Let’s call a function a **near-endomorphism** of if there is a constant such that for all . The set of near-endomorphisms of will be denoted by . We put an equivalence relation on by declaring that iff the function is bounded, and let denote the set of equivalence classes.

It’s not difficult to show that defining in terms of pointwise addition and in terms of composition of functions turns into a commutative ring. And it turns out that this ring has a more familiar name… Before reading further, can you guess what it is?

When I first heard that is isomorphic to the field of real numbers, I found it quite surprising. In this post, I’ll attempt to explain why this is true. I think that this description of the reals deserves to be better known!

**Intuition**

Before attempting a proof, let’s try to demystify the isomorphism . The idea is the following. First, we can identify a real number with the corresponding linear function of slope defined by . But note that we can also recover from the **discretized** function defined by , since . The function is not additive, but it is nearly so: it belongs to the set of near-endomorphisms of defined above. Moreover, if we add a bounded function to the resulting function still satisfies since . It turns out (and this is really the key point of the argument below) that **every** near-endomorphism of is equivalent to a function of the form . Moreover, it is not hard to see that is bounded as a function of . These facts together imply that the map defined by is an isomorphism of rings, where denotes the equivalence class of .

**Exercise:** Show that, under the isomorphism , the real number corresponds to the equivalence class of the function defined by .

**Why ‘Eudoxus reals’?**

In this paper by R.D. Arthan, he calls the **Eudoxus reals**, and justifies the terminology in the following way. In Book 5 of Euclid’s “Elements”, Euclid writes:

Magnitudes are said to be in the same ratio, the first to the second and the third to the fourth, when, if any equimultiples whatever are taken of the first and third, and any equimultiples whatever of the second and fourth, the former equimultiples alike exceed, are alike equal to, or alike fall short of, the latter equimultiples respectively taken in corresponding order.

In modern language, Euclid is saying that iff for all natural numbers and all , the statements and are either both true or both false.

We now quote from Arthan:

*“According to the commentary in Heath’s translation of the Elements, de Morgan gave an interesting rationale for this definition: imagine a fence with equally spaced railings in front of a colonnade of equally spaced columns: *

*Let the distance between the columns be and the distance between the railings be . If the construction is continued indefinitely, an observer can compare with to any degree of accuracy without making any measurements just by counting the columns and railings. For example, in the figure, the 6th railing lies between the 8th and 9th columns, so that , which means that lies between and . If more precision were required, the observer might continue counting to find that the 25th railing lies between the 35th and 36th columns and conclude that lies between and .*

*This picture suggests a way of representing real numbers: construct the colonnade so that the distance between the columns is 1 and represent by the sequence of integers in which the m-th term, say, gives the number of columns to the left of or in line with the m-th railing in the figure. This sequence will represent . In the example in figure 1, the first few are 1,2,4,5,7,8,…. Since Euclid’s book V is generally believed to describe the work of Eudoxus, let us call the real numbers represented in this way the Eudoxus reals.”*

**A proof that **

Following this paper by Theo Grundhöfer, we prove that and are isomorphic as rings (and therefore as fields).

**Proof: **Let . Then there exists such that for all . One sees easily by induction on that . It follows that and thus

for all .

Therefore is a Cauchy sequence, so it has a limit in . The resulting map is easily checked to be additive, and since it is also surjective. The kernel of clearly contains the subgroup of bounded functions. We claim that the kernel is precisely .

To see this, note first that by letting tend to infinity in and multiplying by we obtain

for all . If , then implies that is bounded on . Since by the near-endomorphism property of , it follows that is bounded on all of as claimed. Thus induces an isomorphism of additive groups .

It remains to show that is multiplicative, i.e., that for all . This is clear if , since in that case and thus as well. So we may assume that . In this case, shows that if and if . In the former case, we can write

and we’re done. In the latter case, since is bounded we have and the result follows from a similar calculation. Q.E.D.

**Constructing the real numbers**

The proof we have just given presupposes the existence of the real numbers as a completion of . It is natural to wonder if one could define the real numbers as and deduce from first principles that is a complete ordered field, without referencing another construction of . This can in fact be done; see the papers by Arthan and A’Campo.

**Concluding remarks**

(1) The triple defined above is not quite a ring, since in general (it’s what’s called a near-ring). This is why we first defined the isomorphism at the level of abelian groups and then proved that it’s multiplicative. The distributive law **does** hold once we take the quotient by ; this can be proved directly but it also follows from the argument above.

(2) According to Arthan, the definition of the Eudoxus reals is due to Stephen Schanuel and dates back to the early 1980s. As Arthan writes, “He named the resulting development of the real numbers after Eudoxus, since it seemed to reflect the relationship between the discrete and the continuous apparent in the ancient theory of proportion. Schanuel communicated the idea to many people, but did not publish it.” It seems that the idea was independently discovered by Norbert A’Campo in 2003. In this paper, Ross Street mentions another independent discovery of the Eudoxus reals by Richard Lewis.

(3) The definition of and the isomorphism are reminiscent of John Tate’s definition of the canonical height on an elliptic curve (and more generally the Call-Silverman construction of canonical heights attached to polarized dynamical systems).

(4) Steffen Kionke has shown that equivalence classes of near-endomorphisms, in the spirit of the Eudoxus reals, can be used to construct the completion of any field with respect to an absolute value.

(5) T.J.D. Hermans has explored various generalizations of the Eudoxus reals and proves some interesting results. Given an abelian group , define to be the set of functions such that is finite modulo functions for which is finite. When endowed with the operations of pointwise addition and composition. becomes a ring. Hermans proves that is isomorphic to the field of -adic numbers, and is isomorphic to the rational adèle ring .

(6) For a survey of numerous different constructions of the real numbers, including the Eudoxus reals, see this paper by Ittay Weiss.

Keith Conrad points out the following comments by Steve Schanuel which have been preserved via the Wayback Machine: https://web.archive.org/web/20160424155910/facultypages.ecc.edu/alsani/ct99-00(8-12)/msg00073.html

This provides some depth and color to my remark about the Eudoxus reals being reminiscent of Tate’s work on canonical heights.

And Ben Steinberg notes the following related post: https://ncatlab.org/nlab/show/Eudoxus+real+number?fbclid=IwAR2dsRrUpwlC5gZM-Qla5ZEJTmcXZd1J3y0V53Yg2LvrsL0Lh6Y0Rkj7918