Hayk Sedrakyan
Nairi Sedrakyan
Geometric
Inequalities
Methods of Proving
Problem Books in Mathematics
Series Editor:
Peter Winkler
Department of Mathematics
Dartmouth College
Hanover, NH 03755
USA
Geometric Inequalities
Methods of Proving
B
r1
r3
r2
A C
r1 + r 2 > r 3
Hayk Sedrakyan Nairi Sedrakyan
University Pierre and Marie Curie Yerevan, Armenia
Paris, France
To Ani,
a wonderful daughter and a loving sister
Preface
vii
viii Preface
The authors have tried to find common approaches to different problems. The
goal of the book is to teach the reader new and classical methods for proving
geometric inequalities.
The authors would like to express their gratitude to their family for the support.
Hayk Sedrakyan
Nairi Sedrakyan
Contents
ix
x Contents
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449
Chapter 1
Theorem on the Length of the Broken Line
This chapter consists of two sections. Section 1.1 is devoted to the applications of
one of the most important geometric inequalities, called the triangle inequality.
As a generalization of triangle inequality, Section 1.2 is devoted to the theorem
on the length of the broken line.
Let us recall the triangle inequality: for any triangle, the sum of the lengths of
any two sides is greater than or equal to the length of the remaining side. In other
words, if A, B, C are arbitrary points, then AB AC þ BC. Moreover, the equality
holds true, if and only if point C is on segment AB.
The goal of Section 1.1 is to get the reader acquainted with the triangle
inequality.
This section consists of problems that can be proved by using, if necessary
several times, the triangle inequality.
In Section 1.1 selected problems are those that can be proved using the following
techniques and statements:
1. For any convex quadrilateral ABCD it holds true AC þ BD > AB þ CD.
2. For any triangle with side lengths a, b, c and a median ma drawn to the side with
length a it holds true ma < bþc
2 .
3. If M is an arbitrary point inside of triangle ABC, then MA þ MB < CA þ CB.
4. For any points A, B, C, D it holds true AC BD AB CD þ BC AD.
In some problems, the triangle inequality is not always applied directly. In some
cases, at first one needs to do some geometric constructions or translations and only
after that apply the triangle inequality.
In Section 1.2, selected problems are those that can be proved using the theorem
on the length of the broken line, that is: if A1, A2, . . . , An are arbitrary n points, then
A1An A1A2 þ A2A3 þ . . . þ An 1An. Moreover, the equality holds true if and
only if point Ai is on the segment Ai 1Ai þ 1, for i ¼ 2, 3, . . . , n 1.
Some problems in this chapter were inspired by [7]. Nevertheless, even for these
problems the authors have mostly provided their own solutions.
the medians drawn from vertices A, B, C and a, b, c are, respectively, the lengths
of sides BC, CA, AB of triangle ABC.
1.1.8. Let M be a point inside of triangle ABC. Prove that
(a) MA þ MB < CA þ CB,
(b) min(MA, MB, MC) þ MA þ MB þ MC < ma þ mb þ mc,
(c) MA þ MB þ MC max (AB þ BC, AC þ BC, AC þ AB),
pffiffi
(d) MA þ MB þ MC 23 minðAB þ BC; AC þ BC; AC þ ABÞ.
1.1 Triangle Inequality 3
1.1.9. (a) Let ABCD be a quadrilateral and E,F be the midpoints of sides AB, CD,
respectively. Prove that EF BCþAD
2 .
(b) Let M be the intersection point of medians AD and BE of triangle ABC. Prove
that, if ∠AMB π/2, then AC þ BC > 3AB.
(c) Let C1, A1 be points (different from the vertices) on sides AB, BC of triangle
ABC. Let K be the midpoint of A1C1 and I be the in center of triangle ABC.
Given that A1BC1I is an inscribed quadrilateral. Prove that AKC is an obtuse
angle.
1.1.10. Let ABCD be a quadrilateral such that angles A and C are equal to 90 . Prove
that the perimeter of the inscribed quadrilateral in the quadrilateral ABCD is not
smaller than 2AC.
1.1.11. Let M be a point inside of an equilateral triangle ABC. Prove that
(a) MA þ MB > MC ,
(b) MA2 þ MB2 þ MC2 < 2 AB2.
1.1.12. Let E be a point on side AC of triangle ABC. Prove that
(a) BE AC AE BC þ CE AB,
(b) (EB BA) AC (BC AB) AE.
1.1.13. Let D be a point on side BC of triangle ABC such that AD > BC. Let point
E on side AC be defined by the following condition EC AE
¼ ADBC
BD
. Prove that
AD > BE.
1.1.14. (a) Prove that for any distinct points A, B, C and D it holds true AC BD
AB CD þ BC AD.
(b) Let a square with the center O be externally constructed on the side AB of
triangle ABC. Let M, N be the midpoints of sides BC, AC and the lengths of
these sides be equal to a, b, respectively. Find the possible greatest value of the
sum OM þ ON when angle ∠ACB changes.
OAþOC
(c) Given a rectangle ABCD on a plane. Find the smallest value of OBþOD , where
O is an arbitrary point in that plane.
(d) Prove that for any points A, B, C and D it holds true
AB þ BC þ AC 2AD sin ∠ BDC.
(e) Let D, E, F be points on sides BC, CA, AB of triangle ABC, respectively. Prove
that
1 1 1 AB þ BC þ AC
þ þ ðDE þ EF þ FDÞ ,
AD BE CF R
exist numbers i and j (i 6¼ j), such that 1 aaij λ1n . For n ¼ 4, prove that the estimate
qffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffi
¼ þ 18 ¼ 1, 32471795:::: is not possible to make smaller.
1 3 9 69 3 9þ 69
λ4 18
1.1.21. Given n (n 3) distinct points with the pairwise distances between them
equal to a1 , a2 , :::, anðn1Þ , where a1 a2 ::: anðn1Þ : Given that δn > 0 and δn
2 2
nðn1Þ
ð 1 þ δn Þ 2 1
> 2: Prove that there exist numbers i and j(i 6¼ j), such that
ai
aj 1 < δn :
Solutions
1.1.1. (a) Let D be a point on side BC, such that AB ¼ BD. Then, we have that
∠A > ∠ BAD ¼ ∠ BDA ¼ ∠ C þ ∠ DAC > ∠ C. Therefore ∠A > ∠ C.
(b) Proof by contradiction argument. Assume that AB BC. If AB ¼ BC, then
∠C ¼ ∠ A. This leads to a contradiction. If AB > BC, then according to prob-
lem 1.1.1а we have that ∠C > ∠ A. This leads to a contradiction.
(c) Let D be a point on line AC, such that point C belongs to segment AD and
CD ¼ BC. Hence, as ∠ABD ¼ ∠ B þ ∠ CBD > ∠ CBD ¼ ∠ BDC, then for
triangle ABD, according to problem 1.1.1b, we obtain that AB < AD. Thus, it
follows that AB < AC þ BC.
2 , where p ¼
(d) We have that p a ¼ bþca 2 . Therefore, p > a.
aþbþc
Figure 1.1
6 1 Theorem on the Length of the Broken Line
1
B1 B2 þ B2 B3 þ ::: þ Bn B1 ¼ ðA1 A2 þ A2 A3 þ ::: þ An A1 Þ:
2
AB þ DE ¼ AD þ BE: ð1:1Þ
The initial n-gon is convex; therefore its diagonals AD and BE intersect at some
point P. According to the triangle inequality AD þ BE ¼ AP þ BP þ PD þ
PE > AB þ DE. This leads to the contradiction with (1.1).
In order to construct the required octagon, let us circumscribe an isosceles
trapezoid A1A2A3A4 around the circle, such that its base is A1A4 and the base
angle is equal to 45 . Then, construct it up to octagon A1A2 . . . A8, as it is shown
in Figure 1.3.
way, one can construct n-gon, such that it is possible to cut off by its
In a similar
diagonals n2 circumscribed quadrilaterals.
1.1.6. (a) Let M be the intersection point of diagonals AC and BD of a quadrilateral
ABCD. Note that ΔABM ~ ΔDCM. Therefore,
Figure 1.2
1.1 Triangle Inequality 7
Figure 1.3
AC BD ¼ AM þ MC BM DM ¼ AM þ BM CD BM AM CD ¼
AB AB
AM BM
¼ AB CD AB CD
AB
(see problem 1.1.2). In a similar way, we obtain that |AC BD| |AD BC|, thus
|AB CD| þ |AD BC| 2|AC BD|.
In the last inequality the equality holds true if and only if quadrilateral ABCD is a
rectangle.
(b) Note that ∠MAD ∠ MDA. Therefore, MD MA. As CD AB ¼ MB ¼ MA ¼ k 1
CM DM
(b) We have that ma < bþc 2 , mb < 2 and mc < 2 , thus ma þ mb þ mc < a þ b þ c.
aþc aþb
mc > 34 ða þ b þ cÞ.
8 1 Theorem on the Length of the Broken Line
Figure 1.4
Figure 1.5
Figure 1.6
1.1.8. (a) Let N be the intersection point of lines AM and BC; then AM þ BM <
AM þ MN þ BN ¼ AN þ NB < AC þ CN þ BN ¼ AC þ BC.
Therefore, AM þ BM < AC þ BC.
(b) Let M be a point inside of triangle AGB1 (Figure 1.5), where G is the intersection
point of the medians of triangle ABC.
Using the result of problem 1.1.8а, we obtain that AM þ BM b2 þ mb and
AM þ MC 23 ma þ 23 mc . Therefore,
Case 2 If there does not exist such point M0 (see Case 1), then one can easily prove
that ∠C 120 .
Note that either ∠MCB < 120 or ∠MCA < 120 . Without loss of generality one
can assume that ∠MCB < 120 ; then according to problem 1.1.8а, MA þ MA1
AC þ CA1. Therefore, MA þ MB þ MC MA þ MA1 AC þ BC.
This ends the proof.
1.1.9. (a) Let us denote by K the midpoint of diagonal BD. We have that
EF EK þ KF ¼ AD 2 þ 2 , and the equality holds true if and only if K belongs to
BC
Figure 1.7
Figure 1.8
(c) Let O be the midpoint of segment AC. Note that AC ¼ AC1 þ CA1. According to
problem 1.1.9а, we have that OК AC1 þCA
2
1
¼ AC
2 . Therefore, AKC is an obtuse
angle.
1.1.10. Let a quadrilateral MNPK be inscribed in a quadrilateral ABCD (Figure 1.7).
Let E and F be the midpoints of segments KM and NP, respectively. Since
AE ¼ KM 2 , EF
MNþKP
2 (see problem 1.1.9а) and CF ¼ NP 2 , then
MNþNPþKPþKM
2
AE þ EF þ CF AF þ CF AC. Thus, it follows that MN þ NP þ KP þ
KM 2AC.
Remark If ∠A 90 , ∠ C 90 , then MN þ NP þ KP þ KM 2AC.
1.1.11. (a) Let us consider the Figure 1.8.
Consider segments MA1, MB1, MC1, such that MA1||AB, MB1||BC and MC1||AC.
One can easily prove that AC1MB1, BA1MC1 and CA1MB1 are isosceles trapezoids.
Therefore, MA ¼ C1B1, MB ¼ A1C1 and MC ¼ A1B1. Hence, MA þ MB > MC .
(b) Let MA1 ¼ x, MB1 ¼ y, MC1 ¼ z, AB ¼ a (Figure 1.8). Therefore, x þ y þ z ¼ a
and
1.1 Triangle Inequality 11
Figure 1.9
AE AD ¼ BD EC þ AE BC: ð1:2Þ
pffiffi
that МО þ NO 22þ1 ða þ bÞ. The equality holds true if only ∠ANO ¼
∠ CMO ¼ 90 , that is ∠C ¼ ∠ A þ ∠ B þ 90 . Therefore ∠C ¼ 135 .
(c) According to problem 1.1.14a, for points A, B, C, O we have that
AC BO AB CO þ BC AO . In a similar way, for points A, D, C, O we
have that AC DO AD CO þ AO CD. Summing up these inequalities, we
AOþCO
obtain that BOþDO ABþADAC
. Taking O A, we obtain that the left-hand side of
AC
the last inequality is equal to ABþAD . Thus, the possible smallest value is equal
AC
to ABþAD.
OAþOC
Remark One can prove that OBþOD ABþBC
BD .
AB OC þ AC OB AO BC:
Hence,
BC
AB þ AC þ BC AD ¼ 2AD sin ∠BDC:
OD
!
!
(g) Consider a point M0 , such that MM0 ¼ AD . For points M, C, M0 , D, according to
problem 1.1.14a, we have that MC M0 D þ CM0 MD CD MM0 .
As M0 D ¼ AM, CM0 ¼ BM, MM0 ¼ AD, then MA MC þ MB MD AB AD.
!
!
(h) Consider a point A0 , such that BA0 ¼ AC .
1.1 Triangle Inequality 13
DA DB AB þ DB DC BC þ DC DA AC AB BC AC:
(i) For points A, A1, B1, C1 according to problem 1.1.14a, we have that
and
1 2 2 2 1 2 AB2 þ AD2 BD2
AM ¼ AA1 þ AO A1 O ¼ AA1 þ
2 2 2 2
32 3 9
3 3 2 4
2 A1 B þ A1 D2 BD2 1 1
¼ AA1 2 þ AB2 þ AD2 A1 B2 þ A1 D2 þ BD2 :
9 2 4 3 9
1 1
AM1 2 ¼ AB1 2 þ AD1 2 þ AC2 B1 D1 2 þ D1 C2 þ CB1 2 :
3 9
Therefore
1
AM1 2 AM2 ¼ AB1 2 þ AD1 2 þ AC2 AB2 AD2 AA1 2
3
from this equality and inequalities (1.6), (1.7), (1.8) we obtain that
AB BC AB2 BC2
þ ¼ þ
SA þ SB SB þ SC SA AB þ SB AB SB BC þ SC BC
ðAB þ BCÞ2
:
SA AB þ SC BC þ SBðAB þ BCÞ
Case 2 Assume that there does not exist such a point M0. Let max
(∠A, ∠B, ∠C) ¼ ∠ C. We have that (see the proof of problem 1.1.8d)
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
SA þ SB þ SC > AC þ BC > 2AC2 þ 2BC2 AB2 ¼ 3 minðMA; MB; MCÞ:
Figure 1.10
16 1 Theorem on the Length of the Broken Line
0
If T 2 ΔAMN, then m(T ) AT AA max (AM, AN) ¼ AN ¼ m(N ), as max
0 0
(∠MA A, ∠NA A) 90 .
If T 2 ΔANC, then mðT Þ ATþTC
2 ANþCN
2 ¼ AN ¼ mðN Þ.
Thus, the greatest value of m(T )is equal to m(N ). From the proof, it follows that,
if ∠B ¼ ∠ C, then points M and N are the required points. Otherwise, point N is the
required point.
Remark In the case, if ∠A > 90 , mðT Þ CN < BC
2.
1.1.17. (a) Using the triangle inequality, we obtain that BG þ 12 CG > 12 AB and
CG þ 12 BG > 12 AC. Summing up these inequalities, we deduce that 3(BG þ CG) >
AB þ AC.
2
CG2 AB2 AC2 AB2 AC2
We have that BG CG ¼ BG BGþCG ¼ 3ðBGþCGÞ < ABþAC . Hence, AC þ BG <
AB þ CG.
(b) Let A1 be the midpoint of side BC. According to problem 1.1.7a, we have that
OA1 OBþOC
2 . Point A2 is on the segment OA1 and A1A2 : A2O ¼ 1 : 2.
Then OG A2 G þ A2 O ¼ AO
3 þ 3 OA1 3 þ 3 þ 3 . It is clear that the equal-
2 AO BO CO
ities OG ¼ A2G þ A2O and OA1 ¼ 2 do not hold true simultaneously. Thus, it
OBþOC
A1 A 2 þ A3 A 4 þ A5 A6 A2 A3 A4 A5 A6 A1
< A1 A2 þ A3 A4 þ A05 A6 A2 A3 A4 A05 A6 A1 : ð1:9Þ
0 0
Let A01 be a point on the segment OA1, such that OA 1 ¼ OA3 ¼ OA 5 and let A02 be
0
a point on the segment OA2, such that OA 2 ¼ OA6. Then, we have that
A1 A2 þ A01 A02 < A01 A2 þ A02 A1 , thus A1 A2 þ A01 A6 < A01 A2 þ A1 A6 . Hence,
From (1.9) and (1.10) it follows that A1A2 þ A3A4 þ A5A6< A2A3 þ A4A5 þ A6A1.
1.1 Triangle Inequality 17
1.1.19. Proof by contradiction argument. Let M and N be two points among those n,
such that the distance MN ¼ d1 is the greatest. Consider plane Π, passing through
the midpoint of segment MN and perpendicular to it.
One of the half-spaces with boundary Π contains k points from the given
n points, where k n2. Let us denote these points by N, N1, . . . , Nk 1 and
MN ¼ d1 MN1 ¼ d2 . . . . MNk 1 ¼ dk. According to our assumption, we
have that
2 2 2
d1 1 þ d2 , d2 1 þ d3 , :::, dk1 1 þ d k :
n n n
It is known that the estimate λ10 is not possible to make smaller for 3 n 7.
1.1.20. The proof by contradiction method. Consider given points A1, A2, . . . , An,
such that AiAj A1An ¼ d1, i, j 2 {1, ..., n} and max ðA1 Ai ; Ai An Þ ¼ di .
i2f2;:::;n1g
Without loss of generality one can assume that d2 d3 . . . dn 1. Then
according to our assumption, we obtain that dd12 > λ1n , dd23 > λ1n , :::, ddn2
n1
> λ1n .
Multiplying these inequalities, we deduce that
n d 1 > d n1 :
λn2 ð1:14Þ
hence
nðn1Þ
anðn1Þ ð1 þ δn Þ 2 1 a1 :
2
1.1 Triangle Inequality 19
minðAC;ADÞ
therefore max ðAC;ADÞ 1 < δn . This leads to a contradiction.
1.1.22. Let R and r be the circumradiuses of triangles ABC and MAN, respectively.
∠MAN
sin ∠BAC ¼ 2r BC.
MN 2R
According to the law of sines, we have that sin
2 2
pffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffi 2 < r < pffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffi 2 :
BC R BC
Hence, it is enough to prove that
ð BMCNþ BNCMÞ ð BMCN BNCMÞ
Let the circumcircle of triangle MAN intersect sides AB and AC of triangle ABC
at points K and E, respectively (Figure 1.11).
We have that ∠KEM ¼ ∠ BAM ¼ ∠ CAN ¼ ∠ EMC. Therefore, KE k BC.
Thus, it follows that ΔKAE ΔBAC.
pffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffi
Hence, KE ¼ Rr BC, AK ¼ pBMBN ffiffiffiffiffiffiffiffiffiffiffiffi , AE ¼ pffiffiffiffiffiffiffiffiffiffiffiffi
CMCN
Using the triangle inequal-
r ð r 1Þ r ð r 1Þ:
R R R R
pffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffi
j BMBN CMCN j
ity, we obtain that |AK AE| < KE < AK þ AE. Therefore, <
ffi pffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffirffi pffiffiffiffiffiffiffiffiffiffi BC
BMBN þ CMCN BC2 BC2
1R< . Thus, pffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffi 2 < r < pffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffi 2 :
R
BC
ð BMCNþ BNCMÞ ð BMCN BNCMÞ
Figure 1.11
20 1 Theorem on the Length of the Broken Line
1.1.23. There are 50 correct clocks on the table. Prove that at some moment the sum
of the distances from the center of the table to the end of the minute hands will be
greater than the sum of the distances from the center of the table to the center of the
clocks.
1.1.24. Let M be a point inside of the parallelogram ABCD. Prove that
1.1.25. Prove that the sum of the lengths of the diagonals of a convex pentagon is
greater than its perimeter and is smaller than twice the perimeter.
1.1.26. Prove that for any convex pentagon there are three diagonals that are sides
of a triangle.
1.1.27. Prove that in a convex polygon there are no three sides that are greater than
the largest diagonal of that polygon.
1.1.28. Prove that the arithmetic mean of the lengths of the sides of an arbitrary
convex polygon is less than the arithmetic mean of the lengths of its diagonals.
1.1.29. Prove that h1a < h1b þ h1c , where ha, hb, hc are the altitudes of some triangle.
ABAC ¼ 4 BMþCN .
Hint Prove that BMCN 3 ABþAC
1.1.33. Given that in triangle ABC side AC is the largest side. Prove that for any
point M of plane AM þ CM is not less than BM. When does the equality hold true?
1.1.34. Given that points A1, A2, . . . , An are not on the same line. Let P and Q be
distinct points, such that A1P þ A2P þ . . . þ AnP ¼ A1Q þ A2Q þ . . . þ AnQ ¼ S.
Prove that A1K þ A2K þ . . . þ AnK < S for some point K.
1.1.35. Given a tetrahedron ABCD. Prove that there exists a triangle with sides
AB CD, AC BD and AD CB.
1.1.36. Let a, b, c be the side lengths of some triangle. Prove that
a2 þ2bc
þ bc2þ2ca c2 þ2ab
2
Hint
(a) See problem 1.1.14а.
(b) Note that SD1 AA1 SBA1 B1 SCB1 C1 SDD1 C1 ¼ ðr 1 2 r 2 Þsin 2 ∠A sin 2 ∠B and
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
SABCD ACBD
2 2r 2 sin ∠A sin ∠B 2r 2 sin ∠A sin ∠B.
1.1.44. Given a quadrilateral ABCD, such that ∠A ∠ D and ∠B þ ∠ D 180 .
Prove that AC þ CD AB þ BD.
Hint Let C1 be the intersection point of line AC and a circle passing through points
A, B, D.
Prove that C1 is on segment AC, and see problem 1.1.6b.
1.1.45. Prove that if a1, a2, . . . , an (n 3) are positive numbers, such that
2 2
a1 þ a22 þ ::: þ a2n > ðn 1Þ a41 þ a42 þ ::: þ a4n , then one can construct a tri-
angle with sides ai, aj, ak for any 1 i < j < k n.
2
Hint If m > 3 and a2i1 þ ::: þ a2im > ðm 1Þ a4i1 þ ::: þ a4im , then ðm 2Þ
2
a4im 2a2im a2i1 þ ::: þ a2im1 þ ðm 1Þ a4i1 þ ::: þ a4im1 a2i1 þ ::: þ a2im1
< 0:
Therefore, D > 0 or
2
a2i1 þ ::: þ a2im1 > ðm 1Þðm 2Þ a4i1 þ ::: þ a4im1
2 2 :
ðm 2Þ a2i1 þ ::: þ a2im1 ; a2i1 þ ::: þ a2im1 > ðm 2Þ a4i1 þ ::: þ a4im1
a
Hint We have that bþca ¼ 12 cþab
bþca þ aþbc
bþca .
1.1.51. Let O be the circumcenter of triangle ABC. On sides AB and BC are given
points M and N, respectively, such that 2 ∠ MON ¼ ∠ AOC. Prove that
NB þ BM þ MN AC.
Hint See problem 1.1.14d.
1.1.52. The sum of the distances from point M to the two neighboring vertices of the
square is equal to a. What is the largest value of the sum of the distances from point
M to the other two vertices of the square?
Hint See problem 1.1.14а.
1.1.53. Given that the perimeter of a convex quadrilateral is equal to 2004 and one
of its diagonals is equal to 1001. Can the second diagonal be equal to 1? Can the
second diagonal be equal to 2? Can the second diagonal be equal to 1001?
1.1.54. Let ABC be a triangle. Prove that
(a) maa2 þ mbb2 þ mcc2 mcab þ mabc þ mbac.
Hint According to problem 1.1.14а, we have that a2 23 ma 2c m3b þ b2 m3c .
(b) ma mb þ mb mc þ mc ma 2p2 34 ðab þ bc þ acÞ.
Hint According to problem 1.1.14а, we have that ma mb c 2c þ a2 b2.
(c) ma þ mb þ mc þ min (a, b, c) la þ lb þ lc þ max (a, b, c).
Hint We have that ma la a2 ðp bÞ ¼ bc
2 :
1.1.55. Let A, X, D be points on a line, such that X is in between A and D. Let point
B be such that ∠ABX 120 and point C be in between B and X. Prove that
pffiffiffi
2AD 3ðAB þ BC þ CDÞ.
Hint We have that AX2 AB2 þ BX2 þ AB BX 34 ðAB þ BXÞ2 .
1.1.56. Let K and L be points on side AB of triangle ABC, such that AK ¼ KL ¼ LB.
Prove that
24 1 Theorem on the Length of the Broken Line
If A1, A2, . . . , An are distinct points, then A1An A1A2 þ A2A3 þ . . . þ An 1An.
Note that the equality holds true only if simultaneously hold true the following
conditions: point Ai is on the segment Ai 1Ai þ 1 for all i ¼ 2, 3, . . . , n 1.
1.2 Theorem on the Length of the Broken Line 25
1.2.1. Given two circles with radiuses R1 and R2, such that for the distance between
their centers it holds true the following inequality d > R1 þ R2. Prove that
d R1 R2 XY d þ R1 þ R2, where X and Y are arbitrary points of these
two circles.
1.2.2. Prove that in any polygon there are at least two sides a and b, such that
1 ba < 2.
1.2.3. Given a convex hexagon ABCDEF, such that ∠A 90 , ∠ D 90 . Prove
that BC þ CE þ EF þ FB 2AD.
1.2.4. Given the points A(a, 0), B(0, b), C(c, d) and O(0, 0) on a coordinate plane.
Prove that AB þ BC þ CA 2CO.
1.2.5. Prove the following inequalities:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
(a) a21 þ b21 þ ::: þ a2n þ b2n ða1 þ ::: þ an Þ2 þ ðb1 þ ::: þ bn Þ2 .
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
a21 þ ka1 b1 þ b21 þ ::: þ a2n þ kan bn þ b2n
(b) qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ða1 þ ::: þ an Þ2 þ kða1 þ ::: þ an Þðb1 þ ::: þ bn Þ þ ðb1 þ ::: þ bn Þ2 ,
where |k| 2.
(c) ða þ b þ cÞða b þ cÞ þ ða b þ cÞða þ b cÞ þ ða þ b cÞða þ b þ cÞ
pffiffiffiffiffiffiffi pffiffiffi pffiffiffi pffiffiffi
abcð a þ b þ cÞ, where a, b, c > 0:
1.2.6. On the legs of a right-angled triangle with hypotenuse c and acute angle α are
chosen points P and Q. Let PK and QH be perpendiculars to the hypotenuse drawn
from points P and Q. Prove that KP þ PQ þ QH c sin 2α.
1.2.7. Prove that from all triangles inscribed in an acute triangle ABC, the smallest
perimeter has the orthic triangle (the triangle whose vertices are the endpoints of the
altitudes of triangle ABC).
1.2.8. Given a triangle ABC. Prove that ∠ACB 120 is a necessary and sufficient
condition for the following inequality to hold true MA þ MB þ MC AC þ BC, for
any point M of plane ABC.
1.2.9. (a) Prove that MAA11MA
A2
2
þ ::: þ MAAn1 MAn MA1 MAn , where M, A1, . . . , An are
n1 An A1 An
1.2.17. Inside of the convex polygon with perimeter P are given two rectangles with
the perimeter with sides a, b and c, d which do not have any common interior point.
Prove that min(a, b) þ a þ b þ min (c, d ) þ c þ d < P.
1.2.18. (а) Let points A and B be outside of circle ω. Through points A and B are
drawn tangents AM and BN to circle ω. Prove that if segment AB intersects the
circle ω, then AB > AM þ BN, and if it does not intersect, then AB AM þ BN.
(b) Let points A, B, C be outside of circle ω, such that ω intersects segments AC, BC
and does not intersect segment AB. Through points A, B, C are drawn tangents
AM, BN, CK to circle ω. Prove that AM BC þ BN AC > CK AB.
Solutions
1.2.1. Let point X belong to a circle with center O1 and radius R1.
Let point Y belong to a circle with center O2 and radius R2.
For the broken line O1XYO2 we have that R1 þ XY þ R2 d.
Thus, it follows that XY d R1 R2.
1.2 Theorem on the Length of the Broken Line 27
NC ¼ ðac4Þ þd ¼ AC
2
2 , then AB þ BC þ AC 2OC.
See problem 1.1.14d.
1.2.5. (a) The proof follows straightforwardly from 1.2.5b for k ¼ 0.
qffiffiffiffiffiffiffiffiffiffiffiffi
2
(b) Let us consider the points Ai 1 k4 ða1 þ ::: þ ai Þ; b1 þ ::: þ bi þ 2kða1 þ :::
þai ÞÞ, i ¼ 1, . . . , n.
Note that
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u0sffiffiffiffiffiffiffiffiffiffiffiffiffi 12
u 2
u k 2
k
Ai Aiþ1 ¼ t@ 1 aiþ1 A þ biþ1 þ aiþ1
4 2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ a2iþ1 þ kaiþ1 biþ1 þ b2iþ1 , i ¼ 0, :::, n 1,
2 ,b ¼ 2 ,c ¼ 2 .
(c) Let a þ b þ c ¼ m, a b þ c ¼ n, a þ b c ¼ k, then a ¼ nþk mþk mþn
Therefore, if mn þ nk þ mk 0, then
28 1 Theorem on the Length of the Broken Line
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi !
pffiffiffiffiffiffiffi pffiffiffi pffiffiffi pffiffiffi 1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2
abcð a þ b þ cÞ ¼ ððn þ kÞmÞ2 þ ðn þ kÞ mn þ nk þ mk þ
4
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2ffi
þ ððm þ kÞnÞ2 þ ðm þ kÞ mn þ nk þ mk
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2
þ ððn þ mÞkÞ2 þ ðn þ mÞ mn þ nk þ mk
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
4ðmn þ nk þ mkÞ2 þ 4ðm þ n þ kÞ2 ðmn þ nk þ mkÞ
4
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
4ðmn þ nk þ mkÞ2 þ 12ðmn þ nk þ mkÞ2
4
¼ mn þ nk þ mk,
Figure 1.12
1.2 Theorem on the Length of the Broken Line 29
Let AA1, BB1, CC1 be the altitudes of triangle ABC. As BP BB1, then from
(1.16) we obtain that
Let the altitudes AA1, BB1, CC1 intersect at point H. Note that
2BB1 sin ∠B ¼ BB1 sin ∠B þ BB1 sin ∠B ¼ BB1 sin ∠B þ CC1 sin ∠C ¼¼ BH sin ∠B
þ CH sin ∠C þ B1 H sin ∠B þ C1 H sin ∠C ¼ A1 C1
þ A1 B1 þ þB1 H cos ∠C1 B1 H þ C1 H cos ∠B1 C1 H
¼ A1 C1 þ A1 B1 þ C1 B1 :
ð1:18Þ
MN þ NP þ MP > A1 C1 þ A1 B1 þ C1 B1 :
Figure 1.13
AM þ CM þ BM ¼ AM þ CM þ B1 M1 AM þ MM1 þ B1 M1 AB1
¼ AC þ CB:
AM þ BM þ CM
AM0 þ BM0 þ CM0 :
CB2
Indeed, according to problem 1.2.9a, we have that MCMB2
þ MBMB
B2 B
2
MCMB
BC
or
MB þ MC MB2. Therefore MA þ MB þ MC MA þ MB2 AB2 ¼ AM0 þ M0B2 ¼
AM0 þ BM0 þ CM0.
1.2.9. (а) Let us choose a point Bi on ray MAi, such that MBi ¼ MA
1
i
, i ¼ 1, . . . , n.
AA
Note that Bi Bj ¼ MAiiMA
j
j
.
Indeed, if rays MAi and MAj are not on the same line, we obtain that triangle
MAiAj is similar to triangle MBiBj.
BB
Hence, it follows that Aii Ajj ¼ MB
MAi ¼ MAi MAj .
i 1
AA
Therefore, Bi Bj ¼ MAiiMA
j
j
.
If rays MAi and MAj are on the same line, then
1 1 MA MA Ai Aj
Bi Bj ¼ MBi MBj ¼ ¼ i j
¼ :
MAi MAj MAi MA MAi MAj
1.2 Theorem on the Length of the Broken Line 31
Therefore, points M, A1, A2, A3 are on the same circle. In a similar way, one
obtains that points M, A2, A3, A4, ...,M, An 2, An 1, An are on the same circle. Then,
points M, A1, A2, . . . , An are on the same circle (in the given order).
1
(b) 1. We have that MAMB ¼ MBMC
1
þ MCMD
1
þ MDMA
1
ρðM;CDÞ
¼ ρðM;AB
ρðM;ABÞ
AB
Þ þ 1 is the smallest if ρ(M, AB) is the largest, i.e. M is the
midpoint of arc AB.
pffiffiffi
MAMB ctg 8 ¼ 3 þ 2 2.
Therefore, MCMD 2π
AM þ MB þ MN þ CN þ DN ME þ MN þ NF EF:
2 þ 2 þ 2 A1 C1 :
According to problem 1.2.12а, we have that AB BC CD
pffiffi
thus, p 2 3 3 p1 :
This ends the proof.
1.2.13. Points A, B are on the sides of a polygon and divide its perimeter to two equal
parts. Note that for any point C, belonging to any side of the polygon, we have that
OC OM þ MC ACþBC 2 a2, where O is the center of segment AB (Figure 1.14).
Thus, a circle with radius a2 and center O covers this polygon.
This ends the proof.
1.2.14. (a) Given a tetrahedron SA1A2 . . . An with a vertex S. Consider scanning its
lateral surface, making the “cross section” along the edge SA1 (see Figure 1.15).
Given that the sum of the planar angles at vertex S is equal to 180 . Then
∠A1SA10 ¼ 180 and 2SA1 < A1A2 þ A2A3 þ . . . þ An 1An þ AnA10 ¼ A1A2 þ A2A3
þ . . . þ An 1An þ AnA1.
(b) Let A be the intersection point of line A1S with a side of polygon A1A2 . . . A0n A1,
different from A1 (see the solution of problem 1.2.14а). Hence, if A 2 AkAk þ 1,
(An þ 1 A10 ), then A1A2 þ . . . þ AkA > A1S þ SA, SA þ AAk þ 1 þ . . .
þ AnA10 > SA10 .
Therefore, A1A2 þ . . . þ An 1An þ AnA1 > 2SA1.
1.2.15. Consider the layout of the side surface of the cube (Figures 1.16 and 1.17).
pffiffiffi
Note that AA0 ¼ 3 2 and
Figure 1.14
1.2 Theorem on the Length of the Broken Line 33
Figure 1.15
Figure 1.16
Figure 1.17
34 1 Theorem on the Length of the Broken Line
AB þ BC þ CD þ DE þ EF þ FA ¼ AB þ BC þ CD þ DE þ EF þ FA0 AA0
pffiffiffi
¼ 3 2:
AB þ BC þ CDq þffiffiffiffiffiffiffiffiffiffiffiffiffiffi
DE þ ffiEF þ FA ¼
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ a1 þ a2 þ b21 þ b22 þ c21 þ c22 þ d21 þ d 22 þ e21 þ e22 þ f 21 þ f 22
2 2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ða1 þ b1 þ c1 þ d1 þ e1 þ f 1 Þ2 þ ða2 þ b2 þ c2 þ d 2 þ e2 þ f 2 Þ2 ¼ m,
Figure 1.18
1.2 Theorem on the Length of the Broken Line 35
1.2.17. Lemma Let convex polygons B1B2 . . . Bp and C1C2 . . . Cm not have any
common interior points. Then, there exists a line l passing through one of the sides
of the polygons and separating these polygons.
Indeed, let us choose a point O inside of the polygon B1B2 . . . Bp, such that O is
not on lines BiCj. Let k be the smallest positive integer, such that the image of the
polygon B1B2 . . . Bp by homothety with the center Oand ratio k has a common point
with the polygon C1C2 . . . Cm.
Then, the following two cases are possible (Figure 1.19a, b).
In both cases the proof of the lemma is straightforward.
The proof of the problem follows from the lemma and problem 1.2.16.
1.2.18. (a) Let segment AB intersect circle ω, CD is tangent to the circle ω and CD||
AB (Figure 1.20a).
We have that AC þ AB þ BD > CP þ PD ¼
¼ AC þ AM þ BD þ BN. Therefore, AB > AM þ BN.
If segment AB does not intersect circle ω (Figure 1.20b), then
AB AC þ CD þ BD ¼ AC þ CP þ PD þ DB ¼ AC þ CM þ BD þ DN
¼ AM þ BN
Figure 1.19
36 1 Theorem on the Length of the Broken Line
Figure 1.20
Figure 1.21
1.2.19. Given two circles with radiuses R1, R2, such that the distance between their
centers is equal to d and d þ R1 < R2. Prove that R2 R1 d XY d þ R1 þ R2,
where X and Y are arbitrary points of these two circles.
1.2.20. Prove that in any quadrilateral there are at least two sides a and b, such that
1 ba < 1, 875.
1.2.21. Prove that any hexagonal cross section of a unit cube by a plane passing
pffiffi
through its center has an area not less than 3 4 2.
1.2 Theorem on the Length of the Broken Line 37
1.2.22. In a quadrilateral ABCD angles A and C are not less than 90 . Prove that the
perimeter of the inscribed quadrilateral in the quadrilateral ABCD is not less than
2AC (see problem 2.3.6).
1.2.23. (a) In a tetrahedron SABC all planar angles at vertex S are equal to 60 .
Prove that AB þ BC þ AC SA þ SB þ SC.
(b) In a convex hexagon ABCDEF any two of diagonals AD, BE and CF make
an angle of 60 . Prove that AB þ BC þ CD þ DE þ EF þ FA AD þ BE þ CF.
(c) Among all convex quadrilaterals with given diagonals and given angles between
them find the quadrilateral with the smallest perimeter.
Hint
(а) Let SA ¼ a, SB ¼ b, SC ¼ c; then
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
a 2 pffiffi3ffi 2
b
2 pffiffiffi 2
3
AB þ BC þ AC ¼ b þ a þ c þ b
2 2 2 2
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
c 2 pffiffi3ffi 2
þ a þ c ,
2 2
1.2.25. Prove that for any polyhedron there are three edges, from which you can
construct a triangle.
1.2.26. Given an infinite set of points S on a plane, such that in any 1
1 square
there are finitely many number of points of the set S. Prove that there are two
distinct points A and B from the set S, such that for any other point X belonging to
the set S it holds true min(XA; XB) 0, 999AB.
Hint Proceed the proof by contradiction argument.
1.2.27. Given a triangle ABC, such that ∠BAC 60 . Let M be the midpoint of side
BC and P be any point in plane ABC. Prove that PA þ PB þ PC 2AM.
Hint Let the rotation by the angle 60 and center A, point P moves to point P1 and
point C moves to point C1. Prove that PA þ PB þ PC ¼ PB þ PP1 þ P1C1
BC1 2AM.
1.2.28. On the sides of a unit square, as on hypotenuses, are externally constructed
right-angled triangles. Let A, B, C, D be the vertices of the right angles and O1, O2,
38 1 Theorem on the Length of the Broken Line
O3, O4 be the incenters of these triangles, respectively. Prove that the area of
quadrilateral:
(a) ABCD is not greater than 2.
(b) O1O2O3O4 is not greater than 1.
Hint
(a) Let M and N be the midpoints of the opposite sides of the square. Then
AC AM þ MN þ CN ¼ 2.
(b) Prove that if points O1, O2, O3, O4 are on a circle, circumscribed around the
pffiffiffi
square, then O1 O3 2.
where H is the length of the altitude of tetrahedron SABC, drawn from the vertex S.
Hint See problem 1.2.5а.
1.2.30. Let P be the projection of point M on a plane containing points A, B, C.
Prove that if from segments PA, PB, PC one can construct a triangle, then one can
construct a triangle from segments MA, MB, MC too.
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Hint We have that MA þ MB ¼ AP2 þ MP2 þ BP2 þ MP2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðAP þ BPÞ2 þ 4MP2 (see problem 1.2.5a).
This chapter consists of four sections and is devoted to the projection method.
The projection method is one of the fundamental methods applied in order to deal
with geometric inequalities.
In this chapter we learn some techniques on how the projection method can be
applied and to prepare a background for the application of projection method: in
Section 2.1 we consider several problems and some properties of convex polygon
lying inside of another polygon, in Section 2.2 we consider some problems with a
sufficient condition for comparison of lengths of two broken lines on the plane; in
Section 2.3 we deal with the inscribed polygons with the least perimeter and in
Section 2.4 we consider problems that can be proved using these properties and the
projection method.
The main statement that we are going to use in order to apply the projection
method is the following: if A0 B0 is the orthogonal projection of segment AB onto a
line l or onto a plane α, then AB A0 B0 .
In Section 2.1 selected problems are those that deal with geometric inequalities
related to two figures, such that one of them is inside of the other one. In this
section, problems 2.1.1 and 2.1.10 are used as the main techniques of proofs.
In Section 2.2 is provided a sufficiency condition of comparing the sum of the
length of the segments belonging to two sets of segments on the plane. This
beautiful condition is given in problem 2.2.18. Moreover, the following problems
demonstrate the application of this condition.
In Section 2.3 are given the solutions, for the triangle and quadrilateral, of the
following problem: given a convex n-gon A1A2 . . . An. Inscribe to A1A2 . . . An a
polygon B1B2 . . . Bn, such that B1B2 . . . Bn has the least perimeter (in Section 4.1
are provided some generalizations of this problem, see problem 4.1.18).
The last section of this chapter (Section 2.4) is devoted to the application of
projection method in different problems.
Some problems in this chapter were inspired by [4, 8, 9, 12, 13,
15]. Nevertheless, even for these problems the authors have mostly provided their
own solutions.
2.1.1. Prove that, if a convex polygon lays inside of other polygon, then the
perimeter of an internal polygon is less than the perimeter of the external polygon.
2.1.2. Prove that, if the sum of the plane angles at the top of a pyramid is greater than
180 , then each lateral edge of a pyramid is smaller than the half-perimeter of its base.
2.1.3. Consider a convex quadrilateral with sides a , b , c , d laying inside of a unit
1 1 1 1
square. Prove that þ þ þ > 4.
a b c d
2.1.4. Prove that, if a regular hexagon with side length b lays inside of an equilateral
a
triangle with side length a, then b < .
3
2.1.5. Prove that, if a rectangle is covered by a right-angled triangle, so that its
smaller side is on a hypotenuse, then it is possible to cover this rectangle by the
same right-angled triangle, such that its two sides are situated on the legs.
2.1.6. (a) Let S1 and S2 be squares with sides a and b such that they are inside of a
unit square. Given that they have no common points. Prove that a þ b < 1.
(b) Given that regular n-gons with sides b and c are located inside of a regular n-gon
with side a and have no common points. Prove that b þ c < a.
2.1.7. If a regular polygon B1B2 . . . Bn is inscribed into polygon A1A2 . . . An
(B1 2 A1A2, B2 2 A2A3, . . ., Bn 2 AnA1), where ∠A1 ¼ ∠A2 ¼ . . . ¼ ∠An, then
A1A2 . . . An is also a regular polygon and one can place in polygon B1B2A3 . . .
AnA1 a polygon equal to B1B2 . . . Bn, so that its sides are parallel to the sides of
polygon A1A2 . . . An.
2.1.8. Let two non-intersecting triangles with altitudes h1 , h2 , h3 and h01 , h02 , h03 lay
pffiffiffi
inside of a unit square. Prove that 2 minðh1 ; h2 ; h3 Þ þ min h01 ; h02 ; h03 .
2.1.9. (a) Prove that, if a regular 2n -gon with a side length a2 lays inside of a regular
2n-gon with a side length a and center O, then it covers point O.
(b) A regular n-gon with a side length b is inside of a regular n-gon with a side
a
length a and does not contain its circumcenter. Prove that b < π .
2cos 2 2n
2.1.10. Let a convex polygon M is inside of triangle ABC. Prove that polygon M can
be covered by triangle ABC, so that one of the sides of polygon M lays on one of the
sides of the triangle.
2.1.11. Consider two triangles, such that one triangle has sides a, b, c, the other has
sides a0 , b0 , c0 . Which relations between numbers a, b, c and a0 , b0 , c0 are necessary
and sufficient in order the first triangle to be covered by the second one?
2.1.12. Prove that, if triangle A1B1C1 is covered by a triangle ABC, then
(a) A1 B1 þ B1 C1 þ A1 C1 maxðA1 B1 ; B1 C1 ; A1 C1 Þ
AB þ BC þ AC maxðAB; BC; ACÞ,
2.1 Convex Polygon Lying Inside of Another Polygon 41
Solutions
2.1.1. Let us construct on the sides of the internal polygon, outside of it, half-strips,
with parallel edges perpendicular to corresponding sides of the polygon (Figure 2.1).
It is clear, that the perimeter of the internal polygon does not exceed that part of
the perimeter of the external polygon which is inside of these strips.
Hence, the perimeter of the internal polygon is less than the perimeter of the
external polygon.
2.1.2. Given a pyramid SA1A2 . . . An with a vertex S. Consider the layout of its
lateral surface making the cut along the edge SA1 (Figure 2.2).
Figure 2.1
42 2 Application of Projection Method
Аn
A1
S
A2
A3
Figure 2.3 K
N B C
M A D L
Since according to the statement of the problem, the sum of the plane angles at the
vertex S is greater than 180 , then triangle SA1 A01 lays inside of polygon A1 A2 :::An A01 .
According to problem 2.1.1 2SA1 þ A1 A01 < A1 A2 þ A2 A3 þ :::þ An A01 þ A1 A01 .
2.1.3. According to problem 2.1.1, we have that a þ b þ c þ d < 4. If
a þ bþ 1c þ d1 4, then by summing up these two inequalities, we deduce that
1 1
a þ 1a þ b þ 1b þ c þ 1c þ d þ d1 < 8. This leads to a contradiction, as
x þ 1x 2 (x > 0). Therefore, 1a þ 1b þ 1c þ d1 > 4.
2.1.4. It is sufficient to notice that the radius of the circle inscribed into a hexagon is
less than the radius of the circle inscribed into the triangle.
2.1.5. Without loss of generality we can assume that rectangle is inscribed in a
right-angled triangle (Figure 2.3).
Since AB AD, then we have that MN BC > KC. Thus, it follows that
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
NC ¼ MC2 MN 2 < MC2 KC2 ¼ MK, where MNBA is a rectangle, such
that triangle NCF can be placed inside of a triangle MKL similar to it.
2.1.6. (a) At first, let us prove the following lemma.
Lemma If a square PQRS with a side length a is inscribed in a rectangle ABCD
(Figure 2.4), then ABCD is also a square and the square with a side a can be placed
inside of pentagon ABQRD, such that the sides of the square would be parallel to
AB and AD.
2.1 Convex Polygon Lying Inside of Another Polygon 43
Figure 2.4 B Q C
P1 Q1
Q2
P P2
R2 R
S2
S1 R1
A S D
Figure 2.5
a b
Indeed, consider the rotation of square PQRS around point O by an angle 90 .
Then point R will turn into point Q, hence line CD will turn into a line BC.
Consequently, point O is equidistant from sides CD and BC. Similarly, we obtain
that point O is equidistant from sides AB and BC, AD and AB. From above it follows
that ABCD is a square with center O.
Consider a square P1Q1R1S1 with center O, such that its sides are parallel to AB,
AD, and P1Q1 ¼ PQ (Figure 2.4).
!
Perform a translation by a vector Q1 Q2 . Therefore, square P1Q1R1S1 turns into a
square P0 Q0 R0 S0 that is inside of pentagon ABQRD, as one can easily prove that
Q1Q2 ¼ PP2 ¼ SS2. This ends the proof of the lemma.
Using the lemma, we can place the square with a side a inside of a unit square, so
that their sides are parallel and again the square with a side a would not have
common points with a square with a side b (Figure 2.5).
Using the lemma once again, we obtain that the sides of the squares with sides
a and b, which do not have common points, are parallel to the sides of the unit
square. It is not difficult to prove that a þ b < 1.
(b) Let a regular n-gon B1B2 . . . Bn with a side b is inside of a regular n-gon with a
side length a. Let us draw lines passing through the vertices of B1B2 . . . Bn and
parallel to the sides of the n-gon with side a (Figure 2.6).
44 2 Application of Projection Method
Figure 2.6
а
c B2
B3
A2
B1
b
A1 Bn An
Figure 2.7
l
Πc
c
dc
db b
Πb Πa
Figure 2.8 А3
B3¢
B2 B3
A2 B2¢
O
B1
B1¢
A1 Bn An
Figure 2.9
P1
l
P2
d1
d2 P
Note that OB1 ¼ OB2 ¼ . . . ¼ OBn, or in the other words, O is the center of
regular n-gon B1B2 . . . Bn. Choose points B01 , B02 , :::, B0n on segments OA1,
OA2, . . . , OAn correspondingly, so that OB0 1 ¼ OB0 2 ¼ ¼ OB0 n ¼ OB1. Since
points B1, B2, . . . , Bn, B01 , B02 , :::, B0n lay on the same circumference, then from the
condition B1 B2 ¼ B02 B03 it follows that B1 B02 B2 B03 is an equilateral trapezoid.
Consequently, ρ B02 ; B1 B2 ¼ ρ B2 ; B02 B03 . Similarly, one can prove that
ρ B1 ; B01 B02 ¼ ρ B02 ; B1 B2 ¼ ρ B2 ; B02 B03 ¼ ¼ ρ Bn ; B0n B01 ¼ ρ B01 ; Bn B1 .
!
Let B02 M⊥B1 B2 and M 2 B1B2, then at translation by a vector B02 M the image of
the regular n-gon B01 B02 :::B0n will be in a polygon B1B2A3 . . . AnA1.
2.1.8. Let straight line l contains one of the sides of the given triangles and separates
these triangles (Figure 2.9).
Consider bands Π1, Π2, Π containing these triangles and the square with minimal
widths d1, d2, and d, their boundaries being parallel to line l. It is clear that
pffiffiffi
d d1 þ d2. Since 2 d and d1 min (h1, h2, h3), d2 min h01 ; h02 ; h03 , then
pffiffiffi
2 minðh1 ; h2 ; h3 Þ þ min h01 ; h02 ; h03 .
46 2 Application of Projection Method
Figure 2.10
2.1.9. (a) Suppose that the polygon with a side a2 does not contain point O, then the
polygon symmetric to the given polygon with respect to point O is also inside of the
polygon with a side a. Then, according to problem 2.1.6(b), a2 þ a2 < a. This leads to
a contradiction.
(b) At first, let us prove that the circumcircle of the regular n-gon with side b does
not have points outside of the circumcircle of the regular n-gon with side a.
Indeed, assume that these circles intersect. Let us choose point M on the larger
circle, such that it is not a vertex of the regular n-gon with a side a and is inside
of the smaller circle (Figure 2.10).
Let given regular n-gons be A1A2 . . . An (A1A2 ¼ a) and B1B2 . . . Bn. Assume that
point M is on the smaller arc A1A2.
We have that ∠A1 MA2 ¼ πðn1 n
Þ
and that polygon A1A2 . . . An is inside of angle
∠A1MA2.
Consequently, polygon B1B2 . . . Bn also is inside of that angle.
Hence, ∠A1 MA2 > max ∠Bi MBj > π ðn1 Þ
n . This leads to a contradiction.
i, j
Let us denote the radiuses of these circles by Ra and Rb and their centers by
O and O1. From aforesaid, it follows that Ra Rb þ OO1, and since point O is not
inside of polygon B1B2 . . . Bn, then OO1 > rb, where rb is the radius of the circle
inscribed in B1B2 . . . Bn.
a b a
Thus, Ra > Rb þ rb, or > . We deduce that b < π.
2 sin πn π b 2cos 2 2n
2 sin þ
n 2tg π
n
2.1.10. It is enough to prove the problem for triangle A1B1C1 (Figure 2.11a, b),
where A1B1 k AB, B1C1 k BC, A1C1 k AC.
Consider the following two cases: Figure 2.11a, b.
Let us consider first the case of Figure 2.11a. Perform a rotation around vertex C2
by some angle, so that one of the sides with vertices C2, B2, A2 of polygon
M becomes for the first time parallel to lines A1B1, A1C1, or B1C1, respectively.
Let u 2 [u1, u2], u1 and u2 being the values of u for aforesaid rotations (in positive
and negative directions). Let us denote by Mu the image obtained by a rotation of
2.1 Convex Polygon Lying Inside of Another Polygon 47
Figure 2.11 B
B1
C2 A2
u nf(u) j-u
mf(u) Mu
a g
A1 B2 C1
A C
a
B
B1
u
mf(u)
Mu
a g
A1 B2 C1
A C
b
where a, φ, and φ1 are constants. Since, in case u 2 [u1, u2] f(u) > 0, or equivalently,
sin(u þ φ1) > 0, then on segment [u1, u2], function sin(u þ φ1) accepts its minimal
value at point u1 or u2. Hence, on segment [u1, u2], function f(u) accepts its
minimal value at points u1 or u2.
The proof is similar for the case of Figure 2.11b. This ends the proof.
48 2 Application of Projection Method
a' b'
a b
c
B' c' A'
C'
a' b'
ha'
a b
c
B' c' A'
Π a'
Figure 2.13
2.1.11. According to problem 2.1.10, if the first triangle is covered by the second
triangle, then one can assume that one of the sides of the first triangle lays on the
side of the second triangle. Let us find necessary and sufficient conditions for the
first triangle to be covered by the second one in a way shown in Figure 2.12.
Let ha0 , hb0 , hc0 be the altitudes of the triangle with sides a0 , b0 , c0 drawn to the
sides a0 , b0 , c0 , correspondingly, and Πa0 , Πb0 , Πc0 be the projections of the first
triangle on the straight lines, containing heights ha0 , hb0 , hc0 (Figure 2.13).
Let us prove that, or order that, one can place the triangle with sides a, b, c inside
of the triangle with sides a0 , b0 , c0 in a way shown in Figure 2.12, it is necessary and
sufficient that conditions
hold true.
The necessity of the conditions ha0 Πa0 follows from the fact that the triangle
0 0 0 0 0
sides a, b, c is inside of a band with boundaries B C and l1 , where A 2 l1 ,
with
0 0 0
l1 B C , (see Figure 2.12). Similarly, one can prove the necessity of conditions
hb0 Πb0 and hc0 Πc0 .
2.1 Convex Polygon Lying Inside of Another Polygon 49
C'
a'
a
b b'
β' β α α'
B' B A A'
Figure 2.14
Figure 2.15
C
d j l
E F
l1
Now, if condition (2.1) is satisfied, then it is clear that c c0 , otherwise Πa0 > ha0 .
Let us consider Figure 2.14.
!
If the parallel displacement of point C by a vector BB0 has moved it in the shaded
half-band, then it means that hb0 < Πb0 . This leads to a contradiction. This means
that, if condition (2.1) is satisfied, then the first triangle can be placed inside of the
second one.
Note that, if point C and straight line l (Figure 2.15) are given, then the
projection of segment CE (E 2 l ) on line l1 is equal to CE| cos (φ δ)|.
Thus, we obtain that Πc0 ¼ hc , Πb0 ¼ 12 ða j sin ðα0 þ βÞj þ b j sin ðα α0 Þjþ
c sin α0 Þ, Πa0 ¼ 12 ða j sin ðβ0 βÞj þ b j sin ðβ0 þ αÞj þ c sin β0 Þ (see the proof of
problem 2.2.1a). Then, condition (2.1) can be rewritten as:
1
hc0 hc , hb0 ða j sin ðα0 þ βÞj þ b j sin ðα α0 Þj þ c sin α0 Þ,
2 ð2:2Þ
1
ha ða j sin ðβ0 βÞj þ b j sin ðβ0 þ αÞj þ c sin β0 Þ:
0
2
It is clear that for condition (2.2) one needs only the values of a, b, c, a0 , b0 , c0 .
Note that, for two triangles 18 variants of dispositions (similar to Figure 2.12)
are possible and for the first triangle to be covered by the second triangle, it is
50 2 Application of Projection Method
necessary and sufficient that condition (2.2) is satisfied at least for one of these
18 variants.
2.1.12. We have to prove that f(A1B1C1) f(ABC), if ΔABC covers ΔA1B1C1,
where f(ABC) 2 R and that f(ABC) ¼ f(MNK), if ΔABC ¼ ΔMNK. According to
problem 2.1.10, one can assume that one of the sides of triangle A1B1C1 lays on
one of the sides of triangle ABC. Without loss of generality, we can assume that two
sides of triangle A1B1C1 lay on two sides of triangle ABC, one of the two sides of
triangle A1B1C1 coincides with one of the sides of triangle ABC.
Indeed, in the case of Figure 2.16a, we have that
Therefore, f(A1B1C1) f(ABC), while in the case of Figure 2.16b, we have that
f(A1B1C1) f(A1B1B2) f(A1BB2) f(A1BC) f(ABC). Thus, it follows that
f(A1B1C1) f(ABC).
(a) Let triangles A1B1C1 and ABC are as it is shown in Figure 2.17.
Figure 2.16 C C
A2 B2
C1
C1
B B1 A1 AB B1 A1 A
a) б)
Figure 2.17 Cº C1
g1
g
b=b1 a1 a
a=a1 b1 b
Aº A1 c1 B1 B
c
2.1 Convex Polygon Lying Inside of Another Polygon 51
Let us consider the following three cases: max(α1, β1, γ 1) ¼ γ 1, max(α1, β1, γ 1) ¼ α1,
and max(α1, β1, γ 1) ¼ β1.
I. If max(α1, β1, γ 1) ¼ γ 1. Therefore, max(α, β, γ) ¼ γ. Note that ∠CB1B ¼
180 β1 > 90 , which means that a > a1. Hence, a þ b ¼ a þ b1 > a1 þ b1.
II. If max(α1, β1, γ 1) ¼ α1, then α ¼ α1 β1 > β. Consequently, max(α, β, γ) ¼ α
or max(α, β, γ) ¼ γ.
If max(α, β, γ) ¼ α, then b þ c ¼ c þ b1 > b1 þ c1.
If max(α, β, γ) ¼ γ, then γ α ¼ α1 γ 1. Therefore, ∠CB1B ¼ 180 β1 >
90 . Consequently, a > a1, this means that a þ b ¼ a þ b1 >
a1 þ b 1 b 1 þ c 1.
III. Let max(α1, β1, γ 1) ¼ β1. Consider the following three cases:
1. max(α, β, γ) ¼ β, then a þ c ¼ a þ BB1 þ c1 > a1 þ c1.
2. max(α, β, γ) ¼ α, then β1 α1. Consequently b1 a1, hence b þ c ¼
b1 þ c a 1 þ c > a 1 þ c 1.
3. max(α, β, γ) ¼ γ. In this case, let us notice that if point E moves from
point B1 to point B along segment B1B, then the value of
∠AEC ∠ACE decreases from β1 γ 1 (β1 γ 1 0) to the value of
β γ (β γ 0). Consequently, on segment B1B exists such point
E that ∠AEC ¼ ∠ACE. Then a1 þ c1 AE þ CE ¼ AC þ CE ¼
b þ CE a þ b. Consequently, a1 þ c1 a þ b (∠CEB > 90 , if point
E does not coincide with point B).
This ends the proof of (a).
(b) In order to end the proof, note that if max(α, β, γ) 120 , then T(ABC) ¼
AB þ BC þ AC max (AB, BC, AC) (see problem 1.2.8) and if max(α, β, γ) <
120 , then T(ABC) ¼ BB0 (Figure 2.18).
Indeed, since max(α, β, γ) < 120 , then quadrilateral ABCB0 is convex and point
B is outside of the circumcircle of the equilateral triangle ACB0 . Let point М0 be such
Figure 2.18 B
M0
M
A 600 C
0 0
60 60
M¢
B¢
52 2 Application of Projection Method
that AМ0 ¼ AM and ∠MAM0 ¼ 60 (Figure 2.18). Then ΔAMC ¼ ΔAM0 B0 and conse-
quently, AM þ BM þ CM ¼ MM0 þ BM þ M0 B0 BB0 ¼ BM0 þ M0B0 ¼ BM0 þ
AM0 þ CM0. This means that T(ABC) ¼ BB0 . Note that T(ABC) ¼ BM0 þ AM0 þ
CM0 > BM0 þ AM0 > AB and T(ABC) ¼ BB0 < AB þ AB0 ¼ AB þ AC. Similarly, one
can prove that T(ABC) > BC, T(ABC) > AC, T(ABC) < BC þ AC, T(ABC) <
AB þ BC.
Without loss of generality, we can assume that side A1C1 is on side AC.
Consider the following cases:
I. If max(α, β, γ) < 120 , max(α1, β1, γ 1) < 120 , then convex quadrilateral
A1B1C1B10 has a diameter T(A1B1C1) and is located inside of convex quadri-
lateral ABCB0 with a diameter T(ABC). Hence T(A1B1C1) T(ABC).
II. If max(α, β, γ) 120 , max(α1, β1, γ 1) < 120 , then according to problem
2.1.12a, we have that
Figure 2.19
h1
2.1 Convex Polygon Lying Inside of Another Polygon 53
Figure 2.20 B
B
B1
B1
A≡A1 C1 C A A1 C≡C1
a b
Figure 2.21 B
B1
A C
Figure 2.22 C
C1
A A1 H M B1 B
Now, let AB1 > AC, then ∠AB1C < ∠ACB1. Consequently, ∠BB1 C > π2.
If ∠AB1A1 ∠BCB1, then the proof of the problem is obtained by rotating
triangle A1B1C around point C by angle ∠AB1A1.
Let AB1 > AC and ∠AB1A1 > ∠BCB1, then ∠AB1C1 < ∠ABC. Since
B1C < BC, A1B1 < AB, ∠AB1C1 < ∠ABC, then if triangle A1B1C1 is placed in
such a way that points B1 and B coincide, while point A1 is on side AB, then vertex
C1 of triangle A1B1C1 is inside of triangle ABC. This ends the proof.
Remark If ABC is an acute triangle, then the statement of the problem is wrong
(see the example in Figure 2.21).
2.1.15. According to problem 2.1.10, one can assume that points A1 and B1 lay on
segment AB (Figure 2.22).
54 2 Application of Projection Method
Note that ∠AC1 B ∠ACB π2. Therefore, if M is the midpoint of segment AB,
pffiffiffi
then AB 2C1 M 2C1 H ¼ 3A1 B1 .
Remark Similarly, one can prove that, if the triangle with altitudes h1, h2, h3 is
inside of non-acute triangle ABC, then the inequality max(AB, BC, AC) 2 min (h1,
h2, h3) holds true.
Another proof of the problem can be obtained by using problem 2.1.13.
2.1.16. (Solution of M.D. Kovalev.) First we need to prove that the side of an
equilateral triangle, containing ΔDEF, is not less than p2ffiffi3 cos 10 . According to
problem 2.1.10, it is sufficient to consider the following two cases (see Figure 2.23).
pffiffi
In case (a), we have that 23 a cos 10 . Consequently, a p2ffiffi3 cos 10 .
In case (b), we have that a D1 E ¼ sin 100
sin 60
¼ p2ffiffi3 cos 10 , where ∠FD1E ¼ 60 .
Consider now the minimal equilateral triangle X1Y1Z1 with sides parallel to the
sides of triangle XYZ, which contains triangles XYZ and DEF (Figure 2.24).
a E a a a
1
F
200
D F D1 D E
a b
Figure 2.23
Y1 Y1 Y1
Y x Y y
a x a a
Y
х
X Z X Z
y z
XºX1 Z Z1 X1 Z1 X1 Z1
a b c
Figure 2.24
2.1 Convex Polygon Lying Inside of Another Polygon 55
The area of a convex figure containing ΔXYZ and ΔDEF is not less than:
pffiffiffi pffiffiffi pffiffiffi
3 1 3 3
in case (a): þ x¼ þ ða 1Þ,
4 2 4 4
pffiffiffi pffiffiffi pffiffiffi
3 xþy 3 3
in case (b): þ ¼ þ ða 1Þ,
4 2 4 4
pffiffiffi pffiffiffi pffiffiffi
3 xþyþz 3 3
in case (c): þ ¼ þ ða 1Þ.
4 2 4 4
pffiffi
It remains to note that 43 a 12 cos 10 .
Remark Among all convex figures covering any triangle with the sides not exceed-
ing 1, the least area has a triangle ABC, such that ∠A ¼ 60 , AB ¼ 1, and the altitude
drawn to AB is equal to cos10 . The area of that triangle is equal to 12 cos 10 .
2.1.17. Let sides AB and BC of rectangle ABCD be equal to 1 and 2 respectively,
and squares with sides a, b, c are not mutually intersecting and lay inside of
rectangle ABCD.
Note that, for each side of a rectangle one can find a “good” square, such that,
while moving it in a direction perpendicular to the given side, it does not intersect
with the other squares before intersecting with that side. According to the
Dirichlet’s principle there is a “good” square simultaneously for two sides of the
rectangle. If these are opposite sides, then this ends the proof (see Figure 2.25). In
the case of Figure 2.25a, we have that a x þ y ¼ MN and b ND. Consequently,
a þ b CD ¼ 1. We have that c 1. Thus, a þ b þ c 2.
In the case of Figure 2.25b, according to problem 2.1.6a, we have that
b þ c 2 MN, a MN. Therefore, a þ b þ c 2 MN þ MN ¼ 2.
If these sides are adjacent, then it is possible to assume that the “good” square is
located on one of these sides (see the proof of problem 2.1.6a). By repeating these
reasonings for the new squares we eventually find that one of these squares lies in
one of the corners of the rectangle (see Figure 2.26).
B C B C
x М
y b
2-MN
x N
b c b
A D A M N D
2-MN
a b
Figure 2.25
56 2 Application of Projection Method
Figure 2.26 B P C
A Q D
b a b a c b a
c
c
a b c
Figure 2.27
b x y
c y
c x
а c b
d
a
a
a b
Figure 2.28
If the projections of any two squares on side CD do not intersect, then as it has
been proven above, a þ b þ c 2. Otherwise, there exists a straight line parallel to
AD crossing all three squares. If none of the squares with sides b and c crosses
segment PQ (see Figure 2.26), then as we have already proven, a þ b þ c 2. Thus,
possible positions of the squares are presented in fig 2.27.
From the proof of problem 2.1.6a, it follows that instead of Figure 2.27 one can
consider the variants presented in Figure 2.28. In the case of Figure 2.28a, we have
that a þ b þ c a þ b þ d 2 and, in the case of Figure 2.28b, we have that a þ b þ
c a þ x þ y þ c ¼ (a þ x) þ (c þ y) 1 þ 1 ¼ 2.
2.1.18. At first, note that if point B is on side AD of triangle ACD, then rABC rADC
(by rXYZ we denote the radius of the incircle of triangle XYZ).
Indeed, let O and O1 be the centers of the incircles of triangles ABC and ADC,
respectively. Since, these points belong to bisector AA1 of triangle ADC, then
∠ACO ¼ 12 ∠ACB 12 ∠ACD ¼ ∠ACO1 . Therefore, AO AO1. Let E, E1 2 AC
and OE ⊥ AC, O1E1 ⊥ AC. Then, ΔAOE ΔAO1E1, thus rrADC ABC
¼ AO
AO
1
1.
2.1 Convex Polygon Lying Inside of Another Polygon 57
Let triangle ABC is covered by a unit square 1, and vertex B does not belong to
any of the sides of the square. Then, if ray AB intersects one of the sides of the
square at point D, we have that rABC rADC.
From the aforesaid, it follows that, it is sufficient to prove the inequality
pffiffi
r 541, for triangle ABC, such that all of its vertices are on the sides of the square.
Let the vertices of the triangle are on the sides of square MNPQ and MN ¼ 1.
Consider the following cases:
(a) A, B 2 MN, C 2 PN, with B being on segment AN.
pffiffi pffiffi pffiffi
Since r ¼ r ABC r ACN r APN r MPN ¼ 1 22 < 541, then r < 541.
(b) A, B 2 MN, C 2 PQ with B being on segment AN. Let C0 be the midpoint of the
segment PQ, then
pffiffiffi
1 1 51
r ¼ rABC rACN rMCN ¼ ¼ r MCN0 ¼ :
MC þ CN þ 1 MC0 þ C0 N þ 1 4
Note that
pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 þ x2 þ 1 þ y2 þ ð1 xÞ2 þ ð1 yÞ2
pffiffiffiffiffiffiffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 2
1 þ x 1 þ y þ ð1 x Þ þ ð1 y Þ
2 2
pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ 1 þ x2 1 1 þ y2 1 þ 1
x2 y2
¼ 1 pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi 1 x y,
1 þ x2 þ 1 1 þ y2 þ 1
therefore
pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffiffi
1 þ x2 þ 1 þ y 2 þ ð1 x Þ2 þ ð1 y Þ2 1 þ x 2 1 þ y 2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
þ ð1 xÞ2 þ ð1 yÞ2 þ 1 xy
pffiffiffi
1 þ 5 ð1 xyÞ:
2.1.19. Let a triangle with sides a1, b1, c1 be inside of the triangle with sides a, b, c.
Is it true that a21 þ b21 þ c21 a2 þ b2 þ c2 ?
2.1.20. Given points A1, B1, and C1 on sides BC, CA, AB of triangle ABC, respec-
tively, such that ∠AB1C1 þ ∠BC1A1 þ ∠CA1B1 ¼ 180 . Prove that triangle
A1B1C1 can be placed inside of one of the triangles AB1C1, BC1A1, CA1B1.
2.1.21. Given a point O inside of triangle ABC and points A1, B1, C1 on sides BC,
CA, AB, respectively, such that OB1 ⊥ AC, OA1 ⊥ BC, OC1 ⊥ AB.
Prove that triangle A1B1C1 can be placed inside of one of triangles AB1C1,
BC1A1, CA1B1.
2.1.22. Let a convex n-gon be inside of a unit square. Prove that one can find three
vertices A, B, C of this n-gon, so that the area of triangle ABC does not exceed
(a) 12, for n ¼ 3, 4, (b) 14, for n ¼ 5, (c) 18, for n ¼ 6, (d)n82 , for n 7.
2.1.23. Let a convex quadrilateral with the sum of the lengths of the diagonals equal
to d0 be inside of a convex quadrilateral with the sum of the lengths of the diagonals
equal to d. Prove that d0 < 2d.
2.1.24. Let a convex polygon with area S2 and perimeter P2 be inside of the convex
polygon with area S1 and perimeter P1. Prove that 2PS11 > PS22 .
2.1.25. Let a convex quadrilateral with the sum of the lengths of the pairwise
distances of its vertices (i.e., the sum of all its sides and diagonals) equal to S2 be
inside of a convex quadrilateral with the same sum equal to S1. Prove that S2 < 4S31 .
2.2 Sufficient Conditions for Comparison of Lengths of Two Broken Lines on the Plane 59
2.1.26. Is it possible to place inside of a unit square two regular triangles with sides
qffiffi
greater than 23, so that they do not intersect?
2.1.27. Prove that a regular n-gon with a side a can be placed inside of a regular
n þ 1-gon with a side a.
2.1.28. Let a parallelogram with altitudes h1 and h2 be inscribed in a convex
quadrilateral ABCD. Prove that quadrilateral ABCD can be placed inside of a
rectangle with one of the sides equal to h1 þ h2.
2.1.29. (a) Prove that, if a convex polygon lays inside of a circle, then the perimeter
of the polygon is less than the circumference of the circle.
(b) Prove that, if a circle lays inside of a convex polygon, then the circumference of
the circle is less than the perimeter of the polygon.
Hint Inscribe regular polygons in the circle.
2.1.30. Let O be a given point in triangle ABC with the lengths of sides AB ¼ c,
BC ¼ a, CA ¼ b, a b c. Prove that there exists a vertex of the triangle, such that
the distance from point O does not exceed pbffiffi2.
Hint Consider cases when the point is either in triangle ADC or in ABD, where
AD ⊥ BC(D 2 BC).
2.2.1. (a) Given on a plane two sets of segments A1B1, . . . , AnBn and C1D1, . . . ,
CmDm, the sum of lengths of projections of the segments of the first set on any
straight line of this plane is greater than the sum of the lengths of projections of the
segments of the second set on the same straight line. Prove that the sum of the
lengths of the segments of the first set is greater than the sum of the lengths of the
segments of the second set.
(b) Given on a plane two sets of segments A1B1, . . . , AnBn and C1D1, . . . , CmDm.
Let l1, . . . , ln be a straight line on that plane, such that AiBi ⊥ li, i ¼ 1, . . . , n.
Given that for any i (i ¼ 1, . . . , n) the sum of the lengths of projections of the
segments of the first set on any straight line li is greater than the sum of the
lengths of projections of the segments of the second set on the same straight
line. Prove that the sum of lengths of segments of the first set is greater than the
sum of lengths of the segments of the second set.
(c) Given on a plane two sets of segments A1B1, . . . , AnBn and C1D1, . . . , CmDm.
Let l1, . . . , ln be such straight line on that plane that AiBi ⊥ li, i ¼ 1, . . . , n. It is
known that for any value of i (i ¼ 1, . . . , n) the sum of the lengths of pro-
jections of the segments of the first set on any straight line li is greater than the
60 2 Application of Projection Method
sum of the lengths of projections of the segments of the second set on the same
straight line. Prove that the sum of the lengths of the segments of the first set on
any line on the plane is greater than the sum of the lengths of the segments of
the second set on the same line.
2.2.2. Prove that for points A, B, C, D, E on a plane the following inequality
AB þ CD þ DE þ EC AC þ AD þ AE þ BC þ BD þ BE holds true.
2.2.3. Four points on the straight line are denoted by letters A, B, C, D. Prove that
AE þ ED þ |AB CD| > BE þ CE holds true for any point E outside that line.
2.2.4. Given on a plane vectors ~ a, ~ c, ~
b,~ d, whose sum is equal to ~ 0. Prove that
~
a þ ~
b þ ~
c þ ~
d ~ d þ ~
a þ~ b þ~d þ ~ d .
c þ~
2.2.5. Given on a plane two arbitrary triangles ABC and A1B1C1. Prove that
2.2.6. Given on a plane several segments, whose sum of lengths is equal to π. Prove
that it is possible to choose a straight line, so that the sum of lengths of projections
of the segments on this straight line is (a) less than 2, (b) more than 2.
2.2.7. Given on a plane several vectors, whose sum of lengths is equal to π. Prove
that it is possible to choose several of these vectors, so that the length of their sum is
greater than 1.
Solutions
2.2.1. (a) Given on a plane a segment AB and a straight line l. Consider projection of
segment AB on line l and denote by l(AB) the length of that projection.
If AB ⊥ l, then l(AB) ¼ 0. Similarly, for polygon ϕ denote by l(ϕ) the length of
the projection of polygon ϕ on line l.
It is not difficult to prove that, if point C belongs to segment AB, then l(AB) ¼ l
(AC) þ l(CB). On the other hand, if polygon A1A2 . . . An is convex, then l(A1A2) þ l
(A2A3) þ . . . þ l(An 1An) þ l(AnA1) ¼ 2l(A1A2 ... An) and if segments AB and CD
are either on the same line or on the parallel lines, then
CD l(AB) ¼ AB l(CD).
(b) Let us first prove the following lemma.
Lemma If on a plane are given such segments M1N1, M2N2, . . . , MpNp ( p 2),
that no two of them are on the same or parallel lines, then there exists a convex and
M N M N
centrally symmetric polygon with sides M12N1 , M22N2 , :::, p2 p , M12N 1 , M22N2 , :::, p2 p ,
so that the sides with length M2i Ni are parallel to segments MiNi, for any 1 i p.
We proceed the proof by mathematical induction.
2.2 Sufficient Conditions for Comparison of Lengths of Two Broken Lines on the Plane 61
Figure 2.29 M2 N2
N1
M1
Figure 2.30 d d’
fk Nk+1
fk ¢ A¢
О
A fk ¢¢
Mk+1
Figure 2.31
Figure 2.32 fA
M¢ Bi
M O
Ai
li
E F
segment with length AiBi þ AjBj (parallel to segment AiBi). So, one obtains that
segments A1B1, . . . , AnBn satisfy to the conditions of the lemma.
Let ϕA be a convex polygon with a center of symmetry at point O, with the sides
parallel to segments A1B1, . . . , AnBn and equal to A12B1 , :::, An2Bn , A12B1 , :::, An2Bn .
The polygon (or segment) ϕC, for segments C1D1, . . . , CmDm with the same
symmetry center, is defined similarly.
Let us prove that polygon ϕC is inside of polygon ϕA. Otherwise, there exists
such a point M 2 ϕC, as it is shown in Figure 2.32.
Let M0 be the symmetric point to point M with respect to point O.
We have that
Remark
1. Similarly, one can prove that if among segments A1B1, . . . , AnBn there are two
located on non-parallel (crossing) lines and li(A1B1) þ . . . þ li(AnBn)
li(C1D1) þ . . . þ li(CmDm), i ¼ 1, . . . , n, then A1B1 þ . . . þ
AnBn C1D1 þ . . . þ CmDm.
2. If for any line l it holds true l(A1B1) þ . . . þ l(AnBn) l(C1D1) þ . . . þ l
(CmDm), then A1B1 þ . . . þ AnBn C1D1 þ . . . þ CmDm.
(c) Since ϕC is inside of ϕA (see the solution of problem 2.2.1b), then l(ϕC) < l(ϕA),
so 2l(ϕC) < 2l(ϕA). Therefore, l(A1B1) þ . . . þ l(AnBn) > l(C1D1) þ . . . þ l
(CmDm).
Remark One can prove that, if there are given segments A1B1, . . . , AnBn on a plane
and lines li ⊥ AiBi, i ¼ 1, . . . , n, then for any line l of that plane and for arbitrary
mi > 0, i ¼ 1, . . . , n, holds true the following inequality
2.2 Sufficient Conditions for Comparison of Lengths of Two Broken Lines on the Plane 63
2.2.2. According to remark 2 (see the solution of problem 2.2.1b) it is enough to prove
the inequality for the projections of points A, B, C, D, E on a line, i.e., it is enough to
solve the problem for the case when points A, B, C, D, E are on the same line.
As the inequality is symmetric with respect to points C, D, E, then we can take
that E lays between points C and D, then DE þ EC ¼ CD.
According to the triangle inequality AC þ AD CD, BC þ BD CD,
AE þ BE AB. Summing up these three inequalities, we deduce that
AC þ AD þ AE þ BC þ BD þ BE AB þ CD þ DE þ EC:
2.2.3. Let AB > CD (in the case of AB < CD the proof is similar). One has to prove
that
AE þ ED þ AB > CD þ BE þ CE: ð2:3Þ
Figure 2.33 E
A B C D
N
64 2 Application of Projection Method
! ! !
2.2.4. Consider points A, B, C, D, such that a, BC ¼ ~
AB ¼ ~ b, CD ¼ ~
c, then
! ~
DA ¼ d. We have to prove that
AB þ BC þ CD þ DA AC þ BD þ 2MN, ð2:4Þ
where M and N are the midpoints of segments BD and AC. According to remark
2 (see the solution of problem 2.2.1b) it is sufficient to prove inequality (2.4) for
projections of points A, B, C, D on a line, i.e., it is enough to solve the problem for
the case, when points A, B, C, D are on the same line.
We can assume that A(0), B(b), C(c), D(d), and b, c, d 0, d b, then we must
prove that |c b| þ |c d| c 2b þ |d þ b c|.
If d þ b c, then |c b| þ |c d| |(c b) (c d)| ¼ d b ¼ c 2b þ |
d þ b c|.
If d þ b < c, then |c b| þ |c d| ¼ c b þ c d c 2b þ |d þ b c|.
2.2.5. According to remark 2 (see the solution of problem 2.2.1b) it is sufficient to
prove the inequality for the projections of points A, B, C, A1, B1, C1 on a line, i.e., it
is enough to prove the problem for the case, when points A, B, C, A1, B1, C1 are on
the same line.
As the inequality is symmetric with respect to points A, B, C (A1, B1, C1), without
loss of generality one can assume that point B lays between points A and C and point
B1 lays between points A1 and C1, A1C1 AC.
Then AB þ BC þ AC þ A1B1 þ B1C1 þ A1C1 ¼ 2AC þ 2A1C1. Since AC AB1
þ B1C, A1C1 A1B þ BC1, AC AA1 þ CA1, A1C1 AC AC1 þ CC1, then by
summing up these inequalities we obtain that
2AC þ 2A1 C1 AA1 þ AB1 þ AC1 þ BA1 þ BC1 þ CA1 þ CB1 þ CC1 :
Therefore
AB þ BC þ AC þ A1 B1 þ B1 C1 þ A1 C1
AA1 þ AB1 þ AC1 þ BA1 þ BB1 þ BC1 þ CA1 þ CB1 þ CC1 :
2.2.6. Given segments A1B1, . . . , AnBn. We can assume that any two segments do
not belong to the same or parallel straight lines. For n ¼ 1 the statement of the
problem is evident, let n 2.
Construct a polygon ϕA (see the solution of problem 2.2.1b).
(a) Let O be the center of symmetry of polygon ϕA, and d be the minimal of all
distances between opposite sides (Figure 2.34).
Figure 2.34 l fA
O d
2.2 Sufficient Conditions for Comparison of Lengths of Two Broken Lines on the Plane 65
Figure 2.35
fA
d1
O
l1
Since the circumference with a center O and radius d2 is inside of polygon ϕA, the
length of that circumference is less than the perimeter of polygon ϕA, i.e., πd < π
(see problem 2.1.29b). Then the sum of the lengths of the projections of the
segments A1B1, . . . , AnBn on line l is equal to 2d and 2d < 2 (Figure 2.34).
(b) Consider all diagonals of the ϕA polygon which pass through point
O (Figure 2.35) and let d1 be the length of the largest of them.
Since polygon ϕA is inside of the circumference with a center O and radius d21 , the
perimeter of polygon ϕA is less than the length of that circumference, i.e., πd1 > π
(see problem 2.1.29a). Then the sum of the lengths of the projections of the
segments A1B1, . . . , AnBn on line l1 (Figure 2.35) is equal to 2d1. Thus, it follows
that the sum is larger than 2.
! ! ! !
2.2.7. Let be given vectors A1 B1 , :::, An Bn , such that A1 B1 þ ::: þ An Bn ¼ π.
Consider segments A1B1, . . . , AnBn, then according to problem 2.2.6b there
exists a straight line l, such that l(A1B1) þ . . . þ l(AnBn) > 2 (see notations in the
!
solution of problem 2.2.1). Let ~ e be a unit vector on line l and Ai Bi ,~ e 90 ,
!
i ¼ 1, . . . , k and Ai Bi ,~e > 90 , i ¼ k þ 1, . . . , n. Then, either l(A1B1) þ . . . þ l
(AkBk) > 1 or l(Ak þ 1Bk þ 1) þ . . . þ l(AnBn) > 1. Let l(A1B1) þ . . . þ l(AkBk) > 1.
! !
We have that A1 B1 þ ::: þ Ak Bk lðA1 B1 Þ þ ::: þ lðAk Bk Þ > 1. Therefore,
! !
A1 B þ ::: þ A k Bk > 1.
1
2.2.8. Points C1, A1, B1 are taken correspondingly on sides AB, BC, AC of triangle
ABC, such that BA1 ¼ λBC, CB1 ¼ λCA, AC1 ¼ λAB, where 12 < λ < 1. Prove that for
perimeter P of triangle ABC and for perimeter P1 of triangle A1B1C1 it holds true
(2λ 1)P < P1 < λP.
2.2.9. Point O is inside of triangle ABC with a perimeter P. Prove that
P
2 < AO þ BO þ CO < P.
66 2 Application of Projection Method
2.2.10. Prove that, if the lengths of all sides and diagonals of a convex polygon is
less than d, then its perimeter is less than πd.
2.2.11. Given several convex polygons and it is known that it is impossible to draw
a straight line, so that it does not intersect any polygon and on both sides of it there
is at least one polygon. Prove that all these polygons can be confined within a
polygon with the perimeter not exceeding the sum of their perimeters.
2.2.12. Given on a plane several vectors, whose sum of the length is equal to 1.
Prove that they can be broken into three groups (possibly empty), so that the sum
pffiffi
of the lengths of the vectors in these groups is more than 32π3.
Remark Add to these n vectors another 2n vectors obtained from the vectors of the
given set by rotation by angle 120 clockwise or counterclockwise.
2.2.13. Given on a plane a convex n-gon. Let ak be the length of its k-th side and dk be
the length of the projection of n-gon on a line containing that side (k ¼ 1, 2, . . . , n).
Prove that da11 þ ::: þ dann 4.
Remark First prove the statement of the problem for centrally symmetric polygons.
2.3.1.
Given convex polygon A1A2 . . . An and points B1, B2, . . . , Bn on its sides
B1 2 A1 A2 ; B2 2 A2 A3 ; :::; Bn 2 An A1 ; Bi Aj , so that for any points C1, . . . , Cn on
the side of polygon A1 . . . An (C1 2 A1A2, . . ., Cn 2 AnA1, Ci Aj ) the inequality
B1B2 þ B2B3 þ . . . þ Bn 1Bn þ BnB1 C1C2 þ C2C3 þ . . . þ Cn 1Cn þ CnC1
is true.
Prove that
2.3.2.
Given convex polygon A1A2 . . . An and points B1, B2, . . . , Bn on its sides
B1 2 A1 A2 ; B2 2 A2 A3 ; :::; Bn 2 An A1 ; Bi Aj , so that condition (2.5) of problem
2.3.1 is satisfied. Prove that for any points C1, . . . , Cn on the sides of polygon
A1A2 . . . An (C1 2 A1A2, ..., Cn 2 AnA1) the following inequality is true:
B1B2 þ B2B3 þ . . . þ Bn 1Bn þ BnB1 C1C2 þ C2C3 þ . . . þ Cn 1Cn þ CnC1.
2.3.3. Prove that if A1A2A3 is not an acute-angled triangle, then on its sides do not
exist such points B1, B2, B3 (B1 2 A1A2, B2 2 A2A3, B3 2 A3A1, Bi Aj) for which the
condition (2.5) of problem 2.3.1 is satisfied.
2.3.4. Prove that if A1A2A3 is an acute-angled triangle, then on its sides exist such
points B1, B2, B3 (B1 2 A1A2, B2 2 A2A3, B3 2 A3 A1 , Bi Aj for which the condition
(2.5) of problem 2.3.1 is satisfied.
2.3 Inscribed Polygons with the Least Perimeter 67
2.3.5. Given convex quadrilateral A1A2A3A4 and points B1, B2, B3, B4 on the sides of
that quadrilateral (B1 2 A1A2, B2 2 A2A3, B3 2 A3A4, B4 2 A4A1, Bi Aj ) so that the
condition (2.5) of problem 2.3.1 is satisfied. Prove that quadrilateral A1A2A3A4 is
inscribed and that
2.3.6. Prove that if a convex quadrilateral A1A2A3A4 is inscribed and for it condition
(2.6) of problem 2.3.5 is satisfied, then on the sides of that quadrilateral exist such
points B1, B2, B3, B4 (B1 2 A1A2, B2 2 A2A3, B3 2 A3A4, B4 2 A4A1, Bi Aj), for which
the condition (2.5) of problem 2.3.1 is satisfied.
2.3.7. Given a regular tetrahedron A1A2A3A4 with edge 1 and points B1, B2, B3, B4 on
its faces A2A3A4, A1A3A4, A1A2A4, and A1A2A3, respectively. Prove that
pffiffiffiffiffi
B1 B2 þ B2 B3 þ B3 B4 þ B4 B1 0, 4 10.
2.3.8. Given a cub ABCDA0 B0 C0 D0 with edge 1 and points B1, B2, B3, B4, B5, B6 on its
faces ABCD, AA0 B0 B, BB0 C0 C, A0 B0 C0 D0 , DD0 C0 C, AA0 D0 D, respectively. Prove that
pffiffiffi
B1 B2 þ B 2 B3 þ B 3 B4 þ B 4 B5 þ B 5 B6 þ B 6 B1 2 3 .
Solutions
Figure 2.36 An
Bn
A3
B2
M A2 B B1 A1
B'2
68 2 Application of Projection Method
Figure 2.37 ln
Bn-1
An
Bn
A3 bn l1
B2 b1 C1 A1
C1'
b1 B1
A2
C1''
l2
B02 B01 þ B01 Bn < B02 B1 þ B1 Bn , thus B2 B01 þ B01 Bn < B2 B1 þ B1 Bn , which means
that
hence
Remark
1. The equality holds true if and only if
Figure 2.38 A1
B3
B1
A3 B2 A2
Figure 2.39 A1
B1
B3
A3 B2 A2
70 2 Application of Projection Method
Figure 2.40 A2
B2
B1
P
O A3
A1
B3
B4
A4
Since points B1 and B3 lay on the circumference with a diameter A1H, we have
∠B3B1A1 ¼ ∠B3HA1. Similarly we get that ∠B2B1A2 ¼ ∠B2HA2 and since
∠B3HA1 ¼ ∠B2HA2, then ∠B3B1A1 ¼ ∠B2B1A2.
The other two equalities are proved similarly.
Remark According to problems 2.3.1 and 2.3.3, a triangle with a minimum perim-
eter can be inscribed in the given triangle A1A2A3 if it is an acute triangle (necessary
condition).
On the other hand, according to problems 2.3.2 and 2.3.4 this is also a sufficient
condition.
2.3.5. We have that (Figure 2.40)
∠A2 þ ∠A4 ¼ π ∠B2 B1 A2 ∠B1 B2 A2 þ π ∠B4 B3 A4 ∠B3 B4 A4
¼ π ∠B4 B1 A1 ∠B1 B4 A1 þ π ∠B3 B2 A3 ∠B2 B3 A3 ¼ ∠A1 þ ∠A3 ,
Figure 2.41 A2
B2
A3
B1
O
B3
А1
B4
A4
Since points B1 and B4 are on the circle with diameter A1O, then
∠B4B1A1 ¼ ∠B4OA1. We have ∠B4 OA1 ¼ π2 ∠B4 A1 O ¼ π2 ∠B2 A2 O ¼
∠B2 OA2 .
Since points B1 and B2 are on the circle with diameter A2O, then
∠B2OA2 ¼ ∠B2B1A2. Thus ∠B4B1A1 ¼ ∠B2B1A2. The other three inequalities
one can prove similarly.
Remark
1. According to problems 2.3.1 and 2.3.5, a quadrilateral with a minimum perim-
eter can be inscribed into the given quadrilateral A1A2A3A4 if the latter is
inscribed and condition (2.6) of problem 2.3.5 is satisfied (necessary condition),
and according to problems 2.3.2 and 2.3.6 these are also sufficient conditions.
2. If a quadrilateral with a minimum perimeter can be inscribed into the given
quadrilateral A1A2A3A4 then the number of these would be infinite. According to
remark 1 to problem 2.3.2, corresponding sides of all minimum possible perim-
eter quadrilaterals inscribed into A1A2A3A4 quadrilateral will be parallel to each
other (Figure 2.42).
2.3.7. Let M be the midpoint of edge A1A3. Denote points symmetric to B1, B2, B3,
B4 with respect to plane A2MA4 by B01 , B02 , B03 , B04 , respectively, and the midpoints of
the segments B03 B1 , B2 B02 , B3 B01 , B4 B04 by C1, C2, C3, C4,respectively.
Note that points C1 and C3 are on facets A2A3A4 and A1A2A4 respectively, and
points C2 and C4 are on segments MA4 and MA2, respectively
According to problem 1.1.9a (see the solution) we have that
B 1 B2 þ B 2 B3 B 2 B3 þ B1 B2 B3 B4 þ B4 B1 B1 B 4 þ B3 B 4
¼ þ þ þ
2 2 2 2
¼ B 1 B2 þ B 2 B3 þ B 3 B4 þ B4 B1 ,
Figure 2.42 A2
B1
B2
A1
A3
B4
B3
A4
D1 D2 þ D2 D3 þ D3 D4 þ D4 D1
C1 C2 þ C04 C01 C2 C3 þ C03 C04 C3 C4 þ C02 C03 C1 C4 þ C01 C02
þ þ þ
2 2 2 2
¼ C1 C2 þ C2 C3 þ C3 C4 þ C4 C1 :
Thus D1D2 þ D2D3 þ D3D4 þ D4D1 C1C2 þ C2C3 þ C3C4 þ C4C1, which
means that D1D2 þ D2D3 þ D3D4 þ D4D1 C1C2 þ C2C3 þ C3C4 þ C4C1
B1B2 þ B2B3 þ B3B4 þ B4B1, consequently,
B1 B2 þ B2 B3 þ B3 B4 þ B4 B1 D1 D2 þ D2 D3 þ D3 D4 þ D4 D1
ρðNA3 ; MA4 Þ þ ρðMA4 ; NA1 Þ þ ρðNA1 ; MA2 Þ þ ρðMA2 ; NA3 Þ,
where ρ(l1, l2) is the distance between straight lines l1 and l2.
We shall prove that one can choose on segments MA4 and NA3 points K and E,
respectively, so that KE ⊥ MA4, KE ⊥ NA3, then ρ(NA3, MA4) ¼ KE (Figure 2.43).
! ! !
Denote A3 A ¼ ~
1 a, A3 A ¼ ~
4 b, A3 A2 ¼ ~
c, A3 E ¼ λ, A4 K ¼ μ.
A3 N A4 M
We have that
! ¼ ! ! !
EA 3 þ A3 M þ MK ¼
EK
! ~ a ! λ
~
a ~
a
¼ λA3 N þ þ ð1 μÞMA 4 ¼¼ ~ c þ þ ð1 μ Þ ~
b þ~ b ¼
2 2 2 2
1
Figure 2.43 A4
A2
K b
A1
M c
a
A3
! ~
!
Since aj ¼ ~
j~ b ¼ j~ a~
cj ¼ 1,~b ¼~ c ¼~
b~ a ¼ 12
c~ and EK ⊥ b þ~ c , EK ⊥
2~
b ~a , then μ~ a þ ð2 λ 2μÞ~ c ~
b λ~ b þ~ c ¼ 0 and ðμ~
a þ ð2 λ 2μÞ
~ cÞ 2~
b λ~ b ~a ¼ 0, thus 3λ 2μ þ 3 ¼ 0 and 2λ 3μ þ 3 ¼ 0. Hence, we
obtain that λ ¼ μ ¼ 35.
r ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2ffi
! pffiffiffiffiffi
Then ρðNA3 ; MA4 Þ ¼ EK ¼ EK ¼ 0, 1 3~ ~
a þ b 3~ c ¼ 0, 1 10. Simi-
pffiffiffiffiffi
larly we get that ρðMA4 ; NA1 Þ ¼ ρðNA1 ; MA2 Þ ¼ ρðMA2 ; NA3 Þ ¼ 0, 1 10, thus
pffiffiffiffiffi
B1 B2 þ B2 B3 þ B3 B4 þ B4 B1 0, 4 10.
Another solution of this problem can be obtained by using a problem 4.1.24b.
Remark If points B1, B2, B3, B4 are on segments A3N, A4M, A1N, A2M, respectively,
pffiffiffiffiffi
and AA33BN1 ¼ AA44BM2 ¼ AA11BN3 ¼ AA22BM4 ¼ 35, then B1 B2 þ B2 B3 þ B3 B4 þ B4 B1 ¼ 0, 4 10.
2.3.8. Consider Cartesian coordinate system in the space and let B(0; 0; 0), A
(1; 0; 0), C(0; 1; 0), B0 (0; 0; 1), B1(x; y; 0).
Denote the points symmetric to B1, D1, D2, D3, D4 with respect to planes AA0 B0 B,
BB0 C0 C, A0 B0 C0 D0 , DD0 C0 C, AA0 D0 D by D1, D2, D3, D4, and D5, respectively. Then
D1(x; y; 0), D2(x; y; 0), D3(x; y; 2), D4(x; 2 þ y; 2), D5(2 þ x; 2 þ y; 2).
Note that
B1 B2 þ B2 B3 þ B3 B4 þ B4 B5 þ B5 B6 þ B6 B1 ¼
¼ D 1 B2 þ B2 B3 þ B3 B4 þ B4 B5 þ B5 B6 þ B6 B1
D 1 B3 þ B3 B4 þ B4 B5 þ B5 B6 þ B6 B1 ¼ D 2 B3 þ B3 B4 þ B4 B5 þ B5 B6 þ B6 B1
D 2 B4 þ B4 B5 þ B5 B6 þ B6 B1 ¼ D 3 B4 þ B4 B5 þ B5 B6 þ B6 B1
D 3 B5 þ B5 B6 þ B6 B1 ¼ D 4 B5 þ B5 B6 þ B6 B1 D 4 B6 þ B6 B1 ¼
¼ D 5 B6 þ B6 B1 D 5 B1 ,
74 2 Application of Projection Method
pffiffiffi
consequently B1 B2 þ B2 B3 þ B3 B4 þ B4 B5 þ B5 B6 þ B6 B1 D5 B1 ¼ 2 3.
(See also problem 4.1.24c.)
2.3.9. Given an inscribed quadrilateral ABCD. Prove that the perimeter of the
quadrilateral inscribed into ABCD is greater than or equal to 2AC sin ∠A.
2.3.10. Given straight lines l, m, and n. Find the triangle with the minimum possible
perimeter and the vertices on lines l, m, and n.
2.3.11. Given an inscribed quadrilateral ABCD. Find on lines AB, BC, CD, DA, such
points B1, B2, B3, B4, respectively, for which the sum B1B2 þ B2B3 þ B3B4 þ B4B1
would be minimal.
2.4.1. Prove that radius of the incircle of the right-angled triangle is less than half of
its leg.
2.4.2. Several circles, the sum of lengths of which is equal to 10, are located inside a
unit square. Prove that one can find a straight line crossing at least four of these circles.
2.4.3. Non-self-crossing broken line of length 1000 is located inside a unit square.
Prove that that one can find a straight line parallel to one of the sides of a square
which would cross at least 500 segments of this broken line.
2.4.4. Given two straight lines a and b. A1B1, A2B2, A3B3 are perpendiculars drawn
from three consecutive points A1, A2, A3 on line a to line b.
Prove that A2B2 max (A1B1, A3B3).
2.4.5. Diagonals of a convex quadrilateral ABCD intersect at point O. Let P and Q be
the centers of circumcircles of triangles AOB and COD. Prove that PQ > ABþCD 4 .
2.4.6. Points A1, B1, and C1 are taken on sides BC, CA, and AB of non-obtuse
triangle ABC, respectively. Prove that 2(B1C1 cos α þ C1A1 cos β þ A1B1 cos γ)
BC cos α þ AC cos β þ AB cos γ, where ∠A ¼ α, ∠B ¼ β, ∠C ¼ γ. Give an example
of an obtuse triangle for which the inequality does not hold.
2.4.7. Let circles of unit radiuses have no common internal points and be inside of a
band S. The band is formed by two parallel lines having a distance w. Let us call
these circles a k-cloud, if each line that intersects S, intersects also more than or
pffiffiffi
equal to k circles. Prove that for k-cloud w 2 þ 3 2k , where k 2 N and [a] is the
integer part of the number a.
2.4 Method of Projections 75
2.4.8. Let ON be the radius of the circle with a center at point O, which intersects
chord AB at point M at a right angle. Let P be an arbitrary point on the larger arc AB,
not coinciding with the point diametrically opposite to point N. Straight lines PM
and PN define points Q and R on the circumference and on chord AB, respectively.
Prove that RN > MQ.
2.4.9. In convex pentagon ABCDE, side AB is perpendicular to side CD, and side BC
is perpendicular to side DE. Prove that, if AB ¼ AE ¼ ED ¼ 1, then BC þ CD < 1.
2.4.10. (a) Opposite sides of convex hexagon ABCDEF are parallel (AB||DE, BC||
EF, CD||FA). Prove that RA þ RC þ RE P, where RA, RC, RE are the radiuses of the
circumcircles of triangles FAB, BCD, and DEF, respectively, and p is the half-
perimeter of hexagon ABCDEF.
(b) Let M be an arbitrary point inside triangle ABC, Ra, Rb, Rc the distances of point
M from A, B, C, da, db, dc the distances from point M to lines BC, AC, and AB,
respectively. Prove that Ra þ Rb þ Rc 2da þ 2db þ 2dc.
(c) Let M be a point inside triangle ABC. Prove that one of the angles
∠MAB, ∠MBC, ∠MCA is less than or equal to 30 .
2.4.11. Prove that for an acute triangle ha þ hb þ hc 4R þ r.
2.4.12. Given points U and V on sides AB and CD of square ABCD, respectively.
Let straight lines DU and AV intersect at point P, and lines CU and BV at point Q.
Prove that PQ 12 AB.
2.4.13. Given 110 unit vectors on a plane, the sum of these being a zero vector.
Prove that of these 110 vectors one can choose such 55 vectors, that the modulus of
the sum of which will not be greater than 1.
2.4.14. Let the diagonals of convex quadrilateral ABCD intersect at point P. Points
Q, R, S, and T are the feet of the perpendiculars drawn from point P to lines AB, BC,
CD, and DA, respectively. Prove that
1
PQ þ PR þ PS þ PT ðAB þ BC þ CD þ DAÞ:
2
2.4.15. Let ABCDEF be a convex hexagon. Given that AB ¼ CD ¼ EF. Prove that
AD þ BE þ CF AB þ BC þ CD þ DE þ EF þ FA. For which hexagon does equal-
ity hold true?
2.4.16. The vertices of tetrahedron KLMN lay either (inside) on the facets or on the
edges of another tetrahedron ABCD. Prove that the sum of the lengths of all edges of
tetrahedron KLMN is less than 43 of the sum of the lengths of all edges of
tetrahedron ABCD.
2.4.17. Let M be a point inside a convex n-gon A1A2 . . . An and p be half-perimeter
∠An
of that n-gon. Prove that MA1 cos ∠A ∠A2
2 þ MA2 cos 2 þ ::: þ MAn cos 2 p.
1
2.4.18. (a) Points M, N, P are chosen on edges A1B1, C1C and AD of unit cube
ABCDA1B1C1D1, respectively. Prove that the perimeter of triangle MNP is not less
pffiffiffi
than 3 6=2.
76 2 Application of Projection Method
(b) Given a unit cube ABCDA1B1C1D1. Prove that the distance from the arbitrary
point in the space to one of straight lines A1B1, C1C, AD is not less than p1ffiffi2.
(c) Given two spheres with diameters d and D inside a unit cube so that they do not
pffiffiffi
have common points. Prove that d þ D < 3 3.
Solutions
Figure 2.44 A B
A1 A2 Am
2.4 Method of Projections 77
Figure 2.45
bi ci bi
ai
ai
B A2¢
A3¢
If the straight lines a and b are not in the same plane then we shall consider the
projections of points A1, A2, A3, B1, B2, B3 on the plane perpendicular to line b. Let
those points are A01 , A02 , A03 , and B (Figure 2.46).
Since max ∠BA0 2 A03 ; ∠BA0 2 A01 90 , hence it follows that max(BA0 1, BA0 3) >
BA0 2 and we have that BA0 1 ¼ B1A1, BA0 2 ¼ B2A2, BA0 3 ¼ B3A3. Therefore
A2B2 < max (A1B1, A3B3).
Remark Given segments AB and CD, then inequality SABM max (SABC, SABD)
holds true for any point M of segment CD.
2.4.5. Note that the lengths of the projections of segment PQ on lines AC and BD
2 and 2 , respectively. Thus PQ þ PQ > 2 þ 2 , consequently, PQ
are equal to AC BD AC BD
2.4.6. Let A2C2 be the projection of segment A1C1 on side AC. Since β π2 and
A1C1 A2C2, therefore we deduce that 2A1C1 cos β 2A2C2 cos β ¼ 2 cos β
(AC AC1 cos α CA1 cos γ).
Similarly we get that 2A1B1 cos γ 2 cos γ(AB AB1 cos α BA1 cos β) and
2B1C1 cos α 2 cos α(BC BC1 cos β CB1 cos γ).
78 2 Application of Projection Method
2ðB1 C1 cos α þ C1 A1 cos β1 þ A1 B1 cos γ Þ 2AC cos β þ 2AB cos γ þ 2BC cos α
2cos α cos βðAC1 þ C1 BÞ 2cos β cos γ ðBA1 þ A1 CÞ 2 cos γ cos αðAB1 þ B1 CÞ
¼ 2BC cos α þ 2AC cos β þ 2AB cos γ cos αðAB cos β þ AC cos γ Þ
cos βðAB cos α þ BC cos γ Þ cos γ ðAC cos α þ BC cos βÞ ¼
¼ 2BC cos α þ 2AC cos β þ 2AB cos γ BC cos α AC cos β AB cos γ ¼
¼ BC cos α þ AC cos β þ AB cos γ:
Hence, it follows that 2(B1C1 cos α þ C1A1 cos β1 þ A1B1 cos γ) BC cos α þ
AC cos β þ AB cos γ.
For α ¼ γ ¼ π6, AC1 ¼ AB 3 ¼ 3, and C1B1 k CB, B1A1 k AB we have that C1 B1 ¼ 3,
a a
pffiffi
A1 B1 ¼ 23 a , A1 C1 ¼ 37 a. Thus
2.4.7. Draw through the center of a certain circle of the k-cloud a straight line l,
which is perpendicular to the boundaries of band S. Then that line l should intersect
not less than k 1 other circles. Line l divides the plane into two half-planes. It is
not difficult to understand that one of the half-planes contains not less than 2k þ 1
centers of those circles (including kthe centers which are on line l ). Let O1, O2, . . . ,
On be these centers, where n ¼ 2 þ 1 (Figure 2.47).
Let points O01 , O02 , :::, O0n be the projections of points O1, O2, . . . , On on line l.
Denote by d i ¼ Oi O0i , i ¼ 1, . . . , n. Note that for i ¼ 1, . . . , n 1 we have that
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffi pffiffiffi
O0i O0iþ1 ¼ Oi O2iþ1 ðdi diþ1 Þ2 4 ðd i diþ1 Þ2 4 1 ¼ 3,
pffiffiffi pffiffiffi
since 0 di 1, i ¼ 1, . . . , n. Consequently O01 O0n 3ðn 1Þ ¼ 3 2k , where
O01 and O0n are, respectively, the lowest and the highest points on line l.
Figure 2.47 l
On¢ On
O¢j O j
O¢2 O2
O¢1 O1
2.4 Method of Projections 79
Figure 2.48 L
K
A T M R B
Q N
Since band S should extend, at least by the radius of the circle, on each side of the
ends of segment O01 O0n , to contain the circles with centers O1 and On, then it should
pffiffiffi
have a width w 2 þ 3 2k .
2.4.8. Consider the figure below (Figure 2.48).
Segments MR and KM are projections of segments NR and QM on lines AB and
KM, respectively. We have that ∠QMK ¼ π2 ∠KQM ¼ π2 ∠MNR ¼ ∠MRN.
Therefore, it will be sufficient to prove that MR > MK. Note that
_ _ _ _
∠BRP ¼ BPþAN
2 ¼ BPþBN
2 ¼ ∠MLP.
This means that MLPR is an inscribe quadrilateral. Since ∠MLR ¼
∠MPR ¼ ∠TLM, then MR ¼ MT > MK.
2.4.9. Let ∠CDB ¼ α, ∠CBD ¼ β, α β, and A0 E0 is the projection of segment AE
on line BD. We have that AE A0 E0 ¼ BD þ sin α þ sin β, consequently
Hence
sin ðα þ βÞ
BD < : ð2:8Þ
sin α þ sin β
BD sin α BD sin β
Since BC þ CD ¼ sin ðαþβÞ þ sin ðαþβÞ, then according to (2.8) we get
BC þ CD < 1.
2.4.10. (a) Consider Figure 2.49.
We have that BF MN and BF PK. Thus, we deduce that BF MNþPK
2 or
2RA sin α ða sin βþf sin γÞþ2 ðc sin γþd sin βÞ. Therefore, it follows that
80 2 Application of Projection Method
Figure 2.49 M B b C P
b g
a c
A a
D
f a
d
g b
N F e E K
1 sin β 1 sin γ
RA ða þ dÞ þ ðf þ cÞ : Similarly we get that
4 sin α 4 sin α
1 sin α 1 sin β
RC ðf þ cÞ þ ðb þ eÞ ,
4 sin γ 4 sin γ
1 sin γ 1 sin α
RE ðb þ eÞ þ ða þ dÞ :
4 sin β 4 sin β
Summing up these inequalities and making use of inequality x þ 1x 2ðx > 0Þ,
we obtain that RA þ RC þ RE 12 ða þ dÞ þ 12 ðb þ eÞ þ 12 ðf þ cÞ ¼ p. This ends
the proof.
Remark Equality holds if and only if hexagon ABCDEF is regular.
(b) Draw perpendiculars MB1, MD1, MF1 to lines BC, AC, AB, respectively, and
construct parallelograms B1MF1A1, D1MB1C1, F1MD1E1.
Since the radiuses of the circumcircles of triangles D1MF1, B1MF1 B1MD1 are
equal to R2a , R2b , R2c , with R2a ¼ RE1 , R2b ¼ RA1 , R2c ¼ RC1 , then according to problem
2.4.10a we have that R2a þ R2b þ R2c p ¼ da þ db þ dc . Thus
Ra þ Rb þ Rc 2da þ 2db þ 2d c :
Figure 2.50 B
A1
C1
M
A B1 C
Figure 2.51 B
A2
C1 A1
kc ka
O
kb
A B1 C
2.4.11. Let O be the center of the circumcircle of acute triangle ABC and ka, kb, kc be
the distances from point O to its sides (Figure 2.51).
Lemma Prove that ka þ kb þ kc ¼ R þ r.
Let A1, B1, and C1 be the midpoints of sides BC, CA, and AB, respectively. By
Ptolemy’s theorem akc þ cka ¼ bR akb þ bka ¼ cR, ckb þ bkc ¼ aR. On the other
hand, aka þ bkb þ ckc ¼ 2S ¼ (a þ b þ c)r. By adding these equalities and reducing
a þ b þ c, we obtain that ka þ kb þ kc ¼ R þ r. This ends the proof of the lemma.
Projection of broken line AOA1 on straight line AA2 is equal to ha, where
AA2 ⊥ BC, thus R þ ka ha. Similarly we get that R þ kb hb, R þ kc hc.
By summing up these inequalities we get according to lemma that
4R þ r ¼ 3R þ ka þ kb þ kc ha þ hb þ hc (see problem 5.5.8b).
2.4.12. Let BU CV, then AU DV. Draw a median MN of the square (Figure 2.52)
and P1P2||Q1Q2||AB.
Projections of segments PP1 and QQ1 on side AB do not have common internal
points, thus
P1 P2 Q1 Q2 KN KM
PQ AB ðPP1 þ QQ1 Þ ¼ AB þ AB þ
2 2 2 2
1
¼ AB:
2
82 2 Application of Projection Method
Figure 2.52
Figure 2.53 x
a1 S - r = a1 + ... + an
an
a2
Figure 2.54 B
B1
A2
A1
A P C
C2
D1 C1
2.4.14. We shall make use of the following fact: if XX1 is a bisector of angle X of
triangle XYZ and ∠XX1Z 90 , then the midpoint of side YZ belongs to segment
X1Z. Indeed, since ∠XX1Z 90 , then ∠Y ∠Z.consequently,
YX1 XY
¼ 1:
X1 Z XZ
Let straight line A1C1 contain the bisector of angle APB and points A2, C2 be,
respectively, the midpoints of sides AB, CD (see Figure 2.54).
Let also points A02 , C02 be the projections of points A2, C2 on line A1C1.
We have that A1 C1 A02 C02 , thus it follows that A02 C02 A2 C2 ADþBC 2 (see
problem 1.1.9a), consequently A1 C1 ADþBC 2 . Similarly we get that B D
1 1 ABþCD
2 .
Summing up these inequalities we obtain that
ABþBCþCDþDA
2 A1 C1 þ B1 D1 ¼ A1 P þ B1 P þ C1 P þ D1 P PQ þ PRþ
PS þ PT, which means that ABþBCþCDþDA 2 PQ þ PR þ PS þ PT.
Remark The equality in the last inequality holds true if and only if quadrilateral
ABCD is a rectangular.
2.4.15. Let points B0 , C0 be projections of points B, C on line AD. Then AD ¼
AB cos ∠BAD þ B0 C0 þ CD cos ∠ADC AB cos ∠BAD þ BC þ CD cos ∠ADC.
Similarly we obtain that BE AB cos ∠ABE þ AF þ EF cos ∠FEB and
CF CD cos ∠FCD þ ED þ FE cos ∠CFE.
Summing up these two inequalities we deduce that
AD þ BE þ CF ABð cos ∠BAD þ cos ∠ABEÞ þ CDð cos ∠ADC þ cos ∠FCDÞþ
∠BAD þ ∠ABE
þ EFð cos ∠FEB þ cos ∠CFEÞ þ BC þ DE þ AF 2AB cos
2
∠ADC þ ∠FCD ∠FEB þ ∠CFE
þ 2CD cos þ 2EF cos þ BC þ DE þ AE
2 2
3AB þ BC þ DE þ AE ¼ AB þ BC þ CD þ DE þ EF þ FA:
84 2 Application of Projection Method
Figure 2.55 N C D P
M A F K
AD þ BE þ CF > AB þ BC þ CD þ DE þ EF þ FA:
1
KL þ LM þ KM þ KN þ LN þ MN ¼ KL þ LM þ KM þ ðKN þ LN þ KLÞþ
2
þ ðKN þ MN þ KMÞ þ ðLN þ MN þ LMÞ
4ðKL þ LM þ KMÞ
¼ 2ðKL þ LM þ KMÞ:
2
Project tetrahedron ABCD on the plane of facet KLM. Denote the projections of
vertices A, B, C, D of the tetrahedron by A1, B1, C1, D1, respectively. Let Γ be the
convex envelope of points A1, B1, C1, D1 and PΓ be the perimeter of polygon Γ.
Since triangle KLM is inside Γ, then according to problem 2.1.1 we have
KL þ LM þ KM PΓ. Thus, it follows that
2.4 Method of Projections 85
Figure 2.56 B1 B1 C1
D1
A1
A1 C1
D1
a b
PΓ ¼ A1 B1 þ B1 C1 þ C1 D1 þ D1 A1 ¼
2 1
¼ ðA1 B1 þ B1 C1 þ C1 D1 þ D1 A1 Þ þ ðA1 B1 þ B1 C1 þ C1 D1 þ D1 A1 Þ <
3 3
2 2
< ðA1 B1 þ B1 C1 þ C1 D1 þ D1 A1 Þ þ ðA1 C1 þ B1 D1 Þ
3 3
KL þ KM þ KN þ ML þ NL þ MN 4 n 2n
> : ð2:9Þ
AB þ AC þ AD þ BC þ BD þ CD 3n þ 3
4ðn2Þ
Since at n ! 1 we have that 3nþ3n ! 43, then it is clear that the value 4
3 in
inequality 2.4.16 cannot be lowered.
2.4.17. Let ∠MAiAi þ 1 ¼ βi, ∠MAiAi 1 ¼ γ i, ∠Ai ¼ αi, i ¼ 1, . . . , n, An þ 1 Ai,
A0 An, it is clear that αi ¼ βi þ γ i. Note that
X
n X
n
βi þ γ i β γi
2p ¼ MAi ð cos βi þ cos γ i Þ ¼ 2MAi cos cos i
i¼1 i¼1
2 2
X
n
αi
2MAi cos ,
i¼1
2
∠A1 ∠An
consequently, MA1 cos 2 þ ::: þ MAn cos 2 p.
2.4.18. (a) Let M0, N0, and P0 be the midpoints of edges A1B1, C1C, and DA,
pffiffi
respectively. From right-angled triangles AA1P0 and A1M0P0 we have A1 P0 ¼ 25,
pffiffi
M0 P0 ¼ 26, and cos ∠A1 M0 P0 ¼ p1ffiffi6. Similarly, we get that the cosines of angles
∠B1M0N0, ∠C1N0M0, ∠CN0P0, ∠DP0N0, and ∠AP0M0 are equal to p1ffiffi. 6
Let point X0 be the projection of point X on line M0P0. Then, we have that
B01 M0 þ M0 P0 þ P0 D0 ¼ B01 M0 þ M0 P0 þ P0 D0 , M0 P0
1 1
¼ M0 P0 þ pffiffiffi ðB1 M þ PDÞpffiffiffi :
6 6
1 1
MP M0 P0 þ pffiffiffi ðB1 M þ PDÞpffiffiffi : ð2:10Þ
6 6
1 1
MN M0 N 0 þ pffiffiffi ðA1 M þ CN Þpffiffiffi , ð2:11Þ
6 6
and
1 1
PN P0 N 0 þ pffiffiffi ðAP þ NC1 Þpffiffiffi : ð2:12Þ
6 6
2.4 Method of Projections 87
(b) Let O be the center of the cube and M be an arbitrary point of the space.
Let’s consider projections of point M and straight lines A1B1, C1C, AD on plane
M0N0P0 (see notations in the solution of problem 2.4.18a).
Denote the projection of point X on plane M0N0P0 by X0 .
Since OM0 ⊥ A1B1, we have OM0 ⊥A01 B01 . Let A01 B01 \ A0 D0 ¼ A2 ,
0 0 0 0 0 0 0 0
A1 B1 \ C1 C ¼ B2 , A D \ C1 C ¼ C2 , then A2B2C2 is a regular triangle with
center O.
We have that SM0 A2 B2 þ SM0 B2 C2 þ SM0 A2 C2 SA2 B2 C2 , consequently
M0 M0 2 A2 B2 M0 N 0 2 B2 C2 M0 P0 2 A2 C2 1 A2 B2 þ B2 C2 þ C2 A2
þ þ pffiffiffi ,
2 2 2 2 2
thus maxðM0 M0 2 ; M0 N 0 2 ; M0 P0 2 Þ p1ffiffi2, where MP2 ⊥ AD, MM2 ⊥ A1B1, MN2 ⊥ C1C,
and P2 2 AD, M2 2 A1B1, N2 2 C1C, which means that
0 1
maxðMM2 ; MN 2 ; MP2 Þ max M0 M0 2 ; M0 N 0 2 ; M0 P2 pffiffiffi :
2
(c) Let O1 and O2 be the centers of these spheres and projections of segment O1O2
on the edges emerging from the same vertex are equal to a, b, and c. We have
that a2 þ b2 þ c2 ¼ O1O22. Let a b, a c, in that case a Op1 Offiffi3 2 . Note
(Figure 2.57) that d2 þ a þ D2 1.
Consequently
pffiffiffi
1 Op1 Offiffi3 2 þ d2 þ D2 > dþD
pffiffi þ dþD. Hence, we deduce that d þ D < 3
2 3 2 3.
Figure 2.57
D
О2
2
а
О1
d
2
88 2 Application of Projection Method
2.4.19. Prove that the distance from one of the vertices of the convex quadrilateral
to the opposite diagonal does not exceed the half of that diagonal.
2.4.20. Let parallelogram P2 be inscribed in parallelogram P1 and parallelogram P3
be inscribed in parallelogram P3, such that the sides of P3 are parallel to the sides of
P1. Prove that the length of at least one of the sides of P1 does not exceed the double
length of the corresponding parallel side of P3.
2.4.21. Prove that inside a convex n-gon (n 7) one can find a point, the sum of the
distances from which to the vertices are greater than the perimeter.
2.4.22. Given a unit square and such a broken line, that inside it every line parallel
to the side of the square intersects it in no more than one point. Prove that the length
of the broken line is less than 2.
2.4.23. (a) Let us consider on a plane a finite set of segments, the sum of lengths of
pffiffiffi
which is less than 2. Prove that there exists an infinite net of unit squares, the sides
of which do not intersect with any of these segments.
Remark The statement of the problem holds true also if the number of segments is
not finite.
(b) A figure on the coordinate plane has an area S, where S > 1. Prove that it can be
translated by a vector with the integer number of coordinates, so that the figure
and its image do not intersect.
2.4.24. Let ABCDEF be a convex hexagon, such that AB|| ED, BC||EF, CD||AF, and
AB þ DE ¼ BC þ EF ¼ CD þ AF. Prove that
AD þ BE þ CF AB þ BC þ CD þ DE þ EF þ AF:
BC DE BC AF AF DE
AD þ BE þ CF AB þ CD þ EF þ þ þ :
AF DE BC
2.4.29. Let p be the sum of the lengths of the edges of convex polyhedron and d be
the maximal distance between its vertices. Prove that p > 3d.
Hint Consider the projection of the polyhedron on straight line AB, where A and
B are the vertices of the polyhedron and AB ¼ d.
2.4.30. The closed broken line passes on a surface of a unit cube and has common
points with all its facets. Prove that the length of the broken line is not less than
pffiffiffi
3 2.
Hint Consider the projections of the broken line on three mutually non-parallel
edges of the cube.
2.4.31. A number of segments is located inside a unit cube, so that any plane
parallel to one of the facets of the cube crosses not more than one of them. Prove
that the sum of the lengths of these segments is not greater than 3.
Hint Consider projections of these segments on three edges of a cube with
common vertex.
2.4.32. A number of polygons is located inside a unit cube, so that any line parallel
to one of the edges of the cube intersects with not more than one of them. Prove that
the sum of the areas of these polygons is not greater than 3.
Hint Consider projections of polygons on three facets of a cube with common
vertex.
2.4.33. Prove that for any tetrahedron one can find two such planes that the ratio of
pffiffiffi
the areas of the projections of the tetrahedron on these planes is not less than 2.
Hint For the tetrahedron SABC let MN be the common perpendicular of straight
lines SA and BC. Consider planes α, β, γ, passing through line MN, where
α ⊥ SA, β ⊥ BC, such that the angles between γ and straight lines SA, BC are equal.
90 2 Application of Projection Method
Hint Consider the projections of segments ICIA, IAIB, and IBIC on the straight lines
AC, AB, and BC, respectively. Then, one can easily deduce that
IAIB þ IBIC þ ICIA A1B1 þ B1C1 þ A1C1. Prove that GC GA 13 ðA1 C1 þ ACÞ and
GAGB þ GBGC þ GCGA A1B1 þ B1C1 þ A1C1.
2.4.43. Prove the inequality cos α Rr for a triangle, where α ¼ max (α, β, γ).
Hint Prove that cosβ þ cos γ 1.
Chapter 3
Areas
This chapter is devoted to the inequalities related to areas and it consists of only one
paragraph, that is Section 3.1. One of the methods for proving the inequalities
related to areas (of some figures on the plane) is the following: if the figures with
areas S1 , S2 , . . . , Sk cover a figure with area S, then S1 þ S2 þ . . . þ Sk S.
Let us give a simple example. Prove that the area of parallelogram AMNK
(M 2 AB, N 2 BC, K 2 AC) inscribed to triangle ABC is not greater than the half of
the area of ABC. Consider parallelograms BNKE and NCKF; note that triangles
AEK and KNF are, respectively, equal to triangles MBN and KNC. Moreover,
triangles AEK and KNF cover parallelogram AMNK; therefore the sum of the
areas of triangles MBN and KNCis not smaller than the area of parallelogram
AMNK. This ends the proof of given example.
One of the main methods of proving geometric inequalities related with areas is
rewriting them as algebraic inequalities by introducing some notations.
In order to compare areas of two figures, often one needs to consider consequent
figures, such that the first figure and the last figure are the given figures. Afterwards,
one needs to compare consequently areas of any two of the considered figures.
Some problems in this chapter were inspired by [2, 4, 9, 11, 13–16].
Nevertheless, even for these problems the authors have mostly provided their
own solutions.
3.1.1. (a) In a convex quadrilateral ABCD diagonals intersect at a point O and BC||
AD. Prove that SOCD 14 SABCD .
(b) Given points E and F on parallel sides BC and AD of the convex quadrilateral
ABCD, respectively. Given also that AE, BF intersect at point P and segments
CF, ED at point Q. Prove that SPEQF 14 SABCD .
3.1.12. Let A1, B1, C1 be points on sides BC, CA, AB of a triangle ABC, respectively.
Prove that
(a) SABC S2A1 B1 C1 4SAB1 C1 SBA1 C1 SCA1 B1 ,
(b) minðSAB1 C1 ; SBA1 C1 ; SCA1 B1 Þ SA1 B1 C1 ,
(c) minðSAB1 C1 ; SBA1 C1 ; SCA1 B1 Þ 14 SABC ,
pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi 3 pffiffiffiffiffiffiffiffiffiffi
(d) SAB1 C1 þ SBA1 C1 þ SCA1 B1 2 SABC ,
1 BB1 CC1
(e) SA1 B1 C1 min 14SABC ; 2AAðABþBCþACÞ , if segments AA1, BB1 and CC1 intersect at
one point,
(f) SA1 B1 C1 SAB1 C þ SBA1 C þ SCA1 B 3,
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 1 1 1
ffi 1 1
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
SAB1 C1 SBA1 C1 SCA1 B1 SAB1 C1 SBA1 C1 SCA1 B1
(g) max S S ,
ð AB1 C1 ;SBA1 C1 ;SCA1 B1 Þ A1 B1 C1 minðSAB1 C1 ;SBA1 C1 ;SCA1 B1 Þ
3.1.19. Given a convex polygon in which one cannot place any triangle with area
1. Prove that this polygon can be placed into a triangle with area 4.
3.1.20. Given an acute triangle ABC. Let A1, B1, C1 be the points symmetric to
points A, B, C with respect to straight lines BC, AC, AB, respectively. Prove that
SA1 B1 C1 4SABC .
3.1.21. Given a triangle ABC. Segments BB1, CC1, AA1 are placed on rays AB, BC,
CA starting at points B, C, A, respectively, such that BB1 ¼ AC, CC1 ¼ AB,
AA1 ¼ BC. Prove that SAA1 B þ SBB1 C þ SCC1 A 3SABC .
3.1.22. Three secants are drawn for each internal point X of triangle ABC parallel to
its sides. As a result of this, one obtains three triangles each bounded by two secants
and a side of the triangle. Let the areas of these triangles be S1, S2, S3. Prove that
S1 þ S2 þ S3 13 S.
3.1.23. In a convex hexagon ABCDEF the opposite sides are parallel to each other
(AB||DE, BC||EF, CD||AF). Prove that 2SBDF SABCDEF.
3.1.24. Given points P, Q, R on sides AB, CD, EF of the centrally symmetric convex
hexagon ABCDEF, respectively. Prove that 2SPQR SABCDEF.
3.1.25. The triangle is inscribed into a regular hexagon; one of the sides of the
triangle passes through the center of symmetry of the hexagon. Prove that the area
of the triangle does not exceed 13 of the area of the hexagon.
3.1.26. Points D and E are chosen on the sides AB and AC of the triangle ABC,
respectively. Let segments BE and CD intersect at point P. Prove that
pffiffi
(a) SPDE 5 5211 SABC ,
pffiffiffi
(b) SPDE 5 2 7 SABC , if SBCED ¼ 2SPBC.
3.1.27. The segment PQ passes through the centroid of triangle ABC and points P,
Q are on sides BC, AC, respectively. Prove that SMPQ 29 SABC , where M is the
midpoint of side AB.
3.1.28. Prove that the area of any section of the cube by a plane passing through its
center is not less than the area of the faces of the cube.
3.1.29. The part of a plane between two parallel straight lines is called a “strip”. Let
several strips be given on a plane, and no two of them are parallel, i.e. no two
boundary straight lines of different strips are parallel. How one should move the
strips parallel to themselves to make the area of the polygon formed by the
intersection of the strips the greatest possible?
3.1.30. (a) Given a circle ω with a center A 12; 12 and radius R, where R < 12. Points
M and N are on the positive semiaxes Oy and Ox so that segment MN is tangent to
circle ω. Prove that
3.1 Inequalities with Areas 97
pffiffi 2 pffiffi
(1) SMON 2
2
R , if 0 < R 2
4 ,
pffiffi
(2) SMON 14 R2 , if 42 < R < 12.
pffiffiffi
(b) Prove that any section of the unit cube by a plane has an area not greater than 2.
3.1.31. (a) Given points B2, . . . , Bn and C2, . . . , Cn on the sides of triangle B1AC1
(B2, . . . , Bn 2 AB1, C2, . . . , Cn 2 AC1). The ray with a vertex at A intersects seg-
ments B1C1, B2C2, . . . , BnCn at points D1, D2, . . . , Dn, respectively. Prove that
AD1 þ AD2 þ ::: þ ADn maxðAB1 þ AB2 þ ::: þ ABn ; AC1 þ AC2 þ ::: þ ACn Þ:
(b) Prove that in a convex quadrilateral the length of any segment with ends on the
sides of the quadrilateral, which passes through the point of intersection of
diagonals do not exceed the length of one of the diagonals.
(c) Prove that the area of any cross section of the tetrahedron by a plane does not
exceed the area of one of its faces.
3.1.32. Given a convex quadrilateral ABCD. Prove that
(a) CD2 SABC þ BC2 SACD > AC2 SBCD, if ∠A þ ∠ C > 180 ,
(b) CD2 SABC þ BC2 SACD ¼ AC2 SBCD, if ∠A þ ∠ C ¼ 180 ,
(c) CD2 SABC þ BC2 SACD < AC2 SBCD, if ∠A þ ∠ C < 180 .
3.1.33. In a convex hexagon ABCDEF opposite sides are parallel (AB||DE, BC||EF,
CD||AF). Denote by A1, B1, C1, D1, E1, and F1 the midpoints of sides AB, BC, CD,
DE, EF and FA, respectively. Prove that
(a) segments A1D1, B1E1 and C1F1 can be sides of a triangle.
(b) 12 SBDF < S1 SBDF , where S1 is the area of the triangle with sides having
lengths A1D1, B1E1 and C1F1.
3.1.34. Given that the rectangle with sides a and b is inside the rectangle with sides
c and d, such that max(a, b) > max (c, d ). Prove that 2ab < cd.
3.1.35. O is an internal point of the convex quadrilateral ABCD with area S. Let
points K, L, M and N be on segments AB, BC, CD and DA, respectively. Prove that if
OKBL and OMDN are parallelograms, then
pffiffiffi pffiffiffiffiffi pffiffiffiffiffi
(a) S S1 þ S2 :
pffiffiffi pffiffiffiffiffi pffiffiffiffiffi
(b) 1, 25 S > T 1 þ T 2 :
pffiffiffi pffiffiffiffiffi pffiffiffiffiffi
(c) C0 S T 1 þ T 2 ,
where S1, S2, T1, T2 are the areas of the quadrilaterals ONAK, OLCM, OKBL,
OMDN, respectively and
sin 2α þ π4
C0 ¼ max :
½0;π4 cos α
98 3 Areas
3.1.36. Let A1, B1, C1, D1 be the midpoints of sides BC, CD, DA, AB of a convex
quadrilateral ABCD, respectively, S0 be the area of the quadrilateral, formed by the
segments AA1, BB1, CC1, DD1, and S be the area of the quadrilateral ABCD. Prove
that 16 S < S0 15 S.
3.1.37. Consider a triangle with the area S. Prove that if one places externally
equilateral triangles on each side of the given triangle, then centers of these tri-
angles are vertices of an equilateral triangle with the area not less than S.
3.1.38. Given on a plane mutually nonintersecting triangles all obtained from
triangle ABC using a translation by a certain vector. All these triangles are contin-
ued to become parallelograms, such that any of these parallelograms has a diagonal
parallel to AB. Prove that S0 1, 5S, where S0 is the area of the union of these
parallelograms and S is the sum of the areas of all triangles.
3.1.39. Given that in a convex hexagon ABCDEF triangles ACE and BDF are
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
similar. Prove that SABCDEF Rr SACE SBDF , where r and R are, respectively,
inradius and circumradius of triangle ACE.
3.1.40. Given on a plane five points, so that the area of each of 10 triangles defined
by these points is greater than 2. Prove that among those triangles exists a triangle,
pffiffiffi
such that its area is greater than 1 þ 5.
3.1.41. The vertices of a convex hexagon A1A2A3A4A5A6 are on the sides of the unit
square. Prove that the area of one of triangles A1A2A3, A2A3A4, A3A4A5, A4A5A6,
A5A6A1, A6A1A2 is not greater than 18.
3.1.42. Given on a plane a triangle ABC and a point P, such that ∠A ¼ 600, PA ¼ 1,
pffiffiffiffiffi pffiffi
PB ¼ 2, PC ¼ 3. Prove that SABC 13 þ 73 83.
3.1.43. Given on a plane a triangle ABC and a point P, such that ∠A ¼ α π2,
PA ¼ r1, PB ¼ r2, PC ¼ r3, with 0 < r1 < r2 < r3, where r1, r2, r3, α are constants.
Find the possible highest and lowest values of the area of triangle ABC.
3.1.44. Consider two parallelograms that intersect exactly in eight points. Prove that
the common area of these parallelograms is greater than or equal to half of the area
of one of them.
Solutions
3.1.1. (a) Consider the midpoints of segments AC, CD and BD and let those points
be M, N and P, respectively. We have that MN||AD and NP||BC and also
SOCD SMCDP ¼ SMCN þ SPND ¼ 14 SACD þ 14 SBCD ¼ 14 SACD þ 14 SABC ¼ 14 SABCD .
Consequently, SOCD 14 SABCD .
(b) Making use of problem 3.1.1a we get that
3.1 Inequalities with Areas 99
Figure 3.1 B C
N
M
K
L
1 1 1
SPEQF ¼ SPEF þ SEQF SABEF þ SFECD ¼ SABCD :
4 4 4
1
SB1 B2 :::Bn ¼ SA1 A2 :::An SB1 A2 B2 ::: SBn A1 B1 ¼ SA1 A2 :::An ðSA1 A2 A3 þ ::: þ SAn A1 A2 Þ ¼
4
1 1
¼ SA1 A2 :::An ð2SA1 A2 :::An SC1 A2 C2 ::: SCn A1 C1 2SC1 C2 :::Cn Þ > SA1 A2 :::An :
4 2
Similarly, we deduce that SALD ¼ ðnS1 yÞyðmS2 yÞ. Since SKLMN ¼ SBNC þ SALD,
then x þ y ¼ ðmS1 xÞxðnS2 xÞ þ ðnS1 yÞyðmS2 yÞ or 1 ¼ ðS1 þS mS1 nS2
2 ÞðmþnÞx
þ ðS1 þS
mS1 nS2
2 ÞðmþnÞy
. Now
we have to find the maximal value of x þ y at x < min (mS1, nS2).
Let x y. Denote ðS1 þSmS1 nS2
2 ÞðmþnÞx
¼ sin 2 α, ðS1 þS
mS1 nS2
2 ÞðmþnÞy
¼ cos 2 α, where α0 < α π4,
sin 2 α0 ¼ ðS1 þS2 Þðmþn
mS1 nS2
ÞminðmS1 ;nS2 Þ.
Then x þ y ¼ ðS1 þS24mS
ÞðmþnÞsin 2 2α < ðS1 þS2 ÞðmþnÞsin 2 2α0 .
1 nS2 4mS1 nS2
In the case of nS2 < mS1 the inequality (3.1) is obtained similarly.
3.1.4. We have that SABMND < SABCD,
1 1 1
SABMND ¼ SAMN þ SABM þ SAND ¼ SAMN þ SABC þ SACD ¼ SAMN þ SABCD ,
2 2 2
consequently, SAMN < 12 SABCD . Note that SABCD ¼ SABD þ SBCD ¼ SABD þ 4SCMN.
Therefore, 4SAMN ¼ SABCD þ SABD, so SAMN > 14 SABCD .
3.1.5. (Solution of G. Khotsanyan, 9th grade)
First we shall prove two lemmas which are used at the proof of the inequality.
Lemma 1 A straight line perpendicular to AB is drawn from point H of segment
AB, where AH < BH. Points C and D are taken on that perpendicular so that
CH < DH. Points E and F are the feet of the bisectors drawn from vertices C and D
in triangles ABC and ABD. Prove that AE < AF.
Proof Indeed, according to the interior angle bisector theorem, for triangles ABC
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
CH 2 þAH 2 AF DH2 þAH 2
and ABD we get: AE
EB ¼ AC
BC ¼ p ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi , ¼ AD
¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi .
CH 2 þBH 2 FB BD DH2 þBH 2
pffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
x2 þAH 2
Consider now a function f ðxÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffi BH 2
2
¼ 1 þ AH x2 þBH 2
. Note that it grows in
2x þBH2
the interval x 2 [0 ; 1 ) (since AH2 BH2 < 0). Consequently, from the condition
3.1 Inequalities with Areas 101
CH < DH, it follows that f(CH) < f(DH). Hence, we deduce that AE
EB < AF
FB or
EB < FB . From this inequality we obtain that AF > AE.
EBþAE FBþAF
Lemma 2 Given that points E and F are the feet of the bisectors drawn from
vertices C and D in triangles ABC and ABD, respectively, with CD||AB. Prove that,
if segments CE and DF do not intersect, then CD > EF.
Proof First introduce following notations (see also Figure 3.2) ∠ACB ¼ 2α,
∠ADB ¼ 2β, ∠BCD ¼ u, ∠ADC ¼ v. It is known that ∠ABC ¼ ∠BCD ¼ u,
∠BAD ¼ ∠ADC ¼ v.
From triangles ABC and ABD we deduce that 2α þ u þ v þ ∠CAD ¼ 180 ,
2β þ u þ v þ ∠CBD ¼ 180 . Hence, we have that 2α þ u þ v < 180 ,
2β þ u þ v < 180 . Summing up these two inequalities, it follows that
2α þ 2β þ 2u þ 2v < 360 , or ∠ECD þ ∠FDC < 180 . If segments CE and DF do
not intersect, then CD > EF. This ends the proof of the lemma.
Now, we continue the proof of the problem. Let us introduce following notations
(see Figure 3.3), AE ¼ a, AH ¼ b, DG ¼ c, BF ¼ d, AB ¼ CD ¼ e, BC ¼ AD ¼ f.
We have that
Figure 3.2 C D
u v
a b b
a
v u
A E F B
e-a e-c
E M G
a c
A b H f-b D
102 3 Areas
Figure 3.4 B C1 C
B1 M D1
A A1 D
Thus,
S 1
S0 ¼ ðc aÞðd bÞ: ð3:2Þ
2 2
Figure 3.5 B C
c2 c1
a1
A D
Figure 3.6 A
E
A1 M
O
F B
!
S ef
S0 1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi : ð3:3Þ
2 ef þ ðe þ f Þ e2 þ f 2 þ e2 þ f 2
pffiffiffiffiffi
Since e þ f 2 ef , e2 þ f2 2ef, then
qffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffi
ef þ ðe þ f Þ e2 þ f 2 þ e2 þ f 2 ef 3 þ 2 2 : ð3:4Þ
1 MA MB 1 2 sin ∠A sin ∠B
SMEF ¼ ME MF sin ∠EMF ¼ AB sin 2 ∠O ¼
2 AB2 2 sin ∠O
MA MB
¼ SABO sin 2 ∠O:
AB2
ð3:5Þ
104 3 Areas
(a) According to (3.5) SABO would be minimal if the value of the expression MAMB
AB2
2
¼ 14 14 ðMAMB
AB2
Þ
would be maximal i.e. at MA ¼ MB.
To construct segment AB one has to note that OA1 ¼ A1A, where MA1||OF.
Remark 1. SMEF 14 SABO .
Indeed, according to (3.5), SMEF MAMB
AB2
SABO 14 SABO .
2. SOA1 M 14 SOAB .
(b) According to (3.5) we have that
where ON ⊥ AB.
If the straight line passing through point M perpendicular to line OM intersects
the sides of the given angle at points A1 and B1, then segment A1B1 is what was
required. Indeed, we have that
1 1 sin 2 ∠O sin 2 ∠O 1 1
þ ¼ ON OM ¼ þ :
MA MB SMEF SMEF MA1 MB1
But if the straight line passing through point M perpendicular to line OM does
not intersect the sides of the given angle, then the required segment cannot be
constructed. Indeed, since N 2 OK (Figure 3.7), where KM||OE, then ON does not
have a maximal value.
N K
O F
Figure 3.7
3.1 Inequalities with Areas 105
Figure 3.8
D2
D3
D D'
D1 C1 C
M B1
B'
A A1 B B3 B2
1
SA1 B1 C1 D1 ¼ SMA1 B1 þ SMB1 C1 þ SMD1 C1 þ SMA1 D1 SABC
4
1 1 1
þ SBCD þ SACD þ SABD ¼
4 4 4
1 S
¼ SABCD ¼
2 2
(see the remark 1 of problem 3.1.6a). This means that SA1 B1 C1 D1 S2.
(b) Draw from points C and M straight lines parallel to diagonal BD (Figure 3.8).
1 1 1
SMA1 D1 þ SMB1 C1 SAB3 D3 þ SCB0 D0 < SAB3 D3
4 4 4
1 1 1
þ SB2 B3 D3 D2 ¼ SAB2 D2 ¼ SABCD :
4 4 2
Since from the statement of the problem, it follows that points C and A are
equidistant from line BD, i.e. BD is the midline of triangle AB2D2.
Remark 1. If ABCD is a square and point M coincides with C, then
1
SMA1 D1 þ SMB1 C1 ¼ SCBD ¼ SABCD :
2
106 3 Areas
Figure 3.9 B
B1
M
G
A P C1 C
Figure 3.10 C
E
E' D
A F B
2. In general the inequality does not hold true: take in a square ABCD a point
M close to point C and move vertices B and D along the sides in the direction
of A.
S2ABCD S2ABCD
3. One can prove that, in general case, SMA1 D1 þ SMB1 C1 < 14 max SBDC ; SABD .
SMGN SABC
9 sin ∠B, SNGP 9 sin ∠C. Thus
2 SABC 2
SABC 2 SABC
SMNP ¼ SMGP þ SMGN þ SNGP sin ∠A þ sin 2 ∠B þ sin 2 ∠C :
9 4
Figure 3.11 B
E
K M
D E1
M1
K1
A T P C
have to prove that SAE0 D 14 SABC . Note that, this follows from the solution of
problem 3.1.6a (see remark 2).
3.1.10. Draw from the point A a straight line parallel to DE (Figure 3.11).
Since DK ¼ KM ¼ ME, then AK1 ¼ K1M1 ¼ M1E1. Now by using Menelaus’
theorem for triangle AE1C and straight lines BT, BP, we obtain that
Consequently,
Olympiad.
(b) Assume that SA1 B1 C1 < minðSAB1 C1 ; SBA1 C1 ; SCA1 B1 Þ and let SAB1 C1 SBA1 C1
SCA1 B1 . Then, taking into account the solution of the problem 3.1.12a, we get
that
therefore
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
AA1 BB1 CC1 4SABC ða þ b þ cÞ αβγ ð1 αÞð1 βÞð1 γ Þ ¼ 2SA1 B1 C1
1 BB1 CC1
ða þ b þ cÞ, hence SA1 B1 C1 2AA
ðABþBCþACÞ.
(f) We have to prove that ð1 αð1 γ Þ βð1 αÞ γ ð1 βÞÞ αð1γ 1
Þ þ βð1αÞþ
1
γ ð1βÞÞ 3, or
1
1 1 1
A ¼ ðð1 αÞð1 βÞð1 γ Þ þ αβγ Þ þ þ 3:
αð1 γ Þ βð1 αÞ γ ð1 βÞ
Note that
ð1 αÞð1 βÞ βγ ð1 β Þð1 γ Þ
A¼ þ þ
α 1γ β
αγ ð1 αÞð1 γ Þ αβ 1β
þ þ þ ¼
1α γ 1β α
β 1γ γ
þβ 1 þ βþ þγ1þ γ
1γ β 1 α
1α α 1β α
þ þα1þ α¼ þ
γ 1
β α 1 β
β 1γ γ 1α
þ þ þ þ 3 3,
1γ β 1α γ
110 3 Areas
that
1 sin 2α sin 2β
ða þ bÞ2
SPMN 2 sin ð2α þ 2βÞ 1
¼ ¼ ¼
SABC 1 2 cos α cos β cos ðα þ βÞ
2a cos α 2b cos β sin ðα þ βÞ
2
1 1
¼ < pffiffi2 pffiffi2 ¼ 1,
cos ðα þ βÞ þ cos ðα þ βÞ cos ðα βÞ
2
2
2
þ 22
cos β 2a cos α
because 2bsin β ¼ sin α and 0 α β < α þ β 45 .
αβ
2
because cos 2 2 sin 2γ 0.
3.1 Inequalities with Areas 111
3.1.16. Consider the straight lines which pass through the vertices of a triangle and
are parallel to one of the sides of the parallelogram (Figure 3.12).
Note that it is sufficient to solve the problem for the cases, when the triangle and
the parallelogram have sides belonging to the same straight line (Figure 3.13a, b).
In a case of Figure 3.13a we have that SXYZT SAY 1 Z1 T 1 12 SABC (see the solution
of problem 3.1.6a, remark 2), where Z1T1||AB. In the case of Figure 3.13b, we have
that SABC SXCT 12 SXYZT .
3.1.17. (a) Let points A and B be the most removed vertices of a polygon. Then it is
clear that the polygon is inside the strip formed by the perpendiculars to the
segment AB at points A and B.
Consider the strip with a minimum width having boundaries parallel to the
segment AB and which contains the given polygon (Figure 3.14).
Then the intersection of these two strips contains the polygon and has area
2SABC þ 2SABD 2S.
(b) Let A, B, C be such vertices of the given polygon that SABC is maximal. It is clear
that the given polygon is contained in the half-plane ΠA, the boundary of which
passes through point A and is parallel to CB (Figure 3.15).
Figure 3.12
B Y Z
Y1 Y Z Z1
A X T T1 C X A B T
a b
Figure 3.13
112 3 Areas
The half-planes ΠB and ΠC are defined similarly. The intersection of these three
half-planes contain the given polygon and have the area equal to 4SABC. Thus, if
SABC S2, then this ends the proof. It remains to consider the case, whenSABC > S2.
Consider a triangle A1B1C1 with a minimum area having parallel sides to the
sides of triangle ABC and which contains the given polygon (Figure 3.16).
Figure 3.14
C
A B
Figure 3.15
PA
A
B
Figure 3.16 A1
D
A S2 B1
B
S1 M
F S3
E
C
C1
3.1 Inequalities with Areas 113
Figure 3.17
P
F
l1(t) l2(t)
A B
C
S1 S2
E
MC
It is clear that AA1, BB1 and CC1 intersect at point M. Let CC 1
¼ k, then SABC ¼ k
(S1 þ S2 þ S3), where S1 ¼ SFAC, S2 ¼ SDAB, S3 ¼ SECB, and D, E, F are some vertices
of the given polygon.
Because SABC > S2, consequently k > 1.
We have that
In the absence of some of the points DEF the proof can be done similarly.
3.1.18. (a) Let points A, B be the most removed vertices of the given polygon and
C be any point on segment AB (Figure 3.17).
Consider the function f(t) ¼ l1(t) l2(t) on the segment [0; 1], where t ¼ AC AB ,
while l1(t) and l2(t) are the lengths of the segments, cut off by the given polygon
from the midpoint perpendiculars of segments AC and CB, respectively. It is not
difficult to understand that the graph of the function f(t) is a broken line and as
l1(0) ¼ l2(1) ¼ 0, then we have that f(0) < 0 and f(1) > 0. Hence, it follows that there
exists a certain point C0, such that l1(t0) ¼ l2(t0), where t0 ¼ AC
AB (i.e. f(t0) ¼ 0).
0
Consider the straight line passing through point C and perpendicular to segment
AB. Assume that it divides the given polygon into two polygons with areas S1 and S2
(Figure 3.17).
Let us draw through points E and F support lines,1 then S1 l1(t) AC.
Similarly, we obtain that S2 l2(t) CB. Consequently, l1(t) AC þ l2(t) CB S.
For point C0 the quadrilateral EFPQ is a parallelogram, such that
1
If the line l has at least one common point with a figure F and whole figure F is located on one side
of l, then the line l is called a support line of the figure F.
114 3 Areas
1 1 1
SA1 B1 C1 ¼ SABC þ SABC1 þ SBCA1 þ SACB1 ab sin 3γ bc sin 3α ac sin 3β ¼
2 2 2
1 1 1
¼ 4SABC ab sin 3γ bc sin 3α ac sin 3β:
2 2 2
1 1 1
ab sin 3γ þ bc sin 3α þ ac sin 3β ¼ SABC ð2 cos 2γ þ 2 cos 2α þ 2 cos 2β þ 3Þ ¼
2 2 2
¼ SABC ð4cos 2 γ þ 4cos 2 α þ 4cos 2 β 3Þ 0
1 1 1
SAA1 B þ SBB1 C þ SCC1 A ¼ ac sin α þ ab sin β þ bc sin γ
2 2 2
sin α sin β sin γ
¼ SABC þ þ
sin β sin γ sin α !
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
sin α sin β sin γ
SABC 2 þ2 1 1
sin β sin γ sin α
sr ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffi !
sin α sin γ
SABC 4 1 ¼ 3SABC ,
sin γ sin α
pffiffiffiffiffi
because x þ y 2 xy (x 0, y 0).
2 2 2
3.1.22. We have that SS1 ¼ ðkþmþn
k
, S2 ¼ ðkþmþn
Þ2 S
n
, S3 ¼ ðkþmþn
Þ2 S
m
Þ2
(Figure 3.18).
Consequently
3.1 Inequalities with Areas 115
Figure 3.18 B
S3 S1
X
m k
S2
A n C
Figure 3.19 D E
K M F
N
C
B A
2 2 2
S1 þ S2 þ S3 ¼ S k
kþmþn þ n
kþmþn þ m
kþmþn
2 k 1 2 n 1 2 m 1
S þ þ
3 kþmþn 9 3 kþmþn 9 3 kþmþn 9
2 1 S
¼S 1 ¼ ,
3 3 3
since x2 23 x 19.
3.1.23. Let us draw through points B, D and F lines parallel to segments AF, EF and
AB, respectively (Figure 3.19). Note that SBCD ¼ SBND, SDEF ¼ SDKF, SABF ¼ SBMF.
Consequently 2SBDF ¼ SABCDEF þ SMNK SABCDEF.
3.1.24. Denote the points symmetrical to P, Q and R relative to the symmetry center
0 0 0 0 0 0
of hexagon ABCDEF by P , Q and R , respectively. It is clear that points P , Q and R
0 0 0 0 0
2
belong, respectively, to sides DE, AF and BC. Since Q R#R Q, PQ #P Q and RP #
0
RP , we have 2SPQR ¼ SPR0 QP0 RQ0 SABCDEF (see the solution to problem 3.1.23).
3.1.25. Let ABCDEF be a regular hexagon with a symmetry center O and triangle
MNP an inscribed triangle, with side MN passing through point O. One can assume
2
AB # CD denotes that the segments AB and CD are parallel and equal.
116 3 Areas
that M 2 CD, N 2 AF and P 2 AB. We have that SMNP max (SMNA, SMNB), SMNA
¼ 12 SAMDN 12 SACDF ¼ 13 SABCDEF and SMNB ¼ SBON þ SBOM ¼ SBAO þ SBCO
¼ 3 SABCDEF . Consequently, SMNP 3 SABCDEF .
1 1
AC ¼ λ, AB ¼ μ, SBPC ¼ x, SPDE ¼ y.
3.1.26. Let us introduce the following notations: EC BD
Indeed λμ þ 12 ð1 λÞ þ 12 ð1 μÞ ¼ 1 λμ
2 9. It remains to note that
7
SQMP ¼ S SQCP þ SAQM þ SBMP S 79 S ¼ 29 S, consequently SQMP 29 S.
3.1.28. It is clear that the cross section of the cube by a plane passing through its
center is a centrally symmetric convex polygon, which has an even number of sides.
Thus, it is either a quadrilateral (Figure 3.20a), or a hexagon (Figure 3.20b).
In the first case (Figure 3.20a), we have that Sсеч ¼ bh a a ¼ a2.
3.1 Inequalities with Areas 117
b D
b
B a c
d g
c
h b d E
A
b
a
F
a b
Figure 3.20
Figure 3.21 B S2
C
S1
A
D
yi yi¢
3.1.29. It is clear that one can leave two of the given strips as they are and move the
remaining k strips. Let coordinate straight line be parallel to non of the boundary
lines of these strips.
Denote by xi (i ¼ 1, 2, . . . , k) coordinates of the intersection points of the line li
with the coordinate line, where the straight line li is parallel to the bounding lines of
the i-th strip and is equidistant from them. It is clear that there exist numbers ai and
bi, such that ai xi bi [because the intersection of each of these k strips with
parallelogram ABCD is not empty (Figure 3.21)], where ABCD is the intersection of
two not moved strips.
118 3 Areas
b b
x
x x
a g a 1800-g
x b b=0
a a 1800 -d
d
a b
Figure 3.22
It is clear that to each figure which is the intersection of all these strips,
corresponds a set of numbers x1, x2, . . . , xk. Denote the area of all these k þ 2 strips
by S(x1, x2, ..., xk). If all these strips do not have common points we will set S(x1,
x2, ..., xk) ¼ 0.
We shall prove that there exists a constant number M, such that at any
xi , x0i 2 ½ai ; bi , i ¼ 1, 2, . . . , k the following inequality holds true
Sðx1 ; x2 ; :::; xk ÞS x0 ; x0 ; :::; x0 M jx1 x0 j þ jx2 x0 j þ ::: þ jxk x0 j :
1 2 k 1 2 k
ð3:8Þ
Note that S y1 ; :::; yi1 ; yi ; yiþ1 ; :::; yk S y1 ; :::; yi1 ; y0i ; yiþ1 ; :::; yk S1 þ
S2 (Figure 3.21), and S1 , S2 y0i yi d, where d is the longest diagonal of the
parallelogram ABCD. Then S1 þ S2 2 y0i yi d, and thus
Sðx1 ; x2 ; :::; xk Þ S x0 ; x0 ; :::; x0
1 2 k
Sðx1 ; x2 ; :::; xk Þ S x01 ; x2 ; :::; xk þ S x01 ; x2 ; :::; xk S x01 ; x02 ; x3 :::; xk þ
þ ::: þ S x01 ; x02 ; :::; x0k1 ; xk S x01 ; x02 :::; x0k
2d x1 x01 þ 2d x2 x02 þ ::: þ 2d xk x0k ¼ M jx1 x01 j þ jx2 x02 j þ ::: þ jxk x0k j ,
where M ¼ 2d.
Using the inequality (3.8) one can prove that there exist numbers x01 , x02 , :::, x0k ,
such that for any xi 2 [ai, bi], i ¼ 1, 2, . . . , k, Sðx1 ; x2 ; :::; xk Þ S x01 ; x02 ; :::; x0k .
We shall prove that all strips to which correspond numbers x01 , x02 , :::, x0k have a
common center of symmetry.
Proof by contradiction argument.
Let certain two parallel sides of the intersection with the area S x01 ; x02 ; :::; x0k ¼ S
be not equal. Let a > b (Figure 3.22).
3.1 Inequalities with Areas 119
Figure 3.23 y
E
a
K
O N x
Let us shift the strip, one of the boarders of which contains a, by x (Figure 3.22).
0
Denote the area of intersection of new strips by S . Then for sufficiently enough
small x we have that
SMON ¼ :
cos 2α
pffiffi
pffi 2
pffiffi 2
2 cos αR
2
(1) For 0 < R 4
2
we have to prove that cos 2α 22 R , or
pffiffi cos αpffiffiffiffiffiffiffiffiffi ffi pffiffi cos αpffiffiffiffiffiffiffiffiffi ffi pffiffi
2 ð cos 2αÞ
pffiffiffiffiffiffiffiffiffiffi 2 ð cos 2αÞ
pffiffiffiffiffiffiffiffiffiffi
2 R (α ¼
6 0). We need to prove that 2 2
4 , or
1 cos 2α
pffiffiffiffiffiffiffiffiffiffiffiffiffi 1 cos 2α
2 cos α 1 cos 2α, i.e. (cosα 1)2 0.
(2) We shall prove that for 0 < R < 12 the following inequality holds true
p ffi 2
2 cos αR
2 pffiffiffi 2
cos 2α 14 R2 , or 2R cos α 12 0.
pffiffi
For 42 < R < 12 the equality may hold true, since p1ffiffi2 < 2p1ffiffi2R < 1, hence there
exists such an angle α, for which cos α ¼ p1ffiffi . 2 2R
120 3 Areas
B1 C1 B1 W C1
B1 C1
V
A1 D1 P A1
A1 N D1 D1
S
M N O
K U
B C B C B C
b
M O1
R T
K
A D A D A Q D
a b c
Figure 3.24
(b) According to problem 3.1.29, if one shifts the plane parallel to itself so that it
passes through the center of the cube, then the area of the cross section does not
decrease. Thus, it is sufficient to make the proof for the case of dissection of the
cube by a plane passing through its center. Consider the following cases
(Figure 3.23a, b, c).
pffiffiffi
I. (Figure 3.24a) Ssec ¼ 2SANC1 2maxðSAD1 C1 ; SADC1 Þ ¼ 2.
II. (Figure 3.24b) Ssec ¼ 2SNKP 2maxðSNKC1 ; SNKC Þ ¼ maxðSANC1 K ; SA1 NCK Þ
pffiffiffi
2.
III. Let OO1 ⊥ ABC, where O is the center of the cube, O1K ⊥ QT and O1K ¼ R,
∠O1KO ¼ β (Figure 3.24c).
pffiffi
pffi 2
12STQD 12 22R
If 0 < R 4 , then Ssec ¼ cos β pffiffiffiffiffiffi
2
R (see problem 3.1.30 a,1).
1þR2
pffiffi 2 pffiffi
4
pffiffiffi pffiffi
We need to prove that 1 2 22 R pffiffiffiffiffiffiffi 2R ffi
, or 2 2R 2R2 p2ffiffiffiffiffiffiffiffiffi
2R ffi
,
4þR
1 2
1þ4R2
pffiffiffi pffiffi pffiffiffi p ffiffiffiffiffiffiffiffiffiffi pffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffiffiffiffiffiffiffiffiffi1
2
2 R pffiffiffiffiffiffiffiffiffi
2 ffi
2
, or 2 p1þ4R ffi R, or 4 2R 1 þ 4R2 þ 1 þ 4R2 .
2
1þ4R
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1þ4R
pffiffiffi
Indeed, 1 þ 4R2 þ 1 þ 4R2 > 4R þ 2R ¼ 6R > 4 2R.
pffiffi 12ð14R2 Þ
12S
IV. If 42 < R 12, then Ssec ¼ cosTQD β pffiffiffiffiffiffi
R (see problem 3.1.30a)
1 2 þR
4
pffiffiffi pffiffiffiffiffiffiffiffiffiffi
1 1þ4R2 1
(3) We need to prove that 2 þ 2R R
2
2, or (1 þ x)2(1 þ x) 8x, where
2
Figure 3.25 A
a-x
x
B D C
Using further the inequality (3.9) for numbers ai ¼ ABi sin x and bi ¼ ACi sin
(α x) at 0 < x < α, we deduce that
Figure 3.26 C
B А'
E M
E' F
A B'
D B1
C1
M N1
K1
L
L1
D1
B
C N
К
A A1
a b
Figure 3.27
0 0 0
From problem 3.1.31a we have that ME þ MF max (MC þ MA , MB þ MD),
or EF max (AC, BD).
(c) Let the section of a tetrahedron ABCD be a quadrilateral KLMN (Figure 3.27a).
We need to prove that SKLMN max (SAMC, SDKB). Let us project the tetrahedron
on a plane perpendicular to KM (Figure 3.27b).
Since SKLMN ¼ 12 KM L1 N 1 and from problem 3.1.31b we have that L1N1 max
(A1C1, B1D1), then SKLMN max 12KM A1 C1 ; 12KM B1 D1 maxðSAMC ; SBKD Þ.
It remains to note that SAMC max (SACD, SABC) and SBKD max (SABD, SCBD),
(see remark to the solution of problem 2.4.4.).
In the case of triangular section one can assume that it is passing, e.g., through
vertex D, i.e., points L and M have coincided with point D.
3.1.32. Let points C1, B1, A1, D1 be H images of points C, B, A, D (Figure 3.28).
3.1 Inequalities with Areas 123
B1
D
AC×CD
AB×CD
H- image BD×AC
C А C1 BC×CD
A1
BC×AD
BC×AC
B
D1
Figure 3.28
Figure 3.29 C2 C C1 D
Δ1 B1 D1
B2
B E2 Δ2
A1 E
E1
A2 A F1 F
(a) If ∠A þ ∠C > 180 , then ∠B þ ∠D < 180 . Hence C1B1A1D1 is a convex quad-
rilateral, thus SA1 B1 C1 þ SA1 C1 D1 > SC1 B1 D1 . SA1 B1 C1 ¼ CD2 SABC , SA1 C1 D1 ¼
BC2 SACD , SC1 B1 D1 ¼ AC2 SBCD . Consequently CD2 SABC þ BC2 SACD > AC2
SBCD.
(b) If ∠A þ ∠C ¼ 180 , then SA1 B1 C1 þ SA1 C1 D1 ¼ SC1 B1 D1 . Thus CD2 SABC þ BC2
SACD ¼ AC2 SBCD.
(c) If ∠A þ ∠C < 180 , then ∠B þ ∠D > 180 , hence point A1 is inside the triangle
C1B1D1, consequently SA1 B1 C1 þ SA1 C1 D1 < SC1 B1 D1 . Therefore, it follows that
CD2 SABC þ BC2 SACD < AC2 SBCD.
3.1.33. Consider a hexagon A2B2C2C1E2F1, where CC2 ¼ AF1, AA2 ¼ CC1, C1E2||
A2B2||DE, C2B2||F1E2||BC (Figure 3.29).
We have that the opposite sides of the hexagon A2B2C2C1E2F1 are parallel to
each other and C1 C2 ¼ AFþCD
2 ¼ A2 F1 . Consequently, C1E2 ¼ B2A2, E2F1 ¼ B2C2.
Let points E and E3 be symmetric to each other with respect to point E2, then
ABCE3 is a parallelogram. Therefore, C1 E2 ¼ ABþDE2 , F1 E 2 ¼
BCþEF
2 . One can easily
prove that B2C1||A1D1 and B2F1||B1E1, B2C1 ¼ A1D1, B2F1 ¼ B1E1.
This ends the proof of the point (a) of the problem.
124 3 Areas
Figure 3.30 d
d1
c c1 b
a
Figure 3.31
N d C P
N1 P1 D
c O
B M1 Q1
M A Q
possible.
To conclude the proof it remains to draw through the vertices A, B, C, D of the
rectangle with the sides a and b a straight line, parallel to the sides of the rectangle
with the sides c and d. We have that
ab ¼ SABCD < SBCQ1 þ SADN1 þ SCDP1 þ SABM1 ¼ SBCN þ SADQ þ SCDP þ SABM
¼ cd ab:
Denote the intersection points of a straight line passing through O with lines BA,
AD, CD and BC, respectively, W, X, Y and Z. Assume that W
X
A. Then OW OX ¼ 1,
and OY > 1. Now we rotate the line around point O in such a way that it does not
OZ
2SAOB SBOC
T1 ¼ :
a
sin ð2αþπ=4Þ
2b
pffiffiffiffiffi pffiffiffiffiffi pffiffiffi
cos α ¼ 1, i.e. C0 1. Thus, in all cases T 1 þ T 2 C0 S.
Note that if in a quadrilateral the conditions AB ¼ BC, AD ¼ CD are satisfied and
ABCO is a parallelogram, then using the condition SSACD ABC
¼ tg 2 α0 , where 0 α0 π4
sin ð2α0 þπ=4Þ pffiffiffiffiffi pffiffiffiffiffi pffiffiffi
and C0 ¼ cos α0 we obtain that T 1 þ T 2 ¼ C0 S.
We need to prove that, if 0 α π4, then sin 2α þ π4 < 54 cos α. This ends the
proof of the point (b)
of the problem.
Indeed, let φ 2 0; π4 and cos φ ¼ 45, then at 0 α < φ, we have that
sin 2α þ π4 < 1 ¼ 54 cos φ < 54 cos α. But if φ α π4, then tgφ ¼ 34 >
126 3 Areas
Figure 3.32 A D1 B
M4
M
O M1
C1
М3
A1
K
M2
D
B1
C
pffiffiffi pffiffi
2 1 ¼ tg π8. Therefore, φ > π8 and sin 2α þ π4 sin 2φ þ π4 ¼ 22 31
25 <
pffiffi
2 4 1, 25 cos α.
2 5
Remark
Using
the concept of a derivative it is possible to prove that
tg 2α0 þ π4 tg α0 þ 2 ¼ 0, or tg3α0 þ 3tgα0 2 ¼ 0. Consequently, tg α0 ¼
p ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffi p ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffi
2 þ 1 2 1¼ 0, 59 . . ., then C0 ¼ 1, 11. . ..
3 3
3.1.36. (Solution of O. Sarkisyan, 9th grade) Let the diagonals of the quadrilateral
ABCD intersect at point O (Figure 3.32) and OC AO
¼ λ, OD
BO
¼ μ.
Let us express S0 through S, λ, μ. Since segment AA1, BB1, CC1 and DD1 medians
of triangles ABC, BCD, ACD and ABD, then
S
SAA1 CC1 ¼ SAA1 C þ SCC1 A ¼ SABA1 þ SCDC1 ¼ :
2
Similarly,
S
SBB1 DD1 ¼ SBCB1 þ SDAD1 ¼ :
2
Consequently,
AK A1 K BM þ MO 1 BM
¼ ¼ ¼ þ :
AO MO 2MO 2 2MO
AK AO þ OK OC 1
¼ ¼1þ ¼1þ :
AO AO 2AO 2λ
λþ1
Therefore, 1 þ 2λ
1
¼ 12 þ 2MO
BM
, from which it follows that BM ¼ 2λþ1 BO. From
(3.10) we deduce that
BM1 BM λ þ 1 BO
¼ ¼ ¼
M1 B1 A1 B1 2λ þ 1 BD=2
ð2λ þ 2Þμ
¼ :
ð2λ þ 1Þðμ þ 1Þ
Consequently,
Figure 3.33
B
O2
O3
b
c a
A b C
O1
А2
B¢ B¢ B2
A¢ A¢ B1
A1 B1 A1
C2
А2
C1 B2 C1 l
l C2
C¢ C¢
a b
Figure 3.34
0 0 0 0 0 0
triangles A B C and A2B2C2 have a common point, where A B C is the image of
triangle A1B1C1 under similarity transformation with a center O and similarity
coefficient k. The existence of line l follows from Figure 3.34.
Let A1B1C1 and A2B2C2 be two given triangles. Note that if line l contains sides
A1C1 or B1C1, then parallelograms A1C1B1D1 and A2C2B2D2 do not have any
common internal point, while if line l contains side A1B1, then hexagons
A1M1N1B1K1P1 and A2M2N2B2K2P2 do not have common internal points [here
MiNi and PiKi are midlines of triangles AiBiCi and AiBiDi (i ¼ 1, 2), respectively
(Figure 3.35)].
From the aforesaid it follows that, if given triangles AiBiCi, i ¼ 1, 2, . . . , n do
not have any common internal point, then any two of the hexagons AiMiNiBiKiPi,
i ¼ 1, 2, . . . , n also do not have any common internal point.
P
n P
n
Therefore, S0 SA i M i N i B i K i P ¼ 2 SAi Bi Ci ¼ 2 S.
3 3
i¼1 i¼1
130 3 Areas
Figure 3.35 D2
P2 K2
A2 B2
D1
M2 N2
P1 K1
A1 C2 B1 l
M1 N1
C1
Figure 3.36 N
A C
A1 C1
F E1 D
M E K
3.1.39. Let us draw through points A, C and E straight lines MN, NK and MK parallel
to BF, BD and DF, respectively (Figure 3.36).
Then ΔBFD ΔMNK, and let MN ¼ λBF.
Note that
λ2 SBDF ¼ SMNK ¼
¼ SBDF þ SMNBF þ SNKDB þ SMKDF :
We have that R > r1 (see the solution of problem 7.1.79), where r1 is the radius of
the incircle of triangle BDF. Hence, we obtain that
3.1 Inequalities with Areas 131
Figure 3.37 A
B M E
D
C
R R pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
SABCDEF SBDF ¼ SACE SBDF :
r1 r
Remark 1. If center O of the circumcircle of a triangle ACE is in a convex hexagon
ABCDEF, then SABCDEF ¼ SABOF þ SBCDO þ SDEFO BFR 2 þ 2 þ 2 ¼
BDR FDR
R
r1 SBDF .
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2. If ABCDEF is a convex hexagon, then SABCDEF Rr Rr11 SACE SBDF , where r,
R and r1, R1 are the inradius and circumradius of triangles ACE and BDF,
respectively.
3.1.40. Denote given five points by A, B, C, D, E. If the pentagon with vertices at A,
B, C, D, E is not convex, then we can assume that point D is inside triangle ABC,
pffiffiffi
then SABC ¼ SABD þ SDBC þ SDAC > 6 > 5 þ 1.
Consider now the case when pentagon ABCDE is convex. Let M be the point of
intersection of BE with AC (Figure 3.37).
pffiffi pffiffi pffiffi pffiffiffi
Suppose that BM ME
1þ2 5, then SSACE ¼ BM
ME
1þ2 5, SACE SABC 1þ2 5 > 1 þ 5.
pffiffi ABC
pffiffi
51
ME
In the case BM < 1þ2 5, SMDE min (SCDE, SADE) > 2, SSBMD ¼ BM
ME > 2 , SBMD >
pffiffi p ffiffi
ffi p ffiffi
MDE
ffi pffiffiffi
51
2 SMDE ¼ 5 1. Therefore SBDE ¼ SBMD þ SMDE > 5 1 þ 2 ¼ 5 þ 1.
3.1.41. In the case when four vertices of the hexagon A1A2A3A4A5A6 are on two
adjacent sides of the unit square ABCD, we have that if A1, A2 2 AB, A3, A4 2 BC,
4
then SA1 A2 A3 SA2 A3 A4 ¼ 14 A1 A2 BA3 A2 B A3 A4 14 A1 A2 þBA3 þA
4
2 BþA3 A4
64
1
.
Consequently, minðSA1 A2 A3 ; SA2 A3 A4 Þ 18.
It remains to consider the case when A1, A2 2 AB, A3 2 BC, A4, A5 2 CD,
A6 2 AD.
Let A4C A2B. If DA5 AA1, then we proceed to the proof by contradiction
argument. Suppose that the areas of all these triangles are greater than 18. Then by
approaching point A3 to C, and point A6 to D, one can obtain SA3 A4 A5 ¼ SA4 A5 A6 ¼ 18,
while other areas are greater than 18. Now by drawing together points A1 and A2, we
can reach SA6 A1 A2 ¼ SA1 A2 A3 ¼ 18, while areas of other two triangles are greater than 18.
Then we have that
132 3 Areas
1 þ A1 A2 1 þ A4 A5
SA1 A2 A3 A4 A5 A6 ¼ SA1 A2 A3 A6 þ SA6 A3 A4 A5 ¼ BA3 þ CA3 ¼
2 2
1 A1 A2 BA3 A4 A5 CA3 3
¼ þ þ ¼ :
2 2 2 4
On the other hand
1 A1 A2 þ A4 A5 1
SA1 A2 A3 A4 A5 A6 ¼ SA6 A1 A5 þ SA2 A3 A4 þ SA1 A2 A4 A5 > þ ¼
4 2 4
1 1 1 1 1 1 1 3
þ þ ¼¼ þ þ ¼ :
8 BA3 A3 C 4 8 BA3 A3 C 4 2 ðBA3 þ A3 CÞ 2 4
Figure 3.38 Q A1 A2 R
0 1
45 a
4a
M
a
N
A3
U
A6 1 b V
4b
a
P A5 A4 S O
3.1 Inequalities with Areas 133
i
Since the function f ðxÞ ¼ x cos α sin α þ 1x decreases in 0; pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
cos α sin α
ffi and
aþb< 1
< pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 ffi, ða þ bÞ cos α sin α þ
then > cos α sin αþ
1 1
sin α cos α sin α sin α
pffiffiffi aþb
sin α ¼ cos α þ sin α. Consequently, 2 > sin αþ45 1
ða þ bÞ cos α sin α þ aþb
1
ð Þ
cos αþ sin α
pffiffiffi
> sin ¼ 2 . This leads to a contradiction.
ðαþ45 Þ
This ends the proof.
Remark The number 18 cannot be made smaller, because if A1
A, A4
C,
A1 A2 ¼ A1 A6 ¼ 12 ¼ A3 A4 ¼ A4 A5 , then SA1 A2 A3 ¼ SA2 A3 A4 ¼ ::: ¼ SA6 A1 A2 ¼ 18.
0 0
3.1.42. Let us consider point B and P , such that BB ~ 0 ¼ PP~ 0 ¼ AC. ~ Let O be the
0
midpoint of segment BC and AB ¼ c, BC ¼ a, AC ¼ b, PB ¼ d.
0 0
According to problem 1.1.14a, for points P, B , P , C we have that bc d þ 6.
0 0
Since quadrilateral ABB C is a parallelogram, then O is the midpoint of segment AB .
0 2 0
Consequently, PO2 ¼ 2þ2d 4AB , AB 2 ¼ 2b2 þ 2c2 a2, PO2 ¼ 8þ18a
2 2
4 .
Hence, we obtain that b þ c a ¼ d 12, then d 12 ¼ bc d þ 6. Thus
2 2 2 2 2
pffiffiffiffi pffiffi pffiffi pffiffiffiffiffi pffiffi
d 1þ2 73, therefore SABC ¼ 43 bc 43 ðd þ 6Þ 13 þ 73 83.
3.1.43. First we prove that if given an angle ∠MAN ¼ α, then on sides AM, AN one
can find points B0, C0, respectively, and inside the given angle a point P, such that
PA0 ¼ r1, PB0 ¼ r2, PC ¼ r3 and ∠PB0A ¼ ∠PC0A.
Indeed, let point K be inside angle MAN and ρ(K, AM) ¼ r2, ρ(K, AN) ¼ r3.
Denote by P the intersection point of the circle with a center at A and radius r1
with ray AK. Then AP ¼ r1 and ρρððP;AM Þ
P;AN Þ ¼ r3 , and it remains to use the conditions
r2
r2 > r1 and r3 > r1 to choose on rays AM and AN points B0 and C0, such that
PB0 ¼ r2, PC0 ¼ r3 (Figure 3.39).
Then ΔPEB0 and ΔPFC0 are similar.
Hence ∠PB0A ¼ ∠PC0A. We need to prove that SABC SAB0 C0 .
Let ΔABC and point P satisfy the conditions of the problem.
0 0
Consider points B and P , such that BB ~ 0 ¼ PP~ 0 ¼ AC~ 0 , then (see the solution of
0
problem 3.1.42), if PB ¼ d, we obtain that
Figure 3.39 B0
M
r2
E
P
r1 r3
a
A F C0 N
134 3 Areas
bc dr 1 þ r 2 r 3 ð3:12Þ
d2 r 22 þ r 23
r 21 þ 2ðdr 1 þ r 2 r 3 Þ cos α:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Hence d r 1 cos α þ r 22 þ r 23 þ 2r 2 r 3 cos α r 21 sin 2 α.
Therefore,
1
SABC ¼ bc sin α
2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 2
r 1 cos α þ r 1 r 22 þ r 23 þ 2r 2 r 3 cos α r 21 sin 2 α þ r 2 r 3 sin α ¼ SAB0 C0 ,
2
since for triangle AB0C0 the inequality (3.12) turns into equality.
Now we will find the minimal value of SABC. To do that, let us note that points A,
B, C are on the circles with a center at point P and with radiuses r1, r2, r3, respec-
tively (see Figure 3.40).
0 0
Note that triangle AB C also satisfies the conditions of the problem and
SABC SAB0 C0 ¼ 14 AB AB0 AC AC0 sin 2 α ¼ 14 r 22 r 21 r 23 r 21 sin 2 α.
Thus, it follows that
2
1 2 2 1 r 2 r 21 r 23 r 21 sin 2 α
SABC ¼ r 2 r 1 r 3 r 1 sin α
2 2 2
:
4 SAB0 C0 4SAB0 Co
Figure 3.40 C
a B
A
a
B¢ P
C'
3.1 Inequalities with Areas 135
Figure 3.41
3.1.45. Prove that in any convex polygon one can place a rectangle having the area
not less than 14 of the area of the given polygon.
3.1.46. There are 5 patches placed on the shirt with area 1, the area of each of them
being not less than 12. Prove that one can find two patches so that the area of their
common parts is not less than 15.
3.1.47. Let T1 and T2 be two triangles with sides a1, b1, c1 and a2, b2, c2. Prove that
there exists a triangle T with sides a, b, c, such that if S1, S2 and S are the areas of
triangles T1, T2 and T, respectively, then
qffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffi
S1 þS2 a21 þa22 b21 þb22 c21 þc22
(a) S 2 , where a ¼ 2 , b¼ 2 ,c ¼ 2 ,
pffiffiffiffiffiffiffiffiffiffiffiffiffi
(b) S 4 S1 S2 , where a ¼ a1 þ a2, b ¼ b1 þ b2, c ¼ c1 þ c2.
136 3 Areas
3.1.48. The area of a convex quadrilateral ABCD is equal to S. Prove that the area of
the quadrilateral, with the vertices at the midpoints of segments AC, AD, BD and
BC, is less than S2.
3.1.49. Prove that in a triangle with area S one can inscribe a regular triangle, such
that its area is not greater than S4.
3.1.50. Prove that any acute triangle with area 1 can be placed into right-angled
pffiffiffi
triangle with area not greater than 3.
3.1.51. Given several squares the sum of area of which is equal to S. Prove that
(a) one can place them without overlapping inside a square with area 2S,
(b) with these squares one can always cover the square with area S4.
3.1.52. Let AD be the altitude of the right-angled triangle ABC, ∠A ¼ 90 . The
straight line passing through the centers of the incircles of triangles ABD and ACD
intersects sides AB and AC, respectively, at points K and L. Prove that
SABC 2SAKL.
3.1.53. Bisectors of angles A, B, C of an acute triangle ABC intersect its circumcir-
cle at points A1, B1, C1, respectively. The straight line AA1 intersects the bisectors of
the external angles B and Cof triangle ABC at point A0. Points B0, C0 are defined
similarly. Prove that SA0 B0 C0 4SABC .
3.1.54. Let ABC be an acute triangle, points M, N and P be the feet of the
perpendiculars drawn from the centroid of the triangle to sides AB, BC and CA,
respectively. Prove that SMNP > 27
4
SABC .
3.1.55. Let ABCD be a unit square. For any internal points M and N, such that line
MN does not contain any of the vertices of the square, denote by S(M, N ) the least of
the areas of the triangles with the vertices from the set of the points {A, B, C, D, M,
N}. Find the smallest number k, such that S(M, N ) k for all such points M and N.
Hint Let point N be inside of triangle CMD, then
SAMB þ SCMD ¼ 12, SCMD ¼ SCMN þ SMND þ SCND 3S(M, N ). Therefore
4SðM; N Þ 12. Hence, it follows that SðM; N Þ 18.
3.1.56. Points K, L, M and N are taken on sides AB, BC, CD and DA of a convex
quadrilateral ABCD, respectively. Denote by S1, S2, S3 and S4 the areas of
pffiffiffiffiffi pffiffiffiffiffi
triangles AKN, BKL, CLM and DMN, respectively. Prove that 3 S1 þ 3 S2 þ
p ffiffiffiffi
ffi p ffiffiffiffi
ffi p ffiffiffiffiffiffiffiffiffiffiffi
ffi
3
S3 þ 3 S4 2 3 SABCD .
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3
Hint Prove that SABCD ¼ AK S1
AN SABD
þ þ SABD
3 1 AK AN
AB AD SABCD 3 AB AD SABCD .
Hint Draw a segment connecting the opposite sides of the parallelogram, then one
can assume that ΔA1A2A3 is inside the parallelogram with area 12.
Consequently, SA1 A2 A3 14.
3.1.58. (a) Let ABCDEF be a convex hexagon, such that AB||ED, BC||FE and CD||
AF. Prove that
6
1
SABC SBCD SCDE SDEF SEFA SFAB SABCDEF : ð3:13Þ
6
3.1.62. Prove that none of the triangles inscribed into polygon M can have an area
greater than the maximal area of the triangles, the vertices of which coincide with
some of the vertices of M.
3.1.63. (a) Let M be a convex polygon and l be an arbitrary straight line. Prove that
one can inscribe into M a triangle, оne side of which is parallel to l and the area of
which is not less than 38 of the area of M.
138 3 Areas
(b) Let M be a regular hexagon and l be a line parallel to one of its sides. Prove that
one cannot inscribe into M a triangle, оne side of which is parallel to l and the
area of which is greater than 38 of the area of M.
3.1.64. Solve problem 3.1.43 in the case of α > π2.
Hint Prove that there exists a triangle PB0C0 and a point A inside it, such that
PA ¼ r1, PB0 ¼ r2, PC0 ¼ r3 and ∠BAC ¼ α, ∠AB0P ¼ ∠AC0P. Then
SABC SAB0 C0 .
3.1.65. A triangle with the area 1 is cut out of paper. Prove that one can flex it over
pffiffiffi
the segment of a line, so that area S of obtained figure would be less than 2 2.
Hint Let AB AC BC, if we flex triangle ABC (SABC ¼ 1) over the bisector of
angle C, then S ¼ BCþAC
BC
. On the other hand, if we flex triangle ABC over a segment
perpendicular to side BC, then one can prove that
2
S ∠C
,
3 þ tg
tg ∠B
and that the equality can hold true. Then, prove that
!
1 2 pffiffiffi
min sin ∠B
; tg ∠C
<2 2:
1þ sin ð∠Cþ∠BÞ 3þ tg ∠B
3.1.66. Given a triangle ABC. Prove that there exists a straight line l, such that if
points A1, B1, C1 are symmetric to points A, B, C with respect to line l, then the area
pffiffiffi
of the common parts of triangles ABC and A1B1C1 is greater than 2 2 1 .
Hint See the Hint of problem 3.1.64.
3.1.67. Let diagonals AD, BE and CF of the convex hexagon intersect at one point
and A1 ¼ AD \ BF, D1 ¼ AD \ CE, B1 ¼ BE \ AC, E1 ¼ BE \ DF, C1 ¼ CF \ BD,
F1 ¼ CF \ AE. Prove that
1
SA1 B1 C1 D1 E1 F1 SABCDEF :
4
Hint Let diagonals AD, BE, and CF intersect at a point O. Prove that
AOB SBOC SAOF
SA1 OB1 ¼ ðSAOBSþS BOC ÞðSAOB þSAOF Þ
18 ðSAOF þ SBOC Þ.
3.1.68. The medians of triangle ABC intersect the circumcircle of the triangle for
the second time at points A1, B1, C1. Prove that SA1 B1 C1 SABC .
Remark Let point M be the centroid of triangle ABC. Then
ða2 þb2 þc2 Þ
2
4.1.1. Let R be the radius of the circumsphere of tetrahedron SABC. Prove that
(a) SA2 + SB2 + SC2 + 4R2 AB2 + BC2 + AC2,
(b) SA2 + SB2 + SC2 + AB2 + BC2 + AC2 16R2,
(c) xyAB2 + yzBC2 + xzAC2 + xtSA2 + ytSB2 + ztSC2 (x + y + z + t)2R2, where x, y, z, t
are arbitrary numbers.
4.1.7. (a) Given 7 points on a unit sphere. Prove that among them one can find
pffiffiffi
2 points, such that the distance between them is less than 2.
(b) Given 5 points on a unit sphere. Prove that among them one can find 2 points,
pffiffiffi
such that the distance between them is not greater than 2.
4.1.8. For any points A, B, C, M and arbitrary numbers x, y, z prove that
(a) (x + y + z)(xMA2 + yMB2 + zMC2) xyc2 + yza2 + xzb2, where AB ¼ c, BC ¼ a,
AC ¼ b ,
(b) aMA2 + bMB2 + cMC2 abc,
(c) MA MBc + MB MCa + MC MAb abc,
(d) (x + y + z)(xMB2MC2 + yMA2MC2 + zMA2MB2) xyc2MC2 + yza2MA2 + xzb2MB2,
(e) (a + b + c)(aMB2MC2 + bMA2MC2 + cMA2MB2) a2b2c2,
(f) (MA MB + MB MC + MC MA)(MA + MB + MC) a2MA + b2MB + c2MC.
4.1.9. Prove the following inequality
A1 A2 A 2 A3 An1 An A1 An
þ þ ::: þ ,
MA1 MA2 MA2 MA3 MAn1 MAn MA1 MAn
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
MA þ MB þ MC 2 p
MA MB cos ∠PMA cos ∠PMBþ ffi
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
þ2pMB MC cos ∠QMB cos ∠QMCffi þ
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
þ2 MA MC cos ∠RMA cos ∠RMC:
4.1.12. Let M be a given point inside of triangle ABC. Denote the distances from the
point M to lines BC, CA, AB by da, db, dc and the distances from point M to vertices
A, B, C by Ra, Rb, Rc, respectively. Prove that
(a) Ra + Rb + Rc 2da + 2db + 2dc,
(b) RaRb + RbRc + RcRa 2Rada + 2Rbdb + 2Rcdc.
4.1.13.1 For any points A1, A2, . . ., An, M and any numbers m1, . . ., mn prove the
following inequality m1 MA21 þ ::: þ mn MA2n m1 GA21 þ ::: þ mn GA2n , where
! !
m1 + . . . + mn 0 and m1 GA 1 þ ::: þ mn GA n ¼ ~0.
4.1.14. Let points A1, A2, . . ., An be on the same circle, m1, . . ., mn > 0 and G be
! !
such a point that m1 GA 1 þ ::: þ mn GA n ¼ ~ 0. Given that straight lines GA1, . . .,
GAn intersect the circle for the second time at points B1, . . ., Bn. Prove that
(a) m1GB1 + . . . + mnGBn m1GA1 + . . . + mnGAn,
(b) m1 GB21 þ ::: þ mn GB2n m1 GA21 þ ::: þ mn GA2n .
4.1.15. (а) Let M be a point inside of the convex n-gon with vertices A1, A2, . . ., An.
Perpendiculars MB1, MB2, . . ., MBn are drawn from point M to lines A1A2, A2A3,
. . ., AnA1, respectively, and C1 2 A1A2, C2 2 A2A3, . . ., Cn 2 AnA1 are arbitrary
A2 ¼ A2 A3 ¼ ::: ¼ An A1 .
mn MBn
points. Given positive numbers m1, . . ., mn, such that mA11MB 1 m2 MB2
P P
Prove that mi mj Ci C2j mi mj Bi B2j .
i<j i<j
(b) Let M be a point inside of convex n-hedron. Perpendiculars MB1, . . ., MBn are
draw from point M to planes Π1, . ., Πn containing the faces of the polyhedrons
with areas S1, . . ., Sn, respectively, and C1 2 Π1, . . ., Cn 2 Πn are arbitrary
points. Given positive numbers m1, . . ., mn such that m1SMB
1
1
¼ ::: ¼ mnSMB
n
n
.
P P
Prove that mi mj Ci C2j mi mj Bi B2j :
i<j i<j
4.1.16. (a) Given a triangle ABC and positive numbers x, y, z. Points A1, B1, and C1
are taken on straight lines BC, CA, and AB, respectively. Prove that
xA1 B21 þ yB1 C21 þ zA1 C21 xAB42ðþyBC
xyþyzþzxÞ
2 S2 , and that the equality holds true.
þzAC2 ABC
1
The point Gis called center of mass for the system of points A1, A2, . . ., An with masses m1,
m2, . . ., mn, and the expression IM ¼ m1 MA21 þ ::: þ mn MA2n is called the moment of inertia of the
system of points A1, . . ., An with masses m1, . . ., mn relative to the point M.
142 4 Application of Vectors
O1 O2 p1 þ p2
< ,
A1 A2 þ B1 B2 þ C1 C2 2maxðp1 ; p2 Þ
4.1.23. Let ABCDbe a tetrahedron, such that AC ⊥ BC and AD ⊥ BD. Prove that the
cosine of the angle between straight lines AC and BD is less than CD
AB .
4.1.24. (a) Given points A1 and B1, A2 and B2, . . ., An and Bn on faces Γ1, Γ2, . . ., Γn
of a convex m-hedron (m n 3), respectively. Given that for any i 2 {1, 2, ..., n}
vector mi~ ei mi1~ ei1 is perpendicular to the plane containing face Γi, where m1,
!
A Aiþ1
m2, . . ., mn are given positive numbers and~ ei ¼ Aii Aiþ1 , i ¼ 1, 2, . . ., n, An + 1 A1,
m0 ¼ mn, ~ e0 ¼ ~en . Prove that
Given a tetrahedron C1C2C3C4, such that C1C2 ¼ C2C3 ¼ C3C4 ¼ C4C1. Find on
faces C2C3C4, C1C3C4, C1C2C4, and C1C2C3 points B1, B2, B3, and B4, respec-
tively, such that sum B1B2 + B2B3 + B3B4 + B4B1 is minimal.
4.1 Application of Vectors for Proving Geometric and Trigonometric Inequalities 143
(c) Given a cube ABCDA0 B0 C0 D0 . Find on faces ABCD, AA0 BB0 , BB0 C0 C, A0 B0 C0 D0 ,
DD0 C0 C, and AA0 D0 D points B1, B2, B3,B4, B5, and B6, respectively, such that the
sum B1B2 + B2B3 + B3B4 + B4B5 + B5B6 + B6B1 is minimal.
4.1.25. Given that circumcenter O of the tetrahedron is inside of that tetrahedron.
Prove that
(a) DA2 + DB2 + DC2 + AB2 + BC2 + AC2 > 12 OA2,
(b) DA + DB + DC + AB + BC + AC > 6 OA.
Solutions
or ð~
r B þ~ r C Þ2 0.
r D ~
144 4 Application of Vectors
! ! ! ! !
a, BC ¼ ~
4.1.4. (a) Let us denote AB ¼ ~ c, DE ¼ ~
b, CD ¼ ~ d, EF ¼ ~
e, then one has
to prove that
2
2 ~ 2 ~
a þ b þ~
2 2 ~
c þ d þ~
2
e þ ~
2 ~
a þ b þ~ ~
c þ d þ~e
2 2 2
~ a þ~b þ~ c þ ~ c þ~
b þ~ d þ ~ c þ~d þ~ e ,
! ! ! ! ! !
(b) Let AB ¼ ~ a, BC ¼ ~
b, CD ¼ ~ c, DE ¼ ~
d, EF ¼ ~
e. Note that 2A1 D1 ¼
! ! ! !
BE þ AD , consequently, 4A1 D21 ¼ BE2 þ AD2 þ 2 BE AD . Consider paral-
2 ! ! !2
lelogram BADM. We have that a ~
~ d ¼ EM 2 ¼ BM BE ¼
! !2 2
AD BE , hence4A1 D21 ¼ 2 AD2 þ BE2 ~ a ~
d . Then, we have that
4 A1 D21 þ B1 E21 þ C1 F21 ¼ 4CF2 þ 4BE2 þ 4AD2
2 2 2
~ a ~d ~ a þ~ c þ~
b þ 2~ d þ~e ~ b ~e ¼
2 2 2
¼4 ~ c þ~ d þ~ e þ4 ~ a þ~ b þ~c þ4 ~ c þ~
b þ~ d
2 2 2
~ a ~d ~ a þ~ b þ~d þ~e þ 2~c ~ b ~e ¼
2 2 2
¼3 ~ a ~d þ3 ~ a þ~ b þ~d þ~ c þ3 ~
e þ 2~ e 4ð~
b ~ a þ~ e Þ2
c þ~
2 2 2
3 ~ a ~ d þ3 ~ a þ~ b þ~d þ~e þ 2~c þ3 ~ b ~
e
3ðAB þ DEÞ2 þ 3ðBC þ EFÞ2 þ 3ðCD þ AFÞ2 :
! ! ! !
a, AC ¼ ~
4.1.5. Let AD ¼ ~ b, then DC ¼ ~
b ~ a, and AB ¼ k ~ b ~a , where k > 0.
2
One has to prove that k > 1. We have that ~ a2 > ðk 1Þ~ b k~
a and
0
2 k1 ðk 1Þ~
b k~
2
~ B aÞ
b2 > k~ b ðk þ 1Þ~
a . Therefore, ¼ ~
a 2
@ 2 .
k þ 1‘ ~ ~
b kb ðk þ 1Þ~
2 a >0
Hence, k > 1.
4.1 Application of Vectors for Proving Geometric and Trigonometric Inequalities 145
!
ei ¼ MA
4.1.6. Let ~ MAi , i ¼ 1, 2, 3, 4, and ~
i e ¼~e1 þ~ e2 þ~e3 þ~ e4 . Let us draw through
point M a plane Π perpendicular to vector ~ e. It is clear that, if all points Ai were in
the same half-space with boundary Π, then point M would have been inside the
given tetrahedron. This means that there exists a vector~ ek , such that~e ~
ek 0. Then
we have that ð~ e ~ ek Þ~
ek 1. Hence, we deduce that min cos ∠Ai MAj
i<j
3 ð~ ek Þ~
e ~ 13.
1
ek
Similarly, we obtain that there exists a vector ~ em , such that ~
e ~em 0. In that
case, we have that ð~
e ~em Þ ~
em 1. Thus, it follows that
max cos ∠Ai MAj 13 ð~
e ~em Þ ~em 13. This inequality can be proven as follows:
i<j
P
max cos∠Ai MAj 16 cos∠Ai MAj ¼ 12 1
ð~
e2 ~e2 1 ~ e2 2 ~
e2 3 ~e2 4 Þ 13, beca-
i<j i<j
use ~
e2 0 and ~
e2 1 ¼ ~
e2 2 ¼ ~
e2 3 ¼ ~
e2 4 ¼ 1.
4.1.7. (а) Let points A1, A2, . . ., A7 are on the unit sphere with center O. Let us
pffiffiffi ! 2 2
assume that Ai Aj 2 ði 6¼ jÞ, then Ai Aj 2 ¼ Ai Aj ¼ ~ aj ~
ai 2, where
!
OA i ¼ ~ai , i ¼ 1, . . ., 7. This means that for i 6¼ j we have that
~
ai ~
aj 0: ð4:1Þ
~
ai ~
aj < 0, ði 6¼ jÞ: ð4:2Þ
A1
B
MB×MC
MB×AC
H- image MC×AB
M C M1 C1
MA×MB
MA×BC
MA×MC
A
B1
Figure 4.1
4.1 Application of Vectors for Proving Geometric and Trigonometric Inequalities 147
! !
4.1.9. Let us denote MA i ¼ ~ r i and MA MAi 2
i
¼!ρ i , i ¼ 1, . . ., n. Note that
rffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
! ! 2 ! !
rj j ¼ ð~ri2~r2j Þ ¼
j~ri~
2
ρ i ρ j ¼ ρ i ρ j . We have to prove
Ai Aj
MAi MAj ¼ ~
ri ~ rj ~r i ~
r j
that ! ρ1! ρ 2 þ ! ρ2! ρ 3 þ ::: þ !ρ n1 ! ρ n !ρ1! ρ n . This can be
obtained, if we use the following inequality j~ a1 j þ j~
a2 j þ ::: þ j~
an1 j
a1 þ ~
j~ a2 þ ::: þ ~ an1 j (see Section 1.2) for vectors ~ ! !
ai ¼ ρ iþ1 ρ i , i ¼
1, :::, n 1.
4.1.10. Let us draw tangents to the circle at the vertices of the inscribed n-gon
A1 . . . An, and choose unit vectors ~
e1 , :::,~
en on these tangents (Figure 4.2).
Let B1 2 A1An, B2 2 A1A2, . . ., Bn 2 An 1An be the marked points, then
B1 B2 þ B2 B3 þ ::: þ Bn1 Bn þ Bn B1
! ! ! !
B1 B2~
e1 þ B2 B3~ e2 þ ::: þ Bn1 Bn~
en1 þ Bn B1~ en ¼
! ! ! ! ! !
¼ B1 A1 þ A 1 B2 ~ e 1 þ B2 A2 þ A 2 B3 ~e2 þ ::: þ Bn An þ An B1 ~
en ¼
! ! ! 2S
¼ A1 A2~
e1 þ A2 A3~
e2 þ ::: þ An A1~
en ¼ ,
R
! ! ! !
because B2 A2~
e2 ¼ B2 A2~
e1 , :::, An B1~
en ¼ An B1~
e1 .
Figure 4.2
A2 e2
e1 A3
A1 B2 B3 e3
B1
en
An
148 4 Application of Vectors
MA þ MB þ MC
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ∠AMB pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ∠BMC pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ∠AMC
2 MA MB cos þ 2 MB MC cos þ 2 MA MC cos :
2 2 2
ð4:3Þ
pffiffiffiffiffiffiffi pffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffi
Let A1, B1, C1 be such points that MA1 ¼
MA, MB1 ¼ MB, MC1 ¼ MC and
∠A1 MB1 ¼ ∠AMB , ∠B1 MC1 ¼ ∠BMC
2 , ∠A1 MC1 ¼ π
∠AMC. Then, from inequality
! ! 2 ! 2 2
P
n
where mi ¼ m. Therefore, IM IG.
i¼1
4.1.14. (a) As mR2 ¼ I0 ¼ mOG2 + IG (see the proof of problem 4.1.13), then
m(R2 OG2) ¼ IG, where O is the center and R is the radius of the given circle.
Hence, point Gis inside of the given circle, thus GAi GBi ¼ R2 OG2.
4.1 Application of Vectors for Proving Geometric and Trigonometric Inequalities 149
We have that
! ! !2
X
n X
n n pffiffiffiffiffiffiffiffiffiffiffiffiffi2 X
X n pffiffiffiffiffiffiffiffiffiffiffiffiffi2 X
n pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
mi GBi mi GAi ¼ mi GBi mi GAi mi GAi GBi ¼
i¼1 i¼1 i¼1 i¼1 i¼1
! ! !2
n pffiffiffiffiffiffiffi
X X X
n
pffiffiffiffiffi 2
n
¼ m2 R2 OG2 ¼ mI G ¼ mi 2 ð mi GAi Þ mi GAi
i¼1 i¼1 i¼1
P
n P
n
(see the proof of problem 4.1.15a). Therefore, mi GBi mi GAi .
i¼1 i¼1
P
n P
n
Therefore, mi GB2i mi GA2i (see the proof of problem 4.1.14 a)
i¼1 i¼1
P
n P
n
Remark If 0 < p 2, then mi GBip mi GAip .
i¼1 i¼1
Indeed, we have that
! ! !2
X
n X
n X
n pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffip pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffip 2
mi GBip mi GAip mi GBi GAi ¼ m R2 OG2 ¼
i¼1 i¼1 i¼1
p
¼ m2 mm1 GA21 þ ::: þ mmn GA2n :
p
Because f ðxÞ ¼ x2 at 0 < p 2 is concave function on the interval [0 ; + 1 ) and
P
n
mi
m ¼ 1, then by using the Jensen’s inequality we deduce that
i¼1
m mn 2 m1 2 mn 2
1
f GA21 þ ::: þ
GAn f GA1 þ ::: þ f GAn , or
m m m m
m mn p2 m mn
1 1
GA21 þ ::: þ GA2n GA1p þ ::: þ GAnp , consequently,
m m m m
150 4 Application of Vectors
! !
X
n X
n m mn p
1
mi GBip mi GAip m2 GA21 þ ::: þ GA2n
i¼1 i¼1
m m
2
m1 GA1p þ ::: þ mn GAnp :
P
n P
n
Hence, it follows that mi GBip mi GAip .
i¼1 i¼1
! !
4.1.15. (а) Note that mi MB 1 þ ::: þ mn MB n ¼ ~0. Indeed, let m1MB1 ¼ kA1A2, . . .,
! !
mnMBn ¼ kAnA1. Under rotations by 90 or –90 vectors mi MB 1 , :::, mn MB n trans-
! ! ! !
form into vectors kA1 A2 , :::, kAn A1 , and because kA1 A2 þ ::: þ kAn A1 ¼ ~
0, then
! ! ~
m1 MB 1 þ ::: þ mn MB n ¼ 0.
! !
Let C0 be such a point that m1 C0 C þ ::: þ mn C0 C ¼ ~
1 0, then n
X X ! ! 2 X
mi mj Ci C2j ¼ C0 Cj C0 Ci mi mj ¼ mi mj C0 C2j
i<j i<j i<j
X X ! !
þ mi mj C0 C2i 2mi mj C0 Ci C0 Cj ¼
i<j i<j
!2
X
n X
n
! X
n
¼ ðm1 þ ::: þ mn Þ mi C0 C2i mi C0 Ci ¼ ðm1 þ ::: þ mn Þ mi C0 C2i :
i¼1 i¼1 i¼1
P
n P
n
Similarly, we obtain that mi mj Bi B2j ¼ ðm1 þ ::: þ mn Þ
mi MB2i .
i<j i¼1
It is known that for any a1, . . ., an and b1, . . ., bn > 0 inequality a21 þ ::: þ a2n
2
b1 þ ::: þ b2n ða1 b1 þ ::: þ an bn Þ2 holds true, the equality being reached only if
2
b1 ¼ ::: ¼ bn . Indeed, it is not difficult to check that a1 þ ::: þ a2n
a1 an
2 P 2
b1 þ ::: þ b2n ða1 b1 þ ::: þ an bn Þ2 ¼ ai bj aj bi 0, where the equality
i<j
holds true, only if aibj ajbi ¼ 0, ( i < j), or equivalently, if ab11 ¼ ::: ¼ abnn . Then, we
have that
4.1 Application of Vectors for Proving Geometric and Trigonometric Inequalities 151
X n
n
Ai A 2 X X Ai Aiþ1 2 X pffiffiffiffiffi 2
iþ1
mi mj Ci C2j ¼
pffiffiffiffiffi ð mi Co Ci Þ ðm1 þ ::: þ mn Þ
i¼1
m i i<j i¼1
m i i<j
!2 !2
X
n X n
ðm1 þ ::: þ mn Þ Ai Aiþ1 Co Ci 4ðm1 þ ::: þ mn Þ SCo Ai Aiþ1
i¼1 i¼1
! 2
X
n
4ðm1 þ ::: þ mn ÞS2A1 A2 :::An ¼ 4ðm1 þ ::: þ mn Þ SMAi Aiþ1 ¼
i¼1
!2
Xn
Ai Aiþ1 pffiffiffiffiffi
¼ ðm1 þ ::: þ mn Þ pffiffiffiffiffi mi MBi
i¼1
mi
! !2
X A i A2
n Xn
¼ ðm1 þ ::: þ mn Þ iþ1
mi MBi ¼
i¼1
mi i¼1
X n
Ai A2iþ1 X
¼ mi mj Bi B2j ,
i¼1
m i i<j
P P
where Cn + 1 C1. Therefore, mi mj Ci C2j mi mj Bi B2j .
i<j i<j
! !
(b) It is sufficient to prove that mi MB 1 þ ::: þ mn MB n ¼ ~ 0 and to replace in the
proof of (a) term Ai A2iþ1 by S2i and areas by volumes.
Pn ! ! n ! !
P
Let N be any point inside, then V ¼ 13 NB i MB
MBi Si ¼ 3
i k
NB i mi MB i ,
i¼1 i¼1
where mi SMB
i
i
¼ 1k, i ¼ 1, . . ., n.
! ! !
By making (in the obtained equality) the replacement NB i ¼ NM þ MB i , we
! Pn !
deduce that the expression NM mi MB i is constant for any N. Hence,
i¼1
P
n !
mi MB i ¼ ~
0.
i¼1
4.1.16. (a) Let m1, m2, m3 be positive numbers, such that x ¼ m1m2, y ¼ m2m3,
z ¼ m1m3. Then,
(see the proof of problem 4.1.15a). To prove that the equality holds true, it is
sufficient to prove that inside of triangle ABC there exists point M, such that
m1 MA1 m2 MB1 m3 MC1
BC ¼ AC ¼ AB , where MA1 ⊥ BC, MB1 ⊥ AC, MC1 ⊥ AB, and A1 2 BC,
B1 2 AC, C1 2 AB.
152 4 Application of Vectors
(b) Let us choose points C1, C2, and C3 on edges A1A4, A2A4, and A3A4, respectively,
S2 S2 S2
so that A1 C 1
C1 A 4 ¼ SA2 1 A2 A3 , A 2 C2
C2 A 4 ¼ SA2 1 A2 A3 , A3 C 3
C3 A 4 ¼ SA2 1 A2 A3 . Then, it is not difficult to
A2 A3 A4 A1 A 3 A4 A1 A2 A4
quently, SAMB 4
A A
¼ SAMB 1
A A
. Similarly, we obtain that SAMB 4
A A
¼ SAMB 2
A A
and SAMB 4
A A
¼ SAMB 3
A A
.
1 2 3 2 3 4 1 2 3 1 3 4 1 2 3 1 2 4
Now, by using problem 4.1.15b for numbers m1 ¼ m2 ¼ m3 ¼ m4 ¼ 1, we obtain
that the given sum is minimal at those points.
4.1.17. Lemma 1 Let M0 be such a point, that for any point M the following
P
n P
n
inequality holds true: mi MAi mi M0 Ai . If
i¼1 i¼1
Pn !
M A
(a) M0 6 Ai , i ¼ 1, 2, . . ., n, then mi M00 Ai i ¼ ~
0, (4.4)
i¼1
P !
n M A
(b) M0 Ak, then mi M00 Ai i j mk : (4.5)
i¼1, i6¼k
The Proof
P
n !
M A
(a) Let ei ¼ ~
mi~ S 6¼ ~
0, where ~
ei ¼ M00 Ai i . Take point Mt so that the equality
i¼1
!
M0 Mt ¼ t ~
S 6¼ ~
0 holds true, where t > 0.
!
M t Ai
Let us denote ~
ei ðtÞ ¼ M t Ai , then we have that
4.1 Application of Vectors for Proving Geometric and Trigonometric Inequalities 153
!
X
n X
n
! ! X
n X
n
!
mi M0 Ai mi M0 Ai ~
ei ðtÞ ¼ M0 Mt mi ~
ei ðtÞ þ mi Mt Ai ~
ei ðt Þ
i¼1 i¼1 i¼1 i¼1
X
n
¼ f ðt Þ þ mi Mt Ai : ð4:6Þ
i¼1
Now, we need to prove that there exists t0 > 0, such that f(t0) > 0, then from (4.6)
P
n Pn
it follows that mi M0 Ai > mi Mt0 Ai . This leads to a contradiction the condition
i¼1 i¼1
of the lemma. Hence ~ S ¼~ 0.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Because j~ ei j ¼ ð~
ei ðtÞ ~ ei ðtÞ ~ ei Þ2 ¼ 2 sin ∠Mt2Ai M0 < ∠Mt Ai M0 , then there
j~Sj
exists t0 > 0, such that j~ei ðt0 Þ ~ ei j < 2nmi , for all i ¼ 1, 2, . . ., n. Then,
! X ! X X
n
n n
!
f ðt0 Þ ¼ M0 Mt0 mi~
ei ðt0 Þ ¼ M0 Mt0 mi~
ei þ mi M0 Mt0 ð~
ei ðt0 Þ ~
ei Þ
i¼1 i¼1 i¼1
X
n
!
t0~
S2 mi M0 Mt0 j~
ei ðt0 Þ ~
ei j
i¼1
X n ~
S t0 2
~ ~
> t0 S t0 S
2 mi ¼ ~S > 0:
i¼1
2nm i 2
P n
!
~
(b) Let S ¼ ei > mk . Take point Mt so that the equality Ak Mt ¼ t~
mi~ S
i¼1, i6¼k
holds true, where t > 0. Then, we have that
X
n X
n
!
m i Ak Ai mi Ak Ai~
ei ðtÞ ¼
i¼1, i6¼k i¼1, i6¼k
! X X X
n n n
!
¼ Ak M t mi~ei þ mi Ak Mt ð~
ei ðtÞ ~
ei Þ þ m i M t Ai :
i¼1, i6¼k i¼1, i6¼k i¼1, i6¼k
Let us denote ~ S ¼ mk þ q and choose t0, such that j~ ei j < ðn1q Þmi , i ¼ 1,
ei ðt0 Þ ~
2, . . ., n, i 6¼ k. Then,
Xn X n
q X
n
S2 t0 ~
mi Ak Ai > t0~ S mi þ mi Mt0 Ai
i¼1 i¼1, i6¼k
ðn 1Þmi i¼1, i6¼k
Xn Xn
¼ mk Mt0 Ak þ mi Mt0 Ai ¼ m i M t 0 Ai :
i¼1, i6¼k i¼1
154 4 Application of Vectors
The obtained
inequality
contradicts to the conditions of the lemma. Conse-
P
n
quently, m~e mk . This ends the proof of the lemma.
i¼1, i6¼k i i
Remark By using the Weierstrass theorem one can prove that there exists a point
P
n
M, such that the sum mi MAi is minimal (see [2], problem 19).
i¼1
Lemma 2 If for point M0 the condition (4.4) or (4.5) is fulfilled and points A1, . . .,
An are not on the same line, then for any point M, different from M0, the inequality
Pn P
n
mi MAi > mi M0 Ai holds true.
i¼1 i¼1
X
n X
n
! ! X
n X ! n X n
mi MAi > mi MAi~ei ¼ MM0 mi~
ei þ mi M0 Ai~
ei ¼ mi M0 Ai ,
i¼1 i¼1 i¼1 i¼1 i¼1
X
n X
n
! ! X
n Xn ! !
mi MAi > mi MAi~ei MAk mi~
ei ¼ mi MAi MAk ~
ei
i¼1 i¼1, i6¼k i¼1, i6¼k i¼1, i6¼k
X n
¼ mi Ak Ai :
i¼1
Remark 1. From lemmas 1 and 2 it follows that, if points A1, A2, . . ., An are not on
the same line, then point M0 is unique.
2. If in a statement of lemma 2 the condition that points A1, A2, . . ., An are not on
P
n Pn
the same line was not given, then mi MAi mi M0 Ai
i¼1 i¼1
4.1.18. The proof of the problem follows from the following two lemmas.
Lemma 1 If B1B2 . . . Bn is a polygon inscribed into a polygon A1A2 . . . An,
( Bi 6 Aj and Bi 2 AiAi + 1, An + 1 A1), so that for any polygon C1C2 . . . Cn
(Ci 6 Aj , Ci 2 AiAi + 1) and any positive numbers mi the inequality
Pn Pn
mi Ci Ciþ1 mi Bi Biþ1 , where Cn + 1 C1, Bn + 1 B1, holds true, then the
i¼1 i¼1
following conditions are satisfied:
!
Ai Aiþ1 ðmi1~
ei1 mi~
ei Þ ¼ 0, i ¼ 1, :::, n, ð4:7Þ
!
Bi Biþ1
where ~
ei ¼ Bi Biþ1 , m 0 ¼ m n, ~
eo ¼ ~
en .
4.1 Application of Vectors for Proving Geometric and Trigonometric Inequalities 155
Proof Suppose that, at i ¼ k condition (4.7) does not hold true, that means that
! !
! 0 0 Bk1 B0k 0 B0k Bkþ1
~ ~
0 6¼ Ak Akþ1 ðmk1 ek1 mk ek Þ. Let Bk 2 Ak Akþ1 ,~
ek1 ¼ Bk1 B0 ,~
ek ¼ B0 Bkþ1 .
k k
If points B0k and Bk are different, then we have that
! 0 þ m B !
mk1 Bk1 Bk þ mk Bk Bkþ1 > mk1 Bk1 Bk~e k1 k k Bkþ1~ e0k ¼
! 0 ! 0 ! 0 !
¼ mk1 Bk1 B0k~
ek1 þ mk1 B0k Bk~
ek1 þ mk Bk B0k~ek þ mk B0k Bkþ1~e0k ¼
!
¼ mk1 Bk1 B0k þ mk B0k Bkþ1 þ B0k Bk mk1~e0k1 mk~e0k : ð4:8Þ
! ! ! !
Let B0k Bk ¼ λAk Akþ1 , B0k Bk mk1~ e0k1 mk~e0k ¼ λAk Akþ1 mk1~ e0k1 mk~
e0k .
!
Let us choose λ, so that λ Ak Akþ1 ðmk1~ ek1 mk~ek Þ > 0. We need to prove that
!
0 0
one can choose Bk , so that Bk 2 Ak Akþ1 and λ Ak Akþ1 mk1~ e0k1 mk~e0k > 0.
!
Indeed, let Ak Akþ1 ðmk1~ ek1 mk~ ek Þ ¼ a 6¼ 0. Let us choose B0k , so that
0 a
~
e ~ ei < , at i ¼ k 1 ; k (see the proof of lemma 1 of problem 4.1.17),
i 4mi Ak Akþ1
then
! !
e0k1 mk~
λ Ak Akþ1 mk1~ e0k λ Ak Akþ1 ðmk1~ek1 mk~ ek Þ ¼
! 0 0
¼ λ Ak Akþ1 mk1 ~ ek1 ~ek1 mk ~ ek ~ek
! 0 0 aλ
λ Ak Akþ1 mk1 ~ ek1 ~ ek1 mk ~ ek ~
ek :
2
Therefore,
! ! a λ
0 0
λ Ak Akþ1 mk1~ek1 mk~
ek λ Ak Akþ1 ðmk1~
ek1 mk~
ek Þ
2
aλ
¼ > 0:
2
We have obtained that for polygon B1 B2 :::Bk1 B0k Bkþ1 :::Bn that the considered
sum is less than for polygon B1B2 . . . Bk . . . Bn. This leads to a contradiction.
Remark If m1 ¼ m2 ¼ . . . ¼ mn, then it is possible to prove lemma 1 more simply
than in the general case. On the other hand, Bi Aj is impossible.
Lemma 2 If for the inscribed polygon B1B2 . . . Bn the condition (4.7) is satisfied,
P
n !
then for any inscribed polygon C1C2 . . . Cn the sum mi Ci Ciþ1~
ei is constant and
i¼1
156 4 Application of Vectors
P
n P
n
mi Ci Ciþ1 mi Bi Biþ1 .
i¼1 i¼1
X
n
! X
n ! !
mi Ci Ciþ1~
ei ¼ mi Ci Aiþ1 þ Aiþ1 Ciþ1 ~
ei
i¼1 i¼1
X
n
! X
n
!
¼ mi Ci Aiþ1~
ei þ mi Aiþ1 Ciþ1~
ei ¼
i¼1 i¼1
X
n
! X
n
!
¼ mi Ci Aiþ1~
ei þ mi1 Ai Ci~
ei1
i¼1 i¼1
X
n ! ! X
n
!
¼ mi1 Ai Aiþ1 Ci Aiþ1 ~
ei1 þ mi Ci Aiþ1~
ei ¼
i¼1 i¼1
X
n
! X
n
!
¼ mi1 Ai Aiþ1~
ei1 Ci Aiþ1 ðmi1~
ei1 mi~
ei Þ
i¼1 i¼1
X
n
!
¼ mi1 Ai Aiþ1~
ei1 ,
i¼1
Remark If n is odd, then there exists no more than one polygon B1B2 . . . Bn, while
for even n there can exist an infinite number of polygons B1B2 . . . Bn.
4.1.19. Let O be the incenter of triangle ABC. We need the following property of
point O:
! ! !
BC OA þ AC OB þ AB OC ¼ ~
0: ð4:9Þ
Let us consider points A1, B1, C1 (Figure 4.3). Since, the circle with diameter OA
passes through points B1 and C1, then B1 C1 ¼ OA sin ∠A ¼ BCOA 2R , where R is the
circumradius of triangle ABC. From the above said conditions OA ⊥ B1C1,
OB ⊥ C1A1, OC ⊥ A1B1, it follows that, under the rotation by 90 vectors
! ! ! ! ! !
2RB1 C1 , 2RC1 A1 , 2RA1 B1 transform into vectors BC OA , AC OB , AB OC .
Thus, we have that condition (4.9) is satisfied. According to that condition
4.1 Application of Vectors for Proving Geometric and Trigonometric Inequalities 157
Figure 4.3 B B1
A1
C1
A B1 C
!
! !
!
B1 C1 O1 A1 þ A1 C1 O1 B1 þ A1 B1 O1 C1 ¼ B2 C2 O2 A2
! !
þA2 C2 O2 B2 þ A2 B2 O2 C2 :
ð4:10Þ
! ! ! ! ! !
Taking into account that O2 A2 ¼ O2 O1 þ O1 A1 þ A1 A2 , O2 B2 ¼ O2 O1 þ
! ! ! ! ! !
O1 B1 þ B1 B2 , O2 C2 ¼ O2 O1 þ O1 C1 þ C1 C2 , from (4.10) it follows that
!
!
!
p2 O1 O2 ¼ O1 A1 ðB2 C2 B1 C1 Þ þ O1 B1 ðA2 C2 A1 C1 Þ
! ! ! ð4:11Þ
þO1 C1 ðA2 B2 A1 B1 Þ þ þB2 C2 A1 A2 þ A2 C2 B1 B2
!
þA2 B2 C1 C2 :
Therefore,
p2 O1 O2 O1 A1 B2 C2 B1 C1 þ O1 B1 A2 C2 A1 C1 þ O1 C1 A2 B2 A1 B1 þ
þB2 C2 A1 A2 þ A2 C2 B1 B2 þ A2 B2 C1 C2 :
ð4:12Þ
Let us further note that |B2C2 B1C1| B1B2 + C1C2, |A2C2 A1C1|
A1A2 + C1C2, |A2B2 A1B1| A1A2 + B1B2, then from (4.12) it follows that
p p p
p2 O 1 O 2 1
þ B2 C2 A1 A2 þ 1 þ A2 C2 B1 B2 þ 1 þ A2 B2 C1 C2 <
2 2 2
p1 þ p2
< ðA1 A2 þ B1 B2 þ C1 C2 Þ:
2
Thus, A1 A2 þBO11BO22þC1 C2 < p12pþp2 .
2
It is clear that, from the very beginning we could assume that max( p1, p2) ¼ p2.
Remark If A1 6 A2, B1 6 B2, then (4.13) takes the following form
p2 O1O2 (O1A1 + A1B1 + O1B1)C1C2. Thus, p2 O1O2 < (C1A1 + A1B1 + C1B1)
C1C2 ¼ p1 C1C2.
p1 p2
Hence, it follows that OC11 O
C2
2
< min ;
p p .
2 1
4.1.20. (a) Let P, Q, and R be the tangential points of the unit sphere with the faces
! ! !
Then, ~
of trihedral angle. p ¼ OP, ~ q ¼ OQ , and ~r¼ OR are unit vectors and
arccos3 ~
2 c
p,~ π
q 2, arccos3 ~ 2 c
q,~ π
r 2, arccos3 ~
2 c
p,~r π2 . On the other
!
hand, for vector ~ a ¼ OA , we have that ~ p ¼~
a~ q ¼~
a~ a~r ¼ 1.
pffiffiffi
We needto prove that
j~
a j 3 .
Let max ~ cq ; ~
p,~ cr ; ~
p,~ cr
q,~ ¼ ~ cq ¼ φ, then there exist angles φ1 and
p,~
! !
α, such that φ2 φ1 φ, α < π2, and ~ p ¼ 1; 0; 0 , ~ q ¼ cos φ; sin φ; 0 ,
!
~
r ¼ cos φ1 cos α; sin φ1 cosα; sin α (see Figure 4.4).
pffiffiffi
If α ¼ π2, then a ¼ 3.
y,!z , then we find that x ¼ 1, x cos φ + y sin φ ¼ 1 and x cos φ
If ~ a ¼ x, 1
cos α + y sin φ1 cos α + z sin α ¼ 1. Consequently, x ¼ 1, y ¼ tg φ2 , z¼
φ
φ
φ
cos 2 cos α cos φ1 2 = sin α cos 2 .
2
Then, we have to prove that cos φ2 cos α cos φ1 φ2 sin 2 α 1þ32cos φ.
Figure 4.4
z
r
О a
y
p j1 q
x
4.1 Application of Vectors for Proving Geometric and Trigonometric Inequalities 159
Figure 4.5
Figure 4.6 B¢ C¢
L
A¢ M0 M D¢
e2 L0
e3
B C
e1
K0
K
A D
2
We have that cos φ2 cos φ1 φ2 1. Therefore, f ðtÞ ¼ cos φ2 t cos α
max f cos φ2 ; f ð1Þ , at t 2 cos φ2 ; 1 .
4 4
Note that f cos φ2 ¼ 1þ cos φ
4 sin α2 1þ32cos φ 4 sin α2 1þ32cos φ sin 2 α.
2
2
It remains to prove that f ð1Þ ¼ cos φ2 cos α 1þ32cos φ sin 2 α, or
3ð1 þ cos φÞcos 2 α 4 cos φ2 cos α 2 cos φ 0.
We have that cos ~ cr ¼ cos α cos ðφ φ Þ. Thus, it follows that
q,~ 1
pffiffi
π φ
cos α cos 4 cos α cos 2 cos α cos ðφ φ1 Þ 3. Hence cos α 2 3 2. We have
2
h pffiffii
that t ¼ cos α 2 0; 2 3 2 . We deduce that gðtÞ ¼ 3ð1 þ cos φÞt2 4 cos φ2 t
pffiffi
2 cos φ max g 2 3 2 ; gð0Þ .
pffiffi pffiffi
Note that g(0) ¼ 2 cos φ 0, and g 2 3 2 ¼ 83 ð1 þ cos φÞ 8 3 2 cos φ2
pffiffi hpffiffi i
2 cos φ ¼ 43 cos 2 φ2 8 3 2 cos φ2 þ 2 0, because t ¼ cosφ2 2 22; 1 . Therefore,
160 4 Application of Vectors
(b) Yes, it is possible (see Figure 4.5). First one has to construct “such” tetrahedron
and then add another 96 faces. One should take AB ¼ CD, AD ¼ DB ¼ BC ¼ AC,
where AB is large enough.
4.1.21. Let points K0, L0, M0 be the midpoints of edges AB, CC0 , and A0 D0 ,
! ! !
e1 ¼ KK00M
respectively. Let us denote ~ 0 e ¼ M0 L0 ,~
M0 ,~2
L0 M 0
M0 L0 e3 ¼ L0 M0 (see Figure 4.6).
Note that
! ! !
! ! e þ ! e3 ¼ KA þ AA0 þ A0 M ~
KM~e1 þ ML ~2 LK ~ e1 þ
! ! ! ! ! !
þ MD0 þ D0 C0 þ C0 L ~ e2 þ LC þ CB þ BK ~e3 ¼
! ! ! ! !
¼ AA0~e1 þD0 C0~
e2 þ CB~e3 þ KA~ e1 þ BK~e3 þ
! ! ! !
þ A0 M~ e1 þ MD0~ e2 þ C0 L~ e2 þ LC~e3 ¼
! ! ! ! !
¼ AA0~
e1 þD0 C0~ e2 þ CB~ e3 þ BA~ e1 þ BK ð~
e3 ~
e1 Þ
!
! !
þ þA0 D0~ e2 þ A0 M ð~
e1 ~e2 Þ þ C0 C~e3
!
þ C0 L ð~
e2 ~e3 Þ:
ð4:15Þ
! ! !
Because BK ð~ e3 ~e1 Þ ¼ 0, A0 M ð~ e2 Þ ¼ 0, C0 L ð~
e1 ~ e2 ~
e3 Þ ¼ 0 (see the proof
! ! !
of problem 2.4.18а), then from (4.15) it follows that, the sum KM~ e1 þ ML~ e2 þ LK
~
e3 is constant, this means that
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
þmd ¼ 43 ðAG þ BG þ CG þ DGÞ 83 AG2 þ BG2 þ CG2 þ DG2 , because R2 ¼
! !2 ! !
AO2 ¼ AG þ GO ¼ AG2 þ GO2 þ 2 GO AG . We obtain that
! ! ! ! !
4R2 ¼ AG2 þ BG2 þ CG2 þ DG2 þ 4GO2 þ 2GO AG þ BG þ CG þ DG ¼
¼ AG2 þ BG2 þ CG2 þ DG2 þ 4GO2 AG2 þ BG2 þ CG2 þ DG2 :
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Thus, ma þ mb þ mc þ md 83 AG2 þ BG2 þ CG2 þ DG2 16 3 R (see prob-
lem 4.1.13).
! ! !
4.1.23. Let CA ¼ a ~i, CB ¼ b ~j, CD ¼ x ~i þ y ~j þ z ~ k, where ~i,~j, ~k are
! ! BD
coordinate vectors and z 6¼ 0. We have that
cos φ ¼ CA ACBD
x
¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi, where φ is the angle between lines AC and BD, as
x2 þ ðy bÞ2 þ z2
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
x2 þy2 þz2 ! !
AB ¼
CD p ffiffiffiffiffiffiffiffiffi and 0 ¼ AD BD ¼ xðx aÞ þ yðy bÞ þ z2 . We have to prove
a2 þb2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
that x2 a2 þ b2 < ðx2 þ y2 þ z2 Þ x2 þ y2 þ z2 þ b2 2by , or x2 a2 þ b2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
< ðax þ byÞ ax by þ b2 , b2(x2 + y2) < b2(ax + by), 0 < z2.
Now, it is not difficult to check that the conditions of problem 4.1.24a are
satisfied, where m1 ¼ m2 ¼ m3 ¼ m4 ¼ m5 ¼ m6 ¼ 1. Thus, it follows that
B1 B 2 þ B2 B 3 þ B3 B4 þ B4 B5 þ B5 B6 þ B6 B1
A1 A2 þ A2 A3 þ A3 A4 þ A4 A5 þ A5 A6 þ A 6 A1 :
4.1.25. (a) Let us consider midpoints M, N, P, K, F, E of edges AD, BD, CD, AB,
AC, BC, respectively.
As OM ⊥ AD, OK ⊥ AB, OF ⊥ AC, then AO > AM, AO > AK, and AO > AF.
According to problem 7.1.38a tetrahedron AMKF does not contain point O. Then,
we obtain that point O is inside of the polyhedron with faces MPF, MFK, FKE,
PNE, MNK, FPE, MPN, NKE. For point X 6 O draw a plane α passing through
!
point O and perpendicular to vector OX . It is obvious, that at least one of the points
M, N, P, K, F, E and point X are on the different sides of plane α. If that is the point
! !
M, then OM OX < 0.
! ! ! !
Let ~e1 ¼ OA ,~ e2 ¼ OB ,~e3 ¼ OC ,~ e4 ¼ OD and ~ e ¼~e1 þ~ e2 þ~ e3 þ~ e4 . We
OA OB OC OD
need to prove that
4.1 Application of Vectors for Proving Geometric and Trigonometric Inequalities 163
~
e < 2: ð4:16Þ
e 6¼ ~
Indeed, it is sufficient to prove inequality (4.16) in case ~ 0.
! ! !
Let OX ¼ ~ e, then as it was mentioned
OM OX < 0. Consequently,
ð~
e1 þ~ e < 0. We have that ~
e4 Þ~ e2~e ~
e and ~ e ~
e3~ e. Thus, ð~
e1 þ~
e2 þ
2
~
e3 þ~ e < 2~
e4 Þ~
e . Hence ~
e < 2~
e , then ~
e < 2. Note that
4.1.26. Let ABCD be a quadrilateral circumscribed around a circle. Given that the
opposite sides AB and CD, BC and AD are on the lines intersecting at points M and
N. Prove that cos ∠A + cos ∠B + cos ∠C + cos ∠D + cos ∠M + cos ∠N 2.
4.1.27. Let ABC be an arbitrary triangle. Prove that for any equilateral triangle
A1B1C1 the inequality A1 A2 þ B1 B2 þ C1 C2 16 AB2 þ BC2 þ CA2 p2ffiffi3 SABC
holds true.
4.1.28. Given points A1, . . ., An and positive numbers m1, . . ., mn. For any point
X let us denote by f(X) the expression m1A1X + . . . + mnAnX. Given that point M is
on segment AB, such that BM AB ¼ α. Prove that f(M ) αf(A) + (1 α)f(B).
4.1.29. Given that pointsA1, A2, . . ., An are on the same sphere, m1, . . ., mn > 0 and
! !
G is such a point that m1 GA1 þ ::: þ mn GAn ¼ ~ 0. Let straight lines GA1, . . ., GAn
intersect this sphere (for the second time) at points B1, B2, . . . , Bn. Prove that
m1 GB1P þ ::: þ mn GBnP m1 GA1P þ ::: þ mn GAnP , where 0 p 2.
164 4 Application of Vectors
DA þ DB !
Hint Let 2 ; AB ¼ α 6¼ 90 , then
! ! !
SADB ¼ 12 AB DA þ2 DB jctgαj ¼ 14 DA2 DB2 ctgα.
For α ¼ 90 see problem 1.1.14a (not only α ¼ 90 ).
4.1.32. Let in a tetrahedron ABCD the angles ADB, ADC, BDC be obtuse and the
lengths of edges AD, BD, CD be equal. Prove that ABC is an acute triangle.
! !
Hint Prove that CA CB > 0.
4.1.33. Prove that the difference of the squares of the lengths of the adjacent sides of
a parallelogram is less than the product of its diagonals.
4.1.34. Let n points be inside of the unit sphere. Prove that the sum of the squares of
distances between all possible pairs of these points does not exceed n2.
4.1.35. (a) Consider a convex hexagon, such that the length of each of its sides is
more than 1. Will there always be in it a diagonal with the length more than 2?
(b) Consider a convex hexagon ABCDEF, such that the lengths of the diagonals AD,
BE, CFare greater than 2. Will it always have a side with the length more than 1?
Hint See problem 4.1.4a.
4.1.36. Let M be the intersection point of the diagonals of an inscribed quadrilateral,
N be the intersection point of its midlines and O be its circumcenter. Prove that
OM ON (the midline is a segment connecting the midpoints of the opposite sides).
!
Hint Let ABCD be an inscribed quadrilateral, then ON ¼
! ! ! ! ! ! ! !
2
!2 OA þ OB þ OC þ OD .
4 OA þ OB þ OC þ OD , Prove that OM
1
4
4.1.37. Prove that for any triangle ABC the following inequalities hold true:
(a) a2 þ b2 þ c2 9R2 ; cos 2α þ cos 2β þ cos 2γ 32,
(b) a3 þ b3 þ c3 þ 3abc a2 b þ b2 a þ a2 c þ ac2 þ b2 c þ bc2 ; cos α þ cos β þ cos
γ 32,
(c) c2 a2 þ b2 þ R2 ; cos 2α þ cos 2β cos 2γ 32.
4.1 Application of Vectors for Proving Geometric and Trigonometric Inequalities 165
4.1.38. Let α1, α2, . . ., α6 be (the linear angles of) the dihedral angles of a
tetrahedron. Prove that cosα1 + cos α2 + . . . + cos α6 2.
4.1.39. Use problem 4.1.8a for x ¼ y ¼ z ¼ 13 to prove problem 4.1.4b.
! ! ! ! !
Hint Consider points X, Y, Z, and M, such that XY ¼ BC þ CD EF FA ,
! ! ! ! ! ! ! !
YZ ¼ FA þ AB DE CD , and XM ¼ BC EF .
4.1.40. The rectangular projection of a triangular pyramid to some plane has the
greatest possible area. Prove that this plane is parallel either to one of the sides of a
pyramid or to two skew edges of the pyramid.
4.1.41. (a) Given that in a tetrahedron ABCD the sum of the cosines of all plane
angles at vertex D does not exceed –1. Prove that for any point M, other than D,
holds true MA + MB + MC + MD > DA + DB + DC.
(b) Given that in a tetrahedron ABCD the sum of the cosines of all
plane angles at vertex D does not exceed –1. Prove that inside this
tetrahedron one can find a point M0, such that ∠AM0B ¼ ∠CM0D, ∠AM0C ¼
∠BM0D, ∠AM0D ¼ ∠BM0C.
Moreover, for any point M, other than M0, it holds true MA + MB + MC +
MD > M0A + M0B + M0C + M0D.
Hint See the proof of problem 4.1.17.
4.1.42. Let triangles A1A2A3 and B1B2B3 with orthocenters H1 and H2, respectively,
be inscribed in a circle of radius R. Prove that H1H2 < 4R + A1B1.
Hint Let G1 and G2 be the centroids of those triangles A1A2A3 and B1B2B3,
! 1 ! ! !
respectively. Prove that H1H2 ¼ 3G1G2 and G1 G2 ¼ 3 A1 B1 þ A2 B2 þ A3 B3 :
Chapter 5
Application of Trigonometric Inequalities
This chapter consists of five sections, that is, Sections 5.1, 5.2, 5.3, 5.4, and 5.5.
Many problems in geometry can be solved by applying trigonometry. In partic-
ular, many problems related to geometric inequalities can be deduced to trigono-
metric inequalities.
Section 5.1 mainly provides trigonometric inequalities concerning to angles of a
triangle. Let us emphasize few methods of proving such inequalities: note that, if
πβ πγ
α, β, γ are the angles of some triangle, then πα2 , 2 , 2 are the angles of some
acute triangle. Therefore, if some inequality holds true for angles α, β, γ of some
πβ πγ
triangle, then substituting these angles by πα 2 , 2 , 2 one can obtain a “new”
inequality for angles α, β, γ.
Moreover, if some inequality holds true for the angles of any acute triangle, then
from this inequality one can obtain a “new” inequality for the angles of any triangle.
In Section 5.1 many problems are proved using the maximal (minimal) values of
a quadratic polynomial. One of the crucial methods explained in this section is the
method based on the following statement: if the quadratic coefficient of a quadratic
function is positive (negative), then on any segment that function accepts its
maximal (minimal) value in one of the endpoints of the considered segment.
Section 5.2 selects such inequalities, concerning the angles of a triangle, that
hold true either only for acute triangles or only for obtuse triangles.
One of the most important methods of proving geometric inequalities is based on
modifications of trigonometric expressions. Therefore, Section 5.3 is devoted to
some important mathematical identities related to triangles.
Section 5.4 considers some trigonometric inequalities that are later on applied in
Section 5.5 in order to prove geometric inequalities.
Summarizing the above mentioned, Sections 5.1, 5.2, 5.3, and 5.4 provides
methods related to trigonometry in order to prove geometric inequalities. These
types of geometric inequalities are considered in Section 5.5.
Some problems in this chapter were inspired by [1, 2]. Nevertheless, even for
these problems the authors have mostly provided their own solutions.
Let α, β, and γ be the angles of some triangle. In the problems of this section it is
required to prove the following inequalities.
5.1.1. cos2 α þ cos2 β þ cos2 γ 34.
5.1.2. cosαcosβcosγ 18.
5.1.3. cos2α þ cos2β cos2γ 32.
5.1.4. (a) 1 < cosα þ cosβ þ cosγ 32,
pffiffiffi 2
(b) 3 cos α2 þ cos β2 þ cos 2γ 4 cos α2 þ cos β2 þ cos 2γ þ 2 cos α2 cos β2 cos 2γ .
pffiffiffi
(d) sin α2 þ sin β2 þ sin 2γ 12 sin α2 sin β2 sin 2γ þ 3 cos α2 cos β2 cos 2γ .
(e) cos 2 αβ
4 þ cos
2 βγ
4 þ cos
2 γα
4 2ð cos α þ cos β þ cos γ Þ,
αβ βγ γα
(f) cos 2 þ cos 2 þ cos 2 p2ffiffi3 ð sin α þ sin β þ sin γ Þ,
pffiffiffi
(g) cos α þ cos β þ cos γ þ ctgα þ ctgβ þ ctgγ 1, 5 þ 3,
(h) (1 cos α)(1 cos β)(1 cos γ) cos α cos β cos γ
(4 2 cos α 2 cos β 2 cos γ),
pffiffiffi
(i) sin 2α þ sin 2β þ sin 2γ 2 3ð cos α cos β þ cos β cos γ þ cos γ cos αÞ:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffi
jαβj jβγ j jγαj
5.1.21. sin 2 þ sin 2 þ sin 2 71þ17 17
32 .
5.1.22. (a) 2xy cos α þ 2yzcosβ þ 2xz cos γ x2 þ y2 þ z2, where α þ β þ γ ¼ π and
α, β, γ, x, y, z are arbitrary numbers.
cos α cos β cos γ
(b) þ sin
sin α1 β1 þ sin γ 1 ctg α1 þ ctg β1 þ ctg γ 1 , where α þ β þ γ ¼ π and
α1, β1, γ 1 are angles of some triangle.
(c) a2ctgα1 þ b2ctgβ1 þ c2ctgγ 1 4S, where a, b, c, S are respectively, the sides and
the area of some triangle, and α1, β1, γ 1 the angles of another triangle.
(d) (b2 þ c2)(1 cos φ) þ a2 cos φ 4S| sin φ|, where φ is any angle and S is the
area of the triangle with sides a, b, c.
(e) a2 b21 þ c21 a21 þ b2 a21 þ c21 b21 þ c2 a21 þ b21 c21 16SS1 , where a, b,
c, S are respectively, the sides and the area of some triangle, and a1, b1, c1, S1 of
another triangle.
(f) (xa2 þ yb2 þ zc2)2 16S2(xy þ yz þ zx), where x, y, z are arbitrary numbers and
S is the area of a triangle with sides a, b, c.
(g) k2tgα1 þ l2tgβ1 þ m2tgγ 1 S, where k, l, m are the distances from the circum-
center of triangle ABC to lines BC, AC, AB, respectively, S is the area of triangle
ABC, and α1, β1, γ 1 are the angles of some other acute triangle.
2
5.1.23. cos 2α2 þ cos 2 β2 þ cos 2 2γ sin α2 þ sin β2 þ sin 2γ .
πþα πþβ πþγ
5.1.24. sin 2α þ sin 2β þ sin 2γ cos 8 þ cos 8 þ cos 8 .
Solutions
1 þ 4 8 ¼ 2.
1 3
5.1 Inequalities for the Angles of a Triangle 171
or
pffiffiffi
3 α β γ 1 α β γ
cos þ cos þ cos 1þ sin þ sin þ sin 1 ,
4 2 2 2 4 2 2 2
α π β π γ π 3
cos þ þ cos þ þ cos þ ,
2 6 2 6 2 6 2
π αþβ γ þ π3 αþβ
þ γþπ=3 π
sin α þ sin β þ sin γ þ sin 2 sin þ 2 sin 4 sin 2 2
¼ 4 sin :
3 2 2 2 3
pffiffi
Hence, we obtain that sin α þ sin β þ sin γ 3 sin π3 ¼ 3 2 3.
This ends the proof.
πβ πγ
5.1.7. Using the inequality of the
pffiffi
problem 5.1.6 for angles πα
2 , 2 , 2 , we deduce
α β γ
that cos 2 þ cos 2 þ cos 2 2 .
3 3
follows that
γ π αþβþγ π
α β γ π αþβ þ þ
ctg þ ctg þ ctg þ ctg 2ctg þ 2ctg 2 6 4ctg 4 12 ¼ 4ctg π :
2 2 2 6 4 2 2 6
β γ
p ffiffi
ffi
Therefore, ctg α2 þ ctg 2 þ ctg 2 3ctg π6 ¼ 3 3.
This ends the proof.
πβ πγ
5.1.11. Using the inequality of the problem 5.1.2 for angles πα
2 , 2 , 2 , we obtain
that sin α2 sin β2 sin 2γ 18.
This ends the proof.
5.1.12. Note that
1 1 cos ðx þ yÞ
sin x sin y ¼ ð cos ðx yÞ cos ðx þ yÞÞ ¼
2 2
x þ y
¼ sin 2 :
2
5.1 Inequalities for the Angles of a Triangle 173
3
cos αcosβ þ cosβ cos γ þ cos γ cos α ¼
4
βþγ βγ 1 3
¼ 2 cos α cos cos þ ð cos ðβ γ Þ þ cos ðβ þ γ ÞÞ
2 2 2 4
α 1 3 1 α 2 α
2 cos α sin þ ð1 cos αÞ ¼ 2 sin 1 4 sin þ 3 0:
2 2 4 4 2 2
Similarly, we obtain that ctgβ þ ctgγ 2tg α2 and ctgγ þ ctgα 2tg β2.
Therefore, ctgα þ ctgβ þ ctgγ tg α2 þ tg β2 þ tg 2γ .
This ends the proof.
5.1.18. Note that cos α þ cosβ ¼ 2 sin 2γ cos αβ
2 2 sin 2γ . Thus, it follows that
cosα þ cos β cosβ þ cos γ cos γ þ cosα
cos α þ cosβ þ cos γ ¼ þ þ
2 2 2
γ α β
sin þ sin þ sin :
2 2 2
This ends the proof.
174 5 Application of Trigonometric Inequalities
5.1.19. We have that ctg 2 α þ ctg 2 β 12 ðctgα þ ctgβÞ2 2tg 2 2γ (see the proof of
the problem 5.1.17). Hence ctg 2 α þ ctg 2 β þ ctg 2 γ tg 2α2 þ tg 2 β2 þ tg 2 2γ .
This ends the proof.
5.1.20. (a) If the triangle is not acute angled, then inequality is correct, as
cos α cos β cos γ 0 < sin α2 sin β2 sin 2γq . ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
π pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
If x, y 2 0; 2 , then cos x cos y 1þ cos2ðxþyÞ ¼ cos xþy 2 .
But if the triangle is an acute triangle, then
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
cos α cos β cos γ ¼ cos α cos β cos β cos γ cos γ cos α
αþβ βþγ αþγ γ α β
cos cos cos ¼ sin sin sin :
2 2 2 2 2 2
Because
γ γ π 1 γ π πγ
f ð1Þ ¼ sin
cos 1 þ þ cos þ 2 sin
2 2 6 2 2 6 4
1 γ π πγ πγ πγ π
þ cos þ 2 sin ¼ 2 sin sin þ 1 0,
2 2 6 4 4 4 3
we have that f(1) 0.
(e) Let γ β α. One needs to prove that
3 1 αβ βγ γα
þ cos þ cos þ cos cos α þ cos β þ cos γ:
4 4 2 2 2
γ 1 βα 1 βα π 3γ 3
2 sin cos cos cos þ cos γ 0:
2 4 2 2 4 4 4
prove that
f cos βα4 0, as 0 cos βα 4 1 and for x 2 [0, 1].
We have that f(x) max( f(0), f(1)), then it is sufficient to prove that f(0) 0 and
f(1) 0.
Indeed, we have that
γ 1 1 γ γ 1 1 1
f ð0Þ ¼ cos γ 2 sin ¼ 2sin 2 2sin 2 2 2 < 0:
2 2 2 2 2 2 2 4
γ 1 π 3γ
f ð1Þ ¼ 2 sin 1 cos þ cos γ:
2 2 4
Remark For α þ β þ γ ¼ π the inequality of problem 5.1.20e may not hold true.
For example, if α ¼ π2 , β ¼ 5π
2 , γ ¼ 2π.
or
4 γ αβ π 3γ αβ 2 4 γ
2 pffiffiffi cos 1 cos 2 2 cos cos þ pffiffiffi sin γ pffiffiffi cos þ 1 0:
3 2 4 4 4 3 3 2
Let f ðxÞ ¼ 2 p4ffiffi cos γ 1 x2 2 cos π3γ
þ p2ffiffi3 sin γ p4ffiffi3 cos 2γ þ 1.
3 2 4 x
Note that
pffiffi
2 p4ffiffi3 cos 2γ 1 2 p4ffiffi3 cos π6 1 ¼ 2 and 22 cos αβ
4 1, as 0 αβ π
4 4.
We have that
pffiffiffi
αβ 2
f cos max f ð1Þ; f :
4 2
pffiffi
Let us prove that f 22 0 and f(1) 0, then (5.1) holds true.
We have that
pffiffiffi
2 2 pffiffiffi π 3γ 2 π pffiffiffi π 3γ
f ¼ pffiffiffi sin γ 2 cos pffiffiffi sin 2 cos ¼
2 3 4 3 3 4
pffiffiffi π 3γ pffiffiffi π pffiffiffi π
¼ 1 2 cos < 1 2 cos < 1 2 cos ¼ 0,
4 4 4
pffiffi
therefore f 22 < 0.
Now, let us prove that f(1) 0.
We need to prove that 1 þ 2 cos π3γ γ
4 3 cos 2 þ 3 sin γ.
p4ffiffi p2ffiffi
0 3 π 3γ 2 γ 2 2 γ 2
g ðγ Þ ¼ sin þ pffiffiffi sin pffiffiffi cos γ pffiffiffi sin pffiffiffi cos γ
2 4 3 2 3 3 2 3
2 π 2 π
pffiffiffi sin pffiffiffi cos ¼ 0:
3 6 3 3
π
Hence, gðγ Þ g 3 ¼ 0.
Hence, it follows that f(1) 0.
5.1 Inequalities for the Angles of a Triangle 177
We have that
2
2γ 2 γ
2γ 2 γ γ
0 γ x sin 2 2x cos 2 γ 1 sin 2 2 sin 2 cos 2
f ðxÞ ¼ 2 sin 2 < 2 sin 2 ¼
2 x2 sin 2 γ 2
2 x2 sin 2 γ 2
3γ γ γ
cos 2 sin cos 3 1
¼ sin γ 2 2 2
2 sin γ 2 0:
x2 sin 2 2γ x2 sin 2 2γ
Thus, it follows that for x 2 sin 2γ ; 1 , we have that f 0 (x) < 0. Therefore, f(x) is a
decreasing function on sin 2γ ; 1 .
αβ 3 pffiffiffi
Note that, it is sufficient to prove that f cos þ 3.
2 2
αβ γ i αβ
As cos 2 sin ; 1 , then f cos f ð1Þ.
2 2 2
3 pffiffiffi αβ 3 pffiffiffi
Let us prove that f ð1Þ þ 3, then f cos þ 3.
2 2 2
γ sin γ
We have that f ð1Þ ¼ 2 sin þ cos γ þ þ ctgγ.
2 1 sin 2 2γ
Let γ ¼ π 2φ, then 0 < φ < π3 pffiffiffi and one needs to prove that
2 cos φ cos 2φ þ 2ctgφ ctg 2φ 32 þ 3, or
3 þ tg 2 φ pffiffiffi 1 2
3 2 cos φ :
2tgφ 2
pffiffiffi 2 2
3 tgφ 4tgφ cos φ 12 ,
pffiffiffi 2 2
3 cos φ sin φ 4 cos φ sin φ cos φ 12 ,
2 2 pffiffiffi 2
ð3cos 2 φ sin 2 φÞ 2 sin 2φ cos φ 12 3 cos φ þ sin φ ,
1 pffiffiffi 2
ð2 cos φ 1Þ2 ð2 cos φ þ 1Þ2 sin 2φð2 cos φ 1Þ2 3 cos φ þ sin φ :
2
178 5 Application of Trigonometric Inequalities
1 pffiffiffi 2
ð2 cos φ þ 1Þ2 sin 2φ 3 cos φ þ sin φ :
2
pffiffiffi
The last inequality holds true, as (2 cos φ þ 1)2 4 and 1
2 sin 2φ 3 cos φþ
sin φÞ2 12 1 4 ¼ 2.
(h) Let α β γ, then we need to prove that
on [0; 1].
We need to prove that f cos αβ2 0.
We have that 0 < γ π3 , therefore
γ γ
f 0 ðxÞ ¼ 12 cos γ sin x2 þ 2 1 5 cos γ þ 2cos 2 γ x 2 sin 1 þ cos 2 γ ,
2 2
γ
f 0 ð0Þ ¼ 2 sin 1 þ cos 2 γ < 0,
2
γ
f 0 ð1Þ ¼ 2 sin 6 cos γ 1 cos 2 γ þ 2 10 cos γ þ 4cos 2 γ
2
6 cos γ 1 cos 2 γ þ 2 10 cos γ þ 4cos 2 γ ¼
¼ ð cos γ 1Þð3 cos γ 1Þ < 0:
Then, for 0 < x < π3 , it follows that gðxÞ > g π3 ¼ 0. Thus, it follows that
f ð1Þ ¼ 2 sin 2γ gðγ Þ > 0:
Note that, for 0 < x 1, f(x) min( f(0), f(1)) > 0. Hence, we obtain that
pffiffiffi
sin 2αþ sin 2β þsin 2γ 2 3ð cos α cos β þ cos β cos γ þ cos γ cos αÞ ¼
αβ αβ
¼ f cos > 0, as 0 < cos 1:
2 2
jα β j jβ γ j jγ αj π π
sin þ sin þ sin sin 2x þ sin x þ sin x ¼
2 2 2 2 2
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ sin x þ cos x þ cos 2x ¼ 1 þ sin 2x þ 1 sin 2x: 2
pffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi
Consider the function f ðtÞ ¼ 1 þ t þ 1 t2 on the interval [0; 1]. Since
pffiffiffiffiffiffi
f 0 ðtÞ ¼ 2p1ffiffiffiffiffi
1þt
ffi pffiffiffiffiffiffiffi
t
1t2
¼ 2p 1t 2t
ffiffiffiffiffiffiffi
1t2
¼ 2pffiffiffiffiffiffiffi4t2 þt1
pffiffiffiffiffiffi
1t2 ð 1tþ2tÞ
.
pffiffiffiffi q ffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffi
171
Consequently, max f ðtÞ ¼ f 8 ¼ 7þ 17
8 þ 23þ 17
32 .
½0;1
qffiffiffiffiffiffiffiffiffiffipffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffi
Thus, sin jαβ 2
j
þ sin jβγ j
2 þ sin jγαj
2 7þ 17
8 þ 23þ 17
32 ¼ 71þ17 17
32 .
pffiffiffiffi
Remark The given estimate is exact since at β arcsin 1781, we have that
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffi
sin jαβ
2
j
þ sin jβγ j
2 þ sin jγαj
2 71þ17 17
32 .
Note that
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffi
jα β j jβ γ j jγ αj 71 þ 17 17
sin þ sin þ sin < :
2 2 2 32
(b) Let x, y, z > 0 and 2xy ¼ sin1α1 , 2yz ¼ sin1β , 2xz ¼ sin1 γ , this means that
1 1
sin β1 sin γ 1 sin α1
x ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi, y ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi, z ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi. Then from the
2 sin α1 sin β1 sin γ 1 2 sin α1 sin β1 sin γ 1 2 sin α1 sin β1 sin γ 1
problem 5.1.22a, it follows that
5.1 Inequalities for the Angles of a Triangle 181
(c) Let the angles and circumradius of the triangle with sides a, b, c be equal to u, v,
w and R, respectively. Then
1 1 1 1
S ¼ R2 sin 2u þ R2 sin 2v þ R2 sin 2w ¼ a2 ctgu þ b2 ctgv þ c2 ctgw :
2 2 2 4
Therefore,
(g) Let points A1, B1, and C1 be the midpoints of sides BC, AC, and AB, respec-
tively. Using the problem 4.1.8а for x ¼ tgα1, y ¼ tgβ1, z ¼ tgγ 1, we obtain that
tg α tg β A B2 þtg β1 tg γ 1 B1 C21 þtg α1 tg γ 1 A1 C21
tg α1 OA21 þ tg β1 OB21 þ tg γ 1 OC21 1 1 1 1 tg α1 þtg β þtg γ ,
1 1
2 2 2
or k2 tg α1 þ l2 tg β1 þ m2 tg γ 1 ctg γ 1 c4 þ ctg α1 a4 þ ctg β1 b4 , as
tgα1 þ tgβ1 þ tgγ 1 ¼ tgα1tgβ1tgγ 1.
According to problem 5.1.22c, it follows that 14 a2 ctg α1 þ b2 ctg β1 þ
c2 ctg γ 1 Þ S.
Therefore, k2tgα1 þ l2tgβ1 þ m2tgγ 1 S.
This ends the proof.
5.1.23. In any triangle there are angles α and β, such that either α2 , β2 π6 or α2 , β2 π6,
then 1 2 sin α2 1 2 sin β2 0. Thus, it follows that
2
α β γ α β γ
cos 2 þ cos 2 þ cos 2 sin þ sin þ sin ¼
2 2 2 2 2 2
α β γ α β
¼ cos α þ cos β þ cos γ 2 sin sin 2 sin sin þ sin ¼
2 2 2 2 2
α β γ α β γ α β
¼ 1 þ 4 sin sin sin 2 sin sin 2 sin sin þ sin ¼
2 2 2 2 2 2 2 2
γ α β α β γ
¼ 1 þ sin 1 2 sin 1 2 sin 2 sin sin sin ¼
2 2 2 2 2 2
αβ γ α β
¼ 1 cos þ sin 1 2 sin 1 2 sin 0,
2 2 2 2
2
α β γ α β γ
cos 2 þ cos 2 þ cos 2 sin þ sin þ sin :
2 2 2 2 2 2
α β γ
sin α þ sin β þ sin γ cos þ cos þ cos : ð5:3Þ
2 2 2
In the same way from (5.3), we deduce that
π α π γ
α β γ π β
cos þ cos þ cos cos þ cos þ cos : ð5:4Þ
2 2 2 4 4 4 4 4 4
2
α 3α α α α
f ð1Þ ¼ 1 2 cos α þ 2 sin þ sin 3 sin ¼ 2 sin 1 2 sin þ 1 0:
2 2 2 2 2
This ends the proof.
5.1.29. We need to prove that, if 0 < x, y < π3, then
α β γ β γ þ α 2φ 4
tg þ tg þ tg tg þ tg þ ¼
2 2 2 2 2 3
π
2 sin φ 4 2 cos φ 4
¼ 2 þ þ ¼ 2:
β γ þ α 2φ π 3 1 þ sin φ 3
cos þ cos φ
2 2 2
Therefore, tg 2α2 þ tg 2 β2 þ tg 2 2γ 2.
This ends the proof.
5.1.31. We have that
Note that
cos 2α þ cos 2β
sin 2 α þ sin 2 β ¼ 1 ¼ 1 þ cos γ cos ðα βÞ
2
αβ
¼ 2 cos γcos 2 þ 1 cos γ
and 2
γ αβ
cos α þ cos β ¼ 2 sin cos :
2 2
γ αþβ αβ
Let f(x) ¼ Ax2 þ þ γC.
Bx Since sin 2 ¼ cos 2 cos 2 1 and A 0, then
max f ðxÞ ¼ max f sin 2 ; f ð1Þ ; thus, if we prove that the inequality (5.7)
½ sin 2γ ;1
holds true for cos αβ γ αβ
2 ¼ sin 2 and cos 2 ¼ 1, then it holds true (in the
general case).
If cos αβ γ
2 ¼ sin 2, then (5.7) has the following form:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 cos 2 γ 1 þ cos γ 1 cos ðα1 β1 Þ þ sin γ sin γ 1 2
or
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
sin γ 1 þ cos γ 1 cos ðα1 β1 Þ þ sin γ sin γ 1 2:
pffiffi
Similarly, to the proof given above in this case too it is sufficient to prove the
inequality (5.10) for cos α1 β π
2 ¼ 1. Then under the condition γ, γ 1 3 one has to
1
γ
If cos 2 þ γ21 1 þ 2 sin γþγ 1
4 cos γγ 1
4 > 0, then
γ γ1 γ þ γ1 γ γ γ þ γ1 γ þ γ1 γ γ1
D ¼ cos þ cos þ 4 sin sin 1 cos 1 þ 2 sin cos
2 2 2 2 2 4 4
γ þ γ1 γ γ γ þ γ1 γ þ γ1
1 þ cos þ 4 sin sin 1 cos 1 þ 2 sin
2 2 2 2 4
γ þ γ1 γ þ γ1 γ þ γ1 γ þ γ1
1 þ cos þ 2 1 cos cos 1 þ 2 sin ¼ D0 :
2 2 2 4
2
Then D0 ¼ 2cos2 γ20 þ sin 2 γ 0 4sin 2 γ20 1 sin γ20 , where γ 0 ¼ γþγ
2 .
1
2 γ0
γ0 2
As 2 D0 ¼ 2sin 2 1 2 sin 2 0, then D0 2. Hence, D 2.
This ends the proof of (5.11).
5.1.40. (sin2β þ sin2γ)(1 cos φ) þ sin2α cos φ 2 sin α sin β sin γ sin φ, where φ is
any angle.
5.1.41. sin2α þ sin2β þ sin2γ < 2(sinα sin β þ sin β sin γ þ sin γ sin α).
5.1.42. sin α2 þ β þ sin β2 þ γ þ sin 2γ þ α > sin α þ sin β þ sin γ.
5.1.43. (a) 1
sin α2 sin β2
þ 1
sin β2 sin 2γ
þ 1
sin α2 sin 2γ
12,
pffiffiffi
cos α2 þ cos β þ cos 2γ 2 3,
1 1 1
(b)
2
pffiffi
5.1.45. sin α sin β sin 2γ 2 9 3.
5.1.46. ctg2β ctgαctgγ, if 2sin2β ¼ sin2α þ sin2γ.
5.1.47. (a) cos α cos β cos ðα βÞ þ cos β cos γ cos ðβ γ Þ þ cos γ cos α cos ðγ αÞ
1 cos α cos β cos γ,
pffiffi
(b) sin α4 sin β4 sin 4γ þ cos α4 cos β4 cos 4γ 3 8 6,
α β γ
α β β γ γ α
4 sin 2 þ sin 2 þ sin 2 k sin 2 sin 2 þ sin 2 sin 2 þ sin 2 sin2 , where k 3,
(c) 63k 2
β γ β β γ γ
(d) sin α2 þ sin 2 þ sin 2 k sin α2 sin 2 þ sin 2 sin 2 þ sin 2 sin α2 63k 4 , where 0 <
k 27,
(e) sin α2 þ sin β2 þ sin 2γ þ k sin α2 sin β2 þ sin β2 sin 2γ þ sin 2γ sin α2 sin α2 sin β2 sin 2γ
32 þ 5k
8 , where k 5,
4
(f) 7 k α β γ α β β γ γ α where
þ sin þ sin þ sin þ k sin sin þ sin sin þ sin sin
5 2 2 2 2 2 2 2 2 2 2
4 α β γ
þ 2k sin sin sin
5 2 2 2
k 45.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
5.1.48. (a) sin α þ sin β þ sin γ 15 4 þ cos ðα βÞ þ cos ðβ γ Þ þ cos ðγ αÞ,
Let α, β, and γ be the angles of some acute triangle. Prove the inequalities of the
problems 5.2.1–5.2.8.
pffiffiffi
5.2.1. (a) tgα þ tgβ þ tgγ 3 3,
(b) tg 2α2 þ tg 2 β2 þ tg 2 2γ < 2.
5.2.2. (a) tgα þ tgβ þ tgγ ctg α2 þ ctg β2 þ ctg 2γ ,
Solutions
πβ1 πγ 1
5.2.1. (a) Note that, if α ¼ πα 2 , β ¼ 2 , γ ¼ 2 , then α1 þ β 1 þ γ 1 ¼ π and
1
π
Remark For α ¼ 2π
3 , β ¼ γ ¼ 6, we have that tgα þ tgβ þ tgγ ¼ 3.
p1ffiffi
Then,
α β γ
tg 2 þ tg 2 þ tg 2 ¼
2 2 2
2 α β β γ γ α
¼ tg α2 þ tg β2 þ tg 2γ 2 tg tg þ tg tg þ tg tg :
2 2 2 2 2 2
Remark If γ > π2, then tgα þ tgβ þ tgγ < 0 < ctg α2 þ ctg β2 þ ctg 2γ , (see the
problem 5.2.9).
(b) As tgα þ tgβ þ tgγ ¼ tgαtgβtgγ (see the proof of the problem 5.2.9) and
5.2 Inequalities for the Angles of Acute and Obtuse Triangles 193
then according to the problem 5.2.2a, we have that tgαtgβtgγ ctg α2 ctg β2 ctg 2γ .
Remark If γ > π2, then tgαtgβtgγ < 0 < ctg α2 ctg β2 ctg 2γ ,
5.2.3. (a) Note that
pffiffi
π
Remark For γ ¼ 2π
3 , α ¼ β ¼ 6, we have that sin α þ sin β þ sin γ ¼ 1 þ 2
3
< 2:
1 2
cos 2 α þ cos 2 β þ cos 2 γ ¼ cos α þ cos 2 β þ cos 2 β þ cos 2 γ þ cos 2 γ þ cos 2 α
2
1
ð3 cos α cos β cos γ Þ:
2
Therefore,
2
5.2.6. Note that tg 2 α þ tg 2 β ðtgαþtgβ
2
Þ
2tg 2 αþβ 2γ
2 ¼ 2ctg 2 (see the proof of the
problem 5.2.2a). Thus,
tg 2 α þ tg 2 β þ tg 2 γ ¼
1 1 1 γ α β
¼ tg 2 α þ tg 2 β þ tg 2 β þ tg 2 γ þ tg 2 α þ tg 2 γ ctg 2 þ ctg 2 þ ctg 2 :
2 2 2 2 2 2
Remark For angles β ¼ α, γ ¼ π 2α, where α is a sufficiently small number, the
inequality does not hold true.
5.2.7. We have that
Remark For α ¼ β ¼ π6 , γ ¼ 2π
3 the inequality does not hold true.
2
pffiffiffið cos α þ cos β þ cos γ Þ2 sin α sin β sin γ ¼
3
2
2 γ αβ γ αβ 8 γ αβ
¼ pffiffiffi 2 sin cos þ cos γ 2 cos cos sin γ ¼ pffiffiffisin 2 cos 2 þ
3 2 2 2 2 3 2 2
4 γ γ αβ 2
þ2 pffiffiffi sin cos γ cos cos þ pffiffifficos 2 γ sin γ:
3 2 2 2 3
2 γ 2 γ 2 3 γ γ
f ð1Þ ¼ pffiffiffi 2 sin þ cos γ 2 cos sin γ pffiffiffi 2 sin þ cos γ 2 cos sin γ,
3 2 2 3 2 2 2
5.2 Inequalities for the Angles of Acute and Obtuse Triangles 195
2
since 2 sin 2γ þ cos γ ¼ 32 2 sin 2γ 12 32. Thus, it follows that
pffiffiffi γ γ γ π π
f ð1 Þ 3 2 sin þ cos γ 2 cos sin γ ¼ 2 2 sin þ sin γ ¼
2 2 2 6 3
γ π γ π
¼ 4 sin 1 cos 0: Therefore; f ð1Þ 0:
2 6 2 6
We need to prove that f cos 2γ ¼ p2ffiffi3 ð sin γ þ cos γ Þ2 ð sin γ þ cos γ Þ 1 0.
pffiffiffi π
pffiffiffi
Indeed, we have that 1 ¼ sin 2 γþ cos 2 γ sin γ þ cos γ 2sin γ þ 2.
γ pffiffiffi 2pffiffi2 4
Thus, f cos 2 ¼ ð sinγ þ cosγ Þ pffiffi3ð sinγ þ cosγ Þ1 1 2 pffiffi3 1 1<0.
2
Since 0
αβ
γ , then we have that cos γ cos αβ 1, this means that
2 2 2 2
2
pffiffiffið cos α þ cos β þ cos γ 1Þ2 sin α sin β sin γ ¼
3
αβ γ
¼ f cos h max i f ðxÞ ¼ max f cos ; f ð1Þ 0:
2 γ 2
cos ; 1
2
Therefore,
2
sin α þ sin β þ sin γ pffiffiffi ð cos α þ cos β þ cos γ Þ2 :
3
αβ
ð4 cos γ 2Þcos 2
2
γ αβ
4 sin cos γ cos þ 1 cos γ þ sin 2 γ cos 2 γ 0: ð5:12Þ
2 2
Since 4 cos γ 2 0, then min f ðxÞ ¼ min f cos 2γ ; f ð1Þ , where f ðxÞ ¼
½ cos 2γ ;1
γ
ð4 cos γ 2Þx 4 sin 2 cos γx þ 1 cos γ þ sin 2 γ cos 2 γ:
2
2
We have that f ð1Þ ¼ 2 sin 2γ 1 cos γ 0 and f cos 2γ ¼ 1 sin 2γ 0.
Thus, the inequality (5.12) holds true.
5.2.9. We have that
α β γ α γ þ β 90 45 45
sin sin sin < sin sin sin 45 sin sin sin 45 ¼
2 2 2 2 pffiffiffi 2 2 2
1 cos 45 21 β γ 1 γβ βþγ
¼ sin 45 ¼ , since sin sin ¼ cos cos ,
2 4 2 2 2 2 2
sin γþβ90
2 sin 45 ¼ 12 cos α2 cos βþγ 2
sin α2 sin γþβ90
2 ¼ 12 cos 45 α cos 45 , sin 2452 ¼ 12 1 cos 45 ,
cos(45 α) 1, and 0 < α2 < γβ π
2 < 2. Therefore, cos
γβ
2 < cos α2.
5.2.13. Let γ > π2, then
4R2
,
2
þp2 4R2
(b) cos α cos β þ cos β cos γ þ cos γ cos α ¼ r 4R2
,
¼ pp2 þrð2Rþr
4R
2 2 2
(c) 1
cos α þ 1
cos β þ 1
cos γ Þ2
.
αβ βγ γα
¼ p þ2Rrþr
2 2
5.3.11. cos 2 cos 2 cos 2 8R2
.
5.3.12. tgα þ tgβ þ tgγ ¼ a2 þb2 þc
4S
2 8R2 ¼ 2
2pr
p ð2RþrÞ2
.
Þ2
.
5.3.14. tg α2 þ tg β2 þ tg 2γ ¼ 4Rþr
p .
5.3.15. sin 3 α cos β cos γ þ sin 3 β cos αcos γ þ sin 3 γ cos α cos β ¼
pr 2
¼ 4 2R p2 þ ð2R þ r Þ2 :
4R
βγ αγ αβ
5.3.16. sin 2 α cos 2 þ sin 2 β cos 2 þ sin 2 γ cos 2 ¼
2 2 2
ð2R þ r Þp þ 4Rr þ r
2 2 3
¼ :
8R3
200 5 Application of Trigonometric Inequalities
Solutions
5.3.1. (a) Taking into account the relation sin α þ sin β þ sin γ ¼ 2R
a
þ 2R
b
þ 2R
c
, we
obtain that sin α þ sin β þ sin γ ¼ R.
p
(b) Since sin α sin β sin γ ¼ 8R3 and 4R ¼ S ¼ pr, then sin α sin β sin γ ¼
abc abc pr
2R2
.
αþβ αβ γ γ
sin α þ sin β þ sin γ ¼ 2 sin cos þ 2 sin cos ¼
2 2 2 2
γ αβ αþβ α β γ
¼ 2 cos cos þ cos ¼ 4 cos cos cos :
2 2 2 2 2 2
Therefore,
α β γ 1 p
cos cos cos ¼ ð sin α þ sin β þ sin γ Þ ¼ ,
2 2 2 4 4R
5.3.5. Since
Hence,
1 p2 ð2R þ r Þ2
cos α cos β cos γ ¼ ð1 þ cos 2α þ cos 2β þ cos 2γ Þ ¼ ,
4 4R2
(see the problem 5.3.9a).
5.3.10. (a) We have that
2pr 4S
tgα þ tgβ þ tgγ ¼ ¼ :
p2 ð2R þ r Þ 2
a2 þ b þ c2 8R2
2
(see the problem 5.3.7 and the proof of the problem 5.3.12).
5.3.14. Note that
α β γ r r r
tg þ tg þ tg ¼ þ þ
2 2 2 pa pb pc
ab þ bc þ ac p2 r 2 þ 4Rr r þ 4R
¼ r¼ r¼ ,
ðp aÞðp bÞðp cÞ pr 2 p
(see the proof of the problems 5.3.1c, 5.3.9b and the problems 5.3.1b, 5.3.8b).
5.3.16. We have that
βγ αγ α β sin 2 α
sin 2 αcos 2 þ sin 2 βcos 2 þ sin 2 γcos 2 ¼ ð1 þ cos ðβ γ ÞÞ þ
2 2 2 2
sin β
2
sin γ
2
þ ð1 þ cos ðα γ ÞÞ þ ð1 þ cos ðα βÞÞ ¼
2 2
1 ð1 cos 2 αÞ cos β cos γ ð1 cos 2 βÞ cos α cos γ
¼ sin 2 α þ sin 2 β þ sin 2 γ þ þ þ
2 2 2
ð1 cos 2 γ Þ cos α cos β sin α sin β sin γ
þ þ ð sin α þ sin β þ sin γ Þ ¼
2 2
sin 2 α þ sin 2 β þ sin 2 γ cos α cos β þ cos β cos γ þ cos γ cos α
¼ þ
2 2
cosα cos β cos γ sin α sin β sin γ
ðcosα þ cos β þ cos γ Þ þ ð sin α þ sin β þ sin γ Þ ¼
2 2
p2 r2 4rR r2 þ p2 4R2 p2 ð2R þ r Þ2 R þ r pr p
¼ 2
þ 2
2
þ 2 ¼
4R 8R 8R R 4R R
ð2R þ rÞp þ 4Rr þ r
2 2 3
¼
8R3
(see the problems 5.3.1a, 5.3.1b, 5.3.6, 5.3.8a, 5.3.9b, 5.3.10b).
5.4 Trigonometric Inequalities 205
5.4.9. Prove that sin2α þ sin 2β þ sin 2γ þ sin 2δ 16 sin α sin β sin γ sin δ, if
α, β, γ, δ > 0 and α þ β þ γ þ δ ¼ π.
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
5.4.10. Prove that ab sin α þ cd sin β 2 ðp aÞðp bÞðp cÞðp d Þ, if a, b, c,
d, p a, p b, p c, p d are positive numbers, where p ¼ aþbþcþd 2 and a2 þ b2
2ab cos α ¼ c þ d 2cd cos β.
2 2
206 5 Application of Trigonometric Inequalities
β, γ and α þ β þ γ ¼ π,
(d) sin α2 sin β2 þ sin β2 sin 2γ þ sin 2γ sin α2 12 þ 2 sin α2 sin β2 sin 2γ , where α, β, γ > 0
and α þ β þ γ ¼ π.
5.4.12. Prove that cosα þ cos β 1, if α, β > 0 and 2α þ β π, 2β þ α π.
sin ðβþλαÞ sin β
5.4.13. Prove that sin ðαþλβÞ sin α, if 0 < α β, α þ β < π, and 0 λ 1.
0 < φ < γ β α, α þ β þ γ ¼ π.
5.4.19. Prove that
Solutions
5.4.1. Note that if x, y 2 0; π2 , then
thus
5.4 Trigonometric Inequalities 207
1 1 1
ctgα þ ctgβ þ ctgγ ¼ ðctgα þ ctgβÞ þ ðctgβ þ ctgγ Þ þ ðctgγ þ ctgαÞ
2 2 2
αþβ βþγ γþα
ctg þ ctg þ ctg :
2 2 2
5.4.5. We need to prove that ctg α21 þ ctg β21 þ ctg γ21 ctg α21 þ ctg β1 þγ 1 γ
2 þ
ctg 2γ ctg α2 þ ctg β2 þ ctg 2γ .
As α1 α and γ 1 γ, then β1 β (β1 ¼
6 β). Hence, β1 > β.
We have to prove that
β1 γ β þ γ1 γ γ
ctg þ ctg 1 ctg 1 þ ctg , ð5:14Þ
2 2 2 2
β1 þγ 1 β1 þγ 1
sin sin
β1 γ 1
2
β1 þγ 1
β1 þγ 1 2γ
2
:
cos 2 cos 2 cos 2 cos β1 þγ
2
1
α1 β þ γ1 γ α β
ctg þ ctg 1 ctg þ ctg , ð5:15Þ
2 2 2 2
that is
πγ πγ
sin sin
α1 þγβ1 γ 1
2
πγ
βα
2
πγ
:
cos 2 cos 2
cos 2 cos 2
Figure 5.1
1 1
ai ai
ai
5.4 Trigonometric Inequalities 209
Figure 5.2
an-1 a2
p-2an-1 p-2a2
p-2an p-2a1 a1
an
cos α2 þ . . . þ cos αn 1.
2 2
cosφ1 . . . cos φn increases and if we repeat this action not more than n 1 times,
we deduce that cos φ1 ::: cos φn < cos nπn.
If for some i we have that φi > π2, then at n 2, we obtain cos φ1 :::
cos φn < 0 cos nπn.
5.4.8. (a) Since
A ¼ cos α cos β cos γ ðtgα þ tgβ þ tgγ Þ ¼
¼ sin ðα þ βÞ cos γ þ cos α cos β sin γ,
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
then using the inequality α1 b1 þ α2 b2 a1 þ a22 b21 þ b22 , we obtain that
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
A ðsin 2 ðα þ βÞ þ cos 2 αcos 2 βÞðcos 2 γ þ sin 2 γ Þffi ¼
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
¼ sin 2 ðα þ βÞ þ ð cos ðα þ βÞ þ cos ðα βÞÞ2
4
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 4 3 1 2 2
sin 2 ðα þ βÞ þ ðj cos ðα þ βÞj þ 1Þ2 ¼ j cos ðα þ βÞj pffiffiffi:
4 3 4 3 3
Therefore,
αþβ γþδ αþβþγþδ
ctgα þ ctgβ þ ctgγ þ ctgδ 2ctg þ 2ctg 4ctg ¼ 4:
2 2 4
¼ 12 ðR 2r Þ2 0. Consequently, p2 32 R2 þ 10Rr þ r 2 .
πβ πγ
(c) Since πα
2 þ 2 þ 2 ¼ π, then according to the problems 5.4.11a and b, we
obtain that
πα πβ πγ πα πβ πγ
1 þ 4 cos cos cos cos þ cos þ cos
2 2 2 2 2 2
17 2 πα πβ πγ
þ cos cos cos ,
12 3 2 2 2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
(d) We have that 2 sin α2 sin β2 ¼ tg β2 sin αtg α2 sin β 12 tg β2 sin α þ sin βtgα2 ,
this means that
α β β γ α γ
2 sin sin þ 2 sin sin þ 2 sin sin
2 2 2 2 2 2
1 α 1 β 1 γ
tg ð sin β þ sin γ Þ þ tg ð sin α þ sin γ Þ þ tg ð sin α þ sin βÞ ¼
2 2 2 2 2 2
α βγ β αγ γ αβ
¼ sin cos þ sin cos þ sin cos ¼
2 2 2 2 2 2
α β γ
¼ cosα þ cos β þ cos γ ¼ 1 þ 4 sin sin sin ,
2 2 2
If α þ λβ2 π2, then cos β þ λα λβ λβ λα
2 cos α þ 2 , as α þ 2 β þ 2 < β þ α < π.
We have that sin 2 sin 2 and sin β sin α, consequently, at α þ λβ2 π2 the
λα λβ 1 1
Let α1 and α2 be constant numbers. Consider the expression cos sin α1 þ sin α2 , where
x1 cos x2
α1 > 0, α2 > 0, α1 þ α2 x1 þ x2 π, and x1, x2 0 and let this expression accept its
maximal value at x1 ¼ β1 and x2 ¼ β2.
Let α1 α2, then one can assume that β1 β2. Otherwise, we have that
β1 cos β2
cosβ1 < cos β2. Therefore, (cosβ1 cos β2)(sinα2 sin α1) 0 or cos sin α1 þ sin α2
cos β2 cos β1
sin α1 þ sin α2 .
If β1 ¼ 0, then
Lemma
then λ ¼ 1.
214 5 Application of Trigonometric Inequalities
We proceed the proof by contradiction argument. Let λ 6¼ 1, then one can assume
that λ < 1. Let α1 α2 . . . αn, then it follows from (5.17) that
β1 β2 . . . βn and β1 < α1 , :::, βn1 < αn1 π2.
Note that
sin ðβ1 þ β2 Þ ¼ λð sin α1 cos β2 þ sin α2 cos β1 Þ > λð sin α1 cos α2 þ sin α2 cos α1 Þ ¼
¼ λ sin ðα1 þ α2 Þ: