See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/344265857

Polynomial Problems From Recent Iranian Mathematical Olympiads: Elements

From The Iranian Experience In Teaching Polynomials For Mathematical
Olympiad Curricula

Article · September 2020

CITATIONS READS
0 2,395

1 author:

Navid Safaei
Sharif University of Technology
42 PUBLICATIONS   8 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Epistemological Studies in Evolutionary Economics(ESEE) View project

American Mathematical Monthly(AMM), problems and solutions. View project

All content following this page was uploaded by Navid Safaei on 16 September 2020.

The user has requested enhancement of the downloaded file.

Polynomial Problems From Recent Iranian Mathematical Olympiads: Ele-
ments From The Iranian Experience In Teaching Polynomials For Mathe-
matical Olympiad Curricula.
Navid Safaei
Sharif University of Technology, Tehran, Iran
Email: navid [email protected]
Navid is a senior researcher at the Sharif University of Technology, Tehran, Iran. His
main research interests is evolutionary theory for the Socio-economic domain. He is
the head of mathematical Olympiad department in Salam High schools complex since
2009. He has participated as instructor and problems proposer in Iranian TST, since
2014. He was involved in training curricula for mathematics competitions since 2005. He
wrote more than 20 books(in Persian) for high school mathematics including Calculus,
Mathematics 9, 10, 11, since 2005, at Olgoo publication, Tehran, Iran.
Navid is a local organizer of Iran for European Mathematical Cup(EMC) and Silk Road
Mathematical Competition(SRMC). He was the Iranian team leader for the Romanian
Masters of Mathematics (RMM) competition. As an instructor, He has been invited
in many training camps around the world such as Azerbaijan, Bulgaria, Croatia, Singa-
pore, etc. He also published many articles concerning methods and techniques for solving
mathematical competitions problems in Mathematical reflections(MR), Cibiti Matem-
atiki, i. e. , USM (in Ukrainian), Mathematika (in Bulgarian). He also proposed some
problems for the American mathematical monthly(AMM) journal. Navid is contribut-
ing with XYZ-Press publishing house and his first book about polynomials published by
February 2019.

1. Introduction
Starting from 2015, the Algebra exams of the third round and TST of the Iranian
Mathematical Olympiad have contained some interesting problems about polynomials.
Most of these problems are original while challenging. This article outlines the conceptual
framework concerning ideas about polynomials that touch on some strong points in
tandem with some points concerning my teaching experiences. Having done this, the
kernel of ideas for this article is to introduce the readers to the Iranian approach toward
mathematical competitions with exclusive regard to the Algebra and polynomials.

2. Finding polynomials and examining coefficients

The most generic approach to teach polynomials is to start with coefficients and iden-
tities. Hence, a good entry point into our topic is to start with a polynomial equation.
However the central idea of this problem seems too easy, a few students solved the prob-
lem completely. It was partly because this problem was the fifth problem of an exam
with 6 problems that took 6 hours! The author of this problem was Mojtaba Zare who
received a gold medal in the IMO 2015 and immediately started to propose problems
and teach Algebra for the Mathematical Olympiad curricula. The students can find too
 2

many interesting problems concerning finding polynomials and examining the coefficients
in Andreescu, et al (2019, 2020).

Problem 1. (3rd round Iranian mathematical Olympiad, 2015, Algebra exam,

Problem 5)
Find all polynomials P (x) with real coefficients satisfying P (5x)2 − 3 = P (5x2 + 1),
whenever:
i. P (0) 6= 0,
ii. P (0) = 0.
Solution. Suppose that P (x) is not constant. Let degP (x) = d and P (x) = a0 + a1x +
· · · + ad xd . Comparing the coefficients of both sides we find that ad = 5( − d). Then, we
can prove by induction that all the coefficients of√P (x) are rational. Now, considering
the equation 5x2 + 1 = 5x, we have that r = (5 + 5)/10 is one √ of the roots.√Since P (x)
has rational coefficients, it is easy to deduce P (5r) = P ((5 + 5)/2) = c + d 5, for some
rational numbers c, d. Further, putting x = r in the original equation, we can find √ that
2 2 2
P (5r) − 3 = P (5r + 1). That is P (5r) − P (5r) − 3 = 0. Hence, P (5r) = (1 + 13)/2.
Impossible!
Remark. As you have seen, this solution doesnt need the condition on P (0). During
the meeting of Algebra team, the proposer outlined the following proof.
It is easy to deduce P (5x)2 = P (−5x)2 . Thus, P (x) is either odd or even. Assume
the latter, then P (x) = Q(x2), for some polynomial Q(x). Hence, rewrite the original
equation as !
2 x2
P (x) − 3 = P +1 .
5
2
Since P (x) is a polynomial in x2, we can assume that P (x) = R( x5 + 1). For some
polynomial R(x). Hence
!2  2

x2 ( x + 1)2
R +1 = R 5 + 1 .
5 5

2 2
Putting t = x5 + 1, then R(t)2 − 3 = R( t5 + 1) and degR(x) = 12 degP (x). Hence,
continuing this way, we shall face an infinite descent unless P (0) = 0. Therefore, the
first case would be reduced to the second case.
Assuming1 now the second case, define the following sequence a0 = 0, an+1 = 1 + (a2n )/5.
We can prove by induction that ak ∈ [1, 2), for each positive integer k ≥ 1. Now, define
the sequence bn = P (an ). Putting x = an in the original equation, we can find that
bn+1 = b2n − 3. It is clear that for n = 1, the sequence assumes positive integer values
and bn+1 > bn for each n ≥ 1. Thus limn→∞ bn = +∞. On the other hand, since P (x)
1
This part of the proof is the refined version of a proof we find in an exam paper. Unfortunately,
this student just solved the second part, assuming P (0) = 0. Thanks to Mojtaba Zare for sending the
ideas from the exam paper.
 3

is a polynomial, P ([1, 2)) is bounded. That is, bn should be bounded, which yields a
contradiction.
The next problem is selected from our 2017 TST. I was the author of this problem.
I thought that it was a very easy problem. But, from our 17 gold medalists, only
6 completely solved it! Most of the students had some computations concerning the
coefficients but had not taken this idea further.

Problem 2. ( 2017 Iranian TST, Test 3, Problem 5) Let {ci }∞ i=0 be a sequence of
non-negative rational numbers such that c2017 > 0. Define the sequence of polynomials:

P−1 (x) = 0, P1 (x) = 1, Pn+1 (x) = xPn (x) + cn Pn−1 (x).

Prove that for all n > 2017 there does not exist an integer n such that P2n (x) = Pn (x2 +c)
for some rational number c.
Solution. It could immediately be deduced that:

Pd (x) = xd + (c1 + c2 · · · + cd−1 )xd−1 + (c3 c1 + c4c1 + c4 c2 + · · · + cd−1 cd−2 )xd−2 + · · · .

Pd−1 Pk−2
The coefficient of xd−2 could be written as k=3 ck l=1 cl . This is also equal to
d−1 d−2
!
1 X X
(c1 + c2 + · · · + cd−1 )2 − c2k − ck ck+1 .
2 k=1 k=1

Now, assume by contradiction that there is such an n. By comparing the coefficients of

x2n−1 , x2n−2 we find that:

1 2n−1
X
c = ck ,
n k=1
2n−1 k−2
X X n(n − 1) 2 n−1
X
ck cl = c + ck .
k=3 l=1 2 k=1

Substituting the first equality into the second one, we find that
 2
1 2n−1
X

2n−1
X
2
2n−2
X n−1
X
ck = ck + 2 ck ck+1 + 2 ck (∗).
n k−1 k=1 k=1 k=1

Now we prove that the above equality is wrong!

Note that, by the Cauchy-Schwartz inequality, we have:
 2
1 2n−1
X n−1
X
ck  = c21 + (c2k+1 + c2k )2 .
n k−1 k=1

And:  2
1 2n−1
X n−1
X 2
ck  = c22n−1 + (c2k−1 + c2k ) .
n k−1 k=1
 4

Yielding:  2
2 2n−1
X 2n−1
X 2n−2
X
ck  = 2 c2k + 2 ck ck+1 .
n k−1 k=1 k=1

P2n−2 Pn−1
Applying the resulting inequality to (∗) we find that k=1 ck ck+1 + 2 k=1 ck = 0. Since
ck = 0 and c2017 > 0 this is impossible! we are done!

3. On roots of polynomials
The second important topic in teaching polynomials is: roots. Though most of the stu-
dents solved many problems on Vieta’s formula and Intermediate Value Theorem(IVT),
they found these two problems very hard. It seems that there were some weaknesses in
their arguments or lack of innovative ideas. The first problem was proposed by me for
the final exam of Seemurg Training Camp of Nations (STCN) that was held in January
2019 between 6 countries2 around the world. Only 8 out of 124 students solved this
problem. The reason was they were not brave enough to deal with the complexity which
emerges after writing Vieta’s formula for the roots.

Problem 3. (2019 STCN, Final Exam, Problem 4)3 Assume that all the roots of
P (x) = xd − a1xd−1 + · · · + (−1)d ad lie on [0, 1]. Prove that for all k = 1, 2, . . . , d we have

ak − ak+1 + · · · + (−1)d−k ad = 0.

Solution. Let us consider P (x) = (x − x1)(x − xd ) where xi ∈ [0, 1]. Note that, by
Vietaś formula we have:
X
ai = ai (x1, x2, . . . , xd) = xs1 · · · xsi .
1=s1 ,,si =d

Now we prove the statement of problem by induction on d. The case d = 1 is obvious.

Note that:
P (1) = 1 − a1 + a2 − · · · + (−1)d ad = 0.
Thus:

(−1)k ak + (−1)k+1 ak+1 + · · · + (−1)d ad = P (1) − 1 + a1 − a2 + · · · + (−1)k ak−1 .

Therefore, the desired inequality leads to:

(−1)k (P (1) − 1 + a1 − a2 + · · · + (−1)k ak−1 ) ≥ 0.

Assume the statement of the problem holds true for all positive integers less than or
equal to d. Now we must prove for the polynomial

P (x) = (x − x1)(x − xd )(x − xd+1 ) = xd+1 − b1xd + · · · + (−1)d+1 bd+1

2
The competing countries were Azerbaijan, Bangladesh, Iran, Mexico, Mongolia and Tajikstan
3
Two months later this problem appeared in the problems section of Kvant (2019/1) as Problem
M2544
 5

that: bk − bk+1 + · · · + (−1)d+1−k bd+1 = 0 for all k = 1, . . . , d + 1.

Note that:

bi = bi (x1, . . . , x( d + 1)) = ai (x1, x2 , . . . , xd ) + x(d + 1)a(i − 1)(x1 , x2, . . . , xd ).

According to the above, we must prove:

(−1)k (Q(1) − 1 + b1 − b2 + · · · + (−1)k bk−1 )

= (−1)k (Q(1) − 1 + a1 + xd+1 − a2 − xd+1 a1 + · · · + (−1)k (ak−1 + xd+1 ak−2 )
= 0.

Since Q(1) = (1 − xd+1 )P (1) one can rewrite the above inequality as:

(−1)k (1 − xd+1 )(P (1) − 1 + a1 − a2 + · · · + (−1)k ak−1 ) ≥ 0,

which is true, due to the induction hypothesis. However, the next problem has a simple
statement, it needs insightful knowledge about the IVT and the notion of multiple roots.
Most of the contestants didn’t consider the fact that in the neighborhood of a root r,
we face a change in the sign of polynomial unless the root is double or has even number
of occurrences. That is, P (x) = (x − r)2 aQ(x), for some natural number a and some
polynomial Q(x) such that Q(r) 6= 0.

Problem 4. (3rd round Iranian mathematical Olympiad, 2019, Algebra exam,

Midterm, Problem 3) For each positive integer d, find all open intervals I ⊂ R, of
largest length, such that for any choice of a0 , . . . , a2d−1 ∈ I the polynomial P (x) =
x2d + a2d−1x2d−1 + . . . + a0 , has no real root.
Solution. The answer is I = (1, 1 + 1/d). Assume that the desired interval is of the
form of (b, c). For some q, r in (b, c), put ai = q for odd i and ai = r for even i, where
0 = i = 2d − 1. Then
P (−1) = 1 − dq + dr.
Since P (x) has no real root, we must have P (−1) > 0. Thus, q − r < 1/d. We can
assume that c − b = 1/d. Therefore, consider the interval as (b, b + 1/d). It is easy to
find that b > 0, since by choosing a0 = b + , for some sufficiently small  > 0, we find
that P (0) = a0 = b +  should be positive.
Now, assign the following numbers to the coefficient of the polynomial.

a2d−1 = a2d−3 = · · · = a1 = b + 1/d + , a2d = a2d−2 = · · · = a0 = b + ,

for some sufficiently small  > 0.

It is clear that for all positive real x, P (x) > 0 for all negative real x. Putting x = −t,
where t > 0, then,

P (−t) = t2d − (b + 1/d)t2d−1 + bt2d−2 − (b + 1/d)t2d−3 + · · · − (b + 1/d)t + b + Q(t),

 6

for some polynomial Q(t), degQ(t) = 2d − 1. As  tends to zero, it remains to find all
b > 0 such that
     
2d 1 2d−1 1 2d−3 1
R(t) = t − b + t + bt2d−2 − b + t + ··· − b + t + b ≥ 0.
d d d
Note that R(1) = 0. Therefore, R0 (1) should be zero. That is,
   
1 1
2d − (2d − 1) b + + (2d − 2)b − · · · − b + = 0.
d d
Hence, d(1 − b) = 0. That is, d = 1.
Now, we claim the interval I = (1, 1+1/d) works! It is obvious that P (x) has no positive
real roots. Now, we prove that P (x) > 0 for all negative real x. Put x = −t, t > 0.
Then,
     
2d 1 2d−1 1 2d−3 1
P (−t) > t − 1 + t + t2d−2 − 1 + t + ··· − 1 + t + 1.
d d d
Thus, it remains to prove that;
(t2d + t2d−2 + · · · + t2 + 1) (t2d−1 + t2d−3 + · · · + t3 + t)
≥ .
(d + 1) d

Now, because t2d + 1 ≥ t2d−2k+1 + t2k−1 , t2k + t2k−2 ≥ 2t2k−1), we are done.
No students got complete points for this problem. We had 6 students who got 6/15 from
this problem. They proved that the length of the interval is at most 1/d and proved the
interval I = (1, 1 + 1/d) satisfies the problem conditions. But, they couldnt prove that it
is the only solution. As the author of this problem, I expected more complete solutions
from 80 contestants.
In the following remark, I provide an alternative approach to find the optimal case. It
needs some knowledge about limits along with knowledge about the roots. In effect, this
approach seems more complicated for high school students. I provide this approach here
to show that despite their complexity, sometimes advanced techniques can nevertheless
be reconciled with elementary techniques.
Remark. There is an alternative proof for the segment where we proved b = 1. One
can find that
1
t2d − b(t2d−1 − t2d−2 + · · · + t − 1) − (t2d−1 + t2d−3 + · · · + t) ≥ 0.
d

Hence for all t > 1,

1 dt2d − (t2d−1 + t2d−3 + · · · + t)
b ≤ . 2d−1 .
d t − t2d−2 + · · · + t − 1
And, for all t < 1,
1 dt2d − (t2d−1 + t2d−3 + · · · + t)
b ≥ . 2d−1 .
d t − t2d−2 + · · · + t − 1
 7

Thus,
1 dt2d − (t2d−1 + t2d−3 + · · · + t)
b≤ lim .
d t→1+ t2d−1 − t2d−2 + · · · + t − 1
And
1 dt2d − (t2d−1 + t2d−3 + · · · + t)
b≥ lim ,
d t→1− t2d−1 − t2d−2 + · · · + t − 1
since the function
dt2d − (t2d−1 + t2d−3 + · · · + t)
t2d−1 − t2d−2 + · · · + t − 1
has a limit as t approaches 1. Thus, we find that
dt2d − (t2d−1 + t2d−3 + · · · + t) dt2d − (t2d−1 + t2d−3 + · · · + t)
lim = lim .
t→1− t2d−1 − t2d−2 + · · · + t − 1 t→1+ t2d−1 − t2d−2 + · · · + t − 1
Hence,
1 dt2d − (t2d − 1 + t2d−3 + · · · + t)
b = . lim .
d t→1 t2d−1 − t2d−2 + · · · + t − 1
Finally, note that
dt2d − (t2d−1 + t2d−3 + · · · + t)
t2d−1 − t2d−2 + · · · + t − 1
(t2d − t2d−1 ) + (t2d − t2d−3 ) + · · · + (t2d − t)
=
(t − 1)(t2d−2 + t2d−4 + · · · + 1)
(t − 1)(t2d−1 + t2d−3 (t2 + t + 1) + · · · + t(t2d−2 + · · · + 1)
= .
(t − 1)(t2d−2 + t2d−4 + · · · + 1)
Thus,
dt2d − (t2d−1 + t2d−3 + · · · + t)
lim
t→1 t2d−1 − t2d−2 + · · · + t − 1
t2d−1 + t2d−3 (t2 + t + 1) + · · · + t(t2d−2 + · · · + 1)
= lim
t→1 t2d−2 + t2d−4 + · · · + 1
d2
=
d
= 1.
1
That is, b = .d = 1.
d
4. Absolute value, Triangle inequality, and Complex numbers
The third topic that I use to teach when I have a lecture on polynomials, is using the
triangle inequality. Most of the students know it very well. Meanwhile, they can not
adopt it in their arguments and proof procedures. Further, proofs proceed based on the
triangle inequality need some special regard to the equality case(s), too.
This problem was proposed by the author. 5 students from the 13 gold medalists solved
it. The interesting issue was that all of these 5 students became members of our national
team for the IMO 2019. It seems solving this problem brings fruition!
 8

Problem 5. (Iranian TST, 2019, Exam 3, Problem 1) Let 1 < t < 2 be a

real number. Prove that for all sufficiently large positive integers d, there is a monic
polynomial P (x) of degree d, such that all of its coefficients are either 1 or −1 and:

|P (t) − 2019| ≤ 1.

Solution. At first, we shall prove the following lemma;

Lemma 1. Let bn be a sequence of positive real numbers satisfying bn = 2b0 + b1 + · · · +
bn−1 . Then for each real number z, |z| ≤ 2b0 + b1 + · · · + bn , there are a0, · · · , an ∈ {1, −1}
such that:  
 n
X 
 
z − aibi  ≤ b0 .
 
i=0

Proof. Write the inequality |z| ≤ 2b0 + b1 + · · · + bn−1 in the form

|z − Sgn(z)bn | ≤ 2b0 + b1 + · · · + bn−1 . Then proceed to the proof by induction on n. For
sake of convenience, we also define Sgn(0) = 1. This completes our proof.
Back to our problem. Let us define bi = ti. Then it is easy to deduce that
tn − 1
bn − b0 = tn − 1 ≤ 1 + t + · · · + tn−1 = = b0 + b1 + · · · + bn−1 .
t−1
Moreover, choose d such that td = 2019. Then

2019 ≤ td ≤ 2 + t + · · · + td .

Hence, by our lemma, there are a0, · · · , ad ∈ {1, −1} such that:
 
Xd 
 
 aiti − 2019 = 1.
 
i=0

We are done.

In the next problem, we need a tricky identity. During July 2017, the problems selec-
tion committee told me that they needed a complex number problem. I proposed this
problem. Unfortunately, students were not good enough at polynomials with complex
coefficients. We had only 3 complete scores. I designed the first part of the problem in
a way that its solving works as a hint for the second part. That is, the student could
realize some facts from the relation between
 the roots of a polynomial and its reciprocal
polynomial (i. e., polynomials P (x), x P x1 , and d = degP (x)). The inquisitive readers
d

can find some interesting lines of thought about roots of unity, reciprocal polynomials,
etc in Safaei(2019).
 9

Problem 6. (3rd round Iranian mathematical Olympiad, 2017, Final Algebra

exam, Problem 2) Let P (z) = a0 + a1z + · · · + ad z d be a polynomial with complex
coefficients, we define its reverse as:

P ∗ (z) = a0 z d + a1z d−1 + · · · + ad .

Prove that
 
1
i P ∗ (z) = z d P z
.

ii Let all roots of the polynomial qd−l (z) of degree d lie inside or on the unit circle
(l > 0), (l ∈ Z). Prove that all the roots of the polynomial below lie on the unit
circle:
Q(z) = z l qd−l (z) + qd−l
∗
(z).

Solution i. Assume qd−l (z) = (z − z1) · · · (z − zd−l ) and |zi | ≤ 1 for all i = 1, · · · , d − l.
 
1
Then we can find that P ∗ (z) = z d P z
.
1
Solution ii. If s is a root of P (z) then, s
must be the root of P ∗ (z), which leads to:
∗
qd−l (z) = (1 − zz1) · · · (1 − zzd−l ).

Assume Q(r) = 0 for some complex number r. Then:

rl qd−l (r) + qd−l

∗
(r) = 0.

Thus, rl qd−l (r) = −qd−l

∗
(r) and then:

|rl qd−l (r)| = |qd−l

∗
(r)|.

Hence |rl |.|qd−l (r)| = |qd−l

∗
(r)| or:

|rl |.|(r − z1) · · · (r − zd−l )| = |(1 − rz1) · · · (1 − rzd−l )|.

If |zi| = 1 for some i then |r − zi | = |zi |.|r − zi | = |rzi − 1| = |1 − rzi |. Thereby without
loss of generality, we can assume |zi | < 1 for all i. Now we prove the following lemma:
Lemma 2. The following identity holds for all complex numbers r, zi ,

zi |r − zi|2 − |1 − r|2 = (|r|2 − 1)(1 − |zi|2 ).

Proof.

zi |r − zi |2 − |1 − r|2 = (r − zi )(r − zi ) − (1 − rzi )(1 − rzi )

= |r|2 + |zi|2 − 1 − |r|2 .|zi|2
= (|r|2 − 1)(1 − |zi |2).

This completes our proof.

 10

We conclude that: whenever |zi | < 1 the following corollary holds:

Corollary. If |r| > 1 then |r − zi | > |1 − rzi |, and if |r| < 1 then |r − zi | < |1 − rzi |.
Meanwhile, if |r| > 1 then:

|rl |.|(r − z1) · · · (r − zd−l )| > |rl |.|(1 − rz1) · · · (1 − r(zd−l )| > |(1 − rz1 ) · · · (1 − rzd−l )|,

which leads to contradiction.

Moreover, if |r| < 1 then:

|rl |.|(r − z1) · · · (r − zd−l )| < |rl |.|(1 − r(z1) · · · (1 − rzd−l )| < |(1 − rz1 ) · · · (1 − rzd−l )|,

which, again leads to a contradiction. Thus |r| = 1, and we are done!

Remark. Sometimes, using an identity, taking a norm from both sides, and comparing
the order of magnitudes of both sides can help fruitfully to finish the problems concerning
complex numbers.

5. Lagrange’s Interpolation Formula (LIF)

The fourth topic that I prefer to teach is Lagrange’s interpolation Formula (LIF). By
adopting this formula, it becomes possible to determine or characterize the polynomial
P (x) by 1 + degP (x) distinct points. Most students know this formula, they even write
this formula in their papers, and they effortlessly finish their idea. They are partly right!
This is because most of the problems concerning the LIF need another complementary
innovative idea. The readers can find very interesting ideas about the LIF in books like
Andreescu and Dospinescu (2010, 2013) and Andreescu, et al (2020).
The first problem has a solution that only needs insight from rational and irrational
numbers and examining some coefficients. The problems selection committee hadn’t
found it. Their solution was based on the LIF. I bring the first solution from one of the
exam papers and the second solution is based on the discussions during the July 2016.14
out of 74 students solved this problem. Thus, it seems that this problem was hard and
wasnt a good candidate for being the first problem.
Problem 7. (3rd round Iranian mathematical Olympiad, 2017, Final Algebra
exam, Problem 1) Let P (x) be a polynomial with integer coefficients of degree 2016
with no rational roots. Prove that there exists a polynomial Q(x) with integer coefficients
of degree 1395 such that for all distinct roots r, s of P (x), Q(r) − Q(s) is an irrational
number.
Solution. At first, we will prove the following lemma.
Lemma 3. Assume r, s are irrational numbers. Then, at least one of r − s, r2 − s2 is
irrational.
2 2
Proof. Assume the contrary. We find that r + s = rr−s −s
is rational. Therefore 2r =
r + s + r − s is rational. Contradiction. Our proof is complete.
 11

Back to our problem, let Q(x) = x1395 + x1394 + · · · + ax2 + bx. We will specify a, b in
such a way that for all distinct roots r, s of P (x), Q(r) − Q(s) is an irrational number.
First of all, we fix a and increase the value of b. Now, we prove the following lemma.
Lemma 4. If for a fixed a and some (r, s) the value of r − s is irrational, then there is
at most one b such that Q(r) − Q(s) is rational.
Proof. If for b, c the values Qb(r) − Qb (s), Qc (r) − Qc (s) are rational, then we subtract
them to find (b − c)(r − s) is rational. Absurd. This completes our proof.
Hence, for all large enough b and for all (r, s) such that r − s ∈ Q, Qb (r) − Qb (s) is
irrational. Analogously, if r2 − s2 is irrational, for all large enough a and for all (r, s)
such that r2 − s2 ∈ Q, Qa(r) − Qa(s) is irrational. Now, choose a, b large enough and
we are done.
Second solution. In this solution, we need Lemma 1 in the foregoing solution and
another lemma, provided as follows.
Lemma 5. If a polynomial P (x) of degree d assumes rational values in at least d + 1
rational points then, P (x) has rational coefficients.
Proof. Assume that P (r0 ), · · · , P (rd ) are rational for the subset {r0, · · · , rd } of rational
numbers. Then, by Lagrange’s Interpolation Formula (LIF), we find that:

d
X (x − r0 ) · · · (x − ri−1 )(x − ri+1 ) · · · (x − rd )
P (x) = P (ri ) .
i=0 (ri − r0) · · · (ri − ri−1 )(ri − ri+1 ) · · · (ri − rd )

Considering the above formula, it is clear that P (x) has rational coefficients. This
completes our proof.
Back to our problem, consider the polynomial Qa(x) = (x + a)1395, for some a that will
be determined later. Assume that Qa(x) doesnt work. Then for each a there is a pair
(r, s) from the roots of P (x) such that Qa (r) − Qa (s) = (r + a)1395 − (s + a)1395 is an
integer. Since we have only finitely many pairs of (r, s), then there is a pair (r, s) that
S(a) = (r + a)1395 − (s + a)1395 is rational for all but finitely many a. By the above
lemma ( i.e. Lemma 4), S(a) should have rational coefficients. On the other hand, the
coefficients of a1394, a1393 are r − s, r2 − s2 . Whilst, by Lemma 3 both of them can not
be rational. We are done.
The next problem was proposed by me. I wanted to emphasize the fact that the solutions
of the inequality |P (x)| < C for some polynomial P (x) and some real number C are a
subset of an interval of the form (−a, a). At the very first glance, no one thought about
the LIF. However, later, one important property of the LIF seems helpful. That is,
examining the leading coefficients in both sides of the following identity

d
X (x − r0 ) · · · (x − ri− ))(x − ri+1 ) · · · (x − rd )
P (x) = P (ri )
i=0 (ri − r0 ) · · · (ri − ri−1 )(ri − ri+1 ) · · · (ri − rd )
 12

shows that the leading coefficient of P (x) is equal to

d
X P (ri )
.
i=0 ((ri − r0 ) · · · (ri − ri−1 )(ri − ri+1 ) · · · (ri − rd )

So, one can think about establishing a bound for the denominator. This problem was
solved by 6 students out of 72 contestants and all the 6 students won the gold medal.
Problem 8. (3rd round Iranian mathematical Olympiad, 2018, Final Algebra
exam, Problem 4) Let P (x) be a non-constant polynomial with real coefficients. For
all positive real numbers M, prove that there is a positive integer m such that for any
monic polynomial Q(x) of degree greater than or equal to m, the total number of integer
solutions of the inequality
|P (Q(x))| ≤ M
does not exceed degQ(x).
Solution. It is clear that the solutions of the inequality, |P (x)| ≤ M are a subset of an
interval of the form (−a, a) for some positive real number a. Now, assume degQ(x) =
d ≥ m. Consider integers x0 < · · · < xd. Then by Lagrange’s interpolation formula, one
can find that
Xd Y x − xj
Q(x) = Q(xi) .
i=0 x − xj
i6=j i

Since Q(x) is monic, we can find that

d
X Y 1
1= Q(xi) .
i=0 i6=j xi − xj

The right-hand side is less than or equal to

X 1 2d
max |Q(xi)| < . max |Q(xi)|.
i
i i!(d − i)! d! i
Hence,
d! m!
max |Q(xi)| > ≥ .
i 2d 2m
Now, choose m such that !
−m! m!
, ⊆ (−a, a).
2m 2m
We deduce that, from any d + 1 integers, at least one of themsatisfies that inequality
|Q(x)| > 2m!
m . But, all the integer solutions of the inequality |P (Q(x))| = M must satisfy

the inequality |Q(x)| ≤ 2m!

m . We are finally done.

6. Multi-Variable Polynomials (MVPs)

The last topic I selected to discuss is Multi-Variable Polynomials (MVPs). Whenever I
teach this topic, I start with the similarities between MVPs and Single-Variable Polyno-
mials (SVPs). That is, I found that this strategy is more consistent in light of current
 13

knowledge of the trainees. That is, I explicitly define MVPs and extend upon the analo-
gies. Then, after one or two sessions I start to discuss on disanalogies between the MVPs
and SVPs. Proceeding by this way, students find a better attitude toward and a better
understanding of the MVPs.
Based on the above strategy, I concentrate more on the notion of homogeneous polynomi-
als and representing an MVP as the sum of its homogeneous parts of different degrees.
This was my motivation to write an article about homogeneity (Safaei, 2018). This
problem needs a very basic idea of the growth rate of MVPs. That is, in the SVPs, if
P (x) = a0 + a1 x + · · · + adxd , then for each  > 0, (ad − )xd < P (x) < (ad + )xd for
all but finitely many real x. In this problem, I outline the same idea for the MVPs. 16
students out of 76 students completely solved this problem.
Problem 9. (3rd round Iranian mathematical Olympiad, 2017, Algebra exam,
Midterm, Problem 3) Do there exist infinitely many points (x1, y1 ), (x2, y2), . . . in the
Cartesian plane such that for any sequence b1, b2, . . . of real numbers there exists a
polynomial P (x, y) with real coefficients such that for each i, P (xi , yi) = bi ?
P
Solution. Considering P (x, y) = di=0 Pd (x, y), where Pk (x, y) are homogeneous polyno-
mials of degree k. Then, for all sufficiently large values of |x|, |y|, P (x, y) < C(|x|+|y|+1)d
for some constant C. Now, consider the sequence bi = i(|xi | + |yi | + 1)d . Then, for all
large enough i, the equation P (xi , yi ) = bi = i(|xi | + |yi | + 1)i has only finitely many
solutions.
Remark. One can also consider the sequence bi = (|xi| + |yi | + 1)i .
Remark. At the outset, this problem looks like the Lagrange Interpolation Formula
(LIF) for multi-variable polynomials. That is, assume degP (x, y) = d. Then, the poly-
nomial P (x, y) has d(d+1) 2
unknown coefficients. Assume we have two sets {x0, . . . , xd },
{y0, . . . , yd }. Then, if we know the value of P (xi, yj ) for all 0 = i, j, i + j = d, then, the
polynomial P (x, y) could uniquely be determined as follows:
X (x − x0 ) · · · (x − xi−1 ) (y − y0 ) · · · (y − yj−1 )
P (x, y) = P (xi , yj ) . .
0≤i,j,i+j≤d (xi − x0 ) · · · (xi − xi−1 ) (yj − y0 ) · · · (yj − yj−1 )

However, it is a great achievement to determine a multivariable polynomial through this,

but we can not find a good idea to finish the above-mentioned problem.

7. Concluding Remarks
In this article, I outlined a framework about the important teaching elements of poly-
nomials. For this reason, I used some problems from recent Iranian Mathematical
Olympiads. As has been seen, these problems are challenging but interesting. Pro-
ceeding in this way, readers can find different approaches implemented to solve those
problems, some notes and remarks about the number of complete solutions, and some
notes to expose the weakness of other potential arguments. Further, I introduced some
resources for further reading.
 14

Teaching polynomials needs a priori knowledge from multiple sources. In some degree,
it is akin to teaching Combinatorics or advanced Number Theory. Furthermore, and
quite generally, it also has some degree of idiosyncrasy. Since, as has been previously
shown, it needs insight from the calculation and algebraic expressions. Rather than
being a problem, it is in fact, a great advantage for those who want to solve challenging
problems in Algebra.

8. References
T. Andreescu and G. Dospinescu, Problems from the book, 2nd edition, XYZ-Press,
2010.
T. Andreescu and G. Dospinescu, Straight from the book, XYZ-Press, 2012.
T. Andreescu, N. Safaei, and A. Ventullo, 117 polynomial problems from the Awe-
someMath summer program, XYZ-Press, 2019.
T. Andreescu, N. Safaei, and A. Ventullo, Awesome Polynomials for Mathematics Com-
petitions, XYZ-Press, 2019.
N. Safaei, Discrete Approach to a Result Concerning a Contour Integral, Mathematical
Reflections, 2, 2019.
N. Safaei, Problem M2544 problem, Kvant, 1, 2019.
N. Safaei, Searching for Homogeneity Across Multi-Variable Polynomials, Mathematical
Reflections, 1, 2018.

View publication stats