- #1
- 2,355
- 334
I just bought a copy of Conway's book "On Numbers and Games" which I had read many years ago with great delight. It reminded me of a little construction I did on my own which I thought I'd share.
In Conway's book he defines surreal numbers as (equivalence classes) of ordered pairs of numbers recursively said pairs generalizing Dedekind cuts and further generalizing to "games".
I found another construction based on the binary expansion of integers.
Start with:
I. [tex] 0=\{\}[/tex]
Then define the sets you construct by inclusion and union starting with 0 by its binary expansion:
II. [tex] \{ n_k\} = \sum_k 2^{n_k}[/tex]
Thus for example:
[tex] 9=8+1=2^3+2^0=\{3,0\}=\{(2^1+2^0),0\}=\{\{1,0\},0\}=\{\{\{0\},0\},0\}[/tex]
Every non-negative integer is a finite set and every finite set is a non-negative integer.
[tex] S \equiv \sum_{n\in S} 2^n[/tex]
Addition and multiplication are defined algorithmically as usual binary addition and multiplication. I won't go through the trouble of recasting these algorithms in set notational form as it is an easy exercise.
Now consider then the set of all natural numbers:
[tex] \mathbb{N}=\{0,1,2,\cdots\}[/tex]
Observe that following the algorithm we have:
[tex] 1 + \mathbb{N} = 0[/tex]
(assuming we can carry out the algorithm on a transfinite infinite bit computer we can show that each "bit" is set to zero at some finite step.)
We thus define the set of natural numbers as the number -1. Further using one's complement we can define the entirety of the integers as subsets of the set of natural numbers. Example:
[tex] -5 = \mathbb{N}\oplus 5 + 1 = \{0,1,3,4,\cdots\}=\cdots111011_2[/tex]
Ok, nothing especially exciting here. We are simply doing arithmetic on an infinite computer. However note that we may add and multiply any subsets of -1={0,1,2,...} not all of which represent natural numbers. The question then is which do behave like regular numbers?
(Henceforth I will use binary notation for the most part or at worst one level of brackets.)
Consider for example the "set of even natural numbers":
[tex]E=\{\cdots 6,4,2,0\}=...10101010101_2[/tex]
[tex] E\times 3 = ...1010101_2\times 11_2 = \cdots 11111_2=-1 [/tex]
Thence we may take:
[tex] E = -1/3[/tex]
(And we can take the one's complement to define +1/3!)
Now that is interesting! With some work you can find that all rationals with reduced odd denominators may be so expressed.
Finally note that if we go on to include sets containing negative integers
e.g. [tex] 3.625=11.101_2 = \{1,0,-1,-3\}[/tex]
then we may define in addition fixed point binary numbers and thence include even denominator rationals. In order for the addition and multiplication algorithms to be well defined we must restrict ourselves to finite fixed point binary expansions. In short we stick to subsets of the set of integers which have a least element.
So what other "numbers" lie in the power set of -1={...4,3,2,1,0}?
Since we can express odd fractions, we might for example consider if the limit
[tex] e = \lim_{n\to\infty,n odd} (1+1/n)^n[/tex]
is meaningful as a limit in terms of the algorithmic definition of addition and multiplication. I.e. do the binary values below a given fixed index eventually become fixed themselves?
Lots of possible fun here. Enjoy!
Regards,
James Baugh
P.S. I seem to recall finding square roots of some integers but can't recover the construction...possibly I am only remembering trying to find such and forgetting my failure. It has been over a decade since I played with this idea.
R. J.B.
In Conway's book he defines surreal numbers as (equivalence classes) of ordered pairs of numbers recursively said pairs generalizing Dedekind cuts and further generalizing to "games".
I found another construction based on the binary expansion of integers.
Start with:
I. [tex] 0=\{\}[/tex]
Then define the sets you construct by inclusion and union starting with 0 by its binary expansion:
II. [tex] \{ n_k\} = \sum_k 2^{n_k}[/tex]
Thus for example:
[tex] 9=8+1=2^3+2^0=\{3,0\}=\{(2^1+2^0),0\}=\{\{1,0\},0\}=\{\{\{0\},0\},0\}[/tex]
Every non-negative integer is a finite set and every finite set is a non-negative integer.
[tex] S \equiv \sum_{n\in S} 2^n[/tex]
Addition and multiplication are defined algorithmically as usual binary addition and multiplication. I won't go through the trouble of recasting these algorithms in set notational form as it is an easy exercise.
Now consider then the set of all natural numbers:
[tex] \mathbb{N}=\{0,1,2,\cdots\}[/tex]
Observe that following the algorithm we have:
[tex] 1 + \mathbb{N} = 0[/tex]
(assuming we can carry out the algorithm on a transfinite infinite bit computer we can show that each "bit" is set to zero at some finite step.)
We thus define the set of natural numbers as the number -1. Further using one's complement we can define the entirety of the integers as subsets of the set of natural numbers. Example:
[tex] -5 = \mathbb{N}\oplus 5 + 1 = \{0,1,3,4,\cdots\}=\cdots111011_2[/tex]
Ok, nothing especially exciting here. We are simply doing arithmetic on an infinite computer. However note that we may add and multiply any subsets of -1={0,1,2,...} not all of which represent natural numbers. The question then is which do behave like regular numbers?
(Henceforth I will use binary notation for the most part or at worst one level of brackets.)
Consider for example the "set of even natural numbers":
[tex]E=\{\cdots 6,4,2,0\}=...10101010101_2[/tex]
[tex] E\times 3 = ...1010101_2\times 11_2 = \cdots 11111_2=-1 [/tex]
Thence we may take:
[tex] E = -1/3[/tex]
(And we can take the one's complement to define +1/3!)
Now that is interesting! With some work you can find that all rationals with reduced odd denominators may be so expressed.
Finally note that if we go on to include sets containing negative integers
e.g. [tex] 3.625=11.101_2 = \{1,0,-1,-3\}[/tex]
then we may define in addition fixed point binary numbers and thence include even denominator rationals. In order for the addition and multiplication algorithms to be well defined we must restrict ourselves to finite fixed point binary expansions. In short we stick to subsets of the set of integers which have a least element.
So what other "numbers" lie in the power set of -1={...4,3,2,1,0}?
Since we can express odd fractions, we might for example consider if the limit
[tex] e = \lim_{n\to\infty,n odd} (1+1/n)^n[/tex]
is meaningful as a limit in terms of the algorithmic definition of addition and multiplication. I.e. do the binary values below a given fixed index eventually become fixed themselves?
Lots of possible fun here. Enjoy!
Regards,
James Baugh
P.S. I seem to recall finding square roots of some integers but can't recover the construction...possibly I am only remembering trying to find such and forgetting my failure. It has been over a decade since I played with this idea.
R. J.B.
Last edited: