UKMT British Mathematical Olympiad Round 2

Unofficial Solutions

Krish Nigam

Contents
1 BMO2 2002 2
1.1 Problem 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Problem 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Problem 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2 BMO2 2016 7
2.1 Problem 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Problem 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3 Problem 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

3 BMO2 2020 11
3.1 Problem 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

4 BMO2 2021 13
4.1 Problem 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4.2 Problem 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

1
Krish Nigam ([email protected]) BMO2 UNOFFICIAL SOLUTIONS

1 BMO2 2002
1.1 Problem 1

The altitude from one of the vertices of an acute-angled triangle ABC meets the
opposite side at D. From D perpendiculars DE and DF are drawn to the other
two sides. Prove that the length of EF is the same whichever vertex is chosen.

Solution. We want to show that the length EF is independent of which vertex created
it. I have arbitrarily chosen vertex A.

Let the angles at vertex A, B, C be equal to ∠A, ∠B, ∠C respectively.

We know ∠EDC = 90◦ − ∠C =⇒ ∠ADE = ∠C.
Since ∠DEA = 90◦ and ∠AF D = 90◦ , ∠DEA+∠AF D = 180◦ . Therefore, quadrilateral
AFDE is cyclic and we can circumscribe.
By angles in the same segment, ∠ADE = ∠AF E = ∠C. Now, we can use the sine rule
in 4AF E, in order to find an expression for EF .
EF AE sin ∠A
= =⇒ EF = AE · (1)
sin ∠A sin ∠C sin ∠C
However, we want to show that this expression does not change due to the vertex picked.
Let’s consider the area of the total triangle [ABC] and its circumradius R. By the full
sine rule:
AB BC CA
= = = 2R
sin ∠C sin ∠A sin ∠B
Krish Nigam ([email protected]) BMO2 UNOFFICIAL SOLUTIONS

By the area of a triangle formula ‘ 12 ab sin θ’, we see:

1
[ABC] = AB · BC · sin ∠B
2
1
= (2R sin ∠C) · (2R sin ∠A) · sin ∠B
2
= 2R2 sin ∠A · sin ∠B · sin ∠C
sin ∠A
In (1), we established that EF = AE · sin ∠C
. So, I want to find an expression for AE.

AE = AC − EC , EC = CD · cos ∠C
= (AC cos ∠C) · cos ∠C
= AC cos2 ∠C

AE = AC − AC cos2 ∠C = AC 1 − cos2 ∠C = AC sin2 ∠C

=⇒ EF = AC sin ∠A · sin ∠C
By the full sine rule though, AC = 2R sin ∠B.

[ABC]
∴ EF = 2R sin ∠A · sin ∠B · sin ∠C =
R
Hence, the length EF is independent of the vertex chosen since it only depends on the
area of the triangle and its circumradius which are constants pre-defined to the triangle
itself.
Krish Nigam ([email protected]) BMO2 UNOFFICIAL SOLUTIONS

1.2 Problem 2

A conference hall has a round table with n chairs. There are n delegates to the
conference. The first delegate chooses his or her seat arbitrarily. Thereafter the
(k + 1)th delegate sits k places to the right of the kth delegate, for 1 ≤ k ≤ n − 1.
No chair can be occupied by more than one delegate.
Find the set of values n for which this is possible.

Solution. In this problem, we have n chairs so it will make sense to deal with things
modulo n. This is because after going round all n seats you will end up back at the
original seat and the counting resets.
If we label our seats 0, 1, . . . , n − 1 going counterclockwise (i.e. in the right direction, as
per the question), then we have:
• WLOG, the 1st person sits in seat 0
• the 2nd person sits in seat 1
• the 3rd person sits in seat 3
• and the kth person sits in seat ≡ k(k+1)
2
(mod n)
This is because the kth person sits 0 + 1 + 2 + · · · + (k − 1) seats to the right of the first
person.
We should begin by considering what happens for small values of n. If n = 1 and n = 2
then clearly this works. When n = 3, both the first and the third person will end up
in the same seat which is not good. Our seating however works when n = 4. Exploring
further, we should find that the next time this works is for n = 8. From this, I see a
pattern.
I conjecture that the set of values n for which the seating is possible is when n = 2k
for all non-negative integers k.
It is often a good idea to see where our system goes wrong and how this can narrow down
our possibilities for the values of n. For some n, we will have a problem if there are two
people who have the same seat. Going back to the numbering of the seats above, there
cannot exist two different numbers p, q with 1 ≤ p < q ≤ n − 1 such that:
p(p − 1) q(q − 1)
≡ (mod n) (2)
2 2
While it is tempting to divide through by 12 in our above expression, it is better to pause
and make use of an important idea. If a ≡ b (mod n) ⇐⇒ ka ≡ kb (mod kn) for all
real k.1 This idea instead enables us to make much greater progress since we are then
dealing in modulo 2n which gives us more tools than merely dealing in modulo n.
Multiplying (2) by 2 gives

p(p − 1) ≡ q(q − 1) (mod 2n)

1
There is also another rule when k is an integer that a ≡ b (mod n) ⇐⇒ ka ≡ kb (mod n), but this isn’t so
helpful here.
Krish Nigam ([email protected]) BMO2 UNOFFICIAL SOLUTIONS

Now, q(q − 1) − p(p − 1) = (q − p)(p + q − 1) so we have a clash of seating if and only if

(q − p)(p + q − 1) ≡ 0 (mod 2n) (3)

Let’s suppose that 2n is in fact a power of 2 (based on our conjecture). Let’s also suppose,
for contradiction, that there are two different seats p, q with p < q which end up being
occupied. So, from (3), that means 2n is a factor of (q − p)(p + q − 1).
But, observe that exactly one of p − q or p + q − 1 can be even. Therefore, 2n will divide
either of p − q or p + q − 1, but since p − q < n it must be that 2n divides p + q − 1.
However, p + q − 1 < 2q − 1 < 2n as q ≤ n. So, p + q − 1 < 2n and so 2n cannot divide
p + q − 1. Hence, there exists no such different values of p, q.
But that only shows n = 2k is a possible candidate for values of n. We also want to prove
that powers of 2 are the only solutions for n.
So, suppose for contradiction that 2n = xy is not a power of 2. This means that one of x
or y is odd and the other is a power of 2. Without loss of generality, suppose that x < y.
Now, we can manufacture the incident of our seating going wrong as per (3). We can
pick appropriate values for p and q by considering (q − p)(p + q + 1) = 2n. We can take
p = y−x+1
2
and q = x+y+1
2
; we can do this because it satisfies p < q and both y−x+1
2
and
x+y+1
2
are integers due to only one of x, y being odd as established.

Therefore, there is nothing stopping us having someone in seat p = y−x+1

2
and someone
x+y+1
else in seat q = 2 . Consider (q − p)(p + q − 1):
  
x+y+1 y−x+1 y−x+1 x+y+1
(q − p)(p + q − 1) = − + −1
2 2 2 2
= (x)(y)
= 2n

But, this leads to (3) as (q − p)(p + q − 1) ≡ 0 (mod 2n). We can’t have this occur!
Hence, if 2n is not a power of 2, then we break down. This means that n must be a power
of 2.
Krish Nigam ([email protected]) BMO2 UNOFFICIAL SOLUTIONS

1.3 Problem 3

Prove that the sequence defined by

1 p 
y0 = 1, yn+1 = 3yn + 5yn2 − 4 , (n ≥ 0)
2
consists only of integers.

Solution. First let’s find an explicit formula for yn−1 in terms of yn . Our sequence can
also be written as:
 
1
q
2
yn = 3yn−1 + 5yn−1 −4 for n ≥ 1
2
q
2
∴ 2yn − 3yn−1 = 5yn−1 −4

=⇒ 4yn2 + 9yn−1
2 2
− 12yn−1 yn = 5yn−1 −4
4yn2 + 4yn−1
2
− 12yn−1 yn + 4 = 0
2
− 3yn · yn−1 + yn2 + 1 = 0


∴ yn−1

So, what we have above now is a quadratic in yn−1 . So, solving this gives
p
3yn ± 9yn2 − 4 (yn2 + 1) 1 p 
yn−1 = = 3yn ± 5yn2 − 4
2 2
 p 
But, we must reject yn−1 = 12 3yn + 5yn2 − 4 , since this would imply that yn−1 =
yn+1 for all n, but that’s clearly false by counterexample:

(n = 1) y0 = 1, y2 = 5 =⇒ y0 6= y2
 p 
Hence, it must be that yn−1 = 21 3yn − 5yn2 − 4 . Now, consider yn+1 from the ques-
tion.
1 p
2
 1 p
2

yn+1 = 3yn + 5yn − 4 = 3yn − 3yn − 5yn − 4
2 2
= 3yn − yn−1

∴ yn+1 = 3yn − yn−1 for all n ≥ 1

By this, it is clear that all members of the sequence are integers, provided that the first
two are integers.
Checking the first two terms: y0 = 1, y1 = 1, y2 = 3(1) − 1 = 2 ∈ Z.
Krish Nigam ([email protected]) BMO2 UNOFFICIAL SOLUTIONS

2 BMO2 2016
2.1 Problem 1

Circles of radius r1 , r2 and r3 touch each other externally, and they touch a common
tangent at points A, B and C respectively, where B lies between A and C. Prove
that
16(r1 + r2 + r3 ) ≥ 9(AB + BC + CA)

Solution. The fact we are being asked to prove is not necessarily completely obvious,
and it is difficult to see which direction to head in after having examined the setup. But,
we will deduce what we can from the setup and potentially working backwards will be
helpful in seeing how the result is derived.

As can be seen from the diagram, the centres of each of the circles are labelled O, P and
Q. Clearly, OP = r1 + r2 , P Q = r2 + r3 and OQ = r1 + r3 , since all the circles meet
tangentially. We will now aim to obtain expressions for AB, BC and CA in terms of r1 ,
r2 and r3 .
By Pythagoras’, AB 2 = OP 2 − (r1 − r2 )2 = (r1 + r2 )2 − (r1 − r2 )2 = 4r1 r2 =⇒ AB =
√ √ √
2 r1 r2 . Similarly, BC = 2 r2 r3 and CA = 2 r1 r3 . So, the statement required to prove
reduces to
√ √ √
16 (r1 + r2 + r3 ) ≥ 9 (2 r1 r2 + 2 r2 r3 + 2 r1 r3 )
√ √ √
8 (r1 + r2 + r3 ) ≥ 9 ( r1 r2 + r2 r3 + r1 r3 )

In order to avoid dealing with square roots, it would be helpful to make a substitution
√ √ √
at this point. Let x = r1 r2 , y = r2 r3 , z = r1 r3 .
xz xy yz
=⇒ r1 = , r2 = , r3 =
y z x
We also know from the setup AB + BC = AC =⇒ 2x + 2y = 2x =⇒ x + y = z. Using
these constraints, we have turned our problem into an algebraic inequality that we must
prove. So, we must show that
 
xz xy yz
16 + + ≥ 18(x + y + z)
y z x
8(x2 z 2 + x2 y 2 + y 2 z 2 ) ≥ 9(xyz)(x + y + z)
Krish Nigam ([email protected]) BMO2 UNOFFICIAL SOLUTIONS

But since x + y = z, we can reduce the inequality down two just two variables x and y.
So, we get
4 x2 (x + y)2 + y 2 (x + y)2 + x2 y 2 ≥ 9xy(x + y)2

While it may not be the most elegant thing, what we can do now is fully expand and
simplify the terms on the right hand side and the left hand side. Doing so yields

4x4 + 4y 4 + 12x2 y 2 + 8x3 y + 8xy 3 ≥ 9x3 y + 9xy 3 + 18x2 y 2

So, the expression finally reduces and we are required to prove:

4x4 + 4y 4 ≥ x3 y + xy 3 + 6x2 y 2 (4)

But (4) is not too difficult to prove true. By AM-GM

x4 + y 4 p 4 4
≥ x y =⇒ x4 + y 4 ≥ 2x2 y 2 (5)
2
Also, by rearrangement inequality

x4 + y 4 ≥ x3 y + xy 3 (6)

Now doing 3 × (5) + (6)

3x4 + 3y 4 + x4 + y 4 ≥ 6x2 + y 2 + x3 y + xy 3
4x4 + 4y 4 ≥ x3 y + xy 3 + 6x2 y 2

which demonstrates that (4) is true and hence the logic follows backwards, thus it is true
that
16(r1 + r2 + r3 ) ≥ 9(AB + BC + CA) .
Krish Nigam ([email protected]) BMO2 UNOFFICIAL SOLUTIONS

2.2 Problem 3

Let ABCD be a cyclic quadrilateral. The diagonals AC and BD meet at P , and

DA and CB produced meet at Q. The midpoint of AB is E. Prove that if P Q is
perpendicular to AC, then P E is perpendicular to BC.

Solution.
Krish Nigam ([email protected]) BMO2 UNOFFICIAL SOLUTIONS

2.3 Problem 4

Suppose that p is a prime number and that there are different positive integers u
and v such that p2 is the mean of u2 and v 2 . Prove that 2p − u − v is a square or
twice a square.

Solution. From the question we have,

u2 + v 2
p2 = =⇒ 2p2 = u2 + v 2
2
Now, we want to somehow ‘manufacture’ this 2p − u − v to begin considering how it could
be a perfect square. Multiplying both sides by 2 helps us with this.
4p2 = 2u2 + v 2 = u2 − v 2 + u2 + v 2

∴ 4p2 − (u + v)2 = (u − v)2 =⇒ (2p − u − v) (2p + u + v) = (u − v)2

We only need to examine 2p − u − v from the above. The expression above alludes to a
somewhat ‘obvious’ result: if we have two coprime numbers m, n (i.e. gcd(m, n) = 1)
and if mn is a perfect square, then both m and n are also square numbers. This way, it
is handy to consider common divisors of 2p − u − v and 2p + u + v.
Let q be a common prime factor of both 2p − u − v and 2p + u + v. If q divides both
these, then it must also divide their sum ((2p − u − v) + (2p + u + v) = 4p) and their
difference ((2p + u + v) − (2p − u − v) = 2(u + v)). So,
q|4p , q|2(u + v)
Let’s first suppose the case that q is odd. Since q|4p, it must be that q|p, but further
since p is prime, q = p. As a result, p|2(u + v) which means that p|(u + v).
But, right from the beginning, we established that p|(u2 + v 2 ) (as 2p2 = u2 + v 2 ). Hence
p also divides the following difference,
p|(u + v)2 − (u2 + v 2 ) = 2uv =⇒ p|uv
However, if p for example divides u then it also divides v as p|u + v. So, both u and v
are divisible by p and we can write that u = Ap and v = Bp for some factors A and B.
∴ 2p2 = (Ap)2 + (Bp2 ) = p2 A2 + B 2

Though, because A and B are integer factors it should be clear that A = B = 1, and
thus p = u = v. But we can’t have this as u 6= v by the question.
So, we have established that the common prime factor q cannot be odd. Hence it must
be 2, and the greatest common divisor of 2p − u − v and 2p + u + v must be some power
of 2, say 2n .
Going back to the equation we established,
(2p − u − v)(2p + u + v) = (u − v)2
we must have that 2p − u − v = 2n x2 and 2p + u + v = 2n y 2 for some numbers x and
y. This is because gcd(2p − u − v, 2p + u + v) = 2n and the right hand side is a perfect
square which means the left hand side must also be.
Since 2p − u − v = 2n x2 , if n is even then it is a square, and if n is odd then it is twice a
square.
Krish Nigam ([email protected]) BMO2 UNOFFICIAL SOLUTIONS

3 BMO2 2020
3.1 Problem 2

Describe all collections S of at least four points in the plane such that no three
points are collinear and such that every triangle formed by three points in S has
the same circumradius.
(The circumradius of a triangle is the radius of the circle passing through all three
of its vertices.)

Solution. This problem involves us investigating properties of some points in the

plane. Since we must have at least 4 points, we can start with arbitrarily picking any 3
non-collinear points which form some arbitrary triangle.
We can do this because our three points will initially define a circumcircle (and so a
circumradius) uniquely, and then after we can investigate what we do with the next
fourth point and then the next point and so on. So, we start with the following.

Now, we want to find a fourth point, say X, such that the circumcircle formed by AXB,
BXC and CXA all have the same circumradius.
Of course, if our point X lies on the circumcircle of ABC, then that is good enough,
since then AXB, BXC and CXA all have exactly the same circumcircle so the same
circumradius. Similarly, we can do this for points after that, by simply making all the
points concyclic.
However, we must think - is this the only way to create circumcircles with the same
radius? We can play around with this a little bit. Essentially, what we want now is
that our fourth point X does not lie on the circumcircle of ABC, but it still satisfies the
condition that AXB, BXC and CXA have the same circumradius.
This will give us the following setup.
Krish Nigam ([email protected]) BMO2 UNOFFICIAL SOLUTIONS

Notice that we must have the three circumcircles with the same radius intersecting at the
point X. Now, we want to investigate the properties of this point X. A helpful way to
proceed is by joining the centres of two of the circles to form a line say O1 O2 .2
Now, what we have in our diagram are several rhombuses at play. Due to the circles
being constrained to have the same radii, BO1 XO2 is a rhombus, so it’s diagonals are
perpendicular, hence O1 O2 ⊥ BX.
We will bring rhombuses AO3 XO1 , CO2 XO3 and BO1 XO2 into play. Let’s call ∠AO3 O1 =
α and ∠CO2 O3 = β. By diagonals in a rhombus bisecting the angles we see that
∠AO3 C = 2(α + β) =⇒ ∠O1 O3 O2 = α + β (draw the rhombuses if this is unclear).
But, O1 X = O2 X = O3 X, so 4O1 O3 O2 has center X! Hence, by angle at centre is twice
angle at circumference, ∠O1 O3 O2 = 2∠O1 XO2 , so ∠O1 XO2 = ∠AO3 C. Since opposite
angles in a rhombus are equal, we get ∠AO3 C = ∠O2 BO1 .
Therefore, 4BO1 O2 is congruent to 4AO3 C by SAS. Hence, O1 O2 = AC and AO1 =
CO2 (by the same radii), so ACO2 O1 is a parallelogram and O1 O2 is parallel to AC. It
then follows that BX ⊥ AC as well.
We can follow this argument round using lines O1 O3 and O2 O3 , and we will similarly find
that AX ⊥ BC and CX ⊥ AB. So X is the orthocentre of 4ABC!
So, given three points in the plane A, B, C, the next point can either lie on the circum-
circle of ABC or on the orthocentre of ABC. Since there is only one unique orthocentre,
once a point is on the orthocentre, the other points must fill up on the circumcircle of
ABC.

2
It doesn’t really matter which two centres we choose; the argument remains the same due to the symmetry
of the problem.
Krish Nigam ([email protected]) BMO2 UNOFFICIAL SOLUTIONS

4 BMO2 2021
4.1 Problem 1

A positive integer n is called good if there is a set of divisors of n whose members

sum to n and include 1. Prove that every positive integer has a multiple which is
good.

Solution. We have n ∈ Z+ is good if there exists a set S such that:

S = {1, a1 , a2 , . . . , ak }
where all members of S are divisible by n and 1 + a1 + a2 + · · · + ak = n.
Clearly, n can never be good and prime, since the divisors of n are 1 and n, and 1 6= n
or 1 + n 6= n. Trying small examples, we can immediately see that n = 6 is good, since
when n = 6, S = {1, 2, 3}. It is tempting to suppose that all multiples of 6 are good,
since both 12 and 18 are. However, 66 is not good, and in fact any number of the form
6p, where p ∈ P and p ≥ 11, is never good.
Naturally, it would make sense to try the next best case of multiples of 12, however again
we encounter a similar problem. 12 × 29 = 348 is not good, and in fact any number of
the form 12p, where p ∈ P and p ≥ 29, is never good.
More investigating is required here, and perhaps it is not sensible to go down the route
of simple ‘multiples’.
n = 6 S = {1, 2, 3}
n = 12 S = {1, 2, 3, 6}
n = 24 S = {1, 2, 3, 6, 12}
After looking at some examples of good numbers, we can make a key observation:
If k is a good number, then 2k is also a good number. Further, if k is a good number,
then 2x k is also a good number for all positive integers x.
To prove this, let’s suppose we have our number k which is good. Thus, there must exist
a set S = {1, a1 , a2 , . . . , ar }, with 1, a1 , a2 , . . . , ar being divisors of k and k = 1 + a1 +
a2 + · · · + ar . Now, if we add the number k itself to S, then the sum of members in S
becomes
1 + a1 + a2 + · · · + ar + k = k + k = 2k
and since k is divisible by all of 1, a1 , a2 , . . . , ar , 2k is also divisible by them as well as k,
so 2k is also good. The logic follows repetitively, so multiplying by 2 creates more good
numbers, and hence 2x k is always good too.
In the examples illustrated above, we have considered examples with n = 2x × 3, however
we can generalise this further to n = 2x × p, where p is any odd integer.
I conjecture that all numbers of the form n = 2x × p are good. Clearly, however,
there needs to be some restrictions on the value of x since 2 × 7 = 14 is not good, but
22 × 7 = 28 is good.
Consider n = 2x × p. We take p as being odd. It is a common fact that every number can
be expressed as the sum of distinct powers of two,3 and more specifically if the integer is
3
This is equivalent to saying that every base 10 number can be written in binary.
Krish Nigam ([email protected]) BMO2 UNOFFICIAL SOLUTIONS

odd then when it is expressed as a sum of distinct powers of two, one of the powers is
20 = 1. So, we can express p as

p = 20 + 2a1 + 2a2 + · · · + 2ak = 1 + 2a1 + 2a2 + · · · + 2ak

Now, we can write n = 2x × p as

n = 2x p − p + p = (2x − 1) p + p

Recall the common factorisation (2x − 1) = (2 − 1) (2x−1 + 2x−2 + · · · + 1) = 1+2+· · ·+

2x−2 + 2x−1 .

∴ n = 1 + 2 + · · · + 2x−2 + 2x−1 p + (1 + 2a1 + 2a2 + · · · + 2ak )

= 1 + 2a1 + 2a2 + · · · + 2ak + p + 2p + · · · + 2x−2 p + 2x−1 p

We can clearly see that the above expression for n is a sum of various factors of n. Thus,
n = 2x ×p is good. However, we must have x being large for safety (which isn’t a problem
at all), so that we can have the full sum of factors hold.
Hence, we have proved that numbers of the form n = 2x × p for any large, positive integer
x and positive, odd integer p are good.
Thus, n is obviously a multiple of all odd integers, and it is also a multiple of all even
integers by suitably choosing a p and large enough x.
Krish Nigam ([email protected]) BMO2 UNOFFICIAL SOLUTIONS

4.2 Problem 3

Let ABC be a triangle with AB > AC. Its circumcircle is Γ and its incentre is I.
Let D be the contact point of the incircle of ABC with BC.
Let K be the point on Γ such that ∠AKI is a right angle.
Prove that AI and KD meet on Γ.

Solution. Let the lines AI and KD meet at a point M , and let points E and F
lie on lines AC and AB respectively such that IE is perpendicular to AC and IF is
perpendicular to AB.

We start by proving that 4KEC is similar to 4KF B. Since the point E lies on AC
and point F lies on AB, ∠ECK = ∠F BK by angles in the same segment (chord AK).
Now, note that points A, F , I, E, K are concyclic with diameter AI. This is because
AF IE is cyclic as ∠AF I + ∠IEA = 90◦ + 90◦ = 180◦ , and ∠IKA = 90◦ = ∠AIE so K
lies on this circle too by converse of angles in the same segment.
Thus, by angles in same segment (chord AK), ∠AEK = ∠AF K =⇒ ∠KEC = ∠KF B.

∴ 4KEC ∼ 4KF B (by AA)

Krish Nigam ([email protected]) BMO2 UNOFFICIAL SOLUTIONS

So we can deduce the following relation.

KC CE
=
KB BF
But, CE = CD and BF = BD, since tangents from the same point are of equal length.4
KC CD
∴ = (7)
KB BD
The relation in (7) implies that line KDM is the internal bisector of ∠CKB, by converse
of the angle bisector theorem.
Now, ∠CKB = ∠CAB by angles in the same segment (chord BC). Let ∠CKM = α, so
∠CKB = 2α = ∠CAB.
But, line AM passes through the point I (the incentre) so line AIM is the internal angle
bisector of ∠CAB. This is because 4AIE ∼ = 4AF I by SAS. So, ∠M AB = α.
Hence, since ∠M KB = α = ∠CKM , then by converse of angles in the same segment,
point M lies on the circumcircle Γ of ABC.

4
These lines are tangents to the incircle because the radii points of contact are at 90◦ , and we apply converse
of ‘radius perpendicular to tangent’.