The Geometric Problems in the Bulgarian National

Competitions – Tendency to Development

Sava Grozdev, Svetlozar Doichev, Svetoslav Savchev

Bulgarian Academy of Sciences, Acad. G. Bonchev Str., BI.8

1113 Sofia,Bulgaria

51, Knjaz Batenberg Str., E, apt.125

6003 Stara Zagora,Bulgaria

58 A, Chehov Str.
1113 Sofia,Bulgaria

E-mail: savagroz@banmatpc.math.bas.bg

The Bulgaria National Olympiad is one of the oldest

national Olympiads in the world. It started during the
academic year 1949-50. Some of its main goals are the
following: to popularize Mathematics among rising
generations and to simulate its study; to active
teachers; to discover talented students and give them
possibility of expression; to ensure conditions for
future training of skilled scientists in the field of
Mathematics and its applications, but also in near
scientific domains; to select and prepare a strong and
competitive national team for the International
Mathematical Olympiad. In 1971 the first Spring
National Mathematical Competition took place with
similar goals. It was followed by the Winter National
Mathematical Competition with its first edition in 1982.
Now about 30 national and regional competitions in
Mathematics are organized in Bulgaria annually. The
National Olympiad and both the Winter and the Spring
Competitions are approved as the most important, the
most qualitative and with the most numerous

129
 participation. Moreover, they are used to select the
national teams for both the International and Balkan
Mathematical Olympiads.
This article examines the geometric problems from
these three competitions. The following features are
discovered: the proportion between synthetic and
analytic character is changing through the years and
the analytic one is leading at present; the genre
variety decreases gradually and the problem
formulation becomes more classical on account of it;
problem difficulty increases gradually; aspiration
appears to approach the ideas of the International
Mathematical Olympiad . To prove this idea about 35
problems are discussed by dividing them into two
groups. The first one contains 18 problems from the
80-ies, which accent synthetic character, genre variety
and some comparable easiness of the solutions. The
second group contains 17 problems from the 90-ies
with the accent of analytic character, classical
sounding and comparable difficulty of the solutions.
Examples are given using the idea of Ivan Ganchev to
present problems in systems which turns out to be
quite useful in simplifying solutions and developing
skills in problems solving.

Founded in the academic year 1949/1950, the Bulgarian national Olympiad

is among the oldest mathematical contests in the word. Some of its original
goals are popularizing mathematics at high-school level, discovering and
encouraging students of high mathematical ability, motivating teachers and,
later on, selecting and coaching a competitive team for the International
Mathematical Olympiad. The Spring National Mathematical Competition,
whose goals are similar, was organized for the first time in 1971. It was
followed by the Winter National Mathematical Competition, with its first
edition in 1982. There are about 30 national and regional mathematical contests
in Bulgaria at present. However, the three mentioned here remain the most
important ones, of highest quality and with most numerous participation. They
are also used in the process of team selection for both the International and the
Balkan Mathematical Olympiad.
The subject of this article is the geometry problems posed on these three
contests. Several dozens of problems are considered. We found it best to
classify them mainly with respect to their chronological order by the following

130
reason. It turns out that each stage of development of the Bulgarian
mathematical contests heavily influences their style and mathematical contents.

· Whereas synthetic solutions were dominant in the 1980ies, analytical

approaches prevail after 1990.

· The diversity of geometric settings, so easy to spot in the years 1980--

89, decreases more and more.

· The recent geometry posed on Bulgarian competitions is now of a flavor

much more classical than some 15 years ago.

· In return, the problems are much harder as mathematical challenges.

Following the IMO geometry tradition is the major concern of the
proposers.

Some reasons for and consequences of this state of matters are briefly
touched upon below. Our emphasis is on the most recent editions of the
contests which display some new trends of development.
Examples are given following Ivan Ganchev's idea to present problems in
systems. This approach turns out to be quite useful in simplifying solutions and
developing problem solving skills.

1 The 1980ies: The Fruits of a Strategy

The Bulgarian national olympiad went through substantial changes during the
years of its existence. Mass participation was the main objective in the 1950ies
and 1960ies. The style of problems proposed was basically the one present at
the university entrance exams. This started changing in the late 1960ies and the
early 1970ies.
A co-founder of the International Mathematical Olympiad, Bulgaria was lacking
the traditions of participating countries like Hungary and the former Soviet
Union, for instance. So a number of renowned mathematicians was concerned
with establishing a different kind of mathematical culture in the overall
approach to high-school students. Without changing the olympiad structure,

131
more and more extracurricular elements were being introduced into the
unwritten syllabus of the olympiad.
The general state educational policy was also changing at that time. Studying
mathematics was intensified in the regular schools, and specialized
mathematical schools were opened in all regional centers (1971-75).
Mathematical teachers of high ability created their own schools and circles of
talented students. The main incentive was a good performance at the olympiad,
but new forms of extracurricular activity came into being as well. The authority
of the Central Olympiad Committee increased considerably.
The results became visible already around 1976, when the Bulgarian team
came out fourth at the international olympiad. However, the entire system of
extracurricular activities reached maturity a little later-and the stable
performance of the national IMO team in 1981-1989 is merely one of the many
indications. We mention only the International Olympiad in Prague, 1984,
where Bulgaria achieved the second highest combined score among all
participants.
The work on running high-school mathematical activities was a well-
coordinated effort of the Ministry of Education and the Union of Bulgarian
Mathematicians, assisted by the Mathematical Department of the Sofia
University and the Institute of Mathematics and Mechanics of the Bulgarian
Academy of Sciences. The social prestige of the olympiad movement was
surprisingly high. The centralized higher education in Bulgaria honored
unconditionally the achievements of the IMO contestants, allowing them to
become university students without taking the (quite competitive) entrance
exams. A new factor was the Balkan Olympiad which started out in 1984. The
successes of the young Bulgarian mathematicians at the time in question were
due to a carefully developed orderly system, the opportunities for a broader
selection because of the numerous participants, to the enthusiasm of teachers in
many regional centers.
We pass on to the mathematical contents of the olympiad between 1981 and
1989. More exactly, let us take a closer look at its final rounds, the third and the
fourth. For one thing, the goal of encouraging mass participation was still
present as a codition sine qua non. On the other hand, the final rounds had to
complete a major part of the IMO selection in an evironment where the top
contestants' preliminary preparation greatly exceeded the one of a regular
participant. The typical result was a contest paper of difficulty comparable to
that of an IMO paper, but also reflecting the curriculum of the Bulgarian school.
For instance, solid geometry is practically missing at the international olympiad
but well represented at Bulgarian contests in the eighties.

132
 The following list of problem statements may seem excessive but it reflects
important features of the whole period.

Problem 1 (National Olympiad, fourth round, 1982) Determine the locus of

the centroids of the equilateral triangles whose vertices lie on the sides of a
given square.

Problem 2 (National Olympiad, fourth round, 1983) Find a square of

minimum side length which can contain 5 nonoverlapping unit circles.

Problem 3 (National Olympiad, fourth round, 1984)

In space, n + 1 points P1 , P2 ,..., Pn and Q are given, n ³ 4 , no four of which
are in the same plane. It is known that for each triple of distinct points
Pi , Pj and Pk one can find a point Pl such that Q is interior to the tetrahedron
Pi P j Pk Pl . Show that n must be even.

Problem 4 (National Olympiad, third round, 1985)

Five points in the plane have the following property: Among any 4 of them, one
can choose 3 which are the vertices of an equilateral triangle.
a) Prove that 4 of the points are the vertices of a rhombus with acute
angle 60 .

b) Find the number of equilateral triangles with vertices at the given

points.

Problem 5 (National Olympiad, third round, 1985)

The midpoints of the edges AB and CD of the tetrahedron ABCD and the center
of its inscribed sphere are on the same straight line. Prove that the center of the
circumscribed sphere of the tetrahedron lies on the same line.

Problem 6 (National Olympiad, fourth round, 1985)

In the triangle ABC, the angle Ð ACB = g is acute and AC ¹ BC . A point P
is taken on the median CM with the property that the bisectors of ÐPAC and
ÐPBC intersect at a point Q on CM . Find ÐAPB and ÐAQB .

Problem 7 (National Olympiad, fourth round, 1985)

133
A sphere with center O is inscribed in a quadrangular pyramid ABCDM ,
with base ABCD . The point O lies on the altitude MH of the pyramid. Each
of the planes, ( ACM ) , ( BDM ) and ( ABO ) divides the lateral surface of the
pyramid into two parts of equal area. The areas of the plane sections of the
pyramid with the planes ( ACM ) and ( ABO ) are in ratio ( 2 + 2 ) : 4 . Find
the angle between the planes ( ACM ) and ( ABO ) and the dihedral angle of the
pyramid with edge AB .

Problem 8 (National Olympiad, fourth round, 1986)

Find the least positive integer n for which there exist an n -gon P and a point A
in its interior such that, if a source of light is placed at A , no side of P will be
completely lit by the light. For the value of n obtained, prove that one can find
two points inside the n -gon P with the property that, if sources of light are
placed at them, all sides of P will be completely lit.

Problem 9 (National Olympiad, fourth round, 1986)

A regular tetrahedron T of unit edge is given. Find the maximum volume of a
cube contained in T and such that one of its diagonals lies on an altitude of the
tertahedron.

Problem 10 (National Olympiad, fourth round, 1987)

Let D be the set of all triangles inscribed in a given circle k , whose degree
  
angle measures are integers different from 45 , 90 and 135 . For each triangle
T Î D , define f (T ) as the triangle with vertices the intersections of k with the
altitudes of T (or with their extensions).

a) Prove that there exists a positive integer n such that for each triangle
T Î D at least two among the triangles T , f (T ) ,
f 2 (T ) = f ( f (T )),..., f n (T ) = f ( f n -1 (T )) are congruent.
b) Determine the least value of n such that a) holds true.

Problem 11 (National Olympiad, third round, 1988)

Construct a circle G through the vertices A and B of the triangle ABC such
that the common chord of G and the circumcircle of ABC has a given length
.

134
Problem 12 (National Olympiad, fourth round, 1988)
Each point of the (three-dimensional) space is colored either red or white. Prove
that there exists a tetrahedron whose vertices and centroid are of the same color.

Problem 13 (National Olympiad, fourth round, 1988)

Let A, B and C be noncollinear points. For each point D on the ray AC ® ,
denote by E and F the points of tangency of the incirle of the triangle ABD with
AD and BD . Prove that the line EF passes through a fixed point as D ranges
over AC ® .

Problem 14 (National Olympiad, third round, 1989)

Find all convex n -gons with the following property: There exist n - 2 diagonals
of the n -gon each of which bisects its area.

Problem 15 (Winter Competition, 1983)

The diagonals AC and BD of the cyclic quadrilateral ABCD intersect at E .
Denote by P , Q the feet of the perpendiculars from E to AD , BC , and by M ,
N the midpoints of AB , CD . Prove that the line MN is the perpendicular bisector
of the line segment PQ .

This list provides ample evidence about an impressive genre diversity.

Together with traditional classical settings (problems 6,13,15), we encounter
instances of geometric loci (problem 1), geometric extrema in the plane and in
space (problems 2, 9), Euclidian constructions (problem 11), classical solid
geometry (problems 5,7), combinatorial geometry in the plane (problems
4,8,14) and in space (problems 3,12), even a combination of plane geometry and
divisibility (problem 10).
Since our attention is focused mostly on classical geometry, we present
solutions to problems 6 and 15.

In Problem 6, the bisector property

gives
AP BP QP
= = .
AC BC QC
Hence A and B belong to the locus
of points X such that the ratio
XP : XC is equal to QP QC. It is

135
 well-known that this locus is a circle
k with diameter on the line PC
(the Apollonius circle of the line
segment PC ), and Q is an endpoint
of this diameter. Denote the other
endpoint by R . Since QR divides the
chord AB of k into two equal parts and is not perpendicular to AB (because
AC ¹ BC , AB is a diameter of k , too. Therefore ÐAQB = 90 . Now a standard
computation yields the value of ÐAPB. We have
ÐAPB = ÐACB + ÐPAC + ÐPBC = g + 2(ÐPAQ + ÐPBQ )
= g + 2( ÐAPB - ÐAQB) = g + 2ÐAPB - 180 ,
so ÐAPB = 180 - g.

In Problem 15, it suffices to prove

that P and Q are equidistant from M
(by symmetry, the same will hold for
N , implying the desired conclusion).
To this end, take the midpoints U and
V of AE and BE , respectively. We
claim that DMPU and DQMV are
congruent. Indeed, PU and QV are
medians to the hypotenuses of the right
triangles AEP and BEQ , hence
PU = UE and QV = VE .
Next, ÐCAD = ÐCBD by inscribed angles, so the right triangles just
mentioned are similar, which yields ÐPUE = ÐQVE . Finally, EUMV is a
parallelogram, so ÐEUM = ÐEVM and UE = MV , VE = MU . It follows now
that PU = MV, UM = VQ, ÐPUM = ÐMVQ ,and the proof is complete.

136
 The proofs above are representative. Many classical geometry problems from
the 1980ies do admit of analytic solutions. However, the very intention of the
proposers and the jury was to promote synthetic methods in the spirit of the
ancient Greek tradition. It was unlikely for a problem proposal to make a
contest paper unless a synthetic proof was available in advance. Computational
solutions were of course developed by some contestants. But such solutions
were mostly regarded as an expression of poor geometrical thinking and not
encouraged too much. As a rule, the proofs known to the jury were short and
not particularly hard. Once a contestant reached a key idea, he or she was able
to prepare a neat writeup in a limited amount of time.

Summing up, some major features of the geometry problems from the 1980ies
were the following:

· synthetic approaches prevailing;

· great genre diversity;
· concise and relatively easy solutions.

2 The 1990ies: Fall and Rise

Bulgaria was hit by a severe overall crisis in the early 1990ies. The orderly
system of extracurricular mathematical activities, which seemed to be
established once and for all, suffered many blows and finally collapsed. In
particular, the Ministry of Education gradually resigned from its obligations to
mathematical contests at all levels. Financing was extremely scarce, if available
at all. A multitude of dubious educational reforms decreased the intensity and
the level of teaching high-school mathematics to a critical degree. The
priviledges of the olympiad winners to enter the universities without entrance
exams were abolished.
The Union of Bulgarian Mathematicians was the institution to go on with the
mathematical olympiads despite the extremal conditions. Without state funding
and by many other reasons, it was already an achievement to save the major
contests. However, preserving the entire movement with the same goals and
structure as before was a fiction. Mass participation was unthinkable not only
by trivial financial considerations. Most regional centers with established
traditions had ceased to exist. Neither regular students nor their teachers had
incentives and motivation to continue what had been started some two decades

137
before. The base for selecting the national team was thus drasctically reduced to
the two major mathematical schools in the capital Sofia.
The limited task of preserving Bulgaria's positions at the international level
was probably the only reasonable choice in that environment.
Once the Union took over all the responsibility with respect to the olympiad,
certain changes in the spirit of the competitions occurred. The main goal was
already selecting a strong national team. This circumstance influenced in
particular the geometry on the national contests. Here are examples of geometry
problems from this period.

Problem 16 (National Olympiad, third round, 1996)

Let ABCD be a convex quadrilateral satisfying ÐABC + ÐBCD < 180 . The lines
AB and CD intersect at E . Prove that ÐABC = ÐADC if and only if
AC 2 = CD.CE - AB. AE .

Problem 17 (National Olympiad, third round, 1998)

The convex quadrilateral ABCD satisfies AD = CD and ÐDAB = ÐABC < 90 .
The line through D and the midpoint of BC meets the line AB at E . Prove
that ÐBEC = ÐDAC .

Problem 18 (Spring Tournament,2001)

Let AA1 and BB1 be altitudes of the obtuse nonisosceles triangle ABC . Denote
by O and O1 the circumcenters of the triangles ABC and A1B1C , respectively. A
line through C intersects the line segments AB and A1B1 at D and D1 ,
respectively, and E is a point on the line OO1 such that ÐECD = 90 . Prove that
EO1 : EO = CD1 : CD .

Problem 19 (National Olympiad, third round, 2001)

The diagonals of the cyclic quadrilateral ABCD intersect at E . Prove that if
ÐBAD = 60 and AE = 3CE then the sum of two of the sides of the
quadrilateral is equal to the sum of the other two.

Problem 20 (National Olympiad, fourth round, 2001)

In a nonisosceles triangle ABC , the points of tangency of the incircle G with
the sides AB , BC and CA are C1 , A1 and B1 , respectively. Let

138
AA1  G = A2 , BB1  G = B2 , and let A1 A3 , B1B3 be bisectors in the triangle
A1B1C1 ( A3 Î B1C1, B3 Î A1C1 ) . Prove that:

a) A2 A3 is the bisector of ÐB1 A2C1 ;

b) If P and Q are the common points of the circumcircles of the
triangles A1 A2 A3 and B1B2 B3 then the incenter of the triangle
ABC lies on the line PQ .

Problem 21 (Spring Tournament,1998)

Let I and r the incenter and the inradius of the triangle ABC , respectively,
and let N be the midpoint of its median through C . Prove that if r = CN - IN
then AC = BC or ÐACB = 90 .

Problem 22 (National Olympiad,third round,1997)

The quadrilateral ABCD is cyclic. Let F be the common point of its diagonals
AC and BD , and E the common point of the lines AD and BC . If M and N
are the midpoints of AB and CD , prove that
MN 1 AB CD
= - .
EF 2 CD AB

Problem 23 (Winter competition,2001)

The points A1, B1 and C1 are taken on the sides BC , CA AB of the triangle
ABC , respectively. The point G is the centroid of DABC . Let Ga , Gb , Gc be the
centroids of DAB1C1, DBA1C1, DCA1B1 , respectively, and G1, G2 the centroids of
DA1B1C1 , DGa Gb Gc . Prove that:

a) the points G , G1 and G2 are collinear;

b) the lines AGa , BGb and CGc are concurrent if and only if the lines AA1 ,
BB1 CC1 are concurrent.

Problem 24 (Spring Tournament,2000)

The point A1 is chosen on the side BC of the triangle ABC so that the
triangles ABA1 and ACA1 have congruent incircles. Denote the diameters of
these circles by d a . The diameters db and dc are defined analogously. Let

139
 BC = a , CA = b, AB = c, a + b + c = 2 p . If ha , hb , hc are the altitudes through
A, B, C , respectively, and d is the diameter of the incircle of ABC , prove
that:
a) d a + p( p - a ) d = ha ;
a
b) d a + db + d c + p ³ ha + hb + hc .

Problem 25 (Winter competition,1997)

A triangle ABC is given whose angles with vertices A and B are greater than
or equal to 60 .Let BL be the bisector of ÐABC and H the foot of the
altitude through A. Find ÐAHL if ÐBLC = 3ÐAHL .

Problem 26 (Winter competition,2000)

Given is a convex quadrilateral ABCD whose diagonals meet at M. It is known
that DB = 3DM and AM = MC .

a) Express BC and CD in terms of the sides of the triangle ABD .

b) Prove that if 2ÐADB - ÐABD = 180 then ÐDBC = 2ÐBDC .

Problem 27 (Winter competition,1999)

Let O and R be the circumcenter and the circumradius of the triangle ABC .
The incircle of this triangle is tangent to BC , CA and AB at A1, B1 and C1 ,
respectively. The lines determined by the midpoints of the line segments AB1
and AC1, BC1 and BA1, CA1 and CB1 intersect at A2 , B2 and C2 . Prove that the
triangle A2 B2C2 has circumcenter O and circumradius R + r .
2

Problem 28 (Spring Tournament,1996)

^
The point D lies on the arc BC of the circumcircle of the triangle ABC which
does not contain A, and D ¹ B, D ¹ C . Points E and F are taken on the rays
BD® and CD ® so that BE = AC and CF = AB . Let M be the midpoint of the
line segment EF .
a) Prove that ÐBMC is a right angle.
^
b) Find the locus of the points M as D ranges over the arc BC .

140
We present two solutions of Problem 17 now, a synthetic and a computational
one.

Naturally, a synthetic approach

would not rely on the initial
configuration alone. Additional
constructions are needed for a purely
geometric solution. To begin with,
observe that if AD and BC meet at
N then AN = BN . Also, extend EC
to meet AN at P .
Denote ÐACD = ÐDAC = a and
BAC = b .Then ÐABC = a + b , and
the desired equality ÐBEC = a is
equivalent to ÐBCE = b , or to
ÐNCP = b .
On the other hand, an easy angle computation shows that ÐDCN = 2b . Hence
the problem reduces to proving that CP is the bisector of DCN .

Note that even after these key steps one may very well get stuck here. To move
on, we have to realize that ratios of line segments are likely to help rather than
angle computations. More exactly, a repeated application of Menelaus's theorem
yields the conclusion. Consider DDMN and the line AB first, then DDMN
again and the line CE . By Menelaus's theorem,
DE MB NA DE MC NP
. . = . . = 1.
EM BN AD EM CN PD
Now, taking the equalities AN = BN , BM = CM and AD = CD into account,
we infer that CD CN = PD PN . Hence CP is the bisector of ÐDCN , as
needed.

The problem is by far not among the hardest ones posed on Bulgarian
competitions. Still, a solution like above demands ingenuity and may be a
matter of chance. On the other hand, a straightforward computational proof
takes no inventiveness at all.

Let ÐACD = ÐDAC = a, ÐBAC = b ,

as in the previous proof, and let

141
 ÐAED = u , ÐCDE = v .Then
ÐADC = p - 2a, ÐDCE = p + a - b - u - v
Apply the law of sines for DADE and
DCDE first. In view of AD = CD , we
obtain

sin u AD CD sin v
= = = .
sin(a + b) DE DE sin(b - a + u + v)
Next, we have ÐBCE = a + b - u - v . Since BM = CM , the law of sines for
DBME and DCME gives
sin u BM CM sin v
= = = .
sin(a + b) ME ME sin(a + b - u - v)
Hence
sin(a + b) sin u sin(a + b)
= = ,
sin(b - a + u + v) sin v sin(a + b - u - v)
and it follows that sin(b - a + u + v) = sin(a + b - u - v) . This can be rewritten
as cos b sin(u + v - a) = 0 . Since b is acute, we easily conclude that
u+v= a.

Problem 22 is an appropriate representative of the computational aspect of the

period. The triangles ABE and CDE are similar, because ABCD is cyclic.
Hence AB = EA = EB = l , a positive number which is not 1 (or else AD and
CD EC ED
BC would not intersect). This initial observation already provides the base of a
vector solution for which no figure is necessary.
 ®
Let EC = c, ED = d . Define i as the unit vector collinear with EC and
 ®
having the same direction, that is, i = 1 EC .Likewise, let j = 1 ED . It is now
®

c d
® ®
a standard (but somewhat tedious) task to express EF and MN as linear
  ®  ® 
combinations of i and j .Note first that EA = lcj and EB = ldi . Since F
lies on both AC and BD , there are numbers x and y such that
® ® ®  
EF = x EA+ (1 - x) EC = (1 - x)ci + xlcj ,
® ® ®  
EF = y EB + (1 - y) ED = yldi + (1 - l)dj .

142
  
Equating the coefficients of i and j and eliminating y yields ld - c .
x=
(l2 - 1)c
Therefore
® l
EF = [(lc - d )i + (ld - c) j ].
l2 - 1
On the other hand,
® 1 æ® ® ö
1 æ® ® ® ®ö
MN = çç AD + BC ÷÷ = çç ED - EA+ EC - EB ÷÷
2è ø 2è ø
= [(c - ld )i + (d - lc) j ] .
1  
2
1 2
It is clear now that the ratio of MN 2 and EF 2 equals the ratio of and l ,
4 l2 - 1
that is,
2 2
MN 2 1æ 1ö 1 æ AB CD ö
= çl - ÷ = ç - ÷ .
EF 2 4 è lø 4 è CD AB ø
The claim follows.

Standard as they appear to be, these two third-round problems were solved by
very few contestants. They are examples of seemingly standard configurations,
rich in connections and well-known properties but objectively hard to be
handled synthetically. We presented a synthetic proof of Problem 17 above, and
should add that Problem 22 admits of such a proof as well (far and away harder
to devise).However, it is exactly searching for a synthetic approach and
estimating the difficulty arising that confirms a general claim.The effort to
argue synthetically is neither proportional to the theoretical significance of the
question nor reasonable in the conditions of a time constraint.
The majority of the problems 15-28 are similar.
Not only is their average difficulty higher than the one of Problems 1-15. Their
spirit is different. It was a conscious, continuous effort to impose such a spirit.
There is no pursuit of diversity any more, although no topic is ever permanently
closed. A brief look at the problem statements says enough: these are
exclusively questions in classical geometry, and questions intentionally
difficult. Writing down a solution alone may take quite a while, the exposition
is prone to computational errors of any kinds. Consequently, it happens that no
contestant solves a certain geometry problem sometimes-and the hardest
problem on a contest in Bulgaria is almost never a geometry one.

143
 The resemblance with most IMO style geometry proposals is fairly evident,
which was the intention of the proposers and the leaders of the national team.
Purely geometric proofs are neither prohibited nor impossible to find in
principle. Only the time is long gone when the geometry IMO problems were
the hardest challenge for the Bulgarian participants, on which they could afford
spending a long time during the contest. Arguing synthetically is welcome, if
the insight comes quickly. On the other hand, being on the safe side becomes
increasingly more important. By safe we mean being able to carry out
reasonable amounts of standard computations to get the solution to a geometry
problem as fast as possible. Because the real challenges to decide the IMO golds
are still to be handled, and these are rarely geometrical questions.
The key characteristics of the Bulgarian geometry problems in the 1990ies are
therefore the following:

· analytic approaches prevailing;

· settings almost exclusively classical;

· solutions relatively hard and difficult to expose.

3 The Future: New Trends and a Cautious Hope

The period of the late 1990ies was a relatively stable one for the olympiad
movement. With a lot of effort, the Union of Bulgarian Mathematicians
managed to establish a strict organization of all major initiatives, for proper
selection and coaching the national team. New and encouraging signs can be
observed. Exchange of valuable experience started with Russian and Chinese
colleagues as well as with co-workers from the Balkan countries. This includes
inviting teams from abroad to Bulgarian contests, exchanging visits, and so on.
Forms for early selection of talented young mathematical minds play an
increasingly important role. Such are the Junior Balkan Olympiad and the
annual junior contest in Hong Kong. The latter is regarded by many as
something of a junior IMO despite the differences in the format. Bulgaria is an
active participant in both of these since their beginning.
The Union of Bulgarian Mathematicians resumes publishing appropriate
books and materials for extracurricular mathematical work. Slowly and with
great difficulty, several regional mathematical centers outside Sofia gradually
revive their previous activity. Some universities agreed to accept IMO
participants without entrance exams again, as before. Finding sources of
funding is a constant major concern. Though randomly, foundations would

144
provide the means for certain indispensable needs like travel expenses. Private
and corporate sponsorship is modest but present.
Unfortunately, the Ministry of Education remains impassive to the mathematical
olympiad.
This is a brief outline of the atmosphere in which yet another tendency was
conceived. To justify our observations, we continue with several more geometry
problems.

Problem 29 (National Olympiad,fourth round,1997)

Let G be the centroid of the triangle ABC . Prove the inequality
2
sin ÐCAG + sin ÐCBG £
3
a) if the circumcircle of the triangle ACG is tangent to the line AB;
b) for an arbitrary triangle ABC .

In case the question were naturally stated, it

would have read:
Determine the maximum value of the sum
sin ÐCAG + sin ÐCBG , where G is the centroid
of the triangle ABC .
However, this problem was found too hard by the jury, so a hint for the key step
was explicitly provided in the problem statement. But why is the special case in
a) crucial?
Suppose that DABC is arbitrary, and let M be the midpoint of AB . There
are two circles passing through C and G which are tangent to the line AB . Let
the corresponding points of tangency be A1 , B1 ,on the rays MA® , MB ® ,
respectively. Since MA12 = MG.MC = MB12 by the power-of-a-point theorem
and, in addition, CG : GM = 2 : 1, G is the centroid of A1 B1C as well.
Moreover, A and B are exterior to the two circles unless the two triangles
coincide. It is straightforward now that ÐCAG £ ÐCA1G , ÐCBG £ ÐCB1G .
Thus, assuming Ð CA1G and ÐCB1G acute and the special case a) already
settled, we have
2
sin ÐCAG + sin ÐCBG £ sin CA1G + sin CB1G £ .
3
Thus we are left with the proof of a) and with the subcase of b) where one of
ÐCA1G and ÐCB1G is right or obtuse.

145
 So, let the circumcircle of DACG
touch AB . We use the standard
notation for the elements of DABC .
By the power-of-a-point theorem and
the well-known median formula,

c2 1 1
= MA2 = MG. MC = mc2 = (2a 2 + 2b 2 - c 2 ),
4 3 12
yielding a 2 + b 2 = 2c 2 . The median formula again gives m =
3
b, mb =
3 .
a
a
2 2
Then

2[ ACG ] 2[BCG ] [ ABC ] [ ABC ]

sin ÐCAG + sin ÐCBG = + = +
AC. AG BC.BG bma amb
(a 2 + b 2 ) sin g
= .
3ab
The law of cosines, combined with a 2 + b 2 = 2c 2 , implies a 2 + b 2 = 4ab cos g .
Therefore sin ÐCAG + sin ÐCBG = 2 sin 2g £ 2 , and a) follows.
3 3
Finally, let, for instance, ÐCA1G ³ 90  ; then
ÐCB1G is acute. Denote by
a 1 , b1 , c1 and g 1 the side lengths of DA1 B1C and ÐA1CB1 , respectively. We
obtain from DCA1G that CG 2 > CA12 + A1G 2 , that is,
1 1
(2a 12 + 2b12 - c12 ) > b12 + (2b12 + 2c12 - a 12 ).
9 9
2 2 2
By a), we have a 1 + b1 = 2c1 , and the above inequality takes the form
a 12 > 7b12 . Now set x = b12 a 12 . The argument in the proof of a) also gives

2[B1CG ] b1 sin g 1
sin ÐCB1G = =
B1C . B1G a1 3
2
b1 æ a 12 + b12 ö
= 1- ç ÷ = 1 14 x - x 2 - 1 = f ( x).
ç ÷
a1 3 è 4a 1b1 ø 4 3
Since x < 1 7 , it follows that f ( x) < f (1 7) = 1 7 . Therefore

146
 1 2
sin ÐCAG + sin ÐCBG < 1 + sin ÐCB1G < 1 + < .
7 3

Problem 30 (National Olympiad,fourth round,1999)

The vertices of a triangle have integer coordinates and one of its sides has
length n , where n is a square-free positive integer. Prove that the ratio of the
circumradius and the inradius of the triangle is irrational.
The setting suggests that the question can probably be quickly reduced to
divisibility considerations with a minimum use of geometry, which is indeed the
case at first glance. However, this impression is deceitful. It proves hard (or
impossible) to reach the conclusion without switching back to geometry at a
certain point.
Let R , r and S be the circumradius, the inradius and the area of the triangle,
respectively. It is not hard to observe that S is rational.
Assume that the ratio R r is rational. Without loss of genereality, let one
endpoint of the side with length n be at the origin, and let the remaining two
vertices have coordinates ( x, y) and ( z, t ) , with x, y, z, t integers. Then the
side lengths of the triangle are of the form a = A, b = B , c = C , where
A = n = x 2 + y 2 , B = z 2 + t 2 , C = ( x - z) 2 + ( y - t ) 2 .By hypothesis, the number
R abc a + b + c abc(a + b + c)
= .q = =
r 4R 2S 8S 2
is rational. Then so is also ABC ( AB + BC + CA) = abc (a + b + c ) = 8qS 2 ,
because S is rational. Upon squaring A BC + B CA = 8qS 2 - C AB , we
readily observe that AB is rational; by symmetry, the same holds for BC
and CA. Hence AB, BC, CA are perfect squares. Let
2 2 2
A = a 1 a 2 , B = b1 b2 , C = c1 c 2 , where a 2 , b2 , c 2 are square-free. Since
AB = (a1b1 )2 a 2 b2 , it follows that a 2 = b2 . Likewise, b2 = c 2 , so
A = ma 12 , B = mb12 , C = mc12 for some square-free positive integer m . Now,
A = n = ma 12 and n is square-free, which gives m = n, a 1 = 1 . Thus

x 2 + y 2 = n, z 2 + t 2 = nb12 , ( x - z) 2 + ( y - t ) 2 = nc12 .

It appears that a little more work based on the above equalities will yield the
desired contradiction. However, one more geometric consideration is needed.
Namely, we proved so far that the three side lengths are

147
a = n , b = nb1 , c = nc1 . Then 1 + b1 > c1 and 1 + c1 > b1 by the triangle
inequality, implying b1 = c1 . This last observation proves critical indeed, since
one can write x 2 + y 2 = n, z 2 + t 2 = nd 2 , ( x - z) 2 + ( y - t ) 2 = nd 2 for some d .
These yield n = x 2 + y 2 = 2( xz + yt ) . Now the well-known Lagrange's
identity
( x 2 + y 2 )( z 2 + t 2 ) = ( xz + yt ) 2 + ( xt - yz) 2

suggests considering the number k = 2( xt - yz) . We obtain

[ ]
n 2 + k 2 = 4 ( xz + yt ) 2 + ( xt - yz) 2 = 4( x 2 + y 2 )( z 2 + t 2 ) = 4n 2 d 2 ,
or k 2 = n 2 ( 4d 2 - 1) . But then 4d 2 - 1 turns out to be a perfect square, a
contradiction.

Problem 31 (Spring Tournament,2000)

There are given n ³ 4 points in the plane such that their pairwise distances are
all integers. Prove that at least 1 of these distances are divisible by 3.
6
We first prove the statement for n = 4 : if A, B, C , D are four points such that
their pairwise distances are integers then at least one of these 6 distances is a
multiple of 3. Assume on the contrary that this is false. The square of an integer
not divisible by 3 is congruent to 1 modulo 3.
Thus, AB 2 º AC 2 º AD 2 º BC 2 º BD 2 º CD 2 º 1 (mod 3) .

Suppose that C lies in ÐBAD

and denote ÐBAC = a, ÐDAC = b .
The law of cosines for DABC, DADC
and ÐABD yields
2 AB. AC cos a = AB 2 + AC 2 - BC 2 ,
2 AC. AD cos b = AC 2 + AD 2 - CD 2 ,
2 AB. AD cos(a + b) = AB 2 + AD 2 - BD 2 .

The left-hand sides are integers congruent to 1 modulo 3. Therefore

2
4 AC 2 . AB. AD cos a cos b º 1 (mod 3) . In addition, since 2 AC º 2 (mod 3) , we
obtain 4 AC 2 . AB. AD cos(a + b) º 2 (mod 3) . Subtracting the last two
congruences implies that

148
4 AC 2 . AB. AD[cos a cos b - cos( a + b)] = 4 AC 2 . AB. AD sin a sin b
is an integer congruent to 2 modulo 3.

On the other hand, cos a and cos b are clearly rational. Moreover, if
cos a = , cos b = , with a , b, c, d integers and ( a , b) = (c, d ) = 1 , then none
a c
b d
of a , b, c, d is a multiple of 3. Then, once the number
2 2 (b 2 - a 2 )(d 2 - c 2 )
4 AC . AB. AD sin a sin b = 4 AC . AB. AD
bd
is an integer not divisible by 3 , the same holds for the number
(b 2 - a 2 )(d 2 - c 2 ) . But a 2 º b 2 º c 2 º d 2 º 1 (mod 3) , implying that
(b 2 - a 2 )(d 2 - c 2 ) is a multiple of 3. The contradiction proves the claim for
n = 4.

Now the general case follows as an immediate consequence. Indeed, take

n arbitrary points in the plane, n ³ 4 , whose pairwise distances are all integers.
One can choose n
4 ()
quadruples of points among these, each quadruple
containing a pair of points at distance divisible by 3. Each such pair is counted
( n )times. Therefore the total number of distances determined by the given
2
-2

points and divisible by 3 is at least

(n )
4
=
1
( n ), that is: at least 1 of the total
( )
n-2
2
6
2
6
number of distances in question.

Problem 32 (Winter Competition,2001)

A 30  - 60  - 90  triangle T with hypotenuse 1 is given. A point is chosen on
each side of T so that the traingle with vertices at these points is a right
triangle. Determine the least value of its hypotenuse.

Problem 33 (National Olympiad,fourth round,1998)

A square, a regular n -gon and a regular m -gon are constructed externally on
the sides of the nonobtuse triangle ABC . Their centers are the vertices of an
equilateral triangle. Prove that n = m = 6 and find the angles of ABC .

One may ask here: Why are the last five problems considered separately. Do
the observations in Section 2 not apply to them? True, their difficulty is

149
definitely high, probably too high from a certain point of view. But being
simply ``hard" is not the point.
With most questions in Section 2, computations are quite long and yet
reasonable. The major difficulty is translating the information into a certain
analytic language (trigonometry, vector algebra, complex numbers, standard
coordinate geometry, etc.) Once the translation is done, the rest is a matter of
time and patience. A reasonable amount of effort suffices.
The story is different about Problem 29, for instance. Not having the crucial
hint stated in a), one is probably destined to face more technical trouble than he
or she could possibly take during a contest. It is not the mechanical translation
that matters, nor the subsequent manipulations. Essential geometrical thinking
precedes the computations, thinking hard to be replaced by other means. In this
case, the key idea is a purely geometric observation: The angle subtended by a
fixed line segment AB on one side of a given angle at a variable point C on
the other side of this angle is a maximum if and only if the circumcircle of
DABC is tangent to the second side (at the point C ). Much technical work is
still to come, but work manageable, guided by a proper understanding of the
situation.

A similar consideration is the core of Problem 32. If a right triangle T with

minimal hypotenuse is inscribed in a given triangle T0 then the circumcircle of
T is tangent to the side of T0 containg the vertex of the right angle. As above,
the remaining work is tedious but within reach.
In Problem 30, it proved necessary to amend the insufficient information in a
divisibility argument by an unexpected use of the modest triangle inequality. A
fact which may be regarded as a true fine point of this proof.
In Problem 31, insight is needed that the case of a general n reduces to
n = 4 by a counting argument, which is the combinatorial part of the matter. As
for n = 4 itself, this is where analytic geometric means are essentially involved.
This special case alone is a substantial problem in itself.

The last Problem 33 is also remarkable. It takes little to express the sides of
the triangle known to be equilateral by hypothesis, and to equate the expressions
obtained. However, by doing this we only end up with equalities promising no
more than hard work without any perspective. And indeed, the few solutions
found by contestants rested on a preliminary geometric observation again. It is a
well-known classical fact (Napoleon's theorem):
If equilateral triangles are constructed externally on the sides of a given
triangle then their centers are the vertices of an equilateral triangle.

150
 These five examples are not unique, and we believe that a certain new trend
has come into being. This is posing geometry problems not simply hard but also
natural and hence not isolated from the vast mathematical reality. With them, it
is never clear a priori what additional consideration will be needed-be it a
geometric observation, a combinatorial reasoning, a divisibility consideration,
or a combination of all these. We are flattered to think that such a trend brings
the olympiad movement closer to genuine mathematical work in the proper
sense.

REFERENCES

1. Ganchev.I., (1971), For the mathematical problems, Sofia.(In Bulgarian)

2. Kenderov, (1990), Bulgarian Olympiads, Sofia.(In Bulgarian)

3. Journal "Mathematics plus", 1993-2001 issues, ISSN 0861-8321, Sofia.(In

Bulgarian)

151