Professional Documents
Culture Documents
Ï ×ø Öø Û Ø Ø Òóø
Ï ×ø Öø Û Ø Ø Òóø
SETS
1. INTRODUCTION
We start with the notion of an element or obje
t under
onsideration; e.g., an integer like 2, a
variable like x, a ve
tor in a plane, represented by an arrow, et
. A
olle
tion of su
h obje
ts forms
what we
all a set. And the obje
ts that
omprise the
olle
tion are
alled the elements or members
of that set. [Note: The idea of a set is not as simple as it may sound. To avoid logi
al diÆ
ulties
later, one must in fa
t have to be quite pre
ise in dening exa
tly what a set is. For our purposes
however, we shall be
ontent with our usual day-to-day pedestrian notion of a set℄.
A set may be denoted by an a
tual enumeration of its members, su
h as f1; 2; 3g, the set
onsisting
of the rst three positive integers. [Q: How do you know that this is a set
onsisting of the rst
three positive integers, and not a set
onsisting of the numerals 1, 2, and 3?℄. A set may also
be denoted by a rule whi
h distinguishes its members from the rest of the obje
ts in the world,
i.e., S 1 = fz : jz j = 1g. This latter set is the unit
ir
le in the
omplex plane. On the other
hand, a set's elements need not share a
ommon
hara
teristi
. For example, one
an have a set
A = f; 1; x; banana; g. Sets will be usually symbolized by upper
ase letters: A, B , X , , et
.
When we used the word \
olle
tion" above, we ta
itly assumed that there are obje
ts to
olle
t. But
in fa
t, a
olle
tion, or a set may be empty. We thus denote the empty set by the symbol ? or f g;
it is also
alled the null set. Sin
e there is no restri
tion as to what a set may
ontain, then it may
ontain another set, e.g., f1; 2; f1; 2g; 3; f1; 2; 3g : : : g. Note very well that if N = f g, then N is
empty. However, M = f f g g is not empty. It is a set
ontaining exa
tly one element, ?, whi
h is
the empty set f g. To avoid monotony and enhan
e
omprehensibility, we shall often refer to a set
of sets as a \
olle
tion" of sets or, even better sounding, a \
lass" of sets.
An interesting problem whi
h illustrates the logi
al diÆ
ulties that may
rop up
in set theory is Russel's paradox: sin
e a set may also
ontain a set as one of its
elements, then it may very well
ontain itself. A set whi
h
ontains itself is termed
abnormal, otherwise it is normal. Now
onsider the set N of all normal sets. Is
it normal or abnormal? If N is normal, then it must belong to itself, sin
e it is
the
olle
tion of all normal sets. But if N belongs to itself, then by denition N is
abnormal, in whi
h
ase it is not
ontained in N . Hen
e N does not
ontain itself
and therefore it is normal, and therefore
ontains itself...
We shall use the usual notations: if the element x is in the set A, we shall write x 2 A or,
onversely,
A 3 x to mean that the set A
ontains x. Set equality A = B means that all elements of A are also in
B and vi
e versa. If the set B is in
luded in A, we write B A, whi
h means that x 2 B ) x 2 A.
On the other hand, B A allows for the possibility of the two sets being equal. These
an also be
written in another way: A B or A B .
The important properties of set in
lusion are:
1
2
(1) ? A 8A;
(2) A A 8A;
(3) A B; B A ) A = B ;
(4) A B; B C ) A C:
To avoid logi
al diÆ
ulties, we shall assume that there exists a universal set U of whi
h all sets we
shall dis
uss are subsets. Furthermore, all statements made about sets are within the
ontext of the
existen
e of su
h a universal set.
The power set of a set A, denoted by P(A), is dened as the
olle
tion of all the subsets of A.
2. Set Algebra
Given two sets, there are two basi
binary operations that are dened, set union and set interse
tion,
denoted by [ and \:
(5) A [ B := fx : x 2 A or x 2 B g
(6) A \ B := fx : x 2 A and x 2 B g:
The operations of forming unions and interse
tions are
ommutative and asso
iative
A [ B = B [ A; (A [ B ) [ C = A [ (B [ C )
A \ B = B \ A; (A \ B ) \ C = A \ (B \ C ):
These operations also obey the distributive laws:
(A [ B ) \ C = (A \ C ) [ (B \ C )
(A \ B ) [ C = (A [ C ) \ (B [ C ):
We dene set
omplementation as follows: the
omplement of a set A, denoted by A
, is:
(7) A
:= fx : x 2= Ag:
[You should realize that, had we not assumed the existen
e of the universal set U , of whi
h A is a
subset, this denition would not make any sense.℄
Set
omplementation has the following properties:
(A
)
= A; ?
= U; and U
= ?
A [ A = U; A \ A
= ? and A B () B
A
(A [ B )
= A
\ B
and (A \ B )
= A
[ B
:
The last two equations are equivalent to De Morgan's rules in symboli
logi
:
(a b) a _ b; and (a _ b) a b:
3
Using the two basi
operations dened above, one
an dene still other operations. For example, we
an dene set dieren
e and symmetri
dieren
e as:
(8) A B := A \ B
= fx : x 2 A and x 2 = Bg
(9) A 4 B := (A B ) [ (B A):
Let X and Y be two arbitrary sets. A rule asso
iating a unique element y 2 Y with ea
h element
x 2 X is
alled a mapping of X into Y . Su
h a mapping may be denoted by a symbol, say, f , in
whi
h
ase we write the usual notation y = f (x). X is then
alled the domain of the mapping f ,
and we write X = Dom(f ), while Y is
alled the
o-domain of f . We usually use loosely the term
fun
tion for f ; stri
tly speaking, the \fun
tion" is the triple (X; f; Y ) The notation f : X ! Y is
also used for the mapping f of X into Y , and we also write f : x 7! y for y = f (x).
Y
X f
0011
0011
y=f(x)
x
1100
In the mapping f : x 7! y, the element y 2 Y is the image of x 2 X under the mapping f , while x
is the pre-image of y 2 Y: Let M X ; then the subset ff (x) : x 2 M g Y is also
alled the image
of M under the mapping f; and denoted by f [M ℄: The set of all the images of the elements of X is
the range of the fun
tion f , denoted by Ran(f ), whi
h may also be written as f [X ℄.
The
omposite mapping of two mappings f : X ! Y and g : Y ! Z is the mapping g Æ f : X ! Z ,
given by g Æ f : x 7! g(f (x)).
The mapping f is an extension of g if Dom(f ) Dom(g), and f (x) = g(x) 8x 2 Dom(g). Conversely,
g is a restri
tion of f . The restri
tion of f to a domain M is denoted by f jM .
If no two distin
t elements in X have the same image in Y under the mapping f : X ! Y , then
the mapping is said to be inje
tive or one-one. If Y = f [X ℄, then f is said to be surje
tive or onto.
If f is both inje
tive and surje
tive, then it is
alled bije
tive or a one-to-one
orresponden
e. If f
4
go f
f g
X f Y g g(f(x))
1
0
11
00
x 11
00
f(x) 0
1
00
11 Z
go f
PROOF: We prove this by showing that the left side of the equation is a subset of the right side,
and vi
e versa. Let x 2 f 1 [A [ B ℄. This means f (x) 2 A [ B , so that x must belong to f 1[A℄ or
f 1 [B ℄. Hen
e, f 1 [A [ B ℄ f 1 [A℄ [ f 1 [B ℄.
Now suppose x 2 f 1[A℄ [ f 1 [B ℄. Then x must belong to f 1 [A℄ or f 1 [B ℄, whi
h means that
f (x) must belong to A or B or, f (x) 2 A [ B . But this means that x 2 f 1 [A [ B ℄. Hen
e,
f 1 [A℄ [ f 1 [B ℄ f 1 [A [ B ℄. :)
The next two theorems may be proven as exer
ises:
Theorem 2. The pre-image of the interse
tion of two sets is the interse
tion of the pre-images of
the sets:
f 1 [A \ B ℄ = f 1 [A℄ \ f 1 [B ℄:
Theorem 3. The image of the union of two sets is the union of the images of the sets:
f [A [ B ℄ = f [A℄ [ f [B ℄:
In fa t, we have a whole slew of relations for the mapping f , whose proofs are left as exer ises:
?℄ = ?; f [X ℄ Y ;
f[
A B ) f [A℄ f [B ℄;
f [[i Ai ℄ = [i f [Ai ℄;
f [\i Ai ℄ \i f [Ai ℄:
5
And similarly, for the inverse mapping f 1 :
(10) f 1[?℄ = ?; f 1 [Y ℄ = X ;
(11) A B ) f 1 [A℄ f 1 [B ℄;
(12) f 1 [[i Bi ℄ = [i f 1 [Bi ℄;
(13) f 1 [\i Bi ℄ = \i f 1 [Bi ℄;
(14) f 1 [B
℄ = f 1 [B ℄
:
4. SET PRODUCTS
Given two sets X and Y , we dene their produ t X Y as the set of ordered pairs (x; y):
X Y := f(x; y) : x 2 X; y 2 Y g:
Generalizing this, if we have a
olle
tion of sets indexed by i, fXi g; i = 1 : : : ; n we dene the n-fold
produ
t as the set of ordered n-tuples:
X1 Xn := f(x1 ; x2 ; ; xn ) : x1 2 X1 ; x2 2 X2 ; ; xn 2 Xn g:
Example 1: The \plane" is the set produ
t R R whose element is the ordered pair (x; y).
Example 2: A manifold is a set whi
h has the property of
ontinuity, or dierentiability; thus, we
may talk of a C k -manifold, where k = 0; 1; 2; : : : ; 1. The
ir
le S 1 is a C 1 manifold. The produ
t
S 1 R = f('; z ) : 2 S 1 ; z 2 Rg is the ordinary
ylinder. The produ
t S 1 S 1 = f(; ') : 2
S 1 ; ' 2 S 1 is the 2-torus; T 2 . A
ylindri
al band is the produ
t of S 1 I; where I R .
Can a manifold always be expressed as a produ
t of two sets? The answer is \no". A
ounter-
example is the well-known Mobius band. Lo
ally, a se
tion of the Mobius band is I I; I R .
Globally however the band
annot be expressed as S 1 I be
ause of the twist in it whi
h makes
the band a non-orientable surfa
e. Non-orientable manifolds like the Mobius band and the equally
notorious Klein bottle are taken as bre bundles whi
h are the generalizations of set produ
ts.
To partition a set means to de
ompose the set into pairwise disjoint subsets whi
h exhaust the set,
i.e., the union of the subsets equals the set. Obviously, unless the set has only one element, there is
more than one way to partition a set.
Example 3: A set X = fa; b;
; d; eg may be partitioned into ffa; dg; f
; b; egg; ff
g; fa; eg; fd; bgg;
et
.
Example 4: The set R 3 may be de
omposed into subsets
onsisting of planes normal to the x-axis.
Su
h a partition is also
alled a foliation of R 3 ; and ea
h plane is
alled a folium. Similarly, R 3 may
also be partitioned into a foliation
onsisting of
on
entri
spheres S 2
entered at the origin.
Denition: A (binary) relation in a set X is a verbal phrase or mathemati
al symbol denoted by R
su
h that for any ordered pair (x; y) 2 X X , the statement x R y
an be
ategori
ally
lassied as
\true" or \false". The symbol x R y may be read as: the element x is related by the relation R to
the element y.
6
The assertion that x R y is false may also be written as x R= y.
Example 5: Take the relation > (\greater than") in the set of the real numbers R. For any x 2 R
and y 2 R , the statement x > y
an be denitely
lassied as either \true" or \false". But > is not
a relation in the set of
omplex numbers C sin
e it is meaningless in that set.
Denition: A relation R in a set X is
alled an equivalen
e relation in X (and denoted spe
i
ally
by ) if it satises the following three
onditions:
1. x x for every x 2 X (re
exivity);
2. x y ) y x (symmetry);
3. x y; y z ) x z (transitivity).
Let be an equivalen
e realtion in a set X . We shall denote [x℄ := fy 2 X : y xg i.e., the
olle
tion of elements of X whi
h are related to x by the equivalen
e relation . We
all the set [x℄
the equivalen
e
lass of x.
Theorem 4. The distin
t equivalen
e
lasses of X form a partition of X .
Two sets X and Y are said to be numeri
ally equivalent, or have the same
ardinality if there is a
bije
tion between them. In other words, the two sets have the same number of elements if there is
a one-to-one
orresponden
e between them. This is obvious for nite sets. We use this as denition
of numeri
al equivalen
e for innite sets.
A set X has a greater
ardinality,
ard X >
ard Y than the set Y if there exists an inje
tive map
from Y to X , but there is no inje
tive map from X to Y .
7
A set is said to be
ountable if it is nite, or if there is a bije
tion between itself and the set of
positive integers Z+ or the
ounting numbers.
Example 7: The set of all integers Z is
ountable, for we
an set a bije
tion between Z and Z+ as
follows:
Z: 0; 1; 1; 2; 2; : : :
Z+ : 1; 2; 3; 4; 5; : : : :
Example 8: The set of all positive even numbers is a subset of Z+ , and a proper one at that. Yet
they have the same
ardinality, sin
e we have the
orresponden
e n $ 2n.
Example 9: The set of perfe
t squares is another proper subet of the
ounting numbers; yet the
two sets have the same
ardinality, sin
e we have the bije
tion n $ n2 .
Example 10: The set Q of rational numbers is
ountable. We prove this. First, given a rational
number p=q, dene its height as h = jpj + q. Now, there is only one rational number with height
h = 1 : 0=1, three numbers with height equal to 2: 0=2; 1=1; 1=1; ve numbers with a height of
3: 0=3; 1=2; 1=2; 2=1; 2=1; et
. We then prune the list by eliminating dupli
ates in values, e.g.,
0=1; 0=2; 0=3; et
., whi
h amount to 0; 1=2; 2=4; 3=6; et
., whi
h amount to 1=2; et
. The result is:
0
h=1:
1
1 1
h=2: ;
1 1
1 1 2 2
h=3: ; ; ;
2 2 1 1
1 1 3 3
h=4: ; ; ;
3 3 1 1
::::::
Clearly, one
an devise a route a
ross the rows and down the
olumns of this table, at the same time
establishing a bije
tion with the set Z+ . Hen
e the set Q is
ountable.
The
ardinality of Z+ and all innitely
ountable sets is denoted by 0 .
Example 11: The set R is un
ountable. It is suÆ
ient for us to prove that the open interval (0; 1)
is not
ountable. Suppose the elements of the interval (0; 1) were
ountable. Then we
an put
the elements in a 1-to-1
orresponden
e with the
ounting numbers by listing them down in some
fashion:
1 : 0:213746692 : : :
2 : 0:001934206 : : :
3 : 0:802881890 : : :
4 : 0:592230864 : : :
:::::: :
This innite list ought to
ontain all the elements of the interval (0; 1). Now we show that, in fa
t,
this list is not
omplete, by a
tually
onstru
ting a number x = 0:x1 x2 x3 x4 : : : whi
h is not in the
list. For x1
hoose any digit 0-9 not equal to 2. Then x
annot be the same as item 1 in the list,
sin
e they dier in the rst de
imal digit. For x2
hose any digit 0-9 not equal to 0. Then x
an
8
neither be equal to item 2 in the list sin
e their se
ond de
imal digits are not the same. For x3
hoose any digit 0-9 ex
ept 2. Then x
an neither be equal to item 3, sin
e their third de
imal digits
are not the same.
We
ontinue this indenitely,
hoosing for the value of xn any digit other than the n-th de
imal digit
of item n above. Then obviously, x
annot be equal to any of the listed items, hen
e, it is not in
the list,
ontradi
ting our assumption that the above list is exhaustive. This problem of an x whi
h
is not in the list
annot be eliminated by, say, in
luding x in the list. For then, with x already in
the list, we
an repeat the pro
ess and
ome out with yet another number y whi
h again is not in
the list, and this pro
ess
an be repeated ad innitum. This leads us to two
on
lusions: The set
R is not
ountable, and there exists non-rational, or irrational numbers. Thus we say that the set
R (and, therefore, the set I of irrational numbers are un
ountably innite, or simply un
ountable or
nondenumerable.
The
ardinality of an un
ountably innite set is denoted by
.
Theorem 5. The set of all innite sequen
es of 0 and 1 is un
ountable.
PROOF: Exer
ise. :)
Example 12: The set R has
ardinality
. So has the interval (0; 1) - i.e., the two sets are
numeri
ally equivalent. For the proof, see the following gure. The interval (0; 1) has been bent
into a semi-
ir
le and pla
ed tangent to the straight line representing R. The straight line OP Q
establishes the bije
tion between the sets R and (0; 1).
0 1
O
P
Q R
The example
an be generalized into the numeri
al equivalen
e of R and any subset of R whi
h
ontains an open interval; but this would make use of the Cantor-Bernstein Theorem whi
h says
that if X A Y and Y B X; then X Y .
Re
all that given a set X , we dened its power set, P(X ); as the
olle
tion of all the subsets of X :
P(X ) := fS : S X g.
Example 13: Let X = fa; b;
g. Then P(X ) = f?; fag; fbg; f
g; fa; bg; fa;
g; fb;
g; X g. If
ard Z = n, where n is nite, what is
ard P(Z )?
We now have a hierar
hy of
ardinal numbers:
1 < 2 < 3 < : : : < 0 <
Two questions that arise are: a) are there
ardinal numbers greater than
? and b) is there a
ardinal number between 0 and
? The answer to the rst question is \yes", sin
e
ard P(R ) = 2
.
The se
ond is still an open question. George Cantor (1845-1918) believed that there is none. This
assertion is
alled Cantor's
ontinuum hypothesis (CH). Kurt Godel in 1940 showed that CH is
onsistent with the Zermelo-Fraenkel axioms of set theory in the sense that its negation
annot be
9
proved, starting from the ZF axioms. Subsequently, however, Paul Cohen showed that CH itself
annot be proved either, starting from the same axioms. Thus, the most that
an be said at present
is that CH is independent of the ZF axioms.
But what is known is that 20 =
. Thus,
ard P(Z) = 20 =
. Thus, we nally have the hierar
hy:
1 < 2 < 3 < : : : < 0 <
< 2
< 22 < : : :
Lemma 1. There exists no bije
tive mapping between a set X and P(X ).
PROOF : We will suppose that there is a bije
tion
a $ A; b $ B;
$ C; :::
between all the elements a; b;
; : : : of X , and all the elements A; B; C; : : : of P(X ), and arrive at a
ontradi
tion. In the
orresponden
e a $ A, we say that the subset A is asso
iated with the element
a. We now dene Z to be the set of elements of X whi
h do not belong to their asso
iated subsets.
More pre
isely, we pla
e a in Z if a $ A, but a 62 A. Clearly, Z X , and therefore an element of
P(X ). Now let z be the element of X asso
iated with Z . Is z in Z ? If z 2 Z , then z 62 Z , sin
e by
onstru
tion, Z
ontains only those elements whi
h do not belong to their asso
iated subsets. On
the other hand, if z 62 Z , then z 2 Z sin
e Z
ontains pre
isely those elements whi
h do not belong
to their asso
iated subsets. In either
ase, z 2 Z and z 62 Z at the same time, whi
h is impossible.
Hen
e there is no su
h element z whi
h is asso
iated with Z . :)
Theorem 6.
ard P(X ) >
ard X
PROOF: It is lear that ard P(X ) annot be less than ard X sin e the singletons of X like fxg,
form a subset of P(X ) whi
h already is in one-to-one
orresponden
e with X . Thus we need to show
only that
ard P(X ) 6=
ard X , whi
h the pre
eding lemma establishes. :)
Example 14: One further example is worth mentioning here be
ause, although its origin is in
the remote res
esses of pure mathemati
s, it has gured quite prominently in physi
s in the last
two de
ades, parti
ularly in the study of
haoti
phenomena, strange attra
tors and fra
tals. One
may have the (mistaken) notion that sets of dis
rete points (nite or innite) are
ountable, while
un
ountability holds only for \
ontinuous" sets like open intervals. This example will lay that notion
to rest. We refer to the famous Cantor set:
Starting with the interval F0 = [0; 1℄, remove the middle third ( 13 ; 23 ); giving us F1 = [0; 31 ℄ [ [ 32 ; 1℄:
Then from F1 delete the middle thirds ( 19 ; 29 ) and ( 97 ; 89 ); letting F2 denote the remaining
losed set,
et
:
We have the following de
reasing sequen
e of sets:
F0 F1 F2 Fn
The
antor set is then dened to be the interse
tion
\1
F = Fn
n=0
How many points does the Cantor set have? To begin with, it is obvious that it
ontains the points
0; 1; 13 : 32 ; 19 ; 29 ; 97 ; 89 ; : : : ; the end-points of the deleted intervals. But in fa
t, F
ontains a lot
10
F0
0 1
0 1/3 2/3 1
F1
0 1
F3
of other points. To see this, let any x 2 [0; 1℄ be written in ternary notation:
+ 22 + 33 + + nn + ;
a1 a a a
x=
3 3 3 3
where ea
h an
an take on the values 0, 1 or 2. It is easy to see then that a point x is in F - but
not as an end point - only if none of the numbers aP 0 ; a1 ; a2 ; : : : ; an ; : : : is 1. Hen
e, the point x is
in F i it has a ternary expansion of the form x = 1 n=1 2bn=3n; where bn = 0; 1: In other words,
there is a bije
tion between the Cantor set F and the set of innite sequen
es of 0 and 1. But this
set is un
ountable by Theorem 5 above.
As a matter of fa
t, although F [0; 1℄, one
an
onstru
t dire
tly a bije
tion between F and [0; 1℄,
whi
h means that F has the same
ardinality as [0; 1℄. Sin
e a point in F is dened by a sequen
e
(bn ) of 0 and 1 above, then one
an dene a number:
+ 22 + 33 + + nn + ;
b1 b b b
y=
2 2 2 2
where ea
h bn is either 0 or 1. But y is pre
isely the binary expansion of a point in [0; 1℄.
Now, F
ontains no open interval, for, if su
h an interval exists then, for any n, it must be in
luded
in one of the disjoint sub-intervals of Fn , and its length is at most ( 13 )n . Here one has an example
of a set
ontaining no open interval, yet has
ardinality
. One
an also show that F has vanishing
length. If we
al
ulate the total length of the deleted segments, we nd:
1 2 1 2 2 1 2 3 1
Ld =
3
+
3 3
+ 3 3 + 3 3 + = 1:
Hen
e, the deleted segments, F
, has the full measure, and therefore the Cantor set F has a total
length of 0.
Finally, it
an be shown that F is not exa
tly 0-dimensional like a dis
rete set of points, but neither
is it 1-dimensional be
ause it
ontains no open interval. It has a Hausdor dimension of log 2= log 3.
7. ORDER RELATIONS