of exp
x ln y
=
ln a
ln y
=x
ln a
= x loga y

Theorem 35. Characterisation of exp function.


If f : R → R complies with:
1. f (x + y) = f (x)f (y)
2. f (1) = e

3. For xn → x
lim f (xn ) = f (x)
n→∞

then f = exp

17
Proof. Suppose x ∈ R, x 6= 0 then
 x 2
f (x) = f >0
2
Also f (0) 6= 0, because:

0 < e = f (1 + 0) = f (1)f (0)

By induction we show f (m) = em , step:

f (m + 1) = em e = em+1

Extending this to integers:


1
f (−m) = = e−m
f (m)

If n ∈ N then:
1 n
f( ) = e
n
thus
1 1
en = f( )
n
m
=f m

By similar arguments e n n . By continuity for irrational numbers
we get:
f (x) = lim f (xn ) = lim exp(xn ) = exp(x)
n→∞ n→∞

Theorem 36. If g : R+ → R
1. g(xy) = g(x) + g(y)

2. g(e) = 1
3. limn→∞ g(xn ) = g(limn→∞ xn )
then g = ln
Proof. Similar to previous theorem.

18
4 Notes to chapter IV: Series, exp in com-
plex plane
Theorem 37. k0 ∈ N then following series either both converge or
both diverge:
X∞ X∞
an an
n=0 n=k0

Proof. Cauchy criterion either holds for both or fails for both.
Theorem
P P38. Linearity of series
If an , bn converge and c1 , c2 ∈ R, then:

X ∞
X ∞
X
c1 an + c2 bn = c1 an + c2 bn
n=0 n=0 n=0

Proof. Immediate from the definition and algebra of limits.


Theorem
P 39. Sequence limit and series convergence
Given an if
lim an 6= 0
n→∞
P
then an diverges.
Proof. We
Pprove contraposition.
Suppose an converges, then

lim an = lim sn − sn−1


n→∞ n→∞
= lim sn − lim sn−1
n→∞ n→∞
X X
= an − an
=0

q n diverges
P
Theorem 40. If |q| > 1 then
Proof. Theorem 39.
We should present now the comparison test, but because in it’s basic
form it is very intuitive, we skip it.

Theorem 41. Comparison test, fraction version


Suppose that for large enough n an , bn > 0 and
an+1 bn+1

an bn
then

19
P P
1. From convergence of bn follows convergence of an
P P
2. From divergence of an follows divergence of bn
Proof. Without loss of generality assume n = 1 is large enough. Let N
be arbitrary, then:
N N
Y ai Y bi

a
i=2 i−1
b
i=2 i−1
We simplify both side
aN bN

a1 b1
a1
Set c := b1
aN ≤ cbN
and thesis follows from the basic version of comparison criterion.

P∞
Theorem 42. Sum of n=1 nq n for |q| < 1
Proof. The crucial observation here is that
k
X k
X k
X
1 + 2q 2 + 3q 3 + · · · + kq k = qn + qn + · · · + qn
n=1 n=2 n=k


X k
X
nq n = lim nq n
k→∞
n=1 n=1
k X
X k
= lim qn
k→∞
n=1 m=n
k
X q n − q k+1
= lim
k→∞
n=1
1−q
k
1 X n
= lim q − q k+1
k→∞ 1 − q
n=1
k
! !
1 X
= lim q n − nq k+1
k→∞ 1 − q
n=1
1 − q k+1
 
1 k+1
= lim − nq
k→∞ 1 − q 1−q
1 − q k+1 − nq k+1 (1 − q)
= lim 2
k→∞ (1 − q)
1
= 2
(1 − q)

20
Theorem 43. Condensation test
If an > 0 for all n and an is monotonic and decreasing then

X ∞
X
an 2n a2n
n=0 n=0

are either both converging or both diverging.


We can also write, more generally

X ∞
X ∞
X
an ≤ 2n a2n ≤ 2 an
n=1 n=0 n=1

Proof. Informal argument is easier to follow:



X
an = a1 + a2 + a3 + a4 + a5 + a6 + a7 + . . .
|{z} | {z } | {z }
n=1 First group Second group Third group

≤ a1 + a2 + a2 + a4 + a4 + a4 + a4 + . . .
|{z} | {z } | {z }
First group Second group Third group
X∞
= 2n a2n
n=0
= a1 + a2 + a2 + a4 + a4 + a4 + a4 + . . .
|{z} | {z } | {z }
First group Second group Third group

= a1 + a2 + a2 + a3 + a4 + a4 + . . .
| {z } | {z }
First group Second group

≤ a1 + a1 + a2 + a2 + a3 + a3 + . . .
| {z } | {z }
First group Second group
X∞
=2 an
n=1

Formalizing this argument is a bit tedious and not really time worthy.

Theorem 44. Harmonic series diverges



X 1
=∞
n=1
n

Proof. Using theorem 43.


∞ ∞
X 1 X 1
∼ 2n n = ∞
n=1
n n=1
2

21
Theorem 45. Harmonic and logarithm

X 1
=∞
n=2
n ln n

Proof. Using thm 43.


∞ ∞
X
n1 X 1
2 n n
= ln 2 =∞
n=2
2 ln 2 n=2
n

Definition 6. We define ζ(s) function for s > 1:



X 1
ζ(s) :=
n=1
ns

Theorem 46. For s > 1


ζ(s) ∈ R
Proof. Using thm 43.

∞ ∞
X 1 X 1
2n n s = n(s−1)
n=1
(2 ) n=1
2
∞  n
X 1
=
n=1
2s−1

and because s > 1 1 − s < 0 so 21−s < 1 and because series is geometric
(q = 2s−1 ) it converges.
Theorem 47. Harmonic log and power.
If s ∈ R then
X 1
s
n(ln n)
converges iff s > 1.
Proof. Using thm 43.
X 1 s
X 1 s
2n s = (ln 2) = (ln 2) ζ(s)
2n (ln 2n ) ns
Follows from thm 46.
Theorem 48. I generalization of condensation
If an is positive decreasing sequence and k ∈ Z+ then

X ∞
X ∞
X
an ≤ k n akn ≤ k an
n=1 n=1 n=1

22
or equivalently

X ∞
X
an k k n akn
n=1 n=1
both converge or both diverge.
Proof. Identical to thm 43.

Theorem 49. Let P be a set of prime number and pn the sequence,


then
X∞
pn = ∞
n=1

Proof. Fix some N , then:


 
 
X 1 Y 1
exp = exp
p p
p∈P : p≤N p∈P : p≤N

Using exp(x) ≥ 1 + x
 
Y 1
≥ 1+
p
p∈P : p≤N

Let O := {n ∈ Z : n ≤ N ∧ Doesn’t exists p ∈ P such that p2 divides n}


+
X 1

n
n∈O

Last step is a bit hand-wavy, but after staring at it for a while we realize
that O is indeed a subset of sum acquired
P 1 by opening the brackets.
We’ve already showed that ζ(2) = n2 converges so:
 
X 1  ≥ ζ(2)
X 1
ζ(2) exp
p n
p∈P : p≤N n∈O
N
X 1

n=1
n

Thus: !
N
X 1 X 1
≥ ln − ln ζ(2)
p n=1
n
p∈P : p≤N

by taking N → ∞ we see that series diverges.

Theorem 50. Kummer


P test
If an > 0, then an converges iff there exists θ > 0 and bn > 0 such
that
an
∀ n > n0 : bn ∗ − bn+1 ≥ θ
an+1

23
P 1
P∞
Proof. If an converges then choose bn := an k=n+2 ak . Then
∞ ∞
an 1 X 1 X
bn ∗ − bn+1 = ak − ak
an+1 an+1 an+1
k=n+1 k=n+2
an+1
=
an+1
So choose θ := 1.

On the other hand, if there exists such bn and θ:


an
bn ∗ − bn+1 ≥ θ
an+1
bn an − an+1 bn+1
≥θ
an+1
θ−1 (bn an − an+1 bn+1 ) ≥ an+1

We have a telescoping series:



X
θ−1 bn an − an+1 bn+1 = θ−1 a1 b1
n=0
P
By comparison test an converges.

Theorem 51. Raabe convergence test


If there exists an > 0 and s > 1 such that
 
an
n −1 ≥s
an+1
P
then an converges
Proof. Use Kummer test with bn = n

Definition 7. Convergence in C
Let zn ⊆ C. We say that zn converges to z iff

∀ ε > 0 : ∃ nε : ∀ n ≥ nε : |zn − z| < ε

Definition 8. Disk
Suppose z ∈ C, ε ∈ R, then we define disk:

D(z, ε) := {w ∈ C : |w − z| < ε}

The geometric interpretation of convergence to z is that for all ε > 0


there is a point in sequence after which all terms will land in disk
D(z, ε).

24
Theorem 52. If w = a + ib ∈ C (where a, b ∈ R) then:

max(|a|, |b|) ≤ |w| ≤ |a| + |b|

Proof. Without loss of generality assume |a| ≥ |b|, then

max(|a|, |b|) = |a|


p
≤ a2 + b2
= |w|
p
≤ a2 + 2ab + b2
= |a + b|
≤ |a| + |b|

Natural interpretation is given by imagining complex number in the


plane and looking at sides of the triangle form by described objects.
Theorem 53. Sequence of numbers zn = xn + iyn ∈ C where xn , yn ∈
R converges to z = x + iy iff

lim xn = x ∧ lim yn = y
n→∞ n→∞

Proof. Using previous theorem we write:

max(|xn − x|, |yn − y|) ≤ |zn − z| ≤ |xn − x| + |yn − y|

Thesis follows from squeeze thm.


P
Theorem
P 54.PSeries zn where zn = xn +iyn for xn , yn ∈ R converges
iff xn and yn converge.
Proof.
k
X k
X
lim zn = lim xn + iyn
k→∞ k→∞
n=0 n=0

X ∞
X
= xn + i yn
n=0 n=0

Theorem 55. Sequence of complex numbers converges iff its a Cauchy


sequence.
Proof. This follows from Thm 53 and previous inequality.
Theorem 56. Complex series converges iff it complies with Cauchy
criterion.
Proof. Immediate consequence of previous theorem.

25
Definition
P 9. Absolute convergence and conditional convergence.
Let zn be some series, then:
P
Absolute convergence: |zn | converges.
P P
Conditional convergence: |zn | diverges but zn converges.
Theorem 57. If series converges absolutely it converges.
Proof. By generalized triangle inequality:
n n
X X
zi ≤ |zi |



i=m i=m

Series converges absolutely, so Cauchy criterion tells us that series con-


verges.
P
Theorem 58. Suppose xn ⊆ C converges absolutely and sequence
yn is obtained by rearrangement of terms of xn , then
X X
xn = yn

Proof. Being obtained by re-agreement means there exists bijection be-


tween elements of sequences xn and yn . Lets call it π : N → N. Now we
have:
∀ i : yi = xπ(i)
We will show that

X
∀ ε > 0 : ∃ nε : ∀ p, n > nε : |yk | < ε
k=n+1

We start by fixing ε > 0.P


Then by convergence of |xn | we know that:
r
X
∃ N : ∀ m, r > N : |xn | < ε
k=m+1

We fix that N and notice that {xn : i ≤ N } is finite. That means there
exists N 0 such that

{xn : i ≤ N } ∩ {xπ(i) : i > N 0 } = ∅

Now we set nε := N 0 .
We fix some p, n > nε , assume p > n. From the definition

{yi : n < i ≤ p} = {xπ(i) : n < π(i) ≤ p}


Now, because n, p > N 0
⊆ {xπ(i) : i > N 0 }
⊆ {xi : i > N }

26
We define two extra elements:
s := min{i : n < π(i) ≤ p}
h := max{i : n < π(i) ≤ p}
We define them because now we can find a sequence in xn such that
yn+1 , . . . , yp is its sub-sequence. That is because, the following holds:
{yi : n < i ≤ p} ⊆ {xi : s ≤ i ≤ h}
But now:
p
X h
X
|yi | ≤ |xi |
i=n+1 i=s

On the other hand s, h > N , therefore:


h
X
|xi | < ε
i=s

and by transitivity of order:


p
X
|yi | < ε
i=n+1

From our observation it also follows that:



X
|xi − yi | = 0
i=0

because
n
X n
X
|xi − yi | ≤ |xi |
i=0 i=m
for some tail where m get arbitrary large.

P
Theorem 59. If an ⊆ R and an converges conditionally, then for
any x ∈ R there exists π : N → N such that

X
aπ(n) = x
n=0

Proof. Obviously there are infinitely many negative terms, because oth-
erwise the sequence would converge absolutely.
WLOG assume an 6= 0 for all n.
Let pn , nn represent all positive and all negative terms of an .

X
pn = ∞
X
nn = −∞

27
Otherwise we would have absolute convergence.
We define the new sequence mn inductively.
We take smallest n0 such that
n0
X
pi > x
i=0

and this is our first n0 elements of mn .


New we take n1 such that
n0
X n1
X
pi + nn < x
i=0 i=0

and take n1 elements from nn .


Similarly we select larger elements from both sequences constructing
the bijection. P
Our sum is oscillating around number x, and since an → 0 we know
that an → 0. This means that our approximation of x will get better
and better, thus X
aπ(n) = x

Note that in previous we can easily construct an such that x = ±∞, we


just construct a sequence that is growing reasonable fast (for example
after each turn as seen above select numbers such that result of the
turn increases by 1).

Theorem 60. Abel’s theorem


Let an , bn ⊆ C. Suppose there exists M such that
n
X
∀ n : |An | = An ≤ M


i=0

and

X
|bn − bn+1 | < ∞
n=0

and limn→∞ bn = 0.

Proof.
P Lemma: First wePwill show that if αn is bounded sequence and
|βn | converges, then αn βn converges absolutely. Take M upper
bound, then X X X
|αn βn | < |αn M | = M |αn |

28
Now we try re-write an bn slightly.

n
X n
X
ak bk = b1 A1 + bk (Ak − Ak−1 )
k=1 k=2
n−1
X
= An bn + Ak (bk − bk+1 )
k=1

Since An bn = an bn then an bn → 0 as n → ∞ because an ≤ M and


bn → 0.
Pn−1
And from the lemma it follows that k=1 Ak (bk − bk+1 ) converges
absolutely.
P
Thus an bn converge.

Theorem 61. Dirichlet’s test


bn ⊆ R.
If limn→∞ bn = 0 where P P
And partial sums An of an are bounded then an bn converges.
Proof. Because bn ≥ bn+1 and bn → 0

X ∞
X
|bn − bn+1 | = (bn − bn+1 ) = b1
n=1 n=1

therefore we can apply Abel’s theorem.


Theorem 62. Leibniz test
If bn ⊆ R and limn→∞ bn = 0 then

X n
(−1) bn
n=0

converges.
P n
Proof. Clearly partial sums of = (−1) are bounded and because
limn→∞ bn = 0 we can apply Dirichlet’s test.
Theorem 63. Abel’s test
If bn ⊆ R+ is decreasing (monotonic) and
P
an converges then
X
an bn

converges
Proof. We have
X X X
an bn < an b1 = b1 an

29
Definition 10. Cauchy
P product
P of series
If an , bn ⊆ C and an , bn converge, then we define the Cauchy
product of series C as:
 
X∞ X∞ Xn
C= cn =  aj bn−j 
n=0 n=0 j=0

Theorem
P∞64. Mertens theorem P∞
If A = n=0 an converges and B = n=0 |bn | converges then
 
X∞ X∞ X
∞ n
X
C= cn =  aj bn−j 
n=0 n=0 n=0 j=0

converges and if A converges absolutely, then it converges absolutely.


Pn
Proof. Let Cn stand for i=0 ci . Then:

CN = a0 b0 + (a0 b1 + a1 b0 ) + · · · + (a0 bN + a1 bN +1 + · · · aN b0 )
N
X N
X −1
= b0 an + b1 an + · · · + bN a0
n=0 n=0
= b0 AN + b1 AN −1 + · · · bN A0
N
X
= bn AN −n
n=0

We will show that


lim Cn = AB
n→∞

Let ε > 0 and choose η which we will specify later.


We have:

|CN − AB| = |Cn + ABr + ABr − AB|


≤ |Cn + ABr | + |A||Br − B|

We will estimate each of the terms separately.

Since limn→∞ Bn = B, limn→∞ An = A


There exists such nB and nA that if r > max(nB , nA ) then

|An − A| < η
|Br − B| < η
|A||Br − B| < |A|η

30
We have

N r
X X

|Cn − ABr | = bj AN −j − A bj
j=0 j=0

N N
X X

= bj (AN −j − A) + bj AN −j
j=0 j=r+1
 
r

X X

≤  sup |AN −j − A| ∗ |bj | + sup|An | ∗
bj
0≤j≤r j=0 n∈N j=r+1
≤ η ∗ β + sup|An ||Br − B|
n
≤ η(β + sup|An |)
n

Thus
|CN − AB| ≤ η(β + sup|An | + |A|) < η(β + sup|An | + |A| + 1)
n n

Choose  −1
η := ε β + sup|An | + |A| + 1
n

Definition 11. exp in complex plain.


 z n
exp(z) := lim 1 +
n→∞ n
Theorem 65. Series for exp
For all z ∈ C,

 z n X z n
lim 1 + =
n→∞ n n=0
n!
moreover the series on right side of equality converges absolutely.
Proof. Skipped for now.
Theorem 66. de Moivre
If α ∈ R then for all n ∈ N

cos(nα) + i sin(nα) = (c cos α + i sin α)n


Proof.
cos(nα) + i sin(nα) = exp(inα)
n
= exp(iα)
n
= (cos α + i sin α)

31
Theorem 67. Real trigonometric functions in term of sin, cos
If x ∈ R

sin(x) = =(exp(ix))
cos(x) = <(exp(ix))

and
n
sin(x) = =(exp(ix))
n
cos(x) = <(exp(ix))

Proof. The first two equation are immediate from equation expressing
exp in terms of sin, cos.
Then by de Movair
i sin(nα) =

32