Final Report

InfinityDots

April 12, 2018

InfinityDots MO 2 was a Mock IMO contest held on the AoPS Mock Contests forum from March
28 to April 9, 2018.

Contents
Problems 2
Day 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Day 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

Official Solutions and Comments 4

Overall Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Problem 1 : νp of factorials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Problem 2 : Bijection on the set of integers . . . . . . . . . . . . . . . . . . . . . . . . . 5
Problem 3 : Uniqueness of a point with strange properties . . . . . . . . . . . . . . . . . 6
Problem 4 : P (i) divides P (i + 1)? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Problem 5 : ap − b is always divisible by p . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Problem 6 : Every set is symmetric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

Results 15

Final Remarks 15

1
 Language: English

Day: 1

Problem 1. Determine whether there exists a finite set S of primes such that for all positive integers
m, there exists a positive integer n and prime p ∈ S such that pm | n! but pm+1 - n!.

Problem 2. Determine all bijections f : Z → Z satisfying

f f (m+n) (mn) = f (m)f (n)

for all integers m, n.

Note: f 0 (n) = n, and for any positive integer k, f k (n) means f applied k times to n, and f −k (n)
means f −1 applied k times to n.

Problem 3. Let A, B, C be three distinct points on a line `. Prove that for each pair of distinct
←−−→ ←−→ ←−→
points B1 , C1 such that B1 C1 does not pass through A, and B1 C is not parallel to C1 B, there is a
unique point A1 satisfying:
←−−→
(i) A1 does not lie on B1 C1 ,
←−−→ ←−→ ←−−→
(ii) the projections of A onto B1 C1 , of B onto C1 A1 , and of C onto A1 B1 lie on a line not parallel
to `, and
←−−→ ←−→ ←−−→
(iii) the reflections of A over B1 C1 , of B over C1 A1 , and of C over A1 B1 lie on a line not parallel
to `.

Language: English Time: 4 hours and 30 minutes

Each problem is worth 7 points

2
 Language: English

Day: 2

Problem 4. Let P ∈ Z[x] be a nonconstant polynomial without integral roots. Prove that there is
a positive integer m 6 3 · deg P such that P (m) does not divide P (m + 1).

Problem 5. Let c1 , c2 , . . . , ck be integers. Consider sequences {an } of integers satisfying

an = c1 an−1 + c2 an−2 + · · · + ck an−k

for all n > k + 1. Prove that there is a choice of initial terms a1 , a2 , . . . , ak not all zero satisfying:
there is an integer b such that p divides ap − b for all primes p.

Problem 6. Ana has an n × n lattice grid of points, and Banana has some positive integers
a1 , a2 , . . . , ak which sum to exactly n2 . Banana challenges Ana to partition the n2 points in the
lattice grid into sets S1 , S2 , . . . , Sk so that for all i ∈ {1, 2, . . . , k},

(i) |Si | = ai , and

(ii) the set Si has an axis of symmetry.

Prove that Ana can always fulfill Banana’s challenge.

Note: a line ` is said to be an axis of symmetry of a set S if the reflection of S over ` is precisely S
itself.

Language: English Time: 4 hours and 30 minutes

Each problem is worth 7 points

3
Solutions and Comments
Overall Comments.
On difficulty, we feel that our problems this year are generally easier than last year (especially
P5), but still on a Mock IMO level.
On subject balancedness, we decided to go with a single hard geometry problem as we feel
geometry is declining as a subject, and that problems in other subjects are more interesting. Also,
in our problem selection process, we decided to rate each problem by its “subject-ness” which sum
to 1. We’d also like to report that we rated our contest as follows:

Algebra: 1.88, Combinatorics: 1.47, Geometry: 1.00, Number Theory: 1.65

which we find balanced enough.

Problem 1.
Determine whether there exists a finite set S of primes such that for all positive integers m, there
exists a positive integer n and prime p ∈ S such that pm | n! but pm+1 - n!.
Proposed by TacH
Answer. No.
Solution. We will prove, by induction on n, the following stronger statement: for any set of n
primes {p1 , p2 , · · · , pn }, there exists an integer A such that for any A consecutive integers, there is
at least one not in Si , where Si is defined as {νpi (m!) | m ∈ Z+ }.
S

The base case is clear: choose A = p + 1. For the inductive step, it suffices to exhibit a new
value A0 for {p1 , p2 , . . . , pn+1 } in terms of the value of A for {p1 , p2 , · · · , pn }.
We claim A0 = ApA+1 0 0
n+1 + A suffices. Consider A consecutive integers. If the last A − A does
not contain A consecutive integers not in Sn+1 , then Sn+1 must have at least pA+1 n+1 numbers in that
A+1
range (one for each A consecutive integers), representing pn+1 consecutive values of m. One such
m must be divisible by pA+1 n+1 , and [νpn+1 (m!) − A, νpn+1 (m!) − 1] will be our desired interval.
0
Therefore, any A consecutive integers must contain A consecutive integers not in Sn+1 , so using
the property of A, at least one of them must be not in n+1
S
i=1 Si as required. 

Comments. A better A0 in the inductive step is

A + νpn+1 (pA+1 A A−1

n+1 !) = A + pn+1 + pn+1 + · · · + pn+1 + 1,

but this is a bit harder to prove.

4
Problem 2.
Determine all bijections f : Z → Z satisfying

f f (m+n) (mn) = f (m)f (n)

for all integers m, n.

Note: f 0 (n) = n, and for any positive integer k, f k (n) means f applied k times to n, and
f −k (n) means f −1 applied k times to n.
Proposed by talkon
Answer. The infinite family of functions n 7→ n + c for any integer c, and the function n 7→
(−1)n+1 n.
Solution. We consider two cases, depending on the value of f (0).

Case 1: f (0) = 0.
By plugging in (m, n) = (k, −k), we have −k 2 = f (k)f (−k) for all integers k. Since f is
bijective, by induction on k, {f (k), f (−k)} = {k, −k} for all positive integers k. Hence f (f (n)) = n
for all integers n.
Now suppose that m, n are integers with the same parity, so f (m + n) is even. Hence,

mn = f f (m+n) (mn) = f (m)f (n),

so either both f (m) = m and f (n) = n or f (m) = −m and f (n) = −n. Therefore there are four
solutions left to check: n 7→ n, n 7→ −n, n 7→ (−1)n n, and n 7→ (−1)n+1 n, and by considering m
even and n odd, we can see that only two work: n 7→ n and n 7→ (−1)n+1 n.

Case 2: f (0) 6= 0.
Plug in (m, n) = (f −1 (k), 0) to get f k (0) = kf (0) for all integers k. In particular, when k = −1
we have f −1 (0) = −f (0). Now substitute in (m, n) = (m, −f (0)) to get, for all integers m,

f f (m−f (0)) (−mf (0)) = 0. (1)

Now note that the orbit . . . → f −2 (0) → f −1 (0) → 0 → f (0) → f (f (0)) → . . . contains all
multiples of f (0), so it is unbounded and not periodic. Hence from

f f (n)−n−f (0) (0) = f f (n) f −n−f (0) (0)

= f f ((n+f (0))−f (0)) (−n − f (0)) · f (0)

=0

where the second equation follows from f k (0) = kf (0) and the third equation follows from (1), we
have f (n) − n − f (0) = 0 for all integers n. Hence the function f must be of the form n 7→ n + c
for some constant c, and it’s easy to see that all such functions work. 
Comments. There are several possible ways to proceed in Case 2. For example, another way is to
plug in m = 0, f (0) and 2f (0).

5
Problem 3.
Let A, B, C be three distinct points on a line `. Prove that for each pair of distinct points B1 , C1
←−−→ ←−→ ←−→
such that B1 C1 does not pass through A, and B1 C is not parallel to C1 B, there is a unique point
A1 satisfying:
←−−→
(i) A1 does not lie on B1 C1 ,
←−−→ ←−→ ←−−→
(ii) the projections of A onto B1 C1 , of B onto C1 A1 , and of C onto A1 B1 lie on a line not parallel
to `, and
←−−→ ←−→ ←−−→
(iii) the reflections of A over B1 C1 , of B over C1 A1 , and of C over A1 B1 lie on a line not parallel
to `.

Proposed by TacH
←−−→ ←−→
Solution. Let A⊥ , B⊥ , C⊥ be the projections of A onto B1 C1 , of B onto C1 A1 , and of C onto
←−−→ ←−−→ ←−→ ←−−→
A1 B1 , and let A0 , B 0 , C 0 be the reflections of A over B1 C1 , of B over C1 A1 , and of C over A1 B1
respectively.
Since the lines in (ii) and (iii) are not parallel to `, there exist a spiral similarity λ sending
←−−→
A → A0 , B → B 0 , and C → C 0 . Let S be the center of λ, let `0 = B1 C1 , and let T = ` ∩ `0 .
Note that there is also a spiral similarity χ with center S that sends A → B and A⊥ → B⊥ . Let
the tangent to (ABS) at B cut C1 B⊥ at P . Since ∠A⊥ AB = ∠B⊥ BP , 4AA⊥ T ∼ 4BB⊥ P .
Therefore χ sends T → P . Hence, from B = AT ∩ BP , (BT P S) is cyclic.
Now, S is also the center of a spiral similarity ψ that sends T → A⊥ and P → B⊥ . Since
C1 = T A⊥ ∩ P B⊥ , (ST P C1 ) is cyclic.
Combining this with the above, we have now shown that S must be on (BT C1 ). Similarly, S
must be on (CT B1 ) as well, hence S is unique, and is the intersection of (BT C1 ) and (CT B1 ).
Finally, from this unique S, we can construct A1 as the intersection of B⊥ C1 and C⊥ B1 . Clearly
this A1 works, and it must be unique as well. 
←−−→
Comments. 1. The conditions in the problem are needed; if A is on B1 C1 then `1 , `2 found in
←−→ ←−→
the above solution will coincide with `, and if B1 C is parallel to C1 B, the locus becomes a
line since B⊥ , C1 , C⊥ , B1 lie on the same line.

2. If we remove the not parallel to ` condition, there may be four additional points A formed,
which can be constructed as follows. First, draw a line `0 through A⊥ parallel to `. Then,
draw the circle ΓB with diameter BC1 possibly cutting ` at X1 , X2 , and the circle ΓC with
diameter CB1 possibly cutting ` at Y1 , Y2 . Then the additional points are the points formed
by the intersection of C1 Xi and B1 Yj .

6
Problem 4.
Let P ∈ Z[x] be a nonconstant polynomial without integral roots. Prove that there is a positive
integer m 6 3 · deg P such that P (m) does not divide P (m + 1).
Proposed by ThE-dArK-lOrD and tastymath75025
Solution. Let u be the least positive integer such that P (u) 6= ±P (1). Since P (x) = ±P (1) can
only have 2n roots, 2 6 u 6 2n + 1.
Using finite differences or Lagrange Interpolation, we have
n  
n−i+1 n
X
P (u + n) = (−1) P (u − 1 + i).
i
i=0

If P (u) | P (u + 1) | · · · | P (u + n), the above equation reduces to

P (u + n) ≡ ±P (u − 1) ≡ ±P (1) (mod P (u))

hence P (u) - P (u + n) which is a contradiction. Therefore, P (m) - P (m + 1) for some m ∈

{u, u + 1, . . . , u + n − 1}. 
Comments. 1. For the record, we’d like to note that a more general (and unsolved) version of
this problem was actually posted in the High School Olympiad forums by @ThE-dArK-lOrD:
Given a positive integer n. Determine the largest positive integer m such that there exists a
polynomial P ∈ Z[x] with degree n such that P (i) | P (i + 1) for all positive integer i < m.
The post has been deleted, and since there were no comments, exposure was very low. @tasty-
math75025 has provided a bound n + 2 6 m 6 3n privately, from which we have created this
problem. Note that @tastymath75025 is not a member of InfinityDots.

2. The lower bound n + 2 6 m in the above comment can be shown by taking P (x) = 2x − 1
when n = 1, and taking
n!
P (x) = (x − 1)(x − 2) · · · (x − n) +
n−1
otherwise.
When n = 2k is even, we can also show that n + 3 6 m by considering a polynomial
Q(x) of degree k satisfying Q((t + 1/2)2 ) = ±1 for all t = 0, 1, . . . , k, then take P (x) =
cQ((x − k − 3/2)2 ) where c is a big enough constant to make P ∈ Z[x].

3. One way to get a better lower/higher bound on the general problem is to find the exact
maximum value of m0 where there exists a P ∈ Z[x] such that |P (1)| = |P (2)| = · · · =
|P (m0 )|. It will immediately follow that m0 + 1 6 m 6 m0 + n.

7
Problem 5.
Let c1 , c2 , . . . , ck be integers. Consider sequences {an } of integers satisfying

an = c1 an−1 + c2 an−2 + · · · + ck an−k

for all n > k + 1. Prove that there is a choice of initial terms a1 , a2 , . . . , ak not all zero satisfying:
there is an integer b such that p divides ap − b for all primes p.
Proposed by talkon
Solution (1: Algebraic). If c1 , c2 , . . . , ck are all zero, choose ai = k! for all i = 1, 2, . . . , k. The
recurrence relation gives us an = 0 for all n > k + 1. Not hard to see that this gives p | ap =⇒
ap ≡ 0 (mod p) for all primes p.
Now, suppose c1 , c2 , . . . , ck are not all zero. Let z1 , z2 , . . . , zk be the roots (counting multiplic-
ities) of the characteristic equation λk − ki=1 ci λk−i = 0. The key part is choosing
P

an = z1n + z2n + · · · + zkn

for all n.
By Vieta, we get that the elementary symmetric polynomials
X
ej = zi1 zi2 . . . zij = (−1)j+1 cj
1≤i1 <i2 <...<ij ≤k

is an integer for all j = 1, 2, . . . , k. By the Fundamental Theorem of Symmetric Polynomials, we

get that P (z1 , z2 , . . . , zk ) is an integer for all symmetric polynomials P ∈ Z[x1 , x2 , . . . , xk ]. In
particular, an is an integer for all n ∈ Z+ .
To prove that a1 , a2 , ..., ak are not all zero, note that if ai = 0 for all i = 1, 2, . . . , k, we can
use Newton’s identities to prove by induction on i that σi = 0 for all i = 1, 2, . . . , k. This implies
ci = 0 for all i = 1, 2, . . . , k, which is a contradiction.
Now, for each prime p, we’ve
k
cp1 = (z1 + z2 + ... + zk )p = zip + pT (z1 , z2 , ..., zk ) = ap + pT (z1 , z2 , ..., zk )
X

i=1

for some symmetric polynomial T ∈ Z[x1 , x2 , ..., xk ]. So, T (z1 , z2 , ..., zk ) is an integer. This gives
ap ≡ cp1 ≡ c1 (mod p), so we are done by choosing b = c1 . 
Solution (2: Combinatorial). If ci are all zero, then just choose ai like in the first solution. Else,
consider necklaces of size n that is

• marked, that is, we mark the positions on it as 1, 2, . . . , n, and

• created from parts of size 1, 2, . . . , k.

We assign a value to each necklace defined as the product of ci ’s where i’s are the size of the parts
counting multiplicity.

8
 Now define an as the sum of values of all such necklaces of size n. Clearly an is an integer for
all n, and if j = min{i | ci 6= 0} then aj = jcj is clearly nonzero. Next we’ll show that {an } satisfy
the recurrence relation.
Let’s look at each such necklace. The position 1 of the necklace is the rth position (counting
clockwise) of a part of size i, where 1 6 r 6 i 6 k. Call (r, i) the type of that necklace.
By looking at the size of part containing position n + 1 − r (that is the part immediately
counterclockwise from the part containing position 1), we can see that the sum of values of necklaces
of size n with type (r, i), denoted by an,(r,i) , satisfies

an,(r,i) = c1 an−1,(r,i) + c2 an−2,(r,i) + · · · + ck an−k,(r,i)

when n > k. Summing the above over all types (r, i) implies {an } satisfies the recurrence relation.
Finally, we can see that actually, when n = p is prime, we can remove the marks from the
necklace allowing it to be rotated into p positions. Except when all parts have size 1, where rotation
doesn’t change the necklace, these p positions are all counted as distinct in ap , but have the same
value. Since the necklace where all parts have size 1 has value cp1 , it follows that for all prime p,

sp ≡ cp1 ≡ c1 (mod p)

hence we can choose b = c1 . 

Comments. The sequences {an } in the two solutions actually turn out to be exactly the same
sequence. One way to show this is by noting that the equation about an,(r,i) is still true for all
n 6 k, (taking an,(r,i) = 0 for all n < i) except when n = i where the LHS is ci and the RHS is 0.
This actually implies that
n−1
X
an = ncn + cj an−j
j=1

for all n ∈ Z+ (taking cn = 0 for all n > k), which is exactly the same as Newton’s identities
n−1
X
pn = (−1)n+1 nen + (−1)j+1 ej pn−j
j=1

Pk n
where pn is the power sum symmetric polynomial i=1 zi . Therefore, by induction, an = pn for
all n, so the two solutions result in the same sequence an .

9
Problem 6.
Ana has an n × n lattice grid of points, and Banana has some positive integers a1 , a2 , . . . , ak which
sum to exactly n2 . Banana challenges Ana to partition the n2 points in the lattice grid into sets
S1 , S2 , . . . , Sk so that for all i ∈ {1, 2, . . . , k},

(i) |Si | = ai , and

(ii) the set Si has an axis of symmetry.

Prove that Ana can always fulfill Banana’s challenge.

Note: a line ` is said to be an axis of symmetry of a set S if the reflection of S over ` is precisely
S itself. Proposed by talkon
Solution. Let’s start with some terminology and notations. For collections C, D of positive integers,
and sets S of lattice points,

• C + D means the collection created by adding C and D. That is, an element e that appears
c times in C and d times in D will appear c + d times in C + D.

• C ⊂ D means every element e ∈ C appears at least as much in D as in C.

• if C ⊂ D, then D − C is the collection created by subtracting C from D. That is, an element

e that appears c times in C and d times in D will appear d − c times in D − C.

• s(C) is the sum of elements of C counting multiplicity.

• EC , UC , PC are subcollections with EC + UC + PC = C where EC (‘even’) contains only even

numbers, UC (‘unpaired odds’) contains each odd number at most once, and PC (‘paired
odds’) contains each odd number an even times. Note that EC , UC , PC are uniquely defined
for each C.

• PC /2 is the collection that is exactly half of PC . In formal terms, PC /2 is the collection that
PC /2 + PC /2 = PC

• We call (S, C) colorable iff |S| = s(C) and it’s possible to partition S into k sets S1 , . . . , Sk
so that (i) for each j, |Sj | = cj , and (ii) the set Sj has an axis of symmetry `j .

We’ll need the following easy lemma:

Lemma 1. If S is a set of lattice points with an axis of symmetry `S so that t points of S lies on `S ,
and C is a collection of positive integers summing to |S| with at most t odd numbers, then (S, C)
is colorable. Furthermore, there exists a (proper) coloring where all the axes `j coincide with `S .
Proof. Divide S into t isolated points on `S and (|S| − t)/2 pairs of points symmetric w.r.t. `S .
For each even cj , choose some cj /2 pairs, while for each odd cj , choose an isolated point and some
(cj − 1)/2 pairs. 

10
We will now prove the following statement by induction on n, which clearly implies the problem:

Statement: Let Tn denote the n × n lattice grid {1, 2, . . . , n}2 and A = {a1 , a2 , . . . , ak } be a
collection of positive integers summing to n2 . Then, (Tn , A) is colorable even with the following
added condition:

(iii) for any ak ∈ (EA + UA ), the axis `k is the line x = y.

Proof. The base case n = 1 is obvious. Suppose that the above statement is true for all n 6 m − 1.
We’ll show that it’s true for n = m as well.

I. Adjustment of A
If A contains only even numbers, the problem is trivial by Lemma 1. Else, we can add all
elements of EA to the maximal odd amax to create new collection A0 that contains only odd
elements. If EA 6= ∅, amax + s(EA ) is always unique in A0 (and hence is in UA0 ), hence by
Lemma 1, any proper coloring of A0 can be transformed to that of A as well, and the even
numbers will have x = y as the axis as required. So, from here onward, we’ll consider only
the case when A contains only odd numbers, i.e. EA = ∅. From now on, for brevity, we will
omit the subscript A in UA , PA .
For any two possible sets U, U 0 that |U | = |U 0 | and s(U ) = s(U 0 ), since the axis of all Si s
that ai ∈ U is always the same and by Lemma 1, it’s possible to recolor all points when U is
changed to U 0 . Hence we can assume that U is of the form

{1, 3, 5, . . . , 2u − 3, 2u − 1} ∪ W

where u is the maximum possible that makes W an empty set or {2u + 2γ − 1} for some
γ > 1. Note that any set V that s(V ) > |V |2 can be transform to that form.
Now, let pmax be the largest element in P/2. If pmax > 2u + 3, we can change the pair
pmax , pmax from P to 2u+1, 2pmax −(2u+1) making them both new distinct unique members,
and so belong to set U 0 of new collection. We’ve the following cases

• If W is an empty set, s(U 0 ) = u2 + 2pmax ≥ u2 + 4u + 6, |U 0 | = u + 2.

• If W = {2u + 2γ − 1}, s(U 0 ) = u2 + (2u + 2γ − 1) + 2pmax ≥ u2 + 6u + 9, |U 0 | = u + 3.

In both cases, we get s(U 0 ) ≥ |U 0 |2 , hence we can transform it to the above form. In this
new collection, the new maximal odd in P/2 is at most pmax , and the new 2u0 − 1 of U 0 is
more than the old 2u − 1, so this reduces the value pmax − (2u − 1). Hence, we can make an
additional assumption that pmax 6 2u + 1.
So, now our situation reduces to when A is in this form:

1 3 ··· 2u − 1 2u + 1 2u + 3 2u + 5 · · ·
U / / ··· / ∅ ∅ or only 2u + 2γ − 1
P/2 maybe ∅

11
II. We list some possible ways to reduce to a smaller case for induction
Lemma 2. If there exists an odd integer w, and collections R, X of odd integers such that

(a) R ⊂ P/2 and 2R + X ⊂ A,

(b) |R| 6 m − 1 and |X| = w, and
(c) 2s(R) + s(X) > w(2m − w) > 2s(R) + 2w − 1,

then the Statement is true for (Tn , A).

Proof. Consider the above figure (note that squares in the figure represent points), and use
this procedure:

Step 1. Use the lemma to put a copy of R on the yellow points, using the horizontal black line
as the axis.
Step 2. Use the lemma to put the other copy of R on the mirroring green points, using the
vertical black line as the axis.
Step 3. To each odd in X, assign a point each on the red diagonal.
Step 4. Assign leftover pairs of points (with each pair symmetric w.r.t the line x = y) to a
random odd in X. Repeat this until all points in the figure (points with x > m − w or
y > m − w) are assigned.

After the procedure, all odds in R are used up, and all odds in X will have been reduced to
evens. Hence any unused member of U is still in the new U for n = m − w. Now apply the
inductive hypothesis and we’re done. 
Corollary 3. (w = 1 case of Lemma 2.) If there exists o ∈ A and a collection R that

(a) R ⊂ P/2 and 2R + {o} ⊂ A,

(b) s(R) 6 m − 1, and
(c) 2s(R) + o > 2m − 1,

then the Statement is true for (Tn , A).

Lemma 4. If there is a collection {p, p, o1 , o2 } of odd numbers which is a subcollection of A
and satisfies 2p + o1 + o2 > 4m − 4 then the Statement is true for (Tn , A).

12
 Proof. Put the two copies of p on the border on the (1, m) and (m, 1) corners then assign
(1, 1), (m, m) to o1 and o2 respectively. Use pairs left from o1 , o2 to fill the whole border. Any
leftover are even and covered by the inductive hypothesis. 

III. We consider a certain maximal subcollection of P/2

Consider a subcollection Q ⊂ P/2 that maximises s(Q) < m. Let t = m − s(Q). We make
the following observations:

• If there’s an element abig > 2t − 1 in A − 2Q, we can use Corollary 3 with (o, R) =
(abig , Q). Else, all elements in U are at most 2t − 3, which gives

2t − 3 > 2u(+2γ) − 1 =⇒ t > u(+1) + 1 = |U | + 1.

• If there’s an element asmall < t in P/2 − Q then Q + {asmall } should be Q instead, a

contradiction. Hence, all elements in P/2 − Q are at least t.

IV. We finish the problem.

First, we divide the possible value of t into two cases.
Case 1: t > m/2. Since t > m − t, if P/2 contains any element a that t < a < m, the set
{a} must be Q instead, a contradiction. Hence all elements in P/2 − Q are at least m. We,
then, have the following cases:

• If P/2 is empty set then it follows that there are at most m odd numbers in A = U then
Lemma 1 finish the problem.
• If P/2 contains at least two numbers at least m then we can use Lemma 4.
• Else P/2 − Q contain at most one element which must be less than 2t − 1, this gives

s(P/2) 6 (m − t) + (2t − 3) = m + t − 3 6 2m − 4 =⇒ s(P ) 6 4m − 8,

so s(U ) > m2 − 4m + 8. If U contains no element that is at least 2m − 1, we get

s(U ) 6 1 + 3 + 5 + ... + (2m − 3) = (m − 1)2 . But s(U ) ≡ m2 (mod 2), this means |U |
can’t be m − 1, and so s(U ) 6 (m − 1)2 − (2m − 5) = m2 − 4m + 4 (in case that 2m − 3
is equal to 2u + 2γ − 1), contradiction with s(U ) > m2 − 4m + 8. Hence U contains an
element a that is at least 2m − 1. We, then, use Corollary 3 with (o, R) = (a, ∅).

Case 2: t 6 m/2. We divide into further subcases based on |U | and pmax . Recall from I.
that we have pmax 6 2u + 1 6 2|U | + 1.
When pmax 6 5 or when (|U |, pmax , m) = (3, 7, 8), it is not hard to check that in these cases
we either have enough 5’s to use Lemma 2 with w = 5 and R, X containing only 5’s, or can
use Corollary 3.

13
 In all other cases, we can use Lemma 2 as follows:

Case w X R
|U | odd, pmax 6 (|U | − 1)2 + 1 |U | U
|U | even, pmax 6 (|U | − 2)2 + 1 |U | − 1 U − {1}
(|U |, pmax ) = (3, 7), m > 9 5 U + {7, 7}
(|U |, pmax ) = (4, 7) 5 U + {7, 7} − {1}
see below
(|U |, pmax ) = (4, 9) 5 U + {9, 9} − {1}
(|U |, pmax ) = (5, 11) 7 U + {11, 11}
(|U |, pmax ) = (6, 11) 7 U + {11, 11} − {1}
(|U |, pmax ) = (6, 13) 7 U + {13, 13} − {1}
To create R, we randomly choose an element from F = P/2 − Q − (P/2 ∩ X) (named F for
’finally the end of the solution is near’) to put in R until s(R) > w(2m−w)−s(X)
2 then show
that it’s in the range in condition (c) of Lemma 2. The constraints we need to verify is that
s(x) − (2w − 1)
max F 6 +1
2
and
s(F ) + s(X) > w(2m − w).

One can check that for the cases in the table above, these equations hold, using the facts that
max F 6 pmax , m > 2t > 2|U | + 2, and s(F ) + s(X) > m2 + (−1) − 2s(Q), with (−1) only
in cases that we removed {1} from U . 

Comments. 1. The following, due to @DVDthe1st, is a simple solution for odd n:

Roughly speaking, we can WLOG all ai odd (since we can always pair even terms with any odd
one, then extract it out later via symmetry). Since 1 + 3 + ... + (2n − 1) = n2 , there are at most
n distinct values. Out of all the distinct values, pick out one ai from those with an odd number of
such values (call this set S) and pair the remaining. Now, note the n cells along the middle row.
For each ai in set S, we split it by putting one cell on the middle row, and then we aim to place
ai −1
2 in both the top half and the bottom half. If |S| < n, we add in pairs of equal ai until the
middle row is used up entirely. The remaining pairs we allocate one each to the top half and the
bottom half. So it becomes the problem of splitting the top half and bottom half into a new set of
ai , and we can somewhat induct this.
Upon further thoughts, my solution for P6 can be extended to all odd n. Essentially, we need
to assume that the extra ai that were used on the middle row are maximal, so if we use 1’s then
the remaining pairs are all 1’s and the inducive step is trivial. Otherwise, we realise that the split
ai parts don’t have to be symmetric within the top half, so they become essentially ”spares”, which
we can use to eliminate one row if we so wish.
2. The above idea can be generalized to prove that the result is true when n × n is changed to
m × n for m > n and n odd. It’s also possible that a full solution to the problem can be found from
this idea.

14
Results
In the end, a total of 12 users submitted solutions. Here are the scores, with usernames removed
for anonymity. A score of − indicates that the user had not submitted a solution for that problem.

Rank P1 P2 P3 P4 P5 P6 Σ
1 7 7 7 7 7 − 35
2 7 7 2 7 7 1 31
3 7 7 − 7 6 3 30
4 0 7 7 7 6 1 28
5 − 7 7 7 5 − 26
6 − 7 − 7 6 − 20
7 − 7 − 3 7 − 17
8 5 7 − − − − 12
9 − − − 7 − − 7
9 − 7 − − − − 7
11 0 − − − 6 − 6
12 0 − − − − − 0
Σ 26 63 23 52 50 5 219
Avg. 2.17 5.25 1.92 4.33 4.17 0.42 18.25

Final Remarks
InfinityDots MO 2 has concluded. Thanks to everyone who has been a part of the contest.
Problem proposers: TacH, talkon, tastymath75025, ThE-dArK-lOrD.
PSC Members: ArseneLupin, Awien, Continuum, Diamondhead, Gems98, HimeZ, Mr-T19, TacH,
talkon, ThE-dArK-lOrD.
Graders and Outreach: TacH, talkon, ThE-dArK-lOrD.
Test-solvers: CantonMathGuy, v Enhance.
Sign-ups: ABCDE, ac dc1969, adhikariprajitraj, algebra star1234, Allen4567, anantmudgal09, AnArtist,
Ankoganit, Assmit, atmchallenge, ayan.nmath, B.J.W.T, bobthesmartypants, CinarArslan, den thewhitelion,
don2001, Electron Madnesss, enhanced, fas123456, fastlikearabbit, funstar007, futurestar, green dog 7983,
Hamel, hansu, hua1729, hydrogenhelium, Kayak, kenricksfollower, Ld minh4354, little-fermat, lminsl,
lucasxia01, MarkBcc168, Mate007, Math-Ninja, math90, MATHEMATICS1729, Matir, MEGAKNIGHT,
mickeydomath, MNJ2357, monsterDJ, navi 09220114, nikolapavlovic, Omeredip, Omsai, p square,
pieater314159, pro 4 ever, quangminhltv99, Qwertyphysics, rafayaashary1, randomusername, rmtf1111,
rterte, sa2001, Samuel, SAUDITYA, shinichiman, SHREYAS333, smy2012, Supercali, Superguy,
tenplusten, TheDarkPrince, TheMathsBoy, toto1234567890, Tumon2001, valikk202, william122,
Wizard 32, and all private sign-ups.

This report is compiled by talkon.

15