Sei sulla pagina 1di 10

Theorem 1. The intermediate value theorem If f : [a, b] R is continuous and f (a) 0 f (b) then there exists a c [a, b] such that f (c) = 0.

Proof.

Part A: Set

a 0 := a

b 0 := b

c 0 := a + b

2

We have two options: f (c 0 ) 0 xor f (c 0 ) < 0.

If first option holds, we set a 1 := c 0 , b 1 = b 0 otherwise we proceed symmetrically.

We define a n , b n inductively.

Part B: The sequence a n is monotonic and bounded, thus it converges to some number

c.

On the other hand

b

n

a n

=

b n1 a n1

2

so its not hard to notice that

From this we get

so taking n → ∞, b n c.

b

n

a

n

=

b n = a n +

2 n (b 0

a 0 )

2 n (b 0 a 0 )

Part C: since f is continuous at c and a n c, thus f (a n ) f (c).

Because n : f (a n ) 0, therefore

0.

f (c) = 0 follows.

f (c) 0.

Similarly n : f (b n ) 0, thus f (c)

∀ n : f ( b n ) ≤ 0, thus f ( c ) ≤

Notice the similarity with binary search.

Definition 1. Derivative

Let U

with derivative f (t) if for all ε > 0, there exists a δ(t, ε) > 0, such that

be an open set in R, we say that a function f : U R is differentiable at t U

(t δ(t, ε), t + δ(t, ε)) U

and if x|h| < δ(t, ε) then

f(t + h) f(t)

h

f (t) < ε

Theorem 2. Derivative is linear Suppose U is an open set and f, g : U R are differentiable at t U . If λ, µ R, show

[λf + µg] (t) = λf (t) + µg (t)

1

Proof. Fix ε > 0.

λ f(t + h) f(t)

(λf + µg)(t + h) (λf + µg)(t)

h

h

f(t + h) f(t)

h

λf (t) + µg (t)

f (t) + µ g(t + h) g(t)

h

f (t) + |µ|

g(t + h) g(t)

h

g (t)

g (t)

=

≤|λ|

Choose δ(f, t, ε) :=

|2λ| , δ(g, t, ε) :=

ε

ε

|2λ|

| 2 λ | , δ ( g, t, ε ) := ε ε | 2

Now we turn to mean value inequality, which will be use to prove seemingly obvious theorem about functions with derivative equal zero.

Theorem 3. The mean value inequality Let U be the open interval (α, β). Suppose

K 0, and that a, b U then

: U R is differentiable with t : f (t) K

with a < b.

If f

f(b) f(a) (b a)K

Proof. The proof with follows from the next lemma. (K + ε)(b a) for all ε > 0 than as ε 0 we get

Notice that if f (b) f (a)

f(b) f(a) (b a)K

f ( a ) ≤ f ( b ) − f ( a ) ≤ (

Lemma 1. Using assumption of the previous theorem.

ε > 0: f (b) f (a) (K + ε)(b a)

Proof. Suppose

f (b) f (a) > (K + ε)(b a)

We refer to binary search. Set

We have

a 0 := a

b 0 := b

c 0 := a 0 + b 0

2

(1)

(f(c 0 ) f (a) (K + ε)(c 0 a 0 )) + (f (b 0 ) f(c 0 ) (K + ε)(b 0 c 0 ))

=f (b) f (a) (K + ε)(b a) > 0

So at least one of the expression in a first term must be strictly positive. If (f (b 0 )f(c 0 ) (K + ε)(b 0 c 0 )) > 0 we set a 1 := c 0 , b 1 := b 0 otherwise a 1 := a 0 , b 1 := c 0 . In both cases

f(b 1 ) f(a 1 ) > (K + ε)(b 1 a 1 )

a 0 a 1 b 1 b 0

b 1 a 1 = b 0 a 0

2

2

Extend the definition inductively. As in intermediate value theorem, sequences a n , b n are bounded and converge to the same number c. Since f is differentiable at c, there exists δ such that (c δ, c + δ) U and whenever 0 < |h| < δ

Thus

Since f (c) K we have

f(c + h) f(c)

h

f (c) < ε

2

|f(c + h) f(c) f (c)h| < |h|ε

2

 

for

0 h

< δ

f (c + h) f (c) (K + ε/2)h

for

δ

h

0

f (c) f (c + h) ≤ −(K + ε/2)h

Since

a n , b n c

we can find such N that |a n c| < δ and |b n c| < δ. Therefore

But

for

for

h

h

:=

:=

a N

c

b N c

f(c) f(a N )

f(b N ) f(c)

(K + ε/2)(c a N )

(K + ε/2)(c b N )

f(b N ) f(a N ) = f(b N ) f(c) + f(c) f(a N )

(K + ε/2)(c b N ) + (K + ε/2)(c a N )

= (K + ε/2)(b N a N )

Contradicting (1).

Contradicting (1).

Theorem 4. The constant value theorem Let U R be the open set. If f : U R is differentiable with

 

t U : f (t) = 0

then f is constant.

Proof. From mean value inequality, choose K := 0

 
 

f(b) f(a)

(b a) 0 = 0

and

 

(f )(b) (f )(a) = f (a) f (b) 0

 

thus f (b) = f (a) for any

a, b U .

 
 

Theorem 5. If U R is an open set and function f, g : U R are differentiable with t U : f (t) = g (t) then there exists a constant c such that f (x) = g(x) + c for all x U.

Proof. Apply previous theorem (constant value theorem) to f g

3

f ( x ) = g ( x ) + c for all x ∈ U

Theorem 6. If U R is the open set, a, b U and a > b and g : U R is differentiable with t U : g (t) 0 then

g(b) > g(a)

Proof. The statement is equivalent to:

(g)(b) < (g)(a)

and

(g)(b) (g)(a) < 0

But derivative of g is always negative thus after setting K := 0 in mean value inequality the result follows immediately.

 
 

Definition 2. sup We define supremum as usual: the least upper bound

 

Theorem 7. Every bounded non-empty set of reals has a supremum

 

If =

A R has an upper bound M , then A has a supremum.

Proof. We do binary search again.

 
Proof. We do binary search again.  

Theorem 8. Intermediate value theorem (again) If f : [a, b] R is continuous and

 

f(a) 0 f(b)

 

then there exists a c [a, b] such that f (c) = 0.

Proof.

Part A: Let

 
 

E := {x [a, b]: f (x) 0}

 

E is non-empty because a E and E is bounded by b.

Part B: Let c := sup E.

 

Part C: Fix ε > 0. f is continuous at c so there exists δ > 0 such that

 
 

x [a, b] ∧ |x c| < δ

=

|f (x) f (c)| < ε

We want to show |f (c)| < ε.

Consider three cases of comparing c with a, b.

 

1. a < c < b todo

2. c = b todo

3. c = a todo

 
 

Theorem 9. Bolzano-Weistrass: Every bounded sequence has converging subsequence If a n is a sequence such that for some M R |a i | ≤ M for all i then there exists a sub-sequence a n that converges.

4

Proof. We ask, are there infinitely many elements a i such that

j > i: a i a j

If yes we select them and have bounded monotonic sequence — done. If no, then for any a i there exists a j such that

a i < a j

For every element we select next, smaller one, and get a bounded monotonic sequence — done.

Definition 3. Limit supremum & infimum

lim sup x n = lim

n→∞

n sup{x m : m n}

lim inf x n = lim

n→∞

n inf{x m : m n}

= lim n →∞ n → ∞ inf { x m : m ≥ n }

General remark! There are at least three versions of fundamental axiom. Fundamental axiom, existence of supremum and Bolzano-Weierstrass.

Definition 4. Inner product If x i , y i are vectors there inner product is:

(x) i · (y) i = (x y) i

sometimes denoted by x, y

Definition 5. Euclidean norm If x is a vector.

x := x, x

Theorem 10. The Cauchy-Schwarz inequality If x, y R m , then

|x · y| ≤ x y

Proof. If x = 0 y = 0, the inequality is trivial. Otherwise

Set λ := (x·y)

x 2

Rearranging

0 (λx + y) · (λx + y)

= λ 2 x 2 + 2λ(x · y) + y 2

=

=

λ x + x · y 2 + y 2 (x · y) 2

x

x 2

y 2 (x · y) 2

x 2

0

(x · y) 2

|x · y|

y 2 (x x · y) 2 2

x 2

x y

y 2

5

x · y ) 2 | x · y | ≤ y 2 − ( x

Theorem 11. Standard property of Euclidean norm in R m For all x, y R m :

1.

x 0

2.

x = 0

=

x = 0

3.

λ R then λx = |λ| x

4.

x + y x + y

Proof. The triangle property can be deduced from Cauchy-Schwarz as follows:

x + y 2

=

(x + y) · (x + y)

=

x 2

+ 2(x · y) + y 2

x 2

+

2 x y + y 2

=

( x + y ) 2

x y + y 2 = ( x + y ) 2 Definition 6. Convergence in

Definition 6. Convergence in R m with Euclidean norm We say that a n converges to a as n → ∞ if for any ε > 0 there exists n ε such that for all n n ε

a n a < ε

Theorem 12. Properties of limit in Euclidean space Given R m , following holds:

1. Limit is unique. If a n a and a n b then a = b.

2. In converging sequence every subsequence converges to the same limit. If a n a then a f(n) a.

3. Constant sequence converges. If a n = c for all n then a n c.

4. If a n a and b n b then a n + b n a + b.

5. Suppose a n R m and λ n R then if a n a and λ n λ we have λ n a n λa

6. a n · b n a · b

Lemma 2.

Proof:

x · y = x + y 2 x y 2

4

x + y 2 + x y 2

i (x 2 + 2xy + y 2 x 2 + 2xy y 2 ) i

 

=

4

4

 

i (4xy) i

 

=

 

4

4(x · y)

 

=

 

4

 

=

x · y

Proof.

6

1.

For any ε > 0 following holds:

a b a n a + a n b < ε

Since ε can be arbitrary a b = 0, thus a b = 0, and a = b.

2. Fix ε. There exists n ε such that a n a < ε for all n n ε . For the same n ε , the same must holds for sub-sequence.

3. From the definition of convergence.

4. Immediate consequence of two ε method and triangle inequality.

a n + b n (a + b) a n a + b n b

5. Fix ε > 0. Observe:

λ n a n λa = |λ|

λ n

λ

a n a

We can find n a such that

a n a <

ε

2|λ|

and n λ such that

λ

n

λ a n a n

<

ε

2|λ|

Combining the two and using triangle inequality

ε

λ n a n a <

λ

|λ|

thus

λ n a n λa = |λ|

< |λ|

λ n

λ

ε

|λ|

= ε

a n a

6. Fix ε > 0. From the lemma

a n · b n a · b =

Taking n → ∞ 1

= 0

a n + b n 2 a n + b n 2 ( a + b 2 a b 2 )

4

2 − ( a + b 2 − a − b 2 ) 4 Theorem 13.

Theorem 13. Bolzano-Weierstrass in euclidean space If x n R m and there exists K such that x n K for all n, then there exists sub-sequence that converges.

7

(1)

Proof. Suppose x n := (x Observe that for any n, i

n

so

, x

(2)

n

.

.

.

|x

(i)

n

,

x

(m)

n

| 2

).

x

(i)

n

1im

|x

(i)

n

| ≤ x n

· x

(i)

n

and

|x

(i)

n

| ≤ K

but then x

Let s 1 (n) be a converging sub-sequence for sequence x

taking s (ii) (n) and finding a sub-sequence in it such that x (i) (n) converges. Since x

is always bounded by K such sub-sequence must exists.

Then x s n (n) is converging sub-sequence of x n .

Definition 7. Closed set

A set F R m is closed if for all x n F , if x n x then x F .

Theorem 14. K is closed bounded set in R M iff every sequence in K has a converging subsequence with limit in K.

Proof. Suppose K is closed and bounded. Let a n K. a n is bounded thus by Boltzano-Weiestrass it has converging sub-sequence which itself consists of elements of K thus it converges to a limit in K. suppose every sequence in K has a converging sub-sequence with limit in K.

. We define s m inductively by

(i)

n

is bounded so the Bolzano-Weiestrass in one dimension applies to it.

(1)

n

(i)

n

in one dimension applies to it. (1) n ( i ) n It immediately follows that

It

immediately follows that every converging K-sequence converge to a limit in k since

in

converging sequence all sub-sequences converge to the same limit.

If

it is not bounded we can take sequence running to ±∞ and such sequence doesn’t

have converging sub-sequence.

and such sequence doesn’t have converging sub-sequence. Definition 8. Open set A set U x −

Definition 8. Open set

A set U

x y < ε we have y U .

Definition 9. Balls Consider sets in R m . Let x R m and r R +

Open ball set B(x, r) := {y R m : x y < r}

R m

is open if, whenever x U , there exists ε > 0 such that, whenever

Closed ball set B(x, r) := {y R m : x y r}

Theorem 15. Open ball is open, closed is closed

Consider sets in R m .

Then B(x, r) is open and B(x, r) is closed.

Proof. First we will show that B(x, r) is open. Suppose y B(x, r). Set ε := r x y . Now if for some u

Let x R m and r R + .

y u < ε

then

y u + x y < r

8

by triangle inequality

x u < r

so u B(x, r).

Now we will show that B(x, r) is closed.

Suppose that y n B(x, r) and y n y. Since for any ε

x y n + y n y < r + ε

taking y → ∞ we get

x y r

thus y B(x, r)

Theorem 16. Open set in terms of open balls (in reality—restatement of the definition). The set is open iff every point in that set is a center of an open ball.

Definition 10. Neighborhood The set N is a neighbourhood of the point x if we can find an r > 0 such that

of the point x if we can find an r > 0 such that B (

B(x, r) N

Thus the set is open if and only if it is a neighbourhood of every point it contains.

Theorem 17. A subset E of R m is open iff its complement R m \ E is closed

Proof. Suppose E is open. Let x n R m \ E. Suppose

If x E, it means that there is some ε such that B(x, ε) E. But also for this ε there

is N such that if x N x < ε, but then x N E

Suppose complement of E is closed.

Let x E. If ε > 0 such that B(x, ε) E then for any ε > 0 we can find elements as close as we like to x that are in complement of E and build the sequence out of them.

But complement is closed, so that would mean x E x

exists an open ball around each element of E E is open.

E, contradiction. So there

x n x.

x N

E. Contradiction.

So there x n → x . ∧ x N ∈ E . Contradiction. Theorem 18.

Theorem 18. Consider a collection τ of all open sets in R m Following holds

1.

2. If U αA τ then αA U α τ

, R m τ

3. If U α∈{1,

Proof.

,n}

τ then α{1,

,n}

U α τ

1. From the definition

2. If x αA U α τ then there exists such a that

x U a

and for some r, ball B(x, r) U a thus B(x, r) αA U α τ . The set is open.

3. If x α{1,

smallest one.

,n}

U α

τ

in every set there is a ball around that x, choose a

9

Theorem 19. Consider a collection F of all closed sets in R m Following holds

Theorem 19. Consider a collection F of all closed sets in R m Following holds

1.

2. If F αA ∈ F then αA F α ∈ F

, R m ∈ F

3. If F α∈{1,

,n}

∈ F then α{1,

,n}

F α ∈ F

Proof. Note. It would be also possible to look at complements and use previous theo- rems.

1. From the definition

2. Take x n αA F α , suppose x n x. From the definition of intersection x n F α for any α A, thus all sets contain x and so their intersection does.

F α ∈ F and x n x. If there doesn’t exists a set

F α such that all elements of x n from some points are in this set, then we can find two converging sub-sequence that converge to different limit points. But that contradicts the fact that sequence converges, thus there is such F α . Thus x is the element of this F α and so, the element of the sum.

3. Suppose x n α{1,

,n}

of the sum. 3. Suppose x n ⊆ α ∈ { 1 , ,n } Definition

Definition 11. Continuity in higher dimensions

Let E R m .

given ε > 0, we can find a δ > 0 such that, whenever y E and x y < δ

We say the function f : E R p is continuous at some point x E if,

f(x) f(y) < ε

Theorem 20. The definition of continuity is equivalent to the one given by converging sequences.

Theorem 21. The function f : R m R p is continuous iff f 1 (U ) is open whenever U is an open set in R p .

Proof. Suppose f is continuous

Let U

B(f (x), r) U . Since f is continuous we can find δ such that whenever y B(x, δ)

We know that there exists r such that

R p

be open set.

Let x

f 1 (U ).

y ∈ B ( x, δ ) We know that there exists r such that ⊆

10