(2) Symmetry: If AB CD, then CD AB.

This is a consequence of (C2): Given AB CD, and writing AB AB by

(3) Transitivity: If AB CD and CD EF, then AB EF. This follows by first using symmetry to show CD AB, and
then applying (C2). Notice that Hilbert's formulation of (C2) was a clever way of including symmetry and
transitivity in a single statement.

The third axiom (C3) is the counterpart of Euclid's second common notion, that "equals added to equals
are equal." Let us amplify this by making a precise definition of the sum of two segments, and then
showing that sums of congruent segments are congruent.

Definition

Let AB and CD be two given segments, Choose an ordering A, B of the end- points of AB. Let r be the ray
on the line l= AB consisting of B and all the points of l on the other side of B from A. Let E be the unique
point on the ray r (whose existence is given by (C1)) such that CD BE.

We then define the segment AE to be the sum of the segments AB and CD, depending on the order A, B,
and we will write AE = AB + CD.

Proposition 8.2 (Congruence of sums)

Suppose we are given segments AB A'B' and CD C'D'. Then AB + CD A'B'+C'D'.

Proof Let E' be the point on the line A'B' defining the sum A'E' = A'B' + C'D'. Then A *B*E by construction
of the sum AB + CD, because E is on the ray from B opposite A. Similarly, A'* B'* E!. We have AB A'B by
hypothesis. Furthermore, we have CD C'D' by hypothesis, and CD= BE and C'D' B'E' by construction of E
and E'. From (8.1) we know that congruence is an equivalence relation, so BE BE. Now by (C3) it follows
that AE A'E' as required.

Note: Since the segment AB is equal to the segment BA, it follows in particular that the sum of two
segments is independent of the order A, B chosen, up to congruence. Thus addition is well-defined on
congruence equivalence classes of line segments. So we can speak of addition of line segments or
congruent segments without any danger (cf. also Exercise 8.1, which shows that addition ofline
segments is associative and commutative, up to congruence). Later (Section 19) we will also define
multiplication of segments and so create a field of segment arithmetic

Euclid's third common notion is that "equals subtracted from equals are equal." Bearing in mind that
subtraction does not always make sense, we can interpret this common notion as follows.

Proposition 8.3

Given three points A, B, C on a line such that A * B* C, and given points E, F on a ray originating fom a
point D, suppose that AB DE and AC DF. Then E will be between D andE, and BC EF. (We regard BC as the
difference of AC and AB)

Proof Let F be the unique point on the ray originating at E, opposite to D, such that BC EF'. Then from AB
DE and BC EF' we conclude by (C3) that AC DF. But F and F' are on the same ray from D (check!) and also
AC DE, so by (C2) and the uniqueness part of (C1), we conclude that F = F. It follows that D * E* Fand BC
EF, as required.
Note the role played by the uniqueness part of (C1) in the above proof. We can regard this uniqueness
as corresponding to Euclid's fifth common notion, "the whole is greater than the part." Indeed, this
statement could be interpreted as meaning, if A* B*C, then AB cannot be congruent to AC. And indeed,
this follows from (C1), because B and C are on the same ray from A, and if AB AC, then B and C would
have to be equal by (C1).

So we see that Euclid's common notions, at least in the case of congruence of line segments, can be
deduced as consequences of the new axioms (C1)-(C3). Another notion used by Euclid without definition
is the notion of inequality of line segments. Let us see how we can define the notions of greater and
lesser also using our axioms.

Definition

Let AB and CD be given line segments. We will say that AB is less than CD, written AB < CD, if there exists
a point E in between C and D such that AB CE. In this case we say also that CD is greater than AB, written
CD> AB.

In the next proposition, we will see that this notion of less than is compatible with congruence, and gives
an order relation on congruence equivalence classes of line segments.

Proposition 8.4

(a) Given line segments AB A'B' and CD C'D', then AB < CD if and only if A'B' <C'D'.

(b) The relation < gives an order relation on line segments up to congruence, in the following sense:

(i) IFAB< CD, and CD < EF, then AB < EF.

(ii) Given two line segments AB, CD, one and only one of the three following conditions holds: AB < CD,
AB CD, AB > CD.

Proof (a) Given AB A'B' and CD C'D', suppose that AB < CD. Then there is a point E such that AB = CE and
C* E*D. Let E' be the unique point on the ray CD' such that CE C'E'. It follows from (8.3) that C'* E'* D'.
Furthermore, by transitivity of congruence, A'B' C'E', So A'B' < C'D' as required. The "if and only if"
statement follows by applying the same argument starting with A'B'<C'D'.

(b) (i) Suppose we are given AB<CD and CD < EF. Then by definition, there is a point Xe CD such that AB
CX, and there is a point Ye EF such that CDEY. Let Ze EF be such that CX EZ. Then by (8.3) we have E * Z*
Y. It follows that E + Z* F (Exercise 7.1) and that AB EZ. Hence AB < EF as required.

(ii) Given line segments AB and CD, let E be the unique point on the ray CD for which AB CE. Then either
D =E or C* E *D or C*D*E. We cannot have D *C*E because D and E are on the same side of C. These
conditions are equivalent to AB CD, or AB < CD, or AB > CD, respectively, and one and only one of them
must hold.

9 Axioms of congruence for Angles

Recall that we have defined an angle to be the union of two rays originating at the same point, and not
lying on the same line. We postulate an undefined notion of congruence for angles, written , that is
subject to the following three axioms:
C4. Given an angle Z BAC and given a ray DF , there exists a unique ray DE, on a given side of the line DF,
such that A LBAC L EDE.

C5. For any three angles a,ß. , if u ß and z y, then ß y. Every angle is congruent to it self.

C6. (SAS) Given triangles ABC and DEF, suppose that AB DE and AC DF, and L BAC LEDF. Then the two
triangles are congruent, namely, BC EF, LABC L DEF and LACB LDFE.

Note that Hilbert takes the existence of an angle congruent to a given one (C4) as an axiom, while Euclid
proves this by a ruler and compass constrUC- tion (I.23). Since Hilbert does not make use of the
compass, we may regard this axiom as a tool, the "transporter of angles," that acts as a substitute for
the compass. As with (C2), we can use (C5) to show that congruence is an equivalence relation.

Proposition 9.1

Congruence of angles is an equivalence relation. Proof The proof is identical to the proof of (8.1), using
(C5) in place of (C2). As in the case of congruence of line segments, we would like to make sense of
Euclid's common notions in the context of congruence of angles. This proposition (9.1) is the analogue of
the first common notion, that "things equal to the same thing are equal to each other." The second
common notion, that "equals added to equals are equal," becomes problematic in the case of angles,
because in general we cannot define the sum of two angles.

If Z BAC is an angle, and if a ray AD lies in the interior of the angle L BAC, then we will say that the angle
LBAC is the sun of the angles L DAC and z L BAD. However, if we start with the two given angles, there
may not be an angle that is their sum in this sense. For one thing, they may add up to a straight line. or
"two right angles" as Euclid says, but this is not an angle. Or their sum may be greater than 180°, in
which case we get an angle, but the two original angles will not be in the interior of the new angle. So
we must be careful how we state results having to do with sums of angles.

Note that we do not have an axiom about congruence of sums of angles analogous to the axiom (C3)
about addition of line segments. That is because we can prove the corresponding result for angles. But
in order to do so, we will need (C6).

Hilbert's use of (C6) = (SAS) as an axiom is a recognition of the insufficiency of Euclid's proof of that
result (I4) using the method of superposition. To justify the method of superposition by introducing
axioms allowing motion of figures in the plane would be foreign to Euclid's approach to geometry, so it
seems prudent to take (C6) as an axiom. However, we will show later (17.5) that the (SAS) axiom is
essentially equivalent to the existence of a sufficiently large group of rigid motions of the plane. The
axiom (C6) is necessary, since it is independent of the other axioms (Exercise 9.3). This axiom is
essentially what tells us that our plane is homogeneous: Geometry is the same at different places in the
plane. Now let us show how to deal with sums of angles and inequalities among angles based on these
axioms.

Definition

If L BAC is an angle, and if D is a point on the line AC on the other side of A from C, then the angles L BAC
and L BAD are supplementary.

Proposition 9.2
If LBAC and L BAD are supplementary angles, and if LB'A'C' and LB'A'D' are supplementary angles, and if
L BAC LBA'C', then also L BAD LB'A'D'.

Proof Replacing B',C, D' by other points on the same rays, we may assume that AB A'B', ACA'C, and AD
A'D'. Draw the lines BC, BD, B'C', and B'D'. First we consider the triangles ABC and A'B'C. By hypothesis
we have AB A'B' and AC A'C' and L BAC LB'A'C. So by (C6) we conclude that the triangles are congruent.
In particular, BC B'C' and BCA LB'C'A'.

Next we consider the triangles BCD and B'C'D'. Since AC A'C and AD A'D', and C*A * D and C'* A'*D', we

conclude from (C3) that CD C'D'. Using BC B'C' and L BCA LB'C'A' proved above, we can apply (C6) again

to see that the triangles BCD and B'C'D' are congruent. In particular, BD B'D' and L BDA = LB'D'A'.

Now we consider the triangles BDA and B'D'A'. From the previous step we have BD B'D' and L BDA
LB'D'A'. But by hypothesis we have DA D'A'. So a third application of (C6) shows that the triangles BDA
and B'D'A' are congruent. In particular, Z BADLBA'D', which was to be proved.

Note: We may think of this result as a replacement for (L13), which says that the angles made by a ray
standing on a line are either right angles or are equal to two right angles. We cannot use Euclid's
statement directly, because in our terminology, the sum of two right angles is not an angle. However, in
applications, Euclid's (1.13) can be replaced by (9.2). So for example, we have the following corollary.

Corollary 9.3 Vertical angles are congruent.

Proof Recall that vertical angles are defined by the opposite rays on the same two lines. The vertical
angles z and a are each supplementary to ß, and ß is congruent to itself, so by the proposition, a and a'
are congruent.

Proposition 9.4 (Addition of angles)

Suppose ZL BAC is an angle, and the ray AD is in the interior of the angle L BAC. Suppose LD'A'C' LDAC,
and LB'A'D' L BAD, and the rays A'B and AC are on opposite sides of the line A'D'. Then the rays A'B' and
A'C form an angle, and LB'A'C' LBAC, and the ray A'D' is in the interior of the angle LB'A'C'. For short, we
say "sums of congruent angles are congruent."

Proof Draw the line BC. Then the ray AD must meet the segment BC, by the crossbar theorem (7.3).
Replacing the original D by this intersection point, we may assume that B, D,C lie on a line and B* D*C.
On the other hand, replacing B', C', D' by other points on the same rays, we may assume that AB A'B',
and AC A'C, and AD A'D'. We also have L BAD LB'A'D' and L DAC LD'A'C by hypothesis.

By (C6) we conclude that the triangles ABAD and AB'A'D' are congruent. In particular, BD B'D' and L BDA
B'D'A Again by (C6) we conclude that the triangles ADAC and AD'A'C are congruent. In particular, DC D'C
and LADCLA'D'C'. Let E' be a point on the line B'D' with B'* D' * E'. Then LA'D'E' is supplementary to
LA'D'B, which is congruent to LADB. So by (9.2) and transitivity of congruence, we find that LA'D'E'
LA'D'c'. Since these angles are on the same side of the line A'D', we conclude from the uniqueness part
of (C4) that they are the same angle. In other words, the three points B', D', and C' lie on a line.

Then from (C3) we conclude that BC B'C'. Since LABD LA'B'D' by the first congruence of triangles used in
the earlier part of the proof, we can apply (C6) once more to the triangles ABC and A'B'C'. The
congruence of these triangles implies BAC LBA'C' as required. Since B', D', and C' are collinear and D'A'C'
is an angle, it follows that A', B', C' are not collinear, so B'A'C' is an angle. Since B' and C' are on opposite
sides of the line A'D', it follows that B'* D'* C', and so the ray A'D' is in the interior of the angle LB'A'C', as
required. Next, we will define a notion of inequality for angles analogous to the inequality for line
segments in Section 8.

Definition

Suppose we are given angles Z BAC and LEDF. We say that L BAC is less than LEDF, written Z BAC < LEDE,
if there exists a ray DG in the interior of the angle L EDF such that BAC LGDF. A/ In this case we will also
say that L EDF S greater than L BAC.

Proposition 9.5

(a) If a and B ß, then a <ß <B.

(b) Inequality gives an order relation on angles, up to congruence. In other words:

(i) If z<B and ß < 7, then a <y.

(ii) For any two angles a and B. one and only one of the following holds: a <B: 2B; a>B.

Proof The proofs of these statements are essentially the same as the corresponding statements for line
segments (8.4), so we will leave them to the reader.

Definition

A right angle is an angle a that is congruent to one of its supplementary angles B

Note: In this definition, it does not matter which supplementary angle to we consider, because the two
supplementary angles to a are vertical angles, hence congruent by (9.3). Two lines are orthogonal if they
meet at a point and one, hence all four, of the angles they make is a right angle.

Proposition 9.6

Any two right angles are congruent to each other. Suppose that x = L CAB and a'=

Proof, LC'A'B' are right angles. Then they will be congruent to their supplementary angles ß, B, by
definition. Suppose a and o' are not congruent. Then by (9.5) either a <a or a' <u. Suppose, for example,
a<d. Then by definition of inequality there is a ray A'E in the in- terior of angle ' such that a LEA'B'

It follows (check!) that the ray AC is in the interior of LE'A'D', so that p'<LE'A'D'. But LE'A'D' is
supplementary to L E'A'B', which is congruent to a, so by (9.2), LE'A'D' ß. Therefore, ß' <B. But aß and 'p',
so we conclude that a' <, which is a contradiction.

Note: Thus the congruence of all right angles can be proved and does not need to be taken as an axiom
as Euclid did (Postulate 4). The idea of this proof already appears in Proclus

10 Hilbert Planes

We have now introduced the minimum basic notions and axioms on which to found our study of
geometry.
Definition

A Hilbert plane is a given set (of points) together with certain subsets called lines, and undefined notions
of betweenness, congruence for line segments, and congruence for angles (as explained in the
preceding sections) that satisfy the axioms (I1)-(13), (Bl)-(B4), and (C1)-(C6). (We do not include the
parallel axiom (P).)

We could go on immediately and introduce the parallel axiom and axioms of intersection of lines and
circles, so as to recover all of Euclid's Elements, but it seems worthwhile to pause at this point and see
how much of the geometry we can develop with this minimal set of axioms. The main reason for doing
this is that the axioms of a Hilbert plane form the basis for non-Euclidean as well as Euclidean geometry.
In fact, some people call the Hilbert plane neutral geometry, because it neither affirms nor denies the
parallel axiom.

In this section we will see how much of Euclid's Book I we can recover in a Hilbert plane. With two
notable exceptions, we can recover everything that does not make use of the parallel postulate. Let us
work in a given Hilbert plane. Euclid's definitions, postulates, and common notions have been replaced
by the undefined notions, definitions, and axioms that we have discussed so far (excluding Playfair's
axiom). We will now discuss the propositions of Euclid, Book I.

The first proposition (I.1) is our first exception! Without some additional axiom, it is not clear that the
two circles in Euclid's construction will actually meet. In fact, the existence of an equilateral triangle on a
given segment does not follow from the axioms of a Hilbert plane (Exercise 39.31). We will partially fill
this gap by showing (10.2) that there do exist isosceles triangles on a given segment. Euclid's
Propositions (1.2) and (1.3) about transporting line segments are effectively replaced by axiom (C1).
Proposition (I.4), (SAS), has been replaced by axiom (C6).

Proposition (1.5) and its proof are ok as they stand. In other words, every step of Euclid's proof can be
justified in a straightforward manner within the framework of a Hilbert plane. To illustrate this process
of reinterpreting one of Euclid's proofs within our new axiom system, let us look at Euclid's proof step by
step. Proof of (1.5) Let ABC be the given isosceles triangle, with AB AC(Congruent

line segments). We must prove that the base angles LABC and LACB are congruent. "In BD take any point
F" This is possible by axiom (B2). "On AE cut off AG equal to AF." This is possible by (C1). Now AC AB and
AF AG, and thw enclosed angle Z BAC is the same, so the triangles AAFC and AAGB are congru- ent by a
direct application of (C6). So FC GB and LAFC LAGB and LACF LABG.

Since "equals subtracted from equals are equal," referring in this case to congruence of line segments,
we conclude from (8.3) that BF CG. Then by another application of (C6), the triangles AFBC and AGCB are
congruent. It follows that L CBG L BCF. Now by subtraction of congruent angles (Exercise 9.1l), the base
angles ZABC and LACB are congruent, as required. (We omit the proof of the second assertion, which
follows similarly.)

At certain steps in this proof we need to know something about betweenness, which can also be
formally proved from our axioms. For example, in order to subtract the line segment AB from AF, we
need to know that B is between A and F. This follows from our choice of F. At the last step, subtracting
angles, we need to know that the ray BC is in the interior of the angle LABG. This follows from the fact
that Cis between A and G. So in the following, when we say that Euclid's proof is ok as is, we mean that
each step can be justified in a natural way, without having to invent additional steps of proof, from
Hilbert's axioms and the preliminary results we established in the previous sections. Looking at (I.6), the
converse of (L.5), everything is ok except for one doubtful step at the end. Euclid says, "the triangle DBC
is equal to the triangle ACB, the less to the greater; which is absurd. ”it is not clear what this means,
since we have not defined a notion of inequality for triangles. However, a very slight change will give a
satisfactory proof. Namely, from the congruence of the tri- angles ADBC AACB, it follows that LDCB L
ABC. But also LABC LACB by hypothesis. So DCB LACB, "the less to the greater," as Euclid would say. For
us, this is a contradiction of the uniqueness part of axiom (C4), since there can be only one angle on the
same side of the ray CB congruent to the angle ACB. We conclude that the rays CA and CD are equal, so
A=D, and the triangle is isosceles, as required.

Proposition (.7), as we have mentioned before, needs some additional justification regarding the relative
postions of the lines, which can be supplied from our axioms of betweenness (Exercise 9.4).

For (I.8), (Sss), we will need a new proof, since Euclid's method of superposition cannot be justified from
our axioms. The following proof is due to Hilbert.

Proposition 10.1 (SSs)

If two triangles ABC and A'B'C' have their respective sides equal, namely AB AC A'C, and BC = B'C, then
the two triangles are congruent.

Proof Using (C4) and (C1), construct an angle C’A’B’’ on the order side of the ray A’C’ from B’ that is
congruent to L BAG, and make A'B" congruent to AB. Then AB A'B" by construction, AC A'C by
hypothesis, and L BAC B"A'C by construction, so by (C6) The triangle ABC is congruent to be triangle
A’B’’C’. it follows that BC B’’C

Draw the line B'B. Now A'B' AB A'B", so by transitivity, AB A'B". Thus the triangle A'B'B" is isosceles, and
so by (1,5) its base angles a'b’b and a’b’b are congruent. Similarly, B'C' B"C, so the triangle C'B'B" is
isosceles, and its base angles LB"B'C' and LB'B'C' are congruent. addition of congruent angles (9.4) This
latter triangle was shown congruent to AABC, so LA'B"C' LABC. Now by transitivity of congruence, LABC
LA'B'C, so we can apply (C6) again to conclude that the two triangles are congruent.

Note: This proof and the accompanying figure are for the case where is analogous, and the case where
one of A' or C' lies on the line B'B" is easier., and left to the reader. Starting with the next proposition
(1.9) we have a series of constructions with because we have not yet added axioms to ensure that lines
and circles will meet A'B'.

when they ought to (cf. Section 11), However, we can reinterpret these propositions as existence
theorems, and these we can prove from Hilbert's axioms. Since we do not have the equilateral triangles
that Euclid constructed in (I.1), we will prove the existence of isosceles triangles, and we will use them
as a substitute for equilateral triangles in the following existence proofs.

Proposition 10.2 (Existence of isosceles triangles)

Given a line segment AB, there exists an isosceles triangle with base AB. Proof Let AB be the given line
segment. Let C be any point not on the line AB (axiom (13). Consider the triangle AABC. If the angles at A
and B are equal, then AABC is isosceles (I.6). If not, then one angle is less than the other. Suppose LCAB <
LCBA. Then there is a ray BE in the interior of the angle L CBA such that L CAB EBA.
By the crossbar theorem (7.3) this ray must meet the opposite side AC in a point D. Now the base angles
of the triangle DAB are equal, so by (L6) it is isosceles.

Note: It would not suffice to construct equal angles at the two ends of the interval, because without the
parallel axiom, even if the angles are small, there is no guarantee that the two rays would meet. Now let
us return to Euclid. We interpret (L9) as asserting the existence of an angle bisector. We use the same
method as Euclid, except that we use (10.2) to give the existence of an isosceles triangle ADEF where
Euclid used an equilateral triangle. We may assume that this isosceles triangle is constructed on the
opposite side of DE from A. Then Euclid's proof, using (SSS), shows that LDAF LEAF. It is not obvious from
the construction that the ray AF is in the interior of the angle LDAE, but it does follow from the
conclusion: For if AF were not in the interior of the angle, then AD and AE would be on the same side of
AF, and in that case the congruence of the angles DAF LEAF would contradict the uniqueness in axiom
(C4).

For (1.10) to bisect a given line segment, we again use (10.2) to construct isosceles triangle instead of an
equilateral triangle. The rest of Euclid's proof then works to show that a midpoint of the segment exists.

For (L11) we can also use (10.2) to construct a line perpendicular to a line at a point. By the way, this
also proves the existence of right angles, which is not obvious a priori.

For (L.12), to drop a perpendicular from a point C to a line not containing C, Euclid's method using the
compass does not work in a Hilbert plane. We need a new existence proof (see Exercise 10.4).

Proposition (1.13) has been replaced by the result on congruence of supple- mentary angles (9.2), and
(L14) is an easy consequence (Exercise 10.7). The congruence of vertical angles (I.15) has already been
mentioned above (9.3). The theorem on exterior angles (1.16) is sufficiently important that we will
reproduce Euclid's proof here, with the extra justifications necessary to make it work.

Proposition 10.3 (Exterior angle theorem (1.16))

In any triangle, the exterior angle is greater than either of the opposite interior angles. Proof Let ABC be
the given triangle. We will show that the exterior angle LACD is greater than the opposite interior angle
at A. Let E be the midpoint of AC (L10), and extend BE to F so that BE EF (axiom (C1). Draw the line CE.
Now the vertical angles at E are equal (1.15), so by SAS (C6), the triangles AABE and ACFE are congruent.
Hence LALECE.

To finish the proof, that is, to show that Z ECF is less than LACD, we need to know that the ray CF is in
the interior of the angle LACD. This we can prove based on our axioms of betweenness, Since D is on the
side BC of the triangle extended, B and D are on opposite sides of the line AC. Also, by construction of F,
we have B and F on opposite sides of AC. So from the plane separation property (7.1) it follows that D
and Fare on the same side of the line AC.

Now consider sides of the line BC. Since B E*F, it follows that E and F are on the same side of BC. Since A
EG, it follows that A and E are on the same side of AC. By transitivity (7.1) it follows that A and Fare on
the same side of the line BC = CD, So by definition, F is in the interior of the angle L ACD, and hence the
ray CF is also. Therefore, by definition of inequality for angles, L BAC is less than LACD, as required.

Propositions (1.17)-(1.21) are all ok as is, except that we should reinterpret the statement of (L17).
Instead of saying "any two angles of a triangle are less than two right angles," which does not make
sense in our system, since "two right angles" is not an angle, we simply say; if a and B are any two angles
of a triangle, then z is less than the supplementary angle of ß.

Proposition (1.22) is our other exception. Without knowing that two circles intersect when they ought
to, we cannot prove the existence of the triangle required in this proposition. In fact, we will see later
(Exercise 16.11) that there are Hilbert planes in which a triangle with certain given sides satisfying the
hypotheses of this proposition does not exist! The next proposition (1.23), which Euclid proved using
(1.22), is replaced by Hilbert's axiom (C4), the "transporter of angles." The remaining results that Euclid
proved without using the parallel postulate are ok as is in the Hilbert plane: (1.24), (L.25), (L26) =(ASA)
and (AAS), (L27) "alternate interior angles equal implies parallel," and even the existence of par- allel
lines (L31). Summing up, we have the following theorem.

Theorem 10.4

All of Euclid’s propositions (1.1) through (1.28), except (L1) and (1.22), can be proved in an arbitrary
Hilbert plane, as explained above.

Constructions with Hilbert's Tools

Euclid used ruler and compass constructions to prove the existence of various objects in his geometry,
such as the midpoint of a given line segment. We used Hilbert's axioms to prove corresponding
existence results in a Hilbert plane. However, we can reinterpret these existence results as constructions
if we imagine tools corresponding to certain of Hilbert's axioms. Thus (11), the existence of a line
through two points, corresponds to the ruler. For axiom (C1), imagine a tool, such as a compass with two
sharp points (also called a pair of dividers), that acts as a transporter of segments. For axiom (C4),
imagine a new tool, the transporter of angles, that can reproduce a given angle at a new point. It could
be made of two rulers joined with a stiff but movable hinge. We call these three tools, the ruler, the
dividers, and the transporter of angles, Hilbert's tools. We also allow ourselves to pick points (using (13)
and (B2)) as required.

Now we can regard (10.2) as a construction of an isosceles triangle using Hilbert's tools. Counting steps,
with one step for each use of a tool, we have the construction as follows: Given a line segment AB. Pick
C not on the line AB.

1. Draw line AC.

2. Draw line BC. Suppose L CAB is less than LCBA.

3. Transport L CAB to LABE, get point D. Then ABD is the required isosceles triangle.

11 Intersections of Lines and Circles

In this section we will discuss the intersections of lines and circles in the Hilbert plane, and we will
introduce the further axiom (E), which will guarantee that lines and circles will intersect when they
"ought" to. With this axiom we can justify Euclid's ruler and compass constructions in Book I and Book
IIL. We work in a Hilbert plane (Section 10) without assuming the parallel axiom (P). Because of (10.4)
we can use Euclid's results (L.2)-(1.28) (except (1.22)) in our proofs.

Definition
Given distinct points O, A, the circle r with center O and radius OA is the set of all points B such that OA
OB. The point O is the center of the circle. The segment OA is a radius. From this definition it is clear that
a circle always has points. The point A is on the circle. Moreover, if l is any line through 0, then by axiom
(C1) there will be exactly two points on the line l, one on each side of O, lying on the circle. However, it is
not obvious from the definition that the center is uniquely determined by the set of points of the circle.

Proposition 11.1

Let l' be a circle with center O and radius OA, and let be a circle with center 0" and radius O'A'. Suppose
as point sets. Then O = 0', In other words the center of a circle is uniquely determined

Proof Suppose O 0'. Then we consider the line I through O and 0'. Since it passes through the center O of
I, it must meet I in two points C,D, satisfying C+ O+D and OC OD. Since T= ', the points C,D are also on ,
so we have O'C O'D and C+O'D. We do not know which of O or O' is closer to C, but the two cases are
symmetric, so let us assume C*0* O'. In this case we must have O 0'*D by the properties of betweenness
(!). Then OC < 0'C 0'D < OD, which is impossible, since OC 0D, Hence O0'.

Now that we know that the center of a circle is uniquely determined, it makes sense to define the inside
and the outside of a circle.

Definition

Let I be a circle with center O and radius OA. A point B is inside r (or in the interior of r) if B =0 or if OB <
OA. A point C is outside r (or exterior to I) if OA < OC.

Definition

We say that a line l is tangent to a circle T if l and r meet in just one point A, We say that a circle I is
tangent to another circle & if T and & have just one point in Common.

This definition of tangent circles is a little different from Euclid's: His definition of two circles touching is
that they meet in a point but do not cut each other. Since it is not clear what he means by "cut," we
prefer the definition above, and we will prove that these notions of tangency have the usual properties.

Proposition 11.2

Let T be a circle with center O and radius OA. The line perpendicular to the radius OA at the point A is
tangent to the circle, and (except for the point A) lies entirely outside the circle. Conversely, if a line l is
tangent to I' at A, then it is perpendicular to OA. In particular, for any point A of a circle, there exists a
unique tangent line to the circle at that point.

Proof First, let l be the line perpendicular to OA at A. Let B be any other point on the line 1. Then in the
triangle OAB, the exterior angle at A is a right angle, so the angles at O and at B are less than a right
angle (1.16). It follows (L19) that OB > OA, so B is outside the circle. Thus I meets only at the point A, so
it is a tangent line.

Now suppose that I is a line tangent to ľ at A. We must show that I is perpendicular to OA. It cannot be
equal to OA, because that line meets T in another point opposite A. So consider the line from 0,
perpendicular to l, meeting I at B. If B #A, take a point C on the other side of B from A, so that AB BC
(axiom (C1). The AOBA AOBC by SAS, so we have OA 0C, and hence C is also on I. Since C A, this is a
contradiction. We conclude that B= A, and so I is perpendicular to OA.

Corollary 11.3

If a line l contains a point A of a circle I, but is not tangent to I, then it meets I' in exactly two points.

Proof If l is not tangent to I at A, then it is not perpendicular to OA, in which case, as we saw in the
previous proof, it meets I in another point C. We must show that I cannot contain any further points of r.
For if D were another point of I on I, then OD 0A,OB is congruent to itself, so by (RASS) (Exercise l0.9) we
would have AODB AOAB. Then AB = BD, so by axiom (C1) D must be equal to A or C.

Proposition 11.4

Let O,0',A be three distinct collinear points. Then the circle T with center O and radius OA is tangent to
the circle I" with center O' and radius OA. Conversely, if two circles I, r' are tangent at a point A, then
their centers O,0' are collinear with A

Proof Let 0,0', A be collinear. We must show that the circles and " have no further points in common be-

sides A. The argument of (11.1) shows that there is no other point on the line O0' that lies on both I and
r'. So sup- pose there is a point B not on 00' lying on both T and r'. We divide into two cases depending
on the relative position of O, 0', and A.

Case 1 O*0'*A. Since OA = OB, LOAB LOBA. Also, since O'A = O'B, LO'AB LO'BA, using (L5). 1t follows that
LOBA LO'BA, which contradicts axiom (C4). (This argument also applies if o' *0sA.)

Case 2 OA0'. Again using (L5) we find that LOAB = L OBA and LO'AB = LO'BA. But the two angles at A are
supplementary, so it follows that the two angles at B are supplementary (9.2). But then0, B, and 0'
would be collinear (L14), which is a contradiction.

Conversely, suppose that I and " are tangent at A, and suppose that 0,0, A are not collinear. Then we let
AC be perpendicular to the line 00', and choose B on the line AC on the other side of 00' with AC BC. It
follows by congruent triangles that OA OB and O'A0'B, so B also lies on r and r', contradicting the
hypothesis r tangent to ". We conclude that 0, 0, A are collinear.

Corollary 11.5

If two circles meet at a point A but are not tangent, then they have exactly two points in common,

Proof We have seen above that if they are not tangent, then 0,0,A are not collinear, and they meet in an
additional point B. We must show there are no further intersection points. If D is a third point on T and ',
then OD OA and O'D 0'A, so by (L7), D must be equal to A or B.

In the above discussion of lines and circles meeting, we have seen that a line and a circle, or two circles,
can be tangent (meeting in just one point), or if they meet but are not tangent, they will meet in exactly
two points. There is nothing here to guarantee that a line and a circle, or two circles, will actually meet if
they are in a position such that they "ought" to meet according to the usual intuition. For this we need
an additional axiom (and we will see later (17.3) that this axiom is independent of the axioms of a
Hilbert plane).
E. (Circle-circle intersection property).

Given two circles r,A, if A contains at least one point inside I, and A contains at least one point outside ,
then I' and A will meet. (Note: It follows from Exercise 11.3 and (11.5) that they will then meet in exactly
two points.)

Proposition 11.6 (Line-circle intersection property LCI)

Ina Hilbert plane with the extra axiom (E), if a line l contains a point A inside a circle T, then l will meet T
(necessarily in two points, because of(11.2) and (11.3).

Proof Suppose we are given the line L with a point A inside the circle T. Our strategy is to construct
another circle A, show that A meets I, and then show that the intersection point also lies on l. Let OB be
the perpendicular from O to l (if O is on the line l, we already know that I meets l by (C1). Find a point O'
on the other side of l from 0, on the line OB, with 0'B OB. Let A be the circle with center 0' and radius r=
radius of r. (Here we denote by r the congruence equivalence class of any radius of the circle F.)

Now the line 00' meets A in two points C, D, labeled such that 0, C are on the same side of O, and D on
the opposite side. By hypothesis, A is a point on , inside I. Hence OA <r. In the right triangle OAB, using
(I.19) we see that OB < OA, so OB < r. It follows that O'B<r=0'C, so 0' and C are on opposite sides of l.
Hence ,C are on the same side of I. We wish to show that C is inside I. There are two cases.

Case 1 If O*C* B, then OC < OB <, so C is inside r.

Case 2 If C* 0*B, then also C*0*0', so OC < O'C = r, and again we see that C is inside I. On the other hand,
the point D satisfies O * O'* D, so OD > 0D=r, so D is outside r.

Now we can apply the axiom (E) to conclude that I meets A at a point E. We must show that E lies on I.
We know that OE r o'E and OB O'B by construction, and BE is equal to it self. so by (Sss) AOEB AO'EB. It
follows that the angles at B are equal, so they are right angles, so BE is equal to the line , and so E lies on
l and , as required.

Remark 11.6.1

We will see later (16.2) that in the Cartesian plane over a field, the circle-circle intersection property is
equivalent to the line-circle intersection property. In an arbitrary Hilbert plane, the equivalence of these
two statements follows from the classification theorem of Pejas (cf. Section 43), but I do not know any
direct proof.

Using the new axiom (E) we can now justify Euclid's first construction (L1), the equilateral triangle. Given
the segment AB, let r be the circle with center A and radius AB. Let A be the circle with center B and
radius BA. Then A is on the circle A, and it is inside r because it is the center of T. The line AB meets A in
another point D, such that A * B* D, Hence AD> AB, so D is outside I.

Thus A contains a point inside and a point outside T., so it must meet in a point C. From here, Euclid's
proof shows that AABC is an equilateral triangle.

In a similar way one can justify Euclid's other ruler and compass constructions in Book I. Several of them
depend only on using the equilateral triangle constructed in (1.1). For (1.12) and (L.22) see Exercise 11.4
and Exercise 11.5. Thus we have the following theorem.
Theorem 11.7

Euclid's constructions (1.1) and (L22) are valid in a Hilbert plane with the extra axiom (E). We can also
justify the results of Euclid, Book III, up through (IIL19) (note that (IL.20) and beyond need the parallel
axiom). The statements (IIL10), (IIL11). (IIL12) about circles meeting and (IIL16). (IIL.18), (III.19) about
tangent lines can be replaced by the propositions of this section. (We omit the controversial last phrase
of (III.16) about the angle of the semicircle, also called a horned angle or angle of contingency, because
in our treatment we consider only angles defined by rays lying on straight lines) In (IIL.14) Euclid uses
(1.47) to prove (RASS), but that is not necessary: One can prove it with only the axioms of a Hilbert plane
(Exercise 10.9). For (II17), to draw a tangent to a circle from a point outside the circle, we need the line-
circle intersection property (11.6) and hence the axiom (E). (Note that the other popular construction of
the tangent line using (IIIL31) requires the parallel axiom!) The other results of Book II, up to (IL.19)
(except (IIIL17), are valid in any Hilbert plane, provided that we assume the existence of the intersection
points of lines and circles used in the statement and proofs, and their proofs are ok as is, except as
noted.

Theorem 11.8

Euclid's propositions (III.1) through (.19) are valid in ang Hilbert plane, except that for the constructions
(IL1) and (1I1.17) we need also the additional Axiom (E).

12 Euclidean Planes

Let us look back at this point and see how well Hilbert's axioms have fulfilled their goal of providing a
new solid base for developing Euclid's geometry. The major problems we found with Euclid's method
have been settled: Questions of relative position of figures have been clarified by the axioms of
betweenness; the problematic use of the method of superposition has been replaced by the device of
taking SAS as an axiom; the existence of points needed in ruler and compass constructions is guaranteed
by the circle-circle intersection property stated as axiom (E). Also, in the process of rewriting the
foundations of geometry we have formulated a new notion, the Hilbert plane, which provides a
minimum context in which to develop the beginnings of a geometry, free from the parallel axiom.
Hilbert planes serve as a basis both for Euclidean geometry, and also later, for the non-Euclidean
geometries.

In this section we will complete the work of earlier sections by showing how the addition of the parallel
axiom allows us to recover almost all of the first four books of Euclid's Elements. We will also mention
two more axioms, those of Archimedes and of Dedekind, which will be used in some parts of later
chapters.

Definition

A Euclidean plane is a Hilbert plane satisfying the additional axioms (E), the circle-circle intersection
property, and (P), Playfair's axiom, also called the parallel axiom. In other words, a Euclidean plane is a
set of points with subsets called lines, and undefined notions of betweenness and congruence satisfying
the axioms (I1)-(13), (B1)-(B4), (C1)-(C6), (E), and (P). The Euclidean plane represents our modern
formulation of the axiomatic basis for developing the geometry of Euclid's Elements.
We have already seen in Section 10 and Section 11 how to recover those results of Euclid's Books I and II
that do not depend on the parallel axiom. The first use of the parallel axiom is in (1.29). Since we have
replaced Euclid's fifth postulate by Playfair's axiom, we need to modify Euclid's proofs of a few early
results in the theory of parallels.

So for example, to prove (I.29) we proceed as follows. Given two parallel lines I, m, and a transversal line
n, we must show that the alternate interior angles x and ß are equal. If not, con- struct a line ' through A
making an angle x with n (axiom (C4). By (1.27), r' will be parallel to m. But then I and ' are two lines
through A parallel to m, so by (P), we must have l=I', hence z = .

Proposition (1.30) is essentially equivalent to (P). The existence of parallel lines (1.31) follows from (C4)
and (I.27) as mentioned before, so now we can reinterpret (I.31) in the stronger form that given a point
A not on a line I, there exists a unique parallel to I passing through A. The remaining propositions using
(P), namely (1.32)-(1.34), follow without difficulty. In particular, we have the famous (I32), that "the sum
of the angles of a triangle is equal to two right angles," though if we want to be scrupulous, we would
have to say that sum is not defined, and rephrase the theorem by saying that the sum of any two angles
of a triangle is supplementary to the third angle.

Theorem 12.1

Euclid's theory of parallels, that is, propositions (1.29)-(1.34), hod in any Hilbert plane with (P), hence in
any Euclidean plane.

Starting with (I.35), and continuing to the end of Book I and through Book II, is Euclid's theory of area.
Since Euclid does not define what he means by this new equality, we must presume that he takes it as
another undefined notion, which we call equal content, just as the notion of congruence for line
segments and angles were taken as undefined notions. Since Euclid freely applies the common notions
to this concept, we may say that he has taken the common notions applied to equal content as further
axioms, for example, "figures having equal content to a third figure have equal content to each other,"
or "halves of figures of equal content have equal content."

Hilbert showed that it is not necessary to regard the notion of equal content as an undefined notion
subject to further axioms. He shows instead that it is possible to define the notion of equal content for
figures (by cutting them up, rearranging, and adding and subtracting), and then prove the properties
suggested by Euclid's common notions. To be more precise, we have the following theorem.

Theorem 12.2 (Theory of area)

In a Hilbert plane with (P) there is an equivalence relation called equal content for rectilineal figures that
has the following properties:

(1) Congruent figures have equal content.

(2) Sums of figures with equal content have equal content.

(3) Differences of figures with equal content have equal content.

(4) Halves of figures with equal content have equal content.

(5) The whole is greater than the part.

We will prove this theorem in Chapter 5, (22.5), (23.1), (23.2). For the present you can either accept this
result as something to prove later, or (as Euclid implicitly did) you can regard equal content of figures as
another undefined notion, subject to the axioms that it is an equivalence relation and has these
properties

(1)-(6). For further discussion and more details about the exact meaning of a figure, the notions of sum
and difference, etc., see Section 22 and Section 23.

Using this theory of area, the remaining results (L35)-(1.48) of Book I follow without difficulty. Note in
particular the Pythagorean theorem (I.47), which says that the sum of the squares on the legs of a right
triangle have equal content with the square on the hypotenuse. Also, the results of Book II, (IL. 1) -(IL14),
phrased as results about equal content, all follow easily. Proposition (1.11), how to cut a line segment in
extreme and mean ratio, is used later in the construction of the regular pentagon. Only (I. 14), to
construct a square with content equal to a given rectilineal figure, uses the axiom (E).

Theorem 12.3

In a Hilbert plane with (P), using the theory of area (12.2), Euclid's propositions (L.35)-(1.48) and (11. 1) -
(1L14) can all be proved as he does, using the extra axiom (E) only for (.14). In particular, all these results
hold in a Euclidean plane.

In Book III, the first use of the parallel axiom is in (I.20), that the angle at the center of a circle
subtending a given arc is twice the angle on the circumference subtending the same arc. This result uses
(L.32), that the exterior angle of a triangle is equal to the sum of the two opposite interior angles, and
thus depends on the parallel axiom (P). The following propositions (III.21), (IIL.22), and then (II31)-
(II1.34) follow with no further difficulties. For the propositions (Il.23)-(II30) we need a notion of "equal"
segments of circles, a congruence notion that has not been defined by Euclid, though we can infer from
the proof of (II24) that it means being able to place one segment on the other by a rigid motion. Indeed,
if we take this as a definition of congruence, then the proofs of these results are all ok (Exercise 17.13).
The final propositions (III.35)-(IIIL37) make use of the theory of area for their statements, and depend on
the earlier area results from Books I and II.

Theorem 12.4

in Book III, Euclid's propositions (III.20)-(IIL.37) hold in any Euclidean plane. The last three (II.35)-(IIL37)
make use of the theory of area (12.2). Most of the results of Book IV require the parallel axiom (P), some
need circle-circle intersection (E), and some, notably (IV.10), (IV.11), require (P), (E), and the theory of
area. Thus we may regard the construction of the regular pentagon as the crowning result of the first
four books of the Elements, making use of all the results developed so far.

Theorem 12.5

All the propositions (1V. 1)-(IV. 16) of Euclid's Book IV hold in a Euclidean plane. We end this section
with a discussion of two further axioms that are not needed for Books I-IV, but will be used later. The
first is Archimedes axiom. A. Given line segments AB and CD, there is a natural number n such that n
copies of AB added together will be greater than CD. This axiom is used implicitly in the theory of
proportion developed in Book v, for example in Definition 4, where Euclid says that quantities have a
ratio when one can be multiplied to exceed the other. It appears explicitly in (X.1), in a form reminiscent
of the arguments of calculus: Given two quantities AB and CD, if we remove from AB more than its half,
and again from the remainder remove more than its half, and continue in this fashion, then eventually
we will have a quantity less than CD. In modern texts this would appear as the statement "given any & >
0, there is an integer n sufficiently large that 1/2" < e." Euclid applies this "method of exhaustion" to the
study of the volume of three- dimensional figures in Book XII. When he cannot compare solids by cutting
into a finite number of pieces and reassembling, he uses a limiting process where the solid is
represented as a union of a sequence of subsolids so that the remainder can be made as small as you
like. See Sections 26, 27 for Euclid's theory of volume. Archimedes' axiom is independent of all the
axioms of a Hilbert plane or a

Euclidean plane, so we will see examples of Archimedean geometries that satisfy (A) and non-
Archimedean geometries that do not (Section 18). The other axiom we would like to consider is
Dedekind's axiom, based on Dedekind's definition in the late nineteenth century of the real numbers:

D. Suppose the points of a line are divided into two nonempty subsets S, T in such a way that no point of
S is between two points of T, and no point of T is between two points of S. Then there exists a unique
point P such that for any AeS and any Be T, either A = Por B= Por the point P is between A and B.

This axiom is very strong. It implies (A) and (E), and a Euclidean plane with (D) is forced to be isomorphic
to the Cartesian plane over the real numbers. (See Exercise 12.2, Exercise 12.3, (15.5), and (21.3)-) So if
you want a categorical axiom system, just add (D) to the axioms of a Euclidean plane. From the point of
view of this book, however, there are two reasons to avoid using Dedekind's axiom. First of all, it
belongs to the modern development of the real numbers and notions of continuity, which is not in the
spirit of Euclid's geometry. Second., it is too strong. By essentially introducing the real numbers into our
geometry, it masks many of the more subtle distinctions and obscures questions such as constructibility
that we will discuss in Chapter 6. So we include this axiom only to acknowledge that it is there, but with
no intention of using it.