Académique Documents
Professionnel Documents
Culture Documents
ca
Contents
Contents
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3
3
3
4
4
4
5
6
6
7
2 Limits
2.1 Lateral limits . . . . . . .
2.2 Properties of the limits . .
2.3 Continuity . . . . . . . . .
2.4 Infinites and infinitesimals
2.5 Asymptotic functions . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
9
11
12
13
14
16
.
.
.
.
.
.
.
.
18
20
21
25
26
29
30
30
31
.
.
.
.
.
.
.
.
34
34
35
36
37
38
39
41
41
44
44
45
47
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3 Derivatives
3.1 Properties of the derivatives . . . . . . . .
3.2 Taylors formula . . . . . . . . . . . . . .
3.3 Hopitals rule . . . . . . . . . . . . . . . .
3.4 Fundamental theorems . . . . . . . . . . .
3.5 Graphic study of functions . . . . . . . . .
3.5.1 Global theory . . . . . . . . . . . .
3.5.2 Local theory . . . . . . . . . . . .
3.5.3 Steps of a complete graphic study
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
4 Integrals
4.1 Definite integrals . . . . . . . . . . . . . . . . . .
4.1.1 Properties of definite integrals . . . . . .
4.2 Indefinite integrals . . . . . . . . . . . . . . . . .
4.2.1 Properties of the indefinite integrals . . .
4.3 The fundamental theorems of integral calculus .
4.4 Improper integrals . . . . . . . . . . . . . . . . .
4.5 Numerical integration . . . . . . . . . . . . . . .
4.5.1 Methods based on interpolation functions
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
5.4
5.5
terms
. . . .
. . . .
. . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
48
50
51
53
` lcul i 2016/2017
ca
Sets and functions
1.1
sets
Example. Let X = {1, 2, 3, 4} and Y = {3, 4, 5, 6}. Then, X Y = {3, 4}, X Y = {1, 2, 3, 4, 5, 6},
X \ Y = {1, 2}, and any of the two sets is the others subset.
Example. N Z i, de fet, N Z Q R C.
Example. The numbers of the set R \ Q which we have defined before as F are called irrational.
1.1.2
Intervals of R
A demonstration
Proposition.
2
/ Q.
Proof. We will
see
it
by
reductio
ad
absurdum.
We
suppose,
then,
that
2 Q. Then, by definition
of
Definition. A function
X and
Y is
a law
whichxassociates,
to each
f es fel between
conectortwo
quesets
asocia
a cada
dato
de input
un u
nico dato
de point x X, a
single point y = foutput
(x) fY(x).
. The
y is called
of x, and the
point
x inverseel image of y. The
Es point
la unicidad
de laimage
correspondencia
la que
caracteriza
concepto
funcio))
n. and
Parathe
tener
practico, of
podemos
pensar)).
en The subset of Y
set X is the domain
of f de
(Dom(f
set un
Y ejemplo
is the codomain
f (Codom(f
una caja
que calcule
rea de image
los crculos
partir
radioIm(f
r, area
formed by the images
of allnegra
the points
in X elisacalled
of f ,aand
is del
written
) or f (X).
que sabemos ser r2 . En este sistema cada dato de input es un valor r del
radio del crculo, al cual corresponde 4un u
nico dato de output r2 y la caja
r ! Area
! r2 .
Resulta muy u
til abstraer la definicion y no atarse a un ejemplo concreto y
tener una definicion general, que debemos a Johann Dirichlet, matematico
Remark. We will avoid the word range, which sometimes refers to codomain and other times refers the
image, and can cause confusion.
Notation. A function and the sets on which it acts are usually written as a diagram:
f :X Y
x 7 y
Remark. By definition, two points can have the same image but not the other way round.
Example. X = {1, 2, 3, 4}, Y = {1, 2, 3} and
f :{1, 2, 3, 4}
1
2
3
4
{1, 2, 3}
7
2
7
2
7
3
7
3
o be
f :R R
x 7 y = f (x) = x2
x si x 0
x si x < 0
Graph of a function
1.2.2
Properties
of the
real de
functions
of proyectado
real variablesobre el eje horizontal, corresponde a
Cada
punto
la curva,
un u
nico elemento x0 del dominio [a, b] de f (resaltado en azul). En cambio, si
proyectamos sobre el eje vertical, encontramos un elemento y0 del codominio
Notation. We will often talk about real functions of real variable f : D R R (and not f : R R)
[c, d] de f (resaltado en rojo).
when we are not interested in specifying the domain D of f but neither in forcing the function to be
Gr
aficamente, el dominio de una funci
on real de variable real es un subvalid for all R.
conjunto del eje horizontal, mientras que el codominio es un subconjunto
Definition.del
It is
that a function f : D R R is
ejesaid
vertical.
En la figura de derecha de la im
agen arriba se ve una curva que no
representa
gr
a
una a,
funci
n, porque, por ejemplo, la recta vertical
bounded
if Im(f ) el(a,
b)fico
for de
certain
b oR;
que corta el eje horizontal en x0 intersecta la curva en tres puntos diferentes
monotonic
increasing
if x sobre
< y elf (x)
f (y) forenevery
x, y Dy(strictly
if the is a <);
que se
proyectan
eje vertical
los puntos
1 , y2 , y3 . Si la curva
2 ; a tres im
representara
de
on,for
tendr
agenes ifdistintas,
(3 fel)(x)
( funci
7 g)(x)
=x0every
7x
monotonic
decreasing
ifgr
xa=fico
<3ysin(x),
una
f (x)
f (y)
x, y D (strictly
the is alo>);
cual es contrario
al
car
a
cter
de
unicidad
de
la
correspondencia
entre
input
f
(x)
=Rsin(x)
2 , con x 6= 0.
g
x
periodic
of
period
p
(or
p-peri`
o
dica)
if
f
(x
+
p)
=
f
(x)
for
every
x
D;
y output que caracteriza el concepto de funci
on.
Obviamente, el gr
afico de las funciones constantes es una recta
even if f (x)
= f (x)
for every x
1.5.
Composici
onD;altura
de funciones
horizontal
que tiene una
igual al valor de la constante. En
odd if particular,
f (x) =Pensemos
fel
(x)gr
for
every
x
D; funci
aotra
fico
de en
launa
funci
on onula
coincide
eje horizontal
vez
n como
en una con
cajaelnegra
que conectadado
por los
puntos
(x,
0)
del
plano
real
(a
cualquier
abscisa
corresponde
ordenada
un dato de input con un dato de output (y uno solo) de un cierto sistema.
injective if x 6= y f (x) 6= f (y) for every x, y D;
Formalicemos
la operaci
on def composici
n como
Definition. The composition
of two
functions
: X Y oand
g : Ysigue.
Z is the function
Def.: dadas dos funciones f1 y f2 ,
f1 : gX f :X
! Y Z
x
x 7 ! 7y z==f1(g
(x) f )(x) = g(f (x)).
f2 : W
w
! Z
6
7 ! z = f2 (w)
! Z
That is, you are supposed to chain the calculation of function f with that of function g:
f
g f :X Im(f ) Y Z
x 7 y = f (x) 7 z = g(y) = g(f (x))
Remark.
Dom(g f ) = Dom(f )
Codom(g f ) = Codom(g)
f g 6= g f in general
Example. Let f (x) = x2 and g(x) = sin(x). If we specify their domains and images
f :R [0, +)
x 7
x2
g :R [1, 1]
x 7 sin(x)
we see that both compositions f g and g f are possible, since both chains
g
f g : R [1, 1] R [0, +)
g f : R [0, +) R [1, 1]
g :(0, +) R
x
7 log(x)
f g : (0, +) R R (, 0]
since for the other we would have
f
R (, 0] 6 (0, +) R.
1.2.4
Inverse function
Definition. Let
f :X Y
x 7 y = f (x)
Remark. f 1 (x) 6=
1
f (x)
!!!
f :R R
x 7 y = f (x) = ex
is injective and therefore invertible. As Im(f ) = (0, +), its inverse is the function
f 1 :(0, +) R
y
7 x = f 1 (y) = log(y),
since (f 1 f )(x) = f 1 (f (x)) = f 1 (ex ) = log(ex ) = x, that is, f 1 f = idR , and, likewise, f f 1 =
1
id(0,+) . It is necessary to observe also that f 1 (x) = log(x) 6= ex = e1x = f (x)
.
` lcul i 2016/2017
ca
Limits
Notation. D domain
D = boundary of D. Then, D = D D and D = D {}, where x0 D
means that either x0 D (real points) or x0 = + o x0 = (behaviour of a limit).
xx0
if, for each > 0, exists > 0 such that |f (x) `| < for every x D which satisfies 0 < |x x0 | < .
Remark. The fact that for each > 0 exists > 0 means that depends on and, therefore, the
smallest , theCauchy
smallestha
willpopularizado
have to be taken
forathe
condition
to stillenbesusatisfied.
su teor
sobre
los lmites
Cours dAnalyse
Remark. [1] for every x D which satisfies [2] 0 < |x x0 | ... from the end means: [1] if x0 D, f
" ydefined
se motiva
con
hecho
que el sola abreviar la palabra error con "
need not be
to one of
theelsides
of xde
0 and, therefore, in this case the x D are only the ones on
y la palabra
conthe .value
Como
a therefore,
resultar claro
the other side;
and [2] x distancia
can not take
of deber
x0 and,
f mayde
notsubedefinici
definedon,
by "x0 .
=
0.005
and
then
it
is
true
that
24.9 <dex2 < 25.1
N, Z y Q, es fundamental para estar seguros de que el proceso
provided that
4.995
And so on: No matter how small is , ite will
exist a small
regresi
on <
dex"<y 5.005.
bienalways
definido.
" hacia cero (sin llegar nunca a cero) est
Example. Let us consider the case f (x) = x2 and x0 = 5, and see that
enough
which will satisfy inequalities (and, in fact, it can be proved that it is enough with taking any
on: si f es una funci
on constante k(x) k 8x 2 R, entonces
< 25 + Observaci
5).
f (x) toma siempre el mismo valor, por lo tanto es evidente que el lmite
para x ! x0 de una funci
on constante es igual al valor constante k para
cada x0 2 R. Entonces la operaci
onde Rlmite
aplicada a una funci
on
sgn :R
1 si x > 0
constante es trivial.
0 si x = 0
x 7 sgn(x) =
1 si x < 0
El lmite de una funci
on en un punto de la frontera
de su dominio puede
no ser finito, como vamos a ver en las
9 siguientes definiciones.
1, y se escribe:
l
m sgn(x).
f (x) =
@x!x
lim
x00
1,
If ` R existed such that limx0 sgn(x) = `, taking = 1 we would have to find a > 0 such that
si parathat
cada0 entorno
1), existe
unthat,
entorno
U m (x
D tal
0)
|sgn(x)`| < 1 provided
< |x0| U
<m.( However,
note
no matter
how
small
is,que,
the condition
para
cada
x
2
U
(x
),
x
=
6
x
,
entonces
f
(x)
2
U
(
1).
0
0
m
0 < |x| < always includes points x such
that sgn(x) = 1 and also other points x such that sgn(x) = 1.
m
All in all, ` should be a value such that | 1 `| < 1 and also |1 `| < 1, that is, such that 2 < ` < 0
interpretaci
n de
las definiciones 2. y 3. es la siguiente: a medida que
and also 0 <La` <
2, which is onot
possible.
Definition.valor
Let fabsoluto
: D R (positivos
R and xo0 negativos
D. It isseg
said
the limit
of f when
un that
el signo
del infinito).
Sexvetends
que to
el x0 is +
(or that f diverges
to
+
when
x
tends
to
x
),
and
is
written
0
infinito no es un numero real, sino,
de nuevo, el comportamiento o tendencia
xx0 m
if, for each M > 0, > 0 exists such that f (x) > M for every x D which satisfies 0 < |x x0 | < .
Gr
aficamente,
podemos
definici
de lmite
infinitoof como
In any of the two
cases,
it is said
that therepresentar
straight line la
x=
x0 is aonvertical
asymptote
f.
sigue:
1
Example. LikeSebefore,
f (x) =
(x3)
it can be checked that
2 andxx0
sueleifdecir
que
la recta
==x3,
ntota
0 (paralela al eje y) es una as
Definition. Let
f : Dsea
R
bounded
the=right,
and `a2R.R;Itseisdice
said que
that the limit
Def.:
f :DR,with
R !DR,not
con
D = Ron
oD
(a, +1),
of f when x tends to + is `, and is written
lim f45
(x) = `,
x+
if, for each > 0, K > 0 exists such that |f (x) `| < for every x D which satisfies x > K. In this
case, it is said that the straight line y = ` is a horizontal asymptote of f .
10
M"
"
M"
2.
x2
,
x2 +1
lm xf2(x) = +1 ,
=1
lim
x!+12
x+ x + 1
q UK (+1), existe un entorno UMK (+1) tal que,
si para cada entorno
1
from the definition,
taking
any
K
>
para cada x 2 UM (+1),
se tiene que f (x) 2 UK (+1);
.
K
2.1
lateral 3.
limits
lm f (x) =
x!+1
1,
1.
if, for each > 0, > 0 exists such that |f (x) `| < for every x D which satisfies x0 < x < x0 .
Similarly, it is said that the limit of f when x tends to x0 from the right is `, and it is written
lim f (x) = `,
xx+
0
if, for each > 0, > 0 exists such that |f (x) `| < for every x D which satisfies x0 < x < x0 + .
We talk about lateral limits to refer to limits from the left and from the right.
Example. We reintroduce the example of the symbol function, from which we have already seen that
@ limx0 sgn(x), to demonstrate now that
lim sgn(x) = 1
x0
46
i
lim sgn(x) = 1.
x0+
For the limit from the left, for each > 0 we need to find a > 0 such that |sgn(x) (1)| < provided
that 0 < x < 0, that is, as |sgn(x)+1| < provided that < x < 0. However, note that sgn(x) = 1
for every < x < 0, so the condition |sgn(x) + 1| < is simply 0 < , which always happens no matter
which we choose. The limit from the right is done similarly.
Example. Consider the case f (x) = sin x1 and x0 = 0, and see that
@ lim sin x1
i
@ lim sin x1 .
x0
x0+
For the limit from the left, if ` R existed such that limx0 sin x1 = `, taking = 1 we should be able
to find a > 0 such that |sin x1 `| < 1 provided that 0 < x < 0. However, note that, no matter
how small is , the condition
< x < 0 always includes points x such that sin x1 = 1 and also other
1
points x such that sin x = 1. All in all, ` should be a value such that | 1 `| < 1 and also |1 `| < 1,
that is, such that 2 < ` < 0 and also 0 < ` < 2, which is not possible.
11
Remark. Comparing the definitions of limit and lateral limits, it is deduced that
lim f (x) lim f (x) and lim f (x) and
xx0
xx
0
xx+
0
xx
0
xx+
0
In this case, the limit value coincides with the lateral limits common value.
Notation. Usually the expression limxx0 f (x) = ` is also written as
f (x) `,
xx0
Proposition. Let f, g : D R R and x0 D such that limxx0 f (x) and limxx0 g(x) exist and are
finite. Then,
limxx0 [f (x) g(x)] = limxx0 f (x) limxx0 g(x);
limxx0 [c f (x)] = c limxx0 f (x), for every c R;
limxx0 [f (x) g(x)] = limxx0 f (x) limxx0 g(x);
limxx0
f (x)
g(x)
limxx0 f (x)
limxx0 g(x) ,
`
0+
= + si ` > 0 (and all the other combinations according to the signs of ` 6= 0 and of 0 )
12
=0
The cases
+
0 ()
0
0
1
00
(+)0
are called indeterminacies and will be dealt with in the future.
2.3
continuity
(that is, if [1st] the limit limxx0 f (x) exists and [2nd] its value coincides with f (x0 )). Equivalently, is
we write x = x0 + h for a certain h R, f is continuous in x0 si
lim f (x0 + h) = f (x0 ).
h0
13
Theorem. All elementary functions of calculus (polynomials, powers, exponentials and trigonometric),
their inverses, their compositions and the algebraic operations between them are continuous in their
domains.
Remark. As, if f is continuous in x0 , then limxx0 f (x) = f (x0 ), the calculation of the non trivial limits
of a function reduces to the points x0 of the boundary in its domain and the indeterminacies.
Theorem. Let f : Df R R, g : Dg R R continuous and x0 Df such that g f is well defined
near x0 . Then,
lim g(f (x)) = g lim f (x) .
xx0
xx0
lim ef (x)
xx0
2.4
si limxx0 f (x) =
0
k
e
si limxx0 f (x) = k R
=
+ si limxx0 f (x) = +.
xx0
f (x)
g(x)
... does not exist, it is said that f and g are infinitesimals not comparable in x0 .
... = 0, it is said that f is an infinitesimal of superior order to g in x0 .
... = k R, it is said that f and g are infinitesimals of the same order in x0 .
... = , it is said that f is an infinitesimal of inferior order to g in x0 .
Notation. For its importance, as we will see in the future, the fact that f is an infinitesimal of superior
order to g in x0 is usually written as
f (x) = o(g(x))
and it is said that f is a or small of g when x x0 .
14
Example. The function f (x) = 5x3 is an infinitesimal of superior order to the function g(x) = 2000x2
en x0 = 0 (or 5x3 = o(2000x2 ) when x 0) because
x
5x3
f (x)
= lim
= lim
= 0.
2
x0 400
x0 2000x
x0 g(x)
Example. We reintroduce the example of the casef (x) = x2 sin x1 , which is an infinitesimal in x0 = 0
because, as we have already seen, limx0 x2 sin x1 = 0. On the other hand, it is clear that g(x) = x2
is also an infinitesimal in x0 = 0, because limx0 x2 = 0. If we compare them,
x2 sin x1
f (x)
lim
= lim
= lim sin x1 ,
2
x0 g(x)
x0
x0
x
lim
which does not exist. Therefore, f and g are infinitesimals not comparable in x0 = 0.
loga (x)
loga (x)
x
x
bx
bx
xx
xx
quan x +, where f (x) g(x) means that f is an infinitesimal of inferior order to g when x +.
Example. We want to calculate
lim (log 2)x x40 ,
x+
which, as 0 < log 2 < 1, a priori presents the form of indetermination 0 (+). We can rewrite this limit
as
(log 2)x
lim (log 2)x x40 = lim
=0
1
x+
because
1 3
bx
x+
1 2
x40
quan x + (b =
1
log 2 ,
3 = 1 and 2 = 40).
xx0
f (x)
g(x)
... does not exist, it is said that f and g are infinite not comparable in x0 .
... = 0, it is said that f is an infinite of inferior order to g in x0 .
... = k R, it is said that f and g are infinites of the same order in x0 .
... = , it is said that f is an infinite of superior order to g in x0 .
Example. The function f (x) = 5x3 is an infinite of superior order to the function g(x) = 2000x2 in
x0 = + because
5x3
x
f (x)
= lim
= lim
= +.
lim
x+ g(x)
x+ 2000x2
x+ 400
Example. It can be checked from the definitions of infinite and divergent function that the functions
f (x) = 2x + x sin(x) and g(x) = x are infinites in x0 = +. If we compare them,
lim
x+
f (x)
2x + x sin(x)
= lim
= lim (2 + sin(x)),
x+
g(x) x+
x
which does not exist. Therefore, f and g are infinites not comparable in x0 = +.
15
x+
x+
x40
1
log 2
x = 0
1
log 2 ,
3 = 1 and 2 = 40).
asymptotic functions
Definition. Let f, g : D R R and x0 D . It is said that f and g are asymptotic when x tends to
x0 , and it is written
f (x) g(x),
xx0
if
(x)
limxx0 fg(x)
= 1.
The idea is that both functions behave the same way when x tends to x0 , that is, either they diverge
at the same speed, they converge at the same limit, or the limit does not exist for neither of them. In
particular, if f (x) xx0 g(x), obviously
lim f (x) = lim g(x),
xx0
xx0
but in general the reciprocal is not true if we deal with infinitesimals or infinites.
Example. If f (x) = 2x2 5x3 and g(x) = 2x2 , then f (x) x0 g(x) because
2x2 5x3
2 5x
f (x)
= lim
= lim
= 1.
2
x0 g(x)
x0
x0
2x
2
lim
sin(x)
x
1
x
f (x)
g(x)
xx0
f(x)
.
g(x)
Example. If f (x) = 3x2 , f(x) = 3x2 + x, g(x) = 3x2 and g(x) = 3x2 + x, we have f (x) x+ f(x)
and g(x) x+ g(x) because
3x2
f (x)
3
= lim
= lim
x+ 3x2 + x
x+ f(x)
x+ 3 +
lim
1
x
=1
and similarly for g and g. It can be checked that f (x) g(x) x+ f(x) g(x) and
On the other hand,
f (x)
g(x)
x+
f(x)
g(x) .
f (x) + g(x)
3x2 3x2
0
= lim
= lim
= lim 0 = 0 6= 1,
2
2
x+ f (x) + g
(x) x+ 3x + x 3x + x x+ 2x x+
lim
an xn
an xn + + a1 x + a0
bm xm + + b1 x + b0
17
an nm
x
.
bm
` lcul i 2016/2017
ca
Derivatives
Captulo 3
Derivadas
lim
xx0
f (x) f (x0 )
<
x x0
(x0 )
(that is, if [1st] the limit limxx0 f (x)f
exists and [2nd] it is finite). Equivalently, if we write x = x0 +h
xx0
for a certain h R, f is differentiable in x0 if
f (x0 + h) f (x0 )
lim
.
Definici
on de
de<una
funci
on y su
h0 derivada
h
interpretaci
n value
geom
trica
mec
anica
In this case, the derivative
of f in x0 is othe
of e
this
finiteylimit,
and
is written
3.1.
0
Como ya dijimos, el concepto def derivada
est
a relacionado con el proble(x0 ).
ma de encontrar la recta tangente al gr
afico de una funci
on o, de forma m
as
a una
en un cierto
punto.
Para entender
mejor
problema,
Globally, it isgeneral,
said that
f is curva
differentiable
if it is
differentiable
in every
pointeste
of D.
consideremos la figura aqu debajo.
Remark. By construction,
f 0 (x
) is thesobre
slopeelofgr
the
tangent
straight
line de
at the
graphreal
of f in the point
Llamamos A
el 0punto
afico
de la funci
on real
variable
on the plane y(x=
)).
Therefore,
knowing
its
slope
and
a
point
through
which
it
passes,
0 , ff(x
0
(x) de coordenadas (x0 , f (x0 )) y B el punto que tiene coordenadas the equation
of the tangent
straight
line+at
the
graphhof
(x0 , f (x0 )) ishorizontal y f (x + h) f (x )
(x + h, f (x
h)),
siendo
> f0 in
el incremento
0
Example. Consider the case f (x) = 2x3 and x0 = 1, and see that
f 0 (1) = 6
that is, the limit exists and it is finite, and its value is 6 = f 0 (1). With this, we can also calculate the
tangent straight line at f in x0 = 1:
y = f (1) + f 0 (1)(x (1)) = 2 + 6 (x + 1) = 6x + 4
Finally, with a similar calculation to that of f 0 (1), we see that, in fact, f is differentiable in every point
x R:
f (x + h) f (x)
2(x + h)3 2x3
lim
= lim
= lim (6x2 + 6xh + 2h2 ) = 6x2 = f 0 (x)
h0
h0
h0
h
h
because it exists and it is finite for every x R.
Example. Consider the case f (x) = |x| and see that f is not differentiable in x0 = 0.
f (0 + h) f (0)
|h|
= lim
= lim sgn(h),
h0 h
h0
h
which we have already seen that does not exist.
Example. Consider the case f (x) = 3 x and see that f is not differentiable in x0 = 0.
3
f (0 + h) f (0)
h
1
lim
= lim
= lim 2/3 = +,
h0
h0
h0
h
h
h
that is, the limit exists but it is not finite.
lim
h0
Proceeding like in the first example, from the definition the derivatives of all elemental functions can be
calculated:
f (x) = k R
f 0 (x) = 0
f (x) = x
f (x) = ex
f (x) = ax
f (x) = log(x)
f (x) = loga (x)
f (x) = sin(x)
f (x) = cos(x)
f (x) = tan(x)
f (x) = arcsin(x)
f (x) = arccos(x)
f 0 (x) =
f 0 (x) = sin(x)
f 0 (x) =
f 0 (x) =
f (x) = cosh(x)
1
x
1
x log(a) ,
f 0 (x) = cos(x)
f 0 (x) =
1
cos2 (x)
1
1x2
1
1x
2
1
1+x2
f 0 (x) =
f 0 (x) = cosh(x)
f (x) = arctan(x)
f (x) = sinh(x)
f 0 (x) = ex
19
f 0 (x) =
f 0 (x) = sinh(x)
Example. We want to calculate the derivative of f (x) = x2 sin(x). This function is the product of
two elemental functions, x2 and sin(x), the derivatives of which we can find on the previous list. Using
0
Leibniz rule, f 0 (x) = x2 sin(x) + x2 (sin(x))0 = 2x sin(x) + x2 cos(x).
Theorem (Rule of the chain). Let f : Df R R differentiable in x0 Df and g : Dg R R
differentiable in f (x0 ) Dg . Then, g f is differentiable in x0 and
(g f )0 (x0 ) = g 0 (f (x0 )) f 0 (x0 ).
Remark. If instead of two functions we have three, applying the rule of the chain twice we obtain
(h g f )0 (x0 ) = h0 ((g f )(x0 )) g 0 (f (x0 )) f 0 (x0 ), and so on
Example. Let F (x) = sin2 (x); we want to calculate F 0 4 . Note that the function F calculates first
the sine of x and then squares this result, that is, F = g f , with f (x) = sin(x) and g(x) = x2 . As these
two functions are differentiable on all R, and f 0 (x) = cos(x) and g 0 (x) = 2x,
1
0
0
0
0
0
0
0
0
= (g f )
=g f
f
= g sin
f
=g
=
F
f
4
4
4
4
4
4
4
2
1
2
1
= 2 cos
= = 1.
4
2
2
2
Example. We want to calculate the derivative of F (x) = log |x|. Like before, F = g f , with f (x) = |x|
and g(x) = log(x). We have already seen that the function f is not differentiable in x0 = 0, but anyways
x0 = 0 must stay out of the domain of F because otherwise it would cancel the logarithm. The diagram
we have is
f
g
F = g f :R \ {0} (0, +) R
x
7
|x|
7 log |x|
and, defined this way, the functions f and g are differentiable. However,
x si x > 0
1 si x > 0
0
f (x) = |x| =
f (x) =
x si x < 0
1 si x < 0
20
g 0 (x) f 0 (x)
g 0 (x)
f 0 (x)
si x > 0
si x < 0
1
x
1
x
si x > 0
(1) si x < 0
1
.
x
It could seem that the derivatives of log(x) and log |x| coincide, but the domain of log(x) and its derivative
is (0, +) while that of log |x| and its derivative is R \ {0}, therefore they are different functions.
Theorem. Let f : D R R differentiable in x0 D and such that f 0 (x0 ) 6= 0. Then, f 1 is
differentiable in y0 = f (x0 ) and
0
1
f 1 (y0 ) = 0
.
f (x0 )
Example. Let f (x) = sin(x) = y and, therefore, f 1 (y) = arcsin(y) = x. For the previous theorem,
0
f 1 (y) =
1
f 0 (x)
1
1 (1)
1
p
=p
=
,
cos(x)
1 y2
1 sin2 (x)
3.2
taylors formula
Como muestra la figura superior, estos detalles finos permiten aproximar cada vez mejor la funcion tambien lejos del punto central del
desarrollo de Taylor. Cerca del centro del desarrollo, 0 en este caso,
todos los polinomios dan una21aproximaci
on excelente de la funci
on
y = sin(x) (derivable infinitas veces con continuidad), a medida que
nos alejamos de 0, los polinomios de orden m
as grande aproximan la
funcion con un error cada vez m
as peque
no.
3. El desarrollo de Taylor se usa de forma extensa tambien en los computadores electronicos. Un ordenador sabe hacer solo operaciones al-
f 00 (x0 )
f 000 (x0 )
f (n) (x0 )
(x x0 )2 +
(x x0 )3 + +
(x x0 )n .
2!
3!
n!
f (n) (0) n
f 00 (0) 2 f 000 (0) 3
x +
x + +
x ,
2!
3!
n!
X f (k) (x0 )
0kn
k!
(x x0 )k
(n)
T0 (x) =
X f (k) (0)
0kn
f (n) (x0 )
(x x0 )n for every n 1.
n!
22
k!
xk .
= cos(x) x sin(x)
f (0) = 0
f 0 (0) = 1
f 00 (0) = 0
f 000 (0) = 3
Llavors,
(0)
T0 (x) = f (0) = 0
(1)
(0)
(2)
(1)
f 00 (0)
2!
x2 = x +
0
2!
(3)
(2)
f 000 (0)
3!
x3 = x +
3
3!
x2 = x
x3 = x
x3
2 .
Definition. For each n N, the remainder or the subtraction of Taylors polynomial of order n is the
function
Rn (x) = f (x) Tx(n)
(x).
0
Remark. We started this section saying that we would like to obtain better approximations than a
tangent straight line, and then we have defined Taylors polynomials. If these are to be the approximations
that we are looking for, the remainder of Taylors polynomial should be reasonably small.
Theorem (Taylors theorem). Let f : D R R differentiable n N times in x0 D. Then,
Rn (x) = o((x x0 )n ).
Remark. This expression is what is called Peanos form of the remainder. Other ways of expressing
Rn (x) exist, but here we will only see another one Lagranges form:
Rn (x) =
f (n+1) ()
(x x0 )n+1 ,
(n + 1)!
where hx0 , xi, that is that is between x0 and x but we do not know if [x0 , x] or [x, x0 ].
Lagranges form of the remainder allows us to give a bound of the error that we make when using Tx(n)
(x)
0
as an approximation of f (x), since
f (n+1) ()
f (n+1) ()
(x) = |Rn (x)| =
(x x0 )n+1 =
f (x) Tx(n)
|x x0 |n+1 M |x x0 |n+1 ,
0
(n + 1)!
(n + 1)!
on M = maxhx0 ,xi
|f (n+1) ()|
(n+1)!
(n)
Remark. Using the definitions of Rn (x) and Tx0 (x), we can rewrite Taylors theorem as
f (x) =
X f (k) (x0 )
0kn
k!
(x x0 )k + o((x x0 )n ),
that is, for each n N, f it can be approximated by a polynomial of degree n and the remainder of this
approximation decreases faster than (x x0 )n quan x x0 .
23
Example. We reintroduce the example of the function f (x) = x cos(x), from which we have already
3
3
(3)
(3)
seen that T0 (x) = x x2 . From here we can calculate R3 (x) = f (x) T0 (x) = x cos(x) x + x2 and,
according to Taylors theorem, conclude that
x3
+ o(x3 ).
2
x cos(x) = x
(3)
Imagine now that we use T0 (x) to calculate an approximation of f (1), that is, we consider f (1)
3
(3)
T0 (1) = 1 12 = 0.5. A bound of the error we make is
(4)
f ()
|4 sin() + cos()| (2) |4 sin(1) + 1 cos(1)|
(1)
(3)
=
|1|4 = max
0.16.
f (1) T0 (1) max
4!
24
24
[0,1]
[0,1]
A (1), we have simply calculated f (4) (x) = 4 sin(x) + x cos(x); for (2), we have done a little study of the
function to maximize and we have seen that it is strictly increasing monotonic at the interval [0, 1], that
is that it takes its maximum value at = 1. With this bound we can say that T0(3) (1) 0.16 f (1)
T0(3) (1) + 0.16 0.5 0.16 f (1) 0.5 + 0.16 0.34 f (1) 0.66. Its real value is f (1) 0.54,
which is quite near T0(3) (1) = 0.5.
The Taylor-Maclaurins polynomials of some elemental functions are specially important, because we will
use them in many different fields.
ex =
xk
0kn k!
k+1 xk
1kn (1)
k
log(1 + x) =
(1 + x) =
sin(x) =
cos(x) =
+ o(xn ) = 1 + x +
0kn k
sinh(x) =
cosh(x) =
k x2k
0kn (1) (2k)!
x2k+1
0kn (2k+1)!
+ +
x2k
0kn (2k)!
xn
n!
+ o(xn )
x2
2
+ o(xn ) = x
xk + o(xn ) = 1 + x +
k x2k+1
0kn (1) (2k+1)!
x2
2!
(1)
2!
+ o(x2n+1 ) = x
x3
3!
x3
3
xn
n
+ + (1)n+1
x2 + +
x5
5!
+ o(xn )
xn + o(xn ), where R
2n+1
x
+ + (1)n (2n+1)!
+ o(x2n+1 )
+ o(x2n ) = 1
x2
2!
x4
4!
x
+ + (1)n (2n)!
+ o(x2n )
+ o(x2n+1 ) = x +
x3
3!
x5
5!
+ +
+ o(x2n ) = 1 +
x2
2!
x4
4!
2n
+ +
x2n
(2n)!
x2n+1
(2n+1)!
+ o(x2n+1 )
+ o(x2n )
( 1) ( k + 1)
.
=
k!
k
(n)
Remark. The fact that f (x) = Tx0 (x) + o((x x0 )n ) means that
lim
xx0
f (x)
(n)
Tx0 (x)
(n)
= lim
xx0
Tx0 (x)
= lim
xx0
1+
o((x x0 )n )
(n)
Tx0 (x)
=1+0=1
(n)
by definition of o((x x0 )n ) and of Tx0 (x). All in all, this calculation says that
f (x) Tx(n)
(x)
0
xx0
for any n N very important result that we will use mainly when calculating limits.
24
x2 ex
,
x0 cos(x) 1
lim
which a priori presents the form of indetermination 00 . Using that ex x0 1 + x and that cos(x) x0
2
1 x2 , tenim
x2
x2 e x
x2 (1 + x)
x2 + x3 (1)
=
lim
= lim
=
lim
2
2 = 2.
2
x
x0 x
x0 cos(x) 1
x0
1 x 1 x0
lim
(A (1) we have applied that the addition of infinitesimals is asymptotic to the infinitesimal of inferior
order.)
Remark. Regarding limits, we can replace the expression x x0 with (t) tt0 x0 and then
f ((t)) Tx(n)
((t)).
0
tt0
x0
log(cos(x))
,
x2
x2
2 ,
firstly we have
2
x2 x8
x2
2
log(cos(x))
1
2
lim
= lim
= lim
= lim
= lim 2 = .
2
2
2
2
x0
x0
x0
x0
x0 x
x
x
x
x
2
3.3
pitals rule
ho
Theorem (LHopitals rule). Let f, g : D R R differentiable except maybe in x0 D and such that
(x)
f 0 (x)
limxx0 fg(x)
presents an indetermination of the type 00 or
. If it exists the limit limxx0 g 0 (x) and
g 0 (x) 6= 0 except maybe in x = x0 , then
lim
xx0
f (x)
f 0 (x)
= lim 0
.
g(x) xx0 g (x)
Example. We want to calculate limx0+ x log(x), which a priori presents the form of indetermination
0 (). (We could put limx0 x log(x) but, as the function x log(x) is defined at (0, +), by definition
limx0 x log(x) = limx0+ x log(x).) We can rewrite this limit as
lim x log(x) = lim
x0+
+ .
x0+
log(x)
1
x
f (x)
g(x)
+ .
However, f 0 (x) =
1
x
(so the limit exists), and g 0 (x) = x12 6= 0 for every x (0, +).
limx0+ x log(x) =
0 (x)
limx0 fg0 (x)
= 0.
(x)
Remark. Be careful with overlooking the hypothesis of existence of the limit limxx0 fg0 (x)
. In case it
f (x)
does not exist, the conclusion is not that limxx0 g(x) does not exist either, but that this result simply
cannot be used to calculate it.
x+
x sin(x)
,
x + sin(x)
This does not mean that the original limit does not exist, but that it must be calculated using some other
3
tool. (Note that, in this case, we cannot use that sin(x) x0 x x6 neither because the of the limit
problem is when x +.) Indeed, if we manipulate the expression a little,
1
x sin(x)
= lim
x+ x + sin(x)
x+ 1 +
lim
sin(x)
x
sin(x)
x
sin(x)
x
10
= 1,
1+0
= 0.
(x)
3.4
fundamental theorems
-2
-1
We see firstly that x = 0 is a local maximum of f , that is, > 0 exists such that (x2 1)2 (02 1)2 = 1
for every x R which satisfies 0 < x < 0 + , so, < x < . Let = 1; then, 1 < x < 1
0 x2 < 1 1 x2 1 < 0 0 < (x2 1)2 1, which is what we wanted. Similarly we can check
that x = 1 and x = 1 are local minimums of f . We see now that x = 1 and x = 1 are, furthermore,
global minimums of f . By definition, if f has two global minimums, its value must be the same. Indeed,
f (1) = f (1) = 0. Therefore, we need to see that (x2 1)2 ((1)2 1)2 = (12 1)2 = 0 for every x R,
which is obvious. Finally, f does not have any global maximum because it is not bounded superiorly,
that is, (x2 1)2 takes values as large as we want.
Remark. A function can have none, one or more than one local end of the same type and, if we do not
put any restriction on f or on D, it also fits, one or more than one global ends of the same type.
Theorem (Weierstra (Weierstrass)s theorem). Every continuous function f : [a, b] R has at least
one global minimum and one global maximum.
Theorem (Teorema del valor intermedi). Let f : [a, b] R continuous and u hf (a), f (b)i. Then,
c [a, b] exists such that f (c) = u.
Example. Consider
f :[1, 2]
R
2
x 7 (x 1)2
which has the same analytic expression as that of the previous example. It can be checked that x = 1
and x = 1 are global minimums of f , and that x = 2 is a global maximum of f . Furthermore, for each
u [f (1), f (2)] = [0, 9], c [1, 2] exists such that f (c) = u. In fact, in the drawing we can see clearly
that, for each u (0, 1), three points c [1, 2] exist with this property.
Definition. A stationary point of a function is a point in which the derivative of a function cancels.
Theorem (Fermats theorem on stationary points). Let f : D R R and x0 D \ D an end of f
such that f is differentiable in x0 . Then, x0 is a stationary point of f , that is, f 0 (x0 ) = 0.
2
Example. We reintroduce the example of the function f (x) = x2 1 , from which we have already
seen that it has ends in x = 1, x = 0 and x = 1, and we check that they are also stationary points of f .
We calculate f 0 (x) = 4x(x2 1) and, indeed, if we replace the points we have f (1) = f (0) = f (1) = 0.
27
Remark. The reciprocal of Fermats theorem is not true generally, that is, not every stationary point of
f is necessarily an end of f .
Example. The function f (x) = x3 ex has as a graph
3.11.
Supongamos haber recorrido en coche los 160 km que separan dos puntos
A y B en 2 horas. -6La velocidad
media
con la
cual hemos
viajado
es 80 km/h.
-5
-4
-3
-2
-1 0
1
2
3
Si hemos viajado de forma continua, sin parar o tener otras irregularidades
-1
durante el trayecto, entonces en algun instante t0 2 (0, 2horas) la velocidad
con la cual hemos viajado ha coincidico con la-2 velocidad media de 80 km/h,
o sea:
velocidad instantanea(t0 ) = velocidad media
f 0 (x)
Example. We reintroduce the example of the function f (x) = x3 ex , from which we have already seen
that f 0 (x) = x2 (3 + x)ex . As x2 0 and ex 0 for every x R, the sign of f 0 sign only depends on the
term 3 + x. Therefore, f is decreasing monotonic at (, 3) and increasing monotonic at (3, +),
and x = 3 is a minimum of f .
28
5 (, 3 3 )
f 00 (5) 0.03 < 0 f is concave in (, 3 3 )
2 (3 3 , 3 + 3 )
f 00 (2) 0.54 > 0 f is convex in (3 3 , 3 + 3 )
1 (3 + 3 , 0)
f 00 (1) 0.37 < 0 f is concave in (3 + 3 , 0)
1 (0, +)
The following two sections are a summary of all these concepts (definitions) and tools to study them
(results), plus some observations to make a complete graphic study.
29
3.5.1
Global theory
conceptes
eines
f0
f 00
mc/md
cv/cc
f mc f 0 0
f md f 0 0
f cv f 00 0
f cc f 00 0
Local theory
An end x0 can be: a x0 D, or a point where f is non differentiable, or a x0 D \ D. The first two
types are easy to find and classify; for the last we have the following.
f0
f 00
xt
pi
pith
conceptes
(pe)
eines
x0 xt f 0 (x0 ) = 0
x0 pi f 00 (x0 ) = 0
where xt = end (local), pi = point of inflection, pith = pi with horizontal tangent and pe = stationary
point.
Using the tools of global theory, we can describe exactly the behaviour of a function around a point x0
according to the signs of f 0 and f 00 . (At the table headings, x0 + means that the corresponding function
takes negative values at the left of x0 and positive at its right, and similarly for the other abbreviations,
and afterwards f 0 6= 0 means that f 0 (x0 ) 6= 0, and similarly for the others.)
30
f 00
f0
x0
x0 +
cap
x0
+ x0
+ x0 +
`x
ma
no
f 0 6= 0 and f 00 ?
cap
f 0 = 0 and f 00 ?
f 0 6= 0 and f 00 ?
pi
x0 +
f 0 6=
0 and
f 00
pi
no
=0
f 0 6=
0 and
f 00
f0 =
0 and f 00 = 0
=0
0 and
f 00
pi
no
no
f 0 6= 0 and f 00 ?
f 0 6=
0 and f 00 = 0
=0
cap
+ x0 +
f 00
pith
f0 =
=0
f 0 6=
0 and
pi
+ x0
no
pith
mn
f 0 = 0 and f 00 ?
no
cap
f 0 6= 0 and f 00 ?
1. Calcular D = Dom(f ).
2. Calculate the cut-off points with the axes: the points x D such that f (x) = 0 for the axis of the
x, and the point y = f (0) (if 0 D) for the axis of the y.
3. Calculate the limits limxx f (x) for the x0 D and limx f (x). In particular, we will see if
0
f has asymptotes and we can give its expression.
4. Calculate the stationary points (f 0 (x0 ) = 0) and the sign of f 0 in each interval that these limit, the
points of D and the points where f is non differentiable (to determine increase/decrease).
5. Calculate the points f 00 (x0 ) = 0 and the sign of f 00 sign in each interval that these limit, the points
of D points and the points where f is non differentiable (to determine convexity/concavity).
6. Determine if the points of D and the points where f is non differentiable are the ends of f .
Determine if the points f 0 (x0 ) = 0 are ends and if the points f 00 (x0 ) = 0 are points of inflection,
doing:
31
0
x0 mn
f is x0 +
0
0
0
x0 max
Lemma (T1E: f test for ends). f (x0 ) = 0 and f is + x0
0
f does not change x0 pith
Alternative to T1E:
00
f (x0 ) > 0 x0 mn
00
f is x0 x0 max
f 00 is + x0 + x0 min
f 00 (x0 ) = 0 and
00
f changes
x0 pith
Finally:
f 0 (x0 )
6= 0,
f 00 (x
0 ) = 0 and
f 00 changes
f 00
x0 pi
We can evaluate f in all these points if we want to make a more accurate sketch of the graph f .
Example. Complete graphic study of the function
f (x) =
x2
x
.
1
x
(x)2 1
1. Domain: f only has problems when its domain is cancelled, that is, when x2 1 = 0 x = 1.
Therefore, D = Dom(f ) = R \ {1, 1}.
2. Cut-off points: For the axis of x, we do f (x) = 0 x2x1 = 0 x = 0
(0, 0). For the axis of y,
we do f (0) = 0
(0, 0). Therefore, (0, 0) is the only cut-off point with the axes.
3. Limits: limx1 f (x) = limx1 x2x1 = because x2 1 < 0 when x 1 ; limx1+ f (x) =
limx1+ x2x1 = + because x2 1 > 0 when x 1+ ; and limx+ f (x) = limx+ x2x1 =
limx+ xx2 = limx+ x1 = 0 because x2 1 x+ x2 . As the graph of f is asymmetric,
limx1 f (x) = , limx1+ f (x) = + and limx f (x) = 0. All in all, the straight lines
x = 1 and x = 1 are vertical asymptotes of f , and the straight line y = 0 is a horizontal asymptote
of f .
2
x 1
x +1
+1
2
4. Study of f 0 : f 0 (x) = 1(x(x1)x2x
= (x
f 0 (x) = 0 (xx2 1)
2 1)2
2 1)2 = (x2 1)2
2 = 0 x + 1 = 0,
which does not have real solutions. This means that it will only be necessary to study the sign of
f 0 at the intervals which limit the points of D and the points where f is not differentiable. On
the one hand, D = {1, 1}; on the other hand, and reasoning as for f , Dom(f 0 ) = R \ {1, 1},
therefore the only points where f is not differentiable is x = 1. Therefore,
sgn(f 0 )
(, 1)
(1, 1)
(1, +)
f 0 (2) < 0
f 0 (0) < 0
f 0 (2) < 0
decreasing
decreasing
decreasing
32
+1)2(x 1)2x
+3)
+3)
5. Study of f 00 : f 00 (x) = (f 0 )0 (x) = 2x(x 1) (x
= 2x(x
f 00 (x) = 0 2x(x
=
(x2 1)4
(x2 1)3
(x2 1)3
2
0 2x(x + 3) = 0, which only has the real solution x = 0. Therefore, we need to study the sign
of f 00 at the intervals which limit the points {1, 0, 1}:
(, 1)
sgn(f 00 )
(1, 0)
(0, 1)
(1, +)
f 00 (2) < 0
f 00 ( 12 ) > 0
f 00 ( 12 ) < 0
f 00 (2) > 0
concave
convex
concave
convex
6. Classification: As limx1 f (x) = , the points x = 1 cannot be the ends of f . On the other
hand, we have just seen that f 00 changes sign to x = 0, therefore, for the TPI, the point x = 0 is a
point of inflection of f . As we have already seen, its value is f (0) = 0.
With all this information, we can already represent the graph of f :
-5
-4
-3
-2
-1
-1
-2
-3
q
3
(x3 3x)2 .
33
` lcul i 2016/2017
ca
Integrals
4.1
definite integrals
Calculating the area of a rectangle, a triangle or even any polygon is a task that can be more or less
complicated, depending on the polygon, but it is feasible if we divide it adequately in simpler polygons.
This is not possible when we deal with curvilinear figures, and therefore we need to proceed differently.
Definition. A partition of an interval [a, b] is a finite succession of points
tambien los extremos), y evaluamos la funci
on f en j , como en la figura
P = siguiente.
(x0 , x1 , x.2 , . . . , xn1 , xn )
such that a = x0 < x1 < x2 < < xn1 < xn = b. It is said that the partition is regular if the distance
between consecutive points is always the same, and this common value h = x1 x0 = x2 x1 = =
xn xn1 is the regular partition step.
Remark. A usual way to build a regular partition of an interval [a, b] is dividing it in n subintervals,
such that h = ba
n and xk = a + k h for every 0 k n. (We observe that this gives n + 1 points, not
n.)
Example. We consider the interval [0, 1]. Taking n = 10, we obtain the partition of 11 points (and 10
subintervals) P = (0.0, 0.1, 0.2, . . . , 0.9, 1.0), since h = 10
10 = 0.1 and xk = 0 + k 0.1 for every 0 k 10.
Definition. Let f : [a, b] R a function and P = (x0 , x1 , x2 , . . . , xn1 , xn ) a partition of [a, b]. A
Riemann sum of f regarding P is
n De esta manera, podemos definir la suma de Cauchy-Riemann como:
X
f (k )(xk xk1 ),
k=1
Sn =
n
X
f (j )(xj
xj
1)
n
X
f (j ) x.
j=1
where k [xk1 , xk ], for every 1 k n, is any point of the subinterval.
(Usuallyj=1k is the mean point
x
+x
of the
subinterval,
= k12 lak ,funci
butonot
tambi
en los
extremos), ykevaluamos
n f always.)
en j , como en la figura
Si operamos una partici
on m
as fina de [a, b], por ejemplo esta
siguiente. .
posici
on with
de como:
losstep
j variar
a, y con
el valor de
calculada
en esos
De esta manera,
podemosthat,
definiriflathe
suma
de Cauchy-Riemann
Remark.
We observe
partition
isla regular
h, then
the ella
expression
of fthe
Riemann
puntos
y
la
distancia
x
entre
los
extremos
de
los
sub-intervalos.
El
siguiente
sum is reduced toX
n
n
X
Sn =
f (
n j )(xj
j=1
xj
f (k ) h
1)
j=1
or
teorema, seg
un Riemann, dice que, si f es continua, existe el lmite de n
f (j ) x.n
n
X
X
f (k )
k=1
k=1 esta
Si operamos una partici
on mas fina de [a, b], por ejemplo
34
or
ba
f (k ).
n 117
k=1
Example. We reintroduce the previous example to calculate the Riemann sum of the function
f :[0, 1] R
x 7 x2
x
+xk
n
X
k=1
f (k ) = 0.1
n
X
k=1
Definition. It is said that f : [a, b] R is integrable (Riemann) if the limit of Riemann sums
!
n
ba X
lim
f (k )
n+
n
k=1
exists and is finite. (We can see that this limit does not depend on the choice of the k .) In this case,
the integral of f between a and b is this finite limit value, and is written
Z b
f (x) dx.
a
Notation. The integral symbol with ends a and b will be called definite integral.
Theorem. If f : [a, b] R is continuous, then f is integrable.
Example. We reintroduce the previous example to calculate now the integral of f (x) = x2 between 0
and 1 from the definition (as f is continuous, from the previous theorem we know that it is integrable).
xk1 +xk
1
1
k
10
= 2k1
To calculate the limit, for each n we have
n and k =
2
2n .
Pkn n =
Pn h = n = n , xk = 0 +
2
Furthermore, we will use the formulas k=1 k = n(n + 1)/2 and k=1 k = n(n + 1)(2n + 1)/6 (in the
next topic we will see how to demonstrate this type of identities). Therefore,
!
!
!
Z 1
n
n
n
X
1 X 2k 1 2
1 X (2k 1)2
2
x dx = lim
h
f (k ) = lim
= lim
=
n+
n+ n
n+ n
2n
(2n)2
0
k=1
k=1
k=1
!#
!
"
n
n
n
n
X
X
X
X
1
1
k2 4
k+
1 =
(4k 2 4k + 1) = lim
= lim
4
n+ 4n3
n+ n(2n)2
k=1
k=1
1
n(n + 1)
n(n + 1)(2n + 1)
= lim
4
4
+n
n+ 4n3
6
2
1
1
1
= .
= lim
n+ 3
12n2
3
4.1.1
k=1
= lim
n+
k=1
1 4n3 n
4n3
3
Remark. We observe that, by definition of Riemann sum, at the subintervals where f (x) < 0 (and,
therefore, f (k ) < 0) the rectangles provide a negative area and then
Z a
Z a
Z a
f even
f (x) dx = 2
f (x) dx
i
f odd
f (x) dx = 0.
a
35
f (x) dx = 0.
Notation. In order to understand the integral symbol in case of ends not necessarily ordered, we take
the agreement
Z b
Z a
f (x) dx.
f (x) dx =
a
( f (x) + g(x)) dx =
f (x) dx =
f (x) dx +
f (x) dx +
f (x) dx
Z b
Z b
f (x) dx
|f (x)| dx (triangular inequality)
a
g(x) dx
Example. We reintroduce the previous example and we use that f (x) = x2 is an even function to
calculate
Z 1
Z 1
Z 1
1
3x2 dx = 3
x2 dx = 3 2
x2 dx = 3 2 = 2.
3
1
1
0
4.2
indefinite integrals
x3
3
Remark. In this example we have seen that F (x) = x3 2 is a primitive of f but of course, for example,
3
3
G(x) = x3 + 1 and H(x) = x3 also are. It is easy to see that all the primitives of a same function f differ
only of constants, or, to put it another way, that, if F (x) is a primitive of f , then all the primitives of f
are of the form F (x) + c, with c R.
Notation. The symbol
f (x) dx
is used to denote the set of all the primitives of f , and is called indefinite integral. In the previous
example, we have
Z
x3
x2 dx =
+ c, amb c R.
3
Remark. By indefinite integral definition,
Z
0
f (x) dx = f (x)
36
(abusing a little of notation, since we are derivating a set of functions which have a common derivative)
and therefore somehow the indefinite integral and the derivative are inverse operations.
It is easy to check if a function is anothers primitive, but the definition does not explicit how to find
these primitives not even if the whole function admits them. In fact, this is not true generally, but we
can see that every continuous function admits primitives.
Remark. Going a little further, a function admitting primitives does not mean that these can happen
explicitly as a composition of elemental functions, as we will see later on. A classic example is the one of
2
the family of functions fa (x) = eax , with a 6= 0.
By the definition of primitive, it is enough with observing the list of derivatives of the elemental functions
to deduce the following list:
R
f (x) = k R
f (x) dx = kx + c
R
+1
f (x) = x
f (x) dx = ex + c
R
ax
, per a tot a > 0, a 6= 1
f (x) = ax
f (x) dx = log(a)
R
f (x) = sin(x)
f (x) dx = cos(x) + c
R
f (x) = cos(x)
f (x) dx = sin(x) + c
R
f (x) = cos12 (x)
f (x) dx = tan(x) + c
R
1
f (x) = 1+x
f (x) dx = arctan(x) + c
2
R
f (x) = sinh(x)
f (x) dx = cosh(x) + c
R
f (x) = cosh(x)
f (x) dx = sinh(x) + c
4.2.1
Proposition. Let f, g : D R R functions which admit primitives. Then, for any , R, the
function f (x) + g(x) admits primitives and
Z
Z
Z
( f (x) + g(x)) dx = f (x) dx + g(x) dx.
Example.
(2 5x + 3) dx =
2 dx 5
x dx +
3 dx =
2x
x3
5
+ 3x + c.
log(2)
3
Remark. From the properties for derivatives, such as Leibnizs rule, many properties for the calculation
of primitives are deducted, but they are usually of little use because they only apply to functions with
very specific functions. However, applying the rule which is deduced from the chain rule is usual. If the
functions f and g are differentiable, then
Z
g 0 (f (x)) f 0 (x) dx = g(f (x)) + c.
37
Example.
6x
dx =
x2 5
2x
dx = 3
x2 5
2x
dx = log |x2 5| + c.
x2 5
log(x) dx =
log(x) 1 dx = log(x) x
1
x dx = x log(x)
x
1 dx = x log(x) x + c.
(1)
dx =
e 2t dt = 2
Z
(3)
t
t
t e dt = 2 t e 1 e dt = 2(t + 1)et + c = 2( x + 1)e x + c.
t
(2)
In (1) we have used the change of variable t = x = 1 (x) (t) = t2 = x 2t dt = dx, and in (3)
we have undone it; in (2) we have done an integration by parts.
4.3
f (x) dx =
f (t) dt + c,
is the formula that links the two integral types. If, furthermore, we add the fact that F (x) =
Rwhich
x
f
(t)
dt and that F 0 (x) = f (x), this formula is equivalent to
a
Z
F 0 (x) dx = F (x) + c,
which, again, tells us that somehow the integral and the derivative are inverse operations.
Corollary (Second fundamental theorem of calculus). Let f : [a, b] R a continuous function and let
F : [a, b] R a primitive of f . Then,
Z b
f (x) dx = F (b) F (a).
a
38
x2 dx =
x3
3
1
13 03
1
= .
3
3
3
This way of calculating definite integrals provokes that usually the theorems of integration by parts and
integration by change of variable are written directly in terms of definite integrals, which means:
Theorem (Integration by parts). If f, g : [a, b] R are differentiable functions, then
Z
f (x) g (x) dx =
[f (x) g(x)]ba
f (x) dx =
1 (b)
1 (a)
4.4
improper integrals
Definition. It is said that f : (a, b] R is (Riemann) integrable (or that the integral of f in (a, b] is
convergent) if the limit
Z b
Z b
lim
f (t) dt
or
lim
f (x) dx
xa+
h0+
a+h
Rb
exists and is finite. In this case, the integral of f between a and b, a f (x) dx, is defined as this limit
value. Similarly, it is said that f : [a, b) R is (Riemann) integrable (or that the integral f integral in
[a, b) is convergent) if the limit
lim
xb
f (t) dt
o be
lim
h0
b+h
f (x) dx
Rb
exists and is finite. In this case, the integral of f between a and b, a f (x) dx, is defined as the this limit
value. In any of the two cases, it is said that the integral is divergent if the limit exists but it is not finite.
39
a
a
+
xa
x (t a)
+
However, if 6= 1,
lim
xa+
(t a)
dt = lim
xa+
= lim
xa+
(t a)
dt = lim
xa+
(b a)1 (x a)1
134
1
1
and this limit is finite if and only if 1 > 0 < 1. All in all,
Z b
(ba)1 si < 1
1
1
lim
dt =
+
xa+ x (t a)
si 1
Rb
exists and is finite. In this case, the integral of f between and b, f (x) dx, is defined as this
limit value. Similarly, it is said that f : [a, +) R is (Riemann) integrable (or that the integral of f
in [a, +) is convergent) if the limit
Z
x
lim
x+ a
f (t) dt
R +
exists and is finite. In this case, the integral of f between a and +, a f (x) dx, is defined as this
limit value. Finally, it is said that f : (, +) R is (Riemann) integrable (or that the integral of f
in (, +) is convergent) if the integrals of f in (, 0] and in [0, +) are convergent. In this case,
the integral of f between and + is defined as
Z +
Z 0
Z +
f (x) dx =
f (x) dx +
f (x) dx.
In any of the three cases, it is said that the integral is divergent if the limit(s) exist(s) but is(are) not
finite.
40
an
alogamente, si f : ( 1, a] ! R es continua, definimos
Z
numerical integration
f (x)dx =
1
lm
m! 1 m
f (x)dx ,
We have already seen that not every primitive can be expressed explicitly as a composition of elemental
functions.
of the
guarantees us that the function
y si the
f : (first
1,fundamental
+1) ! R theorem
es continua
encalculation
todo
R definimos
R xFor example,
2
2
F (x) = a ex dx is a primitive of f : [a, b] R, f (x) = ex , but F (x) cannot appear in a simpler
Z +1
Z have
c
+1
form than the one of the integral
function, as we
already Zexplained
in the indefinite integrals
f
(x)dx
=
f
(x)dx
+
f
(x)dx
section. Therefore, in this case,
the second fundamental theorem of the calculation. does not simplify the
R 1 x
2 1
1
and it is necessary
to turn to cmethods of numerical integration or
calculation of, for example, 0 e dx,
(numerical) quadrature to approximate its value.
Ejemplos notables:
k=0
+1
+1
1
where xk = a + k h for every 0 k n,
and1 the wk are
fixedsicoefficients
(for each method) which are
dx =functions
1
called weights. The methods based on interpolation
that
we
will
see
next are a particular case
x
si > 1
1
1
of these formulas, while, as we will see, the idea of the methods based on differential equations in totally
Z +1
Z +1
different.
(x )2
p
4.5.1
Methods based on
x2
1
interpolation
dx =
functions
1
e
2
2 2
dx = 1 ,
de una
primitive typically
a
polynomial.
This
is
done
in
each
subinterval
of
the
regular
partition
and we
distribuci
on de probabilidad debe dar 1 (probabilidad del 100 %!).
obtain a general formula of integration for each type of function g that we use.
135
Rectangular method In each subinterval [xk1 , xk ], the function g is a polynomial of degree 0 (conx
+x
stant) which coincides with f in the mean point, k12 k . Then,
Z xk
xk1 + xk
f (x) dx (xk xk1 ) f
2
xk1
41
n
baX
xk1 + xk
.
f (x) dx
f
n
2
k=1
We can see that a bound of the error made using this approximation is
ba
max |f 00 ()| h2 .
24 [a,b]
Example. We approximate
Z
ex dx
R1
0
1 0 0.052
2
2
e
+ e0.15 + + e0.95 = 0.747130877747997 . . .
10
Trapezoidal method In each subinterval [xk1 , xk ], the function g is a polynomial of degree 1 which
coincides with f in the ends, xk1 and xk . Then,
Z xk
f (xk1 ) + f (xk )
f (x) dx (xk xk1 )
2
xk1
and the formula to calculate the integral of f between a and b is
Z
baX
ba
f (x) dx
(f (xk1 ) + f (xk )) =
2n
n
k=1
!
n1
f (x0 ) + f (xn ) X
+
f (xk ) .
2
k=1
We can see that a bound of the error made using this approximation is
ba
max |f 00 ()| h2 .
12 [a,b]
ex dx
10
10
R1
Example. We approximate
Simpsons method In each subinterval [xk1 , xk ], the function g is a polynomial of degree 2 that
x
+x
coincides with f in the points xk1 , k12 k and xk (the ends and the mean point). Then,
Z xk
xk xk1
xk1 + xk
f (x) dx
f (xk1 ) + 4 f
+ f (xk )
6
2
xk1
42
Z b
n/2
n/21
X
X
ba
f (x) dx
f (x0 ) + 4
f (x2k1 ) + 2
f (x2k ) + f (xn ),
3n
a
k=1
k=1
where n is even, since we subdivide [a, b] in the double of subintervals than before. We can see that a
bound of the error made using this approximation is
ba
max |f (4) ()| h4 .
180 [a,b]
Example. We approximate
Z
ex dx
R1
0
1 0 0.002
2
2
2
e
+ 4 e0.05 + 2 e0.10 + + e1.00 = 0.746824183875915 . . .
3 10
43
` lcul i 2016/2017
ca
Successions and series
5.1
mathematical induction
inducci
o (Mathematical) is a procedure to demonstrate a certain property P(n) for each integer number
n n0 , that is, to demonstrate P(n0 ), P(n0 + 1), P(n0 + 2), etc. It consists of two steps:
1. Basis case: Demonstrate P(n0 ).
2. Inductive step: Demonstrate P(n + 1) supposing certain P(n) (induction hypothesis).
Example. Using induction, we will demonstrate that
1 + 2 + 3 + + n =
n(n + 1)
2
P(n)
1=
1+
3
..
.
1+
..
.
1 + 2 + 3 + + n =
n+1
..
.
1(1+1)
2
2 = 2(2+1)
2
2 + 3 = 3(3+1)
2
(1)
n(n+1)
2
1 + 2 + 3 + + n + (n + 1) =
..
.
? 1(1+1)
2
(2)
(n+1)((n+1)+1)
2
(3)
1 + 2 + 3 + + n + (n + 1) =
n(n + 1)
(n + 1)(n + 2)
(n + 1)((n + 1) + 1)
+ (n + 1) =
=
3
2
2
2
n+1
P(n)
4 | 51 1
4 | 52 1
4 | 53 1
4 | 5n 1
4 | 5n+1 1
44
1. 4 | 51 1 3
2. 5n+1 1 = 5 5n 1 = (4 + 1) 5n 1 = 4 5n + 5n 1. Of course 4 | 4 5n , and 4 | 5n 1 by
induction hypothesis; therefore, 4 | 4 5n + 5n 1 = 5n+1 1 3
5.2
successions of numbers
1
La sucesion ( 12 ) es definitivamente menor de 10
porque lo es para cada
((1)n )nN = (1, 1, 1, 1,n. . .)
n 4.
n2 nN
= (0, 1,
4,para
9, . dar
. .) la definici
Estamos
listos
on de lmite de una sucesi
on.
Remark. Many times we will be interested in slightly modifying the domain of a function because, for
se dice que
la sucesi
on (an ) converge a ` 2 R o que
example, writing Def.:
(n log(n))
n1 is more practical than writing ((n + 1) log(n + 1))nN . We will allow this
little abuse of notation instead of complicating the definition of succession for it to consider these cases.
lm an = `
n!+1
Definition. It is said that the limit of (an ) is ` R (or that (an ) converges to `), and is written
si 8" > 0 la distancia de an a ` es definitivamente menor que ", o sea 9 n0 (")
tal que |an `| < " 8 n n0 ("). lim an = `,
n+
se ve en la figura siguiente:
n!+1
(resp. lm an =
n!+1
1)
1) o que
3n + 1
3
=
2n + 5
2
lim
n+
5
13
4 2
n>
13 5
4 2
from the definition. If ` R tal que limn+ (1)n = ` existed, taking = 1 we should be able to find a
n0 such that |(1)n `| < 1 for each n > n0 . But, no matter how big is n0 , the even n > n0 will always
make (1)n = 1 and the odd ones (1)n = 1. Therefore, ` should be a value such that |1 `| < 1 and
also | 1 `| < 1, that is, such that 0 < ` < 2 and also 2 < ` < 0, which is not possible.
Example. The number e comes precisely from the limit of a succession, since we can see that
1 n
lim
1+
= e.
n+
n
Definition. It is said that the limit of (an ) is + (or that (an ) diverges at +), and it is written
lim an = +,
n+
if, for each M > 0, exists n0 N such that an > M for every n > n0 . Similarly, it is said that the limit
of (an ) is (or that (an ) diverges at ), and it is written
lim an = ,
n+
if, for each M > 0, n0 N exists such that an > M for every n > n0 .
Example. Like before, we can check that
lim n2 =
n+
M.
xx0
lim f (an ) = `
n+
n+
46
1+
1
n
1+x
1x
= 1 to conclude that
1+n
= 1.
n+ 1 n
lim
5.3
series of numbers
Definition. The series (of numbers) of general term an is the succession (sn )nN of partial additions
sn =
n
X
ak .
k=0
nN an
or simply
an .
Example. We reintroduce the example of the succession (an ) = (3, 1, 4, 1, 5, 9, 2, . . .), for which s0 =
a0 = 3, s1 = a0 + a1 = 4, s2 = a0 + a1 + a2 = 8, etc., that is, (sn ) = (3, 4, 8, 9, 14, 23, 25, . . .).
P
Example. The series of general term (1)n , (1)n , is the succession (sn ) = (1, 0, 1, 0, . . .).
P
P
Definition. It is said that
an converges at S R (or simply that
an converges), and is written
X
X
an = S o be simplement
an < +,
if limn+ sn = limn+
kn ak
Definition.P
Let (an ) a succession of nonnegative terms (that is, such that an 0 for every n D). It
is said that
an diverges at +, and is written
X
an = +,
if (sn ) diverges at +.
Remark. The reason why a the divergence of a series is only defined if the series if on nonnegative terms
is that, if a series of any terms does not converge, its terms can be associated or reordered so that the
addition takes different values. (!)
Example. The behaviour of the geometric series of ratio r R is
a
1r si |r| < 1
X
n
a r = + si r 1
n0
@
si r 1
Example. The behaviour of the generalized harmonic series, for p (0, +), is
(
X 1 < + si p > 1
np = + si 0 < p 1
n1
47
Remark. If
1
n
= 0 and, in contrast,
X 1
1 1 1 1 1 1 1
1 1 1 1 1 1 1
= 1 + + + + + + + + 1 + + + + + + + + =
n
2 3 4 5 6 7 8
2 4 4 8 8 8 8
n+
1
1
1
1 1 1
+ 2 + 4 + = 1 + + + + = +.
2
4
8
2 2 2
P
P
P
Theorem. If
an < + and
bn < +, then ( an + bn ) < + and
X
X
X
( an + bn ) =
an +
bn .
= 1+1
Remark. Be careful with rejecting the hypothesis of convergence of the series, since, if it is not taken
into account, we can obtain, for example,
X
X
X
X
X
X
(n + 1) 1 +
n =
n=
(n + 1)
(n + 1 n) =
1=
n1
n1
(n + 1) 1 +
n1
which is absurd.
5.3.1
n1
n1
n1
n1
(n + 1) =
n2
(n
+
1) 1
n1
(n
+
1) = 1,
n1
P
P
Theorem
(Criterion
of
comparison).
Let
a
and
terms such that an bn .
n
P
P
P bn series of nonnegative
P
Then,
bn < +
an < + or, equivalently,
an = +
bn = +.
Example. The series
n1
converges, since
n1
n3
n3
1
+ 4n
(1) X 1 (2)
1
< +,
+ 4n
n3
n1
1
n13 and the criterion of comparison,
where in (1) we have used that, for all n 1,
n3 n3 +4n
and in (2) that the generalized harmonic series with p = 3 converges.
n3 +4n
n1
n2
3
3n + 5
P
1
1
converges. Like before, we want to compare it with n1 n2 (because, when n is large, n2 3n+5
behaves
1
3
1
like n2 ), but, in this case, n2 3n+5 n2 and not the other way round. This is solved displacing the
additions index, since
n1
X
X
X
3
3
3
1
=
=
=
3
n2 3n + 5
(n + 2)2 3(n + 2) + 5
n2 + n + 3
n2 + n + 3
n1
n1
48
n1
1
and now it is true that n2 +n+3
n12 . As index n 1 is not good for us, because the generalized
harmonic series starts at n 1, we separate the first two terms of the addition (n = 1 and n = 0) and
we have
X
X
X
X
3
1
1
1
1 1
3 2 +
=3
= 3 + +
< +.
n2 3n + 5
n2 + n + 3
3 3
n2 + n + 3
3
n2
n1
n1
n1
n1
X
n
n1
n+
q
n
n
n 1
n
n 1
an .
n
= lim
n+
(1)
n 1 = 0 < 1,
therefore, according to the root criterion, the series is convergent. In (1) we have used that
lim
n+
1/n )
n+
log(n)
n
= e0 = 1.
P
Theorem (Quotient
or
dAlemberts
criterion).
Let
an a series of nonnegative terms and = lim an+1
an .
P
P
Then, < 1
an < + and > 1
an = +.
Example. Let us see that the series
X n+1
n!
n0
(n+1)+1
(n+1)!
lim
n+1
n+
n!
= lim
n+
n+2
= 0 < 1,
(n + 1)2
nn0
n2
1
n logp (n)
< + si p > 1
= + si p 1
According to the previous theorem, the behaviour of this series is the same that that of the improper
integral
Z +
1
dx.
x logp (x)
2
If p = 1,
1
dx = lim
x+
x log(x)
1
dt = lim [log(log(t))]x2 = +.
x+
t log(t)
49
Si p 6= 1,
Z x
1
logp (t)
dt
=
lim
dt =
p
x+ 2
t
2 t log (t)
p+1 x ( logp+1 (2)
log
(t)
p+1
si p + 1 < 0
= lim
=
x+
p + 1 2
+
si p + 1 > 0
1
dx = lim
x+
x logp (x)
successions of functions
n1
Definition. It is said that (fn : D R R)s point limit is f : D R (or that (fn ) converges pointy at
f ), and it is written
fn f,
n+
if limn+ fn (x) = f (x) for every x D, that is, if the limit of the succession of numbers (fn (x)) is the
number f (x), for each x D.
Example. The previous succession (enx )n0 converges particularly at the function
f :[0, +) R
1 si x = 0
x
7 f (x) =
0 si x > 0
since
si x = 0, limn+ fn (0) = limn+ en0 = limn+ 1 = 1
n
si x > 0, limn+ fn (x) = limn+ enx = limn+ (ex ) = 0 because 0 < ex < 1 (x > 0)
50
Example. The previous succession (cos(nx)/n)n1 converges particularly at the identically null function
f :R R
x 7 f (x) = 0
since, for any x R, cos(nx) it oscillates between 1 and 1 when n + and limn+
lim fn (x) = lim
n+
n+
1
n
= 0, so,
cos(nx)
= 0.
n
Definition. It is said that (fn : D R R)s uniform limit is f : D R (or that (fn ) converges
uniformly at f ), and it is written
f,
fn
n+
if, for each > 0, n0 N exists such that |fn (x) f (x)| < for every n > n0 and every x D, that is,
if fn s graph is between that of f and that of f + for every n > n0 .
Example. We reintroduce the example (enx )n0 to see that the convergence is not uniform. For = 21 ,
we should be able to find a n0 0 such that |enx f (x)| < 12 for every n > n0 and every x [0, +);
in particular, such that |enx 0| < 21 for every n > n0 and every x > 0. But, for every n > n0 , the
point x = log(2)
provokes
n
1 1
n log(2)
log(2)
n = e
e
= .
2
2
Example. We reintroduce the example (cos(nx)/n)n1 to see that the convergence is uniform. For each
> 0,
cos(nx)
| cos(nx)|
1
|fn (x) f (x)| =
0 =
n
|n|
n
and
1
n
f fn f .
Remark. fn
n+
n+
5.5
series of functions
Definition. The series of functions of general term fn is the succession (Fn )nN of partial additions
Fn (x) =
n
X
fk (x).
k=0
nN fn (x)
or simply
n0
n0
51
fn (x).
X cos(nx)
n1
= cos(x) +
cos(2x) cos(3x)
+
+
2
3
F ); and then F is called addition of the series in a particular sense (resp. uniif Fn F (resp. Fn
form).
Remark. All series of numbers convergences properties have their analogous versions in terms of particular convergence of series of functions, but not of uniform convergence.
P
Theorem
(Weierstra
(Weierstrass)s
criterion).
Let
fn (x) a series of functions
P
P fn : D R
PR and
an a series of numbers such that |fn (x)| an , for every x D. Then, if
an converges,
fn (x)
converges uniformly.
Example. Lets see that the series
X cos(nx) + sin(nx)
n1
n2
fn (x) is continuous;
P
P
si fn are differentiable F (x) = fn (x) is differentiable and F 0 (x) = fn0 (x);
Rb
P
PRb
if fn are integrable F (x) = fn (x) is integrable and a F (x) dx =
a fn (x) dx.
52
5.5.1
Power series
Among the many properties they have, we have that the power series converge uniformly in any closed
interval contained in the interval (x0 R, x0 + R), where the value
R=
1
1/n
lim |an |
n+
n0
(x x0 )n ,
which is the series that has as partial additions the Taylors polynomials of f in x0 . Remember that
(n)
Taylors polynomials, Tx0 (x), approximate a function f around a point x0 by means of polynomials of
arbitrary degree n; therefore, it is logical to think that the limit of approximations
n
X
f (k) (x0 )
lim Tx(n)
(x) = lim
0
n+
n+
k!
k=0
(x x0 )k =
X f (n) (x0 )
n0
n!
(x x0 )n
should be the best possible approximation, that is, the same function f . Unfortunately, this intuitive
argument is not true for all functions but it is for a large number of elemental functions and their
compositions. Functions with this property, that is, that can be written as a power series, are called
analytic functions.
Example. We can check that the following functions are analytic in the domains indicated.
ex =
xn
n0 n!
sin(x) =
cos(x) =
n0 n
x2
2!
sinh(x) =
cosh(x) =
n x2n
n0 (1) (2n)!
x2n+1
n0 (2n+1)!
x3
3!
+ for every x R
=x
x2n
n0 (2n)!
x2
2
x3
3
(1)
2!
xn = 1 + x +
n x2n+1
n0 (1) (2n+1)!
n+1 xn
n1 (1)
n
log(1 + x) =
(1 + x) =
=1+x+
x4
4
x2 +
(1)(2)
3!
=x
x3
3!
x5
5!
x7
7!
x2
2!
x4
4!
x6
6!
+ for every x R
=1
+ for every x R
=x+
x3
3!
x5
5!
x7
7!
x2
2!
x4
4!
x6
6!
+ for every x R
=1+
+ for every x R
53
xn
n0
n+
so the series converges uniformly in closed intervals of (1, 1). Therefore, if |x| < 1, the adding function
is
X
1
xn =
1x
n0
1
is analytic in |x| < 1 because it can be written
as it is a geometric P
function. All in all, the function 1x
n
as the power series n0 x . Note that this function has a singularity in x = 1, since its denominator is
cancelled in it, therefore it makes complete sense that the radius of convergence is R = 1.
nxn
n1
n+
so the series converges uniformly in closed intervals of (1, 1). Therefore, if |x| < 1, the adding function
is
X
X
nxn = x
nxn1
n1
n1
P
This way, we will calculate first the adding function g(x) = n1 nxn1 and afterwards we will and the
addition of the original series. For |x| < 1, we have that
Z
Z X
XZ
X
X
1
n1
g(x) dx =
nx
dx =
nxn1 dx =
xn = 1 +
xn = 1 +
1x
n1
Therefore,
n1
n1
g(x) = 1 +
X
n1
1
1x
0
nxn = x g(x) =
n0
1
(1 x)2
x
.
(1 x)2
Example. We see that function f (x) = e1/x is indefinitely differentiable but not analytic. Therefore,
we calculate its Taylor series in x0 = 0,
X f (n) (0)
xn ,
n!
n0
for which we need to evaluate f (0), f 0 (0), f 00 (0), etc. At first it looks like the function has a singularity
in x = 0, but if we do
2
lim f (x) = lim e1/x = 0,
x0
x0
54
x0
2
x3
2 1/x2
e
=0
x3
using the infinitesimals hierarchy. Proceeding the same way, we can check that, in fact, f (n) (0) = 0 for
2
every n 0. Therefore, f (x) = e1/x en x0 = 0s Taylor series is identically null,
X 0
xn = 0,
n!
n0
55