Non-Borel set/Advanced: Difference between revisions

From Citizendium
Jump to navigation Jump to search
imported>Boris Tsirelson
(New page: {{subpages}} Usually, it is rather easy to prove that a given set is Borel (see below). It is much harder to prove that the set ''A'' is non-Borel; see Advanced if you are acquainted with...)
imported>D. Matt Innis
(make a fake edit to update merged history)
 
(4 intermediate revisions by 2 users not shown)
Line 1: Line 1:
{{subpages}}
{{subpages}}


Usually, it is rather easy to prove that a given set is Borel (see
below). It is much harder to prove that the set ''A'' is non-Borel;
see Advanced if you are acquainted with descriptive set theory. If you
are not, you may find it instructive to try proving that ''A'' is
Borel and observe a failure.


A. The set of all numbers ''x'' such that <math> a_0=3 </math> is an
Usually, it is rather easy to prove that a given set is Borel (see below). It is much harder to prove that the set ''A'' is non-Borel; see [[Non-Borel_set/Advanced]] if you are acquainted with descriptive set theory. If you are not, you may find it instructive to try proving that ''A'' is Borel and observe a failure.
interval, therefore a Borel set.


B. The condition "<math> a_1=3 </math>" leads to a countable union of
A. The set of all numbers ''x'' such that <math> a_0=3 </math> is an interval, therefore a Borel set.
intervals; still a Borel set.


C. The same holds for the condition "<math> a_2=3 </math>" and, more
B. The condition "<math> a_1=3 </math>" leads to a countable union of intervals; still a Borel set.
generally, "<math> a_k=n </math>" for given ''k'' and ''n''.


D. The condition "<math> a_k<n </math>" leads to the union of finitely
C. The same holds for the condition "<math> a_2=3 </math>" and, more generally, "<math> a_k=n </math>" for given ''k'' and ''n''.
many sets treated in C; still a Borel set.


E. The condition "<math> a_k>n </math>" leads to the complement of a set
D. The condition "<math> a_k<n </math>" leads to the union of finitely many sets treated in C; still a Borel set.
treated in D; still a Borel set.


F. The condition "<math> a_k>n </math> for all ''k''" leads to the
E. The condition "<math> a_k>n </math>" leads to the complement of a set treated in D; still a Borel set.
intersection of countably many sets treated in E; still a Borel
set. The same holds for the condition "<math> a_k>7 </math> for all
<math> k>3 </math>" and, more generally, "<math> a_k>n </math> for all
<math> k>m </math> for given <math> m,n. </math>


G. The condition "<math> a_k>7 </math> for all ''k'' large enough"
F. The condition "<math> a_k>n </math> for all ''k''" leads to the intersection of countably many sets treated in E; still a Borel set. The same holds for the condition "<math> a_k>7 </math> for all <math> k>3 </math>" and, more generally, "<math> a_k>n </math> for all <math> k>m </math>" for given <math> m,n. </math>
leads to the union of countably many sets treated in F; still a Borel
set.


H. The condition "the sequence <math> a_1, a_2, a_3, \dots </math>
G. The condition "<math> a_k>7 </math> for all ''k'' large enough" leads to the union of countably many sets treated in F; still a Borel set.
tends to infinity" leads to the intersection of countably many sets of
the form treated in G ("7" being replaced with arbitray natural
number). Still a Borel set!


This list can be extended in many ways, but never reaches the set
H. The condition "the sequence <math> a_1, a_2, a_3, \dots </math> tends to infinity" leads to the intersection of countably many sets of the form treated in G ("7" being replaced with arbitrary natural number). Still a Borel set!
''A''. Indeed, the definition of ''A'' involves arbitrary
subsequences. For given <math> k_0 < k_1 < k_2 < \dots </math> the
corresponding set is Borel. However, ''A'' is the union of such sets
over all <math> k_0 < k_1 < k_2 < \dots </math>; a uncountable union!


Do not think, however, that uncountable union of Borel sets is always
This list can be extended in many ways, but never reaches the set ''A''. Indeed, the definition of ''A'' involves arbitrary subsequences. For given <math> k_0 < k_1 < k_2 < \dots </math> the corresponding set is Borel. However, ''A'' is the union of such sets over all <math> k_0 < k_1 < k_2 < \dots </math>; a uncountable union!
non-Borel. The matter is much more complicated since sometimes the
 
same set may be represented also as a countable union (or countable
Do not think, however, that uncountable union of Borel sets is always non-Borel. The matter is much more complicated since sometimes the same set may be represented also as a countable union (or countable intersection) of Borel sets. For instance, an interval is a uncountable union of single-point sets, which does not mean that the interval is non-Borel.
intersection) of Borel sets. For instance, an interval is a
uncountable union of single-point sets, which does not mean that the
interval is non-Borel.

Latest revision as of 20:47, 30 June 2009

This article is developing and not approved.
Main Article
Discussion
Related Articles  [?]
Bibliography  [?]
External Links  [?]
Citable Version  [?]
Advanced [?]
 
An advanced level version of Non-Borel set.


Usually, it is rather easy to prove that a given set is Borel (see below). It is much harder to prove that the set A is non-Borel; see Non-Borel_set/Advanced if you are acquainted with descriptive set theory. If you are not, you may find it instructive to try proving that A is Borel and observe a failure.

A. The set of all numbers x such that is an interval, therefore a Borel set.

B. The condition "" leads to a countable union of intervals; still a Borel set.

C. The same holds for the condition "" and, more generally, "" for given k and n.

D. The condition "" leads to the union of finitely many sets treated in C; still a Borel set.

E. The condition "" leads to the complement of a set treated in D; still a Borel set.

F. The condition " for all k" leads to the intersection of countably many sets treated in E; still a Borel set. The same holds for the condition " for all " and, more generally, " for all " for given

G. The condition " for all k large enough" leads to the union of countably many sets treated in F; still a Borel set.

H. The condition "the sequence tends to infinity" leads to the intersection of countably many sets of the form treated in G ("7" being replaced with arbitrary natural number). Still a Borel set!

This list can be extended in many ways, but never reaches the set A. Indeed, the definition of A involves arbitrary subsequences. For given the corresponding set is Borel. However, A is the union of such sets over all ; a uncountable union!

Do not think, however, that uncountable union of Borel sets is always non-Borel. The matter is much more complicated since sometimes the same set may be represented also as a countable union (or countable intersection) of Borel sets. For instance, an interval is a uncountable union of single-point sets, which does not mean that the interval is non-Borel.