Contents

1 Affine and Euclidean Geometry 3

1.1 Points and vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Linear operations on vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Convention on identifying points with vectors . . . . . . . . . . . . . . . . . 6
1.4 Algebraic conditions expressing collinearity . . . . . . . . . . . . . . . . . . . 6
1.5 The ratio (AC~ : CB)~ for some collinear points . . . . . . . . . . . . . . . . . 8
1.6 First applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.7 Menelaus’ theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.8 Barycentric coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.9 Ceva’s theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.10 Desargues’ theorem—a few affine versions . . . . . . . . . . . . . . . . . . . 19
1.11 Desargues triangles in intersecting planes . . . . . . . . . . . . . . . . . . . . 23
1.12 Dot product: algebra and geometry . . . . . . . . . . . . . . . . . . . . . . . 25
1.13 Altitudes of a triangle are concurrent . . . . . . . . . . . . . . . . . . . . . . 27
1.14 The Feuerbach circle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.15 Angle sum of a triangle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.16 Law of cosines, law of sines . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.17 Angle bisectors, perpendicular bisectors . . . . . . . . . . . . . . . . . . . . . 32
1.18 Rotation, applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
1.19 Wedge product of two plane vectors . . . . . . . . . . . . . . . . . . . . . . . 38
1.20 A 3D view of plane geometry, triple product . . . . . . . . . . . . . . . . . . 42

2 Projective Geometry 45
2.1 Projective plane as extended affine plane . . . . . . . . . . . . . . . . . . . . 45
2.2 Incidence properties of P2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.3 Remarkable incidence theorems . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.4 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

1
2 CONTENTS

3 Spherical Geometry 49
3.1 Points, lines, triangles, polarity . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.2 Length, angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.3 Sides and angles of polar triangles . . . . . . . . . . . . . . . . . . . . . . . . 52
3.4 Laws of Cosines and Sines . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.5 Anglesum, area, Girard formula . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.6 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Chapter 1

Affine and Euclidean Geometry

1.1 Points and vectors

First we recall coordinate plane geometry from Calculus. The set R2 will be called the plane.
Elements of R2 , that is ordered pairs (x, y) of real numbers, are called points.

Consider directed segments (also called “arrows”) between points of the plane. We allow the
start point and the end point of an arrow to coincide. Arrows up to translation are called
(plane) vectors. That is, the arrow from A = (1, 3) to B = (5, 6) represents the same vector
−→ −−→ −→
as the arrow from C = (−4, −4) to D = (0, −1). We write AB = CD. The vector AA is
called the zero-vector and denoted by 0.

−→
We can represent a vector by an ordered pair of real numbers as well: the vector AB where
A = (a1 , a2 ) and B = (b1 , b2 ) will be represented by hb1 − a1 , b2 − a2 i. This is a fair definition,
−→ −−→ −→
because if AB = CD then b1 − a1 = d1 − c1 and b2 − a2 = d2 − c2 . The vector AB of the
paragraph above is h4, 3i.

3
4 CHAPTER 1. AFFINE AND EUCLIDEAN GEOMETRY

1.2 Linear operations on vectors

The sum and difference of two vectors are defined geometrically in Figure 1.1. In this context
real numbers will also be called scalars. A scalar multiple of a vector is defined in Figure 1.2.

a b u−v

u
a+b

Figure 1.1: Sum and difference of plane vectors

2a
a
−a

Figure 1.2: Scalar multiple

The operations above (addition, subtraction, multiplication by a scalar) are called the linear
operations on vectors. The geometric definitions above translate to the following algebraic
expressions.

• ha1 , a2 i + hb1 , b2 i = ha1 + b1 , a2 + b2 i

• ha1 , a2 i − hb1 , b2 i = ha1 − b1 , a2 − b2 i

1.2. LINEAR OPERATIONS ON VECTORS 5

• λ · ha1 , a2 i = hλa1 , λa2 i

Proposition 1.2.1 (Vector space “axioms”). The linear operations on vectors satisfy the
following properties.

• a+b=b+a

• (a + b) + c = a + (b + c)

• a+0=a

• a + (−a) = 0

• λ · (a + b) = λ · a + λ · b

• (λ + µ) · a = λ · a + µ · a

• λ · (µ · a) = (λµ) · a

• 1·a=a

Proof. The properties follow from the algebraic expressions for the linear operations.

Proposition 1.2.2 (2-dimensionality). Let a and b be non-parallel vectors (algebraically

a1 b2 − a2 b1 6= 0). For a vector c there are unique real numbers λ, µ such that c = λa + µb.

The proof of this proposition is left to the readers (see the problems below).

Problems

1.2.1. Let a = ha1 , a2 i and b = hb1 , b2 i be plane vectors. Prove that they are not parallel
(that is, neither one is a multiple of the other) if and only if a1 b2 − a2 b1 6= 0.

1.2.2. Let a, b be not parallel, and let c ∈ R2 be arbitrary. Prove that there are unique real
numbers λ, µ such that c = λa + µb.

1.2.3. Let a, b be not parallel. Prove that a + b, a − b are also not parallel.

1.2.4. For which real numbers k1,1 , k1,2 , k2,1 , k2,2 is it true that if a, b are not parallel then
k1,1 a + k1,2 b and k2,1 a + k2,2 b are also not parallel.
6 CHAPTER 1. AFFINE AND EUCLIDEAN GEOMETRY

1.3 Convention on identifying points with vectors

−→
To a point A ∈ R2 we can associate its “position vector” OA where O = (0, 0) is the origin.
−→
To a vector v we can associate a point P by considering an arrow OP representing v.
The above two associations are inverses of each other, they define a one-to-one corre-
spondence between points and vectors. Algebraically this one-to-one correspondence is
(a, b) ↔ ha, bi.
Throughout this text we will build in this identification in our notation, without further
explanation. For example, if A is a point, and we write 5A then we really mean either the
−→
vector 5OA or its endpoint. Or, if we say A/2 + B/2 is the midpoint of the segment AB
then here is how to read it precisely: the midpoint of the segment AB is the endpoint of the
−→ −−→
vector 21 OA + 12 OB.

1.4 Algebraic conditions expressing collinearity

The word “collinear” is a shorthand expression for “on the same line”. The word “concur-
rent” is a shorthand expression for “intersecting in one point”.

Proposition 1.4.1. Let A and B be two different points. Point C is on the line through A
and B if and only if there is a real number t such that

C = (1 − t)A + tB. (1.1)

Moreover, t and C uniquely determine each other (i.e. for any C on the AB line there is a
unique real number t satisfying (1.1), and for any real number t there is a unique point C
on the AB line satisfying (1.1).)

Proof. To obtain the position vector of a point C on the line AB we need to add the vector
→
− −→
A and a multiple of AB, see Figure 1.3.
Conversely, the end point of any such vector is obviously on the line AB. Observe that
→
− −→
A + t · AB = (1 − t)A + tB, which proves the proposition.
The proof that C and t determine each other is left as an exercise.

If A = B then for any t the point (1 − t)A + tB obviously coincides with A and B.
A useful rephrasing of Proposition 1.4.1 is that if A 6= B then C is on their line if and only
if there exist numbers x and y such that

C = xA + yB, x + y = 1.
1.4. ALGEBRAIC CONDITIONS EXPRESSING COLLINEARITY 7

B
−→
0.6AB C2
−→
−0.2AB

A
C1 −−→
OC2
−→
OA
−−→
OC1

O
Figure 1.3: C1 and C2 on the line through A and B

Proposition 1.4.2. The points A, B, C are collinear if and only if there exist real numbers
x, y, z not all 0, such that
xA + yB + zC = 0, x + y + z = 0.
Proof. Suppose A, B, C are collinear.
If A and B are different points, then C is on their line. According to Proposition 1.4.1 then
there is a t such that C = (1 − t)A + tB. After rearrangement we obtain (1 − t)A + tB +
(−1)C = 0 and hence 1 − t, t, −1 serve as x, y, z.
If A = B then x = 1, y = −1, z = 0 satisfy the requirements.
To prove the opposite direction let us now assume that xA + yB + zC = 0, x + y + z = 0,
and not all x, y, z are 0. Let us pick one non-zero among x, y, z. Without loss of generality
we may assume that it is z. Rearrangement gives C = (−x/z)A + (−y/z)B. The condition
x+y+z = 0 translates to (−x/z)+(−y/z) = 1. If A and B are different then Proposition 1.4.1
implies that C is on the AB line. If A and B coincide then the remark after Proposition 1.4.1
implies that all A, B, C are the same point, so they are collinear.
An important logical consequence of Proposition 1.4.2 is the following Corollary.
Corollary 1.4.3. If A, B, C are not collinear (i.e. they form a triangle), and x, y, z are real
numbers with
xA + yB + zC = 0 and x + y + z = 0,
then x = y = z = 0.
8 CHAPTER 1. AFFINE AND EUCLIDEAN GEOMETRY

Problems

1.4.1. In the formula C = (1 − t)A + tB, trace the position of C on the line as t varies from
−∞ to ∞.

1.4.2. Prove that in Proposition 1.4.1 the point C and the real number t uniquely determine
each other.

1.4.3. In ABC4 let U be the midpoint of AB and let V be the midpoint of AC. Prove
−−→ −−→
that 2U V = BC.

1.5 ~ : CB)
The ratio (AC ~ for some collinear points
Let A and B be different points, and let C be on their line. Let us assume that C 6= B. We
−→ −−→ −→ −−→ −→
will write that (AC : CB) = λ if AC = λCB. Such a λ exists (and is unique) since AC and
−−→ −−→
CB are collinear vectors with CB 6= 0.

Lemma 1.5.1. Let A, B, C be collinear, A 6= B 6= C, and write C = xA + yB with x + y = 1

−→ −−→
(cf. Proposition 1.4.1). Then we have (AC : CB) = y/x.
−→ −→ −−→ −→
Proof. From C = xA + yB we obtain AC = y AB and CB = xAB (verify!). Also, since
−→ −→ −−→
B 6= C we know that x 6= 0. Thus we have AC = y AB = y x1 CB, what we wanted to
prove.
−→ −−→
Remark 1.5.2. If we extended the notion of (AC : CB) to the case of C = B, by defining
−→ −−→
(AB : BB) = ∞, then the last proposition would say 1/0 = ∞, which agrees with our
intuition.
−→ −−→
In fact we can interpret the ratio (AC : CB) without mentioning vectors. It is the ratio of
the length of the segment AC over the length of the segment CB, with a sign convention.
−→ −−→
The sign convention is that if C is in between A and B, then (AC : CB) is positive, and if
−→ −−→
C is outside of the segment AB then (AC : CB) is negative.
−→ −−→
Proposition 1.5.3. Let A 6= B be fixed. The ratio (AC : CB) uniquely determines C.

The proof is left as an exercise.

1.6. FIRST APPLICATIONS 9

−→ −−→ −→
Remark 1.5.4. We may be sloppy in notation and decide to write AC/CB instead of (AC :
−−→
CB), but we must be careful that this ratio is only defined in the very special situation
where A, B, C are collinear (and some coincidences do not happen). In general there is no
such operation where we divide a plane vector by another plane vector!

Problems

1.5.1. Prove Proposition 1.5.3.

1.6 First applications

−→ −−→
A quadrilateral ABCD is a parallelogram if AB = DC. This condition can be phrased as
−−→ −−→
B − A = C − D, or rearranged to D − A = C − B, which means AD = BC also holds.

Proposition 1.6.1. The diagonals of a parallelogram bisect each other.

D C

A B
Figure 1.4: The diagonals of a parallelograph bisect each other.

Proof. Since ABCD is a parallelogram, we have B − A = C − D, or equivalently A + C =

B + D. Consider the point P = (A + C)/2 = (B + D)/2. The first defining expression
implies that P is the midpoint of A and C. The second expression implies that P is the
midpoint of B and D. Since they agree, P is the intersection of AC and BD, and it bisects
both diagonals.

The median of a triangle is a segment connecting a vertex to the midpoint of the opposite
side. A triangle has three medians.
10 CHAPTER 1. AFFINE AND EUCLIDEAN GEOMETRY
B

2x x
A

Figure 1.5: The medians of a triangle are concurrent.

Proposition 1.6.2. The medians of a triangle are concurrent. Moreover they divide each
other by 2:1.

Proof. Let ABC be a triangle. Consider the point P = (A + B + C)/3, and its equivalent
expressions
2 A+B 1 2 B+C 1 2 C +A 1
P = · + ·C = · + ·A= · + · B.
3 2 3 3 2 3 3 2 3
The first expression claims that P is on the segment connecting the midpoint of A and B
with C, that is, on the median corresponding to C. The second expression claims that P is
on the median corresponding to A, and the third expression claims that P is on the median
corresponding to B. Since they are all equal, there is a point, namely P , that in on all three
medians; and we proved that the medians are concurrent.
A byproduct of the above argument is that the intersection P of the three medians is ex-
pressed as 2/3 of the midpoint of a side plus 1/3 the opposite vertex. According to Section 1.5
this proves that P cuts the median 2 : 1.
Remark 1.6.3. In the above two propositions we needed to make arguments about the in-
tersections of certain lines. In our proofs we used a trick: we did not “compute” the in-
tersections, but rather we “named” a point and then proved that this point is on the lines,
and hence this point must be the intersection. You will find this trick useful when solving
exercises.

PROJECT 1. Invent and prove the 3D, 4D, . . . versions of Proposition 1.6.2.
1.6. FIRST APPLICATIONS 11

Problems
−−→ −→ −−→ −−→
1.6.1. For the points O, A, A0 , B, B 0 assume that OA0 = λOA, OB 0 = λOB, see Fig-
−−→ −→
ure 1.6(a). Prove that A0 B 0 = λAB.

1.6.2. Assume that in Figure 1.6(b) the lines x, y are parallel. Prove that there exists a real
−−→ −→ −−→ −−→
number λ such that OA0 = λOA, OB 0 = λOB.
−−−→ −−→ −−−→ −−→ −−−→ −−→ −−−→ −−→
1.6.3. Assume O1 A0 = λO1 A, O1 B 0 = λO1 B and O2 B 0 = µO2 A, O2 A0 = µO2 B, see the
Figure 1.6(c). Prove that λ = −µ.

Figure 1.6: Figures for Problems 1.6.1, 1.6.2, 1.6.3

1.6.4. In a quadrilateral let U and V be the midpoints of two opposite sides. Prove that the
segment U V and the segment connecting the midpoints of the diagonals bisect each other.
−→ −→ −→
1.6.5. Let S be the centroid of the ABC4. Calculate SA + SB + SC.

1.6.6. Let R be an arbitrary point in the plane and ABCD a parallelogram. Prove that
−→ −→ −→ −−→
RA + RC = RB + RD.

1.6.7. Let ABCD four point of the plane, and let U be the midpoint of AB, and V the
−−→ −−→ −−→
midpoint of CD. Express the vector U V in terms of the vectors AD and BC.
−→ −−→
1.6.8. In the ABCD quadrilateral let AB = a, DC = b. Let the points A, X1 , X2 , X3 , D
divide the AD side into four equal parts. Let the points B, Y1 , Y2 , Y3 , C divide the BC
−−−→ −−−→ −−−→
segment into four equal parts. Express X1 Y1 , X2 Y2 , X3 Y3 in terms of a and b.
12 CHAPTER 1. AFFINE AND EUCLIDEAN GEOMETRY

1.6.9. Let ABCDA0 B 0 C 0 D0 be a cube (ABCD is a square and A0 B 0 C 0 D0 is a translated

−→ −−→ −−→
copy of ABCD). Let a = AB, b = AD, c = AA0 . Let P be the midpoint of C 0 D0 . Let Q be
−→ −→ −−→ −−→
the center of the BCC 0 B 0 square. Express AP , AQ, AD0 , BD in terms of a, b, c.

1.6.10. Let ABCD be a parallelogram. Let the points A, X1 , X2 , B divide AB into three
equal parts. Let C, Y1 , Y2 , D divide CD into three equal parts. Let X2 , U, V, Y2 divide X2 Y2
−→ −→ −−→
into three equal parts. Express AV in terms of AB and AD.

1.6.11. Let P1 , . . . , Pn be points, and µ1 , . . . , µn be real numbers with ni=1 µi = 1. For a

P
−−→
point O consider v = ni=1 µi OPi . Let S be the end point of the vector v if it is measured
P
from O. Prove that the point S does not depend on the choice of O. (Hint: choose two
different O1 and O2 and calculate the vector between the obtained “two” S points. You
should get 0.)

1.6.12. Let the reflection of the point A over the point B be C. Express C in terms of A
and B.

1.6.13. (cont.) Let ABC4 be a triangle. The reflection of A over B is A0 . The reflection
of B over C is B 0 . The reflection of C over A is C 0 . Prove that the centroid of ABC4 and
the centroid of A0 B 0 C 0 4 coincide.

1.6.14. Let P be different from the vertices of the triangle ABC4. Let P BLC, P CM A,
P AN B be parallelograms. Prove that the segments AL, BM , CN bisect each other.

1.6.15. Points P, Q, R lie on the sides of the ABC4 and are such that
−−→ −→ −→ −→ −→ −→
(BP : P C) = (CQ : QA) = (AR : RB).

Prove that the centroids of the triangles P QR and ABC coincide.

1.7 Menelaus’ theorem

Theorem 1.7.1 (Menelaus’ theorem). Let ABC be a triangle and let a transversal line `
intersect the lines of the sides AB, BC, CA in M , K, L, respectively. We assume that none
of K, L, M coincide with A, B, or C. Then
−−→ −−→ −−→ −−→ −→ −→
(AM : M B) · (BK : KC) · (CL : LA) = −1.
1.7. MENELAUS’ THEOREM 13

Proof. We have K = xB +x0 C, L = yC +y 0 A, M = zA+z 0 B with x+x0 = y+y 0 = z +z 0 = 1.

Since K, L, M are collinear, according to Proposition 1.4.2 we know that there are real
numbers p, q, r not all 0 such that p + q + r = 0 and pK + qL + rM = 0. We have

p(xB + x0 C) + q(yC + y 0 A) + r(zA + z 0 B) = 0,

or rearranged, we have

(rz + qy 0 )A + (px + rz 0 )B + (qy + px0 )C = 0. (1.2)

The sum of the coefficients in this last expression is

(rz + qy 0 ) + (px + rz 0 ) + (qy + px0 ) = p(x + x0 ) + q(y + y 0 ) + r(z + z 0 ) = p + q + r = 0. (1.3)

According to Corollary 1.4.3 (1.2) and (1.3) can only hold if all three

rz + qy 0 = px + rz 0 = qy + px0 = 0.

Therefore we have

y 0 /z = −r/q, z 0 /x = −p/r, x0 /y = −q/p

(verify that from the hypotheses we know that none of the x, y, z, p, q, r are 0). Therefore

−−→ −−→ −−→ −−→ −→ −→ z 0 x0 y 0 z 0 x0 y 0 −p −q −r

(AM : M B) · (BK : KC) · (CL : LA) = · · = · · = · · = −1
z x y x y z r p q

what we wanted to prove.

14 CHAPTER 1. AFFINE AND EUCLIDEAN GEOMETRY

Theorem 1.7.2 (reverse Menelaus’ theorem). Let ABC be a triangle and let M, K, L be
points on the lines AB, BC, CA (but different from A, B, C) such that
−−→ −−→ −−→ −−→ −→ −→
(AM : M B) · (BK : KC) · (CL : LA) = −1. (1.4)

Then K, L, M are collinear.

Proof. Let ` be the line connecting L and K. We claim that ` is not parallel with the AB
line—this will be proved in Problem 1.7.1 below.
Let the intersection point be M 0 . According to Menelaus’ theorem we have
−−→ −−→ −−→ −−→ −→ −→
(AM 0 : M 0 B) · (BK : KC) · (CL : LA) = −1.

Comparing this with the assumption (1.4) on K, L, M we conclude that

−−→ −−→ −−→ −−→
(AM 0 : M 0 B) = (AM : M B).

Proposition 1.5.3 then implies that M = M 0 , hence the fact that K, L, M are collinear.

Problems
1.7.1. Let ABC be a triangle and let M, K, L be points on the lines AB, BC, CA (but
−−→ −−→ −−→ −−→ −→ −→
different from A, B, C) such that (AM : M B) · (BK : KC) · (CL : LA) = −1. Prove that
the line ` connecting L and K is not parallel with AB. [Hint: Problem 1.6.3 may be of use.]
1.7.2. Let D be a point on the line of BC, and let E be a point on the line of AC of the
ABC4 (but let D, E be distinct from the vertices). Assume the lines AD and BE intersect
−−→ −−→ −−→ −→
in a point P . Prove that if (BD : DC) · (CE : EA) = −1 then CP is parallel with AB.

1.8 Barycentric coordinates

Theorem 1.8.1. Let A, B and C be non-collinear points in the plane. For any point P we
may write
P = xA + yB + zC
where the real coefficients x, y, z satisfy

x + y + z = 1.

Moreover, x, y, z are uniquely determined by the point P .

1.8. BARYCENTRIC COORDINATES 15

We call x, y, z the barycentric coordinates of P with respect to the triangle ABC.

−→ −→
Proof. The vectors AB and AC are not parallel. Hence any vector can be written as a linear
−→ −→ −→
combination of them, for example AP = pAB + q AC (cf. Proposition 1.2.2). Using that
−→ −→ −→
AP = P − A, AB = B − A, AC = C − A we can rearrange it to

P = (1 − p − q)A + pB + qC,

and hence x = 1 − p − q, y = p, z = q satisfy the requirements.

To prove the uniqueness of barycentric coordinates assume that x, y, z and x0 , y 0 , z 0 are such
that
P = xA + yB + zC, x + y + z = 1,
P = x0 A + y 0 B + z 0 C, x0 + y 0 + z 0 = 1.
Then we have

0 = (x − x0 )A + (y − y 0 )B + (z − z 0 )C, (x − x0 ) + (y − y 0 ) + (z − z 0 ) = 0.

Corollary 1.4.3 implies that x − x0 = y − y 0 = z − z 0 = 0 which proves uniqueness.

PROJECT 2. Observe the similarity between Proposition 1.4.1 and Theorem 1.8.1. They
are the 1D and 2D cases of a general n-dimensional theorem. If you learned linear algebra
(specifically the notions of linear independence, generating set, basis) then find and prove
this general n-dimensional theorem.
16 CHAPTER 1. AFFINE AND EUCLIDEAN GEOMETRY
C

E
D
P

A F B
Figure 1.7: The Ceva configuration

1.9 Ceva’s theorem

Theorem 1.9.1 (Ceva’s Theorem1 ). Let ABC be a triangle and let P be a point in the plane
which does not lie on any of the sides of 4ABC. Suppose the lines AP , BP and CP meet
the opposite sides of ABC at D, E and F , respectively. Then
−→ −−→ −−→ −−→ −−→ −→
(AF : F B) · (BD : DC) · (CE : EA) = 1.

Note that P does not need to lie inside the triangle.

Proof. Using barycentric coordinates, we write P as

P = xA + yB + zC,

where x + y + z = 1. From our assumptions it follows that none of x, y, z, x + y, x + z, y + z

are 0 (verify!). Let us consider the point
x y
V = A+ B. (1.5)
x+y x+y

This expression implies that V is on the line AB. Calculation shows that
1 −z
V = P+ C,
x+y x+y
1
Geometer Giovanni Ceva (1647–1734) is credited with this theorem
1.9. CEVA’S THEOREM 17

and the sum of the coefficients 1/(x+y)+(−z)/(x+y) = (1−z)/(x+y) = (x+y)/(x+y) = 1.

Hence V is also on the line CP . We conclude that the point V is the intersection of AB and
CP , hence V = F . Moreover, from (1.5) we obtain that
y
−→ −−→ x+y y
(AF : F B) = x = .
x+y
x

Similarly, we find that

−−→ −−→ z
(BD : DC) = , (1.6)
y
−−→ −→ x
(CE : EA) = . (1.7)
z
Hence
−→ −−→ −−→ −−→ −−→ −→ yzx
(AF : F B) · (BD : DC) · (CE : EA) = = 1.
xy z

Theorem 1.9.2 (Reverse Ceva’s theorem). Suppose ABC is a triangle, D, E, F are points
on the lines of the sides (but none of them coincide with a vertex) such that
−→ −−→ −−→ −−→ −−→ −→
(AF : F B) · (BD : DC) · (CE : EA) = 1.

Then AD, BE, CF are either concurrent, or they are pairwise parallel.
Proof. If AD, BE, CF are pairwise parallel, then the theorem is proved. Assume that two
of these three lines intersect. Without loss of generality we assume that it is AD and BE.
Let P = AD ∩ BE, and assume that CP intersects AB in the point F 0 (verify that our
assumpitons imply that AB is not parallel with CP !). By Ceva’s Theorem 1.9.1 we have
−−→ −−→ −−→ −−→ −−→ −→
(AF 0 : F 0 B) · (BD : DC) · (CE : EA) = 1.

Comparing this with the condition in the theorem we obtain that

−−→ −−→ −→ −−→
(AF 0 : F 0 B) = (AF : F B).

Proposition 1.5.3 then implies that F = F 0 , hence the fact that AD, BE, CF are concurrent.

For fun, let us include here another “high-school style” proof of Ceva’s Theorem 1.9.1. This
proof does not use vectors at all. Instead it uses the notion of area, and the obvious fact
that the area of a triangle is half the product of base and height.
18 CHAPTER 1. AFFINE AND EUCLIDEAN GEOMETRY

Proof. For simplicity let D, E, F be on the sides (and not outside) of the triangle ABC, and
let P = AB ∩ CF = BC ∩ AD = CA ∩ BE. The triangles AF C and F BC have “bases”
AF and F B and they share the same height mC . Hence the ratio of their areas equals the
ratio of their bases:
Area(AF C) AF
= .
Area(F BC) FB
Similar argument for the triangles AF P and F BP gives

Area(AF P ) AF
= .
Area(F BP ) FB

From the two equations above simple algebra implies

Area(AF C) − Area(AF P ) AF
= .
Area(F BC) − Area(F BP ) FB

The difference of the triangles AF C and AF P is the triangle CAP . The difference of the
triangles F BC and F BP is the triangle BCP . Hence we obtained

Area(CAP ) AF
= . (1.8)
Area(BCP ) FB

We obtained (1.8) by considering the AB side of the triangle the “base”. Repeating the
same argument but now considering the BC side or the CA side to be the “base” we obtain
the equations.
Area(ABP ) BD
= ,
Area(CAP ) DC
Area(BCP ) CE
= .
Area(ABP ) EA
From the last three equations we obtain

−→ −−→ −−→ −−→ −−→ −→ AF BD CE

(AF : F B) · (BD : DC) · (CE : EA) = · · =
F B DC EA
Area(CAP ) Area(ABP ) Area(BCP )
= · · = 1,
Area(BCP ) Area(CAP ) Area(ABP )
which proves Ceva’s theorem in the case when P is inside the triangle. Similar arguments
work when P is outside.
1.10. DESARGUES’ THEOREM—A FEW AFFINE VERSIONS 19

Problems
0
1.9.1. Consider the ABC4 and non-zero real numbers k1 , k2 , k3 . Let PAB and PAB be on
the AB line, the first in the AB segment, the second outside of the AB segment, such that

|APAB | |AP 0 | k1
= 0 AB = .
|PAB B| |PAB B| k2
0 0
Define PBC and PBC on the BC line similarly with the ratio k2 /k3 ; and define PCA and PCA
on the CA line similarly with the ratio k3 /k1 . Show that the lines APBC , BPCA , CPAB are
concurrent.
0
1.9.2. (cont.) Prove that the line PCA PAB contains the point PBC .
0 0 0
1.9.3. (cont.) Prove that PAB , PBC , and PCA are collinear.

1.9.4. Let P be an interior point of ABC4. The lines connecting P with the vertices cut
ABC4 into six smaller triangles. We color every second of these six triangles with red, the
rest with blue. Prove that the product of the areas of the red triangles is the same as the
product of the areas of the blue triangles.

1.10 Desargues’ theorem—a few affine versions

Desargues’ theorem is a remarkable theorem on incidences of certain lines and points involv-
ing two triangles. The key notions are as follows.

• For ABC4 and A0 B 0 C 0 4 we may consider the three lines connecting the corresponding
vertexes: AA0 , BB 0 , and CC 0 . We will consider the condition that these three lines
are concurrent (or are pairwise parallel). If concurrent, we call the intersection point
the center of perspectivity.

• For ABC4 and A0 B 0 C 0 4 we may consider the intersections of the corresponding sides
AB ∩ A0 B 0 , BC ∩ B 0 C 0 , and CA ∩ C 0 A0 . We will consider the condition that these
three points (exist and) are collinear—or none of the three exist. If collinear, we call
the obtained line the axis of perspectivity.

Theorem 1.10.1. Let ABC and A0 B 0 C 0 be triangles such that AB||A0 B 0 , AC||A0 C 0 , BC||B 0 C 0 .
Then the three lines AA0 , BB 0 , CC 0 are either concurrent or pairwise parallel.
20 CHAPTER 1. AFFINE AND EUCLIDEAN GEOMETRY

Proof. Because of the conditions on parallel lines we can write

B − C = k1 (B 0 − C 0 ), C − A = k2 (C 0 − A0 ), A − B = k3 (A0 − B 0 ), (1.9)

for some real numbers k1 , k2 , k3 . Adding together these three equalities (and rearranging the
right hand side) we obtain

0 = (k2 − k3 )A0 + (k3 − k1 )B 0 + (k1 − k2 )C 0 . (1.10)

The coefficients of (1.10) add up to 0, hence Corollary 1.4.3 implies that

k2 − k3 = k3 − k1 = k1 − k2 = 0, and hence k1 = k2 = k3 .

Let k be the common value of k1 , k2 and k3 . Then from (1.9) we can deduce

A − kA0 = B − kB 0 = C − kC 0 . (1.11)

We can consider two cases. If k = 1 then (1.11) implies that the lines AA0 , BB 0 , CC 0 are
pairwise parallel. If k 6= 1 then (1.11) can be rearranged to
1 −k 0 1 −k 0 1 −k 0
P = A+ A = B+ B = C+ C
1−k 1−k 1−k 1−k 1−k 1−k
showing that the point P is on all three lines AA0 , BB 0 , CC 0 —proving that AA0 , BB 0 , CC 0
are concurrent.
Theorem 1.10.2. Let ABC and A0 B 0 C 0 be triangles such that AA0 , BB 0 , CC 0 are concur-
rent. Assume that the point K = AB ∩ A0 B 0 , L = BC ∩ B 0 C 0 , M = CA ∩ C 0 A0 exist. Then
the points K, L, M are collinear.
Proof. The intersection point on AA0 , BB 0 , CC 0 can be written as

k1 A + (1 − k1 )A0 = k2 B + (1 − k2 )B 0 = k3 C + (1 − k3 )C 0 . (1.12)

Rearranging, say, the first equality we obtain

k1 A − k2 B = (1 − k2 )B 0 − (1 − k1 )A0 .

If k1 = k2 then this formula says k1 (A − B) = (1 − k1 )(B 0 − A0 ), meaning that AB||A0 B 0

which is not the case. So we know that k1 6= k2 . Hence, we may divide by k1 − k2 and write
k1 −k2 1 − k2 0 −(1 − k1 ) 0
A+ B= B + A.
k1 − k2 k1 − k2 k1 − k2 k1 − k2
1.10. DESARGUES’ THEOREM—A FEW AFFINE VERSIONS 21

The sum of the coefficients on the left hand side is 1, and the sum of the coefficients on the
right hand side is also 1 (check it!). Therefore the left hand side expression is a point on the
AB line, and the right hand side expression is a point on the A0 B 0 line. Hence the common
value must be the intersection AB ∩ A0 B 0 . We obtained that

k1 −k2
K= A+ B,
k1 − k2 k1 − k2

equivalently
(k1 − k2 )K = k1 A − k2 B. (1.13)

We deduced (1.13) from the fact that the first expression and the second expression in
(1.12) are equal. Similarly, the fact that the second and third, as well as the first and third
expressions in (1.12) are equal we obtain

(k2 − k3 )L = k2 B − k3 C, (k3 − k1 )M = k3 C − k1 A. (1.14)

Adding together all three equalities in (1.13) and (1.14) we get

0 = (k1 − k2 )K + (k2 − k3 )L + (k3 − k1 )M.

Observe that none of the three coefficients are 0, and they add up to 0. According to
Proposition 1.4.2 this means that K, L, M are collinear, what we wanted to prove.

Theorem 1.10.3. Let ABC and A0 B 0 C 0 be triangles such that K = AB ∩ A0 B 0 , L =

BC ∩ B 0 C 0 , M = CA ∩ C 0 A0 exist and are collinear. Then AA0 , BB 0 , CC 0 are either
concurrent or are pairwise parallel.

It is possible to prove this theorem with the techniques we used in the last two proofs—and
it may be a good practice for students to write down such a proof. However, for a change we
are going to prove it by reduction to Theorem 1.10.2—showing that in some sense Desargues’
theorem and its reverse are the same, in other words, Desargues’ theorem is “self-dual”.

Proof. If AA0 , BB 0 , CC 0 are pairwise parallel, then we are done. If not, then two of them
intersect, say, AA0 intersects BB 0 .
22 CHAPTER 1. AFFINE AND EUCLIDEAN GEOMETRY

Consider the triangles AA0 M and BB 0 L. By looking at the picture one can see that the
lines AB, A0 B 0 , M L connecting the corresponding vertexes are concurrent. Theorem 1.10.2
can be applied to the triangles AA0 M and BB 0 L, and we obtain that C = AM ∩ BL,
C 0 = M A0 ∩ LB 0 , and AA0 ∩ BB 0 are collinear. That is, CC 0 passes through the intersection
of AA0 ∩ BB 0 , and hence AA0 , BB 0 , and CC 0 are concurrent.

In a later chapter we will see a simple and elegant way of phrasing Desargues’ theorem—in
projective geometry. All of the three theorems above (and more) are some special cases of
that projective Desargues’ theorem.

PROJECT 3. We can connect two points A and B of the plane if we have a straightedge.
Now suppose that B is “hidden”, it is only given by portions of two intersecting lines, but
we cannot go close to the intersection point B; for example it is outside of the margin of
our paper. How can we connect A and B with a straightedge? How to connect two hidden
points?
1.11. DESARGUES TRIANGLES IN INTERSECTING PLANES 23

Problems

1.10.1. Assume that for ABC4 and A0 B 0 C 0 4 we have AA0 ||BB 0 ||CC 0 , and that the points
K = AB ∩ A0 B 0 , L = BC ∩ B 0 C 0 , M = CA ∩ C 0 A0 exist. Prove that K, L, M are collinear.

1.10.2. The triangles A1 B1 C1 4, A2 B2 C2 4, and A3 B3 C3 4 have their corresponding sides

parallel. Hence each pair of triangles has a center of perspectivity (assume that these centers
of perspectivities exist). Prove that the three centers of perspectivities are collinear.

1.10.3. Use a search engine to learn about “Monge’s theorem”. Try to relate it to Prob-
lem 1.10.2.

1.10.4. A line drawn through the vertex A of a parallelogram ABCD cuts CB in P and
CD in Q. A line through C cuts AB in R and AD in S. Prove that P R and QS are parallel.

1.11 Desargues triangles in intersecting planes

In this section we go out of our way again and show an interesting “high-school style”
argument in relation with Desargues’ theorem.
Consider two planes P1 and P2 in 3 dimensions, intersecting in the line `. Let ABC4 be in
P1 and let A0 B 0 C 0 4 be in P2 . We will analyse the conditions and the claim of Theorem 1.10.3
for these two triangles.
24 CHAPTER 1. AFFINE AND EUCLIDEAN GEOMETRY

The condition is about the three points K = AB ∩ A0 B, L = BC ∩ B 0 C 0 , M = CA ∩ C 0 A0 .

Observe that AB ⊂ P1 , A0 B 0 ⊂ P2 , hence K ∈ P1 ∩ P2 = `. Similarly, L and M must also
lie on `. So the assumption of Theorem 1.10.3 that K, L, M are collinear does not even have
to be assumed! It automatically holds.

The statement of Theorem 1.10.3 is about the three lines AA0 , BB 0 , CC 0 . It will be useful
to consider three more planes. Let SK be the plane containing the intersecting lines AB,
A0 B 0 . Let SL be the plane containing the intersecting lines BC, B 0 C 0 . Let SM be the plane
containing the intersecting lines AC, A0 C 0 .

Observe that both A and A0 are contained in SK and in SM . If two point are contained in
two planes, then their connecting line must be the intersection of the two planes. We have
AA0 = SK ∩ SM . Similarly BB 0 = SK ∩ SL , and CC 0 = SL ∩ SM .

We obtained that the three lines AA0 , BB 0 , CC 0 are the pairwise intersections of three
planes in space. Let’s see how can three (pairwise intersecting) planes look like in three
space. There are two possibilities: (i) either the third one is parallel with the intersection
line of the first two, or (ii) the third one intersects the intersection line of the first two.
Theses two configurations are illustrated in the picture.
1.12. DOT PRODUCT: ALGEBRA AND GEOMETRY 25

In the first case the three intersection lines are pairwise parallel: AA0 , BB 0 , CC 0 are pairwise
parallel. In the second case the three intersection lines are concurrent: AA0 , BB 0 , CC 0 are
concurrent.
What we found is that the 3D version of Desargues theorem 1.10.3 is a tautology.
PROJECT 4. Analyses 3D versions of the other two versions of Desargues’ theorem above.
PROJECT 5. Find a no-calculation proof of e.g. Theorem 1.10.3, by first moving one of
the triangles out of plane into 3D.

1.12 Dot product: algebra and geometry

Let us recall the notion of dot product from Calculus. The dot product of two vectors a and
b is a number denoted by a · b or ab.
Geometrically ab = |a||b| cos φ, where |x| denotes the length of a vector x and φ is the
angle between the vectors a and b. Especially, ab = 0 if and only if a and b are orthogonal.
Algebraically (a1 , a2 ) · (b1 , b2 ) = a1 b1 + a2 b2 .
26 CHAPTER 1. AFFINE AND EUCLIDEAN GEOMETRY

Problem: recall from calculus why the above geometric and algebraic definition agree.
The following properties are easily verified from the algebraic definition.

• ab = ba

• (a + b)c = ac + bc

• (λa)b = λ(ab)

• aa ≥ 0, and aa = 0 if and only if a = 0

• ab = 0 for all b implies that a = 0.

The power of dot product that we will repeatedly use in geometry is the duality: its clear
geometric meaning and its simple algebraic properties. (What we will not use any further is
the a1 b1 + a2 b2 expression.)
√
Definition 1.12.1. The length of a vector a is defined to be |a| = aa. (The square
root makes sense because of the the non-negativity property above. Also this definition is
consistent with the geometric interpretation of dot product above.) The length of a segment
−→
AB is defined to be d(AB) = |AB|. The distance of two sets P, Q ⊂ R2 is defined to be
inf{d(A, B) : A ∈ P, B ∈ Q}.

Problems

1.12.1. Verify the algebraic properties of the dot product above.

1.12.2. Prove that for a parallelogram the sum of squares of the sides is equal to the sum
of the squares of the diagonals.

1.12.3. Let O be the center of the ABCDEF regular hexagon whose side length is 1. Find
−→ −→ −→ −→ −−→ −→ −→ −−→ −→ −→
AB · AO, AB · AC, BC · EF , F C · BD, F C · EF .

1.12.4. Let A, B, C, D be points in the plane, let A0 be the midpoint of BC, let B 0 be the
midpoint of CA, and let C 0 be the midpoint of AB. Prove that

(D − A0 )(C − B) + (D − B 0 )(A − C) + (D − C 0 )(B − A) = 0.

1.13. ALTITUDES OF A TRIANGLE ARE CONCURRENT 27

1.13 Altitudes of a triangle are concurrent

Let ABC be a triangle. A line passing through the vertex and perpendicular to the opposite
side is called an altitude. A triangle has three altitudes.

Theorem 1.13.1. The three altitudes of a triangle are concurrent.

Proof. Let D be the intersection of the altitudes containing the vertexes A and B. Then
AD ⊥ BC and BD ⊥ AC. Hence we have

(D − A)(B − C) = 0, (D − B)(C − A) = 0.

Adding these two equations together, and using the algebraic properties of dot product we
obtain
0 = (D − A)(B − C) + (D − B)(C − A) = ... = −(D − C)(A − B).
Therefore D − C ⊥ A − B, that is, the line DC is the altitude containing C. All three
altitudes pass through D.
The intersection of the three altitudes is called the orthocenter of the triangle.
Consider the vertexes and the orthocenter. It is remarkable that each of theses four points
is the orthocenter of the triangle formed by the other three points. Such a set of four points
will be called an orthocentric tetrad.

Problems
1.13.1. Suppose that the segment connecting the midpoints of AB with CD, and the seg-
ment connecting the midpoints of BC with DA are of the same length. Prove that AC is
perpendicular to BD.
28 CHAPTER 1. AFFINE AND EUCLIDEAN GEOMETRY

1.13.2. Suppose that the segment connecting the midpoints of AB with CD, and the seg-
ment connecting the midpoints of BC with DA are of the same length. Prove that AC is
perpendicular to BD.

1.14 The Feuerbach circle

Lemma 1.14.1. If A, B, C, D is an orthocentric tetrad then

(A + B − C − D)2 = (A − B + C − D)2 = (A − B − C + D)2 =

(−A + B + C − D)2 = (−A + B − C + D)2 = (−A − B + C + D)2 .

Proof. The six numbers above are in fact three pairs, e.g. (A + B − C − D)2 and (−A −
B + C + D)2 are clearly equal, because they are length squares of a vector and its opposite
vector. What we need to prove is that two numbers not in the same pair are also equal.
Without loss of generality let us choose the first two. Calculation shows that

(A + B − C − D)2 − (A − B + C − D)2 = 4(A − D)(B − C).

Since A, B, C, D is an orthocentric tetrad A − D is orthogonal to B − C, and hence (A −

D)(B − C) = 0, showing that (A + B − C − D)2 = (A − B + C − D)2 .

Theorem 1.14.2 (Feuerbach circle). Let D be the orthocenter of the ABC triangle. Con-
sider the following nine points (a) the midpoints of the sides, (b) the midpoints of the seg-
ments connecting vertexes to the orthocenter, (c) the feet of the altitudes. These nine points
are on one circle.
1.14. THE FEUERBACH CIRCLE 29

Proof. We will only prove that points (a) and (b) are on one circle. The fact that points (c)
are also on the same circle is left as an exercise.
Observe that the six points in (a) and (b) are the midpoints of the six segments connecting
two of A, B, C, D where A, B, C, D form an orthocentric tetrad. Hence these points are
(A + B)/2, (A + C)/2, (A + D)/2, (B + C)/2, (B + D)/2, (C + D)/2.
Let N = (A + B + C + D)/4. The vectors connecting N to the six points are
(−A − B + C + D)/4, (−A + B − C + D)/4, (−A + B + C − D)/4,
(A − B − C + D)/4, (A − B + C − D)/4, (A + B − C − D)/4.
These six vectors have the same length because of Lemma 1.14.1. Therefore all six points
are of the same distance from N : they are on one circle.

Problems
1.14.1. (Thales’ theorem) Let O be the midpoint of AC. Prove that ∠(ABC) = π/2 if and
only if d(AO) = d(BO). (That is ∠(ABC) = π/2 if and only if B is on the circle with center
O and radius d(AO).)
1.14.2. Finish the proof of Feuerbach’s Theorem 1.14.2, i.e. prove that points (c) are also
on the same circle as points (a) and (b). [Hint: Thales’ theorem is useful.]
30 CHAPTER 1. AFFINE AND EUCLIDEAN GEOMETRY

1.15 Angle sum of a triangle

Lemma 1.15.1. Let the line m intersect a pair of parallel lines `, `0 ; and let α and α0 be
the angles obtained as in Figure 1.8 (a). Then α = α0 .
Proof. Take unit vectors x, −x and u in the lines `, `0 , and m as in the picture. Then

α = arccos(x · u), α0 = arccos((−x) · (−u)),

so they obviously agree.

Figure 1.8: The sum of the angles of a triangle

Theorem 1.15.2. The sum of the angles of a triangle is π.

Proof. Let the line `0 be parallel to AB and pass through the point C, see Figure 1.8 (b).
According to Lemma 1.15.1 α = α0 and β = β 0 hence we have

α + β + γ = α0 + β 0 + γ = π.

1.16 Law of cosines, law of sines

For a triangle ABC4 the sides opposite to A, B, C will be denoted by a, b, c respectively,
and the angles at A, B, C will be called α, β, γ respectively.
Theorem 1.16.1 (Law of Cosines). We have

c2 = a2 + b2 − 2ab cos γ.
1.16. LAW OF COSINES, LAW OF SINES 31

Proof. We have c2 = c2 = (a − b)2 = a2 + b2 − 2ab = a2 + b2 − 2ab cos γ.

Corollary 1.16.2 (Pythagorean theorem). In a right triangle with hypothenuse c we have
c 2 = a2 + b 2 .
Proof. This is the Law of Cosines for γ = π/2.
Corollary 1.16.3 (Triangle inequality). For three points A, B, C in the plane we have
d(AB) ≤ d(BC) + d(CA).
The proof follows from the Law of Cosines, details are left as an exercise.
Theorem 1.16.4 (Law of Sines). We have
sin α sin β sin γ
= = .
a b c
We will give two proves: the first one shows that the Law of Sines is a formal consequence
of the Law of Cosines. The second proof is geometric.
Proof.
p Proof1. From the Law of Cosines we get cos γ = (a2 + b2 − c2 )/(2ab). Using sin γ =
1 − cos2 γ we have
s  2 2
a + b 2 − c2
sin γ = 1 − =
2ab
r √
4a2 b2 − (a4 + b4 + c4 + 2a2 b2 − 2a2 c2 − 2b2 c2 ) −a4 − b4 − c4 + 2a2 b2 + 2b2 c2 + 2a2 c2
= .
4a2 b2 2ab
Dividing both sides by c we obtain
sin γ
= an expression symmetric in a, b, c.
c
Therefore we will get the same expression on the right hand side, if we start with α or β,
not γ. This proves that sin γ/c = sin α/a = sin β/b.
Proof. Proof2. The altitude m passing through vertex C is a side of two right triangles (See
picture) yielding the two expressions: m = b sin α, m = a sin β. Putting the right hand sides
equal and rearranging gives sin α/a = sin β/b.

Problems
1.16.1. Prove Corollary 1.16.3, the triangle inequality.
32 CHAPTER 1. AFFINE AND EUCLIDEAN GEOMETRY

1.17 Angle bisectors, perpendicular bisectors

Consider an angle α less than π. The ray inside the angle that cuts α into two angles of
measure α/2 is called the angle bisector.
The distance d(P, `) of a point P to a line ` is the the infimum of the distances between P
and A where A ∈ `.
Lemma 1.17.1. The distance of a point to a line is obtained on the perpendicular segment
dropped from the point to the line.
Proof. The Pythagorean theorem proves that x < x0 on picture ? (a).
Lemma 1.17.2. The points in an angle that are of the same distance from the two rays of
the angle are exactly the points of the angle bisector.

Proof. For a point P as in Picture ? (b) its distance to the two sides is u and v according
to Lemma 1.17.1. We have u = c sin(β), v = c sin(γ). Hence u = v holds if and only if
sin(β) = sin(γ). Well known properties of the sin function imply that this holds if and only
if β = γ.
Theorem 1.17.3. The three angle bisectors of a triangle are concurrent.
1.17. ANGLE BISECTORS, PERPENDICULAR BISECTORS 33

Proof. Let xa , xb , xc be the angle bisectors through the vertexes A, B, C. Let P be the
intersection of the xa and xb (verify that they cannot be parallel!). Then

P ∈ xa ⇒ d(P, b) = d(P, c)
⇒ d(P, a) = d(P, b) ⇒ P ∈ xc ,
P ∈ xb ⇒ d(P, a) = d(P, c)

where three of the four ⇒ implications above use Lemma 1.17.2. Since the intersection of
xa and xb is on xc , we have that xa , xb , xc are concurrent.

A byproduct of the theorem is that the intersection of the angle bisectors has the same
distance to the sides. In other words there is a circle with this center that touches the sides
of the triangle: the so-called circle inscribed in the triangle.

For a segment AB, the line passing through the midpoint of AB and perpendicular to AB
is called the perpendicular bisector.

Theorem 1.17.4. The three perpendicular bisectors of the sides of a triangle are concurrent.

The proof is left as an exercise.

Problems

1.17.1. Find and prove a lemma analogous to Lemma 1.17.2 but it it about the perpendicular
bisector of a segment.

1.17.2. Using Problem 1.17.1 find a proof of Theorem 1.17.4 (logically similar to the proof
of Theorem 1.17.3).

1.17.3. Find a byproduct of your proof in Problem 1.17.2, analogous to the byproduct of
the proof of Theorem 1.17.3.
34 CHAPTER 1. AFFINE AND EUCLIDEAN GEOMETRY

1.17.4. Prove that a triangle is equilateral if two of its circumcenter, centroid, and ortho-
center coincide.

1.17.5. Let C be a circle in the plane, and let λ be a number. Prove that λ C = {λx ∈ R2 :
x ∈ C} is also a circle.

1.17.6. (cont.) Reflect the orthocenter of ABC4 over the midpoints of the sides. Prove
that the obtained three points are on the circumscribed circle. (Hint: use Problem 1.17.5
for a well chosen origin and λ, as well as the Feuerback circle.)

1.17.7. Let O be the center of circumscribed circle of ABC4. Let a, b, c be the vectors
pointing from O to the vertexes. Let M be the endpoint of a + b + c measured from O.
Prove that M is the orthocenter of ABC4.

1.17.8. (cont.) Let O be the center of the circumscribed circle of ABC4. Let M be the
orthocenter, and let S be the centroid of ABC4. Prove that O, M, S are collinear. (The
−→ −−→
obtained line is called the Euler-line of ABC4.) Find OS/SM .

1.17.9. Let D be a point on the side BC of ABC4. Prove that

BD |AB| sin(DAB∠)
= .
DC |AC| sin(DAC∠)

1.17.10. (cont.) Prove the Angle Bisector Theorem: If the angle bisector from A intersects
BC in D, then
|BD| |AB|
= .
|DC| |AC|

1.17.11. (cont.) Reprove that angle bisectors are concurrent by using the Angle Bisector
Theorem and Ceva’s theorem.

1.17.12. Let AB = 6, BC = 10 in the ABC4. The angle bisector at B intersects AC in

D. Connect D with the midpoint of AB, let the intersection point with BC be E. What is
the length |BE|?

1.17.13. Let the lengths of the sides of ABC4 be a, b, c (a is opposite of A, etc). If O is

the center of the inscribed circle then prove that

aA + bB + cC
O= .
a+b+c
1.18. ROTATION, APPLICATIONS 35

1.18 Rotation, applications

−→
For α an angle and v = OA a plane vector let Rα (v) denote the vector obtained from rotating
−→
v = OA around the origin in the counterclockwise direction, see Picture ? (a). Thus Rα is
a map from vectors to vectors. Calculation shows that algebraically

Rα : hx, yi 7→ hx cos α − y sin α, x sin α + y cos αi.

−→
If a vector is given by an arrow v = P Q where P is not the origin, then to get Rα (v) we
−→ −→ −→
formally need to translate P Q to OA, then rotate this OA by α. Picture ? (b) shows that
−→
this procedure is not necessary: Rα (v) is also obtained by rotating P Q around P by α.

Proposition 1.18.1. The rotation operator on vectors is consistent with vector operations
as follows,
• Rα (a + b) = Rα (a) + Rα (b),

• Rα (a − b) = Rα (a) − Rα (b),

• Rα (λa) = λRα (a),

• Rα (a)Rα (b) = ab,

• Rβ (Rα (a)) = Rα+β (a).

Proof. All can be calculated from the algebraic description of the operations.
Theorem 1.18.2. Let ABC4 be a triangle, and let T 1 and T 2 squares on the sides AC and
BC outside the triangle. Let K and L be the centers of T 1 and T 2, and X is the midpoint
of AB. Then XK and XL have the same length.
36 CHAPTER 1. AFFINE AND EUCLIDEAN GEOMETRY

−→ −−→
Proof. Let a = AC, b = BC, and let R = Rπ/2 be the rotation by π/2 operator. Observe
−−→
that we can express all relevant vectors in our picture using a, b, R: for example AC 0 = R(a),
−−→00
BC = −R(b).
We have
−→ −→ −−→
−−→ −−→ −−→ BA AC + AC 0 −a + b a + R(a) R(a) + b
XK = XA + AK = + = + = ,
2 2 2 2 2
−→ −−→ −−→
−−→ −−→ −→ AB BC + BC 00 −b + a b − R(b) a − R(b)
XL = XB + BL = + = + = .
2 2 2 2 2
The idea of the proof is that we suspect that not only XK and XL are of the same length
but one is the π/2 rotation of the other. Hence we calculate
 
−−→ a − R(b) R(a) − R(R(b))
R(XL) = R = .
2 2

Observe that applying R twice on a vector is the same as multiplication by −1. Indeed,
Rπ/2 Rπ/2 = Rπ =multiplication by (−1). Hence

−−→ R(a) + b −−→

R(XL) = = XK,
2
what we wanted to prove.
The proof above is not the shortest or most elegant proof, but illustrates the main point:
naming sufficient vectors and operations (but not more) that determine the picture we can
express any other vectors in terms of the named ones. Then we can make comparisons among
any two. A more “elegant” version of the same proof will be given in Problem 1.18.3.
1.18. ROTATION, APPLICATIONS 37

Theorem 1.18.3 (Napoleon Bonaparte2 ). Let ABC4 be an arbitrary triangle and let Ta ,
Tb , Tc by regular (a.k.a. equilateral) triangles on the sides of a, b, c, outside of ABC. Let
A0 , B 0 , C 0 be the centers of Ta , Tb , Tc . Then A0 B 0 C 0 4 is a regular triangle.

−→ −→
Proof. Let b = AB and c = AC, and let R = Rπ/3 be the rotation by π/3 operator. Our
−−→ −−→
goal is to express relevant vectors, namely A0 B 0 and A0 C 0 in terms of these.
First observe that if x is a side vector of a regular triangle then the vector pointing from a
vertex to its center as drawn in Picture ? (b) is (x + R(x))/3. Therefore we have
−−→0 c + R(c) −−→0 b − c + R(b − c) −−→0 −b + R(−b)
AB = , CA = , BC = ,
3 3 3
see Picture ? (c). Now we can express
−− → b − c + R(b − c) c + R(c) −1 1 1 2
A0 B 0 = − −c+ = b − c − R(b) + R(c),
3 3 3 3 3 3
−− → b − c + R(b − c) −b + R(−b) 1 2 2 1
A0 C 0 = − −c+b+ = b − c − R(b) + R(c).
3 3 3 3 3 3
2
yes, him
38 CHAPTER 1. AFFINE AND EUCLIDEAN GEOMETRY

−−→ −−→
What we want to prove is R(A0 B 0 ) = A0 C 0 so let us calculate

−−→
 
0 0 −1 1 1 2 −1 1 1 2
R(A B ) = R b − c − R(b) + R(c) = R(b) − R(c) − R(R(b)) + R(R(c)).
3 3 3 3 3 3 3 3

Looking at Picture ? (d) we see that R(x) = x + R(R(x)), and hence R(R(x)) = R(x) − x.
We further have
−−→ −1 1 1 2 1 2 2 1
R(A0 B 0 ) = R(b) − R(c) − (R(b) − b) + (R(c) − c) = b − c − R(b) + R(c).
3 3 3 3 3 3 3 3
−−→ −−→ −−→
This last expression is the same as the expression for A0 C 0 above, hence R(A0 B 0 ) = A0 C 0
what we wanted to prove.

Problems

1.18.1. Let R denote the rotation by 30o (= π/6) counterclockwise. Simplify the expressions
R12 (x), R13 (x), R100 (x), R6 (x). Consider the polynomial p(x) = x0 + x6 . What is p(R)(x)
for any x?

1.18.2. Let R denote the rotation by 4/9 · π. Write down a polynomial p(R) in R, such that
for all x we have p(R)(x) = 0.
−−→
1.18.3. Consider the situation of Theorem 1.18.2. Express the vector XL in terms of the
−→ −−→ −−→
vectors AC and BC 00 (e.g. using Problem 1.6.7). Express the vector XK in terms of the
−−→ −−→ −−→
vectors AC 0 and BC. Look at the expression you got for XL, and term by term apply the
−−→
rotation R. You should get your expression for XK. Write down the obtained proof for
Theorem 1.18.2.

1.18.4. Let P, Q, R, S be the centers of the squares that are described externally on the
sides of a quadrilateral (in this order). Prove that P R and QS are of the same length, and
are perpendicular to each other.

1.19 Wedge product of two plane vectors

In Calculus we learn the geometry and the algebra of the notion of cross product a × b of
two vectors a, b ∈ R3 in 3-space. Geometrically a × b has length |a||b| sin θ (where θ is the
1.19. WEDGE PRODUCT OF TWO PLANE VECTORS 39

angle between a and b), it lies in the line orthogonal to the plane spanned by a and b and
its direction satisfies the right-hand rule. Algebraically
ha1 , a2 , a3 i × hb1 , b2 , b3 i = ha2 b3 − a3 b2 , a3 b1 − a1 b3 , a1 b2 − a2 b1 i.
Plane vectors ha1 , a2 i can be considered space vectors by ha1 , a2 , 0i. If we take the cross
product of two such plane-space vectors then we obtain a vector of the form h0, 0, ∗i. We do
not want to keep carrying the (0, 0)-part, hence we give a new definition capturing only the
third coordinate.
Definition 1.19.1. The wedge product a ∧ b of two plane vectors a = ha1 , a2 i, b = hb1 , b2 i
is the third coordinate of the cross product ha1 , a2 , 0i × hb1 , b2 , 0i.
From the arguments above we obtain that
• (algebra) ha1 , a2 i ∧ hb1 , b1 i = a1 b2 − a2 b1 ,
• (geometry) a ∧ b = ±|a||b| sin θ. Since |a||b| sin θ is the area of a the parallelogram
spanned by a and b (see figure below), we have
a ∧ b = ± Area(parallelogram spanned by a and b).
Analysing the right-hand rule mentioned above we can determine wether + or - stand
in the formula above: if the direction of b is obtained from the direction of a by
a counterclockwise rotation by not more than π then the sign is positive, otherwise
negative.

The algebraic interpretation easily proves the following properties.

Proposition 1.19.2. We have
(i) (anti-symmetry) a ∧ b = −b ∧ a, a ∧ a = 0,
(ii) (bilinearity)
(a + b) ∧ c = a ∧ c + b ∧ c, c ∧ (a + b) = c ∧ a + c ∧ b,
(λ · a) ∧ b = λ · a ∧ b, a ∧ (λ · b) = λ · a ∧ b.
40 CHAPTER 1. AFFINE AND EUCLIDEAN GEOMETRY

(iii) We have a ∧ b = 0 if and only if a and b are parallel.

Again, the power of this operation is the duality between the properties just listed (proved by
algebra) and the geometric interpretation. Here is an application of the the wedge product.
Proposition 1.19.3. The points A, B, C are collinear if and only if A∧B+B∧C+C∧A = 0.
−→ −−→
Proof. The points A, B, C are collinear if and only if the vectors BA and CB are in one line.
They are in one line if and only if their spanned parallelogram degenerates to a segment, i.e.
has area 0. Hence A, B, C are in one line if and only if
0 = (A − B) ∧ (B − C) = A ∧ B − A ∧ C − B ∧ B + B ∧ C = A ∧ B + B ∧ C + C ∧ A.

Before Proposition 1.19.3 our algebraic interpretations of collinearity were Propositions 1.4.1
and 1.4.2. All the theorems proved using those two propositions have alternative proofs using
our new algebraic interpretation Proposition 1.19.3. We will not reprove earlier theorems
though (students may find it a good exercise), but rather give some new incidence theorem,
and as a change we will prove them using Proposition 1.19.3.
Theorem 1.19.4 (Newton-Gauss line). Let a, b, c, d be four pairwise intersecting lines (this
configuration is called a complete quadrilateral). The pairwise intersections are six points.
Three pairs of these six points are not connected by the lines a, b, c, d, these are called diag-
onals (P S, RQ, U V in the picture). The midpoints of the three diagonals are collinear.
1.19. WEDGE PRODUCT OF TWO PLANE VECTORS 41

Proof. Using the notation of the picture consider the sum of the following twelve terms

P ∧R P ∧Q S∧R S∧Q
R∧U Q∧V R∧V Q∧U
U ∧P V ∧P V ∧S U ∧S

We will view this sum in two different ways.

First: The sum of the terms in each column is zero, because the triples of points (P, R, U ),
(P, Q, V ), (S, R, V ), (S, Q, U ) are collinear, see Proposition 1.19.3. Hence the total sum is
zero.
Second: The first row is 4 times (P + S)/2 ∧ (R + Q)/2, the second row is 4 times (R +
Q)/2 ∧ (U + V )/2. The third row is 4 times (U + V )/2 ∧ (P + S)/2.
We conclude that
P +S R+Q R+Q U +V U +V P +S
∧ + ∧ + ∧ = 0,
2 2 2 2 2 2
and hence according to Proposition 1.19.3 (P + S)/2, (R + Q)/2, (U + V )/2 are collinear.

Theorem 1.19.5 (“parallel case of Pappus’ theorem”). Let A, B, C be collinear points and
let A0 , B 0 , C 0 be collinear. Then two of

AB 0 ||A0 B, BC 0 ||B 0 C, AC 0 ||A0 C

imply the third.

Proof. Consider the following three numbers

(A − B 0 ) ∧ (A0 − B), (B − C 0 ) ∧ (B 0 − C), (C − A0 ) ∧ (C 0 − A).

42 CHAPTER 1. AFFINE AND EUCLIDEAN GEOMETRY

The vanishing of these three numbers is equivalent to the three parallelity conditions of the
theorem, according to Proposition 1.19.2(iii).
However, one can distribute the sum of these three numbers, use the anti-symmetry property
of ∧ and conclude that the total sum is 0. Hence the vanishing of two of them indeed implies
the vanishing of the third one.

We will learn more on Pappus’ theorem in Section 2.

Problems

1.19.1. If A0 , B 0 , C 0 are the midpoints of BC, CA, AB respectively, then show that

4A0 ∧ B 0 ∧ C 0 = A ∧ B ∧ C.

Deduce that Area(A0 B 0 C 0 4) = 14 Area(ABC4).

1.19.2. Let D, E, F be points on the sides AB, BC, CA of a triangle, dividing the sides in
the ratios k1 : 1, k2 : 1, k3 : 1. Show that

Area(DEF 4) 1 + k1 k2 k3
= .
Area(ABC4) (1 + k1 )(1 + k2 )(1 + k3 )

1.20 A 3D view of plane geometry, triple product

In Calculus we learned the geometry and algebra of the triple product (a × b) · c of three
space vectors a, b, c. It will be convenient to use the following notation a ∧ b ∧ c = (a × b) · c,
and we may call it the triple product, or triple wedge product of a, b, c. In particular the
following hold

• a ∧ b ∧ c = (a × b) · c = a · (b × c);

• a ∧ b ∧ c is plus or minus the volume of the parallelepiped spanned by the space vectors
a, b, c;

•  
a1 a2 a3
(a1 , a2 , a3 ) ∧ (b1 , b2 , b3 ) ∧ (c1 , c2 , c3 ) = det  b1 b2 b3  ;
c1 c2 c3
1.20. A 3D VIEW OF PLANE GEOMETRY, TRIPLE PRODUCT 43

• (3-linearity) a ∧ b ∧ c is linear in each of the variables. Linearity in the first variable

means
(a ± a0 ) ∧ b ∧ c = a ∧ b ∧ c ± a0 ∧ b ∧ c,

(λa) ∧ b ∧ c = λ · a ∧ b ∧ c,
and linearity in the second and third variables are similar;

• (antisymmetry) a ∧ b ∧ c = b ∧ c ∧ a = c ∧ a ∧ b = −a ∧ c ∧ b = −c ∧ b ∧ a = −b ∧ a ∧ c.

In view of the second property above it makes no sense of considering the triple product
of vectors lying in the (x, y, 0) plane. A useful tool, however is considering our plane as
the z = 1 plane in 3-space. That is, if a point was (a1 , a2 ) earlier, now we consider it as
(a1 , a2 , 1).
What we gained is a new operation: we can form the triple product A ∧ B ∧ C for three
points. What we lost is that we partially lost our earlier operations: for example A + B does
not make sense any more since (a1 , a2 , 1) + (b1 , b2 , 1) = (a1 + b1 , a2 + b2 , 2) is not in the z = 1
plane any more. However, for example xA + yB makes sense if x + y = 1.

Proposition 1.20.1. For a triangle ABC4 in the z = 1 plane we have

1
Area(ABC4) = ± A ∧ B ∧ C,
2
and the sign is positive if and only if going around the triangle in the order A, B, C is
counterclockwise.

The proof is left as an exercise.

Corollary 1.20.2. The three points A, B, C of the z = 1 plane are collinear if and only if
A ∧ B ∧ C = 0.

Proof. Both conditions are equivalent to the condition that the volume of the parallelepiped
spanned by the space vectors A, B, C is zero.

Corollary 1.20.2 is now our 4th algebraic interpretation of collinearity of three points in the
plane—however this in new settings. Again, all theorems that were proved using any of the
earlier three interpretations (Propositions 1.4.1, 1.4.2, 1.19.3) can be proved with this new
one too—just we need to be careful using operation that make sense in the z = 1 plane. Let
us illustrate this with the following new proof Menelaus’ theorem 1.7.1.
44 CHAPTER 1. AFFINE AND EUCLIDEAN GEOMETRY

Proof. Consider our triangle in the z = 1 plane. We have K = xB + x0 C, L = yC + y 0 A,

M = zA + z 0 B with x + x0 = y + y 0 = z + z 0 = 1. According to Corollary 1.4.3 we have

0 = K ∧ L ∧ M = (xB + x0 C) ∧ (yC + y 0 A) ∧ (zA + z 0 B).

Using 3-linearity and antisymmetry of the triple product we can distribute the above expres-
sion and obtain
0 = (xyz + x0 y 0 z 0 )A ∧ B ∧ C.
Since ABC is a triangle (with non-zero area) we have A∧B∧C 6= 0. Therefore xyz+x0 y 0 z 0 = 0
which is a rearrangement of Menelaus’ theorem.

Problems

1.20.1. Give a proof of Proposition 1.20.1. [Hint: Let T be the tetrahedron with vertexes
(0, 0, 0), A, B, C. Use the geometric interpretation of the triple product to conclude that the
volume of T is plus or minus one sixth of the triple product ABC. Finish the proof by
observing that the volume of T is one third of the area of the ABC4.]

1.20.2. Reprove Ceva’s theorem, using the triple wedge operation. Hint: Try to rephrase
the “high-school style” proof from the end of Section 1.9.

1.20.3 (P). Let the side lengths of the ABC4 be a, b, c (a is opposite with A etc), and let
the angles be α, β, γ. Let the foot of the altitude from A be D. Prove that

aD = (b cos γ)B + (c cos β)C.

Deduce that the area of the triangle formed by the feet of the altitudes is

2 cos α cos β cos γ · Area(ABC4).

Chapter 2

Projective Geometry

2.1 Projective plane as extended affine plane

In relation with some exercises in Chapter 1 we developed the intuitive idea that we would
like to send a point in a configuration to “infinity”. This idea is made precise by defining
a new object, the projective plane P2 . Just like the affine plane has points and lines and
the incidence relation between them, the projective plane will also have points and lines
and incidence relation between them. However, the projective plane will have “more” points
and lines. Moreover the incidence relation will be somewhat more “complete” and more
“symmetric” than in the affine case.
Let a be a line on the Euclidean plane. The collection of all lines parallel with a will be called
a “parallelism class” of lines on R2 , and let’s call this object Ia . So, if a||b then Ia = Ib . One
way of thinking of the I∗ ’s as “symbols” associated with lines in such a way that parallel
lines have the same symbol.

Definition 2.1.1. Let points of P2 be

• points of R2 (these will be called affine points),

• all the Ia ’s (these will be called ideal points).

Let the lines of P2 be

• extended lines of R2 : extend each line in R2 with its own parallelism class, that is
extend a with Ia ,

• the collection of all ideal points is declared to be on one new line, the ideal line.

45
46 CHAPTER 2. PROJECTIVE GEOMETRY

Hence we added to the points of the plane R2 infinitely many new points, one for each
parallelism class of lines on R2 . We also added one new line. Observe that the lines coming
from lines in R2 are exactly one point “longer” on P2 .
PROJECT 6. One way of imagining Ia is imagining it at the “end” of a. Since a has
two “ends”, we must visualize it at both ends of a at the same time. In this visualization,
if a point “goes to infinity” in one direction of the projective line a ∪ {Ia }, then after it
disappears in one direction, it comes back at infinity at the other “end” of a. Let A, B, C be
the vertices of a triangle in a counterclockwise order. Imagine that we send this triangle to
“infinity” in one direction, and as described above, it reappears from the opposite direction.
Will the reappearing triangle be named ABC clockwise or counterclockwise?

2.2 Incidence properties of P2

The following observation can be established by checking all possible cases.
Proposition 2.2.1. For any two points there is a unique line passing through them. For
any two lines there is a unique point contained in both of them.

2.3 Remarkable incidence theorems

Desargues’ theorem holds in P2 .
Theorem 2.3.1 (Desargues’ theorem—projective). Two triangles are perspective with re-
spect to a point if and only if they are perspective with respect to an axis.
PROJECT 7. Learn the “vector calculus” of projective geometry—by searching for “pro-
jective coordinates”. Then prove the projective Desargues’ theorem. The calculation should
be similar to the calculation in the proof of the affine Desargues’ theorems in Chapter 1.
PROJECT 8. Let S1 and S2 be planes in R3 intersecting in a line. Let A be a point in R3 ,
not on these planes. There is a natural notion of projecting points of S1 to points of S2 from
A (see Figure ?). We obtain a map pA from a subset of S1 to S2 . Observe that this map is
not defined for all points of S1 and that it is not surjective. Find a natural extension
p̃A : P2 (S1 ) → P2 (S2 )
where P2 (Si ) is the projective plane obtained by extending Si , and conclude that p̃A is
bijective. Interpret the outcome as a visualization of the ideal line of S1 in S2 . Use this idea
to prove the projective Desargues’ theorem, based on the fact that we already proved the
affine Desargues’ theorem in Chapter 1.
2.4. PROBLEMS 47

Pappus’ theorem holds in P2 .

Theorem 2.3.2 (Pappus’ theorem—projective version). If A, B, C are collinear and A0 , B 0 , C 0
are collinear, then K = AB 0 ∩ A0 B, L = BC 0 ∩ B 0 C, and M = CA0 ∩ C 0 A are also collinear.
PROJECT 9. Prove the projective Pappus’ theorem using projective coordinates, c.f. Pro-
ject 7.
PROJECT 10. Prove the projective Pappus’ theorem using the idea of Project 8.

2.4 Problems
1. Prove Proposition 2.2.1.
The next three problems refer to the following part of the projective Desargues theorem: If
ABC4 and A0 B 0 C 0 4 are perspective w.r.t. a point P , then K = AB ∩A0 B 0 , L = BC ∩B 0 C 0 ,
and M = CA ∩ C 0 A0 are collinear.
2. Phrase the affine geometry theorem that we obtain from this theorem if K is an ideal point
(and none of the other points are ideal). You do not need to prove this theorem, just phrase
it.
3. Phrase the affine geometry theorem that we obtain from this if theorem A is an ideal point
(and none of the other points are ideal). You do not need to prove this theorem, just phrase
it.
4. Phrase the affine geometry theorem that we obtain from this if theorem P, A, A0 is an ideal
point (and none of the other points are ideal). You do not need to prove this theorem, just
phrase it.
The next two problems refer to the projective Pappus’ Theorem 2.3.2.
48 CHAPTER 2. PROJECTIVE GEOMETRY

5. Phrase the affine geometry theorem that we obtain from this theorem if K and L are ideal
points (and none of A, B, C, A0 , B 0 , C 0 are ideal). You do not need to prove this theorem, just
phrase it.

6. Phrase the affine geometry theorem that we obtain from this theorem if A, B, C are ideal
points (and none of the other points are ideal). You do not need to prove this theorem, just
phrase it.
Chapter 3

Spherical Geometry

3.1 Points, lines, triangles, polarity

In spherical geometry the role of the “plane” will be played by the sphere S 2 = {x ∈ R3 :
||x|| = 1}. When we say point, we mean a point on S 2 .
The role of straight lines will be played by the “great circles” in S 2 . A great circle is defined
to be the intersection S 2 ∩ P where P is a plane in R3 containing the origin.
Now we have a geometry of “points” and “lines”. The following point-line incidence proper-
ties are obvious.

Proposition 3.1.1. In spherical geometry

• given two non-antipodal points there is a unique line passing through them;

• the intersection of any two lines is exactly two points.

Definition 3.1.2.

• For two (not antipodal) points of S 2 there is exactly on great circle connecting them.
The two points divide this great circle to two arcs. The shorter of the two will be called
a spherical segment connecting the two points.

• For three spherical points A, B, C (that are not contained in a spherical line) the union
of the three spherical segments AB, BC, CA is called a spherical triangle.

49
50 CHAPTER 3. SPHERICAL GEOMETRY

• For a spherical line a = S 2 ∩ P there are two points X, Y on S 2 with X ⊥ P , Y ⊥ P .

We will call these two points the poles of a.

• Let ABC be spherical triangle. Let C 0 be a pole of AB contained in the same hemi-
sphere as C. Let B 0 be a pole of CA contained in the same hemisphere as B. Let A0
be a pole of BC contained in the same hemisphere as A. The spherical triangle A0 B 0 C 0
is called the polar triangle of ABC.

Theorem 3.1.3 (Bipolar Theroem). The polar triangle of the polar triangle of ABC4 is
itself.

Proof. For the purpose of this proof, if X and Y are on S 2 and X ⊥ Y then we will call the
spherical segment XY a quadrant.
We have that AB 0 is a quadrant because B 0 is a pole of AC. The segment AC 0 is a quadrant,
because C 0 is a pole of AB.
Since AB 0 and AC 0 are quadrant, A must be a pole of B 0 C 0 . What remains to be proved is
that A is on the same side of B 0 C 0 as A0 .
The points A and A0 are on the same side of BC by definition of A0 . Moreover A0 is a pole
of BC. Hence the segment AA0 is less than a quadrant.
Since AA0 is less than a quadrant, and A is a pole of B 0 C 0 (proved above), we must have
that A and A0 is on the same side of B 0 C 0 .

3.2 Length, angle

Definition 3.2.1. A vector v ∈ R2 is called tangent to S 2 at X ∈ S 2 if v ⊥ X. The

collection of tangent vectors to S 2 at X is called the tangent plane TX S 2 .

It is natural to imagine arrows representing tangent vectors in such a way that the arrows
start at X. Thus TX S 2 coincides with the usual notion of a plane tangent to S 2 at X.
3.2. LENGTH, ANGLE 51

Given a spherical segment connecting X and Y there is a unique unit vector v ∈ TX S 2 which
is determined by Y = αX + βv with β > 0. The vector v will be called the unit tangent
vector at X to the segment from X to Y .

Definition 3.2.2. The distance d(X, Y ) between points X and Y is defined to be

d(X, Y ) = arccos(X · Y ).
Because of the geometric interpretation of dot product, d(X, Y ) is in fact the angle between
the vectors X and Y . Also, since the radius of S 2 is 1, this angle is the same as the length
of the spherical segment XY . Thus the odd looking definition of distance above is natural:
it is the length measured on the sphere.
Let v ∈ TX S 2 be a unit vector and consider the parameterized curve
s(t) = (cos t)X + (sin t)v. (3.1)
First observe that this curve lies in S 2 . Indeed, for any t we have
s(t)s(t) = ((cos t)X + (sin t)v)((cos t)X + (sin t)v) =
XX cos2 t + 2 |{z}
|{z} vv sin2 t = cos2 t + sin2 t = 1.
Xv cos t sin t + |{z}
1 0 1
The curve is also in the plane spanned by X and v. Hence the curve is in the intersection
of a plane with the sphere, that is in a spherical line.
It will be important to study the parametrization (3.1) with respect to the distance notion
defined above. Let a be in (0, π). Then s([0, a]) is a spherical segment and v is the unit
vector tangent to the spherical segment Xs(a) at X.
We claim that the length of s([0, a]) is a. Indeed,
d(X, s(a)) = arccos(X · ((cos a)X + (sin a)v)) = arccos(cos a) = a.
Moreover, for any t1 , t2 with |t1 − t2 | < π we have that the length of the s(t1 )s(t2 ) segment
is |t1 − t2 | (Ex. 7 below.) This remarkable property of the parametrization (3.1) is called
“arc-length parametrization”, or “natural parametrization”.
52 CHAPTER 3. SPHERICAL GEOMETRY

Definition 3.2.3. The XY Z∠ is defined to be the angle of the following two vectors: the
unit tangent vector at Y to the segment Y X, and the unit tangent vector at Y to the segment
Y Z.

3.3 Sides and angles of polar triangles

Theorem 3.3.1. Let a, b, c be the sides and α, β, γ be the angles of a triangle ABC4. Let
a0 , b0 , c0 be the sides and α0 , β 0 , γ 0 be the angles of the polar triangle A0 B 0 C 0 4. Then

a0 = π − α, b0 = π − β, c0 = π − γ,

a = π − α0 , b = π − β 0, c = π − γ 0.

Proof. Because of symmetry and the Bipolar Theorem 3.1.3 it is enough to prove one of the
six statements. Let K = AB ∩ B 0 C 0 , L = AC ∩ B 0 C 0 . Since A is a pole of B 0 C 0 we have
KL = α.

Then B 0 C 0 = B 0 L + KC 0 − KL = π/2 + π/2 − α, proving a0 = π − α.

3.4. LAWS OF COSINES AND SINES 53

3.4 Laws of Cosines and Sines

Theorem 3.4.1 (Spherical Law of Cosines). In ABC4 on the sphere (with usual notations)
we have
cos c = cos a cos b + sin a sin b cos γ.
Proof. Let u be the unit tangent vector of CA at C. Let v be the unit tangent vector of
CB at C. Then A = cos b · C + sin b · u, B = cos a · C + sin a · v. Therefore c = AB =
arccos(A · B) = arccos (cos b cos a + sin b sin a u · v), but u · v = cos γ.
Corollary 3.4.2 (Spherical Pythagorean theorem). In a right angled triangle with hy-
pothenuse c on S 2 we have cos c = cos a cos b.
Proof. This is the Spherical Law of Cosines for γ = π/2.

Project. In Spherical Law of Cosines and in Spherical Pythagorean do the following. Write
a · t for a etc for all side measurements. Then expand in Taylor series with respect to t.
Study the coefficients of 1, t, t2 .

Theorem 3.4.3 (Spherical triangle-inequality). In a spherical triangle we have c < a + b.

Proof. We have
cos c = cos a cos b + sin a sin b cos γ;
cos(a + b) = cos a cos b − sin a sin b.
The first line is the Law of Cosines, the second line is a trig identity. From trig we have
sin a sin b ∈ (0, 1), cos γ ∈ (−1, 1). Therefore the RHS of the first line is larger than the RHS
of the second line. Hence the same holds for the LHSs, i.e. cos c > cos(a + b). Since cos is
strictly monotone decreasing on (0, π) this implies c < a + b.
Theorem 3.4.4 (Dual Law of Cosines). In ABC4 on the sphere (with usual notations)

− cos γ = cos α cos β − sin α sin β cos c.

Proof. Apply the Law of Cosines to the polar triangle.

Remark 3.4.5. The dual Law of Cosines calculates γ if α, β and c are given. In Euclidean
geometry one can calculate γ as long as α and β are given, no need for c! Also, this Theorem
shows that the sum of the angles of a triangle is not constant, keeping α and β the same,
but changing c does change γ. More on this is coming in the next section.
54 CHAPTER 3. SPHERICAL GEOMETRY

Theorem 3.4.6 (Spherical Law of Sines). In ABC4 on the sphere (with usual notations)
sin α sin β sin γ
= = .
sin a sin b sin c
Proof. From the Law of Cosines we have
cos c − cos a cos b
cos γ = .
sin a sin b
From this, using sin2 + cos2 = 1 we have
√
1 − cos2 a − cos2 b − cos2 c + 2 cos a cos b cos c
sin γ = .
sin a sin b
Divide by sin c and see that the RHS becomes symmetric in a, b, c.
Theorem 3.4.7. The common value of sin α/ sin a = sin β/ sin b = sin γ/ sin c is equal
6 Vol(OABC-tetrahedron)
,
sin a sin b sin c
√
in other words the -expression in the proof above is equal 6 times the volume of the par-
allelepiped.
Proof. We can move the ABC4 into the special position A = (1, 0, 0), B = (?, positive, 0), C =
(?, ?, positive). Then a good picture and calculation shows that
A = (1, 0, 0), B = (cos c, sin c, 0), C = (cos b, sin b cos α, sin b sin α).
Using the determinant formula for the volume of the OABC tetrahedron we obtain the
theorem.

3.5 Anglesum, area, Girard formula

If two great circles intersect in angle θ, then the domain between the two (on both sides)
will be called a θ-lune.
3.5. ANGLESUM, AREA, GIRARD FORMULA 55

Lemma 3.5.1. The area of a θ-lune is 4θ.

Proof. The area of a θ-lune is a linear function of θ. For θ = π we get the surface area of
S 2 , which is 4π.
Theorem 3.5.2 (Girard’s formula). The sum of the angles of a spherical triangle is π plus
the area of the triangle.
Proof. Consider a spherical triangle T with angles α, β, γ. The α-lune, β-lune, and γ-lune
together cover the sphere 1-sheeted, except they cover T 3 times, as well as a congruent copy
T 0 of T “on the other side” also 3 times.

Hence
Area(α-lune) + Area(β-lune) + Area(γ-lune) = Area(S 2 ) + 2 Area(T ) + 2 Area(T 0 ).
And therefore, using Lemma 3.5.1 we have
4α + 4β + 4γ = 4π + 4 Area(T ),
56 CHAPTER 3. SPHERICAL GEOMETRY

which proves the theorem.

Corollary 3.5.3. The sum of the angles of a spherical triangle is always strictly larger
than π.

Remark 3.5.4. In Euclidean geometry the sum of the angles is π, so the Girard formula says
that in spherical geometry the sum of angles is the Euclidean value plus the area.
A spherical triangle is always contained in a hemisphere. Thus, the area of a spherical
triangle is at most half the area of S 2 , that is 2π. Hence, from Girard’s formula we obtain
an upper bound for the angle sum as well, namely

π < α + β + γ < 3π.

Proposition 3.5.5. For the perimeter p of a triangle on the sphere we have 0 < p < 2π.

Proof. Applying the π < α + β + γ < 3π estimate for the polar triangle we obtain the
statement.

Theorem 3.5.6. The sum of the angles of a spherical quadrilateral is 2π plus the area.

Proof. Cut the quadrilateral by a diagonal to two triangles T1 and T2 . For the two triangles
we have Girard’s formula
α + β1 + δ1 = π + Area(T1 ),

γ + β2 + δ2 = π + Area(T2 ).

Adding together we obtain

α + β + γ + δ = 2π + Area(T1 ∪ T2 ).

Theorem 3.5.7. The sum of the angles of a spherical n-gon is the Euclidean value (n − 2)π
plus the area of the n-gon.

Proof. The idea of the preceding proof can be applied iteratively.

3.6. PROBLEMS 57

3.6 Problems
7. Consider the parametrization (3.1). Prove that for |t1 − t2 | < π we have that the length
of the s(t1 )s(t2 ) segment is |t1 − t2 |.

8. The coordinates of Chapel Hill NC are 35.93o N, 79.03o W. Look up what these coordinates
mean. Look up the coordinates of Paris, France. Find the distance between Chapel Hill and
Paris (call the radius of the Earth R).

9. Let L and W be points on the Earth (radius R), with coordinates 38.8o N, 9.15o W and
38.8o N, 77o W (why L and W?). What is their spherical distance?

10. The points L and W of the exercise above are on the same latitude (38.8o N). What is
their distance measured on that latitude?

11. Prove Theorem 3.5.7 in detail.

12. We saw above that π < angle sum of a spherical triangle < 3π. Prove that this estimate
is sharp.

13. Let P be a plane not containing the origin. Prove that if P ∩ S 2 is not empty, then
it is a circle in spherical geometry. That is, prove that there exist a point C ∈ S 2 and a
nonnegative number r such that P ∩ S 2 = {A ∈ S 2 : d(A, C) = r}.

14.
(a) Let the sides of a triangle in spherical geometry be .3, .4, and .5. Find its angles.
(b) Let the sides of a triangle in spherical geometry be .03, .04, and .05. Find its angles.
(c) Let the sides of a triangle in spherical geometry be .003, .004, and .005. Find its angles.

15. Consider Chapel Hill, Chicago, and Seattle. Find the distances between any two of these
cities online. Using these distances find the angle between the Chapel Hill-Chicago and the
Chapel Hill-Seattle segments.

16. Consider Chapel Hill, Chicago, and Seattle. Find the distances between any two of these
cities online. Calculate the area of the triangle with these vertices on the surface of the
Earth.

17. Let a, b, c and α, β, γ be the side lengths and angles of a spherical triangle. Assume
γ = π/2 (that is, we have a right triangle). Express sin α in terms of the sides. Verify that
for small sides your result is close to the Euclidean value of a/c.
58 CHAPTER 3. SPHERICAL GEOMETRY

18. Give a proof of the spherical Law of Sines based on the result of Exercise 17 along the
lines of Proof-2 of the Euclidean Law of Sines (Theorem 1.16.4).

19. Let a, b, c and α, β, γ be the side lengths and angles of a spherical triangle. Assume
γ = π/2 (that is, we have a right triangle). Express cos α in terms of the sides. Verify that
for small sides your result is close to the Euclidean value of b/c.

20. Let a, b, c and α, β, γ be the side lengths and angles of a spherical triangle. Assume
γ = π/2 (that is, we have a right triangle). Express tan α in terms of the sides. Verify that
for small sides your result is close to the Euclidean value of a/b.