NICE Journal

NICE Committee

Second Edition
Preface
Hello everyone!

This is the second NICE Journal, and we’re glad that you’ve decided to read us. Below is
a short list of notes that may enhance your reading experience.

Chapter Authors and Descriptions

You may be interested in knowing the authors of the journal entries or some information
about them. Here they are:

Entry Name Author Description

Basic algebraic manipu-
lations covering simple
Algebraic Manipulation Elijah Liu
ideas like variables and
SFFT.
How to count only a spe-
Roots of Unity Filter Raymond Feng & Dylan Yu cific group of coefficients
and when to use the filter.
Some rich configurations
arising from the foot of al-
Incenter Miquel Kazi Aryan Amin
titude in a contact trian-
gle.
Number-theoretic func-
Multiplicative Functions Paul Hamrick tions satisfying f (mn) =
f (m) f (n), e.g. φ and τ.
The factorization of x3 +
Titu’s Favorite Factoring Trick Arul Kolla y3 + z3 − 3xyz and its
uses.

Acknowledgements
We would like to thank everyone involved in this journal:

• Kazi Aryan Amin • Paul Hamrick • Dylan Yu

• Sanjana Das • Arul Kolla
• Raymond Feng • Elijah Liu
2
NICE Committee (Second Edition) NICE Journal

Final words before you start reading

Again, we’re very excited that you’re taking your time to read this body of work. Let us
know if you have any suggestions, and have fun learning math!

Sincerely,
Dylan Yu

3
Contents
1 Algebraic Manipulation 6
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2 Variable Assignment and Flexibility . . . . . . . . . . . . . . . . . . . . . . 7
1.3 Simon’s Favourite Factoring Trick . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Various Useful Identities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 Power Laddering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.6 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.6.1 Easier problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.6.2 Harder problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2 Roots of Unity Filter 12

2.1 Lecture notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.1.1 Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.1.2 Why is ROUF useful? . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.1.3 Themes of ROUF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3 Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.4 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

3 Incenter Miquel 19
3.1 Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2 Spiral Similarity at R . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.3 The Mixtillinear Touchpoint . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.4 More Interesting Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.5 Involving the A-Antipode . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.6 A very interesting concyclicity . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.7 Some Nice Contest Examples . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.7.1 A problem related to IMO 2019/6 . . . . . . . . . . . . . . . . . . . 29
3.7.2 A beautiful Isogonal Mittenpunkt problem . . . . . . . . . . . . . . 31
3.8 Changing an incenter problem to an orthocenter one . . . . . . . . . . . . . 32
3.9 An induced bicentric quadrilateral and its properties . . . . . . . . . . . . 36
3.10 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

4 Multiplicative Functions 42
4.1 Dictionary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.2 An instructive example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.3 Dirichlet Convolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.4 Operations with convolutions . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4
NICE Committee (Second Edition) NICE Journal

5 Titu’s Favorite Factoring Trick 50

5.1 “Theory” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.2 A Potpourri of Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

A Key Parts 62
A.1 List of Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
A.2 List of Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
A.3 Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

B Selected Solutions 67
B.1 Solution 4 (Euclid 2021/2.c) . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
B.2 Solution 5 (Euclid 2014/3) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
B.3 Solution 6 (blackpenredpen) . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
B.4 Solution 7 (AoPS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
B.5 Solution 8 (Instagram) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
B.6 Solution 12 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
B.7 Solution 13 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
B.8 Solution 15 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
B.9 Solution 16 (AMC 12A 2021/15) . . . . . . . . . . . . . . . . . . . . . . . . . 71
B.10 Solution 17 (AIME II 2016/12) . . . . . . . . . . . . . . . . . . . . . . . . . . 71
B.11 Solution 65 (AMC 12A 2021/25) . . . . . . . . . . . . . . . . . . . . . . . . . 72
B.12 Solution 68 (PUMaC Number Theory 2019/A2) . . . . . . . . . . . . . . . 73
B.13 Solution 69 (PUMaC Number Theory 2016/A5) . . . . . . . . . . . . . . . 73
B.14 Solution 70 (PUMaC Number Theory 2015/A6) . . . . . . . . . . . . . . . 74
B.15 Solution 73 (ISL 1989/11) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
B.16 Solution 74 (HMMT February Algebra-NT 2019/8) . . . . . . . . . . . . . . 75
B.17 Solution 80 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
B.18 Solution 81 (AHSME 1999/30) . . . . . . . . . . . . . . . . . . . . . . . . . . 75
B.19 Solution 82 (MR S499, Titu Andreescu) . . . . . . . . . . . . . . . . . . . . . 76

5
1 Algebraic Manipulation
You might want to review factoring and Vieta’s formulas.

â1.1 Introduction
Let’s start with two relatively simple techniques:

Fact 1 (Completing the Square). Start with arbitrary quadratic ax2 + bx + c = 0. Then,
b 2 b2 − 4ac
 
one can rearrange that to a x + − = 0.
2a 4a

Proof. Observe the following manipulations:

ax2 + bx + c =0
b c
 
2
a x + x+ =0
a a

b b 2 b 2 c

2
a x + x+ 2 − 2 + =0
a 4a 4a a
2 2
!
b b c

a x+ − 2+ =0
2a 4a a
2
b b2

a x+ +c− =0
2a 4a
2 2
b b − 4ac

a x+ − =0
2a 4a

as shown.

It is best not to memorize the result of this, but rather the steps; as completing the square
is a very useful technique.
This result can be extended to our beloved quadratic formula:

Theorem 2 (Quadratic Formula) √

−b ± b2 − 4ac
In an arbitrary quadratic ax2 + bx + c = 0, x = .
2a
6
NICE Committee (Second Edition) NICE Journal

Proof. Continue with the “completing the square” from before. Then:
2
b b2 − 4ac

a x+ − =0
2a 4a
b 2 b2 − 4ac
 
x+ =
2a 4a2
√
b b2 − 4ac
x+ =±
2a 2a
√
−b ± b2 − 4ac
x=
2a

b
Remark 3. If we let b0 = , then we can rewrite the quadratic formula as
2

−b0 ± b0 2 − ac
p
x= .
a
This representation has fewer coefficients, so it’s a reasonable idea to keep this in the back of
your head.

Now, on to the actual content.

â1.2 Variable Assignment and Flexibility

The great thing about variables is that it is very easy to assign them in problems. Let’s
take a look at this Euclid problem as an example:

Example 4 (Euclid 2021/2.c)

Qing is twice as old as Rayna. Qing is 4 years younger than Paolo. The average age of
Paolo, Qing, and Rayna is 13. Determine their ages.

The idea is to take advantage of the structure of the problem (all the phrases can be
represented as equations and the ages as unknowns). This isn’t really regarded as a
technique, but it works wonders – when you’re stuck, try assigning variables randomly
and delving further.
A corollary of this is that variable substitutions are very easy to perform, and if you
can do one, it is advisable to do so (if it’s beneficial, of course).
I’m also a huge fan of using Vieta’s formulas backwards when given the chance. This
is best illustrated in the following example:

7
NICE Journal NICE Committee (Second Edition)

Example 5 (Euclid 2014/3)

Determine all ordered pairs ( a, b) that satisfy the following system of equations:

a + b = 16
1 1 4
+ =
a b 7

Walkthrough. Very simple, but also very practical.

1. Find ab.

2. Construct the polynomial x2 − ( a + b) x + ab = 0 via reverse Vieta’s and solve.

Let’s also take a look at this comical example:

Example 6 (blackpenredpen)
√
Solve the equation 5 − x = 5 − x2 .

Walkthrough. Let’s look at the restrictions for the problem.

√
1. Show that | x | ≤ 5. This will help check for extraneous roots.

2. Here comes the trick: square both sides and treat 5 as the variable. You’ll end up
with two quadratics.

3. Solve the quadratics and check the bounds.

â1.3 Simon’s Favourite Factoring Trick

Also known as “completing the rectangle.” Let’s start with the canonical example:

Example 7 (AoPS)
Find all pairs of positive integers (m, n) that satisfy mn + 3m − 8n = 59.

Walkthrough. If you’ve never seen this before, you should just read the solution.

1. Factor to get m(n + 3) − 8n = 59.

2. Find a way to turn the “−8n” part to −8(n + 3) and get (m − 8)(n + 3) = 35.

3. Note m, n are positive integers: do casework to finish.

The main reason why this is a useful trick is because m and n have to both be integers.
If they had to be rational numbers, or even real numbers, it would be much harder or
downright impossible to get precise values for m and n.
8
NICE Committee (Second Edition) NICE Journal

â1.4 Various Useful Identities

Below is a list of useful algebraic identities, in no particular order. Proofs are not included
as they will bloat the handout (although it’s nice to explore them for yourselves!)

• SFFT: xy + jx + ky = a =⇒ ( x + k)(y + j) = a + jk

• a2 ± 2ab + b2 = ( a ± b)2

• a2 − b2 = ( a + b)( a − b)

• a3 ± b3 = ( a ± b) a2 ∓ ab + b2

• TFFT: a3 + b3 + c3 − 3abc = ( a + b + c) a2 + b2 + c2 − ab − bc − ca

• Sophie-Germain Identity: a4 + 4b4 = a2 + 2ab + 2b2 a2 − 2ab + 2b2

• x − y is always a factor of x n − yn for nonnegative integer n

• x + y is always a factor of x n + yn for odd positive n

â1.5 Power Laddering

Let’s take this problem that I definitely did not steal from Instagram:

Example 8 (Instagram)
x4 − 6x3 − 2x2 + 18x + 23 p √
Without using a calculator, evaluate when x = 19 − 8 3.
x2 − 8x + 15

Walkthrough. The nested radical looks funny.

p √ √
1. Show that 19 − 8 3 = 4 − 3.
√ √
2. Form a quadratic with 4 − 3 and its conjugate 4 + 3. Use it to show the denomi-
nator of the expression in question is 2.

3. Multiply the quadratic you formed in the above step by x2 and subtract it off the
numerator. Note that this does not affect your answer because you are subtracting
0.

4. Multiply the quadratic by 2x and do the same thing. Then again by multiplying the
quadratic with some constant. You’ll be left with a constant in the numerator and
you can finish from there.
9
NICE Journal NICE Committee (Second Edition)

â1.6 Problems
â1.6.1 Easier problems
Problem 1 (Fermat 2019/24). Consider the quadratic equation x2 − (r + 7) x + r + 87 =
0 where r is a real number. This equation has two distinct real solutions x which are
both negative exactly when p < r < q for some real numbers p and q. Find the value of
p2 + q2 .

Problem 2 (Own). Determine the sum of all possible x values such that x is a positive
integer and xy = 317 + 2x + 3y for some integer y.
372
Problem 3 (Own). Let p, q, and r be the roots of x3 − 16x2 + 84x − 112 = 0. Find
p3 + q3 + r3 − 3pqr.
64
3
Problem 4 (Own). If x +
x
= 1, find x5 − x2 . 18

Problem 5. Express φn in terms of φ and the Fibonacci numbers, where n is a positive

integer and φ is the larger solution to x2 − x − 1 = 0.
√5an+1/√5[1-√5/2]ⁿ
Problem 6 (Jeffrey Qin). Let x, y, and z be real numbers such that 2x2 + 4y2 + 9z2 +
a
4xy + 2x + 6z = −2. The sum of all possible values of | x + y + z| is , a fraction to the
b
simplest form. Compute a + b.

Problem 7 (AIME 1987/14). Compute the following (without a calculator, of course):

(104 + 324)(224 + 324)(344 + 324)(464 + 324)(584 + 324)

(44 + 324)(164 + 324)(284 + 324)(404 + 324)(524 + 324)

â1.6.2 Harder problems

373
These problems are a step up from the previous ones, so I recommend you gather a
couple friends (and maybe some pizza) and tackle them together!

10
NICE Committee (Second Edition) NICE Journal

Problem 8 (Purple Comet HS 2019/19). Find the remainder when ∏33

 4 3 2
n=3 2n − 25n + 33n


is divided by 2019.

Problem 9 (Purple Comet HS 2018/22). Positive integers a and b satsify a3 + 32b +

2c = 2018 and b3 + 32a + 2c = 1115. Find a2 + b2 + c2 .

Problem 10 (NICE Spring 2021/13, David Altizio). Suppose x and y are nonzero real
numbers satisfying the system of equations

3x2 + y2 = 13x,
x2 + 3y2 = 14y.

Find x + y.

11
2 Roots of Unity Filter
â2.1 Lecture notes
â2.1.1 Statement
A fun story: roots of unity filter (ROUF) was said to be invented by Nikolai Nikolov
when he solved IMO 1995/6 in contest via roots of unity. He was awarded a special prize
(which is given to people with beautiful solutions) for this discovery.

Theorem 9 (Roots of Unity Filter)

2πi
For integers m and n, if we define ω = e n , then the following holds:

n −1 
(
k
m 0 n-m
∑ ω = n n|m.
k =0

Proof. Two proofs.

Solution via symmetric sums It suffices to show the result for 0 ≤ m < n, since we may
take the exponent mod n without changing the value of the sum. The result is obvious for
m = 0, so discard that case. Now, as 1, ω, ω 2 , . . . , ω n−1 are roots of the polynomial x n − 1,
this means that all symmetric sums of the nth roots of unity, except for their product, are
0. However, note that the given sum is of degree less than n, so it can be written in terms
the symmetric sums without using the product. But all these symmetric sums are 0, so
the entire sum evaluates to 0, as desired.

Solution via induction Strong induct on n. The result is obvious for n = 1 and n a prime
(why?), then split the nth roots of unity into equally spaced groups of size gcd(nn,m) and
apply the inductive hypothesis to finish.

â2.1.2 Why is ROUF useful?

The reason: if we have some function1 P( x ), then as the name suggests, by taking
P(1) + P(ω ) + P(ω 2 ) + · · · + P(ω n−1 ), the theorem essentially allows us to filter out all
coefficients except for those which are a multiple of n. This is why ROUF is usually
written in the following form:
1 Usually a rational function, i.e. a polynomial divided by a polynomial.
12
NICE Committee (Second Edition) NICE Journal

Corollary 10 (Alternate Form of ROUF)

2πi
Let n be a positive integer, and let ω = e n . For any polynomial P( x ) = a0 + a1 x +
2
a2 x + . . . (with finitely many terms), then

1 
a0 + an + a2n + . . . = P (1 ) + P ( ω ) + P ( ω 2 ) + . . . + P ( ω n −1 ) .
n

Proof. Left as an exercise; just use the first form of ROUF.

2πi
Remark 11 (Which root of unity?). The nth root of unity we use doesn’t have to be ω = e n ; it can
be ω k for any k relatively prime to n, but I’ve never seen a question where ω won’t work when
ω k will.

Here is a prototypical example:

Example 12
Evaluate
10 10 10 10 10 10
           
+ + + + + .
0 2 4 6 8 10

Walkthrough.

1. What should we have as the value of n here? What is our P( x )? (Hint: use binomial
theorem!)
(1+1)10 +(1−1)10
2. Show that the desired sum is equivalent to 2 , and conclude.

The above example uses the 2nd roots of unity, but −1 and 1 aren’t enough to justify the
use of complex numbers. We’ll see a better use of roots of unity in action below.

Example 13
Evaluate
10 10 10
     
+ + .
2 5 8

Walkthrough.

1. What should we have as the value of n here?

2. Why doesn’t the P( x ) in our previous problem work? There is an additional reason
besides “the common differences are different.”

3. How can we edit P( x ) so that the coefficients of the terms with degree a multiple of
n are precisely (10 10 10
2 ), ( 5 ), ( 8 )? (Hint: consider dividing by a power of x.)
13
NICE Journal NICE Committee (Second Edition)

2πi
4. Prove that if ω = e 3 , then the desired sum is equivalent to

1 · (1 + 1)10 + ω (1 + ω )10 + ω 2 (1 + ω 2 )10

.
3
Of course, both of the above examples were small enough that they could just be bashed
out, but the walkthroughs just serve to illustrate the process of applying the roots of unity
filter; the answer extraction is not really the interesting part.

Exercise 14 (Generating function for binomial ROUF). Show that

N 1 n
 
∑ m + kn xm+kn = n ∑ (ω k )−m (1 + ω k · x) N ,
k ≥0 k =1

where ω is an nth root of unity.

Moral
The real power of ROUF comes when evaluating how many ways for something to
happen such that some parameter is a multiple of some number n. These problems
can be tackled using a generating function and then applying ROUF.

For example, the examples above could have instead been rephrased as:
• Reworded from example 12: How many ways are there to choose a subset with
even size out of 10 distinct objects?
• Reworded from example 13: How many ways are there to choose a subset of k
objects out of a set of 10 distinct objects such that k ≡ 2 (mod 3)?
Although the computations may become very messy in the following problem, can you
see the idea behind the solution for it?
Illustrate an idea How many ways are there to choose a subset of k objects out of a
set of 2021 distinct objects such that k ≡ 2 (mod 3) or k ≡ 1 (mod 4)?

It’s just two-set PIE: add the ways for each modulo, then subtract the overlap.

â2.1.3 Themes of ROUF

ROUF is technically an algebraic technique, but its applications to C/G/N are what
actually make it useful. In particular, the idea is often to express a non-algebraic problem
via the subject’s relation to the roots of unity, then use ROUF to do the computation.
Common themes of ROUF: for geometry, it can appear when we’re dealing with a
regular polygon (because the roots of unity are a regular polygon on the unit circle). For
combinatorics and number theory, we’ll see it when we’re asked to count something
constrained by a number-theoretic property, e.g. the large amount of IMO 1995/6 copies.
14
NICE Committee (Second Edition) NICE Journal

â2.2 Examples
This next problem is a repeat of the ones before but just for insurance:

Example 15
Compute
1000
 
∑ 3k
.
k ≥0

2πi
Walkthrough. Let f (n) equal 1 when 3 | n and 0 otherwise, and let ω = e 3 be a 3rd root
of unity.

1. Characterize f (n) with ω.

2. Use 1 + ω + ω 2 = 0 to simplify your answer and finish. For example, we can
arrange the expression to get 1 + ω = −ω 2 .
This next example doesn’t require roots of unity filter, but when 6 tenors and 8 basses is
replaced with a and b of them, respectively, the method becomes more helpful.

Example 16 (AMC 12A 2021/15)

A choir director must select a group of singers from among his 6 tenors and 8 basses.
The only requirements are that the difference between the number of tenors and basses
must be a multiple of 4, and the group must have at least one singer. Let N be the
number of different groups that could be selected. What is the remainder when N is
divided by 100?

Walkthrough. Let f ( x, y) = (1 + x )8 (1 + y)6 .

1. Show that coefficient amn of x m yn in the expanded form of f is the number of groups
of m basses and n tenors. Thus, we want to sum up all values of amn for which
4 | m − n, except for a00 = 1.

2. For what value of y (in terms of x) can we change x m yn to x m−n ? Let g( x ) = f ( x, •),
where • is the new value of y.
3. Apply ROUF on g with the 4th roots of unity.

Example 17 (AIME II 2016/12)

The figure below shows a ring made of six small sections which you are to paint on a
wall. You have four paint colors available and you will paint each of the six sections a
solid color. Find the number of ways you can choose to paint the sections if no two
adjacent sections can be painted with the same color.

15
NICE Journal NICE Committee (Second Edition)

Walkthrough. Suppose that the colors are 0, 1, 2, 3. Clearly the difference between the
colors in adjacent sections is 1, 2, or 3 modulo 4. Define the number at each border
between sections to be this difference.

1. Use a generating function to represent each border.

2. What is the generating function for all 6 borders then? In this function, the co-
efficient of x n should represent the total number of colorings where the colors’
numbers are increased by n as we go around the ring.

3. If we go around the ring and the colors have increased by n, what must n be
divisible by to ensure that the color of the section we started with is still the right
color? Let the number n is divisible by be m.
4. Apply ROUF on A( x ) with the mth roots of unity to finish.

See the solutions for other interesting ways of solving the problem, one using linear
algebra, and another using chromatic polynomials.

â2.3 Further Reading

1. Things Fourier, Evan Chen
2. Dirichlet’s theorem on APs, Evan Chen
3. Cauchy Integral Formula, Altheman

4. Generating Functions in CP, zscoder

5. More Examples, Kevin Sun
6. Jacobsthal numbers, Wikipedia

â2.4 Problems
Silly example:

16
NICE Committee (Second Edition) NICE Journal

Problem 11 (AMC 10B 2021/16). Call a positive integer an uphill integer if every digit is
strictly greater than the previous digit. For example, 1357, 89, and 5 are all uphill integers,
but 32, 1240, and 466 are not. How many uphill integers are divisible by 15?

Problem 12 (BMT Analysis 2015/7). Evaluate ∑37 k 75

k =0 (−1) (2k ).

Problem 13 (AMC 12A 2017/25). The vertices V of a centrally symmetric hexagon in

the complex plane are given by
√ √ 1 1 1 1

V= 2i, − 2i, √ (1 + i ), √ (−1 + i ), √ (1 − i ), √ (−1 − i ) .
8 8 8 8
For each j, 1 ≤ j ≤ 12, an element z j is chosen from V at random, independently of the
other choices. Let P = ∏12 j=1 z j be the product of the 12 numbers selected. What is the
probability that P = −1?

Problem 14. Three regular 7-sided dice, two regular 5-sided dice, and one regular 4-
sided die are rolled. The probability that the 6 dice sum to a number divisible by 3 can be
expressed as m n , where m and n are relatively prime positive integers. Find m.

Problem 15 (IMC 1999/8). We roll a regular die n times. What is the probability that
the sum of the numbers shown is a multiple of 5?

Problem 16 (AIME I 2018/12). For every subset T of U = {1, 2, 3, . . . , 18}, let s( T ) be the
sum of the elements of T, with s(∅) defined to be 0. If T is chosen at random among all
subsets of U, the probability that s( T ) is divisible by 3 is m
n , where m and n are relatively
prime positive integers. Find m.

Problem 17 (HMMT February Team 2015/7). Let f : [0, 1] → C be a nonconstant

complex-valued function on the real interval [0, 1]. Prove that there exists ε > 0 (possibly
depending on f ) such that for any polynomial P with complex coefficients, there exists a
complex number z with |z| ≤ 1 such that | f (|z|) − P(z)| ≥ ε.

Problem 18 (HMMT February Combinatorics 2012/10). Jacob starts with some com-
plex number x0 other than 0 or 1. He repeatedly flips a fair coin. If the nth flip lands
heads, he lets xn = 1 − xn−1 , and if it lands tails he gets xn = x 1 . Over all possible
n −1
choices of x0 ; what are all possible values of the probability that x2012 = x0 ?

17
NICE Journal NICE Committee (Second Edition)

Problem 19 (Titu). For positive integers n, define

n −1
kπ
 
2n
f (n) = ∑ cos n
.
k =0

Compute
∞
f (k)
∑ k
.
k =2 k · 2

Problem 20 (PUMaC Live 2018/4.3). Let 0 ≤ a, b, c, d ≤ 10. For how many ordered
quadruples ( a, b, c, d) is ad − bc a multiple of 11?

Problem 21 (HMIC 2021/3). Let A be a set of n ≥ 2 positive integers, and let√f ( x ) =

∑ a∈ A x a . Prove that there exists a complex number z with |z| = 1 and | f (z)| = n − 2.

Problem 22 (Putnam 1974/B6). Let S be a set with 1000 elements. Find a, b, c, the
number of subsets R of S such that | R| ≡ 0, 1, 2 (mod 3), respectively. Find a, b, c if
|S| = 1001 instead.

Problem 23 (MOP 1999). There are n points on a unit circle such that the product of the
distances from any point on the unit circle to the given points is at most 2. Prove that the
given n points must be vertices of a regular n-gon.

18
3 Incenter Miquel
Basic knowledge of inversion and projective geometry is assumed.

â3.1 Notation
We will introduce some notation that we will use for the rest of the article. Although we
would eventually be forced to use some of the point names for different purposes, we
hope the purpose is clear from the context.

Notation
• We denote the circle passing through the n concyclic points A1 , . . . , An as
( A1 · · · An ), the circumcircle. If there are only two points ( AB), then it denotes
the circle with diameter AB.
• For some 4 ABC, let M A denote the midpoint of arc BC not containing A
in ( ABC ). Furthermore, let MBC denote the midpoint of arc BAC in ( ABC ).
Similarly define MB , MC , MCA , M AB .
• Let M denote the midpoint of side BC in 4 ABC.

• Let O denote the circumcenter of 4 ABC.

• Let I denote the incenter of 4 ABC.
• Let 4 DEF be the contact triangle of 4 ABC (with the convention D ∈ BC, E ∈
AC, and F ∈ AB).

• Let A0 denote the A-antipode in ( ABC ), and let D 0 denote the D-antipode in
( DEF ).
• Let R = ( AI ) ∩ ( ABC ) (the intersection of the two circles) distinct from A.
• Let S ∈ EF (S lies on the line EF) satisfy DS ⊥ EF (DS and EF are perpendicu-
lar).
• Let P = DS ∩ ( DEF ) distinct from D, and Q = AP ∩ ( DEF ) distinct from P.
• Let T = AP ∩ ( ABC ) distinct from A.

19
NICE Journal NICE Committee (Second Edition)

• Let G = AR ∩ BC.

• Let J ∈ EF satisfy MJ ⊥ EF. Furthermore, let M0 be the midpoint of EF, and

X1 = EF ∩ BC. Finally,let Z = AR ∩ EF.
• Let the A-mixtillinear incircle touch AB and AC at K and L.

The diagram featured on the title page conveys all of these points.

â3.2 Spiral Similarity at R

MBC
A

R
P 0 J E
F S M

I
B
D M

MA

Consider the spiral similarity taking EF to BC. Note the similarity is centered at R by
using the Miquel point of EFBC. Now, this implies it sends ( AFIE) to ( ABC ), so as I is
the midpoint of arc EF, I goes to M A . Similarly, A goes to MBC .
Now, we can easily angle chase to get

1
∠ PFE = ∠ PDE = ∠SDE = 90◦ − ∠SED = 90◦ − ∠ FED = ∠ B = ∠ IBC
2
and similarly with ∠ PEF gives that 4 PFE ∼ 4 IBC. In particular, if the spiral similarity
f centered at R maps F 7→ B and E 7→ C, then

f ( P) = I f (S) = D f ( I ) = MA
20
NICE Committee (Second Edition) NICE Journal

However, we know that SD k I M A (both are perpendicular to EF). Since R is the spiral
center taking SI to DM A , hence it is also the center of a spiral similarity g which takes
SD to I M A .So g is a homothety for the two mentioned lines. Inverting about the incircle,
we note that we switch R and S, so R, S, I are collinear. Thus, we have the following
fact:

Lemma 18
In the said configuration, R, D, M A are collinear.

Proof. From our above discussion on g being a homothety and R, S, I being collinear,
their images are also collinear, so R, D, M A are collinear.
In addition, we have the following claim:

Lemma 19
I, MBC , J are all collinear.

Proof. We will use phantom points. Define I MBC ∩ EF = J⊥ . We note that

MMBC AM0 I J⊥
= =
MM A M0 I J⊥ MBC

where the first relation follows by the first spiral similarity mentioned and the second
one follows by using Thales’ theorem on EF k AMBC . Thus, we get that MJ⊥ k AI, but
as AI ⊥ EF, we have MJ⊥ ⊥ EF. Thus, J = J⊥ .
The reader might want to take a look at the following exercises:

Exercise 20 (USAJMO 2014/6). Let ABC be a triangle with incenter I, incircle γ and circumcircle
Γ. Let M, N, P be the midpoints of sides BC, CA, AB and let E, F be the tangency points of γ
with CA and AB, respectively. Let U, V be the intersections of line EF with line MN and line
MP, respectively, and let X be the midpoint of arc BAC of Γ.
(a) Prove that I lies on ray CV.
(b) Prove that line XI bisects UV.

The above problem is incinerated by the Iran lemma and our mentioned claim.

Exercise 21 (EMMO Seniors 2016). In 4 ABC, let point D be the tangency point of incircle ω with
side BC. Analogously define E and F for sides AC and AB respectively. Let P be the feet of
_
the perpendicular from D to EF. Let N be the midpoint of arc ABC of circumcircle Γ of 4 ABC.
Prove that the lines AP and ND concur on Γ.

21
NICE Journal NICE Committee (Second Edition)

Exercise 22 (CMIMC Tiebreaker 2018/G3). Let ABC be a triangle with incircle ω and incenter I.
The circle ω is tangent to BC, CA, and AB at D, E, and F respectively. Point P is the foot of
the angle bisector from A to BC, and point Q is the foot of the altitude from D to EF. Suppose
AI = 7, IP = 5, and DQ = 4. Compute the radius of ω.

â3.3 The Mixtillinear Touchpoint

MBC
A

R
D0
P J E
F
L
K
I
G
B
D M
T

MA

T wasn’t named by accident: let’s look at the following claim:

Lemma 23
T is the A-mixtillinear touchpoint.

Proof. Note that P and D 0 are reflections of each other in AI, as PD 0 k EF. This means that
AP and AD 0 are isogonal, so as AD 0 is the A-nagel cevian. This implies the mixtillinear
touchpoint lies on AP, so T is indeed the said touchpoint.

Remark 24. This means that T, I, J, MBC are collinear.

Lemma 25
Recall that G := AR ∩ BC. Then

G ≡ RA ∩ BC ∩ KL ∩ TM A

22
NICE Committee (Second Edition) NICE Journal

Proof. By using radical axes on ( AI ), ( BIC ), ( ABC ), we see that AR, KL, BC all concur.
As AR ∩ BC = G, we have that G ∈ KL. Similarly, using radical axis on ( I M A ), ( BIC ),
and ( ABC ), we see that BC, KL, TM A concur. As G = BC ∩ KL, we have that G ∈ TM A ,
which completes the proof.

Remark 26. Note that ( GRIDT ) is cyclic.

â3.4 More Interesting Properties

MBC
A

R
P Y
J E
S
Z X F
L
X1 K
I
G
B
D M

MA

Now, let’s talk about X1 . We have the following concurrencies and concyclities:

Lemma 27
X1 , R, MBC are collinear and X1 , R, S, D are concyclic.

Proof. Notice that (SX1 D ) is the S-Apollonian circle of 4SBC, so taking a spiral similarity
of R gives
RB FB BS
= =
RC EC SC
which implies that R is also on the S-Appolonius incircle. Thus, X1 , R, S, D are concyclic.
In addition, DX1 is the diameter, so we have DR ⊥ RX1 , As DR = DM A ⊥ RMBC , we
see that X1 , R, MBC are collinear.
23
NICE Journal NICE Committee (Second Edition)

Now, let’s turn to Z. We have the following claim:

Lemma 28
4 RDX1 , 4 RIG, and 4 RSZ are similar.

Proof. Notice that ] RIG = ] RDG = ] RDX1 and from our above discussion, ]X1 RD =
]GRI = π2 . Thus the first two triangles are similar. Furthermore, SZ = EF k IG, which
means that the last two triangles are similar.
We have the following claim:

Lemma 29
GI is tangent to ( X1 GR).

Proof. Note that ] IGR = ]GX1 R.

We proceed with our next claim:

Lemma 30
Let f is the spiral similarity at R discussed before. Then

f ( Z ) = X1 .

Proof. To see this, just note that Z ∈ EF. Thus, as f ( EF ) = BC, we see that f ( Z ) ∈ BC.
However, we also have that ]X1 RZ = ] IRD, so as ] IRD is the angle of the spiral
similarity, f ( Z ) = X1 .
Let’s see another tangent line:

Lemma 31
BC is tangent to ( RZX1 ).

Proof. Note that ]( EF, BC ) = ]( X1 R, RZ ), which immediately implies the tangency at

X1 = EF ∩ BC.
To wrap this section up, we complete it with this final claim:

Lemma 32
If K1 be the midpoint of DX1 , then ZK1 ⊥ OI.

Proof. Note that ZR · ZA = ZE · ZF imples Z lies on the radical axis of ( ABC ) and ( DEF ).
Also by midpoint of Harmonic Bundle Lemma, we have

K1 D2 = K1 B · K1 C =⇒ K1 lies on the radical axis of ( ABC ) and ( DEF )

which implies the perpendicularity.

24
NICE Committee (Second Edition) NICE Journal

â3.5 Involving the A-Antipode

MBC
A

R
P Y
0 J E
S M
Z X F
Z1 L
K
I
G
B
D M

A0
MA

This configuration is heavily based on ELMO SL 2019/G3. Let ZA0 ∩ ( ABC ) = Z1 .

The main claim is the following:

Lemma 33
S lies on the radical axis of ( AZZ1 ) and ( A0 EF ).

Proof. Note that Z1 ∈ ( A0 EF ), ZRM0 I is cyclic, ZZ1 M0 A is cyclic, and S is the orthocenter
of 4 AZI. Thus, ZS · SM0 = IS · SR = SE · SF, because the reflection of S over M0 lies on
( AI ) and the inversion about the incircle sends S to R and A to M0 , so PoP on ( AI ) yields
the desired.

Exercise 34 (ELMO SL 2019/G3). Let 4 ABC be an acute triangle with incenter I and circumcenter
O. The incircle touches sides BC, CA, and AB at D, E, and F respectively, and A0 is the reflection
of A over O. The circumcircles of ABC and A0 EF meet at G, and the circumcircles of AMG and
A0 EF meet at a point H 6= G, where M is the midpoint of EF. Prove that if GH and EF meet at
T, then DT ⊥ EF.

25
NICE Journal NICE Committee (Second Edition)

Exercise 35 (ELMO 2010/6). Let ABC be a triangle with circumcircle ω, incenter I, and A-excenter
I A . Let the incircle and the A-excircle hit BC at D and E, respectively, and let M be the midpoint
of arc BC without A. Consider the circle tangent to BC at D and arc BAC at T. If TI intersects ω
again at S, prove that SI A and ME meet on ω.

â3.6 A very interesting concyclicity

J0

MBC
A

R
P Y
M0 J E
S
Z X F

X1 I
G Q
B
D M
T

MA

We present the following beautiful lemma:

Lemma 36
If EF ∩ ( ABC ) = { X, Y }, then we have ( XYQDM ) is concyclic.

Proof. We first prove that M ∈ ( DXY ). Note that RMBC DM is cyclic by the shooting
lemma. Note that
X1 D · X1 M = X1 R · X1 MBC = X1 X · X1 Y
Showing that Q ∈ ( DXY ) is a bit more trickier:

26
NICE Committee (Second Edition) NICE Journal

Claim —
Q ∈ ( DXY )

We will introduce two very special points which are very closely related to the standard
orthocenter configuration, namely Queue and Ex Points. Surprisingly, the will aid a lot of
our proofs that follow in this article. We wil now define them.

Remark 37 (The A-Queue point of 4 ABC). If we let B0 , C 0 to be the feet of altitudes from B, C
onto CA, AB, with BB0 ∩ CC 0 ≡ H, then the A-Queue point of 4 ABC, Q A is defined as
Q A ≡ ( AB0 HC”) ∩ ( ABC ).

Remark 38 (The A-Ex point of 4 ABC). If we let B0 , C 0 to be the feet of altitudes from B, C onto
CA, AB, with BB0 ∩ CC 0 ≡ H, then the A-Ex point of 4 ABC, Q A is defined as X A ≡ B0 C 0 ∩ BC.

√
Remark 39. Q A and X A are inverses of each other in the circle ( A, AB0 · AC )

If the reader is interested, they should have a look at The Ex-points and the Queue-
points Part One and Part Two for more interesting properties.
Approach 1; Using Orthic Axis. Note that −1 = (SQ : EF ) =⇒ Q is the D-Queue point
of triangle DEF. Let DQ ∩ EF = XD .
Then it is well known that XD is the D-Ex point in 4 DEF. So XD lies on the orthic
axis of 4 DEF, which is the radical axis of ( DEF ) and the circumcircle of its tangential
triangle, that is ( ABC ). Thus we have

XD X · XD Y = XD E · XD F = XD Q · XD D

This implies Q ∈ ( DXY ).

Remark 40. Radical axes on ( ARSM0 ), ( AI ),and ( DQSM0 ) give that Z = XD (!) So Z is the
D-Ex point of 4 DEF

Approach 2; Coaxility Lemma. We use phantom points. Let ( DXY ) ∩ ( DEF ) = Q0 6= D.

First note that since OI is the Euler Line of 4 DEF, (say by an incircle inversion).This
implies that ( ABC ), ( DEF ),and the nine point circle of 4 DEF are coaxal. Thus, by the
forgotten lemma of coaxality, we have

Pow( DEF) M0 M0 F · M0 E SF · SE Pow( DEF) S

= = =
Pow( ABC) M0 M X·MY
0 0 SX · SY Pow( ABC) S

Now, we also have

M0 F · M0 E Pow( DEF) M0
=
M0 X · M0 Y Pow( DQ0 XY ) M0
27
NICE Journal NICE Committee (Second Edition)

and similarly for S, implying that ( DSM0 ), ( DQ0 XY ), and ( DEF ) are coaxal. Thus,
Q0 ∈ ( DSM0 ) ≡ ( DM0 ). We present two ways to finish from the second approach:

First Finish. Note that Q0 becomes the D-Queue point of 4 DEF. So Q = Q0 (!)
Second Finish. We want AD 0 and AQ0 are isogonal. In other words, if K = AMBC ∩ M0 Q0 ,
then we need −1 = (KM0 ; Q0 D 0 ). This is again equivalent to proving that ( M0 K ) is
orthogonal to the incircle.
To do this note that A ∈ ( M0 K ), so since we have I M0 · I A = r2 (where r is the radius
of the incircle), we are done!

Having done this, lets take a look at some contest problems destroyed by the lemma. We
will suitably change the notation in the problems presented to suit the notation used so
far in the article.

Example 41 (Taiwan TST 2019/2/2)

Prove that if MJ hits ( DXY ) at J 0 , then J 0 ∈ AT.

Solution. Angle chase:

] MJ 0 Q = ] MDQ = ] BDQ = ] DPQ.

Example 42 (MR O451)

Recall that J = I MBC ∩ EF. Prove that the radical axis of ( AI J ) and ( DXY ) passes
through T.

Solution. Note that we have ]QIT = ] BDQ = ]QJ 0 J. This implies QJ 0 J I is cyclic. So T
lies on the radical axis of ( AI J ) and ( DXY ).

Example 43
Prove that the line I MBC is the perpendicular bisector of QD.

Solution. This follows from the fact that TA and TD are isogonal in ] BTC and IQ =
ID.
These exercises will be left for the reader.

28
NICE Committee (Second Edition) NICE Journal

Exercise 44. Prove that AS,BC, I MBC concur

Exercise 45. Prove that AS, DMBC concur on ( ABC ).

Exercise 46. Prove that the midpoint of MB MC is the center of ( XQDMY ).

From now on, we will call AS ∩ BC = U and AS ∩ ( ABC ) ∩ V 6= A.

Exercise 47. Prove that RBVC is a harmonic quadrilateral.

â3.7 Some Nice Contest Examples

â3.7.1 A problem related to IMO 2019/6
Example 48
Let ABC be a triangle with AC > AB. Let N = DI ∩ AMBC . D1 is the BC-extouch
point. Suppose that U 0 ≡ ( AMBC D1 ) ∩ BC. Show that U 0 , M0 , and N are collinear.

Solution. Consider the following figure:

O0 R A
M 0 MBC
F P N
S E
B I
Q K1
U
D
T L

D1

MA
29
NICE Journal NICE Committee (Second Edition)

In the notation, you can probably guess the first claim:

Claim —
U ≡ U0

Proof. Note that

]U MBC A = ]TMBC A = ]TCA = π − ] B − ]CAT

and
]UD1 A = ] BD1 A = π − ] B − ] BAD1
Now, we note AD1 , AT are isogonal in ] BAC, so ]UD1 A = ]U MBC A, so indeed
U 0 ≡ U.
All poles and polars are taken with respect to ( DEF ). Now note that, by La Hire, U, M0 N
are collinear if and only if their polars are concurrent.
It is easy to note that the polar of M0 is AMBC . By a previous example, we have that
the polar of U is DQ. Also the polar of N is just the line through M’ perpendicular to DI.
(Remark that by La hire M0 lies on the polar of N).
Let K = DQ ∩ AMBC . Note that KAM0 Q is cyclic with diameter (KM0 ).We need to
prove that KM0 ⊥ DI. Note that M0 Q and IT are parallel, so we have

∠ AKM0 = ∠ AQM0 = ∠ ATI = ∠ AM A MBC

So if KM0 cuts M A MBC at some point K1 , then KAK1 M A is cyclic, so KM0 is perpendicular
M A MBC which gives KM0 ⊥ DI.
Now we present some easy facts for the reader to ponder on. Let O0 denote the midpoint
of KM0

Exercise 49. Prove that O0 , Q, U are collinear

Exercise 50. Prove that ( M0 K ) is orthogonal to ( DEF )

The point N is famous as the IMO 2019/6 concurrency point. We present an auxiliary
lemma which is helpful in solving the IMO problem itself.

Lemma 51
If L = AI ∩ BC, then ( ARQDLN ) is a cyclic hexagon

Proof. Note that N ADL is cyclic due to obvious reasons. By Shooting Lemma,

M A L · M A A = M A D · M A R =⇒ ARDLN is cyclic

To finish, note that ] AQD = ] PQD = ] RDB = ] ALD.

30
NICE Committee (Second Edition) NICE Journal

â3.7.2 A beautiful Isogonal Mittenpunkt problem

Example 52 (Fake USAMO 2020/3)
Let ( AVE) ∩ ( DEF ) = { E, Y1 } and ( AVF ) ∩ ( DEF ) = { F, Z1 }. Prove that FY1 , EZ1 , OI
are concurrent.

Solution. We present a motivated and rather “easy-to-come-up-with" solution for this

very beautiful and difficult problem. However, given the artillery we’ve developed so far,
this is going to be a walk in a park!

MBC

A
R
P E

F S
W1
Y1
Z Z1 I

Q
U O
B C
D

MA

The presence of so many circles motivates looking at radical axes. Indeed,note that by
31
NICE Journal NICE Committee (Second Edition)

Radical axes on ( AVE), ( AVF ) and ( DEF ), we have that FZ1 , EY1 and AV concur at a
point, say W.
Since FZ1 Y1 E seems to be our chief quadrilateral, we will introduce W1 = FY1 ∩ EZ1
and Z 0 = Y1 Z1 ∩ EF. The reader might have already guessed what’s going on. Since
EFZ1 Y1 is a cyclic quadrlateral, we have IW1 ⊥ Z 0 W. So we need to prove that OI ⊥ Z 0 W.
However note that because we have WY1 · WE = WV · WA, this implies that W lies on
the radical axis of ( ABC ) and ( DEF ).
Hence,we realize that Z 0 must also lie on this radical axis. But since Z 0 ∈ EF, so Z 0
must actually be Z. So we now seek to prove that Z 0 = Z. It turns out that proving
Z 0 = Z, or equivalently Z 0 , R, A are collinear, is just a matter of cross ratio chasing. All
poles and polars are taken with respect to the incircle.
Remark that we have, by definition,

R
−1 = ( A, I; E, F ) = ( Z, S; E, F )

By La Hire, we also know that A lies on the polar of Z 0 . So the polar of Z 0 is just AV. So S
lies on the polar of Z 0 . Hence,we may write

−1 = ( Z 0 , S; E, F )

completing the problem.

We now leave the following exercise to the reader, which justifies the title of this prob-
lem.

Exercise 53. Prove that W1 is the Isogonal Mittenpunkt Point ( X57 ) of 4 ABC.

â3.8 Changing an incenter problem to an orthocenter one

It turns out that since Q is the D-queue point in 4 DEF, so it might be useful to rephrase
the problem in terms of the the contact triangle. In some rather nice cases (like the
problem about to be presented), it turns out that we can completely convert the problem
into an orthocenter configuration, which might be easier to handle. So now we present a
really beautiful problem.
We will stick to the original wording of the problem this time:

Example 54 (CAMO 2020/3)

Let ABC be a triangle with incircle ω, and let ω touch BC, CA, AB at D, E, F, re-
spectively. Point M is the midpoint of EF, and T is the point on ω such that DT is a
diameter. Line MT meets the line through A parallel to BC at P and ω again at Q. Lines
DF and DE intersect line AP at X and Y respectively. Prove that the circumcircles of
4 APQ and 4 DXY are tangent.

32
NICE Committee (Second Edition) NICE Journal

Solution. Let T 0 be the reflection of T in AI. Let AT ∩ ( DEF ) = G. It is obvious that G, Q

are reflections of each other about AI.

X A K P Y

T
T0
M E

F
I
G

B C
D D0

TA

MA

We claim that G is the desired concurrency point We will prove a series of claims:

Claim — EFXY is cyclic with center A.

Proof. Note that we have

] FED = ] BDF = ] FXY
33
NICE Journal NICE Committee (Second Edition)

This implies that EFXY is cyclic. Next, note that we also have

B
] AXD = ] FDB = π −
2

We also have

B
] AFX = ]XFA = ] BFD = π −
2

Similarly, we can prove that AX = AE = AF = AY. The claim is hence proved.

Claim — T is the orthocenter of 4 DXY

Proof. Note that we have ]XED = ]TED = π

2 This implies that X,T,E are collinear.Similarly
Y,T,F are collinear.

Claim — G ∈ ( DXY )

Proof. Let D ∗ denote the D-antipode in ( DXY ). We have ] DGT = ] DGA = π2 . But
since A, T, D ∗ are collinear, we have that ] DGD ∗ = π2 . Thus, G ∈ ( DXY ). Hence, the
claim is proven.

Now note that G is the D-Queue point in 4 DXY. So we completely rephrase the problem
in terms of 4 DXY as follows.

Restated in terms of 4 DXY Let 4 ABC be a triangle with orthic triangle as 4 DEF.
Let G and Q be the D-Queue points of 4 ABC and 4 AEF. Let I denote the midpoint
of AH. Let QH ∩ BC = P. Prove that ( MPQ) and ( ABC ) are tangent to each other
at G, where M is the midpoint of BC.

34
NICE Committee (Second Edition) NICE Journal

A
Q

E
G
N

F
H L

K B P D M C

HA

Let M and N denote the midpoints of BC and EF respectively. Note that H, N and Q are
collinear.
First note that since AQDP and AGDM are cyclic quadrilaterals, hence, we can use a
simple Power of Point trick.
We have
PH · HQ = AH · HD = HG · HM
so MPQG is cyclic. Remark that we had proved earlier that G, Q are reflections of each
other in MI. In particular, we have EF k GQ. Let the tangent to G at ( ABC ) hit BC at K.
Let H A be the reflection of H in BC.
Since we have −1 = ( GH A ; BC ), hence KH A is tangent to ( ABC ) at H A . Further, since
( BHC ) and ( ABC ) are reflections of each other about BC, hence KH is tangent to( ABC ).
Hence KH is anti-parallel to BC and hence parallel to EF.
Now to finish off, we will proceed by radical axes. We want that KH A is tangent to
( MPH A ). Thus, we we basically need ]KH A P = ] BMH A .
35
NICE Journal NICE Committee (Second Edition)

We will just do an angle chase. Note that G is the Miquel point of BFEC and the spiral
similarity centered at G that sends FE 7→ BC also sends N 7→ M.
So we have
]KH A P = ]KHP = ] FNH = ]GQH = ]GQN = ] NGQ = ]GNF = ]GMB
]GMB = ] HMK = ]KMH A = ] BMH A
so by radical axes on ( ABC ), ( PMH A ), ( MPQ), we win!

â3.9 An induced bicentric quadrilateral and its properties

We start off this section by introducing some new points.
Define Y = AC ∩ BS and X = AB ∩ CS. We investigate various properties of the points
X and Y.

Z2
A
Y
R

E
X P
F S
Z1

G B D C

MA

Lemma 55
XYBC has an incircle.

Proof. Note that we have

−1 = (SY ∩ ( I ), SB ∩ ( I ); F, D )
Projecting this through S onto ( I ) =⇒ PY is tangent to ( I ). Similarly, we can show for
X and conclude, XY is tangent to ( I ) at P.

36
NICE Committee (Second Edition) NICE Journal

Lemma 56
XYBC has a circumcircle.

Proof.

∠ BXY = 180◦ − ∠ FIP = 180◦ − 2∠ FDS = 2∠ DFE = ∠ DIE = 180◦ − ∠YCB

So yes, the title of the section does make sense. XYCB is our very own induced bicentric
quadrilateral.

Lemma 57
RAYX is a cyclic quadrilateral.

Proof using phantom points. Redefine: ( AXY ) ∩ ( ABC ) = R0 and XY ∩ BC = G 0 . By

Radical Axes, R0 ∈ AG 0 . Notice R0 is the Miquel Point of BXYC =⇒ R0 , S, I are collinear
and R0 SI ⊥ AG 0 . This is enough to imply that R0 = R and G 0 = G.

Now we have a look at a few examples:

Example 58 (HMMT November Team 2016/10)

Let R1 denote the point on ( AXY ) such that AR1 ⊥ BC. Prove that S, I, R1 are collinear.

Solution. Remark that since XY antiparallel to BC =⇒ A-antipode in ( AXY ) is R1 . The

rest is a trivial angle chase. Details are left to the reader.

Example 59 (SORY 2019/6, Aditya Khurmi)

Prove that if the tangents from M A to ( DEF ) meet ( ABC ) at Z1 , Z2 , then Z1 , X, Y, Z2
are collinear. Z1 lies closer to B than C.

Solution. We will use a small fact that inversion with center A and radius AI swaps the
A-mixtilinear incircle and the incircle of 4 ABC. We will also use another small claim,
which is indeed trivial. Yet we will state it.

Claim (Rephrased Fact 5) — In any triangle 4 XYZ with K as the midpoint of arc YZ
not containing X in ( XYZ ), the incenter I of the triangle is the intersection of XK and
(KYZ ) which lies closer to X.

This is just a trivial application of Fact 5.

We now show that, if ( A, AI ) meets the circumcircle of 4 ABC at Z10 , Z20 , then Z10 Z20
is tangent to the incircle at S. This combined with the “Claim" implies the conclusion,
as then Z10 Z20 is just YX and MZ10 , MZ20 are tangents to the incircle. This is true as A is
37
NICE Journal NICE Committee (Second Edition)

midpoint of arc Z10 Z20 As I is the incenter of 4 MZ10 Z20 , therefore 4 MZ10 Z20 and 4 ABC
would share the same incircle.
Perform an inversion with center A and radius AI. It swaps the A-mixtilinear incircle
with the incircle. So, P and T are swapped. Now the conclusion trivially follows as
( ABC ) (image of Z10 Z20 ) is tangent to the A-mixtilinear incircle (image of the incircle) at T
(image of S). The problem is thus solved.
The reader might want to have a look at the following exercises:

Exercise 60 (OMO Spring 2020/15). Let ABC be a triangle with AB = 20 and AC = 22. Suppose
its incircle touches BC, CA, and AB at D, E, and F respectively, and P is the foot of the
perpendicular from D to EF. If ∠ BPC = 90◦ , then compute BC2 .

Exercise 61. Prove that ( AXY ) is tangent to the A-mixtilinear incircle at a point on AD.

Exercise 62 (Due to math-pi-rate). Let tangents from M A to ( I ) hit ( I ) at U1 and V1 .Prove that
4 PU1 V1 and 4 DEF share the same nine point circle.

â3.10 Problems
We finally present a bunch of practice problems for the reader. Try invoking the machinery
we have developed throughout the article. The problems are quite nice and we encourage
the reader to try them. Enjoy!
Remark 63. A few of the following problems might require knowledge the “Iran incenter lemma.”
If the reader is unfamiliar with the lemma, then the article here is a fantastic one to learn it.

Problem 24 (CJMO 2019/3). Let I be the incenter of 4 ABC, and M be the midpoint
of BC. Let Ω be the nine-point circle of 4 BIC. Suppose that BC intersects Ω at a point
D 6= M. If Y is the intersection of BC and the A-intouch chord, and X is the projection of
Y onto AM, prove that X lies on Ω, and the intersection of the tangents to Ω at D and X
lies on the A-intouch chord of 4 ABC.
Note. The nine-point circle of 4 ABC is the circumcircle of its medial triangle, and if
the incircle touches AC and AB at E and F, respectively, then EF is the A-intouch chord.

Problem 25 (GGG 2.3, Andrew Wu). Acute, scalene 4 ABC is given with circumcircle
Γ and intouch triangle DEF; D, E, F lie on BC, CA, AB respectively. L is the midpoint of
_ ←→ ←→ ←→
arc BAC. G is the intersection of EF and BC; T lies on EF with DT ⊥ EF. P lies on LG
→ ←
← →
with PT k BC. K lies on Γ such that KT bisects ∠EKF; K and A are not on the same side
←
→ ←→
of EF. If A0 is the antipode of A on Γ, then show that A0 K, the line passing through T

38
NICE Committee (Second Edition) NICE Journal

←
→
and the midpoint of DG, and the perpendicular from P to EF concur.

Problem 26 (GGG 2.6, Andrew Wu). Let ABC be an acute, scalene triangle with incenter
_
I and circumcircle Γ; A0 is the antipode of A on Γ, and L is the midpoint of arc BAC. The
←→ ←→ ←→
incircle of 4 ABC meets BC at D. Suppose that A0 I and BC meet at P, and that AP and
←→ ←→
DI meet at Q; X 6= L is the intersection of LQ and Γ. Show that the circumcircles of
4 XDQ and 4 IDP are orthogonal.

Problem 27 (GGG 3.3, Andrew Wu). Let ABC be a scalene triangle with incenter I
and incircle ω. Let K be a point lying on the circumcircle of triangle ABC such that
∠ AKI = 90◦ , and suppose that ω meets BC, CA, AB at D, E, F, respectively. Suppose that
←→
KD meets ω at points D and X. Show that the line joining the circumcenters of triangles
BFX and CEX is parallel to BC.

Problem 28 (GGG 3.6, Andrew Wu). Let ABC be a scalene triangle with incenter I and
circumcircle Ω; L is the midpoint of arc BAC and A0 is diametrically opposite A on Ω. D
←
→ ←
→
is the foot of the perpendicular from I to BC. LI meets BC and Ω at X and Y, respectively,
←→ ← → ←→ ←→
and LD meets Ω again at Z. XZ meets A0 I at T. DI meets the circle with diameter AI
←
→
again at P. Show that the second intersection between PT and the circle with diameter
AI lies on AY.

Problem 29 (Romania JBMO TST 2019/3). Let ABC a triangle, I the incenter, D the
contact point of the incircle with the side BC and E the foot of the bisector of the angle A.
If M is the midpoint of the arc BC which contains the point A of the circumcircle of the
triangle ABC and { F } = DI ∩ AM, prove that MI passes through the midpoint of EF.

Problem 30 (Iran TST 2008/10). In the triangle ABC, ∠ B is greater than ∠C. T is the
midpoint of the arc BAC from the circumcircle of ABC and I is the incenter of ABC. E is
a point such that ∠ AEI = 90◦ and AE k BC. TE intersects the circumcircle of ABC for
the second time in P. If ∠ B = ∠ IPB, find the angle ∠ A.

Problem 31 (Sharygin CR 2020/21). The diagonals of bicentric quadrilateral ABCD

meet at point L. Given are three segments equal to AL, BL, CL. Restore the quadrilateral
using a compass and a ruler.

39
NICE Journal NICE Committee (Second Edition)

Problem 32 (trumpeter’s PoTD Forum 1/10/16). 4 ABC has its incircle ω tangent to
BC, CA, AB at D, E, F respectively. Let ` be the line through A parallel to BC, and let DE,
DF intersect ` at M, N respectively. Let P be the intersection of NE with MF, and let Γ be
the circumcircle of 4 NPM; finally, let AD intersect Γ at a point X such that X, P are not
on the same side of `. Prove that P, X are antipodal points with respect to Γ.

Problem 33 (tworigami). Let ABC be a triangle with incircle centered at I and tangent
to BC, AC, AB at D, E, F respectively. Let K denote the foot of the altitude from D to EF,
and let Ω denote the circumcircle of 4 BIC. Points P and Q are chosen on AC and AB
respectively so that ∠ AIP = ∠ AIQ = 90◦ . Let the circumcircle of 4KDP intersect Ω at
points B1 and B2 and define points C1 and C2 similarly. Furthermore, let the circle with
diameter DK intersect Ω at points A1 and A2 . Prove that B1 B2 , C1 C2 , A1 A2 concur on
DK.

Problem 34 (buratinogigle). Let ABC be a triangle, incircle touches BC, CA, AB at

D, E, F respectively. Prove that orthocenter of triangle DEF is radical center of circles
( A, AD ), ( B, BE), (C, CF ).

Problem 35 (POGCHAMP 2020/3). Scalene triangle ABC has incenter I, incircle ω, and
circumcircle Ω. Let ω touch BC, CA, AB at D, E, F respectively. Let PA be the point on Ω
such that APA ⊥ PA I, and let Q A be the point on ω such that AI k DQ A . Lines AD and
PA Q A meet at S A . Points PB , PC , Q B , QC , SB , and SC are defined analogously. Prove that

DS A ESB FS 9
+ + C < .
AS A BSB CSC 2

Problem 36 (POGCHAMP 2020/4). Let 4 ABC have incircle ω, and suppose ω touches
sides BC, CA, and AB at D, E, and F, respectively. P is the point on ω such that DP ⊥ EF,
and let OB and OC be the circumcenters of 4 BFP and 4CEP, respectively. Prove that
OB OC = 12 BC.

Problem 37. Given a acute, scalene 4 ABC with incenter I and contact triangle 4 DEF
R ∈ ( DEF ) such that DR ⊥ EF. Let DR ∩ EF = P. Let (Y ) and ( Z ) denote the
circumcircles of ( BPF ) and (CPE).Let U, V denote the points of intersection of EF with
(Y ) and ( Z ) respectively, Finally, let EF intersect BI, CI at K, L respectively. Prove that
FL EV
= .
FU EK

40
NICE Committee (Second Edition) NICE Journal

Furthermore, if AD ∩ ( I ) = X and XF, XE intersects (Y ) and ( Z ) at Y 0 , Z 0 respectively,

then prove that Y 0 Z 0 k EF.

Remark 64. An interested reader should check out On a rich configuration related to tangent circles
in a triangle by Navneel Singhal.

41
4 Multiplicative Functions
Basic knowledge of some number theory and sum and product notations is assumed.

â4.1 Dictionary
To motivate this handout, here are a few common multiplicative functions, some more
trivial than others.

• The identity function id on the naturals maps each positive integer to itself.

• The unit function e on the naturals is defined as

(
1 n=1
e(n) = .
0 n 6= 1

• The constant function 1 mapping each natural to 1.

• The divisor function τ (n) on the naturals counts the number of divisors of a given
positive integer.
e
If the integer n can be written as ∏i pi i for distinct primes pi and exponents ei , then
τ (n) = ∏i (ei + 1) by constructive counting on the individual primes.

• The sum of divisors function σ (n) on the naturals taking each positive integer to the
sum of its positive integral divisors including itself.
e
Similar to the divisor function, if the integer n can be written as ∏i pi i for distinct
e e −1
primes pi and exponents ei , then σ(n) = ∏i ( pi i + pi i + · · · + 1) by considering
all possible exponents of pi in each divisor.

• The Euler phi function φ(n) on the naturals taking each integer n to the number of
integers less than n and relatively prime to it.
e
If the integer n can be written as ∏i pi i for distinct primes pi and exponents ei ≥ 1,
then one can show that
1
 
φ(n) = n · ∏ 1 − .
i
pi

In particular, when n = 1 the product is empty and thus φ(1) = 1.

This result follows from establishing the result for each prime power pk and then
applying the Chinese Remainder Theorem as many times as is necessary.
42
NICE Committee (Second Edition) NICE Journal

• The Möbius function µ(n) on the naturals is defined as follows: if n is 1 then µ(n) =
1, if n = ∏ik=1 pi for distinct primes pi then µ(n) = (−1)k , and if n is divisible by a
particular prime twice, µ(n) = 0.
This function is clearly multiplicative by its definition.

• A positive integer n is square-free if it is divisible by no prime twice, that is, if

µ(n) 6= 0.

What all of these functions have in common is that they are multiplicative: if f is one of
these functions and m, n are relatively prime positive integers, then f (mn) = f (m) f (n).
In fact, the first function is totally multiplicative because the relatively prime part can be
ignored: for all positive integers m, n, we have id(mn) = id(m)id(n). Thus, if we want to
understand a multiplicative function, we only have to look at its values on prime powers
pk .

â4.2 An instructive example

Example 65 (AMC 12A 2021/25)
τ (n)
Let f (n) = √ . Compute the value of N that maximizes f ( N ).
3
n

Walkthrough.

1. Prove f is multiplicative.

2. Look at how f behaves on prime powers pk with p small and determine the maxima.

3. Prove that for large primes p, a prime power pk has maximum f ( pk ) for k = 0.

4. Extract the answer.

The main idea in the above solution was that f being multiplicative gave us the ability to
maximize f for each individual prime and thus limit our scope in a useful way.

Exercise 66 (PUMaC Number Theory 2010/A2). Find the largest positive integer n for which
σ(n) = 28.

â4.3 Dirichlet Convolution

In the following examples, multiplicativity will be used more deeply, so we need a
method to tell if functions are multiplicative. One useful fact is the following:
43
NICE Journal NICE Committee (Second Edition)

Theorem 67 (Multiplicative Convolution)

If two functions f and g are multiplicative, then their Dirichlet convolution
n
( f ∗ g)(n) = ∑ f (d) g
d|n
d

is also multiplicative.

Proof. To show this function is multiplicative, note that for relatively prime m and n we
have  mn  m n
 
( f ∗ g)(n) = ∑ f (d) g = ∑ ∑ f ( d1 d2 ) g ·
d|mn
d d1 | m d2 | n
d1 d2
  
m  n
   
=  ∑ f ( d1 ) g ∑ f ( d2 ) g d2 
d |m
d 1 d |n
1 2

= ( f ∗ g)(m) · ( f ∗ g)(n).

Here are some examples of Dirichlet convolutions so that the definition does not appear
to come out of nowhere:
• ( f ∗ e)(n) = f (n) for all multiplicative f
• τ (n) = (1 ∗ 1)(n)
• σ(n) = (id ∗ 1)(n)
This is enough to solve the following problem.

Example 68 (PUMaC Number Theory 2019/A2)

Let f be a function over the naturals so that

• f (1) = 1,
e e
• If n = p11 p2e2 · · · pkk for distinct primes pi and nonnegative integers ei , then
f (n) = (−1)e1 +···+ek .

Find
2019
∑ ∑ f ( d ).
i =1 d | i

Walkthrough.
1. Prove f is multiplicative.
44
NICE Committee (Second Edition) NICE Journal

2. What function is f ∗ 1 and what is it doing in this problem?

3. Extract the answer.

Here is a slightly harder example based on the same premise (although it avoids citing
Dirichlet convolutions).

Example 69 (PUMaC Number Theory 2016/A5)

gcd(m,n)
Let k = 26 · 35 · 52 · 73 · 53. Let S be the sum of lcm(m,n) over all ordered pairs of positive
integers (m, n) where mn = k. If S can be written in simplest form as rs , compute r + s.

Walkthrough.
gcd(m,n)
1. Write lcm(m,n) in a way more conducive to multiplicativity shenanigans, that is,
breaking the problem into dealing with individual primes.
gcd(m,t/m)2
2. Determine why t is a useful expression.

3. Answer the problem for each prime power factor.

4. Finish.

Here is yet another PUMaC example.

Example 70 (PUMaC Number Theory 2015/A6)

Determine the smallest positive integer n for which

35 | ∑ τ ( t )3 .
t|n

Walkthrough.

1. What is the multiplicative function here?

2. Worry about 5 and 7 dividing ∑t|n τ (t)3 separately, looking at prime powers for n.

3. Extract the minimum.

â4.4 Operations with convolutions

Now, we discuss some more theory concerning the Dirichlet convolution. Since f ∗ e = f
for all multiplicative f , it must be true that e is the identity. Thus, if f ∗ g = e for functions
f and g then f and g are inverses.

• Clearly the convolution is commutative.

45
NICE Journal NICE Committee (Second Edition)

• Interestingly, it is also associative:

( f ∗ ( g ∗ h))(n) = (( f ∗ g) ∗ h)(n) = ∑ f ( d1 ) g ( d2 ) h ( d3 ).
d1 d2 d3 = n

• It is also distributive over addition because f ∗ ( g + h) = f ∗ g + f ∗ h.

Another identity dealing with these functions is
φ ∗ 1 = id,
for which it suffices to check the result on prime powers:

∑ φ ( d ) = p k − p k −1 + p k −1 − p k −2 + · · · + p − 1 + 1 = p k .
d| pk

Remark 71. The idea of characterizing a multiplicative f in terms of its primes shows up heavily
in problems involving NT functions.

Theorem 72 (Inverse of µ)
The Möbius function µ is the inverse of 1.

Proof. It suffices to check µ ∗ 1 = e on prime powers pk . Indeed, k = 0 yields a sum of

1, k = 1 yields a sum of 0, and summands corresponding to divisors p` with ` ≥ 2 are
zero.
As a corollary, f ∗ 1 = g implies
g ∗ µ = f ∗ 1 ∗ µ = f ∗ (1 ∗ µ ) = f ∗ e = f ,
so µ essentially recovers f from g. Moreover, this result does not actually care if f and g
are multiplicative, as exemplified below.

Example 73 (ISL 1989/11)

Define a sequence ( an )n≥1 so that ∑d|n ad = 2n . Show n | an .

Walkthrough.
1. Determine an with Möbius inversion.
2. For a particular prime p | n with p dividing n a total of e times, it is enough to show
pe | n.
3. Take the whole expression modulo pe and recall Euler’s totient theorem, this should
finish.
We conclude this section with a final difficult example.

46
NICE Committee (Second Edition) NICE Journal

Example 74 (HMMT February Algebra-NT 2019/8)

There is a unique function f : N → R such that f (1) > 0 and such that
n
∑ f ( d ) f
d
=1
d|n

for all n ≥ 1. What is f 20182019 ?

Walkthrough.
1. What would be a good assumption on f ?
2. Look at prime powers.

3. The condition naturally rewrites in terms of generating functions.

4. What is the generating function of √1 ?

1− x

5. Extract the answer.

â4.5 Problems
Problem 38 (AMC 12B 2021/7). Let N = 34 · 34 · 63 · 270. What is the ratio of the sum
of the odd divisors of N to the sum of the even divisors of N?

Remark 75. It is complete overkill to use multiplicative properties on the problem above. But
why not?

√
Problem 39. Show that τ (n) ≤ 3n.

Problem 40 (Italy MO 1998/6). Let f : N → N be a completely multiplicative, increas-

ing function. Prove that if f (2) = 2, then f (n) = n for all n ∈ N. Does this remain true if
the word “completely” is omitted?

Problem 41 (NEMO Team 2018/9). Find the sum of all N such that

n = 7τ (n).

47
NICE Journal NICE Committee (Second Edition)

Problem 42 (Elementary Number Theory, David Burton). Let F (n) = ∑d|n ( f (d)).
Prove that f (d) is multiplicative if and only if F (n) is multiplicative.

Problem 43. For any positive integer n, prove that

n 
∑(σ(d)) = ∑ d τ (d) .
d|n d|n

Problem 44 (Baltic Way 2019/16). For a positive integer N, let f ( N ) be the number of
ordered pairs of positive integers ( a, b) such that the number

ab
a+b
is a divisor of N. Prove that f ( N ) is always a perfect square.

a a
Problem 45 (Number Theory, George E. Andrews). Prove that if n = p11 p2a2 p33 ...prar ,
then
2a +1
p2a r +1
! !
p1 1 + 1 r +1
∑ dφ(d) = p1 + 1
·...·
pr + 1
.
d|n

Problem 46 (PAMO SL 2018/A5). Let g : N → N be a function satisfying:

• g( xy) = g( x ) g(y) for all x, y ∈ N,

• g( g( x )) = x for all x ∈ N, and

• g( x ) 6= x for 2 ≤ x ≤ 2018.
Find the minimum possible value of g(2).

Problem 47. Prove that for any positive integer n, the number of divisors of n of the form
4k + 1 is larger than the number of divisors of the form 4k + 3, where k is a nonnegative
integer.

Problem 48 (HMMT February Algebra-NT 2021/5). Let N be the product of the first
10 primes, and let
S = ∑ φ( x ) · y.
xy|n

48
NICE Committee (Second Edition) NICE Journal

Compute Sn .

Problem 49 (ISL 2004/N2). The function f from the set N of positive integers into itself
is defined by the equality
n
f (n) = ∑ gcd(k, n), n ∈ N.
k =1

(a) Prove that f (mn) = f (m) f (n) for every two relatively prime m, n ∈ N.
(b) Prove that for each a ∈ N the equation f ( x ) = ax has a solution.
(c) Find all a ∈ N such that the equation f ( x ) = ax has a unique solution.

Problem 50 (AMM 12003, Nikolai Osipov). Given an odd positive integer n, compute
n
gcd(k, n)
∑  .
k =1 cos2
πk
n

49
5 Titu’s Favorite Factoring Trick
â5.1 “Theory”
This entire article revolves around this single factorization, hereby dubbed “Titu’s Fa-
vorite Factoring Trick” (you’ll see why in the problem set).

Theorem 76 (Titu’s Favorite Factoring Trick)

For all real numbers a, b, c,
 
a3 + b3 + c3 − 3abc = ( a + b + c) a2 + b2 + c2 − ab − bc − ca . (5.1)

Proof. By the sum of cubes formula,

a3 + b3 + c3 − 3abc = ( a + b)3 − 3a2 b − 3ab2 + c3 − 3abc

= ( a + b + c)3 − 3( a + b)2 c − 3( a + b)c2 − 3ab( a + b + c)
= ( a + b + c)3 − 3( a + b)c( a + b + c) − 3ab( a + b + c)
 
= ( a + b + c) a2 + b2 + c2 − ab − bc − ca .

This seemingly innocent factorization actually results in tons of nontrivial corollaries.

For example:

Corollary 77
For all real numbers a, b, c,
1  
a3 + b3 + c3 − 3abc = ( a + b + c ) ( a − b )2 + ( b − c )2 + ( c − a )2 .
2

Proof. Note that

( a − b)2 + (b − c)2 + (c − a)2 = a2 − 2ab + b2 + b2 − 2bc + c2 + c2 − 2ca + a2

= 2( a2 + b2 + c2 − ab − bc − ca),

which is twice the last factor of (5.1).

50
NICE Committee (Second Edition) NICE Journal

Corollary 78
For all real numbers a, b, c,

a3 + b3 + c3 − 3abc = ( a + b + c)( a + bω + cω 2 )( a + bω 2 + cω ),

where ω is a complex number satisfying ω 2 + ω + 1 = 0.

Proof. Just check that

1 
( a + bω + cω 2 )( a + bω 2 + cω ) = ( a − b )2 + ( b − c )2 + ( c − a )2
2
through direct expansion.

Corollary 79
For all real numbers a, b, c,

a + b + c = 0 =⇒ a3 + b3 + c3 = 3abc.

Proof. Since a3 + b3 + c3 − 3abc = ( a + b + c) a2 + b2 + c2 − ab − bc − ca , if a + b + c =

0, then a3 + b3 + c3 − 3abc = 0, or a3 + b3 + c3 = 3abc.

This last corollary is extremely useful, and it shows up in a ton of problems. Let’s see
some examples in action.

Example 80 √
√ 3 √
Reals a, b, c satisfy abc = 1 and 3 a + b + 3 c = 0. Find a + b + c.

Walkthrough. This is one of those problems that seems very hard to do if you don’t know
Titu’s Favorite Factoring Trick, but becomes straightforward after seeing it.
√ √3 √
1. What do we know if 3 a + b + 3 c = 0? (Explicit hint: use the corollary above!)

2. Extract the answer.

Example 81 (AHSME 1999/30)

How many ordered pairs of integers (m, n) satisfy mn ≥ 0 and

m3 + n3 + 99mn = 333 ?

Walkthrough.

1. Rewrite the equation to be in the form a3 + b3 + c3 − 3abc = 0.

51
NICE Journal NICE Committee (Second Edition)

2. One of the three factorizations above will help us weed out integer solutions quickly.
Which one?
3. Extract the answer.

Example 82 (MR S499, Titu Andreescu)

Let a and b be distinct real numbers. Prove that
√ √3
27ab( 3 a + b)3 = 1 if and only if 27ab( a + b + 1) = 1.

Walkthrough. This, along with many other problems in the Mathematical Reflections
Journal proposed by Titu Andreescu, inspired the titular “Titu’s Favorite Factoring
Trick”.

1. The problem tries to hide the cubes by instead using cube roots. Make the substitu-
tion a = x3 , b = y3 . What alternative statement appears?
2. The first equation will be a perfect cube, so take its cube root. Which of the equations
is (probably) easier to write in the form a3 + b3 + c3 − 3abc?

3. Factor the equation.

4. Where does the hypothesis that a and b are distinct real numbers come in to play?
Conclude.

â5.2 A Potpourri of Problems

You might ask yourself, why so many problems? Well, one major reason for creating
this article was just to be a collection of these problems (especially since there isn’t much
theory), but hopefully you also get to see just how widely this trick is used!
Problems are arranged in very roughly increasing order of difficulty.

Problem 51 (GGMT Speed 2020/2). Let p, q, r be the real roots of the polynomial x3 −
7x − 13. Evaluate p3 + q3 + r3 .

Remark 83. The problem is probably not intended to be solved with Titu’s Favorite Factoring
Trick, but I think it’s funnier that way.

Problem 52 (Vietnam MO 1985/1). Find all pairs of integers ( x, y) such that x3 − y3 =

2xy + 8.

52
NICE Committee (Second Edition) NICE Journal

Problem 53 (expii). An integer is called circular if it can be expressed as a3 + b3 + c3 −

3abc for some integers a, b, and c. Show that the product of any two circular numbers is
circular.

Problem 54 (Putnam 2015/B1). An equilateral triangle has all of its vertices on the
graph of x3 + 3xy + y3 = 1. Find its area.

Problem 55. Prove that the triangle ABC is isosceles if and only if
√3
√
3
√
a − b + b − c + 3 c − a = 0.

(As usual, a = BC, b = CA, and c = AB.)

Problem 56. Find all ordered pairs of real numbers ( x, y) satisfying

y3 1
 
3
ln x + ln y = ln x + + .
3 9

Problem 57 (BMT Algebra 2020/7). Let a, b, c be real numbers such that a + b + c =

1 1 1
a + b + c and abc = 5. Find the value of
3
1 1 3 1 3
    
a− + b− + c− .
b c a

Problem 58 (BMO 2007/2/1). Find the minimum value of x2 + y2 + z2 where x, y, z are

real numbers such that x3 + y3 + z3 − 3xyz = 1.

Problem 59. Let x and y be nonzero real numbers satisfying x3 + y3 + 3x2 y2 = x3 y3 .

Find all possible values of 1x + y1 .

Problem 60 (RMO/INMO Preparation Thread). Find all real x satisfying

1
2b x c3 + b x c{ x } + 2{ x }3 = .
108

53
NICE Journal NICE Committee (Second Edition)

Problem 61. Let x, y, z be integers such that

( x − y)2 + (y − z)2 + (z − x )2 = xyz.

Prove that x3 + y3 + z3 is divisible by x + y + z + 6.

Problem 62 (AMSP Admissions Test 2021/C.9). Find all positive integers m, n satisfy-
ing   
m3 + 196(28 − 3m) n3 + 256(32 − 3n) = 2021.

Problem 63. Suppose that x, y, z are distinct real numbers such that
q
3 3
p p
( x − y) 1 − z3 + (y − z) 1 − x3 + (z − x ) 3 1 − y3 = 0.

Prove that
(1 − x3 )(1 − y3 )(1 − z3 ) = (1 − xyz)3 .

Problem 64 (Purple Comet HS 2020/16). Let a, b, c be nonzero real numbers satisfying

r
3 a 3 b c
r r
a +b + c 3 = 0.
b c a

Find the maximum possible value of

2
a3 b3 c3

2
+ 2 + 2 .
b c c a a b

Problem 65 (Purple Comet HS 2021/21). Let a, b, and c be real numbers satisfying the
equations
a3 + abc = 26
b3 + abc = 78
c3 − abc = 104.
Find a3 + b3 + c3 .

√
3 √ √
3
Problem 66 (arqady). Let a, b and c be real numbers such that ab + 3
ac + bc ≥ 0.
Prove that ab + ac + bc ≥ 0.

54
NICE Committee (Second Edition) NICE Journal

Problem 67 (Purple Comet HS 2013/27). Suppose a, b, c are real numbers satisfying

a + b + c = 5 and 1a + 1b + 1c = 15 . Find the greatest possible value of a3 + b3 + c3 .

Problem 68 (Purple Comet HS 2018/19). Suppose that a and b are positive real num-

1,030,301− a−b
 √ √
bers such that 3 log101 3ab = 3 − 2 log101 ( ab). Find 101 − 3 a − 3 b.

Problem 69. Find all triples ( a, b, c) of integers satisfying

a3 − b3 − c3 = 3abc,
(

a2 = 2( a + b + c ).

Problem 70 (Purple Comet HS 2019/23). Find the number of ordered pairs of integers
( x, y) such that
x2 y2 1
 
− = 3 2+ .
y x xy

Problem 71 (JMC 12 2021/22). For positive reals a1 , a2 , . . . , an , the function D( a1 , a2 , . . . , an )

denotes the difference between the arithmetic mean and the geometric mean of those n
numbers. Suppose positive real numbers x, y, and z satisfy

 xyz√= 1, √

√
D( x, y, z) = 74 13 , √
√ √ √ √ √
D( x, y) + D( y, 3 z) + D( 3 z, 3 x ) = 37
3
9 .

3 3


√ √ √
What is the value of D( 6 x, 6 y, 6 z)?

Problem 72 (David Altizio). Suppose ABC is a triangle with angles measures A, B, and
C such that

tan A + tan B + tan C = 4 and cot A + cot B + cot C = 5.

Find tan3 A + tan3 B + tan3 C.

Problem 73 (MR J479, Titu Andreescu). Let a, b, c be nonzero real numbers, not all
equal, such that
3 3 3
a2 b2 c2 a2 b2 c2 bc ca ab
    
−1 + −1 + −1 =3 + + − 2− 2− 2 .
bc ca ab bc ca ab a b c

55
NICE Journal NICE Committee (Second Edition)

Prove that a + b + c = 0.

Remark 84 (Samuel Zhou). This (the above problem) is the best problem of all time. Its comedic
value is astronomical.

Problem 74 (Putnam 2019 A1). Determine all possible values of the expression

A3 + B3 + C3 − 3ABC

where A, B, and C are nonnegative integers.

Problem 75 (MR S546, Titu Andreescu). Find all triples of real numbers ( x, y, z) satis-
fying the system below.
3 3 1

 x − 2xyz + y = 2

y3 − 2xyz + z3 = 1
z − 2xyz + x3 = − 23
 3


√ √ √
Problem 76. Let a, b, c be the three real roots of x3 + x2 − 2x − 1. Find
3
3
a+ b+ 3
c.

Problem 77 (SMMT Champion 2021/8). Let ( x, y, z) be a solution to the following

simultaneous system of equations:

x + 2z + 3 y + 2x + 3 z + 2y + 3
+ + = x+y+z
y−z z−x x−y
1 1 1 420
+ + = .
y−z z−x x−y (y − z)(z − x )( x − y)

Given that x + y + z is not an integer, find ( x − y)3 + (y − z)3 + (z − x )3 .

Problem 78. Is 5123 + 6753 + 7203 prime?

3
p √ 3
p √
Problem 79. Let k be an integer and let n = k+ k2 + 1 + k− k2 + 1 + 1. Prove
that n3 − 3n2 ∈ Z.

56
NICE Committee (Second Edition) NICE Journal

Problem 80 (IWYMIC 2018/1/11). Find all pairs of real numbers ( x, y) satisfying the
system below. √ √
 3 x − 3 y = 12

x+y+8 3
 
 xy =
6


Problem 81 (HOMC Team 2018/J8). Let a, b, c be real numbers with a + b + c = 2018.

Suppose x, y, and z are distinct positive real numbers which satisfy a = x2 − yz − 2018,
b = y2 − zx − 2018, and c = z2 − xy − 2018. What is
√
a3 + b3 + c3 − 3abc
?
x + y3 + z3 − 3xyz
3

Problem 82 (CMIMC NT 2018/8). Find the unique set of positive primes { p, q, r } such
that
p3 + q3 + r 3
= 249.
p+q+r

Problem 83 (CMIMC Algebra-NT 2020/9). Let p = 10009 be a prime number. De-

termine the number of ordered pairs of integers ( x, y) such that 1 ≤ x, y ≤ p and
x3 − 3xy + y3 + 1 is divisible by p.

Problem 84 (MR J469). Prove that

(3a + 1)(3b + 1) = 3a2 b2 + 1

if and only if
√ √
3
2
3
a+ b = a2 b2 .

Problem 85 (USAMTS 3/4/16). Find a polynomial f ( x, y, z) in three variables with

integer coefficients
√ such
√ that for all integers a, b, c, the sign of f ( a, b, c) is the same as the
sign of a + b 3 2 + c 3 4.

Problem 86 (Mock AIME I 2015/11, David Altizio). Let a, b, c be complex numbers

57
NICE Journal NICE Committee (Second Edition)

satisfying the system 

 a + b + c = 6,

a3 + b3 + c3 = 87,
( a + 1)(b + 1)(c + 1) = 33.



1
Find a + 1b + 1c .

Problem 87 (Mock AIME 2017/10, Ankan Bhattacharya). Find the number of triples
of integers ( a, b, c) satisfying −100 ≤ a, b, c ≤ 100 and
3
a+b+c

= abc.
3

Problem 88 (US Algebra Math Open Winter 2015/14). How many functions f :
Z → Z are there such that f ( a3 ) = f ( a)3 for all a ∈ Z and f ( a3 ) + f (b3 ) + f (c3 ) =
3 f ( a) f (b) f (c) for all a, b, c ∈ Z satisfying a + b + c = 0?

Problem 89 (OMO Spring 2014/18). Find the number of pairs (m, n) of integers with
−2014 ≤ m, n ≤ 2014 such that x3 + y3 = m + 3nxy has infinitely many integer solutions
( x, y).

Problem 90 (MR J517). Let ( an )n≥1 be a sequence of positive real numbers such that
a1 = 1, a2 = 2 and
a3n+1 + a3n−1
+ an+1 an−1 = 3a2n .
9an
Find a closed form for an .

Problem 91 (Rioplatense MO 2007/3/4). Find all functions f : Z → Z with the

following property: if x + y + z = 0, then f ( x ) + f (y) + f (z) = xyz.

Problem 92 (MR S393). How many positive integers n exist such that n2 + 11 is prime
and n + 4 is a perfect cube?

Problem 93. Find the least real number r such that for each triangle with side lengths a,

58
NICE Committee (Second Edition) NICE Journal

b, and c, we have
max( a, b, c)
√
3
< r.
a3 + b3 + c3 + 3abc

Problem 94 (China TST 2019/3/1). Given complex numbers x, y, z satisfying | x |2 +

|y|2 + |z|2 = 1, prove that  
 3
x + y3 + z3 − 3xyz ≤ 1.


Problem 95 (OMO Spring 2018/24). Find the number of ordered triples ( a, b, c) of

integers satisfying 0 ≤ a, b, c ≤ 1000 for which

a3 + b3 + c3 ≡ 3abc + 1 (mod 1001).

Problem 96. Reals α, β, γ ∈ (0, π ) satisfy

cos α + cos β + cos γ = cos 2α + cos 2β + cos 2γ = cos 3α + cos 3β + cos 3γ = 0.

Find all possible values of sin α + sin β + sin γ.

Problem 97 (MR J489, Titu Andreescu). Prove that in any triangle ABC,
q q
8r ( R − 2r ) r (16R − 5r ) ≤ a3 + b3 + c3 − 3abc ≤ 8R( R − 2r ) (2R + r )2 + 2r2 .

(As usual, a = BC, b = CA, c = AB, r is the inradius, and R is the circumradius.)

√
4
√
3
Problem 98 (math.SE). Is 2+ 3 rational?

Problem 99 (Vietnam TST 2008/6). Consider the set M = {1, 2, . . . , 2008}. Paint every
number in the set M with one of the three colors blue, yellow, red such that each color is
used at least once. Define two sets:
S1 = {( x, y, z) ∈ M3 | x, y, z have the same color and 2008|( x + y + z)}.
S2 = {( x, y, z) ∈ M3 | x, y, z have three pairwise different colors and 2008|( x + y + z)}.

Prove that 2|S1 | > |S2 |.

59
NICE Journal NICE Committee (Second Edition)

Problem 100 (ELMO 2017/6). Find all functions f : R → R such that for all real
numbers a, b, and c:
• If a + b + c ≥ 0 then f ( a3 ) + f (b3 ) + f (c3 ) ≥ 3 f ( abc).

• If a + b + c ≤ 0 then f ( a3 ) + f (b3 ) + f (c3 ) ≤ 3 f ( abc).

Problem 101 (Putnam 1939/B7). Prove that a30 + a31 + a32 − 3a0 a1 a2 = 1, where
∞
x3n+i
ai = ∑ (3n + i )!
.
n =0

60
Editor’s Notes
Congrats, you made it to the end!

A short note
You may have noticed that the Incenter Miquel chapter includes its solutions to the
examples in the chapter rather than at the end (unlike the other chapters). This is because
many of the solutions are well-motivated and thus are worth reading in conjunction with
the rest of the material. The other chapters present their motivation in the walkthroughs
and therefore the solutions should only be read after the content is introduced.

Tell us your thoughts

We are always open to suggestions about the journal. Please let us know how you feel
about it through our Contact Us page.

Submit your work to the journal

Those in the NICE committee aren’t the only people who can submit entries. If you
are interested in contributing an article, please email nicecontest21@gmail.com with the
PDF and LATEX file attached. In the title of the email please include some identifying
information, like “NICE Journal Submission on [Topic].” Please also include your own
information in the email.

That’s it for now! See you guys in the next issue.

Dylan Yu

61
A Key Parts
âA.1 List of Problems

Chapter 1
4 Example – Euclid 2021/2.c . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
5 Example – Euclid 2014/3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
6 Example – blackpenredpen . . . . . . . . . . . . . . . . . . . . . . . . . . 8
7 Example – AoPS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
8 Example – Instagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1 Problem – Fermat 2019/24 . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2 Problem – Own . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3 Problem – Own . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4 Problem – Own . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
5 Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
6 Problem – Jeffrey Qin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
7 Problem – AIME 1987/14 . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
8 Problem – Purple Comet HS 2019/19 . . . . . . . . . . . . . . . . . . . . . 11
9 Problem – Purple Comet HS 2018/22 . . . . . . . . . . . . . . . . . . . . . 11
10 Problem – NICE Spring 2021/13, David Altizio . . . . . . . . . . . . . . . 11

Chapter 2
12 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
13 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
14 Exercise – Generating function for binomial ROUF . . . . . . . . . . . . . 14
15 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
16 Example – AMC 12A 2021/15 . . . . . . . . . . . . . . . . . . . . . . . . . 15
17 Example – AIME II 2016/12 . . . . . . . . . . . . . . . . . . . . . . . . . . 15
11 Problem – AMC 10B 2021/16 . . . . . . . . . . . . . . . . . . . . . . . . . 17
12 Problem – BMT Analysis 2015/7 . . . . . . . . . . . . . . . . . . . . . . . 17
13 Problem – AMC 12A 2017/25 . . . . . . . . . . . . . . . . . . . . . . . . . 17
14 Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
15 Problem – IMC 1999/8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
16 Problem – AIME I 2018/12 . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
17 Problem – HMMT February Team 2015/7 . . . . . . . . . . . . . . . . . . 17
18 Problem – HMMT February Combinatorics 2012/10 . . . . . . . . . . . . 17
19 Problem – Titu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
20 Problem – PUMaC Live 2018/4.3 . . . . . . . . . . . . . . . . . . . . . . . 18
21 Problem – HMIC 2021/3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
62
NICE Committee (Second Edition) NICE Journal

22 Problem – Putnam 1974/B6 . . . . . . . . . . . . . . . . . . . . . . . . . . 18

23 Problem – MOP 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

Chapter 3
20 Exercise – USAJMO 2014/6 . . . . . . . . . . . . . . . . . . . . . . . . . . 21
21 Exercise – EMMO Seniors 2016 . . . . . . . . . . . . . . . . . . . . . . . . 21
22 Exercise – CMIMC Tiebreaker 2018/G3 . . . . . . . . . . . . . . . . . . . 22
34 Exercise – ELMO SL 2019/G3 . . . . . . . . . . . . . . . . . . . . . . . . . 25
35 Exercise – ELMO 2010/6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
41 Example – Taiwan TST 2019/2/2 . . . . . . . . . . . . . . . . . . . . . . . 28
42 Example – MR O451 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
43 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
44 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
45 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
46 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
47 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
48 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
49 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
50 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
52 Example – Fake USAMO 2020/3 . . . . . . . . . . . . . . . . . . . . . . . 31
53 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
54 Example – CAMO 2020/3 . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
58 Example – HMMT November Team 2016/10 . . . . . . . . . . . . . . . . 37
59 Example – SORY 2019/6, Aditya Khurmi . . . . . . . . . . . . . . . . . . 37
60 Exercise – OMO Spring 2020/15 . . . . . . . . . . . . . . . . . . . . . . . . 38
61 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
62 Exercise – Due to math-pi-rate . . . . . . . . . . . . . . . . . . . . . . . . . 38
24 Problem – CJMO 2019/3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
25 Problem – GGG 2.3, Andrew Wu . . . . . . . . . . . . . . . . . . . . . . . 38
26 Problem – GGG 2.6, Andrew Wu . . . . . . . . . . . . . . . . . . . . . . . 39
27 Problem – GGG 3.3, Andrew Wu . . . . . . . . . . . . . . . . . . . . . . . 39
28 Problem – GGG 3.6, Andrew Wu . . . . . . . . . . . . . . . . . . . . . . . 39
29 Problem – Romania JBMO TST 2019/3 . . . . . . . . . . . . . . . . . . . . 39
30 Problem – Iran TST 2008/10 . . . . . . . . . . . . . . . . . . . . . . . . . . 39
31 Problem – Sharygin CR 2020/21 . . . . . . . . . . . . . . . . . . . . . . . . 39
32 Problem – trumpeter’s PoTD Forum 1/10/16 . . . . . . . . . . . . . . . . 40
33 Problem – tworigami . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
34 Problem – buratinogigle . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
35 Problem – POGCHAMP 2020/3 . . . . . . . . . . . . . . . . . . . . . . . . 40
36 Problem – POGCHAMP 2020/4 . . . . . . . . . . . . . . . . . . . . . . . . 40
37 Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

Chapter 4
65 Example – AMC 12A 2021/25 . . . . . . . . . . . . . . . . . . . . . . . . . 43
63
NICE Journal NICE Committee (Second Edition)

66 Exercise – PUMaC Number Theory 2010/A2 . . . . . . . . . . . . . . . . 43

68 Example – PUMaC Number Theory 2019/A2 . . . . . . . . . . . . . . . . 44
69 Example – PUMaC Number Theory 2016/A5 . . . . . . . . . . . . . . . . 45
70 Example – PUMaC Number Theory 2015/A6 . . . . . . . . . . . . . . . . 45
73 Example – ISL 1989/11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
74 Example – HMMT February Algebra-NT 2019/8 . . . . . . . . . . . . . . 47
38 Problem – AMC 12B 2021/7 . . . . . . . . . . . . . . . . . . . . . . . . . . 47
39 Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
40 Problem – Italy MO 1998/6 . . . . . . . . . . . . . . . . . . . . . . . . . . 47
41 Problem – NEMO Team 2018/9 . . . . . . . . . . . . . . . . . . . . . . . . 47
42 Problem – Elementary Number Theory, David Burton . . . . . . . . . . . 48
43 Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
44 Problem – Baltic Way 2019/16 . . . . . . . . . . . . . . . . . . . . . . . . . 48
45 Problem – Number Theory, George E. Andrews . . . . . . . . . . . . . . . 48
46 Problem – PAMO SL 2018/A5 . . . . . . . . . . . . . . . . . . . . . . . . . 48
47 Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
48 Problem – HMMT February Algebra-NT 2021/5 . . . . . . . . . . . . . . 48
49 Problem – ISL 2004/N2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
50 Problem – AMM 12003, Nikolai Osipov . . . . . . . . . . . . . . . . . . . 49

Chapter 5
80 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
81 Example – AHSME 1999/30 . . . . . . . . . . . . . . . . . . . . . . . . . . 51
82 Example – MR S499, Titu Andreescu . . . . . . . . . . . . . . . . . . . . . 52
51 Problem – GGMT Speed 2020/2 . . . . . . . . . . . . . . . . . . . . . . . . 52
52 Problem – Vietnam MO 1985/1 . . . . . . . . . . . . . . . . . . . . . . . . 52
53 Problem – expii . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
54 Problem – Putnam 2015/B1 . . . . . . . . . . . . . . . . . . . . . . . . . . 53
55 Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
56 Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
57 Problem – BMT Algebra 2020/7 . . . . . . . . . . . . . . . . . . . . . . . . 53
58 Problem – BMO 2007/2/1 . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
59 Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
60 Problem – RMO/INMO Preparation Thread . . . . . . . . . . . . . . . . 53
61 Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
62 Problem – AMSP Admissions Test 2021/C.9 . . . . . . . . . . . . . . . . . 54
63 Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
64 Problem – Purple Comet HS 2020/16 . . . . . . . . . . . . . . . . . . . . . 54
65 Problem – Purple Comet HS 2021/21 . . . . . . . . . . . . . . . . . . . . . 54
66 Problem – arqady . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
67 Problem – Purple Comet HS 2013/27 . . . . . . . . . . . . . . . . . . . . . 55
68 Problem – Purple Comet HS 2018/19 . . . . . . . . . . . . . . . . . . . . . 55
69 Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
70 Problem – Purple Comet HS 2019/23 . . . . . . . . . . . . . . . . . . . . . 55
64
 71 Problem – JMC 12 2021/22 . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
72 Problem – David Altizio . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
73 Problem – MR J479, Titu Andreescu . . . . . . . . . . . . . . . . . . . . . . 55
74 Problem – Putnam 2019 A1 . . . . . . . . . . . . . . . . . . . . . . . . . . 56
75 Problem – MR S546, Titu Andreescu . . . . . . . . . . . . . . . . . . . . . 56
76 Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
77 Problem – SMMT Champion 2021/8 . . . . . . . . . . . . . . . . . . . . . 56
78 Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
79 Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
80 Problem – IWYMIC 2018/1/11 . . . . . . . . . . . . . . . . . . . . . . . . 57
81 Problem – HOMC Team 2018/J8 . . . . . . . . . . . . . . . . . . . . . . . 57
82 Problem – CMIMC NT 2018/8 . . . . . . . . . . . . . . . . . . . . . . . . . 57
83 Problem – CMIMC Algebra-NT 2020/9 . . . . . . . . . . . . . . . . . . . 57
84 Problem – MR J469 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
85 Problem – USAMTS 3/4/16 . . . . . . . . . . . . . . . . . . . . . . . . . . 57
86 Problem – Mock AIME I 2015/11, David Altizio . . . . . . . . . . . . . . 57
87 Problem – Mock AIME 2017/10, Ankan Bhattacharya . . . . . . . . . . . 58
88 Problem – US Algebra Math Open Winter 2015/14 . . . . . . . . . . . . . 58
89 Problem – OMO Spring 2014/18 . . . . . . . . . . . . . . . . . . . . . . . 58
90 Problem – MR J517 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
91 Problem – Rioplatense MO 2007/3/4 . . . . . . . . . . . . . . . . . . . . . 58
92 Problem – MR S393 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
93 Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
94 Problem – China TST 2019/3/1 . . . . . . . . . . . . . . . . . . . . . . . . 59
95 Problem – OMO Spring 2018/24 . . . . . . . . . . . . . . . . . . . . . . . 59
96 Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
97 Problem – MR J489, Titu Andreescu . . . . . . . . . . . . . . . . . . . . . . 59
98 Problem – math.SE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
99 Problem – Vietnam TST 2008/6 . . . . . . . . . . . . . . . . . . . . . . . . 59
100 Problem – ELMO 2017/6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
101 Problem – Putnam 1939/B7 . . . . . . . . . . . . . . . . . . . . . . . . . . 60

âA.2 List of Theorems

Chapter 1
1 Fact – Completing the Square . . . . . . . . . . . . . . . . . . . . . . . . . 6
2 Theorem – Quadratic Formula . . . . . . . . . . . . . . . . . . . . . . . . . 6

Chapter 2
9 Theorem – Roots of Unity Filter . . . . . . . . . . . . . . . . . . . . . . . . 12
10 Corollary – Alternate Form of ROUF . . . . . . . . . . . . . . . . . . . . . 13
65
NICE Journal NICE Committee (Second Edition)

Chapter 3
18 Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
19 Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
23 Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
25 Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
27 Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
28 Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
29 Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
30 Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
31 Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
32 Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
33 Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
36 Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
51 Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
55 Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
56 Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
57 Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

Chapter 4
67 Theorem – Multiplicative Convolution . . . . . . . . . . . . . . . . . . . . 44
72 Theorem – Inverse of µ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

Chapter 5
76 Theorem – Titu’s Favorite Factoring Trick . . . . . . . . . . . . . . . . . . 50
77 Corollary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
78 Corollary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
79 Corollary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

âA.3 Index

antipode, 19 Dirichlet convolution, 44 Möbius function, 43

divisor function, 42 mixtillinear incircle, 20
circumcenter, 19 Euler phi function, 42 square-free, 43
circumcircle, 19 sum of divisors function, 42
constant function, 42 identity function, 42
contact triangle, 19 incenter, 19 unit function, 42

66
B Selected Solutions
âB.1 Solution 4 (Euclid 2021/2.c)
Let P, Q, and R be the ages of Paolo, Qing, and Rayna respectively. Then, we have the
following system of 3 equations:

Q
R=
2
P = Q+4
P+Q+R
= 13
3

The answer is (18, 14, 7) .

Q + 4 + Q + Q2
Substitute the first two equations into the third to get that = 13. We
3
can then simplify that to get Q = 14. Then, substitute this into the first two equations to
get that P = 18 and R = 7.

âB.2 Solution 5 (Euclid 2014/3)

Clearing denominators, we get that 7( a + b) = 4ab, or ab = 28. Then, applying Vieta’s
formulas, we note that a and b are the roots of the quadratic x2 − 16x + 28 = 0, which we
can easily factor to give the solutions for a and b.

âB.3 Solution 6 (blackpenredpen)

Let’s look at the restrictions for the problem. Clearly, 5 − x ≥ 0 (you can’t take the square
root of a negative number and get a real answer). Moreover, 5 − x2 ≥ 0, because the
range of the square
√ root√ function is strictly nonnegative. Combining these restrictions,
we get that − 5 ≤ x ≤ 5.
Instead of treating x as the variable, let’s square both sides and treat 5 as the variable:

5 − x = 52 − (2x2 )5 + x4
52 − (2x2 + 1)5 + x4 + x = 0
67
NICE Journal NICE Committee (Second Edition)

Use the quadratic formula on 5:

2x2 + 1 ± (−2x2 − 1)2 − 4( x4 + x )(1)

p
5=
√ 2
2 2
2x + 1 ± 4x − 4x + 1
5=
2
2x2 + 1 ± (2x − 1)
5=
2
This gives us two cases, 5 = x2 + x and 5 = x2 − x + 1. The final answer is left as an
exercise to the reader (you should get only 2 roots).

âB.4 Solution 7 (AoPS)

Let’s try to group factor naively:

mn + 3m − 8n = 59
m(n + 3) − 8n = 59

If only we could get a n + 3 factor somewhere. Let’s artificially force it to happen:

m(n + 3) − 8n = 59
m(n + 3) − 8n − 24 = 59 − 24
m(n + 3) − 8(n + 3) = 35
(m − 8)(n + 3) = 35

Thus, 35 is a product of these two factors, and we can do casework to check. The final
answer is left as an exercise to the reader.

âB.5 Solution 8 (Instagram)

The nested radical looks kind of funny, and it can normally be simplified. Let’s do some
magic:
q √ q √
19 − 8 3 = 16 − 8 3 + 3
√  √ 2
r
= 42 − 2 × 4 × 3 + 3
√ 2
r

= 4− 3
√
= 4− 3
68
NICE Committee (Second Edition) NICE Journal
√
So we have simplified the artificially complex condition to x = 4 − 3. Then:
√
x = 4− 3
√
x−4 = − 3
 √ 2
( x − 4)2 = − 3
x2 − 8x + 13 = 0
Conveniently, the denominator of the expression is x2 − 8x + 15, which is just equal to 2.
Now for the numerator, we will “power ladder.” Since we know that x2 − 8x + 13 = 0,
we can basically multiply both sides by anything we want and it will still equal 0. In other
words,
f ( x )( x2 − 8x + 13) = 0
for practically any function f . So let’s multiply both sides by f ( x ) = x2 and subtract that
off the numerator of the question to reduce the degree of the numerator. Repeating this
process gives us that the numerator is simply equal to x2 − 8x + 23, which is just 10. We
10
have simplified the expression to , which is just equal to 5 .
2

âB.6 Solution 12
By binomial theorem,
10
10
 
10
(1 + 1) = ∑ ,
k =0
k
10
10
 
(1 − 1)10 = ∑ (−1)k ,
k =0
k
and averaging the two quantities gives us the desired result of 512 .
P(1)+ P(−1)
Remark 85. For reference, we applied ROUF on P( x ) = (1 + x )10 , i.e. computed 2 .

âB.7 Solution 13
2πi
Let ω = e 3 . Then by applying ROUF on P( x ) = x −2 (1 + x )10 , we note that we have
shifted the coefficients by two (i.e. the kth coefficient of (1 + x )10 is now the k − 2th
coefficient of x −2 (1 + x )10 ), so (3k10
+2) for nonnegative integers k are now included in our
count. Thus,
10 10 10
     
+ + = P (1) + P ( ω ) + P ( ω 2 )
2 5 8
= 1 · (1 + 1)10 + ω −2 (1 + ω )10 + ω −4 (1 + ω 2 )10 .
69
NICE Journal NICE Committee (Second Edition)

Using ω 3 = 1, we get that

1 · (1 + 1)10 + ω (1 + ω )10 + ω 2 (1 + ω 2 )10 = 1026,

1
implying the answer is 3 · 1026 = 342 .

âB.8 Solution 15
Solution by Evan Chen.
We can rewrite the sum as
1000
 
∑ n
f (n)
n ≥0

where (
1 n ≡ 0 (mod 3)
f (n) =
0 otherwise.

The trick is that we can take

1 n 
f (n) = 1 + ω n + ω 2n
3
2πi
where ω = e 3 is a cube root of unity, satisfying the relation ω 2 + ω + 1 = 0. Thus, we
have
1000 1 1000
   
∑ n f (n) = 3 ∑ n (1 + ωn + ω2n ).
n ≥0 n ≥0

We can swap the order of summation now and instead consider

1000 1 1000 1 1000 1 1000

       
∑ n f (n) = 3 ∑ n + 3 ∑ n ω + 3 ∑ n ω2n . n
n ≥0 n ≥0 n ≥0 n ≥0

By the binomial theorem, the expression in question is

1000 1h
  i
∑ n f (n) = 3 (1 + 1)1000 + (1 + ω )1000 + (1 + ω2 )1000
n ≥0
1 h 1000 i
= 2 + (−ω 2 )1000 + (−ω )1000
3
1 h 1000 i
= 2 + ω + ω2
3
1 h 1000 i
= 2 −1 .
3

70
NICE Committee (Second Edition) NICE Journal

âB.9 Solution 16 (AMC 12A 2021/15)

Solution by AoPS user lawliet163.
Let f ( x, y) = (1 + x )8 (1 + y)6 . By expanding the binomials and distributing, f ( x, y) is
the generating function for different groups of basses and tenors. That is,

8 6
f ( x, y) = ∑ ∑ amn xm yn
m =0 n =0

where amn is the number of groups of m basses and n tenors. What we want to do is
sum up all values of amn for which 4 | m − n except for a00 = 1. To do this, define a new
function
8 6
g( x ) = f ( x, x −1 ) = ∑ ∑ amn xm−n = (1 + x)8 (1 + x−1 )6 .
m =0 n =0

Now we just need to sum all coefficients of g( x ) for which 4 | m − n. Consider a monomial
h( x ) = x k . If 4 | k,

h(i ) + h(−1) + h(−i ) + h(1) = 1 + 1 + 1 + 1 = 4

otherwise,
h(i ) + h(−1) + h(−i ) + h(1) = 0.

g( x ) is a sum of these monomials so this gives us a method to determine the sum we’re
looking for:
g(i ) + g(−1) + g(−i ) + g(1)
= 212 = 4096
4
(since g(−1) = 0 and it can be checked that g(i ) = − g(−i )). Hence, the answer is
4096 − 1 with the −1 for a00 which gives 95 .

âB.10 Solution 17 (AIME II 2016/12)

Three interesting solutions are presented below. You can alternatively do the problem
with easier methods, e.g. PIE, casework, or recursion.

Solution via generating functions and ROUF We use generating functions. Suppose that
the colors are 0, 1, 2, 3. Then as we proceed around a valid coloring of the ring in the
clockwise direction, we know that between two adjacent sections with colors si and si+1 ,
there exists a number di ∈ {1, 2, 3} such that

s i +1 ≡ s i + d i (mod 4).
71
NICE Journal NICE Committee (Second Edition)

Thus, we can represent each border between sections by the generating function x +
x2 + x3 , where x, x2 , x3 correspond to increasing the color number by 1, 2, 3 (mod 4),
respectively. Thus the generating function that represents going through all six borders is

A ( x ) = ( x + x 2 + x 3 )6 ,

where the coefficient of x n represents the total number of colorings where the colors’
numbers are increased by a total of n as we proceed around the ring. But if we go through
all six borders, we must return to the original section, which is already colored. Therefore,
we wish to find the sum of the coefficients of x n in A( x ) with n ≡ 0 (mod 4).
Thus, the sum of the coefficients of A( x ) with powers congruent to 0 (mod 4) is

A(1) + A(i ) + A(−1) + A(−i ) 36 + (−1)6 + (−1)6 + (−1)6 732

= = .
4 4 4
We multiply this by 4 to account for the initial choice of color, so the answer is 732 .

Solution via linear algebra (Allen Wang) Consider the graph K4 and its corresponding
adjacency matrix A. The answer to the problem is therefore the trace of A6 by definition
of matrix multiplication. Since B = A + I is a 4 × 4 matrix full of 1s, B has a rank of 1,
and therefore has 3 eigenvalues of 0. Since the trace of B is 4, the last eigenvalue is 4.
Thus, A has eigenvalues −1, −1, −1, 3.
The trace of A6 is the sum of the 6th powers of its eigenvalues, so we find that the trace
is 36 + 3 = 732 , as desired.

Solution via chromatic polynomials The chromatic polynomial for a cycle Cn is ( x −

1)n + (−1)n ( x − 1), where x is the number of colors and n is the number of sections in
the cycle. Clearly x = 4 and n = 6, so the answer is

(4 − 1)6 + (−1)6 (4 − 1) = 732 .

Remark 86. Chromatic polynomials are actually what this problem is based on, and thus not
the intended solution. Contest problems that can be destroyed with chromatic polynomials
usually succumb to casework as well, but I recommend you learn them as they can be useful
for timed competitions.

âB.11 Solution 65 (AMC 12A 2021/25)

Note f is multiplicative, so it suffices to determine for each prime p which value of k
maximizes
k+1
f ( pk ) = √ k .
3 p

72
NICE Committee (Second Edition) NICE Journal

√ √
Note that for p > 8, 3 p > 2 so k = 0 is optimal by the bound 3 pk > 2k = (1 + 1)k ≥ 1 + k
for k ≥ 1. We now compute k for p ∈ {2, 3, 5, 7}. Note that

k+1 k+2 √ k+2

f ( pk ) > f ( pk+1 ) ⇐⇒ √ k > √ k+1 ⇐⇒ 3 p > ,
3 p 3 p k+1

so we need to find the first k with

√ k+2
3 p>
k+1
2
because kk+ 3 3
+1 is a decreasing function. For p ∈ {5, 7}, it turns out that 2 > p > (3/2) =
27/8. For p = 3, it is the case that (3/2)3 = 27/8 > p > (4/3)3 = 64/27. For p = 2, it is
the case that (4/3)3 = 64/27 > p > (5/4)3 = 125/64. Hence

N = 23 32 51 71 = 2520 .

For fun, this maximal value f ( N ) is

τ (2520) 4·3·2·2 24
√
3
= √3
= √
3
≈ 3.527.
2520 2 315 315

âB.12 Solution 68 (PUMaC Number Theory 2019/A2)

Note ∑d|i f (d) is multiplicative because f is multiplicative and this function is ( f ∗ 1)(n).
Hence it suffices to compute the function for prime powers pk . We have

∑ f (d) = (−1)0 + (−1)1 + · · · + (−1)k ,

d| pk

which is 1 if k is even and zero otherwise. Hence the problem simply asks us for the
√
number of squares at most 2019, which is b 2019c = 44 .

âB.13 Solution 69 (PUMaC Number Theory 2016/A5)

Note that each prime divisor of k functions “independently” in the expression. Hence it
is sufficient to compute

gcd(d, t/d) gcd(d, t/d)2

∑ lcm(d, t/d) = ∑ t
d|t d|t

for each of t ∈ {26 , 35 , 52 , 73 , 53} and multiply the expressions together.

73
NICE Journal NICE Committee (Second Edition)

• t = 26 : The expression is
1 + 22 + 24 + 26 + 24 + 22 + 1 106 53
= 6 = 5.
26 2 2
• t = 35 : The expression is
1 + 32 + 34 + 34 + 32 + 1 2 · 91
= 5 .
35 3
• t = 52 : The expression is
1 + 52 + 1 33
2
= 2.
5 5
• t = 73 : The expression is
1 + 72 + 72 + 1 22 · 52
= .
73 73
• t = 53: The expression is
1+1 2
= .
53 53
Hence
53 2 · 91 33 22 · 52 2 1 13 1 13
S= 5
· 5 · 2· 3
· = · 2· 2 = .
2 3 5 7 53 2 3 7 882
The answer is then 13 + 882 = 895 .

âB.14 Solution 70 (PUMaC Number Theory 2015/A6)

Clearly τ (n)3 is multiplicative, so (τ (n)3 ∗ 1)(n) is multiplicative as well. Hence we need
to compute this for prime powers pk . At prime powers,

(i + 1)(i + 2) 2
 
3 3
∑ τ ( t ) = ∑ ( i + 1) = 2
.
k
t| p 0≤ i ≤ k

Thus 5 divides this if i is 3 or 4 mod 5 and 7 divides this if i is 5 or 6 mod 7. Then it must
be optimal to take n = 25 · 33 = 864 .

âB.15 Solution 73 (ISL 1989/11)

e e
By Möbius inversion, an = ∑d|n µ(n/d)2d . Let n = p11 p2e2 · · · pkk for positive integral ei .
e
Of course, the result is immediate for n = 1. For each pi i , note the expression disappears
e e −1
by considering instances of pi i and pi i in the numerator of 2d and using Euler’s totient
theorem.
74
NICE Committee (Second Edition) NICE Journal

âB.16 Solution 74 (HMMT February Algebra-NT 2019/8)

Note the condition is f ∗ f = 1. We require f to be multiplicative: this is clearly okay
because then checking f ∗ f = 1 follows from checking that for prime powers. We now
compute f on prime powers pk . Let f ( pk ) = g(k). We have

∑ g( a) g(b) = 1 ∀k, a, b ≥ 0.
a+b=k

Then let
∞
G(x) = ∑ g (i ) x i
i =0
and observe
1
G ( x )2 = 1 + x + x 2 + · · · = .
1−x
Hence
1
G(x) = √ .
1−x
1 (2019) (0)
Then g(2019) is the value of 2019! G where 2019 denotes 2019 differentiations.
That is,
1 1 3 4037 4037!!
g(2019) = · · ··· = .
2019! 2 2 2 4038!!
This implies that
4037!!2
f (20182019 ) = .
4038!!2

âB.17 Solution 80
√ √
3 √ √
3
√3
√
3
Since 3 a + √b + 3√c = 0,√we have a3 + b3 + c3 = 3abc by corollary 1.4. But
3 3 3
3abc = 3, so a3 + b3 + c3 = a + b + c = 3 .

âB.18 Solution 81 (AHSME 1999/30)

Rewrite the equation as
m3 + n3 − (−33)3 − 3 · m · n · −33 = 0.
Factoring, we have
1  
(m + n − 33) (m − n)2 + (m + 33)2 + (n + 33)2 = 0.
2
Thus, either m + n − 33 = 0 or (m − n)2 + (m + 33)2 + (n + 33)2 = 0. The answer is
evidently 35 .
75
NICE Journal NICE Committee (Second Edition)

âB.19 Solution 82 (MR S499, Titu Andreescu)

Let a = x3 , b = y3 . Then the first equation becomes

27x3 y3 ( x + y)3 = 1,

3xy( x + y) = 1,
and the second equation becomes

27x3 y3 ( x3 + y3 + 1) = 1,

(3x2 y)3 + (3xy2 )3 + (−1)3 − 3(3x2 y)(3xy2 )(−1) = 0,

and so we clearly must have that

3x2 y + 3xy2 − 1 = 0

as desired. Note that there is some unrigorousness to this proof, because we have not
shown these steps are bidirectional; this is left as an exercise to the reader.

76