Preface

As a sequel to 113 Geometric Inequalities from the AwesomeMath

Summer Program, this book extends the themes discussed in the former
book and broadens a problem-solver’s competitive arsenal. The beauty
of geometry is explored from another perspective, namely the mechanics
associated with defining and breaking down the aspects of such visual and
numerically-involved problems. One may wonder how complex figures can
be described through linear transformations; nevertheless, many creative
and powerful techniques are featured. Strategies from multiple fields, such
as Algebra, Calculus, and pure Geometry provide the reader with a varied
subset of methods useful in mathematics competitions. Starting with the
fundamentals such as the triangle inequality and “broken lines”, the book
progresses increasingly to more sophisticated machinery such as the Averaging
Method, Quadratic Forms, Finite Fourier Transforms, Level Curves, the
Erdös-Mordell and Brunn-Minkowski Inequalities, and the Isoperimetric
Theorem, to name a few. Despite the focus on contest-style problems, rich
theory and generalizations accompany the aforementioned topics to supply the
curious reader with a rigorous exploration of fields associated with geometric
inequalities. To help the student better assimilate these many techniques,
intuitive motivations and well-organized solutions are listed in relative order
of difficulty. Frequently, even three solutions are provided to link the many
interconnected areas of mathematics that converge in unexpected geometric
themes. The study of geometric inequalities will also indirectly strengthen
the reader’s ability to analyze, dissect, and invent creative methods, all skills
that are necessary to succeed in mathematics competitions. We hope that
this book will illuminate some underappreciated geometric arguments and
fortify the reader’s mathematical toolkit.

Many thanks to Mircea Becheanu, Gabriel Dospinescu, and Christian Yankov

for several useful discussions on some of the topics featured in the book. And
thank you to Chris Jeuell and Adrian Andreescu for significantly improving
and polishing the text.
vi

Enjoy the book!

Titu Andreescu, Oleg Mushkarov
Contents

1 The Triangle Inequality 1

1.1 Broken Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Shortest Paths . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.3 The Averaging Method . . . . . . . . . . . . . . . . . . . . . . . 30
1.4 Practice Problems . . . . . . . . . . . . . . . . . . . . . . . . . 39

2 Algebraic Methods 45
2.1 Algebraic Inequalities . . . . . . . . . . . . . . . . . . . . . . . 45
2.2 The Dot Product . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.3 Complex Numbers . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.4 Finite Fourier Transforms . . . . . . . . . . . . . . . . . . . . . 73
2.5 Quadratic Forms . . . . . . . . . . . . . . . . . . . . . . . . . . 85
2.6 Practice Problems . . . . . . . . . . . . . . . . . . . . . . . . . 100

3 Analytic Methods 105

3.1 Employing Calculus . . . . . . . . . . . . . . . . . . . . . . . . 105
3.2 Partial Variations . . . . . . . . . . . . . . . . . . . . . . . . . . 114
3.3 Level Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
3.4 Distances Between Sets . . . . . . . . . . . . . . . . . . . . . . 140
3.5 Practice Problems . . . . . . . . . . . . . . . . . . . . . . . . . 155

4 The Fundamental Inequality in Triangles 161

4.1 An Analytic Proof of The Fundamental Inequality . . . . . . . 162
4.2 An Algebraic Proof of The Fundamental Inequality . . . . . . . 165
x Contents

4.3 A Geometric Proof of The Fundamental Inequality . . . . . . . 168

4.4 Sharp Inequalities in Triangles . . . . . . . . . . . . . . . . . . 169
4.5 Proving Algebraic Inequalities . . . . . . . . . . . . . . . . . . . 176
4.6 Practice Problems . . . . . . . . . . . . . . . . . . . . . . . . . 179

5 The Erdös-Mordell Inequality 183

5.1 Seven Proofs of the Erdös-Mordell Inequality . . . . . . . . . . 183
5.2 Examples of Erdös-Mordell Type Inequalities . . . . . . . . . . 190
5.3 A Generalization of the Erdös-Mordell Inequality . . . . . . . . 201
5.4 The Erdös-Mordell Inequality for Polygons . . . . . . . . . . . 205
5.5 Practice Problems . . . . . . . . . . . . . . . . . . . . . . . . . 208

6 Area Inequalities 211

6.1 Inscribed and Circumscribed Polygons . . . . . . . . . . . . . . 211
6.2 Malfatti’s Marble Problem . . . . . . . . . . . . . . . . . . . . . 225
6.3 The Brunn-Minkowski Inequality . . . . . . . . . . . . . . . . . 238
6.4 Practice Problems . . . . . . . . . . . . . . . . . . . . . . . . . 250

7 Isoperimetric Problems 253

7.1 The Isoperimetric Theorem . . . . . . . . . . . . . . . . . . . . 253
7.2 Isoperimetric Problems . . . . . . . . . . . . . . . . . . . . . . . 258
7.3 Practice Problems . . . . . . . . . . . . . . . . . . . . . . . . . 263

8 Hints and Solutions 265

8.1 The Triangle Inequality . . . . . . . . . . . . . . . . . . . . . . 265
8.2 Algebraic Methods . . . . . . . . . . . . . . . . . . . . . . . . . 287
8.3 Analytic Methods . . . . . . . . . . . . . . . . . . . . . . . . . . 307
8.4 The Fundamental Inequality in Triangles . . . . . . . . . . . . . 346
8.5 The Erdös-Mordell Inequality . . . . . . . . . . . . . . . . . . . 371
8.6 Area Inequalities . . . . . . . . . . . . . . . . . . . . . . . . . . 387
8.7 Isoperimetric Problems . . . . . . . . . . . . . . . . . . . . . . . 405

Bibliography 419
Chapter 1

The Triangle Inequality

The triangle inequality and its generalization for broken lines is one of the
most basic tools for proving distance inequalities in the plane and space. In
this chapter we consider several classical examples of such inequalities as well
as some related practical problems for shortest paths in the plane. The last
section is devoted to the so-called averaging method for proving inequalities
for the lengths of broken lines.

1.1 Broken Lines

The triangle inequality says that for any three points A, B, C, we have

AB + BC ≥ CA.

It follows by induction that for any points A1 , A2 , . . . , An , n ≥ 3 in the

plane (space) (Fig. 1.1) the following generalized triangle inequality holds true:
A1 A2 + A2 A3 + · · · + An−1 An ≥ A1 An . (1.1)
The equality in (1.1) occurs if and only if points A2 , . . . , An−1 lie on the seg-
ment A1 An in this order.
We will also use the vector analog of the triangle inequality which says that
for any vectors →
−
a1 , →
−
a2 , . . . , −
a→
n in the plane (space), we have:

|→
−
a1 | + |→
−
a2 | + · · · + |−
a→ →
− → − −
→
n | ≥ |a1 + a2 + · · · + an |.
2 Chapter 1

A2 A n -1

A1 An

Figure 1.1

Equality occurs if and only if the vectors →

−
a1 , →
−
a2 , . . . , −
a→
n are collinear and equally
oriented.

Example 1.1. Let M be a point inside triangle ABC. Prove that:

(a) M A + M B < CA + CB;

(b) M A + M B + M C < max(AB + BC, BC + CA, CA + AB).

Solution. (a) Let N be intersection of lines AM and BC (Fig. 1.2). Then by

the triangle inequality, BM < M N + N B and AN < CA + CN . Hence

AM + BM < AM + M N + BN = AN + BN < CA + CN + N B = CA + CB.

(b) Without loss of generality, let AB ≤ BC ≤ CA. Draw the lines

through M parallel to the sides of the triangle and denote by A1 and A2 , B1
and B2 , C1 and C2 the points where they intersect BC, CA, AB, respectively
(Fig. 1.3). Then triangles A1 A2 M , M B1 B2 , C2 M C1 are similar, and their
shortest sides are M A1 , M B2 , C1 C2 , respectively. This together with the
triangle inequality implies

M A + M B + M C < (AB2 + B2 M ) + (M A1 + A1 B) + (M A2 + A2 C)
< (AB2 + B2 B1 ) + (A1 A2 + A1 B) + (CB1 + A2 C)
= AC + BC.

Here we have used the fact that M A2 = B1 C. (Why?) 

The Triangle Inequality 3

C C

B1
A2
N

M M
B2 A1

A B A C1 C2 B

Figure 1.2 Figure 1.3

Example 1.2. Let M be a point on a segment AB and K a point in the plane

(space). Prove that:
(a) If M is the midpoint of AB, then
KA + KB
KM ≤ .
2

MB
(b) If = λ, 0 < λ < 1, then
AB
KM ≤ λ · KA + (1 − λ)KB.

Equality in (a) and (b) is attained if and only if the points A, B, K are
collinear and K lies outside the segment AB.

(c) If G is the centroid of triangle ABC, then

KA + KB + KC
KG ≤ .
3
4 Chapter 1

Equality is attained if and only if the points A, B, C, K are collinear and

K lies outside the segments AB, BC, and CA.

Solution. (a) Consider the point N such that AN BK is a parallelogram

(Fig. 1.4). Then

1 1 1
KM = KN ≤ (KB + BN ) = (KB + KA).
2 2 2
1
Note also that this inequality is a special case of (b) for λ = .
2
B N

K A

Figure 1.4

(b) Let A1 and B1 be points on the segments KA and KB such that

M A1 k KB and M B1 k KA (Fig. 1.5). Then

MB MA
M B1 = · KA = λ · KA, M A1 = · KB = (1 − λ)KB
AB AB
and the desired inequality follows from the triangle inequality:

KM ≤ M B1 + B1 K = M B1 + M A1 = λ · KA + (1 − λ)KB.

This inequality also follows from the identity

−−→ −−→ −−→
KM = λKA + (1 − λ)KB
The Triangle Inequality 5

B1 M

K A1 A

Figure 1.5

and the triangle inequality for vectors. Equality in (a) and (b) is attained if
−−→ −−→
and only if the vectors KA and KB are collinear, i.e., the points A, B, K are
collinear and K lies outside the segment AB. 
GM 1
(c) Let M be the midpoint of segment AB. We know that = .
CG 3
Hence from (b) and (a), it follows that
1 1
KG ≤ (KC + 2KM ) ≤ (KC + KA + KB).
3 3
Another way to prove the desired inequality is to use the identity
−−→ −−→ −−→ −−→
KA + KB + KC = 3KG

and the triangle inequality for vectors. Equality in (c) is attained if and only if
−−→ −−→ −−→
the vectors KA, KB, KC are collinear, i.e., the points A, B, C, K are collinear
and K lies outside the segments AB, BC, and CA. 
Example 1.3. (Heron’s problem) Two points A and B lie on one side of a
straight line l. Find a point C on l such that CA + CB is minimized.
Solution. Let B 0 be the reflection of B in l (Fig. 1.6). Then BG = B 0 G and
the triangle inequality for 4ACB 0 implies

AC + CB = AC + CB 0 ≥ AB 0 .
6 Chapter 1

C C

Figure 1.6

Equality occurs precisely when C is the intersection point C0 of l and the line
segment AB 0 . 
Remark. The above problem has been considered about 2000 years ago by
Heron who stated that the shortest distance between A and B via a line l is
exactly the path traversed by a ray of light emitted from A and observed at
B. From here he deduced that when light is reflected in a mirror, the angle of
incidence is equal to the angle of reflection.
Example 1.4. (Ptolemy’s inequality) For any four points A, B, C, D in the
plane, we have
AC · BD ≤ AB · CD + BC · AD.
Equality holds if and only if ABCD is a cyclic quadrilateral.
−−→
First Solution. We may assume that B lies inside ∠ADC. On rays DA,
−−→ −−→
DB, DC, consider the points A1 , B1 , C1 , respectively, such that
1 1 1
DA1 = , DB1 = , DC1 = .
DA DB DC
Then 4ABC ∼ 4A1 B1 C1 and so
AB BC CA
A1 B 1 = , B1 C1 = , C1 A1 = .
DA · DB DB · DC DC · DA
The Triangle Inequality 7

The desired inequality follows from the triangle inequality:

A1 B1 + B1 C1 ≥ A1 C1 .

Equality holds if and only if B1 lies on the segment A1 C1 , that is, when

∠BAD + ∠BCD = ∠A1 B1 D + ∠C1 B1 D = 180◦ . 

Second Solution. See the solution of Example 2.18 using complex numbers.


Example 1.5. (Pompeiu’s theorem) Let ABC be an equilateral triangle and

let M be a point in its plane. Prove that the segments AM , BM , CM are
side lengths of a triangle. Also prove that this triangle is degenerate if and
only if M lies on the circumcenter of triangle ABC.

First Solution. By Ptolemy’s inequality for points A, M, B, C, it follows that

AB · CM ≤ AM · BC + BM · AC.

Since AB = BC = CA, we get CM ≤ AM +BM . Similarly, BM ≤ CM +AM

and AM ≤ BM + CM . We have equality in one of these inequalities, say in
the first one, if and only if AMBC is a cyclic quadrilateral. 
Second Solution. Consider the rotation of 60◦ about A, and let M1 be the
image of M (Fig. 1.7). Then AM = M M1 , CM1 = BM , and 4M M1 C is the
desired triangle. 
Note that it degenerates if and only if points M1 , C, M are collinear which
implies that M lies on the circumcircle of triangle ABC. (Why?)

Example 1.6. (IMO 1995) Let ABCDEF be a convex hexagon with AB =

BC = CD and DE = EF = F A, such that ∠BCD = ∠EF A = 60◦ .
Suppose G and H are points in the interior of the hexagon such that
∠AGB = ∠DHE = 120◦ . Prove that AG + GB + GH + DH + HE ≥ CF .

Solution. Triangles BCD and EF A are equilateral, and hence BE is an axis

of symmetry of ABDE. Let C 0 , F 0 respectively be the points symmetric to
C, F with respect to BE. Points G and H lie on the circumcircles of ABC 0 and
8 Chapter 1

M1

B
A

Figure 1.7

DEF 0 , respectively, because, for instance, ∠AGB = 120◦ = 180◦ − ∠AC 0 B.

Hence from Ptolemy’s Theorem, we have AG + GB = C 0 G and DH + HE =
HF 0 . Therefore

AG + GB + GH + DH + HE = C 0 G + GH + HF 0 ≥ C 0 F 0 = CF,

with equality if and only if both G and H lie on C 0 F 0 . 

Remark. By Ptolemy’s inequality (Example 1.4),

AG + GB ≥ C 0 G, DH + HE ≥ HF 0 ,

so the result holds without the condition ∠AGB = ∠DHE = 120◦ .

Example 1.7. Among all quadrilaterals ABCD with AB = 3, CD = 2, and

∠AM B = 120◦ , where M is the midpoint of CD, find the one of minimal
perimeter.
The Triangle Inequality 9

Solution. Let C 0 and D0 be the reflections of C and D in the lines BM and

AM , respectively (Fig. 1.8). Then triangle C 0 M D0 is equilateral because

1
C 0 M = D0 M = CD and ∠C 0 M D0 = 180◦ − 2∠CM B − 2∠DM A = 60◦ .
2

D M C

C0

A B

D0

Figure 1.8

Hence
1
AD + CD + CB = AD0 + D0 C 0 + C 0 B ≥ AB.
2
1
It follows that AD + CB ≥ AB − CD = 2. Thus AB + BC + CD + DA ≥ 7,
2
with equality if and only if C 0 and D0 lie on AB.
In the latter case, ∠ADM = ∠AD0 M = 120◦ , ∠BCM = ∠BC 0 M = 120◦ ,
and ∠AM D = 60◦ − ∠CM B = ∠CBM . Hence triangles AM D and M BC
are similar, implying
CD 2
 
AD · BC = = 1.
2
On the other hand, AD + BC = 2, and we conclude that AD = BC = 1.
Therefore the quadrilateral ABCD of minimum perimeter is an isosceles trape-
zoid with sides AB = 3, BC = AD = 1, and CD = 2 (Fig. 1.9). 
The next problem was first stated by Giovanni Fagnano in 1775.

Example 1.8. (Fagnano’s problem) Prove that of all triangles inscribed in a

given acute triangle, the orthic triangle has the least perimeter.
10 Chapter 1

D M C

A D0 C0 B

Figure 1.9

First Solution. Let ABC be the given triangle and let M , N , P be arbitrary
points on the sides AB, BC, CA, respectively. Denote by E and F the
respective feet of the perpendiculars from M to AC and BC (Fig. 1.10).
C

N
P

Q R

E F

A M B

Figure 1.10

Then the quadrilateral M F CE is inscribed in the circle with diameter CM

and therefore EF = CM sin ∠C. Let Q and R be the midpoints of M P and
M N , respectively. Then
M N + N P + P M = 2F R + 2QR + 2QE ≥ 2EF = 2CM sin ∠C.
The Triangle Inequality 11

Let AA1 , BB1 , CC1 be the altitudes of triangle ABC, and let E1 and F1 be
the feet of the perpendiculars from C1 to AC and BC, respectively (Fig. 1.11).
Then E1 F1 = CC1 sin ∠C. Denote by Q1 and R1 the midpoints of C1 B1 and
C1 A1 , respectively.
C

A1

B1 F1
R1
Q1
E1

A C1 B

Figure 1.11

Then
∠E1 Q1 B1 = 2∠E1 C1 B1 = 2∠C1 B1 B = ∠C1 B1 A1 ,
which shows that E1 Q1 k A1 B1 . Similarly, F1 R1 k A1 B1 . It follows that E1 ,
Q1 , R1 , F1 are collinear and

A1 B1 + B1 C1 + C1 A1 = 2Q1 R1 + 2Q1 E1 + 2R1 F1 = 2E1 F1 = 2CC1 sin ∠C.

Thus

M N + N P + P M = 2CM sin ∠C ≥ 2CC1 sin ∠C = A1 B1 + B1 C1 + C1 A1 .

Hence of all triangles M N P inscribed in triangle ABC, the orthic triangle

A1 B1 C1 has the least perimeter. 
Remark. Fagnano’s problem can also be solved in the case when the given
triangle is not acute-angled. Assume, for example, that ∠ACB ≥ 90◦ . It is
12 Chapter 1

not difficult to see that in this case, triangle M N P with minimal perimeter
occurs when N = P = C and M is the foot of the altitude of triangle ABC
through C. In this case, triangle M N P is degenerate.
We now consider the analog of Fagnano’s problem for convex polygons.
Note that for any n ≥ 4, there are convex n-gons that have no inscribed
n-gons of minimal perimeter.

Example 1.9. Let A be a convex n-gon with vertices A1 , A2 , . . . , An and let

B be an inscribed n-gon with vertices Bi ∈ Ai Ai+1 , 1 ≤ i ≤ n, An+1 = A1 .
Then B has minimal perimeter amongst all inscribed n-gons in A if and only
if ∠Bn B1 A1 = B2 B1 A2 , ∠B1 B2 A2 = B3 B2 A3 , . . . , ∠Bn−1 Bn An = ∠B1 Bn A1 .

Solution. Assume that B has minimal perimeter amongst all inscribed n-gons
in A but the given condition is not satisfied. Let, for example, ∠Bn B1 A1 6=
∠B2 B1 A2 . Consider point B20 symmetric to B2 with respect to the line A1 A2
(Fig. 1.12). Denote by B the intersection of lines B20 Bn and A1 A2 . Since
∠B2 BM = ∠B20 BM = ∠Bn BA1 6= ∠B2 B1 A2 points B and B1 differ.
A3
An

B2
Bn

B1
M
A1 B A2

B20

Figure 1.12

Take a point B10 on the segment BB1 so that B10 lies on the side A1 A2 . Consider
points B10 , B2 , . . . , Bn .
 The Triangle Inequality 13

According to Example 1.1, (i), B20 B10 + B10 Bn < B20 B1 + B1 Bn ,

thus B2 B10 + B10 Bn < B2 B1 + B1 Bn , which means that

B10 B2 + B2 B3 + · · · + Bn−1 Bn + Bn B10 < B1 B2 + B2 B3 + · · · + Bn−1 Bn + Bn B1 ,

a contradiction.
Conversely, let

∠Bn B1 A1 = B2 B1 A2 = β1 , ∠B1 B2 A2 = B3 B2 A3 = β2 , . . . ,

∠Bn−1 Bn An = ∠B1 Bn A1 = βn .
Consider an arbitrary inscribed n-gon C1 C2 . . . Cn in A. Draw through vertices
A1 , A2 , . . . , An lines l1 , l2 , . . . , ln parallel to sides Bn B1 , B1 B2 , . . . , Bn−1 Bn , re-
spectively. Let C10 and C100 be the orthogonal projections of C1 on lines l1 and
let l2 , C20 and C200 be the orthogonal projections of C2 on lines l2 and l3 , etc.,
and let Cn0 and Cn00 be the orthogonal projections of Cn on lines ln and l1 . We
have

C1 C2 + C2 C3 + · · · + Cn−1 Cn + Cn C1 ≥ C100 C20 + C200 C20 + · · · + Cn−1

00
Cn0 + Cn00 C10

= (A2 C1 cos β1 + A2 C2 cos β2 ) + · · · + (C1 A1 cos β1 + Cn A1 cos βn )

= A1 A2 cos β1 + A2 A3 cos β2 + · · · + An A1 cos βn
= (A2 B1 cos β1 + B1 A1 cos β1 ) + · · · + (A1 Bn cos βn + An Bn cos βn )
= B1 B2 + B2 B3 + · · · + Bn−1 Bn + Bn B1 .
Hence

C1 C2 + C2 C3 + · · · + Cn−1 Cn + Cn C1 ≥ B1 B2 + B2 B3 + · · · + Bn−1 Bn + Bn B1 .

The equality is attained if and only if

C1 C2 k B1 B2 , C2 C3 k B2 B3 , . . . , Cn−1 Cn k Bn−1 Bn , Cn C1 k Bn B1 .

For odd n, the equality holds if C1 = B1 , C2 = B2 , . . . , Cn = Bn . 

The next problem has been first raised by Fermat in a private letter to
Torricelli, who solved it.
14 Chapter 1

Example 1.10. (Fermat’s problem) Given points A, B, C in the plane, find

all points X such that the sum of distances from X to A, B, C is minimized.

First Solution. For every point X in the plane, we set

t(X) = XA + XB + XC.

It is easy to see that if X is outside triangle ABC, then there is a point X 0

such that t(X 0 ) < t(X). Indeed, in this case, one of the lines AB, BC, CA,
say AB, is such that triangle ABC and X lie on different sides of this line
(Fig. 1.13).
C

X0

A B Y
X

Figure 1.13

Consider the reflection X 0 of X in AB. We have AX 0 = AX, BX 0 = BX.

Also, segment CX intersects line AB at some point Y , and XY = X 0 Y . Now
the triangle inequality gives

CX 0 < CY + X 0 Y = CY + XY = CX,

implying t(X 0 ) < t(X).

So we may restrict our attention to points X in the interior or on the boundary
of triangle ABC. Without loss of generality, we assume that ∠C ≥ ∠A ≥
∠B. Then ∠A and ∠B are both acute angles. Denote by ϕ the 60◦ rotation
counterclockwise about A. For a point M in the plane, let M 0 = ϕ(M ). Then
triangle AM M 0 is equilateral. In particular, triangle ACC 0 is equilateral.
The Triangle Inequality 15

Consider an arbitrary point X in triangle ABC. Then AX = XX 0 , while

ϕ(X) = X 0 and ϕ(C) = C 0 imply CX = C 0 X 0 . Consequently,

t(X) = BX + XX 0 + X 0 C 0 ,

that is, t(X) equals the length of the broken line BXX 0 C 0 .
We now consider three cases.
Case 1. ∠C < 120◦ . Then ∠BCC 0 = ∠C + 60◦ < 180◦ . Since ∠A < 90◦ , we
have ∠BAC 0 < 180◦ , so the segment BC 0 intersects side AC at some point D
(Fig. 1.14).
C0

X00

C
X0 D

X0
X

A B

Figure 1.14

Denote by X0 the intersection of BC 0 with the circumcircle of triangle ACC 0 .

Then X0 lies in the interior of the line segment BD and X00 lies on C 0 X0 since
∠AX0 C 0 = ∠ACC 0 = 60◦ . Moreover,

t(X0 ) = BX0 + X0 X00 + X00 C 0 = BC 0 ,

so t(X0 ) ≤ t(X) for every point X in triangle ABC. Equality occurs only if
both X and X 0 lie on BC 0 , which is possible only when X = X0 . Notice that
the point X0 constructed above satisfies

∠AX0 C = ∠AX0 B = ∠BX0 C = 120◦ .

16 Chapter 1

It is called Fermat’s first point or the Fermat-Torricelli point of triangle ABC.

Case 2. ∠C = 120◦ . In this case, the line segment BC 0 contains C and

t(X) = BX + XX 0 + X 0 C 0 = BC 0

precisely when X = C.
Remark. Cases 1 and 2 also follow by Pompeiu’s theorem (Example 1.5).
Indeed, triangle ACC 0 is equilateral and

t(X) = AX + BX + CX ≥ C 0 X + BX ≥ C 0 B.

Case 3. ∠C > 120◦ . Then BC 0 has no common points with the side AC
(Fig. 1.15).
C0

X0 C

A B

Figure 1.15

If AX ≥ AC, then the triangle inequality gives

t(X) = AX + BX + CX ≥ AC + BC.

If AX < AC, then X 0 lies in triangle ACC 0 and

t(X) = BX + XX 0 + X 0 C 0 ≥ BC + CC 0 = BC + AC

since C lies in the quadrilateral BC 0 X 0 X (Fig. 1.11). In both cases, equality

occurs precisely when X = C.
The Triangle Inequality 17

In conclusion, if all angles of triangle ABC are less than 120◦ , then t(X)
is minimal when X coincides with Fermat-Torricelli point of triangle ABC. If
one of the angles of triangle ABC is not less than 120◦ , then t(X) is minimized
when X coincides with the vertex of that angle. 
Second Solution. The following elegant solution of Fermat’s problem is
attributed to Torricelli. It is based on the well-known fact that the sum of
distances from an interior point of an equilateral triangle to its sides is equal
to the altitude of the triangle.
Assume that all angles of triangle ABC are less than 120◦ and denote by P
its Fermat-Torricelli point. The perpendiculars to AP, BP , CP through A, B,
C, respectively, determine an equilateral triangle DEF (Fig. 1.16) since, for
example, ∠F DE = 180◦ − ∠AP B = 60◦ .
F C E

A B

Figure 1.16

Denote by h the length of the altitude of this triangle. Then we know that

P A + P B + P C = h.

Let M be an arbitrary point on the boundary or in the interior of triangle

ABC. Then the sum of distances from M to the sides of triangle DEF is equal