BMOS MENTORING SCHEME (Senior Level)

December 2012 (Sheet 3)

Solutions

These solutions represent one way of solving the questions. They are not the only way,
and just because your solution is different does not mean that it is worse or wrong. You
will get most from the solutions if you have at least attempted the questions. Only then
will you appreciate the difficulties that a problem posed and how they were addressed by
the solution. Relevant theorems have been identified where possible, and references to
further information on the Web have been given. If you are still unclear, ask your mentor.
A. Rzym UKMT
c 2012.

1. Suppose we have a set of integers summing to 2009 with the largest possible product. We
will make some deductions about the properties of the set that we can combine to tell us
what the maximum product is1 .
Suppose there are two terms whose values are k and 1. We can combine them into a single
term – preserving the sum but increasing the product (since k + 1 > k). This contradicts
our assumption that the set had the maximum possible product. We conclude that no
terms are equal to 1.
The set can’t contain any number k ≥ 5, since we would be able to replace it with the
two numbers 2 and k − 2. This preserves the sum but increases the product (because
2(k − 2) > k for k ≥ 5).
Thus all of the numbers are between 2 and 4 (inclusive).
Any term taking the value 4 can be replaced by 2 and 2, preserving both the sum and
the product. Thus there exists a set, achieving the maximum product P , consisting only
of 2’s and 3’s.
In this set, there cannot exist three 2’s (since they can be replaced by two 3’s). Thus there
must be at least d 2009−6
3
e = 668 threes, and the remaining sum of 5 can only be achieved
with one 3 and one 2. We conclude that the maximum product is given by 2 × 3669 . 

2. We observe the trivial solutions {k, k} for any positive integer k (00 being undefined). In
what follows, we assume that m, n are distinct integers greater than 1.
We can write m, n in terms of their prime factorisations:
Y Y
m= pm
i
i
, n = pni i

(where the pi are the consecutive primes) and therefore mn = nm implies

Y Y
pnm
i
i
= pmn
i
i
.

By the Fundamental Theorem of Arithmetic, we can equate powers of primes, i.e. nmi =
mni ∀i. Without loss of generality, we can assume that m > n and therefore mi > ni ∀i.
1
The use of calculus in this question is unlikely to be of much help since we are interested in variables that
can only take discrete values (integers).

For more information about the mentoring schemes, and how to join, visit
http://www.mentoring.ukmt.org.uk/
 But this implies m is a multiple of n, i.e. m = kn for some integer k > 1. Substituting
this into the given equation gives

(kn)n = nkn
k = nk−1 .

Since n, k ≥ 2, the only solution to this equation is n = k = 2 (for all other values, the
RHS is greater than the LHS), whence m = 4.
We conclude that the non-negative solutions to the given equation are {k, k} (for all
positive integer k) and {2, 4} (without regard to order). 

3. Suppose p(x) = pn xn + · · · + p0 is a polynomial of degree n with integer coefficients, and

α is an integer such that p(α) = 0.
Then

p(x) = (x − α)q(x) (1)

where q(x) = qn−1 xn−1 + · · · + q0 is a polynomial of degree n − 1.

By equating descending powers of x in turn in (1) we deduce that all of the qi are integers.
This argument can be repeated to allow us to conclude that if p(x) has integer roots
α, β, γ, δ then

p(x) = (x − α)(x − β)(x − γ)(x − δ)q(x)

where q(x) has integer coefficients.

If we now define p(x) = f (x) − 7, we know that p(a) = p(b) = p(c) = p(d) = 0 and

p(x) = (x − a)(x − b)(x − c)(x − d)q(x). (2)

Does there exist an x such that p(x) = 7 (i.e. f (x) = 14)?

We know that q(x) is an integer (since all of its coefficients are integers) and hence, from
(2), p(x) is a product of five integers. But the first four cannot be the same (since a, b, c, d
are distinct). Thus p(x) must have at least two factors (other than ±1) and therefore
cannot be prime. We conclude that p(x) cannot take a prime value and therefore f (x)
cannot take the value 14. 

4. (a) We observe that while the LHS of the given equation is symmetric in a, b, the RHS
is not. Thus the given equation is unlikely to be true. We can construct a coun-
terexample by setting a = d = 2 and b = c = 3. Then

(a, b)(c, d) = (2, 3)(3, 2) = 1 × 1 = 1,

(ac, bd) = (6, 6) = 6.

We conclude that the given equation is not true for all a, b, c, d. 

(b) Write the prime factorisation of a, b, c as
Y Y Y
a= pai i , b= pbi i , c= pci i

where the pi are the primes.

2
 Then
Y min(ai ,bi )
Y min(ai ,bi ,ci )
(a, b) = pi , (a, b, c) = pi

and
Y max(ai ,bi )
Y max(ai ,bi ,ci )
[a, b] = pi , [a, b, c] = pi .

Consequently
[a, b, c]2 Y 2 max(a ,b ,c )−max(a ,b )−max(b ,c )−max(c ,a )
i i i i i i i i i
= pi
[a, b][b, c][c, a]
and
(a, b, c)2 Y 2 min(a ,b ,c )−min(a ,b )−min(b ,c )−min(c ,a )
i i i i i i i i i
= pi .
(a, b)(b, c)(c, a)
Therefore, the given equation is always true if and only if
2 max(ai , bi , ci ) − max(ai , bi ) − max(bi , ci ) − max(ci , ai ) =
2 min(ai , bi , ci ) − min(ai , bi ) − min(bi , ci ) − min(ci , ai ). (3)
for all i.
We will prove that (3) is always true by considering the various orderings of ai , bi , ci .
• Suppose ai , bi , ci are all equal. Then both sides of (3) equal −ai .
• Suppose exactly two of ai , bi , ci are equal. Given the symmetry of the problem,
we just need to consider two cases:
If ai = bi > ci then both sides of (3) are equal to −ai .
If ai = bi < ci then both sides of (3) are equal to −ai .
• If ai , bi , ci are distinct, without loss of generality we can assume that ai > bi > ci ,
in which case both sides of (3) are equal to −bi .
We have shown that (3) is always true and therefore so is the given equation. 
5. (a) We will use proof by contradiction to derive the required result. If Q(x, y) is a point
on the plane then we will write C(Q) as shorthand for C(x, y).
√
Pick an arbitrary point, O, on the plane. Let γ1 , γ2 denote the circles of radius 1, 3,
both with centre O.
Without loss of generality, assume C(O) = Red. Pick any point, W , on γ2 . Then
there exist exactly two points Wa , Wb on γ1 such that
W WA = W Wb = Wa Wb = Wa O = Wb O = 1.
Now C(Wa ), C(Wb ) must be Green or Blue (in any order), otherwise in the equilateral
triangle (which has unit side length) Wa Wb O there would be two vertices of the same
colour.
If C(Wa ), C(Wb ) are Green and Blue then (by considering the equilateral triangle
W Wa Wb of unit side length) W must be Red. But W is any point on γ2 . We
conclude that all points on γ2 are the same colour.
But this means that there are two points on γ2 which are unit distance apart of the
same colour. This is a contradiction, and we conclude that there always exist two
points, A and B such that AB = 1 and C(A) = C(B). 

3
 (b) We will prove the required result by construction. Denote the eight possible values
by the integers 1 . . . 8. We will distinguish between the mutually exclusive points of
a triangle: in the triangle; on an edge; and on a vertex.
Consider a tesselation of the plane with equilateral triangles of unit side length.
Number the insides of the triangles such that the clockwise ordering of the colourings
(about any vertex) is 6, 7, 8, 3, 2, 1; 8, 5, 6, 1, 4, 3; 1, 2, 5, 8, 7, 4 or 2, 3, 4, 7, 6, 5. If two
triangles (whose inside numbers are x and y) share a common side, number the edge
max(x, y).
If the clockwise numbering about a vertex is 6, 7, 8, 3, 2, 1 (respectively 8, 5, 6, 1, 4, 3;
1, 2, 5, 8, 7, 4; 2, 3, 4, 7, 6, 5) then number that vertex 3 (respectively 1; 5; 7).
By inspection, no two points unit distance apart have the same numbering. 
6. We will illustrate two proofs. The first is fairly straightforward – the second is harder but
very elegant.

(a) Applying the Arithmetic Mean–Geometric Mean (AM–GM) inequality to the collec-
tion {x, yz} gives
x + yz √
≥ xyz ≥ 1.
2
similarly
y + zx √
≥ xyz ≥ 1,
2
z + xy √
≥ xyz ≥ 1.
2
Summing the equations gives
(x + y + z) + (yz + zx + xy) ≥ 6
1 + (x + y + z) + (yz + zx + xy) + xyz ≥ 8
i.e. (1 + x)(1 + y)(1 + z) ≥ 8. 

(b) If ai , bi , ci are positive, then (by the Arithmetic Mean-Geometric Mean inequality)
a3i + b3i + c3i
≥ ai b i c i .
3
Summing over the i gives
a3i + b3i + c3i
P P P X
≥ ai b i c i . (4)
3
Next define
âi b̂i ĉi
ai =
, bi = P 3
1/3 , ci = (5)
3 1/3
P P 3
1/3
âi b̂i ĉi

where the âi , b̂i ĉi are positive. Substituting (5) into (4) gives
P
1+1+1 âi b̂i ĉi
≥ 1/3
1/3 P 3
1/3
3


â3i b̂3i
P P
ĉi
X
X
X
X 3
â3i b̂3i ĉ3i ≥ âi b̂i ĉi . (6)

4
 Let ai , bi , ci take the values {1, x1/3 }, {1, y 1/3 }, {1, z 1/3 }, whence (6) becomes
(1 + x)(1 + y)(1 + z) ≥ (1 + (xyz)1/3 )3 ) ≥ 23 = 8. 

The converse is not true. For example, set x = 1, y = 10−4 , z = 102 . Then
(1 + x)(1 + y)(1 + z) ≥ 8
but xyz < 1. 
7. Number the vertices 1, . . . , 29 (clockwise) in the obvious fashion. If two vertices are
numbered a, b, we define their ‘distance’ apart to D(a, b) = a − b mod 29 or D(a, b) =
b − a mod 29 (whichever is the smaller).
The problem can now be restated: ‘Take 7 vertices in a regular 29-gon. Show that it is
always possible to pick two pairs of vertices, (p1 , q1 ) and (p2 , q2 ) such that D(p1 , q1 ) =
D(p2 , q2 ), where there may be zero or one points in common between the pairs, but not
two’.
We will assume the contrary and look to derive a contradiction.
Consider the 7 numbers generated by computing the distances between the consecutive
chosen clockwise vertices. These are distinct and sum to 29. If the smallest distance is
2, the other distances must be 3, 4, 5, 6, 7, 8 (or greater). But these distances sum to 35,
so we conclude that one of the distances must be 1. Similarly, the next smallest distance
cannot be 3 or greater, and therefore must be 2. Similarly, the next smallest distance
must be 3, and repeating the argument we see that the only possible combination of
differences must be {1, 2, 3, 4, 5, 6, 8}.
Label the chosen vertices a1 , . . . , a7 . If D(a1 , a2 ) = 1 then D(a2 , a3 ) 6= 2 (otherwise
D(a1 , a3 ) = 3 and we know that this distance also occurs between adjacent vertices).
Similarly, D(a2 , a3 ) 6= 3, 4, 5.
We conclude that four consecutive chosen vertices have adjacent distances 6, 1, 8 (or the
reverse order).
Now 2 cannot be next to 3,4 or 6 (otherwise duplicate distances 5,6 or 8 are created).
Thus 2 is next to 5 and 8 and therefore we must have adjacent distances 6, 1, 8, 2, 5.
But this is also not possible - because the distance 7 appears twice.
We conclude that three of these points form an isosceles triangle or four of these points
form an isosceles trapezium. 
8. We are given
g((x − y)2 ) = (g(x))2 − 2xg(y) + y 2 . (7)
Setting x = y = 0 in (7) gives g(0) = (g(0))2 , i.e. g(0) = 0 or g(0) = 1. We will now
consider each of these cases in turn.

• g(0) = 0: Setting x = y in (7) gives

g(0) = [g(x)]2 − 2xg(x) + x2
0 = [g(x) − x]2
g(x) = x
and substituting in the original equation shows that it is, indeed, a solution.

5
 • g(0) = 1: Proceeding as above gives

1 = [g(x) − x]2
g(x) = x ± 1.

Note that we may (or may not) have to choose a different sign for each value of x.
Setting y = 0 in (7) gives

g(x2 ) = (g(x))2 − 2xg(0) (8)

x2 ± 1 = x2 ± 2x + 1 − 2x
±1 = ±2x + 1 − 2x. (9)

Equation (9) is telling us that


 x ± 1 if x = 0,
g(x) = x ± 1 if x = 12 ,
x + 1 otherwise


however our supposition that g(0) = 1 allows us to refine these possibilities to

x ± 1 if x = 12 ,

g(x) =
x + 1 otherwise.

Finally, substituting x = 21 in (8) forces us to conclude that we must take g(x) = x+1
everywhere, and substituting this in the original equation shows that it is, indeed, a
solution.

We conclude that the two solutions to the given equation are g(x) = x and g(x) = x+1. 

References

Fundamental Theorem of Arithmetic:

http://en.wikipedia.org/wiki/Fundamental theorem of arithmetic

AM–GM inequality:
http://en.wikipedia.org/wiki/Inequality of arithmetic and geometric means

Golomb Ruler:
http://en.wikipedia.org/wiki/Golomb ruler