MT

UK
MT

UK
UKMT

United Kingdom
Mathematics Trust

British Mathematical Olympiad

Round 2

© 2022 UK Mathematics Trust

Solutions
British Mathematical Olympiad Round 2 2022 Solutions

1. For a given positive integer 𝑘, we call an integer 𝑛 a 𝑘-number if both of the following
conditions are satisfied:
(i) The integer 𝑛 is the product of two positive integers which differ by 𝑘.
(ii) The integer 𝑛 is 𝑘 less than a square number.
Find all 𝑘 such that there are infinitely many 𝑘-numbers.

Solution
Note that 𝑛 is a 𝑘-number if and only if the equation

𝑛 = 𝑚 2 − 𝑘 = 𝑟 (𝑟 + 𝑘)

has solutions in integers 𝑚, 𝑟 with 𝑘 ≥ 0.

The right-hand equality can be rewritten as

𝑘 2 − 4𝑘 = (2𝑟 + 𝑘) 2 − (2𝑚) 2 ,

so 𝑘-numbers correspond to ways of writing 𝑘 2 − 4𝑘 as a difference of two squares, 𝑁 2 − 𝑀 2

with 𝑁 > 𝑟 and 𝑀 even (which forces 𝑁 to have the same parity as 𝑘).
Any non-zero integer can only be written as a difference of two squares in finitely many ways
(because each gives a factorisation, and a number has only finitely many factors).
If 𝑘 ≠ 4 then 𝑘 2 − 4𝑘 ≠ 0, and as a result, if 𝑘 ≠ 4 then there are only finitely many 𝑘-numbers.
Conversely, if 𝑘 = 4 then setting 𝑚 = 𝑟 + 2 for 𝑟 ≥ 0 shows that there are infinitely many
4-numbers.

© 2022 UK Mathematics Trust www.ukmt.org.uk 2

British Mathematical Olympiad Round 2 2022 Solutions

2. Find all functions 𝑓 from the positive integers to the positive integers such that for all
integersa 𝑥, 𝑦 we have:

2𝑦 𝑓 ( 𝑓 (𝑥 2 ) + 𝑥) = 𝑓 (𝑥 + 1) 𝑓 (2𝑥𝑦).

a The final instance of the word integers was added retrospectively to avoid ambiguity.

Solution
2 𝑓 ( 𝑓 (1)+1)
First substitute 𝑥 = 1 to see that 𝑓 (2𝑦) = 𝑘 𝑦 for all positive integers 𝑦, where 𝑘 = 𝑓 (2) .
By taking 𝑦 = 1, we get 𝑓 (2) = 𝑘, so 𝑘 is a positive integer.
Next, substitute 𝑥 = 2𝑧 and 𝑦 = 1 to see that 𝑓 (2𝑧 + 1) = 𝑘 𝑧 + 1 for all positive integers 𝑧.
Then substitute 𝑥 = 2𝑧 + 1 and 𝑦 = 1 to find that 𝑘 = 2. So 𝑓 (𝑥) = 𝑥 for all integers 𝑥 ≥ 2.
2 𝑓 ( 𝑓 (1)+1)
Using 𝑘 = 𝑓 (2) we find that 𝑓 (1) = 1, and so 𝑓 is the identity.
This is easily checked to satisfy the functional equation.

Remark
The question was originally posed without the final instance of the word integers. This gave
rise to an alternative interpretation of the problem where 𝑥 and 𝑦 can be any numbers such that
𝑥 2 , 𝑓 (𝑥 2 ) + 𝑥, 𝑥 + 1 and 2𝑥𝑦 are integers. In this case 𝑥 must be a positive integer, but 𝑦 can be
a positive half integer. This variant can be solved in a similar, though slightly quicker, way.

© 2022 UK Mathematics Trust www.ukmt.org.uk 3

British Mathematical Olympiad Round 2 2022 Solutions

3. The cards from 𝑛 identical decks of cards are put into boxes. Each deck contains 50
cards, labelled from 1 to 50. Each box can contain at most 2022 cards. A pile of boxes
is said to be regular if that pile contains equal numbers of cards with each label. Show
that there exists some 𝑁 such that, if 𝑛 ≥ 𝑁, then the boxes can be divided into two
non-empty regular piles.

Solution
Suppose a pile of boxes contains 𝑎𝑖 copies of card 𝑖. We label the pile with the tuple (𝑑2 , 𝑑3 , . . .)
where 𝑑𝑖 = 𝑎𝑖 − 𝑎 1 . So a pile is regular if and only if its label is (0, 0, . . . , 0).
It is enough to construct one regular pile, since the remaining boxes form another regular pile.
Suppose we have enough cards to ensure that there are 𝑃 non-empty boxes, where 𝑃 is some
large number to be chosen later. We may view each of these boxes as pile. This is our first
collection of piles.
Their labels all have the property that for all 𝑖, |𝑑𝑖 | ≤ 2022.
Í
We also have 𝑑2 = 0 where the sum is taken over all the piles.
Now suppose that the maximum value of |𝑑2 | = 𝑀. We aim to form a new collection of piles
such that each new pile is either one of the old piles, or is formed by combining exactly two old
piles. If we have some old piles with 𝑑2 = 𝑀 and others with 𝑑2 = −𝑀 we pair these up to
form new piles with 𝑑2 = 0. Once we have done this as many times as possible, the remaining
piles with |𝑑2 | = 𝑀 all have 𝑑2 with the same sign. Consider such a pile: if it has 𝑑2 = 𝑀 we
combine it with any old pile with a negative value of 𝑑2 . There are sure to be enough of these,
since the 𝑑2 values sum to zero. The case where the signs are reversed is identical.
After this process we have at least 𝑃/2 piles. For these piles the maximum value of |𝑑2 | has
decreased (by at least one) and the maximum value of |𝑑𝑖 | for each other 𝑖 has at most doubled.
Thus if we repeat this process (up to) 2022 times we will reach a situation where we have at
least 𝑃/(22022 ) piles and each pile will have 𝑑2 = 0 and |𝑑𝑖 | ≤ 2022 × 22022 for all other 𝑖.
Now we may run this argument again working with 𝑑3 instead of 𝑑2 , then again with 𝑑4 and so
on. More formally, we proceed by induction.
Suppose that for some 𝑘 we have a collection of 𝑃 𝑘 piles such that:
• For each pile 𝑑2 = 𝑑3 = · · · = 𝑑 𝑘 = 0 and
• For all piles and all 𝑖 > 𝑘 we have |𝑑𝑖 | ≤ 𝑀𝑘 for some fixed 𝑀𝑘
Then, by combining the piles as described above, we can reach a situation where we have at
least 𝑃 𝑘 /(2 𝑀𝑘 ) piles, each of which has 𝑑 𝑘+1 = 0 and |𝑑𝑖 | ≤ 2 𝑀𝑘 for all 𝑖.
Setting 𝑃 𝑘+1 = 𝑃 𝑘 /(2 𝑀𝑘 ) and 𝑀𝑘+1 = 2 𝑀𝑘 we have the same situation as before but with 𝑘 + 1
in place of 𝑘.
Thus, if we take 𝑃 large enough, we can ensure that 𝑃50 ≥ 2 which is enough to solve the
problem.

© 2022 UK Mathematics Trust www.ukmt.org.uk 4

British Mathematical Olympiad Round 2 2022 Solutions

4. Let 𝐴𝐵𝐶 be an acute angled triangle with circumcircle Γ. Let 𝑙 𝐵 and 𝑙𝐶 be the lines
perpendicular to 𝐵𝐶 which pass through 𝐵 and 𝐶 respectively. A point 𝑇 lies on the
minor arc 𝐵𝐶. The tangent to Γ at 𝑇 meets 𝑙 𝐵 and 𝑙𝐶 at 𝑃 𝐵 and 𝑃𝐶 respectively. The
line through 𝑃 𝐵 perpendicular to 𝐴𝐶 and the line through 𝑃𝐶 perpendicular to 𝐴𝐵 meet
at a point 𝑄. Given that 𝑄 lies on 𝐵𝐶, prove that the line 𝐴𝑇 passes through 𝑄.
(A minor arc of a circle is the shorter of the two arcs with given endpoints.)

Solution
Note that 𝑄 is sufficient information to construct 𝑃 𝐵 and 𝑃𝐶 .
Let 𝑇 ′ be the second intersection of the line 𝐴𝑄 and Γ.

Denote the foot of the perpendicular from 𝑃 𝐵 to 𝐴𝐶 by 𝑈. Then 𝑃 𝐵 𝐵𝑈𝐶 is cyclic, as is 𝐴𝐵𝑇 ′𝐶.
Consequently: ∠𝑄𝑃 𝐵 𝐵 = ∠𝑈𝑃 𝐵 𝐵 = ∠𝐶 = ∠ 𝐴𝑇 ′ 𝐵 so 𝑃 𝐵 𝐵𝑄𝑇 ′ is also cyclic.
In particular, ∠ 𝐴𝑇 ′ 𝑃 𝐵 = 90◦ .
But the same holds for ∠𝑃𝐶 𝑇 ′ 𝐴 = 90◦ .
So 𝑃 𝐵 , 𝑇 ′, 𝑃𝐶 are collinear. This implies that 𝑇 ′ = 𝑇 since the conditions in the question mean
there is only one point on both the line 𝑃 𝐵 𝑃𝐶 and the circle Γ.

Remark
All successful synthetic solutions to this problem began by defining a new point 𝑇 ′ with some
useful additional properties, and then proving 𝑇 ′ = 𝑇. In the solution above 𝑇 ′ is on the line
𝐴𝑄 and on Γ. A variation on this theme is to define 𝑇 ′ to be on the line 𝐴𝑄 and the line 𝑃 𝐵 𝑃𝐶 .
The solutions below provide two further alternatives.

© 2022 UK Mathematics Trust www.ukmt.org.uk 5

British Mathematical Olympiad Round 2 2022 Solutions

Alternative
Let 𝑇 ′ be the second intersection of circles 𝑃𝐶 𝐶𝑄 and 𝑃 𝐵 𝐵𝑄.

The right angles in the question show that 𝐶𝑈𝐵𝑃 𝐵 is cyclic (with diameter 𝐶𝑃 𝐵 ) so ∠𝑈𝑃 𝐵 𝐵 =
∠𝐶 = 𝑄𝑃 𝐵 𝐵.
Similarly ∠𝐶𝑃𝐶 𝑄 = ∠𝐵.
The right angles show that 𝑇 ′ is on the line 𝑃 𝐵 𝑃𝐶 .
The two angle facts established above show that 𝐶𝑇 ′ 𝐵 = ∠𝐵 + ∠𝐶, so 𝑇 ′ lies on the arc 𝐵𝐶 of
circle Γ. Thus 𝑇 ′ = 𝑇.
Now ∠𝐶𝑇 𝐴 = ∠𝐵 using the cyclic quad 𝐴𝐵𝑇𝐶, while ∠𝐶𝑇𝑄 = ∠𝐶𝑃𝐶 𝑄 using the cyclic quad
𝐶𝑄𝑇 𝑃𝐶 .
We have already shown ∠𝐶𝑃𝐶 𝑄 = ∠𝐵 so we are done.

Alternative
Let 𝑇 ′ be the point on Γ diametrically opposite 𝐴.
Let ∞⊥ℓ denote the point at infinity on the line perpendicular to ℓ.
By (the converse of) angles in a semi circle, 𝑇 ′ = 𝐵∞⊥𝐴𝐵 ∩ 𝐶∞⊥𝐴𝐶 .
Applying Pappus’ theorem to lines 𝐵𝑄𝐶 and ∞⊥𝐴𝐶 ∞⊥𝐵𝐶 ∞⊥𝐴𝐵 gives us that 𝑃 𝐵 𝑃𝐶 𝑇 ′ are
collinear. Thus 𝑇 ≡ 𝑇 ′.
If 𝑄˜ = 𝐴𝑇 ′ ∩ 𝐵𝐶 then from 𝐵𝑃 𝐵 𝐴′𝑄,
˜ 𝐶𝑃𝐶 𝐴′𝑄˜ cyclic we get 𝑃 𝐵 𝑄⊥𝐴𝐶
˜ ˜
and 𝑃𝐶 𝑄⊥𝐴𝐵 so in
˜
fact 𝑄 ≡ 𝑄 and thus 𝐴, 𝑄, 𝑇 are collinear on the diameter passing through 𝐴.

© 2022 UK Mathematics Trust www.ukmt.org.uk 6