A Student's Thoughts on the Balkan

Mathemati al Olympiad
Gabriel Gendler
UNK2
May 23, 2012

Abstra t
This report is intended to give the reader an insight into the atmosphere, goings on and mathemati s at the 29th Balkan Mathemati al
Olympiad, whi h took pla e between the 26th April and the 2nd May
2012 in Antalya, Turkey. The ompetition was ex ellently run, and
the trips were both thoroughly enjoyable  one was a boat trip on the
mediterranean and the other involved tourist attra tions su h as an
an ient ruins and a waterfall.
In the exam, the team did well  we a hieved two silver medals,
two bronze and two honourable mentions. Question 3 was our most
well answered ( olle tively we s ored 48), followed by Q1 (41), Q2 (33)
and nally Q4 (16). The problems were all rewarding, and this report
spends some time dis ussing them. However, the main fo us of this
report is on the various theorems and hallenges the team worked on
olle tively in our spare time (of whi h there were many)  from theory
of innite sets to geometry and from inequalities to ombinatori s,
there seemed to be no end to the stream of problems supplied both by
our leaders, Geo and Gerry, and by ourselves.

Introdu tion
Like a fairytale, with its beginning, middle and end, most student reports
are omprised of three elements, ea h with its own degree of interestingness.
First, mathemati al events; se ond, non-mathemati al events; and third, stories about Adam Gou her. In his absen e, I am left not with three elements,
but with the relatively puny two.

The reader is advised not to ignore the

dieren e between three and two  indeed, if

3=2

then

1=0

(subtra t-

ing 2 from both sides) and as su h I do not exist, whi h would be a rying

shame for my similarly non-existent parents.

Rather than bemoaning the

la k of a third element, however, I shall embra e the opportunity to alter the

stru ture of the student report entirely, by s rapping the `non-mathemati al
event' se tion voluntarily and fo ussing entirely on mathemati al events (effe tive from the next paragraph). Su e to say that the hotel was luxurious,
the weather impe able, and our trip marred only by the la k of Geo Smith
(he was kidnapped by the organisers at the airport and taken to a se ret
lo ation, wherein he was for ed to work around the lo k in unbearable onditions for 48 hours to prepare the paper, and only returned to us after the
exam). From this point forth, minimal attention will be paid to all things
non-mathemati al  any parents reading this report should hen e skip over
se tions densely populated with squiggles and remain in the realm of the
ordinary. I have also appendi ised solutions to some questions I set  some
of these are in the answer se tion be ause you should a tually try to nd
the solution rst, and others simply so that readers with lesser mathemati al
ability (su h as my mum) don't need to s roll through pages of equations to
get to something they are about.
Although this report and that of Geo Smith on ern the same event,
they are wildly dierent for three reasons.

Firstly, he was in a dierent

pla e about half of the time, either on the jury, marking or o-ordinating;
se ondly, his ontains no maths whatsoever and mine ontains very little
else; and thirdly, while he handles the matter of our fellow passengers on the
outward ight deli ately and sensitively, I onsider it my duty to make my
disgust known.

Thursday 26th April

In whi h we journey to Antalya, dis uss mathemati s, are separated from
Geo, meet our guide and are given various freebies
Gatwi k Airport on a working Thursday is, onsistent with expe tations,
quiet.

This result is easily derived from some basi set theory, starting

with the set

of people in the UK. Allowing for a handful of mis ellaneous

ex eptions, su h as ourselves, we dene the following:

the set of working people,

the set of noisy people,

the set of retired people,

N
R
2

the set of people at Gatwi k on a working Thursday,

g G, g
/ W , soSg S \ W .
R = S , so
After the harmless but perhaps derogatory assumption that W
T
S \ W S \ R = , we get g
/ S \ R, so g R. Therefore g
/ N and
Now

R=R

W =W

G=.

So

we have explained the quietness of Gatwi k.

ight would be quiet by the same logi .
aw in the logi is the premise that

One might predi t that our

One would be very wrong.

G = .

In fa t,

G=C

The
 the

non-empty set of people who, while they do work, have neither interest in
or ommitment to their o upations, nor desire to ontribute to so iety, in
su ient quantities to pay slightly more for their va ations by waiting for
s hool holidays  indeed,

is the set of havs.

Despite the adverse onditions (my seat, for example, is an armrest for
two fat men, both of whom smell of beer and urine) we manage to get some
maths going.

Gerry (deputy team leader) grapples with some beautifully

drawn geometry while Geo (team leader) tells us stories of past ompetitions.

Matthew (UNK4) works on a sele tion of old BalkanMO problems

and later some innite set theory. Matei (UNK5) and Daniel (UNK3), with
o asional help from Robin (UNK1) and I (UNK2), work on an interesting
set of geometry problems involving evians.

P oplanar with triangle ABC , where

AC at E and CP meets AB at F ,

It is well known that for any point

AP

meets

BC

at

D , BP

meets

AF BD CE

=1
F B DC EA
The question on erns situations where

EA = 0.

AF F B + BD DC + CE

Of ourse, the trivial examples were the entroid, the gergonne

point and the nagel point, but interesting points to onsider were the in entre, ortho entre, symmedian point or ir um entre. Given that P is ea h of
these, determining the properties of triangles whi h omply with the above
ondition is a worthwhile exer ise. It generally involves nding expressions
for ea h relevant line segment (I use areal o-ordinates), then forming a huge
polynomial (generally in

a, b

and

c)

and fa torising.

Importantly, all the

terms in the polynomial had the same order (here the word means the sum
of the powers of ea h of

a, b

and

 for example,

a5 bc

is 7, as is

a4 c3 ),

and

Matei tea hes us a ni e way of fa torising su h polynomials. If the order of

ea h term is

n,

draw a triangle of isogonal points with

side. For example, for

n = 3,

ten-pin bowling setup. Label ea h vertex

term. The distan e from a dot

n+1

points on ea h

10 dots should be arranged in the form of a

a, b

and

c;

ea h point refers to one

to ea h edge (as a stepping-stone distan e,

not as the row ies) is the power of the letter assigned to the opposite vertex
in the term referred to by

D.

Over ea h dot, write the oe ient of that

term (in luding the sign). Now it is possible to pi k out fa tors by noti ing
repeating patterns  for example,
written as

a3 a2 b + ab2 b2 c + bc2 c3

would be

a
1

1
1

0
0

0
1

and it is relatively easy to spot the fa tor

a
1
0

and its negative version, whi h gives the fa tor

a c.

To someone good at

fa torisation, this may seem like a waste of time, but as kings and queens
have said, I am not good at fa torisation.
Sitting in the middle on the six, I have the privilege of oating from the
geometri al musings of Daniel and Matei to Matthew's set theory.

He is

onsidering pairs of un ountable sets of reals between whi h a bije tion is

possible that preserves order. We agree that su h bije tions exist regardless
of size, but only provided the intervals' ends are losed and open in the same
way (intervals starting or ending with

or

are taken as open on that

end). We turn our attention to the antor set, agreeing  falsely  that the
map is possible; write ea h number in the antor set base three, and only 0s
and 2s are used (1 is written as 0.222. . .

and 0.1 as 0.0222. . .

and so on).

Then turn ea h two into a 1 and we have 0 to 1 in binary. Unfortunately,

as I realised at lun h on Friday (as I write) a third and two thirds both
map to a half, as 0.0111. . . =0.1 in base two. We now turn our attention to
sets that we will suspe t we will get nowhere with  for example, a bije tion
between

and

R\Q

preserving order  I annot nd a way through this at

all. We ask Geo, and he onrms that it's hard (he later proves that no

su h bije tion exists, as does Robin ) and steers the onversation towards
maths that's way over our heads  a problem about aeroplanes. We are to

estimate the distan e to the horizon in a plane at ruising altitude, and I

al ulate that it's somewhere between 200 and 300 miles. Geo nods, and
on ludes that when ying over London, Man hester is visible. Matei and
Daniel have turned their attention to the Buttery Theorem, a result whi h
they would fo us on throughout the following day. On the plane it interested
me as well, but at the time of writing I am si k to the ba k teeth with this
useless, ran id, boring and patheti result, so you an look it up.
Our rowdy ight-mates have added *angry* to the list of unpalatable
traits they exhibit, as the plane has run out of al ohol in the rst half of the
ight (I wonder if they realise that this was their doing entirely?) but are
su iently intoxi ated to not really are. A pair of totally hammered girls
strike up a onversation, and are happy to dis uss the impressive a hievements of Pythagoras when they nd out who we are and where we're going.
I breathe a sigh of relief when the plane lands, and say a short prayer that
in their drunken stupor they will forget my name and therefore be unable to
follow up on their oath to add me on Fa ebook. We say a teary farewell to
Geo, and meet Salih, a friendly Turkish student with impressive English.

Gerry onstru ts the reals on the journey to the hotel, starting with
Peano's su essors and ulminating in the Dedekind ut whi h laun hes us
from rational to real (the only part of the pro ess we hadn't foreseen). Soon
we pull up at the hotel.
At this point the reader should be ome a quainted with the notes do u-

ment that the team was sent a few days before we left  based, a ording to
Geo, on observations from Moldova two years before  su h tips as `bring
towels', `bring loo roll', `do not drink tap water' and `do not leave your
valuables lying around obviously in hotel rooms whilst you are out'.
We are staying in a ve star hotel. There are six main dining areas, ea h
heaving with mouthwatering turkish spe ialities. There are three swimming
pools, the main one omprised of two large se tions joined by a de orative
hannel under a bridge, and in luding four water slides, a statue of a mermaid
and a fountain.

Hoteliers are on hand to arry our luggage to our rooms,

ea h of whi h ould house another swimming pool, and in ludes a large en

suite bathroom and a bal ony (I beat Robin to the double bed; selsh sin e
I am noti eably smaller).

If you hoose, at this point, to berate me for  kly forsaking my promise

to avoid non-mathemati al events, then you la k foresight. It's time for a
hat about Hilbert's Hotel. A hotel with

rooms is simply not impressive

to us any more, having seen our enormous hotel. Admittedly it does not have

rooms, but we agree that it is only a few dozen short. The Continuum

Crib, on the other hand, is a hotel (designed by ourselves) with

(ea h assigned a real number between 0 and 1).

20

rooms

Cantor regularly re eives

ontinuum oa hes, and gives ea h passenger a room (guests never leave

 the ontinuum rib should really be alled Gabriel Gendler's great big
graveyard).

Of ourse, this is no problem; Cantor an house

20

of these

 oa h ohorts by a ri h hoi e of

R2 R

bije tions, so

of them (they do

not arrive ontinuously and so are ountable) shouldn't be a problem. Cantor

has always operated as follows  ea h time a oa h arrives, he assigns it an
integer ID (starting at 1 and ounting up). He then assigns ea h passenger a
new real number between

1
1 2n1

and

1 21n ,

and sends ea h guest to these

rooms. Unfortunately, Cantor's ghoulish guests are getting upset. Although

Cantor knows that ea h guest has its own real number, and that the reals
are evenly spread, the guests seem to think that as the interval shrinks, they
be ome more rammed. Can Cantor hange his a ommodation system to
avoid this per eption, without making his guests hange rooms after they
have arrived?

We unpa k our suit ases and rummage through our BalkMO freebies,
in luding turkish delight, whi h is turkish, and a pair of ompasses, whi h
are not (also in luded is a list of these freebies whi h does not in lude itself,
but Cantor

and

Russell seems too mu h for one evening).

We get to bed

at 2 in the morning, asserting that onsidering the luxury that is our hotel,
Geo must be in Paradise.

Friday 27th April

In whi h we explore the hotel, attend the opening eremony, dis uss mathemati s and swim

Despite today's date being the

33

th

of the

22

th

, very few mathemati al

events o ur. We wake at 7.45 and view the hotel in daylight for the rst

time  indeed, it is a thing of beauty, and it is vast.

Robin and I have the best view from our window, overlooking the pool(s)
and the mediterranean. Bizarrely, a river runs along the oast without meeting it for quite some distan e, but nobody really ares.

An early transfer

moves us to the opening eremony, whi h takes pla e in a third hotel, even
more luxurious than ours  the whole thing is an enormous golng resort,
mu h like most of Antalya. I will not detail the events at the opening eremony, sin e they all ome under the ategory of non-mathemati al event.

We now have 10 hours of free time  a wonderfully relaxed, alm day.

Matei, Robin and I swim for a while, although this too is not very mathemati al, but for most of the afternoon all of us (ex ept Harry) sit by the pool
doing maths. Daniel and Matei independently solve the Buttery Theorem,
and Gerry shows us some beautiful problems involving symmedians.

The

rst is a known result of symmedians, but Gerry wants a geometri proof.

A triangle

T.

ABC is ir ums ribed by . Tangents to at B and C

T AB = CAM , where M is the midpoint of BC .3

meet at

prove that

We all marvel at the beautifully symmetri proof he gives  I am parti ularly impressed with the onstru tion of two new points, sin e I am rarely
able to use onstru tions in my proofs.

Now Gerry shows us a wonderful

ABC ,
M is the midpoint of BC , H is the ortho entre and D
AH . M H meets above BC at X ; XD meets again at Y .

symmedian problem from the Australian Olympiad: in a triangle

is the ir um ir le,
is the foot of

Show that

AY

is a symmedian!

I spend most of the day writing the student report so far, and updating
the others with my progress every so often. As I write (today) the Australian
problem remains unsolved.

Saturday 28th April

In whi h we sit the exam, are reunited with Geo, dis uss our solutions, play
hess and dis uss mathemati s

With our joyous rest day a distant memory, we wake up bright and early
for the exam. This is to begin at 9 and to last 4.5 hours, mu h like FSTs,
NSTs, XSTs, IMOs, RMMs and plane journeys to Turkey.

There will be

four questions (1 more than in the exams listed above, and as we know, the
dieren e between 4 and 3 should not be ignored) and as su h the exam
is a daunting undertaking. Geo mentions that last year was a tough one
 set in Romania, a ountry of talented mathemati ians and perhaps more
talented problem setters  and hopes that this year would be easier.

He

is orre t  the problems (as follows) result in unusually high marks a ross

most ountries  the Turkish team, for example, s ore 40, 40, 40, 39, 36 and
31. The problems are as follows:

A, B and C be points lying on a ir le with entre

ABC > 90 . Let D be the point of interse tion of the
line AB with the line perpendi ular to AC at C . Let be the line through
D whi h is perpendi ular to AO. Let E be the point of interse tion of
with the line AC , and let F be the point of interse tion of with that lies
between D and E . Prove that the ir um ir les of BF E and CF D are
tangent at F .
Question 1  Let

O.

Assume that

Having the pro ien y in Geometry of a o onut, I manage to miss this

problem ompletely, failing to noti e even some of the most blatantly obvious properties of the onguration  however, my able olleagues are less
unsu essful. Matthew runs out of time half way through his proof, but the
other UNKs laim omplete solutions. Harry's solution (without doubt his

best) is detailed in the answer se tion.

Question 2  show that for all

x, y, z R+ ,

(x + y) x + y y + z 4(xy + xz + yz)
y

Robin, Matei and I have the most omplete solutions here, all using the
substitution

a2 = x + y , b2 = x + z

and

c2 = y + z .

Robin then employs a

useful inequality from S hur, followed by Muirhead. His solution is akin to

using a nu lear bomb to apprehend a shoplifter  it is ridi ulously powerful,
and he obliterates the inequality like a psy hoti pyromania . In other words,
it is in redibly professional.

Matei's solution is extremely insightful  after the initial

tution, he arries out yet another, with

d + e = a, d + f = b

a, b, c substie + f = c.

and

After an enormous expansion (whi h he doesn't arry out line by line, to the
dismay of Gerry who must therefore slog it out himself when marking the
paper) this be omes a sum of squares and Matei is done. I am en hanted by
this solution. Like a demented wombat in omparison to Matei and Robin,

10

the elegant kangaroos, I opt for a bizarre and unusual proof, whi h is imaginative but drasti ally more ompli ated.

Nevertheless, it makes for some

wonderful omedy, so I have in luded it in the answer se tion.

On the subje t of solutions to question 2, the answer se tion below on-

7 that afternoon and the other from Daniel

tains two more  one from AoPS

the following day

8 (he is disappointed as Q2 is the only one he doesn't get,

and then after the exam he nds the best solution of them all).

Pn = {2n , 2n1 3, . . . , 2nr

3r , . . . , 3n }. For ea h subset X of Pn , we write SX for the sum of elements of
X, with the onvention that S = 0. Suppose that y R with 0 y SY
3n+1 2n+1 . Prove that there is a subset Y of Pn su h that 0 y SY < 2n .
Question 3  let n be a positive integer. Let

This  ombinatori s problem (it was the only one on the shortlist, so
we suspe t that the organisers were desperate) proves to be our strongest
question, and we all submit roughly the same solution. Matthew's is re orded

in the answer se tion.

Question 4  Let

f:

Z+

be the set of positive integers. Find all fun tions

Z+

Z+ su h that the following onditions hold:

+
i) f (n!) = f (n)! for all n Z
+
ii) m n | f (m) f (n) m 6= n Z
Daniel laims the only solution here (although Matei gets a long way).

The answer, of ourse, is

f (n) = 1, f (n) = 2

or

f (n) = n

 an observation

whi h would have earned me a mark had I bothered to write it down  I

suppose this kind of mistake is what the Balkans are for. I annot envisage
myself oming up with Daniel's proof in a million years  the number of steps
is ridi ulous.

10

Upsettingly for anyone who ares about the environment, we are only
allowed to write on one side of the paper (they may as well have given
us mbius strips).

Another interesting observation is that the rules sheet

in ludes a prohibition against having toys in the exam hall  I wonder how
mu h help even the most intelligent of teddy bears ould realisti ally be? We
troop out of the hall (whi h lost power on two o asions during the exam)

11

and are reunited with Geo, who debriefs us. We now have a free afternoon
(followed by three free days) in whi h to do some maths.

After a spot of

giant hess, in whi h I pose as a knight, we onsider some olouring problems.

Daniel re alls FST1 Q2  whi h I must be vague about sin e the paper is not
yet on line, so this will only make sense for those who sat the paper  and
wonders whether an optimal solution is possible with a diagonal pair; the
upper bound required in a solution theoreti ally allows it. Two 3- olourings
ontradi t ea h other ni ely, giving the desired result.
Over dinner, Daniel shows us a beautiful problem wherein inversion is
a tually intrinsi ally useful, rather than simply a reframing of the problem.

AB is the diameter of a ir le . P is a
su h that QP A = 90 . The largest
ts inside the bise ted segment QP B . This is
that AG = AQ. The reader is en ouraged to

This is from Josh Lam on AoPS:

point on

AB

and

is a point on

possible ir le is drawn that

tangent to

AB

at

G.

Show

11

think before he king the answer.

Sunday 29th April

In whi h we dis uss mathemati s, travel to Perga, see a waterfall and visit
the ity of Antalya
Today we journey to Perga, or as Geo refers to it, Apollonius's rib.
Quite how and when he learned jive remains a mystery. The outing will, of
ourse, disrupt our mathemati al investigations, so we get some maths going
very early. In bed before breakfast, I onsider the geometry problem from

P for triangle ABC with evians AP D , BP E and CP F

AF F B + BD DC + CE EA = 0. I attempt to nd an

the plane  points

su h that

equation for the lo us of points that meet the ondition given a triangle with

a, b and c  P with areal o-ordinates (x, y, z) is ee tive if and only if

a(z y)(y + x)(x + z) + b(x z)(z + y)(y + x) + c(y x)(x + z)(z + y) = 0

sides

whi h I will mess around with later. Harry observes that the gergonne point

is the symmedian point of the intou h triangle by Gerry's symmedian proof

from Friday. Over breakfast Daniel explains his fth solution to question 2,
whi h I mentioned in yesterday's report (answer 8)  it uses a remarkable
yet well known fa torisation, and I wonder why none of us spotted this in
the exam.

12

tan A + tan B + tan C =

sin 2A + sin 2B + sin 2C =

We do an awful lot of trigbashing, showing that

tan A tan B tan C

4 sin Asin B sin C
3

that

where

A + B + C = 180,

that

and (for posterity, nobody found this useful or non-trivial)

a2 b2 c2 sin A sin B sin C

is the area of a triangle.
2

Geo gives a ute geometri representation of my identity that

sin B + sin C sin 2A + sin 2B + sin 2C

sin A +

for a ute triangles (read my solution

to Q2 from yesterday to understand how this ts in). In an a ute triangle,

DEF . Simple trig gives the sides of DEF

R sin 2A, R sin 2B and R sin 2C , whereas the larger triangle has sides
2R sin A, 2R sin B and 2R sin C  we show that BC F D + DE y li ally
we onsider the orthi triangle

as

(by BMO1 Q6 2011) and we're done. Attention swit hes to whether this is
also the ase for obtuse triangles  my trigonometri proof involves the fa t

that

sin x

in reases and

cos x

de reases as

in reases between

and

90 ,

whereas Geo 's orthi triangle proof doesn't hold for more omplex reasons
whi h I won't detail. Instead, Geo gives a lovely proof  starting with an
a ute triangle (not equilateral) the inequality holds stri tly and this triangle
an be transformed ontinuously until it is any other triangle  the dieren e
between the LHS and the RHS hanges ontinuously, never rosses 0 (this
would require it to be ome equilateral) and starts positive, so the inequality
must hold no matter what the triangle. I love this proof as it is in redibly

13

 lassy and simple, although slightly more work is required to show that an
equilateral triangle is indeed the only one for whi h equality holds. A safer

12

proof (also from Geo ) is re orded in the answer se tion.

Some thought is given to the pedal triangle and the wonderful formula

P to
DEF ,

for its area based solely on the power of

oplanar with triangle

ABC ,

the area of

feet of the perpendi ulars from

P,

[DEF ] =

the ir um ir le.
where

D, E

and

Given

are the

is given by

R2 OP 2
[ABC]
4R2

I remark that a result of this is the well known fa t that the feet of
perpendi ulars from a point on the ir um ir le are olinear. Geo outlines
a proof that goes over my head.
The trip itself is fairly medio re, so I won't bore you with the details,
but let it be known that there was a greek ruins, a waterfall, a restaurant, a
ity and a bus.

14

There's surprisingly little time to relax when we return to the hotel, but
enough to play a few games of pool (we observe, in the spirit of a maths
ompetition, that not all points on the table are oplanar and that not all
points on the ue are olinear, whi h ruins things somewhat). We also nd
out that Salih's name is not Salih.

I make the utterly useless observation

that a triangle and it's medial triangle's orthi triangle's intou h triangle
share a ir um entre.
As I ome to this realisation, I begin to experien e a stoma h a he.

feel in reasingly unwell over the next hour, and the pain builds to omplete
agony.

Sensing imminent danger, I un url myself from a ball-like position

on my bed and assume, as a safety pre aution, a more appropriate position

in the ne essarium. I re-emerge 10 minutes later having dealt a serious blow
to the toilet's self esteem. My stoma h a he is ompletely relieved and I fall
asleep very qui kly; the only health s are of the trip for any of the UNKs
has passed.

Monday 30th April

In whi h I almost kill Robin, we play poker and pool, swim, dis uss mathemati s and nd out our marks and medals
Today is ompletely free. I am awoken at a luxurious 9 o' lo k by the
usual pillow to the fa e. In my half-awoken haze my pillow-throwing response
is less rened than usual and I ome lose to unhooking a large painting

15

from the wall, whi h would have made my least professional pillow-throw
additionally my most deadly, sin e Robin is dire tly below.

We breakfast

at the similarly luxurious 9.15, allowing for a not so luxurious 15 minutes

to get dressed, wash and shower. Over breakfast we re eive our marks for
question 1  Matei, Harry and Daniel s ore 10; Robin makes 8 after being
do ked 2 for failing to show that

is the ortho entre.

Matthew takes a

respe table 4 after not nishing his proof, and I am surprised to pi k up a

mark just for re ognising the relevan e of the point opposite

over

O.

This

seems extremely generous onsidering Geo 's warning that I may be ned a
mark on Q2 for not showing that the triangle I have onstru ted is real by
triangle inequalities, despite showing that it is a ute. How there ould be an
a ute triangle that is not a triangle I do not know, but o-ordinators will be
o-ordinators.
After an extensive game of poker, whi h is admittedly quite boring, we
relax for an hour and then regroup for lun h. Geo meets us and tells us our
s ores for Q2  10 for Robin and Matei, the Cran h Expe ted S ore of 0 for
Daniel and Matthew, and 4 for Harry for his substantial progress. My s ore
is still unde ided, and Geo explains that while it was a epted that very
few a ute triangles are not triangles, my nal step using the rearrangement
inequality was not symmetri al.

The problem was that instead of using

sin A cos B + sin B cos A sin A cos A + sin B cos B as I wrote in this report,
I had foolishly written sin A cos C + sin B cos B + sin C cos A sin A cos A +
sin B cos B + sin C cos C and the introdu tion of the middle variable means
that it doesn't work y li ally. Lu kily, as I nd out during a far i al game
of pool, my ne is just 1 mark sin e rearrangement would work with su h a
minor adjustment. As su h I have gained an undeserved mark and arelessly
lost one, leaving me on tra k for my predi ted 20 marks  the Lord giveth and
the Lord taketh away (Job 1:21). We await our marks for Q3, whi h should
be strong  interestingly, we noti e that while we all found Q3 easier than
Q2, suggesting that the jury were mistaken to pla e them in that order, the
marks of other ountries prove us wrong. Perhaps the British squad needs
more work on inequalities?
While the swimming pool itself oers little in the way of maths besides
aerodynami s and the me hani s of water slides, the de k hairs are a hub
of dis ussion over the pinna le of mathemati al hallenges  AQA C2. Su h
tests of the mind as given

sin + cos = 0, show that tan = 1.

Whoever

dened this as mathemati s was wrong. In view of this fa t we onsider establishing an A-level in Olympiad mathemati s, but eventually de ide that it
would be far too hard for the ordinary mathemati ian. Our s ores for question 3 arrive  Robin s ores 8, Matei 9, and the rest 10 other than Harry, who

16

missed Q3. Q4 gives Daniel 10, Matei 4 and Robin and Matthew 1. We're
all fairly happy with our s ores, but predi t that ea h medal boundary will
be pre isely 1 above an UNK s ore. Geo tells me that my Q3 proved a pain
in the ba kside again, due to a language issue, but learly this has been resolved without a ne. Attention then turns to whether any interesting OEIS
sequen es ontain our s ores  then we attempt to nd a quinti polynomial

p su h that p(x) is the s ore of UNK x for all integer x from 1 to 6.

We an't

be bothered, and it seems that Wolfram Alpha an't be bothered either. At

this point the patient reader will nd out why the report is so long  it is
a trial of patien e for our benevolent di tator, James Cran h.

If, in fa t,

he has read this far, then we ongratulate him and apologise for mentioning
that Guernsey FC urrently reside in the 9th level of the English Football
League.
We head to the bea h, dis ussing an Advan ed Mentoring S heme problem involving an innite number of points in general position oloured green
and bla k (with at least 1 m between ea h to avoid a ontinuum of points) 
we are required to show that in no su h olouring an every triangle of green
points ontain a bla k point and vi e versa. My proof is a bash, but Matei's
is gorgeous.

13

Seeing that kayaking was losed from 5 o' lo k (the time now being 5
past 5) I head ba k to the poolside, where Gerry is talking about dodgy
tea hing of maths at A-level, su h as poor denitions of integration and
dierentiation, and sloppiness with inequality manipulation.
he solves:

17

For example,

1
1n
<
2
1000000
log 1

1n
1
< log 1
4 1000000
2

n<

1
1000000
log 1 12
4

log 1
4

whi h, of ourse, is wrong (the sign is the wrong way round)  but whi h

> 0 so no multipli ation by negatives has o urred.14

He then integrates f (x) = 1/x between 2 and 5 (tut tut) by lim0 of
the integral between 2 and plus that between and 5 (dening ln x
orre tly). Using 4, or r r R in the positive version gives a dierent

step went awry?

log 1
4

1
2

result, whi h explains part of the denition of dierentiability  that

lim

f (x + ) f (x )
f (x + r) f (x )
= lim
rR
0
2
(1 + r)

or something along those lines.

Finally he ompletes the sequen e 3, 0, 0, 4, 2, 0 with 12. Genius.
Robin and I have another ha k at least year's inequality  for

X x(x + 2)
y

2x2 + 1

x+y+z = 0:

To my shame, I multiply by the denominators and expand out, but this

doesn't help.

I try substitutions whi h also fail.

Although I do not solve

the problem on the trip, I have more su ess a few days after getting home

15 We spend the evening playing lots

 see the half- ooked answer below.

of pool, even more table tennis and a ridi ulous amount of maths. At the
entre of attention, among several other integer series with little founding in
traditional mathemati s, is this ghastly thing:

1, 1, 2, 1, 1, 2, 2, 2, 3, 1, 1, 2, 1, 1, 2, 2, 2, 3, 2, 1, 1, . . .
in whi h ea h number is found by ounting the number of full repetitions
of the longest repeating sequen e nishing on the previous number, starting
with 1.

With

f (n)

dened as the term of the above sequen e at whi h

18

rst appears (so

f (n)

f (1) = 1, f (2) = 3, f (3) = 9

and so on) it is agreed that

grows rapidly. Very rapidly. Matei onje tures that 10 has not been

rea hed by term G, where G is Graham's number (whose best approximation

is the number of rooms in our hotel, although it may be slightly greater if
we're being honest).

I agree that

f (10)

is large, but not that large.

This

dispute is qui kly settled when Harry orre ts Matei's denition of Graham's
number, whi h was too small by a fa tor of roughly Graham's number (for
anyone who ares, Matei is thinking about

g(64),

g(1),

and Graham's number is

so Matei is out by at least 63). Geo and Gerry nally appear with

the medal boundaries  30 is Silver, 20 is Bronze and Gold is 39 due to the

relatively easy paper.

This means that Daniel and Matei take Silver, and

Robin and I Bronze. Contrary to our expe tations, two of us land lu kily on
a medal boundary, whereas we suspe ted that multiple UNKs would fall a
point shy. I note that giving ea h team member a number between 1 and 6
a ording to relative rank, all three bedrooms add to 7. Further, in reverse
order of UNK, ea h value os illates towards 3.5  6, 1, 5, 2, 4, 3. These two
are losely related, as rooms were assigned alphabeti ally, and so were UNK
numbers.

Just before bed, Robin explains a fantasti trig identity  for a

ABC , there is a point P (the Bro ard point) su h that P AB =

P BC = P CA = , and for this point cot A + cot B + cot C = cot ,

triangle

whi h is impressive.

Tuesday 1st May

In whi h we dis uss mathemati s, relax on a boat, attend the losing eremony, es ape from the losing gala and write Salih a letter
Today is our last full day, and our se ond road trip  or rather, sea trip 
we will spend the day on a boat. As we ongregate outside re eption to board
oa hes to the harbour, Daniel, Robin and I ta kle the otangent/bro ard
point problem (as well as its orollary that

30 ).16

We onsider the inequality from Balkans 2010:

X a2 b(b c)
y

a+b

I solve this the predi table, reliable and fast way.

17 Daniel, as usual, puts

18 (I advise that you take a look). Despite the beauty

my method to shame
of Daniel's solution, I do not regret doing what I did.

19

It is e ient, it is

safe and it is easy. I see no reason to bother with beautiful solutions to ugly
inequalities. They are horrible, and deserve to be hit around the head with
a ma hete, and that's what I endeavour to do.

In the spirit of this entire

event, geometry qui kly be omes the main subje t of the day (at this point
we are on a boat ruising through the extraordinarily blue mediterranean).

ABC =
[ABC]2 +[CBD]2 +[DBA]2 = [ACD]2 .

Gerry sets a problem involving an orthogonal tetrahedron with

CBD = DBA = 90

 show that

This yields to a qui k bash.

Geo then sets two problems  rst, of a

ABCD with any point P on its ir umsphere  show

AP 4 + BP 4 + CP 4 + DP 4 is xed. The se ond involves a triangle with
in ir le . The tangent to parallel to BC meets AB at A1 and AC at A2 ;
B1 , B2 , C1 and C2 are dened similarly. It is required to prove that
X
a2 + b2 + c2
AA1 AA2
9
y
regular tetrahedron
that

These two prove more di ult, but we are soon joined in our quest by a band
of Romanians and Turks with an average BalkMO s ore of 40. I am proud
to show them something they don't know  Matei's fa torising triangle. The
se ond is qui kly dealt with but the rst is still on the table as the team is
disbanded by an enormous waterfall.

Over lun h, Geo reveals his motives  he has been demonstrating the
useful nature of the Huygens-Steiner theorem, whi h he asserts an solve

20

anything whi h involves even powers of lengths.

He then explains what

Huygens-Steiner is (this is the se ond time I've heard an explanation of

Huygens-Steiner and I am beginning to feel as though after ve more explanations I will understand it). Although the rst solves itself in o-ordinates
and the se ond an be ta kled with plane eu lidean geometry, I will omit
these proofs be ause the point of the exer ise was to use Huygens-Steiner,
and I annot write up a Huygens-Steiner proof be ause I don't understand
Huygens-Steiner. The onversation ips to useless geometry, and Geo mentions Adam Gou her's favourite  the produ t of the four tritangential radii
is the square of the area. In other words,

[ABC]2 = r rA rB rC
This is proved with Heron. I also realise that
of

IA IB IC

ABC

is the orthi triangle

 a pleasing but even more useless result. Gerry shows us a ni e

tri k  writing Pas al's triangle mod 2 and reading ea h row as an integer in
binary, we get the sequen e 1, 3, 5, 15, 17, 51, 85. . .
of onstru tible

n-gons

whi h is the sequen e

up to multipli ation by a power of 2.

Unfortunately the boat ride must end and we are for ed from the ool
sea onto stiing buses with little air- on. Lu kily, Gerry is always a sour e
of fas inating maths, and he tea hes me about ordinals,
indu tion.

, 0

and transnite

Some of it goes over my head, but what I understand, I really

enjoy. I won't go into detail here, but I re ommend that the reader ask Gerry
for some very ni e observations. Orders of

2 , 3

and

are not too hard

to nd, but proves more di ult. Eventually, Matei identies one: in the
ontinuing sequen e

a1 +

Where

ak+1

an

starts at 0,

ak

1
a2 +

1
a3 +

1
1
a4 + ...

is in reased to innity followed by a repeat with

in reased by 1, starting with

k = 1.

Jumping ahead to Gatwi k on

the way home I nd another  the integers are ordered alphabeti ally, so
in order of rst digit, followed by se ond and so on, with the de imal point
(these are integers so I mean the end of the integer) its own hara ter.
The losing eremony is brief and non-mathemati al  most notably, I
hold up my union ag in orre tly and we observe that the ag has rotational
but not ree tive symmetry, making our ag unique among those on show.

21

Moreover, Geo gives a moving spee h on the importan e of volunteerism

in the mathemati al ommunity. After the eremony we have time for several
games of pool before the losing gala. These games are all fairly medio re,
but we manage a tri k shot or two.
The losing gala involves food we didn't order, musi we an't stand
and a onversation we an't hear, due to the ridi ulous volume of the musi
(see Geo 's report for a more livid a ount).

Nevertheless Matei and I

ABC
P meet opposite sides at D , E and F with
AF F B + BD DC + CE EA = 0  the ondition from the plane
manage some maths, looking at the lo us of points for any triangle

su h that the evians through

journey. Our expressions annot be fa torised, so we're still in limbo here.

Among other lo i I have been ontemplating is the lo us of points for all

r R with perpendi ular heights from ea h edge x, y and z that minimise

xr + y r + z r . Finally we write Salih a goodbye ard, whi h ontains various
rubbish limeri ks su h as the following (by myself ):
There on e was a person from Turkey,
This limeri k just doesn't work-ey.
He was very ool,
But avoided the pool;
We thank you for being so perky.

22

Wednesday 2nd May

In whi h we journey home, dis uss mathemati s, Daniel learns German, we
say our goodbyes and are reunited with our families
We wake, pa k and breakfast without mu h maths, say our goodbyes and
hop on the oa h. The trip to the airport is quiet  we dis uss some geometry
but are generally more ree tive. The airport is empty, and we move onto
the plane qui kly (at the gate Geo has time to explain the role of the jury in
a maths ompetition). On the plane, I tea h Daniel some German (the ight
is Airberlin so we have plenty of German to read) and ontinue work on our
various problems. A lot of what I've des ribed as `solved later' throughout
this report is a hieved on these ights, and Matei manages to ome up with
a rough shape for the lo us we have been looking at (for s alene triangles).
I will not attempt any ASCII art to reprodu e the diagram; you an do it
yourself, you lazy frog.

Were we a less serious bun h, our goodbyes may

have been more animated, but they were moderately heartfelt nonetheless.
More upsetting, as we realised upon being reunited with our parents, was
the prospe t of having to tidy our own rooms.

Remarks

I'd like to thank Geo, for ghting our orner with the o-ordinators;
Gerry, for keeping an eye on us and supplying us with many problems;

23

UKMT, for oughing up so mu h money; Bev, for organising everything;

Tbitak, the Turkish government's s ien e trust; the havs from the plane,
for entertaining us so greatly; the hotel, for feeding us with wonderful nourishment and providing us with su h luxuries as a pool, leaners and loo roll;
and most of all, Salih the GUNK (Guide for the UNited Kingdoms) who was
helpful, aring, funny, friendly and knowledgeable, and with whom I shared
many a onversation onsisting solely of quoting Shakespeare, Milton and
Leonard Cohen.

Dis laimer: literally no eort was made whatsoever to make this report
truthful in any way, shape or form. If you nd a fa tual error, please write
your orre tion on A4 paper along with your onta t details and put it in the
bin.

Appendix

24

Competitors from the UK

#

Name

Robin Elliott

Gabriel Gendler

Daniel Hu

Matthew Jasper

5
6

Q1

Q2

Q3

Q4

Total

10

27

Award
Bronze

10

20

Bronze

10

10

10

30

Silver

10

15

Honerable* Mention

Matei Manda he

10

10

33

Silver

Harry Metrebian

10

14

Honerable* Mention

*Honerable is the spelling used on the erti ate

Answers
1  Suppose, for a ontradi tion that su h a fun tion exists ( all it

1
and f

f : R R\Q
of P and P ,

rational. Sin e

and

f 1 (P )

where

f ),

so

R. The domain of f 1 is the disjoint union

: R\Q
P is the set

of positive real numbers whi h are not

preserves order, so does its inverse, so the sets

are a Dedekind ut of the reals  either

f 1 (P )

f 1 (P )

ontains a

1 (P ) ontains a minimum element, implying that

maximum element or f
either

or

has a minimum or a maximum respe tively, whi h is absurd.

2  the answer is not, as some might suggest, the Cantor set.

While

the Cantor set does over an innitely small interval, when it is applied
over an area one third the size it's interval redu es by a half, so the guests
ould still feel rammed. Instead, use an interla ing  assign ea h bus a real

number between 0 and 1 ( n will do), assign ea h passenger on the bus a real
2
number between 0 and 1, write out the binary expansion and interla e the
two sequen es (e.g. if one begins 0.01101110. . . and the other 0.00011011. . .
the end result will be 0.0010100111101101. . . ) sin e just

of the rst real

numbers are needed, the resulting rooms never form an interval at all, so
nobody feels rammed.

3  Gerry labels D on AB su h that DT = BT and B 6= D . Now

DT = BT = CT . DT is extended to meet AC at E . He notes that
BCA = T BA by the alternate segment theorem (note that the use of

dire ted angles renders this true rather than that T BA + BCA = 180
whi h is wrong). T BA = T BD and T BD = BDT sin e BDT is

25

BCA = BDT . Therefore ABC and AED are similar,

T EC = ABC . But ABC = ACT by the
alternate segment theorem, and ACT = ECT . Sin e ECT = T EC ,
T E = T C . Sin e T C = T B = T D , T E = T D , so T is the midpoint of DE .
Re all that ABC is similar to AED , so AM C is similar to AT D
and T AD = CAM  QED.
isos eles, so

although mirrored, and

G at the interse tion of AO and . M is the

AB . 2ACB = AOB = 2AOM so ACB = AOM , and
AM O is trivially 90 so AOM = 90 M AO = 180 DGA
DAG = ADG, so BDE = BCE . Therefore BDCE is on y li , and

sin e DCE = 90 , EBD = 180 90 = 90 . Sin e A, C, F and B are

on , ACF + F BA = 180 , so F BA = 180 ACF = ACD

ACF = F CD . Then let the midpoint of EF be N and that of F D be
P . Let the ir um entres of EBF and F CD be O1 and O2 respe tively.
Then 2F BE = F O1 E = 2F O1 N . Similarly 2F BE = 2F CD =
F O2 D = 2F O2 P and F O1 N = F O2 P . Sin e O1 N is perpendi ular to ,
whi h in turn is perpendi ular to P O2 , O1 N is parallel to O2 P and by the
onverse of alternate angles O1 F O2 is ollinear and we're home and dry.
4  Harry labels point

midpoint of

5  Robin expands out the

a, b, c

inequality for

a2 bc + ab2 c + abc2 2a2 b2 + 2b2 c2 + 2a2 c2 a4 b4 c4

Then he move the fourth powers to the LHS and requires to prove that:

Now

a2 bc + ab2 c + abc2 + a4 + b4 + c4 2a2 b2 + 2b2 c2 + 2a2 c2

P
t
he ites S hur's inequality (
y x (x y)(x z) 0) with
X
y

t=2:

a2 (a b)(a c) 0

whi h expands out to

a4 + b4 + c4 + a2 bc + ab2 c + abc2

a3 b

sym

The LHS is the LHS of his RTP (required to prove) inequality. Now it
su es to show that:

26

X
sym

a3 b

a2 b2

sym

Whi h yields trivially to Muirhead.

6  Noting that

a+b > c

and so on, I onsider the triangle with sides of

length a, b and . The inequality at hand is

a2 bc + ab2 c + abc2
Now by the osine law

X
(a2 + b2 c2 )(a2 + c2 b2 )
y

a2 + b2 c2 = 2abcosC ,

so this be omes

a2 bc + ab2 c + abc2 4a2 bc cos B cos C + 4ab2 c cos A cos C + 4abc2 cos A cos B
The area of a triangle (whi h I all

formula gives

2/ sin C = ab

2
and so on, so after this substitution, division by 4 and multipli ation by

sin A sin B sin C

I get

sin A+sin B+sin C 4 sin A cos B cos C+4 sin B cos A cos C+4 sin C cos A cos B
The RHS is rewritten as

2 sin A cos B cos C+2 sin B cos A cos C = 2

cos C(sin A cos B+sin B cos A)

sin A cos B + sin B cos A = sin(A + B) by the ompound angle

sin(A + B) = sin(180 A B) = sin C , so the RHS is just

formula,

and

2 sin A cos A

This is the double angle formula, so I am required to prove:

sin A + sin B + sin C sin 2A + sin 2B + sin 2C

Now I re all that
and

cos A cos B .

ABC

is a ute, so for any

A B , 2 sin A 2 sin B

Hen e by the rearrangement inequality

2(sin A cos B + sin B cos A) 2 sin A cos A + 2 sin B cos B

27

2 sin C sin 2A + sin 2B

Taking the sum of this inequality y li ally and dividing by 2 gives

sin A + sin B + sin C sin 2A + sin 2B + sin 2C

as required.

7 

x + y 2 xy

(by simple AM-GM)

(x + y)z 2z xy
(multiplying by z)

z 2 + (x + y)z + xy z 2 + 2z xy + xy
(adding

z 2 + xy )
(z + x)(z + y) (z +

2
xy)

(fa torising)

z + x z + y z + xy
(rooting)

(x + y) z + x z + y (x + y)(z + xy)
(multiplying by (x+y)).

This applies y li ally, so:

LHS

LHS

Sin e

x+y

xy

(x + y)(z +

xy)

xy +

X
y

by AM-GM

28

(x + y) xy

LHS

xy + 2

LHS

LHS

RHS

X
xy xy
y

xy

as required.

8  Daniel starts with the standard

a, b, c substitution,

then expands out

the RHS, fa torises the LHS and requires:

(a + b + c)abc = 2

X
y

a2 b2

a4

(a + b + c)(a + b c)(a + c b)(b + c a),

a, b and c,

like Robin. The RHS fa torises to

so for a triangle with sidelengths

2abc(a + b + c)/2 (a + b + c)(a + b c)(a + c b)(b + c a)

The RHS is Heron's formula squared multiplied by 16, so

2abc(a + b + c)/2 162

We write one

as

abc
4R and the other as

2abcs 16

rs,

so

abcrs
4R

Whi h simplies to

R 2r ,

Euler's inequality.

9  Matthew will prove by indu tion, rst showing that for any given

n + 1 also works and then looking at the base ase when n = 1.

Pn is good from a to b if:
a y b Y Pn su h that 0 y SY 2n
n+1 2n+1 then it's great.
If Pn is good from 0 to 3

29

n,

Claim 1:

Pn

SPn = 3n+1 2n+1

Proof: I did this by indu tion but sin e I did MPC2 yesterday and I like
geometri series:
n+1

SPn = a 1r
1r

n+1

= 2n

1 32
1 32

= 3n+1 2n+1 .

great Pn , U0 , U1 , U2 , . . . , Um Ui Pn su h that SUi

U0 = , Um = Pn and SUj SUj1 2n 0 < j m.

Claim 2:
in reasing

are

Proof: Only the nal ondition needs to be proved as all an be made to be

true or have been shown before. Suppose

2n .

SUj 1 SUj1 <

n
LHS is an integer so SUj SUj1 < 2 .
and

y = SUj 1

Claim 3: This is also a su ient ondition for

Pn

then

y SUj1 < 2n

being great

SUj1 y < SUj , whi h an be true for all relevant y. Subtra t

0 < y SUj1 < SUj SUj1 and it follows that 0 y SUj1 2n .

Proof: Let

SUj1

Pn1 is great implies Pn is great. n 2

X0 , X1 , . . . Xm be the list of subsets des ribe above for Pn1 . Let
V0 , V1 , . . . Vm be subsets of Pn with elements twi e that of orresponding X
n
and let W0 , W1 , . . . Wm be the same subsets but also ontaining 3 .
Claim 4:

Proof: Let

Now:

9 3n2 8 2n2 > 0(n 2)

2 (9 3n2 ) 8 2n2 > 9 3n2
2 (3n 2n ) > 3n
2 (SXm ) > 3n P
SVm > SW0
This means the set ontaining all

Vi

and

Wi is suitable to show Pn is good

Xi whi h is suitable and the two
of y and overlap so this makes Pn

be ause dieren es of sums is twi e that of

sets have sums overing the whole range
great.
Finally note P1 is great:
, {2}, {3}, {2, 3} Pn
and give values of SY 0,2,3,5. whi h
and by indu tion all Pn are great.

shows

P1

is great (0

y 32 22 = 5)

10  Daniel starts by supposing that the fun tion is eventually onstant.

Then

f (n) = k

for all

n n0 ,

for some onstant

30

k.

We prove by downwards indu tion that f (n) = k for all n. If n0 > 1 then
n0 Z+ . If n0 1 3, then k = f ((n0 1)!) = f (n0 1)! as n0 1 > n0
for all n0 1 3. We also have f (n0 !) = k sin e n! n n, so f (n0 )! = k .
Moreover f (n0 ) = k so f (n0 )! = k!, so k! = k . We have k = f (n0 1)!, so
k! = f (n0 1)!, k = f (n0 1) whi h ontradi ts the denition that n0 is
the lowest number su h that f (n) = k n n0 . Therefore n0 1 3 is
impossible, so n0 1 < 3, so n0 1 is 1 or 2. Therefore (n0 1)! = n0 1,
so f (n0 1)! = f (n0 1) so f (n0 1) is 1 or 2.
By the se ond rule, f (n0 + 1) f (n0 1) is a multiple of 2. f (n0 + 1) = k ,
(mod 2). Moreover both k and f (n0 1) are 1 or 2, and
so k f (n0 1)
1 6= 2 (mod 2), k = f (n0 1). Therefore the assumption that n0 > 1 was
wrong, but n0 1 so n0 = 1  in other words, if the fun tion is eventually
onstant it must be onstant throughout. f (n) = k n where k is 1 or
2. Both are trivially solutions, so we have all su h fun tions where f (n) is
eventually onstant.
If

f (n).

is not eventually onstant, the

We have

n, k Z+

su h that

f (n + k) 6=

f (n + k)! f (n)!
f ((n + k)!) f (n!)
=
Z
(n + k)! n!
(n + k)! n!
Suppose for ontradi tion that f (n) < n for some n. However, n < n + k
a n! = (n + k)! for some a n + 1. Therefore (a 1)n! = (n + k)! n!,
so n! | (n + k)! n!. (n + k)! n! | f (n + k)! f (n)! by the se ond rule
so n! | f (n + k)! f (n)!. f (n + k)! f (n)! 6= 0 sin e f (n + k) 6= f (n) by
denition. Therefore either f (n + k)! f (n)! n! or it is negative.
If f (n + k)! f (n)! is negative, it is at least f (n)! sin e f (n + k)! is
positive. Moreover it is at most n! sin e it divides n!, so f (n)! n!,
whi h implies that n! f (n)!, so n f (n) whi h ontradi ts our denition.
Hen e f (n + k)! f (n)! n!, so f (n + k)! > n!, so f (n + k) > n. Therefore
n! | f (n + k)!, but also n! | f (n + k)! f (n)!, so n! | f (n)! whi h implies that
n f (n)  ontradi tion.
Therefore f (n) n n. As we saw in the ase where f (n) is eventually
onstant, f (2) = f (2!) = f (2)! so f (2) is its own fa torial, so f (2) is 1 or 2.
However sin e f (2) 2, f (2) = 2.
Now suppose there exists n 3 su h that f (n) n! 2. Sin e
so

f (n!) f (2!)
Z
n! 2!
f (n!) 2
Z.
n! 2
31

n! 2 f (n) so n! 2 | f (n)!, whi h implies from the two divisibilities that

n! 2 | 2. However, n > 3, so n! 2 > 2, ontradi tion. Therefore n 3,
f (n) n! 3  so f (3) 3! 3 = 3. f (3) 3, so f (3) = 3.
f (1) = f (1!) = f (1)! so f (1) is 1 or 2, but by the se ond rule f (3) f (1)
(mod 2), so f (1) 1 (mod 2) and f (1) = 1.
So we have:

f (1) = 1
f (2) = 2
f (3) = 3
n, n | n! n, and n! n | f (n!) f (n), so n | f (n)! f (n). Now
f (n) n, so n | f (n)!. Hen e n | f (n).
We now use indu tion to show that f (n) = n n. We know this is true
when n is 1, 2 and 3 as our base ases. We suppose that f (n) = n n k
for some k 3. Then
f (k + 1)! f (k!)
Z
(k + 1)! k!
Hen e

f (k + 1)! k!
Z
(k + 1)! k!
f (k!) = f (k)! = k!.
k(k!), so

sin e

The denominator is

(k + 1)! k! = (k + 1)k! k! =

f (k + 1)! k!
Z
k(k!)
f (k + 1) 2k  then (2k)! | f (k + 1)!
k | 2k and k! | (2k 1)!, k(k!) | 2k(2k 1)!, so k(k!) | 2k!, and k(k!) |
f (k + 1)!. Therefore from the two divisibilities, k(k!) | k! but k(k!) > k! > 0,
so this is a ontradi tion. Therefore f (k + 1) < 2k . k + 1 | f (k + 1) from
above, so k +1 f (k +1) < 2k , and the only multiple of k +1 in this interval
is k + 1. Hen e k + 1 = f (k + 1) and indu tion is omplete. Therefore the
only f whi h suits the onditions and whi h is not eventually onstant is
f (n) = n. This works trivially, so the three solutions are f (n) = 1, f (n) = 2
and f (n) = n.
Suppose for a ontradi tion that

and as

11  Invert the plane about

that

is the line

P Q,

A,

AQ. It is not hard to show

B is P and P is B . The new

with radius

and vi e versa. Hen e

32

 ir le must also be in exa tly the same pla e, so it is tangential to

G , whi h is also

G.

Hen e

G=

2R =

12  we start with

G and

a
sin A .

AG = AQ.

Multiplying by

sin A

at

and adding up

y li ally gives:

2R(sin A + sin B + sin C) = a + b + c

Then multiplying by

r
2

Rr(sin A + sin B + sin C) = rs =

=

R2
2 (sin 2A

+ sin 2B + sin 2C),

Rr(sin A + sin B + sin C) =

r(sin A+sin B+sin C) =

R2
(sin 2A + sin 2B + sin 2C)
2
R
(sin 2A + sin 2B + sin 2C)
2

r(sin A + sin B + sin C) =

Euler tells us that

so:

R
2 , so

R
(sin 2A+sin 2B+sin 2C) r(sin 2A+sin 2B+sin 2C)
2

sin A + sin B + sin C sin 2A + sin 2B + sin 2C

and in the words of the great Jzsef Pelikn, we are ready.

13  take any ve points forming a onvex pentagon, and onsider the
set of points inside the pentagon, in lude the 5 that form the onvex hull.
Let the number of bla k points be

and the number of green points be

g.

b-gon, whi h an be triangulated into b 2 triangles,

b 2 green points inside of the onvex hull. Similarly
g 2 blue points inside of the onvex hull, and a total of

The bla k points form a

so there are at least
there are at least
at least

b+g4

points inside of the onvex hull. In luding the onvex hull

of 5, there are at least

b+g+1

points, but by denition there are

ontradi tion.

33

b+g

14  taking everything

log1/4

reverses the inequality, sin e it is a de-

1 and

reasing fun tion. This demonstrates the fa t that multipli ation by

nding the re ipro al are not the only fun tions that reverse the sign  they
are just two ases of the general rule that applying a de reasing fun tion
hanges the sign. This fa t, while intuitive, seems to be skipped in s hools.

15  I start with a mental ate hism (really, it helps).

Me: Why is this problem di ult?
Me: It doesn't allow for heavy ma hinery su h as Muirhead's.
Me: Why doesn't it respond to Muirhead's?
Me: Be ause dierent terms have dierent orders  the expression is not
homogeneous.
Me: How an these orders be adjusted?
Me: If an identity in

x, y

and

exists with dierent orders on the two

sides.
Me: Does su h an expression exist?
Me: No, be ause

x + y + z = 0.
x+y+z = a

Me: Wait a se ond  if

for some non-zero onstant

a,

we're

sorted, right?
Me: Right  we ould multiply everything with lower order by

(x+y+z)/a

repeatedly, in reasing the order indenitely.

Me: Great!
Me: But
Me: But
Me: Oh,

x + y + z = 0.
(x + 1) + (y + 1) + (z + 1) = 3. . .
I see  we substitute in t = x + 1, u = y + 1

and

v = z+1

and

we an make it homogeneous.
Me: And then we an bash it!
Me: The expansion would involve over 900 terms.
Me: Fine, but at least I have a way of solving it.
Me: You don't know it will work.
Me: Five pounds if it works.
Me: Who does the expansion?
Me: Nobody needs to do the expansion.
Me: Why not?
Me: Be ause we have the same wallet, you and I, so whoever is wrong
may as well have just transferred the money.
Me: Well, whi hever way it is, I just pi ked up a ver, so I'm going to
Tes o to buy a elebratory ho olate bar with my hard earned ash.

34

16  Daniel takes us halfway by identifying several similar triangles found

by drawing evians through

and hen e nding an expression that doesn't

look far o what we're required to prove  that

1
1
1
1
+
+
=
2
2
2
sin A sin B sin C
sin2
The re ipro al of the sine is the ose ant, and the square of the ose ant
is 1 more than the otangent, so we have

cot2 A + cot2 B + cot2 C = cot2 2

Bear this equation in mind for a moment. We know that

tan C = tan A tan B tan C ;

tan A + tan B +

dividing this by its own RHS and multiplying by

two gives

cot A cot B = 2

Adding this equation to the earlier one gives

cot2 A + 2

cot A cot B = cot2

Square rooting gives the desired inequality.

Jensen's (along with some dierentiation) gives the orollary.

17  Multiply away the denominators and expand to require

X
y

a3 b3

X
y

X
y

a 3 b3

a 3 b2 c 0

a3 b2 c

Whi h yields to rearrangement with sets

{a2 b2 , a2 c2 , b2 c2 } and {ab, ac, bc}.

18  Dividing by ab , Daniel requires

X a(b c)
y

c(a + b)
35

He now adds 3. Wow.

X a(b c)
y

c(a + b)

+13

X b(a + c)
y

c(a + b)

Whi h is AM-GM sin e the terms multiply to 1.

36