The Foundations of Analysis

arXiv:1303.6576v1 [math.HO] 26 Mar 2013

Larry Clifton
larry@cliftonlabs.net
March 22, 2013

Contents
1 Introduction 2

2 Magnitude Spaces 2

3 Functions 4

4 Basic Equalities and Inequalities 5

5 Magnitude Space Embeddings 8

6 Classification of Magnitude Spaces 10

7 Well Ordered Magnitude Spaces 12

8 Natural Numbers and Integral Multiples 16

9 Embeddings and Ratios 17

10 Classical Theory of Ratios 19

11 Embeddings and the Fourth Proportional 27

12 Continuous Magnitude Spaces 29

13 Real Numbers 35

1
14 Magnitude Embedding Spaces 36

15 Magnitude Endomorphism Spaces 38

16 Generalized Multiple 40

17 Generalized Product Operator 42

18 Symmetric Magnitude Spaces 44

19 Power Functions 45

1 Introduction
Here we introduce the real numbers. On the one hand, this is a modern
introduction based on morphisms between objects in an algebraic category.
On the other hand, it is an ancient introduction with 24 of the theorems
dating back to about 300 B.C. It is hoped that the complete and elementary
nature of this work will show that it is practical to introduce the real numbers
from a categorical perspective to students who may never study abstract
algebra, but will use real numbers on a regular basis in their future course
work and professional careers.
In this endeavor we have been inspired by Landau’s little book Grundla-
gen der Analysis[1] from which we have appropriated the title of the present
work.

2 Magnitude Spaces
Definition 2.1 A magnitude space is a nonempty set M together with
a binary operation on M, which we usually denote by +, such that for any
a, b, c ∈ M
(i)(associativity) a + (b + c) = (a + b) + c,
(ii)(commutativity) a + b = b + a, and
(iii)(trichotomy) exactly one of the following is true: a = b + d for some
d ∈ M, or a = b, or b = a + d for some d ∈ M.

Definition 2.2 (The whole is greater than the part.) If a and b are el-
ements of a magnitude space M and b = a + d for some d ∈ M, we say a is

2
less than b and write a < b or equivalently we say b is greater than a and
write b > a.

Remark 2.1 Since a and b are smaller than a + b and hence not equal to
a + b, a magnitude space does not have an additive identity or zero element.

The concept of magnitude spaces has a long history. That the concept
was recognized as a formal abstraction before 300 B.C. is evidenced by the
following two quotes. The first is Proposition 16 from Book V of Euclid’s
Elements.

Given–Four proportionate magnitudes a, b, c, d with a to b

the same as c to d.
To be Shown–The alternates will also be a proportion, a to
c the same as b to d.

The second, dating from a generation or two before Euclid, is a comment

by Aristotle (Posterior Analytics I, 5, 20).

Alternation used to be demonstrated separately of numbers,

lines, solids, and time intervals, though it could have been proved
of them all by a single demonstration. Because there was no single
name to denote that in which numbers, lines, solids, and time
intervals are identical, and because they differed specifically from
one another, this property was proved of each of them separately.
Today, however, the proof is commensurately universal, for they
do not possess this attribute as numbers, lines, solids, and time
intervals, but as manifesting this generic character which they are
postulated as possessing universally.

We do not have an explicit list of the properties which the various kinds
of magnitudes are postulated as possessing universally from this time period.
All we have are a number of “Common Notions” such as “the whole is greater
than the part” and “if equals are subtracted from equals, the remainders
will be equal.” The only complete work on proportions surviving from this
period is Book V of Euclid’s Elements. Here we find rigorous reasoning
without an explicit foundation. We do not know if there was an explicit
foundation generally known at the time. What we do know is that, starting
with the definition of magnitude spaces above, we can establish a foundation

3
by which the propositions in Book V of Euclid can be proved in accordance
with present day standards of rigor.[2] And from the propositions in Book
V of Euclid there follows the theory establishing the real numbers in the
modern categorical sense.

3 Functions
Here we review function terminology and a few basic theorems.
Definition 3.1 To indicate that ϕ is a function from a set S into a set S ′ ,
we write ϕ : S → S ′ . We also refer to functions as mappings and if ϕa = b
we say that ϕ maps a into b.
Definition 3.2 Two functions ϕ and ψ from a set S into a set S ′ are equal
if for all a ∈ S, ϕa = ψa. And in this case we write ϕ = ψ.
Definition 3.3 A function ϕ : S → S ′ is one-to-one if for all a, b ∈ S,
a 6= b implies ϕa 6= ϕb.
Definition 3.4 A function ϕ : S → S ′ is onto if for each a′ ∈ S ′ there is
some a ∈ S such that ϕa = a′ . And in this case we say that ϕ maps S onto
S ′.
Definition 3.5 If ϕ1 : S1 → S2 and χ : S2 → S3 are functions, then we
define the composition χ ◦ ϕ : S1 → S3 by (χ ◦ ϕ) a = χ (ϕa) for all a ∈ S1 .
In most cases, we omit the ◦ symbol between the functions and define the
composition χϕ : S1 → S3 by (χϕ) a = χ (ϕa)
Definition 3.6 The identity function iS : S → S maps every element of
a set S to itself. I.e. for all a ∈ S, iS a = a.
Theorem 3.1 Composition of functions is associative
Proof Assume S, S ′ , S ′′ and ϕ : S → S ′ , χ : S ′ → S ′′ , ψ : S ′′ → S ′′′ are
functions. If a ∈ S then
(ψ (χϕ)) a = ψ ((χϕ) a) (Definition 3.5)
= ψ (χ (ϕa)) (Definition 3.5)
= (ψχ) (ϕa) (Definition 3.5)
= ((ψχ) ϕ) a (Definition 3.5)
and therefore ψ (χϕ) = (ψχ) ϕ by Definition 3.2.

4
Theorem 3.2 The identity function iS : S → S is one-to-one and onto.

Proof Suppose a, b ∈ S and a 6= b. Now iS a = a and iS b = b by Definition

3.6 and hence iS a 6= iS b. Therefore iS is one-to one according to Definition
3.3.
And for any a ∈ S there is some c ∈ S (namely c = a) such that iS c = a
by Definition 3.6. Therefore iS is onto according to Definition 3.4.

Theorem 3.3 If ϕ : S → S ′ , ψ : S ′ → S, and ψϕ = iS , then ψ is onto.

Proof For any a ∈ S

a = iS a (Definition 3.6)
= (ψϕ) a (Definition 3.2)
= ψ (ϕa) . (Definition 3.5)

Thus there is an element in S ′ (namely ϕa) which ψ maps into a. Therefore

ψ is onto according to Definition 3.4.

Remark 3.1 In the preceding theorem, we could also conclude that ϕ is one-
to-one.

4 Basic Equalities and Inequalities

Definition 4.1 A binary relation < on a set S is trichotomous if for all
a, b ∈ S, exactly one of the following is true: b < a, or a = b, or a < b.

Definition 4.2 A binary relation < on a set S is transitive if for all

a, b, c ∈ S, a < b and b < c implies a < c.

Definition 4.3 A binary relation is a strict linear order if it is trichoto-

mous and transitive.

Definition 4.4 Binary relations < and > on a set S are inverses (to each
other) if for all a, b ∈ S, a < b if and only if b > a.

Theorem 4.1 If S and S ′ are sets with trichotomous relations < (and in-
verse relations >) and ϕ : S → S ′ is a function such that for all a, b ∈ S,
a < b implies ϕa < ϕb, then for all a, b ∈ S, ϕa has to ϕb the same relation
(<, =, or >) as a has to b, and ϕ is one-to-one.

5
Proof For a, b ∈ S the three mutually exclusive cases

a < b, or a = b, or a > b

imply, by assumption, the three mutually exclusive cases

ϕa < ϕb, or ϕa = ϕb, or ϕa > ϕb

respectively. The three converse implications follow from trichotomy. For

instance, assume ϕa < ϕb. If a = b, then ϕa = ϕb which contradicts the
assumption. If a > b, then ϕa > ϕb which also contradicts the assumption.
Since a = b and a > b are incompatible with our assumption, a < b by
trichotomy. Thus ϕa < ϕb implies a < b.
I say ϕ is one-to-one. For if a, b ∈ S and a 6= b, then a < b or a > b by
trichotomy and hence ϕa < ϕb or ϕb < ϕa and hence ϕa 6= ϕb by trichotomy.
We have now shown that if a 6= b then ϕa 6= ϕb. Therefore ϕ is one-to-one
according to Definition 3.3.

In the remainder of this section lower case variables a, b, c, and d are

elements of a magnitude space M.

Theorem 4.2 The < relation in a magnitude space is trichotomous.

Proof From Definition 2.1, exactly one of the following is true: a = b + d

for some d ∈ M, a = b, or b = a + d for some d ∈ M. From Definition 2.2,
exactly one of the following is true: b < a, or a = b, or a < b. Therefore the
< relation in a magnitude space is trichotomous according to Definition 4.1.

Theorem 4.3 (Translation Invariance) If b < c, then a + b < a + c.

Proof If b < c, then c = b + d for some d by Definition 2.2. Hence

a + c = a + (b + d) = (a + b) + d

by Definition 2.1 (associativity) and therefore a + b < a + c according to

Definition 2.2.

Theorem 4.4 (If equals are added to equals or unequals ...) a+b has
to a + c the same relation (<, =, or >) as b has to c.

6
Proof Fix a and define a function ϕ : M → M by ϕb = a + b. If b < c, then
ϕb < ϕc by the preceding theorem. Therefore ϕb = a + b has to ϕc = a + c
the same relation (<, =, or >) as b has to c by Theorem 4.1.

Remark 4.1 If a < b, then, from Definition 2.2, there is some d such that
b = a + d. In fact there is only one such element. For if a + d = a + d′ , then
d = d′ by the preceding theorem.

Definition 4.5 If a < b, we define b − a to be the unique element d such

that b = a + d.

Theorem 4.5 If a < b, then b − a < b.

Proof Definitions 2.2 and 4.5

Remark 4.2 If a < b, then (b − a) + a = a + (b − a) = b since (b − a) is, by

definition, the unique element which when added to a yields b.

Theorem 4.6 (Transitivity of <) If a < b and b < c, then a < c and
c − a = (c − b) + (b − a).

Proof If a < b and b < c, then c = (c − b) + b = (c − b) + (b − a) + a by

Definition 4.5. Hence a < c by Definition 2.2 and c − a = (c − b) + (b − a)
according to Definition 4.5.

Theorem 4.7 The < relation in a magnitude space is a strict linear order.

Proof The < relation in a magnitude space is trichotomous by Theorem 4.2.

And it is transitive by the preceding theorem and Definition 4.2. Therefore
the < relation in a magnitude space is a strict linear order according to
Definition 4.3.

Theorem 4.8 (If equals are subtracted from equals or unequals ...)
If b > a and c > a, then b − a has to c − a the same relation (<, =, or >)
as b has to c.

Proof a + (b − a) has to a + (c − a) the same relation (<, =, or >) as (b − a)

has to (c − a) by Theorem 4.4. But a + (b − a) = b and a + (c − a) = c by
Definition 4.5. Therefore b − a has to c − a the same relation (<, =, or >)
as b has to c.

7
Theorem 4.9 If a > b, then a has to b + c the same relation (<, =, or >)
as a − b has to c.

Proof b + (a − b) has to b + c the same relation (<, =, or >) as a − b has to

c by Theorem 4.4. And a = b + (a − b) by Definition 4.5. Therefore a has to
b + c the same relation (<, =, or >) as a − b has to c.

5 Magnitude Space Embeddings

In this section M, M ′ , and M ′′ are magnitude spaces.

Definition 5.1 A mapping ϕ : M → M ′ is a homomorphism if ϕ (a + b) =

ϕa + ϕb for all a, b ∈ M. A homomorphism which is one-to-one is an em-
bedding.

Definition 5.2 An embedding ϕ : M → M ′ which is onto as a mapping is

an isomorphism. If there is an isomorphism from M onto M ′ we say M
is isomorphic to M ′ .

Definition 5.3 A homomorphism ϕ : M → M of a magnitude space into

itself is an endomorphism and an endomorphism which is one-to-one and
onto as a map is an automorphism.

Definition 5.4 If ϕ : M → M ′ and χ : M → M ′ are two functions, their

sum is the function (ϕ + χ) : M → M ′ defined by (ϕ + χ) a = ϕa + χa for
all a ∈ M.

The next two theorems show that homomorphisms between magnitude

spaces are always embeddings.

Theorem 5.1 If ϕ : M → M ′ is a homomorphism and a < b, then

ϕa < ϕb and ϕb − ϕa = ϕ (b − a) .

Proof If ϕ is a homomorphism and a < b, then

ϕb = ϕ (a + (b − a)) (Definition 4.5)

= ϕa + ϕ (b − a) . (Definition 5.1)

Hence ϕa < ϕb according to Definition 2.2 and ϕb−ϕa = ϕ (b − a) according

to Definition 4.5.

8
Theorem 5.2 If ϕ : M → M ′ is a homomorphism, then ϕa has to ϕb the
same relation (<, =, or >) as a has to b and ϕ is an embedding.

Proof From the preceding theorem, a < b implies ϕa < ϕb. Hence ϕa has
to ϕb the same relation (<, =, or >) as a has to b and ϕ is one-to-one by
Theorem 4.1. And since ϕ is a homomorphism and is one-to-one, ϕ is an
embedding according to Definition 5.1.

Theorem 5.3 The identity function iM in a magnitude space M is an au-

tomorphism.

Proof For any a, b ∈ M, iM (a + b) = a + b = iM a + iM b by Definition 3.6.

Hence iM is a homomorphism according to Definition 5.1. And iM is one-to-
one and onto by Theorem 3.2. Therefore iM is an automorphism according
to Definitions 5.1 and 5.3.

Theorem 5.4 The sum of two embeddings is an embedding.

Proof If ϕ : M → M ′ and χ : M → M ′ are embeddings and a, b ∈ M, then

(ϕ + χ) (a + b) = ϕ (a + b) + χ (a + b) (Definition 5.4)
= (ϕa + ϕb) + (χa + χb) (Definition 5.1)
= (ϕa + χa) + (ϕb + χb) (commutativity and associativity of +)
= (ϕ + χ) a + (ϕ + χ) b. (Definition 5.4)

Therefore ϕ + χ is an embedding by Definition 5.1 and Theorem 5.2.

Theorem 5.5 The composition of two embeddings is an embedding.

Proof If ϕ : M → M ′ and χ : M ′ → M ′′ are embeddings and a, b ∈ M, then

(χϕ) (a + b) = χ (ϕ (a + b)) (Definition 3.5)

= χ (ϕa + ϕb) (Definition 5.1)
= χ (ϕa) + χ (ϕb) (Definition 5.1)
= (χϕ) a + (χϕ) b. (Definition 3.5)

Therefore ϕχ is an embedding by Definition 5.1 and Theorem 5.2.

9
6 Classification of Magnitude Spaces
Definition 6.1 Let < be a strict linear order with inverse >. By a ≤ b we
shall mean a < b or a = b. By a ≥ b we shall mean a > b or a = b.

Definition 6.2 Let S be a set with a strict linear order < and let A be a
nonempty subset of S. We say b ∈ S is a lower bound of A if b ≤ a for
every a ∈ A. We say that b is a smallest or least element of A if b is a
lower bound of A and b ∈ A. We say b ∈ S is an upper bound of A if a ≤ b
for every a ∈ A. We say that b is a largest or greatest element of A if b is
an upper bound of A and b ∈ A.

Definition 6.3 A magnitude space is discrete if it has a smallest element;

otherwise it is nondiscrete.

Definition 6.4 A magnitude space M is well ordered if every nonempty

subset of M has a smallest element.

Definition 6.5 A magnitude space is complete if every nonempty subset

with an upper bound has a least upper bound or, in other words, the set of
upper bounds has a least element.

Definition 6.6 A magnitude space is continuous if it is complete and

nondiscrete.

Definition 6.7 (Hölder) A magnitude space M is Archimedean if for

any element a ∈ M and any nonempty subset A with an upper bound, there
is some element ζ ∈ M such that ζ ∈ A and ζ + a ∈
/ A.

Remark 6.1 The property of being Archimedean is usually defined in terms

of integral multiples.[3]

Theorem 6.1 If a ≤ b and b < c, then a < c. And if a < b and b ≤ c, then
a < c.

Proof Assume a ≤ b and b < c. Then a < b or a = b by Definition 6.1. If

a < b then since also b < c, a < c by Theorem 4.6. If a = b then since also
b < c, a < c. Thus in both cases a < c. The second part of the theorem
follows by similar reasoning.

10
Theorem 6.2 If an element is greater than an upper bound of a set, then it
is an upper bound of the set but not an element of the set.

Proof Let S be a linearly ordered set, A a nonempty subset of S, a an upper

bound of A, and a < b. I say b is an upper bound of A and b ∈/ A.
If c ∈ A, then c ≤ a by Definition 6.2. And a < b by assumption. Thus
c ≤ b by Theorem 6.1 and Definition 6.1. We have now shown that if c ∈ A,
then c ≤ b. Therefore b is an upper bound of A.
Now suppose b ∈ A. Then b ≤ a by Definition 6.2. But this contradicts
the assumption a < b by Definition 6.1 and Theorem 4.2. Therefore b ∈ / A.

Theorem 6.3 If an element is less than a lower bound of a set, then it is a

lower bound of the set but not an element of the set.

Proof Similar to proof of previous theorem.

Theorem 6.4 If M is a discrete magnitude space with smallest element a,

and b ∈ M, then there is no c ∈ M such that b < c < b + a.

Proof Suppose there is such a c. Then (b + a) − b = ((b + a) − c) + (c − b)

by Theorem 4.6. But (b + a) − b = a by Definition 4.5. Hence c − b < a by
Definition 2.2. But a is the smallest element of M by assumption and so we
have a contradiction.

Theorem 6.5 A well ordered magnitude space is discrete and complete.

Proof Let M be a well ordered magnitude space. Since M is a subset of

M, M has a smallest element by Definition 6.4. Therefore M is discrete
according to Definition 6.3.
Now let A be any nonempty subset of M with an upper bound. If B is
the set of upper bounds of A, then B is a nonempty subset of M and hence
B has a least element by Definition 6.4. Therefore M is complete according
to Definition 6.5.

Theorem 6.6 (Hölder) A complete magnitude space is Archimedean.

11
Proof Let M be a complete magnitude space, a ∈ M, and A a nonempty
subset of M with an upper bound. Since M is complete, A has a least upper
bound ζ ′ by Definition 6.5.
Case 1: ζ ′ ≤ a. Since Let ζ be any element of A. Then a < ζ + a by
Definition 2.2 and so ζ ′ < ζ + a by Theorem 6.1. But then ζ + a is greater
than an upper bound of A and hence ζ + a ∈ / A by Theorem 6.2.
Case 2: a < ζ . Since ζ − a < ζ by Theorem 4.5, ζ ′ − a is less than the
′ ′ ′

least upper bound of A and hence is not an upper bound of A by Theorem

6.3. Thus there is some ζ ∈ A such that ζ ′ − a < ζ by Definition 6.2. And
ζ ′ < ζ + a by Theorem 4.9 and so ζ + a is greater than an upper bound of A
and hence ζ + a ∈ / A by Theorem 6.2.
We have shown, in both cases, that there is some ζ ∈ M such that ζ ∈ A
and ζ + a ∈ / A. Therefore M is Archimedean according to Definition 6.7.

Theorem 6.7 A discrete Archimedean magnitude space is well ordered.

Proof Assume M is a discrete Archimedean magnitude space with smallest

element a and A is a nonempty subset of M. I say A has a smallest element.
Let B be the set of all lower bounds of A. If b ∈ A, then a ≤ b by
Definition 6.2. Therefore a ∈ B and so B is nonempty. Now let c be any
fixed element of A. If b ∈ B, then b ≤ c. Thus c is an upper bound of B
according to Definition 6.2 and so B is a nonempty subset of M with an
upper bound.
But M is Archimedean and hence there is some element ζ ∈ M such that
ζ ∈ B and ζ + a ∈ / B by Definition 6.7. Or, in other words, ζ is a lower
bound of A and ζ + a is not a lower bound of A. And because ζ + a is not
a lower bound of A, there is some b ∈ A such that b < ζ + a by Definitions
6.2 and 6.1 and trichotomy. And since ζ is a lower bound of A, ζ ≤ b by
Definition 6.2. But ζ < b and b < ζ + a is impossible by Theorem 6.4. Thus
ζ = b by Definition 6.1. Therefore ζ ∈ A and ζ is a lower bound of A and
hence ζ is the smallest element of A according to Definition 6.2.

7 Well Ordered Magnitude Spaces

We next prove a form of mathematical induction for well ordered magnitude
spaces.

12
Theorem 7.1 If M is a well ordered magnitude space with smallest element
a, and A is subset of M containing a such that c ∈ A implies c + a ∈ A, then
A = M.

Proof Suppose A 6= M. Then the set B consisting of those elements in

M which are not in A is nonempty and hence has a smallest element by
Definition 6.4. Let b be the smallest element of B. Then b is a lower bound
of B and b ∈ B by Definition 6.2. Now a is the smallest element of M by
assumption and hence a ≤ b by Definition 6.2. And b is not equal to a since
a ∈ A by assumption and b ∈ B, a set having no element in A. And from
a ≤ b and a 6= b follows a < b by Definition 6.1. But then b − a < b by
Theorem 4.5. Thus b − a is less than a lower bound of B and hence b − a ∈/B
by Theorem 6.3. Hence b − a ∈ A. But, by assumption, b − a ∈ A implies
(b − a) + a ∈ A. And (b − a) + a = b by Definition 4.5. Thus b is an element
of A and of B which is impossible.
In the following theorem an embedding of a well ordered magnitude space
into an arbitrary magnitude space is constructed inductively. The general
approach is that of Dedekind.[4]

Theorem 7.2 If M is a well ordered magnitude space with smallest element

a, M ′ is an arbitrary magnitude space, and a′ ∈ M ′ , then there exists a
unique function ϕ : M → M ′ such that
(i) ϕa = a′ and
(ii) ϕb = ϕ (b − a) + a′ for all b > a.

Proof First, there can be at most one function satisfying the two conditions
above. For if there are two distinct functions ϕ and ψ each satisfying the
two conditions, then there must be a smallest b ∈ M for which ϕb 6= ψb by
Definition 6.4. Now ϕa = a′ = ψa since each of ϕ and ψ are assumed to have
property (i) above. Thus a 6= b since ϕb 6= ψb. And a is the smallest element
of M by assumption and hence a ≤ b by Definition 6.2. And from a 6= b and
a ≤ b follows a < b by Definition 6.1. Hence b − a < b by Theorem 4.5. And
since b is the smallest element of M such that ϕb 6= ψb, ϕ (b − a) = ψ (b − a).
Therefore

ϕb = ϕ (b − a) + a′ (property (ii) above)

= ψ (b − a) + a′ (Theorem 4.4)
= ψb (property (ii) above)

13
which is a contradiction.
It remains to show that there exists a function ϕ with the specified prop-
erties. To this end, for each b ∈ M let Mb be the set of those elements in M
which are less than or equal to b. I say that for each b ∈ M there is a unique
function ϕb : Mb → M ′ such that
(i) ϕb a = a′ and
(ii) ϕb c = ϕb (c − a) + a′ for a < c ≤ b.
That there cannot be two distinct functions with these properties for a
given b ∈ M can be shown by the same argument as given in the beginning
of the proof. To prove the existence of one such function for each b ∈ M,
let A be the set of all elements b in M for which there is a unique function
ϕb : Mb → M ′ as described above. In the case of b = a, Ma = {a} and the
function ϕa : Ma → M ′ defined by ϕa a = a′ has the required properties. Thus
a ∈ A. Now suppose b ∈ A. We can then define a function ϕb+a : Mb+a → M ′
in terms of the unique function ϕb : Mb → M ′ according to

ϕb c if c≤b
ϕb+a c = .
ϕb b + a′ if c = b + a

And since b ∈ A, (i) ϕb+a a = a′ and (ii) ϕb+a c = ϕb+a (c − a) + a′ for all
c ∈ Mb+a . We have now shown that if b ∈ A, then b + a ∈ A. Hence A = M
by the preceding theorem and for each b ∈ M there exists a unique function
ϕb satisfying the two conditions above.
Now define the function ϕ : M → S according to ϕb = ϕb b. We then
have
ϕa = ϕa a = a′
and for any b > a

ϕb = ϕb b = ϕb (b − a) + a′ = ϕb−a (b − a) + a′ = ϕ (b − a) + a′ .

Theorem 7.3 If M is a well ordered magnitude space, M ′ is any magnitude

space, a is the smallest element in M, and a′ is any element in M ′ , then
there exists a unique embedding ϕ : M → M ′ such that ϕa = a′ .

Proof From the preceding theorem there is a unique function ϕ : M → M ′

such that (i) ϕa = a′ and (ii) ϕb = ϕ (b − a) + a′ for any b > a. Note that

14
from the second property it follows that
ϕ (b + a) = ϕ ((b + a) − a) + a′ (definition of ϕ)
= ϕb + a′ . (Definition 4.5)
for any b ∈ M since b + a > a.
I say that ϕ is an embedding. To see this, fix c ∈ M and define A to be
the set of all elements d in M such that ϕ (c + d) = ϕc + ϕd. Now a ∈ A
since ϕ (c + a) = ϕc + a′ = ϕc + ϕa by note above. And if d ∈ A, then
ϕ (c + (d + a)) = ϕ ((c + d) + a) (associativity)
= ϕ (c + d) + a′ (note above)
= ϕc + ϕd + a′ (d ∈ A)
= ϕc + ϕ (d + a) (note above)
and so d + a ∈ A. Therefore, A = M by Theorem 7.1. And since our choice
of c ∈ M was arbitrary, ϕ (c + d) = ϕc + ϕd for all c, d ∈ M. Thus ϕ is an
embedding by Definition 5.1 and Theorem 5.2.
Now suppose ψ is some other embedding of M into M ′ which maps a into
a . Then ψa = a′ and, if b > a then
′

ψb = ψ ((b − a) + a) (Definition 4.5)

= ψ (b − a) + ψa (Definition 5.1)
= ψ (b − a) + a′ . (assumed property of ϕ)
But by the preceding theorem, there is only one such mapping and hence
ψ = ϕ.
Theorem 7.4 Any two well ordered magnitude spaces are isomorphic.
Proof Assume that M and M ′ are well ordered magnitude spaces with
smallest elements a and a′ respectively. There is an embedding ϕ : M → M ′
such that ϕa = a′ and an embedding ϕ′ : M ′ → M such that ϕ′ a′ = a from
the preceding theorem. The composition ϕϕ′ : M ′ → M ′ is an embedding by
Theorem 5.5 and the identity mapping iM ′ : M ′ → M ′ is also an embedding
by Theorem 5.3. But
(ϕϕ′ ) a′ = ϕ (ϕ′ a′ ) = ϕa = a′ = iM ′ a′
and so ϕϕ′ = iM ′ by the preceding theorem. Therefore ϕ is onto by Theorem
3.3 and hence is an isomorphism according to Definition 5.2.

15
8 Natural Numbers and Integral Multiples
Definition 8.1 Pick any well ordered magnitude space and denote it by N
and denote the smallest element of N by 1. Since all well ordered magnitude
spaces are isomorphic (Theorem 7.4) and we will use only the algebraic prop-
erties that the well ordered magnitude spaces have in common, it does not
matter which well ordered magnitude space is chosen to play the role of the
“number system” N. We call N the natural numbers.

Definition 8.2 If n ∈ N and a is an element of a magnitude space M, then

the integral multiple na is given by na = ϕ1,a n where ϕ1,a is the unique
embedding of N into M which maps 1 into a (Theorem 7.3).

Theorem 8.1 For each a ∈ M, (n + 1) a = na + a.

Proof Let ϕ1,a be the unique embedding of N into M which maps 1 into a.
Then

(n + 1) a = ϕ1,a (n + 1) (Definition 8.2)

= ϕ1,a n + ϕ1,a 1 (Definition 5.1)
= na + a. (Definition 8.2)

Remark 8.1 For each a in a magnitude space M

1a = ϕ1,a 1 = a
(1 + 1) a = ϕ1,a (1 + 1) = ϕ1,a 1 + a = a + a
(1 + 1 + 1) a = ϕ1,a (1 + 1 + 1) = ϕ1,a (1 + 1) + a = a + a + a
..
.

Theorem 8.2 If χ : M → M ′ is an embedding, then χ (na) = n (χa) for

every a ∈ M and n ∈ N.

Proof Let a be fixed. There exists a unique embedding ϕ1,a of N into M

which maps 1 into a and a unique embedding ϕ1,χa of N into M ′ which maps

16
1 into χa by Theorem 7.3. Now χϕ1,a : N → M ′ is an embedding by Theorem
5.5 and

(χϕ1,a ) 1 = χ (ϕ1,a 1) (Definition 3.5)

= χ (a) . (definition of ϕ1,a )

Therefore each of ϕ1,χa and χϕ1,a are embeddings of N into M ′ which map
1 (the smallest element in N) into the same element χa ∈ M ′ and hence
ϕ1,χa = χϕ1,a by Theorem 7.3. Therefore

n (χa) = ϕ1,χa n (Definition 8.2)

= (χϕ1,a ) n (Definition 3.2)
= χ (ϕ1,a n) (Definition 3.5)
= χ (na) . (Definition 8.2)

Theorem 8.3 If ϕ : M → M ′ is an embedding and a, b ∈ M, then for each

pair m, n ∈ N, m (ϕa) has to n (ϕb) the same relation (<, =, or >) as ma
has to nb.

Proof First, m (ϕa) = ϕ (ma) and n (ϕb) = ϕ (nb) by the preceding theorem.
Second, ϕ (ma) has to ϕ (nb) the same relation (<, =, or >) as ma has to nb
by Theorem 5.2. Therefore m (ϕa) has to n (ϕb) the same relation (<, =, or
>) as ma has to nb.

9 Embeddings and Ratios

The preceding theorem is the connecting point between modern algebraic
definitions of number systems and the classical theory of ratios. The classical
theory is based on the following two definitions.

Definition 9.1 (Euclid V, Definition 4) Two elements a, b in a magni-

tude space are said to have a ratio if there are natural numbers m, n such
that ma > b and nb > a.

Definition 9.2 (Euclid V, Definitions 5, 6, and 7) Let M and M ′ be

magnitude spaces. We say that a pair a, b ∈ M has the same ratio as

17
(or are proportional to) a pair a′ , b′ ∈ M ′ if ma has to nb the same rela-
tion (<, =, or >) as ma′ has to nb′ for every m, n ∈ N. And in this case we
write a : b = a′ : b′ . And if for some m, n ∈ N, ma > nb and ma′ ≤ nb′ , then
we say a has to b a greater ratio than a′ has to b′ and we write a : b > a′ : b′
or a′ : b′ < a : b.

We are particularly interested in magnitude spaces in which every pair

of elements have a ratio. The next two theorems establish that a magnitude
space has this property if and only if it is an Archimedean magnitude space.

Theorem 9.1 If every pair of elements of a magnitude space have a ratio,

then the magnitude space is Archimedean.

Proof Assume M is a magnitude space in which every pair of elements have

a ratio. Let a ∈ M and let A be a nonempty subset of M which has an upper
bound. I say there is some ζ ∈ M such that ζ ∈ A and ζ + a ∈ / A.
Let B be the set of all natural numbers n such that na is an upper bound
of A. If b is any upper bound of A, then there is some natural number n such
that na > b because a and b have a ratio by assumption and Definition 9.1.
Thus B is nonempty. And the natural numbers are a well ordered magnitude
space by Definition 8.1. Hence B has a smallest element by Definition 6.4.
Let n be the smallest element of B.
Case 1: n = 1. Pick any ζ ∈ A. Then na < ζ + a.
Case 2: n > 1. Since n is the smallest natural number such that na is an
upper bound of A and n − 1 < n by Theorem 4.5, it follows that (n − 1) a
is not an upper bound of A. Therefore there must be some ζ ∈ A such that
(n − 1) a < ζ by Definition 6.2. Thus (n − 1) a + a < ζ + a by Theorem 4.3
and

(n − 1) a + a = ((n − 1) + 1) a (Theorem 8.1)

= na. (Definition 4.5)

Therefore na < ζ + a.
In both cases, ζ ∈ A and ζ + a is greater than an upper bound of A and
hence ζ + a ∈
/ A by Theorem 6.2. Therefore M is an Archimedean magnitude
space according to Definition 6.7.

Theorem 9.2 If a and b are elements of an Archimedean magnitude space,

then there is some natural number n such that na > b.

18
Proof Suppose na ≤ b for all natural numbers n. Let A be the set of all
integral multiples of a. Then b is an upper bound of A by Definition 6.2.
Thus there is some element ζ ∈ M such that ζ ∈ A and ζ + a ∈ / A by
Definition 6.7. But if ζ ∈ A, then ζ = na for some natural number n and
ζ + a = na + a. And na + a = (n + 1) a by Theorem 8.1. Hence ζ + a is an
integral multiple of a and hence ζ + a ∈ A which is a contradiction.

Theorem 9.3 Any two elements of an Archimedean magnitude space have

a ratio.

Proof Previous theorem and Definition 9.1.

Theorem 9.4 If M and M ′ are Archimedean magnitude spaces and ϕ :

M → M ′ is an embedding, then ϕa : ϕb = a : b for every a, b ∈ M.

Proof Any two elements of M have a ratio from the previous theorem. And
likewise any two elements of M ′ have a ratio. Fix a, b ∈ M. For any two
natural numbers m and n, m (ϕa) has to n (ϕb) the same relation (<, =, or
>) as ma has to nb by Theorem 8.3. Therefore ϕa : ϕb = a : b according to
Definition 9.2.

10 Classical Theory of Ratios

Henceforth we shall consider only Archimedean magnitude spaces.
In this section variables a, b, c, d, e, f are all elements of a magnitude space
M, a′ , b′ , c′ , d′, e′ , f ′ are all elements of a magnitude space M ′ , and so on.
Variables j, k, m, n are natural numbers.
The propositions in this section appear in the exact order as the propo-
sitions in Book V of The Elements. The proofs are likewise similar with the
following exceptions.

1. Proofs for Propositions 1, 2, and 3 are based on Definition 8.2 and

Theorem 7.3.

2. The proofs of Propositions 10 and 18 follow those of Robert Simson.[5]

Proposition 1 n(a + b) = na + nb.

19
Proof Let a and b be fixed. There exists a unique embeddings ϕ1,a , ϕ1,b , and
ϕ1,a+b of N into M which map 1 into a, b, and a + b respectively by Theorem
7.3. Note that ϕ1,a + ϕ1,b is an embedding of N into M by Theorem 5.4 and

(ϕ1,a + ϕ1,b ) 1 = ϕ1,a 1 + ϕ1,b 1 (Definition 5.4)

= a + b. (definitions of ϕ1,a and ϕ1,b )

Thus ϕ1,a+b and ϕ1,a + ϕ1,b are each embeddings of N into M which map 1
into a + b and hence ϕ1,a+b = ϕ1,a + ϕ1,b by Theorem 7.3. Therefore

n(a + b) = ϕ1,a+b n (Definition 8.2)

= (ϕ1,a + ϕ1,b ) n (Definition 3.2)
= ϕ1,a n + ϕ1,b n (Definition 5.4)
= na + nb. (Definition 8.2)

Proposition 2 (m + n) a = ma + na.

Proof Let a be fixed. There exists a unique embedding ϕ1,a of N into M

which maps 1 into a by Theorem 7.3 and

(m + n) a = ϕ1,a (m + n) (Definition 8.2)

= ϕ1,a m + ϕ1,a n (Definition 5.1)
= ma + na. (Definition 8.2)

Proposition 3 (mn) a = m (na).

Proof Let a be fixed. There exists a unique embedding ϕ1,a of N into M

which maps 1 into a by Theorem 7.3 and

(mn) a = ϕ1,a (mn) (Definition 8.2)

= m (ϕ1,a n) (Theorem 8.2)
= m (na) . (Definition 8.2)

20
Proposition 4 If a : b = a′ : b′ , then ja : kb = ja′ : kb′ .

Proof If a : b = a′ : b′ , then for any two numbers m and n

m (ja) > n (kb) =⇒ (mj) a > (nk) b (Proposition 3)

=⇒ (mj) a′ > (nk) b′ (Definition 9.2)
=⇒ m (ja′ ) > n (kb′ ) . (Proposition 3).

And the same argument applies with > replaced by = or <. Therefore ja :
kb = ja′ : kb′ according to Definition 9.2.

Proposition 5 If a > b, then na > nb and na − nb = n (a − b). And, more

generally, na has to nb the same relation (<, =, or >) as a has to b.

Proof Fix n and let ϕ : M → M be the function defined by ϕa = na.

Then ϕ (a + b) = ϕa+ ϕb for any a and b by Proposition 1 and hence ϕ is
a homomorphism according to Definition 5.1. Hence Theorems 5.1 and 5.2
apply.

Proposition 6 If m > n, then ma > na and ma − na = (m − n)a. And,

more generally, ma has to na the same relation (<, =, or >) as m has to n.

Proof Fix a and let ϕ : N → M be the function defined by ϕn = na. Then

ϕ (m + n) = ϕm+ ϕn for any m and n by Proposition 2 and hence ϕ is
a homomorphism according to Definition 5.1. Hence Theorems 5.1 and 5.2
apply.

Proposition 7 If a = b, then a : c = b : c and c : a = c : b.

Proof If a = b, then ma = mb by Proposition 5. Thus for any m and n,

ma has the same relation to nc (<, =, or >) as mb has to nc. Therefore
a : c = b : c according to Definition 9.2. A similar argument shows that
c : a = c : b.

Proposition 8 If a > b, then a : c > b : c and c : b > c : a.

Proof Assume a > b. Since magnitude spaces are assumed to be Archimedean,

there is some m such that m (a − b) > c by Theorem 9.2. And from Propo-
sition 5, ma > mb and m (a − b) = ma − mb. Therefore ma − mb > c. And
ma > mb + c by Theorem 4.9.

21
 Likewise, the set A of all natural numbers k such that kc > mb is
nonempty by Theorem 9.2. And since the natural numbers are well ordered
by Definition 8.1, A has a smallest element by Definition 6.4. Let n be the
smallest element of A.
I say ma > nc. Suppose, on the contrary, that nc ≥ ma. Then, since it
was shown that ma > mb + c, it follows that nc > mb + c by Theorem 6.1
and

nc > mb + c =⇒ nc > mb + 1c (Definition 8.2)

=⇒ nc − 1c > mb (Theorem 4.9)
=⇒ (n − 1) c > mb (Proposition 5)
=⇒ (n − 1) ∈ A. (definition of A)

But (n − 1) < n by Theorem 4.5 and since n is the smallest element of A, n

is a lower bound of A by Definition 6.2, and hence (n − 1) ∈ / A by Theorem
6.3 which is a contradiction. Therefore, indeed, ma > nc.
And n was chosen so that nc > mb. Therefore, ma > nc and mb ≤ nc by
Definition 6.1 and hence a : c > b : c according to Definition 9.2.
Likewise, nc > mb and nc ≤ ma by Definition 6.1 and hence c : b > c : a
according to Definition 9.2.

Proposition 9 If a : c = b : c, then a = b. And if c : a = c : b then a = b.

Proof Assume a : c = b : c. If a > b, then a : c > b : c by the preceding

theorem. But this is not consistent with the assumption. And for the same
reason b > a is not consistent with the assumption. Therefore a = b. The
second part of the theorem is proved in a similar manner.

Proposition 10 If a : c > b : c, then a > b. And if c : a > c : b, then b > a.

Proof Assume a : c > b : c. Then there are numbers m and n such that
ma > nc and mb ≤ nc by Definition 9.2. And from ma > nc and nc ≥ mb
follows ma > mb by Theorem 6.1. Therefore, a > b by Proposition 5. The
second part of the theorem is proved in the same way.

Proposition 11 If a : b = a′ : b′ and a′′ : b′′ = a′ : b′ , then a : b = a′′ : b′′ .

22
Proof If a : b = a′ : b′ and a′′ : b′′ = a′ : b′ , and m and n are any two numbers

ma > nb =⇒ ma′ > nb′ (Definition 9.2)

=⇒ ma′′ > nb′′ . (Definition 9.2)

And the same argument applies with > replaced by = or <. Therefore a :
b = a′′ : b′′ by Definition 9.2.

Proposition 12 If a : b = c : d then a : b = (a + c) : (b + d).

Proof Assume a : b = c : d. If ma > nb, then mc > nd by Definition 9.2

and hence

m (a + c) = ma + mc (Proposition 1)
> nb + mc (Theorem 4.4)
> nb + nd (Theorem 4.4)
= n (b + d) . (Proposition 1)

Therefore ma > nb =⇒ m (a + c) > n (b + d). And the same argument

applies with > replaced by = or <. Therefore a : b = (a + c) : (b + d) by
Definition 9.2.

Proposition 13 If a : b = a′ : b′ and a′ : b′ > a′′ : b′′ , then a : b > a′′ : b′′ .

And if a : b = a′ : b′ and a′′ : b′′ > a′ : b′ , then a′′ : b′′ > a : b.

Proof Assume a : b = a′ : b′ and a′ : b′ > a′′ : b′′ . Then there are numbers
m and n such that ma′ > nb′ and ma′′ ≤ nb′′ by Definition 9.2. And
because ma′ > nb′ , also ma > nb by Definition 9.2. Therefore ma > nb and
ma′′ ≤ nb′′ and hence a : b > a′′ : b′′ by Definition 9.2. The second part of
the theorem is proved in a similar manner.

Proposition 14 If a : b = c : d, then a has the same relation to c (<, =, or

>) as b has to d.

Proof Assume a : b = c : d. Then

a > c =⇒ a : b > c : b (Proposition 8)

=⇒ c : d > c : b (Proposition 13)
=⇒ b > d (Proposition 10)

23
and by the same argument a < c =⇒ b < d. Likewise
a = c =⇒ a : b = c : b (Proposition 7)
=⇒ c : d = c : b (Proposition 11)
=⇒ b = d. (Proposition 9)

Proposition 15 a : b = ka : kb.
Proof The theorem is true for k = 1 since 1a = a and 1b = b by Definition
8.2. Now assume the theorem is true for k. Then a : b = ka : kb and hence
(ka + a) : (kb + b) = a : b by Proposition 12. But ka + a = (k + 1) a and
kb + b = (k + 1) b Theorem 8.1. Therefore(k + 1) a : (k + 1) b = a : b and the
theorem is true for k + 1. Therefore the theorem is true for all k by Theorem
7.1.
Proposition 16 If a : b = c : d, then a : c = b : d.
Proof Assume a : b = c : d and let m, n be any two natural numbers. Then
ma : mb = a : b and nc : nd = c : d.
by the preceding theorem and hence
ma : mb = nc : nd
by Proposition 11. Therefore ma has the same relation to nc (<, =, or >)
as mb has to nd by Proposition 14 and hence a : c = b : d by Definition 9.2.

Remark 10.1 An alternate formulation of the following theorem is that a :

b = a′ : b′ implies (a − b) : b = (a′ − b′ ) : b′
Proposition 17 If (a + b) : b = (a′ + b′ ) : b′ , then a : b = a′ : b′ .
Proof Assume (a + b) : b = (a′ + b′ ) : b′ . For any m and n
ma > nb =⇒ ma + mb > nb + mb (Theorem 4.4)
=⇒ m (a + b) > (n + m) b (Propositions 1 and 2)
=⇒ m (a′ + b′ ) > (n + m) b′ (Definition 9.2)
=⇒ ma′ + mb′ > nb′ + mb′ (Propositions 1 and 2)
=⇒ ma′ > nb′ . (Theorem 4.4)

24
And the same argument applies with > replaced by = or <. Therefore a :
b = a′ : b′ according to Definition 9.2.

Proposition 18 If a : b = a′ : b′ , then (a + b) : b = (a′ + b′ ) : b′ .

Proof Assume a : b = a′ : b′ and consider the integral multiples m (a + b)

and nb.
Now a + b > b by Definition 2.2 and hence m (a + b) > mb by Proposition
5.
If m = n then m (a + b) > nb. And if m > n, then mb > nb by Propo-
sition 6 and from m (a + b) > mb and mb > nb follows m (a + b) > nb by
Theorem 4.6. In summary, if m ≥ n, then m (a + b) > nb and also by the
same argument m (a′ + b′ ) > nb′ .
Now suppose m < n. In this case, mb < nb by Proposition 6. Thus

m (a + b) < nb =⇒ ma + mb < nb (Proposition 1)

=⇒ ma < nb − mb (Theorem 4.9)
=⇒ ma < (n − m) b (Proposition 6)
=⇒ ma′ < (n − m) b′ (Definition 9.2)
=⇒ ma′ < nb′ − mb′ (Proposition 6)
=⇒ ma′ + mb′ < nb′ (Theorem 4.9)
=⇒ m (a′ + b′ ) < nb′ . (Proposition 1)

And the same argument as above applies with < replaced with = or >.
We have now shown that for all m and n, m (a + b) has to nb the same re-
lation (<, =, or >) as m (a′ + b′ ) has to nb′ and hence (a + b) : b = (a′ + b′ ) :
b′ according to Definition 9.2.

Proposition 19 If (a + b) : (c + d) = a : c, then b : d = a : c.

Proof

(a + b) : (c + d) = a : c =⇒ (a + b) : a = (c + d) : c (Proposition 16)
=⇒ b : a = d : c (Proposition 17)
=⇒ b : d = a : c. (Proposition 16)

25
Proposition 20 If a : b = a′ : b′ and b : c = b′ : c′ , then a has to c the same
relation (<, =, or >) as a′ has to c′ .

Proof Assume a : b = a′ : b′ and b : c = b′ : c′ . Then c : b = c′ : b′ by

Definition 9.2 and hence

a > c =⇒ a : b > c : b (Proposition 8)

=⇒ a′ : b′ > c : b (Proposition 13)
=⇒ a′ : b′ > c′ : b′ (Proposition 13)
=⇒ a′ > c′ . (Proposition 10)

And similar arguments apply with > replaced by = or <.

Proposition 21 If a : b = b′ : c′ and b : c = a′ : b′ , then a has to c the same

relation (<, =, or >) as a′ has to c′ .

Proof Assume a : b = b′ : c′ and b : c = a′ : b′ . Then c : b = b′ : a′ by

Definition 9.2 and hence

a > c =⇒ a : b > c : b (Proposition 8)

=⇒ b′ : c′ > c : b (Proposition 13)
=⇒ b′ : c′ > b′ : a′ (Proposition 13)
=⇒ a′ > c′ . (Proposition 10)

And similar arguments apply with > replaced by = or <.

Proposition 22 If a : b = a′ : b′ and b : c = b′ : c′ , then a : c = a′ : c′ .

Proof Assume a : b = a′ : b′ and b : c = b′ : c′ and let m and n be any two

natural numbers. Then

(ma) : (mb) = (ma′ ) : (mb′ ) and (mb) : (nc) = (mb′ ) : (nc′ )

by Proposition 4. And ma has to nc the same relation (<, =, or >) as ma′

has to nc′ by Proposition 20. Therefore a : c = a′ : c′ according to Definition
9.2.

Proposition 23 If a : b = b′ : c′ and b : c = a′ : b′ , then a : c = a′ : c′ .

26
Proof Assume a : b = b′ : c′ and b : c = a′ : b′ and let m and n be any two
natural numbers. Then

(ma) : (nb) = (mb′ ) : (nc′ ) and (mb) : (nc) = (ma′ ) : (nb′ ) .

by Proposition 4. And ma has to nc the same relation (<, =, or >)as ma′

has to nc′ by Proposition 21. Therefore a : c = a′ : c′ according to Definition
9.2.

Proposition 24 If a : b = c : d and e : b = f : d, then (a + e) : b = (c + f ) :

d.

Proof Assume a : b = c : d and e : b = f : d or, equivalently,

a : b = c : d and b : e = d : f .

Then a : e = c : f by Proposition 22 and (a + e) : e = (c + f ) : f by

Proposition 18. And e : b = f : d by assumption. Therefore (a + e) : b =
(c + f ) : d by Proposition 22.

11 Embeddings and the Fourth Proportional

In this section a, b, c are elements of a magnitude space M and a′ , b′ , c′ are
elements of a magnitude space M ′ .

Definition 11.1 If a : b = a′ : b′ , then we say that b′ is a fourth propor-

tional to a, b, and a′ .

Theorem 11.1 If there is a fourth proportional to a, b, and a′ , then it is

unique.

Proof Suppose b′ and c′ are each fourth proportionals to a, b, and a′ . Then

a : b = a′ : b′ and a : b = a′ : c′ by Definition 11.1. Hence a′ : b′ = a′ : c′ by
Proposition 11 and b′ = c′ by Proposition 9.

Theorem 11.2 If ϕ : M → M ′ is an embedding which maps a into a′ , then

ϕb is the fourth proportional to a, b, a′ for each b.

Proof Theorem 9.4 and Definition 11.1.

27
Theorem 11.3 If a and a′ are fixed and there is, for each b, a fourth pro-
portional to a, b, a′ , then there is an embedding ϕ : M → M ′ which maps a
into a′ .

Proof Let a and a′ be fixed and suppose that there is, for each b, a fourth
proportional to a, b, a′ . Then for each b there is exactly one fourth propor-
tional to a, b, a′ by Theorem 11.1. Thus we can define a function ϕ : M → M ′
such that ϕb is the fourth proportional to a, b, and a′ .
I say ϕ is an embedding. For any two elements b and c of M, we have by
assumption

a : b = a′ : ϕb, a : c = a′ : ϕc, and a : b + c = a′ : ϕ (b + c)

or equivalently by Definition 9.2

ϕb : a′ = b : a, ϕc : a′ = c : a, and ϕ (b + c) : a′ = b + c : a.

Thus
ϕb + ϕc : a′ = b + c : a
by Proposition 24,
ϕ (b + c) : a′ = ϕb + ϕc : a′
by Proposition 11, and
ϕ (b + c) = ϕb + ϕc
by Proposition 9. We have now shown that ϕ is homomorphism according
to Definition 5.1 and hence ϕ is an embedding by Theorem 5.2. And ϕa is
the fourth proportional to a, a, and a′ , or in other words a : a = a′ : ϕa.
And since 1a = 1a, it follows that 1a′ = 1 (ϕa) by Definition 9.2. Therefore
a′ = ϕa by Definition 8.2.

In fact the embedding constructed in the preceding theorem is the unique

embedding of M into M ′ which maps a into a′ . This and a bit more is
established in the next theorem.

Theorem 11.4 If ϕ and χ are embeddings from M into M ′ , then ϕa has to

χa the same relation (<, =, or >) as ϕb has to χb. (In particular if ϕa = χa
for one a, then ϕb = χb for all b and ϕ = χ.)

28
Proof Assume ϕ and χ are embeddings from M into M ′ . Then

ϕa : ϕb = a : b and χa : χb = a : b

by Theorem 9.4 and hence

ϕa : ϕb = χa : χb

by Proposition 11. Therefore ϕa has to χa the same relation (<, =, or >)

as ϕb has to χb by Proposition 14.

Theorem 11.5 If ϕ and χ are endomorphisms of M, then ϕχ = χϕ.

Proof Assume ϕ and χ are endomorphisms of M. Then for any a ∈ M,

ϕ (χa) : ϕa = χa : a and ϕa : a = χ (ϕa) : χa

by Theorem 9.4,
ϕ (χa) : a = χ (ϕa) : a
by Proposition 23, and
ϕ (χa) = χ (ϕa)
by Proposition 9. Therefore (ϕχ) a = (χϕ) a by Definition 3.5 and ϕχ = χϕ
by Definition 3.2.

12 Continuous Magnitude Spaces

In this section we come to the central theorem: If a is an element of an
Archimedean magnitude space M and a′ is an element of a continuous mag-
nitude space M ′ , then there is a unique embedding of M into M ′ which maps
a into a′ . Before tackling the main theorem we need to establish some results
concerning ratios which, although very basic, have not been required prior
to this section.

Theorem 12.1 If a : b > a′ : b′ and ma ≤ nb, then ma′ < nb′ .

Proof ma : mb = a : b by Proposition 15 and a : b > a′ : b′ by assumption.

Hence ma : mb > a′ : b′ by Proposition 13. Likewise, a′ : b′ = ma′ : mb′

29
by Proposition 15 and hence ma : mb > ma′ : mb′ by Proposition 13. Thus
there exist j and k such that
j (ma) > k (mb) and j (ma′ ) ≤ k (mb′ )
by Definition 9.2. And by assumption, ma ≤ nb and so j (ma) ≤ j (nb) by
Proposition 5. And also, from above, j (ma) > k (mb) and hence k (mb) <
j (nb) by Theorem 6.1 and hence km < jn by Propositions 3 and 6. Hence
k (mb′ ) < j (nb′ ) by Propositions 6 and 3. And also, from above, j (ma′ ) ≤
k (mb′ ) and hence j (ma′ ) < j (nb′ ) by Theorem 6.1 and hence ma′ < nb′ by
Proposition 5; the very thing to be shown.
Theorem 12.2 If a : b > a′ : b′ and a′ : b′ > a′′ : b′′ , then a : b > a′′ : b′′ .
Proof Assume a : b > a′ : b′ and a′ : b′ > a′′ : b′′ . From the first ratio
inequality there are natural numbers m and n such that ma > nb and ma′ ≤
nb′ by Definition 9.2. And also ma′′ < nb′′ by the preceding theorem. Hence
ma > nb and ma′′ < nb′′ and therefore a : b > a′′ : b′′ according to Definition
9.2.
Theorem 12.3 If there are natural numbers j, k such that ja > kb and
ja′ = kb′ , then there are natural numbers m, n such that ma > nb and
ma′ < nb′ .
Proof Assume ja > kb and ja′ = kb′ . Since magnitude spaces are assumed
to be Archimedean, there is some natural number p such that p (ja − kb) >
1a by Theorem 9.2. Therefore
p (ja − kb) > a =⇒ p (ja) − p (kb) > 1a (Proposition 5)
=⇒ (pj) a − (pk) b > 1a (Proposition 3)
=⇒ (pj) a > (pk) b + 1a (Theorem 4.9)
=⇒ (pj) a − 1a > (pk) b (Theorem 4.9)
=⇒ (pj − 1) a > (pk) b. (Theorem 8.1)
And since ja′ = kb′ ,
(pj − 1) a′ < (pj) a′ (Theorem 4.5 and Proposition 6)
= p (ja′ ) (Proposition 3)
= p (kb′ ) (ja′ = kb′ )
= (pk) b′ ((Proposition 3)

30
and hence (pj − 1) a′ < (pk) b′ . We now have two natural numbers m = pj−1
and n = pk such that ma > nb and ma′ < nb′ .

Theorem 12.4 If a : b 6= a′ : b′ , then a : b > a′ : b′ or a′ : b′ > a : b.

Proof If a : b 6= a′ : b′ , then there are two natural numbers j, k for which at

least one of following six cases is true:

1) ja < kb and ja′ = kb′

2) ja < kb and ja′ > kb′
3) ja = kb and ja′ < kb′
4) ja = kb and ja′ > kb′
5) ja > kb and ja′ < kb′
6) ja > kb and ja′ = kb′

In each of the cases 2, 4, 5, and 6, a : b > a′ : b′ or a′ : b′ > a : b according

to Definition 9.2. In case 1, there are two natural numbers m, n such that
na < mb and na′ > mb′ by the preceding theorem and hence a′ : b′ > a : b.
In a similar fashion we can show that in case 3 a : b > a′ : b′ .

Theorem 12.5 If M is nondiscrete, then for any a ∈ M and n ∈ N there

is some b ∈ M such that nb < a.

Proof Assume M is nondiscrete. Then the theorem is true for n = 1 since

there is some element b < a by Definition 6.3 and 1b = b by Definition 8.2.
Now suppose the theorem is true for n and nb < a. There is some c ∈ M
such that c < b by Definition 6.3. Let d be the smaller of (b − c) and c so
that d ≤ b − c and d ≤ c. There is some e ∈ M such that e < d by Definition
6.3 and hence e < b − c and e < c by Theorem 6.1. Hence

e + e < (b − c) + e (e < b − c and Theorem 4.4)

< (b − c) + c (e < c and Theorem 4.4)
= b. (Definition 4.5)

31
And hence e + e < b by Theorem 4.6. Now 1 ≤ n by Definition 8.1 and hence
n + 1 ≤ n + n by Theorem 4.4. Therefore

(n + 1) e ≤ (n + n) e (Proposition 6)
= ne + ne (Proposition 2)
= n (e + e) (Proposition 1)
< nb (Proposition 5)
< a (assumption)

and so (n + 1) e < a by Theorem 6.1 and hence the theorem is true for n + 1.
Therefore the theorem is true for all n by Theorem 7.1.

Theorem 12.6 If M and M ′ are magnitude spaces, M is nondiscrete, a, b ∈

M, a′ , b′ ∈ M ′ , then if a : b > a′ : b′ there exists some c ∈ M such that
a : b > c : b > a′ : b′ , and if a : b < a′ : b′ there exists some c ∈ M such that
a : b < c : b < a′ : b′ .

Proof If a : b > a′ : b′ , then there are m, n ∈ N such that

ma > nb and ma′ ≤ nb′

by Definition 9.2. From the preceding theorem, there is some d ∈ M such

that
ma − nb > md.
And

ma − nb > md =⇒ ma > md + nb (Theorem 4.9)

=⇒ ma − md > nb (Theorem 4.9)
=⇒ m (a − d) > nb. (Proposition 5)

Letting c = a − d we have

mc > nb and ma′ ≤ nb′

and hence c : b > a′ : b′ . And since a > c, from Proposition 8, a : b > c : b.

The second part of the theorem is proved in a similar manner.

With these preliminaries out of the way, we are ready for the main theo-
rem in this section.

32
Theorem 12.7 If M is a magnitude space, M ′ is a continuous magnitude
space, a ∈ M, and a′ ∈ M ′ , then for each b ∈ M there is a fourth proportional
b′ ∈ M ′ to a, b, and a′ .

Proof Assume M ′ is continuous and hence complete and nondiscrete by

Definition 6.6. Let a, b ∈ M and a′ ∈ M ′ be given. Let

A = {c′ ∈ M ′ | c′ : a′ < b : a} and B = {c′ ∈ M ′ | c′ : a′ > b : a} .

We first show that each of these sets is nonempty. Since M is Archimedean,

there is some natural number m such that mb > 1a by Theorem 9.2. And
since M ′ is nondiscrete, there is some c′ ∈ M ′ such that mc′ < 1a′ from
Theorem 12.5. Then mb > 1a and mc′ < 1a′ ; hence b : a > c′ : a′ by
Definition 9.2 and c′ ∈ A. Likewise, there is some natural number n such
that 1b < na by Theorem 9.2. Let c′ = na′ + a′ . Then 1c′ > na′ and 1b < na;
hence c′ : a′ > b : a by Definition 9.2 and c′ ∈ B.
If c′ ∈ A and d′ ∈ B, then c′ : a′ < b : a and b : a < d′ : a′ . Thus
c : a′ < d′ : a′ from Theorem 12.2 and hence c′ < d′ from Proposition 10.
′

Thus every element of B is an upper bound of A and every element of A is

a lower bound of B according to Definition 6.2.
Since M ′ is complete, A has a least upper bound b′ by Definition 6.5. I
say b′ is a fourth proportional to a, b, and a′ . That is, a : b = a′ : b′ or
equivalently b′ : a′ = b : a. Suppose that b′ is not a fourth proportional to a,
b, and a′ . Then b′ : a′ < b : a or b′ : a′ > b : a from Theorem 12.4, or what is
the same b′ ∈ A or b′ ∈ B.
If b′ ∈ A (i.e. b′ : a′ < b : a), then there is a c′ ∈ M ′ such that
b′ : a′ < c′ : a′ < b : a from Theorem 12.6 and hence c′ ∈ A and b′ < c′ . But
b′ is an upper bound for A which is a contradiction. Therefore b′ ∈ / A.
If b′ ∈ B (i.e. b′ : a′ > b : a), then from Theorem 12.6 there is a c′ ∈ M ′
such that b′ : a′ > c′ : a′ > b : a and hence c′ ∈ B and b′ > c′ . But since
c′ ∈ B, c′ is an upper bound of A and it is smaller than the least upper bound
of A which is a contradiction. Therefore b′ ∈ / B.
We have now shown that b is not an element of A or B which contradicts
′

b′ not being the fourth proportional to a, b, and a′ . Therefore, indeed, b′ is

the fourth proportional to a, b, and a′ .

Theorem 12.8 If M is a magnitude space, M ′ is a continuous magnitude

space, a ∈ M, and a′ ∈ M ′ , then there exists a unique embedding ϕ : M →
M ′ such that ϕa = a′ .

33
Proof Preceding theorem, Theorem 11.3, and Theorem 11.4.

Theorem 12.9 If M and M ′ are continuous magnitude spaces, a ∈ M and

a′ ∈ M ′ , then there exists a unique isomorphism ϕ : M → M ′ such that
ϕa = a′ .

Proof Since M ′ is continuous, there exists a unique embedding ϕ : M → M ′

such that ϕa = a′ by the preceding theorem. And since M is continuous,
there is also a unique embedding χ : M ′ → M such that χa′ = a by the
preceding theorem. But then ϕ◦χ is an embedding of M ′ into M ′ by Theorem
5.5. And ϕ ◦ χ maps a′ into a′ . But the identity map iM ′ on M ′ is an
embedding by Theorem 5.3 and also iM ′ maps a′ into a′ . Hence ϕ ◦ χ = iM ′
by Theorem 11.4. Hence ϕ is onto by Theorem 3.3 and therefore is an
isomorphism according to Definition 5.2.

Theorem 12.10 Any two continuous magnitude spaces are isomorphic.

Proof Let M and M ′ be any two continuous magnitude spaces. Pick any
a ∈ M and a′ ∈ M ′ . There is an embedding ϕ : M → M ′ such that ϕa = a′
by Theorem 12.8. The embedding ϕ is an isomorphism by the preceding
theorem and therefore M is isomorphic to M ′ according to Definition 5.2.

34
13 Real Numbers
Let us now review the definition of the natural numbers and define the pos-
itive real numbers.
1. Well ordered magnitude spaces are complete (Theorem 6.5) and hence
Archimedean (Theorem 6.6).
2. If M is a well ordered magnitude space, M ′ is any magnitude space (not
necessarily Archimedean), a is the smallest element in M, and a′ is any
element in M ′ , then there exists a unique embedding ϕ : M → M ′ such
that ϕa = a′ (Theorem 7.3).
3. Any two well ordered magnitude spaces are isomorphic (Theorem 7.4).
Because a well ordered magnitude space can be embedded into any mag-
nitude space, we say that well ordered magnitude spaces are minimal mag-
nitude spaces. And for the same reason we also say that the well ordered
magnitude spaces are minimal Archimedean magnitude spaces. We defined
the natural numbers as an arbitrary representative of the well ordered mag-
nitude spaces.
1. Continuous magnitude spaces are complete (Definition 6.6) and hence
Archimedean (Theorem 6.6).
2. If M is an Archimedean magnitude space, M ′ is a continuous magnitude
space, a ∈ M, and a′ ∈ M ′ , then there exists a unique embedding
ϕ : M → M ′ such that ϕa = a′ (Theorem 12.8).
3. Any two continuous magnitude spaces are isomorphic (Theorem 12.10).
Because any Archimedean magnitude space can be embedded into a con-
tinuous magnitude space, we say that continuous magnitudes are maximal
Archimedean magnitude spaces. We now define the positive real numbers as
an arbitrary representative of the continuous magnitude spaces.
Definition 13.1 Pick any continuous magnitude space and denote it by R+
and pick any element of R+ and denote it by 1. Since all continuous magni-
tude spaces are isomorphic (Theorem 12.10) and we will use only the algebraic
properties that the continuous magnitude spaces have in common, it does not
matter which continuous magnitude space is chosen to play the role of the
“number system” R+ . We call R+ the positive real numbers.

35
 When we defined the natural numbers, we immediately defined an integral
multiple na where n is a natural number and a is an element of a magnitude
space. The remainder of this work examines similar constructions of multiples
where the multiplier is not necessarily a natural number. This will lead us to
real multiples of real numbers which is the familiar binary product operator
in R+ .

14 Magnitude Embedding Spaces

Definitions of multiples and products are based on embeddings of one mag-
nitude space into another. It is useful at the onset to consider the general
case.

Definition 14.1 H (M, M ′ ) is the set of all embeddings from M into M ′ .

It may happen that H (M, M ′ ) is the empty set. For example, if M is

nondiscrete and M ′ is discrete, there are no embeddings of M into M ′ . If
H (M, M ′ ) is not empty and ϕ, χ ∈ H (M, M ′ ), then we have already shown
(Theorem 5.4) that ϕ + χ ∈ H (M, M ′ ).

Theorem 14.1 If H (M, M ′ ) is nonempty, then H (M, M ′ ) is a magnitude

space.

Proof Let ϕ, χ, ψ ∈ H (M, M ′ ).

(i)For any a ∈ M

(ϕ + (χ + ψ)) a = ϕa + (χ + ψ) a (Definition 5.4)

= ϕa + (χa + ψa) (Definition 5.4)
= (ϕa + χa) + ψa (Definition 2.1)
= (ϕ + χ) a + ψa (Definition 5.4)
= ((ϕ + χ) + ψ) a (Definition 5.4)

and hence ϕ + (χ + ψ) = (ϕ + χ) + ψ according to Definition 3.2.

(ii)For any a ∈ M

(ϕ + χ) a = ϕa + χa (Definition 5.4)
= χa + ϕa (Definition 2.1)
= (χ + ϕ) a (Definition 5.4)

36
and hence ϕ + χ = χ + ϕ according to Definition 3.2.
(iii)Fix a ∈ M. Since ϕa, χa ∈ M ′ , exactly one of the following is true:

ϕa = χa,
ϕa > χa, or
χa > ϕa

by Theorem 4.2. In the first case, Theorem 11.4 shows that ϕb = χb for all
b ∈ M and hence ϕ = χ according to Definition 3.2. In the second case,
ϕb > χb for all b ∈ M by Theorem 11.4. We can then define a function
d : M → M ′ by
δb = ϕb − χb.
For any b, c ∈ M

ϕ (b + c) = ϕb + ϕc (Definition 5.1)
= ((ϕb − χb) + χb) + ((ϕc − χc) + χc) (Definition 4.5)
= ((ϕb − χb) + (ϕc − χc)) + (χb + χc) (associativity and commutativity of + )
= (δb + δc) + (χb + χc) (definition of δ)
= (δb + δc) + χ (b + c) (Definition 5.1)

and hence ϕ (b + c) − χ (b + c) = (δb + δc) by Definition 4.5. But δ (b + c) =

ϕ (b + c) − χ (b + c) by the definition of δ. Hence δ (b + c) = δb + δc and δ is
an element of H (M, M ′ ) by Definition 5.1 and Theorem 5.2. And, of course,

ϕ (b) = χb + (ϕb − χb) (Definition 4.5)

= χb + δb (definition of δ)
= (χ + δ) b (Definition 5.4)

for all b ∈ M and hence ϕ = χ + δ according to Definition 3.2. A similar

argument applies in the third case where χa > ϕa. So, in summary, if
ϕ, χ ∈ H (M, M ′ ), then exactly one of the following is true:

ϕ = χ,
ϕ = χ + δ for some δ ∈ H (M, M ′ ) , or
χ = ϕ + δ for some δ ∈ H (M, M ′ ) .

We have now shown that H (M, M ′ ), with the usual addition of embeddings,
is a magnitude space according to Definition 2.1.

37
15 Magnitude Endomorphism Spaces
A special case of magnitude space embeddings are endomorphisms; that is
embeddings of a magnitude space into itself.

Definition 15.1 If M is a magnitude space, then we denote the set of en-

domorphisms of M by E (M).

Of course E (M) = H (M, M). Now in general H (M, M ′ ) can be empty.

But E (M) always has elements. In fact for each natural number n the
mapping ϕ : M → M defined by ϕa = na is an embedding of M into M and
hence is an element of E (M).
If ϕ, χ ∈ E (M) then ϕ + χ and ϕ ◦ χ are each elements of E (M) by
Theorems 5.4 and 5.5. We have already shown that E (M) = H (M, M) is a
magnitude space with the + operator on elements of E (M) (Theorem 14.1).
Also we know that the composition operator ◦ is commutative (Theorem
11.5) and associative (Theorem 3.1).

Theorem 15.1 If ϕ, χ, ψ ∈ E (M) then ϕ ◦ (χ + ψ) = ϕ ◦ χ + ϕ ◦ ψ and

(ϕ + χ) ◦ ψ = ϕ ◦ ψ + χ ◦ ψ.

Proof If a ∈ M then

(ϕ ◦ (χ + ψ)) a = ϕ ((χ + ψ) a) (Definition 3.5)

= ϕ (χa + ψa) (Definition 5.4)
= ϕ (χa) + ϕ (ψa) (Definition 5.1)
= (ϕ ◦ χ) a + (ϕ ◦ ψ) a (Definition 3.5)
= (ϕ ◦ χ + ϕ ◦ ψ) a (Definition 5.4)

and hence ϕ ◦ (χ + ψ) = ϕ ◦ χ + ϕ ◦ ψ by Definition 3.2. The second part of

the theorem can be proved in a similar manner.

Theorem 15.2 If ϕ, χ, ψ ∈ E (M) then ϕ ◦ χ has to ϕ ◦ ψ the same relation

(<, =, or >) as χ has to ψ and χ ◦ ϕ has to ψ ◦ ϕ the same relation (<, =,
or >) as χ has to ψ.

38
Proof Fix ϕ ∈ E (M) and let the mapping Ψ : E (M) → E (M) be defined
by Ψχ = ϕ ◦ χ. If χ, ψ ∈ E (M), then

Ψ (χ + ψ) = ϕ ◦ (χ + ψ) (definition of Ψ)
= ϕ ◦ χ + ϕ ◦ ψ (Theorem 15.1)
= Ψχ + Ψψ (definition of Ψ)

and hence Ψ is a homomorphism by Definition 5.1. Therefore ϕ ◦ χ = Ψχ

has to ϕ ◦ ψ = Ψψ the same relation (<, =, or >) as χ has to ψ (and Ψ is an
embedding) by Theorem 5.2. The second part of the theorem can be proved
in a similar manner.

Theorem 15.3 If ϕ ∈ E (M), then ϕ ◦ iM = iM ◦ ϕ = ϕ.

Proof If a ∈ M, then

(ϕ ◦ iM ) a = ϕ (iM a) (Definition 3.5)

= ϕa (Definition 3.6)

and therefore ϕ ◦ iM = ϕ by Definition 3.2. And ϕ ◦ iM = iM ◦ ϕ since

endomorphisms commute by Theorem 11.5.
At this point we would like to point out that the magnitude space E (M)
for an arbitrary (Archimedean) magnitude space has exactly those properties
that we associate with addition and multiplication of positive numbers.

1. E (M) with the addition operator is a magnitude space (Theorem 14.1).

(a) ϕ + (χ + ψ) = (ϕ + χ) + ψ
(b) ϕ + χ = χ + ϕ
(c) Exactly one of the following is true: ϕ = χ, or ϕ = χ + δ for some
δ ∈ E (M), or χ = ϕ + δ for some δ ∈ E (M).

2. The composition operator is associative and commutative (Theorems

3.1 and 11.5).

(a) ψ ◦ (χ ◦ ϕ) = (ψ ◦ χ) ◦ ϕ
(b) ϕ ◦ χ = χ ◦ ϕ

39
 3. Composition distributes over addition (Theorem 15.1).

(a) ϕ ◦ (χ + ψ) = ϕ ◦ χ + ϕ ◦ ψ
(b) (ϕ + χ) ◦ ψ = ϕ ◦ ψ + χ ◦ ψ

4. There is an identity element for the composition operator (Theorem

15.3).

(a) ϕ ◦ iM = ϕ
(b) iM ◦ ϕ = ϕ

5. Composition on left or right preserves order relations in the magnitude

space E (M) (Theorem 15.2)

(a) ϕ ◦ χ has to ϕ ◦ ψ the same relation (<, =, or >) as χ has to ψ

(b) χ ◦ ϕ has to ψ ◦ ϕ the same relation (<, =, or >) as χ has to ψ

16 Generalized Multiple
In Section 14 we showed H (M, M ′ ) is itself a magnitude space. The question
naturally arises of how H (M, M ′ ) might be related to the magnitude spaces
M and M ′ . In this section we consider the special case in which M is a
magnitude space with some distinguished element 1, and M ′ is any magnitude
space such that for each element a′ ∈ M ′ there is an embedding of M into
M ′ which maps 1 into a′ . We already have two examples of this special case:

1. M is a well ordered magnitude space and 1 is the smallest element in

M; and M ′ is an arbitrary (Archimedean) magnitude space (Theorem
7.3).
2. M is an arbitrary (Archimedean) magnitude space and 1 is any element
in M; and M ′ is a continuous magnitude space (Theorem 12.8).

In this section variables a, b are elements of M and a′ , b′ are elements of

M.′

Definition 16.1 Let the mapping Ψ : M ′ → H (M, M ′ ) be defined such that

for each a′ ∈ M ′ , Ψa′ is the unique element of H (M, M ′ ) which maps 1 into
a′ . For any a ∈ M, a′ ∈ M ′ we define aa′ to be (Ψa′ ) a. Note that aa′ ∈ M ′ .

40
Theorem 16.1 The map Ψ : M ′ → H (M, M ′ ) is an isomorphism.

Proof Note that H (M, M ′ ) is a magnitude space by Theorem 14.1. Let

a′ and b′ be any two elements of M ′ . By assumption there exist elements
Ψa′ , Ψb′ , Ψ (a′ + b′ ) ∈ H (M, M ′ ) which map 1 into a′ , b′ , a′ + b′ respectively.
Now Ψa′ + Ψb′ ∈ H (M, M ′ ) by Theorem 5.4 and

(Ψa′ + Ψb′ ) 1 = (Ψa′ ) 1 + (Ψb′ ) 1 (Definition 5.4)

= a′ + b′ . (definition of Ψa′ and Ψb′ )

Thus Ψ (a′ + b′ ) and Ψa′ + Ψb′ are each embeddings which map 1 into a′ + b′
and hence Ψ (a′ + b′ ) = Ψa′ + Ψb′ by Theorem 11.4. Therefore Ψ is an
embedding by Definition 5.1 and Theorem 5.2.
Now if χ ∈ H (M, M ′ ), then χ maps 1 into χ1. And also Ψ (χ1) ∈
H (M, M ′ ) maps 1 into χ1. Therefore χ = Ψ (χ1) by Theorem 11.4. Hence Ψ
is onto and therefore Ψ is an isomorphism from M ′ onto H (M, M ′ ) according
to Definition 5.2.

Theorem 16.2 1a′ = a′

Proof 1a′ = (Ψa′ ) 1 = a′ by Definition 16.1.

The following two theorems were proved separately for integral multiples
in Propositions 1 and 2.

Theorem 16.3 a (a′ + b′ ) = aa′ + ab′

Proof

a (a′ + b′ ) = (Ψ (a′ + b′ )) a (Definition 16.1)

= (Ψa′ + Ψb′ ) a (Definition 5.1)
= (Ψa′ ) a + (Ψb′ ) a (Definition 5.4)
= aa′ + ab′ (Definition 16.1)

Theorem 16.4 (a + b) a′ = aa′ + ba′

41
Proof

(a + b) a′ = (Ψa′ ) (a + b) (Definition 16.1)

= (Ψa′ ) a + (Ψa′ ) b (Definition 5.1)
= aa′ + ba′ (Definition 16.1)

The following two theorems were proved separately for integral multiples
in Propositions 5 and 6.

Theorem 16.5 If a′ > b′ , then aa′ > ab′ and aa′ − ab′ = a (a′ − b′ ). And,
more generally, aa′ has to ab′ the same relation (<, =, or >) as a′ has to b′ .

Proof Fix a and let ϕ : M ′ → M ′ be the function defined by ϕa′ = aa′ .

Then ϕ (a′ + b′ ) = ϕa′ + ϕb′ for any a′ and b′ by Theorem 16.3 and hence ϕ
is a homomorphism according to Definition 5.1. Hence Theorems 5.1 and 5.2
apply.

Theorem 16.6 If a > b, then aa′ > ba′ and aa′ − ba′ = (a − b)a′ . And, more
generally, aa′ has to ba′ the same relation (<, =, or >) as a has to a.

Proof Fix a′ and let ϕ : M → M be the function defined by ϕa = aa′ .

Then ϕ (a + b) = ϕa+ ϕb for any a and b by Theorem 16.4 and hence ϕ is
a homomorphism according to Definition 5.1. Hence Theorems 5.1 and 5.2
apply.

17 Generalized Product Operator

In this section we assume that M is a magnitude space with some distin-
guished element 1 such that for each element a ∈ M there is an embedding
Ψa ∈ E (M) = H (M, M) which maps 1 into a. In other words, we assume
the same thing as in the previous section but, in addition, M = M ′ . In this
case we will write a product ab = (Ψb) a as a · b to emphasize that we have
here a binary operator. Here are two examples of this special case.

1. M is a well ordered magnitude space and 1 is the smallest element in

M (for instance, M = N).

42
 2. M is a continuous magnitude space and 1 is an arbitrary element in M
(for instance, M = R+ ).

From the preceding section Ψ : M → E (M) is an isomorphism and, since

Ψ is an isomorphism,
Ψ (a + b) = Ψa + Ψb
where the addition operator on the right hand side represents the addition
of the endomorphisms Ψa and Ψb and the result is another endomorphism.

Theorem 17.1 Ψ1 = iM

Proof Ψ1 is an embedding of M into M which maps 1 into 1. And the

identity function iM is also an embedding of M into M which maps 1 into 1.
Therefore Ψ1 = iM by Theorem 11.4.

Theorem 17.2 Ψ (a · b) = Ψa ◦ Ψb

Proof First (Ψ (a · b)) 1 = a · b from the definition of Ψ. Second

(Ψa ◦ Ψb) 1 = (Ψb ◦ Ψa) 1 (Theorem 11.5)

= Ψb ((Ψa) 1) (Definition 3.5)
= (Ψb) a (definition of Ψ)
= a · b (Definition 16.1)

and hence also (Ψa ◦ Ψb) 1 = a · b. Therefore Ψ (a · b) = Ψa ◦ Ψb by Theorem

11.4.

Remark 17.1 By means of the preceding theorem, we can show that the
product binary operator · on M has analogous properties as the binary oper-
ator ◦ on E (M) as summarized at the end of Section 15.

Theorem 17.3 a · b = b · a

Proof

Ψ (a · b) = Ψa ◦ Ψb (preceding theorem)
= Ψb ◦ Ψa (Theorem 11.5)
= Ψ (b · a) (preceding theorem)

Thus Ψ (a · b) = Ψ (b · a) and therefore a · b = b · a by Theorem 5.2.

43
Theorem 17.4 (a · b) · c = a · (b · c)

Proof

Ψ ((a · b) · c) = Ψ (a · b) ◦ Ψc (Theorem 17.2)

= (Ψa ◦ Ψb) ◦ Ψc (Theorem 17.2)
= Ψa ◦ (Ψb ◦ Ψc) (Theorem 3.1)
= Ψa ◦ Ψ (b · c) (Theorem 17.2)
= Ψ (a · (b · c)) (Theorem 17.2)

Thus Ψ ((a · b) · c) = Ψ (a · (b · c)) and therefore (a · b) · c = a · (b · c) by

Theorem 5.2.

18 Symmetric Magnitude Spaces

From above, we have the definitions of binary product operators for the natu-
ral numbers and the positive real numbers and we have shown the properties
which these two binary operators have in common. There are, of course,
differences between the product operators for the natural and positive real
numbers. In particular, given any x, y ∈ R+ there is a q ∈ R+ (called the
quotient of y and x) such that y = q ·x. Once again it is useful to develop this
additional property of the real numbers from a general perspective; for there
are also noncontinuous magnitude spaces in which every pair of elements has
a quotient.

Definition 18.1 A magnitude space M is symmetric if there is, for every

pair a, b ∈ M, an endomorphism of M which maps a into b.

Remark 18.1 Alternatively, a magnitude space M is symmetric if, for

every three elements a, b, c ∈ M, there is a fourth proportional (i.e. an
element d ∈ M such that a : b = c : d).

Proof Theorem 12.8.

Theorem 18.1 If M is a symmetric magnitude, then for each ϕ ∈ E (M),

there is a corresponding χ ∈ E (M) such that χ ◦ ϕ = ϕ ◦ χ = iM and each
of ϕ and χ are onto (and thus automorphisms).

44
Proof Let ϕ ∈ E (M) and pick any element a ∈ M. Since M is a symmetric
magnitude space, there is another embedding χ ∈ E (M) such that χ maps
ϕa into a. Then χ◦ϕ maps a into a. And also iM maps a into a by Definition
3.6. Hence χ ◦ ϕ = iM by Theorem 11.4. But also χ ◦ ϕ = ϕ ◦ χ by Theorem
11.5 and so ϕ ◦ χ = iM . Therefore ϕ and χ are each onto by Theorem 3.3
and so ϕ and χ are automorphisms according to Definition 5.3.
In the remainder of this section M is a symmetric magnitude space with
one element designated by 1. Then for any a ∈ M, there is a unique em-
bedding of M into M which maps 1 into a. From the preceding section, we
can define an isomorphism Ψ : M → E (M) such that for a ∈ M, Ψa is the
unique element in E (M) which maps 1 into a. And, as before, we can define
a binary operator on M according to a · b = (Ψb) a. A property of this binary
product which differs from the product of two natural numbers is given in
the following theorem.

Theorem 18.2 If a, b ∈ M, there exists a unique d ∈ M such that b = d · a.

Proof The embedding Ψa from M into M is an automorphism by the pre-

ceding theorem and hence is one-to-one and onto by Definition 5.3. Hence
there is a unique d ∈ M such that (Ψa) d = b by Definitions 3.3 and 3.4.
Therefore there is a unique d ∈ M such that d · a = b by Definition 16.1.

Definition 18.2 For a, b ∈ M, we denote by b/a the unique element d of M

such that b = d · a.

Theorem 18.3 b has to a the same relation (<, =, or >) as b/a has to 1.

Proof (b/a) · a has to 1 · a the same relation (<, =, or >) as b/a has to 1 by
Theorem 16.6. But b = (b/a) · a by Definition 18.2 and 1 · a = a by Theorem
16.2. Therefore b has to a the same relation (<, =, or >) as b/a has to 1.

19 Power Functions
We have mentioned above that it is worthwhile to view the formation of
products in a general way. As a concrete example, let us consider the defini-
tion of xy where x and y are positive real numbers and x > 1. In this section
· is the product operator in R+ .

Definition 19.1 Let R>1 be the elements of R+ which are greater than 1.

45
Theorem 19.1 If x, y ∈ R>1 , then x · y ∈ R>1 .

Proof If x, y ∈ R>1 , then x > 1 and y > 1 by Definition 19.1. And x·y > 1·y
by Theorem 16.6 and 1 · y = y by Theorem 16.2. Thus x · y > y and y > 1.
Therefore x · y > 1 by Theorem 4.6 and x · y ∈ R>1 by Definition 19.1.

Theorem 19.2 R>1 with the multiplicative operator · is a magnitude space.

Proof The multiplicative operator · is a binary operator on R>1 by the

preceding theorem. And · is associative and commutative by Theorems 17.4
and 17.3.
It remains to show that the binary operator · on R>1 is trichotomous. Let
x, y ∈ R>1 . Note that y = (y/x) · x = x · (y/x) and x = (x/y) · y = y · (x/y)
by Definition 18.2 and Theorem 17.3. From trichotomy in R+ , exactly one
of the following is true: y < x, or y = x, or y > x.
Now in each case, there is a unique d ∈ R+ , namely d = x/y, such that
x = y · d by Theorem 18.2. But d = x/y ∈ R>1 only if y < x. Similarly, in
each case, there is a unique d ∈ R+ , namely d = y/x such that y = x · d. But
d = y/x ∈ R>1 only if y > x. Thus there are three mutually exclusive cases:
x = y · d for some d ∈ R>1 , or x = y, or y = x · d for some d ∈ R>1 . Thus R>1
with the · binary operator has the trichotomy property in Definition 2.1.

Theorem 19.3 R>1 with the multiplicative operator · is a continuous mag-

nitude space.

Proof We begin with an elementary observation. If A is a nonempty set

with some element a greater than 1 and b is an upper bound of A, then
b > 1. For if b is an upper bound of A, then a ≤ b by Definition 6.2. And
from 1 < a and a ≤ b follows 1 < b by Theorem 6.1.
Definition 2.2 defines < and > in a magnitude space in terms of the binary
operator of the magnitude space. In the proof of the preceding theorem, it
may be observed that for x, y ∈ R>1 , x has to y the same relation (<, =,
or >) in the order defined by + binary operator as x has to y in the order
defined defined by the · operator.
Now let A be a nonempty subset of R>1 with a nonempty set B of upper
bounds with respect to the order defined by the · operator. From the prelimi-
nary observation, B coincides with the set of upper bounds of A with respect
to the order defined by the + operator. But R+ is continuous by Definition
13.1 and hence B has a smallest element with respect to the order defined by

46
the + operator by Definition 6.6. Therefore B has a smallest element with
respect to the order defined by the · operator. Therefore R>1 is continuous
by Definition 6.6.

Up to the preceding theorem, we have consistently used the symbol +

for the binary operator in a magnitude space. To be precise, if we have
two magnitude spaces M and M ′ , the binary operators are in general not
the same. If a, b ∈ M and a′ , b′ ∈ M ′ then the + sign in the expression
a + b is understood to be the binary operator in M and the + sign in the
expression a′ + b′ is understood to be the binary operator in M ′ . Admittedly
it would be more precise to denote the binary operator in M ′ by +′ and
to write a′ +′ b′ but we have left it up to reader to make this distinction.
In particular we defined a map ϕ : M → M ′ to be a homomorphism if
for all a, b ∈ M, ϕ (a + b) = (ϕa) + (ϕb) and the first + sign refers to the
binary operator in M while the second + sign refers to the binary operator
in M ′ . And if we had a different symbol for the binary operator in M ′ , say
×, we would of course say that ϕ : M → M ′ is an homomorphism if for all
a, b ∈ M, ϕ (a + b) = (ϕa) × (ϕb). In the present case, we denote the binary
operator in the magnitude space R>1 by · and so by a homomorphism R+
into R>1 we mean a function ϕ : R+ → R>1 such that for all x, y ∈ R+ ,
ϕ (x + y) = (ϕx) · (ϕy).
Having shown that R>1 is a continuous magnitude space, we know that
for each y ∈ R>1 there exists a unique embedding of R+ into R>1 which
maps 1 into y by Theorem 12.8. The assumptions of Section 16 are satisfied
with M = R+ and M ′ = R>1 . And, in accordance with Section 16, for
each x ∈ R>1 we define Ψx to be the unique embedding of R+ into R>1
which maps 1 into x. For y ∈ R+ and x ∈ R>1 we then have, as before, a
product yx = (Ψx) y ∈ R>1 . To avoid confusing this definition of yx with
multiplication in R+ we make the following definition.
Definition 19.2 If x ∈ R>1 then we denote by Ψx the unique embedding of
R+ → R>1 which maps 1 into x. And for y ∈ R+ , we denote (Ψx) y by xy .
Theorem 19.4 (x1 · x2 )y = xy1 · xy2
Proof Theorem 16.3.
Theorem 19.5 xy1 +y2 = xy1 · xy2
Proof Theorem 16.4.

47
References
[1] Edmund Landau, Grundlagen der Analysis, Akademische Verlagsge-
sellschaft M.B.H., Leipzig, 1930.

[2] Howard Stein, “Eudoxus and Dedekind: on the Ancient Greek Theory of
Ratios and its Relation to Modern Mathematics,” Synthese 84:163-211,
1990.

[3] Otto Hölder, “Die Axiome Quantität und die Lehre vom Mass,” Berichte
über die Verhandlungen der Königlich Sächsischen Gesellschaft der Wis-
senschaften zu Leipzig, Mathematisch-Physikaliche Classe, Bd. 53, S.1-64,
1901.

[4] Richard Dedekind, Was sind und was sollen die Zahlen?, Friedrich
Vieweg & Sohn, Braunschweig, 1888.

[5] Robert Simson, The Elements of Euclid; the First Six Books together with
the Eleventh and Twelfth, Sixth Edition, Edinburgh, 1781.

48