Recap: it would be nice to add a number such as as an "infinite number", and then to determine the behaviour of a sequence as "goes to infinity" just look at .
Adding an "infinite number" to a system in this way is similar to the process when a system is "completed" by adding all values of limits (such as how each irrational real number is a limit of rationals and added to to form ). The problems with the example of are that (1) we need to define the sequence's value at and (2) there are, in fact, many ways of "going to infinity" and each should get a different infinite number like .
With that said, Robinson's method for constructing nonstandard numbers does have a lot of similarities with "completions" in mainstream mathematics.
This section covers the ultrafilter construction: it is rather concrete and straightforward. We start with a sequence and want to investigate its behaviour "at infinity".
A real-valued process is a sequence of real numbers We will make all real valued processes into numbers with the help of an ultrafilter. An ultrafilter is a subset of where is an infinite indexing set (here we take ). The ultrafilter must satisfy certain axioms.
Think of as specifying the large subsets of . (They are so large that the intersection of any two large sets is large.)
The easiest example of an ultrafilter is the principal ultrafilter for some , defined as . An ultrafilter is nonprincipal if it is not of this type.
For an example of a nonprincipal ultrafilter one must work a bit harder. Start with . This is a filter, i.e. satisfies the first three axioms. Now using Zorn's lemma extend to a maximal filter in . This will be an ultrafilter: the mathematical work is done when given a filter not containing nor one observes that
is a filter properly extending . To see that an ultrafilter extending is not principal just note that it contains for all .
We now fix an ultrafilter for extending and look at real-valued processes . The trick is to say two such sequences and are equivalent (essentially equal) if the set of indices where they are equal is "large" i.e. in .
Exercise: show this is an equivalence relation.
I will notate the equivalence class of the sequence by .
The new extended set of numbers, denoted , or , is the set of real-valued processes modulo this equivalence. It includes all normal real numbers because each gives rise to a constant process , all , whose equivalence class we identify with .
Exercise: show that all distinct constant real-valued processes are inequivalent under the equivalence relation.
For an example of an infinite number, consider the process given by . This is not equal to any since is in . The number corresponding to the original is typically a new number too. To picture where this arises we need to put some new structure on .
Other operations such as and can be defined in an analogous way. In all cases, these definitions agree with the usual one on because of the identification of with a constant sequence and .
Exercise. For show that for all . (We write this as .)
The system contains infinitesimal numbers too, such as where . It is easily seen that for all positive . (Exercise.)
Here is an important definition: write when is zero or infinitesimal.
Proposition. For , for at most one .
When for some we say is finite and write .
Theorem. For the sequence and we have, iff for all ultrafilters .
One direction goes by assuming and showing for each positive . This is rather standard in the literature. The other direction (which is less often presented) goes by assuming is not a limit and so for some in there are with and for . Then there is an ultrafilter containing and
showing for this .
We haven't quite got to where we want to be: we have infinite numbers such as the number for the process but we haven't made sense of yet, and the characterisation of convergence involves a quantifier over ultrafilters .
The first is easily fixed. Since is a function it extends naturally to the nonstandard universe. First
The set is the set of where . For any such we set
Note that only the matter because there is a -large set of such indices . We could set for all other , or indeed anything else.
Now we have for any , including infinite such .
Theorem. iff for all infinite .
For one direction, suppose and . then for each there is so that , and as is infinite, . Therefore .
For the other, suppose is false and take and natural numbers with . So is infinite (because the sequence is increasing) and .
Remark. It is certainly remarkable that the notion (which involves three quantifiers, ) is equivalent to " for all infinite " which involves only one quantifier, quantifying over infinite . Of course the "infinite" is doing some work as is . Note that neither of these two key notions are defined in the "usual" way by saying some set of indices is in . You could ask, given that gives a function with we should be able to obtain a similar function from the nonstandard characterisation. Indeed we can: take
It is easy to check that this definition gives finite for positive standard real .