0% found this document useful (0 votes)
28 views16 pages

Real Polynomials Winter2024

Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
28 views16 pages

Real Polynomials Winter2024

Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 16

2024 Winter Camp Real Polynomials Victor Rong

Real Polynomials
Victor Rong
January 5, 2024

Polynomials are one of the most ubiquitous mathematical objects, and as such, appear frequently
in math olympiads. In this handout, we focus on polynomials with real coefficients and techniques
to handle them. This is in contrast to integer polynomials which come with additional number-
theoretic structures and ideas, though many of the fundamentals are the same. For practice with
integer polynomials, I highly recommend Yufei Zhao’s Integer Polynomials handout.

1 The Essentials
We generally analyze polynomials in two forms:
• As a linear combination of powers xi
• By the roots ri
The Fundamental Theorem of Algebra tells us that a polynomial has roots and factors into them,
while Vieta’s Formulas give relations between the roots and the coefficients. It is also important
to always consider the degree of the polynomials. If we are able to reduce the degree, for example,
we can obtain contradictions based on the number of roots.

Theorem (Fundamental Theorem of Algebra). Let P (x) be a complex polynomial of degree n ≥


1. Then PQ(x) has exactly n complex roots r1 , r2 , . . . , rn , counting with multiplicity. Furthermore,
P (x) = c ni=1 (x − ri ) for some constant c.

Lemma 1.0.1. Let P (x) be a polynomial with real coefficients. Then its non-real roots form
pairs of complex conjugates.

Theorem (Vieta’s Formulas). Let polynomial P (x) = cn xn + cn−1 xn−1 + . . . + c0 have roots
r1 , r2 , . . . , rn . Then we have
cn−1
r1 + r2 + · · · + rn = − ,
cn
cn−2
r1 r2 + r1 r3 + · · · + rn−1 rn = ,
cn
..
.
c0
r1 r2 · · · rn = (−1)n .
cn

Proof. We can obtain these equations by expanding


P (x) = cn (x − r1 )(x − r2 ) · · · (x − rn )
and comparing coefficients.

Additionally, the real polynomials form a Euclidean domain, meaning that the greatest common
divisor can be defined and satisfies certain properties.

1 of 16
2024 Winter Camp Real Polynomials Victor Rong

Definition. For two real polynomials P (x), Q(x), the greatest common divisor D(x) := gcd(P, Q)
is defined as a real polynomial of largest degree which divides both P (x) and Q(x). Additionally,
deg D ≤ min(deg P, deg Q).

Remark. Note that the D can be explicitly written based on the shared roots of P, Q, and is unique
up to a scalar multiplication.

Theorem (Bezout’s Identity). Let P (x) be a real polynomial of degree n and Q(x) be a real
polynomial of degree m. Then there exist real polynomials A(x), B(x) such that A(x)P (x) +
B(x)Q(x) = D(x), where D(x) = gcd(P, Q).

Lemma 1.0.2. Let P (x) and Q(x) be real polynomials which are relatively prime. We have

P (x) | Q(x)R(x) =⇒ P (x) | R(x).

Proof. We have P S = QR for some polynomial S. By Bezout’s Identity, we also have AP +BQ = 1.
Then

P S = QR
BP S = BQR
BP S = (1 − AP )R
(BS + AR)P = R.

So P | R.

2 Handling Conditions
Many problems ask for polynomials satisfying given conditions. In this case, it’s always a good idea
to consider the degree as it constrains how many conditions the polynomial can reasonably satisfy.
The next lemma makes this rigorous.

2.1 Intersecting Polynomials

Lemma 2.1.1. Let P (x) and Q(x) be polynomials which intersect at points x1 , x2 , . . ., xk . That
is, P (xi ) = Q(xi ) for i = 1, 2, . . . , k. Then
k
Y
P (x) − Q(x) = A(x) (x − xi ) = A(x)(x − x1 )(x − x2 ) . . . (x − xk )
i=1

for some polynomial A(x) with deg A = deg(P − Q) − k. If k > deg P and k > deg Q, P ≡ Q.

This fact is quite useful. One particular application is moving from results about integer or discrete
inputs to all numbers. For example, if we determine that P (n) = n for all integers n, then we can
conclude that P (x) = x for all real numbers x.

2 of 16
2024 Winter Camp Real Polynomials Victor Rong

k
Example 1. A polynomial P (x) of degree 10 satisfies P (k) = k+1 for
k = 0, 1, . . . , 10. Find P (11).

Proof. If the right hand side were some polynomial expression in terms of k instead, then the
problem would be trivial. With this in mind, we transform the problem to force the equation to
be a polynomial. Let Q(x) = (x + 1)P (x) − x. This is clearly a polynomial of degree 11. Then the
given equations now become
Q(k) = 0
for k = 0, 1, . . . , 10. So Q(x) = ax(x − 1)(x − 2) . . . (x − 10) for some constant a. To determine
this constant, we need one more point on Q. We can find this point by plugging x = −1 into
Q(x) = (x + 1)P (x) − x since this zeros out the unknown P (x) term. Thus,

Q(−1) = 1 = a(−1)(−2)(−3) . . . (−11).


−1
a= .
11!

Finally, we can compute Q(11). We obtain Q(11) = −1, and so P (11) = 65 .

2.2 Finite Differences


If the inputs happen to be at fixed intervals, we can use the method of finite differences.

Lemma 2.2.1. For any polynomial P (x) of degree k ≥ 1 and constant a, P (x + a) − P (x) is a
polynomial of degree k − 1.

Directly applying Lemma 2.2.1 to P taken at fixed points allows us to recursively form a table of
differences.

Theorem (Method of Finite Differences). Let P (x) = cn xn + . . . + c0 . Say we know the values
of P (1), P (2), . . . , P (m) for m ≥ n + 1. Define the difference table so that Dr,j = P (j) for r = 0
and Dr,j = Dr−1,j+1 − Dr−1,j for r ≥ 1. Then row r’s values correspond to a polynomial of
degree n − r. In particular, Dn,j = n!cn and Dn+1,j = 0 for all j.

Example 2 (HMMT 2015). The complex numbers x, y, z satisfy

xyz = −4
(x + 1)(y + 1)(z + 1) = 7
(x + 2)(y + 2)(z + 2) = −3.

Find, with proof, the value of (x + 3)(y + 3)(z + 3).

3 of 16
2024 Winter Camp Real Polynomials Victor Rong

Solution. Let P (t) = (t + x)(t + y)(t + z). Then we know t is a monic cubic such that P (0) =
−4, P (1) = 7, P (2) = −3 and we wish to compute P (3). Forming the difference table, we can
immediately fill in the 6 = 3! in row 3 as we know P ’s leading coefficient:

Row 0 -4 7 -3 ?
Row 1 11 -10 ? -
Row 2 -21 ? - -
Row 3 6 - - -

Then we can work our way up to get

Row 0 -4 7 -3 -28
Row 1 11 -10 -25 -
Row 2 -21 -15 - -
Row 3 6 - - -

Thus, (x + 3)(y + 3)(z + 3) = −28.

A more clever application of Lemma 2.2.1 yields the following result:

Theorem (Prouhet-Tarry-Escott). Let k be some natural number. There exists e1 , e2 , . . . , e2k ∈


{−1, 1} such that for any polynomial P (x) of degree < k,

e1 P (1) + e2 P (2) + . . . + e2k P (2k ) = 0.

Proof. The base case for k = 1 is obvious. We will show the induction step from k − 1 to k. Let
Q(x) = P (x) − P (x + 2k−1 ). By Lemma 2.2.1, Q(x) is a polynomial of degree < k − 1. By the
k−1
induction hypothesis for k−1, there exists e1 , e2 , . . . , e2k−1 such that e1 Q(1)+. . .+e2 Q(2k−1 ) = 0.
Then
k−1
e1 Q(1) + . . . + e2 Q(2k−1 ) = 0
k−1
e1 (P (1) − P (2k−1 + 1)) + . . . + e2 e2k−1 (P (2k−1 ) − P (2k )) = 0
e1 P (1) + . . . + e2k−1 P (2k−1 ) − e1 P (2k−1 + 1) − . . . − e2k P (2k ) = 0.

So this new sequence works for k as desired. By induction, we are done.

Remark. The sequence can be defined explicitly and is known as the Thue-Morse sequence.

4 of 16
2024 Winter Camp Real Polynomials Victor Rong

2.3 Lagrange Interpolation

Theorem (Lagrange Interpolation). Given n points (xi , yi ) for i = 1, . . . n and xi pairwise dis-
tinct, there is a unique polynomial L(x) of degree ≤ n − 1 such that P (xi ) = yi for i = 1, . . . , n.
This polynomial can be written as
n
X Y x − xj
L(x) = yi .
xi − xj
i=1 j̸=i

Proof. Let us first check that the given polynomial does indeed satisfy the constraints. For any
1 ≤ k ≤ n, we have
n
X Y xk − xj
L(xk ) = yi
xi − xj
i=1 j̸=i
Y xk − xj X Y xk − xj
= yk + yi
xk − xj xi − xj
j̸=k i̸=k j̸=i
X xk − xk Y xk − xj
= yk · 1 + yi · ·
xi − xj xi − xj
i̸=k j̸=i,k

= yk .

So this is indeed an interpolating polynomial, and the degree is clearly ≤ n − 1. It remains to prove
that L is the unique interpolating polynomial of degree ≤ n−1. This is immediate by Lemma 2.1.1,
so we are done.

Remark. If L(x) is the Lagrange interpolating polynomial, then the interpolating polynomials (of
any degree) are exactly L(x) + A(x) ni=1 (x − xi ) for A(x) any polynomial.
Q

Example 3 (PFTB). Let a1 , a2 . . . an be distinct integers. Prove that for any natural number k
Pn aki
the value Q is also an integer.
i=1 j̸=i (ai − aj )

Proof. Let P (x) = xk . Let L(x) be the Lagrange interpolating polynomial


n
X Y x − aj
L(x) = aki .
ai − aj
i=1 j̸=i

The value n−1 in L(x). We know that P (x) = L(x) +


Qn in the problem is exactly the coefficient of x
A(x) i=1 (x − ai ) where A(x) is some Q polynomial. Note that A(x) is an integer polynomial as
is an integer polynomial and ni=1 (x − ai ) is a monic integer polynomial. As L(x) = P (x) −
P (x) Q
A(x) ni=1 (x − ai ) ∈ Z[x], its coefficients are integers so we are done.

Remark. If k < n, the proof above tells us that the value in question is exactly 0.

5 of 16
2024 Winter Camp Real Polynomials Victor Rong

2.4 Algebraic Conditions


These tend to be trickier to handle, and the method depends very much on the exact problem.
Here are some general tips
• A lot of the tips for FEs apply! Things like surjectivity/injectivity aren’t (usually) relevant,
but you should try to
– Figure out the solution set first
– Try to show the polynomial is even/odd (if the solution set reflects this)
– Plug things in to set terms to zero or cancel terms
• Extend the condition set if necessary. For example, if the problem has a condition for all
x, y, z such that x, y, z are positive, you should be able to show that it’s true for all x, y, z
complex.
• Reduce the condition to a monovariate problem if it’s multivariate. This may involve treating
all but one of the variables as constants and interpreting the problem as a single-variable
polynomial. Then you can use the typical ideas for handling monovariate polynomials.
Let’s take a look at a recent USAMO P6.

Example 4 (USAMO 2019). Find all polynomials P with real coefficients such that

P (x) P (y) P (z)


+ + = P (x − y) + P (y − z) + P (z − x)
yz zx xy
holds for all nonzero real numbers x, y, z satisfying 2xyz = x + y + z.
Solution. Define the multivariate polynomial
F (x, y, z) = xP (x) + yP (y) + zP (z) − xyz (P (x − y) + P (y − z) + P (z − x)) .
Let d be the degree of x in F . Note that (2yz − 1)x ≡ y + z mod (2xyz − x − y − z). Then
(2yz − 1)d F (x, y, z) = (2xyz − x − y − z)G(x, y, z) + H(y, z)
for some polynomial G(x, y, z) and H(y, z). Then we can choose any y, z such that y, z ̸= 0,
y + z ̸= 0, and yz ̸= 12 , and compute the corresponding x such that 2xyz = x + y + z to obtain
H(y, z) = 0. As H is 0 over a non-empty open set, we can conclude that H is identically 0.
Furthermore, 2yz − 1 and 2xyz − x − y − z are relatively prime, so we can conclude that for some
G2 (x, y, z),
F (x, y, z) = (2xyz − x − y − z)G2 (x, y, z).
Now, we can use any x, y, z ∈ C such that 2xyz = x + y + z. Plugging in (x, −x, 0) gives xP (x) −
xP (−x) = 0 so P is even. Now let us plug in (x, √i2 , − √i2 ). We have

i

i

i

i

x
 
i
 
i
 √ 
xP (x) + √ P √ − √ P −√ = P x− √ + P −x − √ +P 2i
2 2 2 2 2 2 2
x
 
i
 
i
 √ 
xP (x) = P x− √ +P x+ √ +P 2i
2 2 2

i
 
i
 √ 
P x− √ − 2P (x) + P x + √ = −P 2i .
2 2
The LHS is a second-order finite difference, so we can conclude that deg P ≤ 2. Using P (x) = ax2 +b
gives P (x) = a(x2 + 3) which can be checked.

6 of 16
2024 Winter Camp Real Polynomials Victor Rong

3 Popular Polynomials
3.1 Sum of Powers

Theorem (Faulhaber’s Formula). Let k be a non-negative integer. Prove that there is some
rational polynomial Pk (x) of degree k + 1 such that for all natural numbers n,
n
X
Pk (n) = jk.
j=1

Proof. We will use strong induction. The base case of k = 0 is obvious. Assume we have proved
that there exists such rational polynomial Pj (x) for all 0 ≤ j < k. For any natural number r, let
Qr (x) = x(x−1)···(x−r+1) . Note that for natural numbers n, Qr (n) = nr . We have

r!
n
X
Qk+1 (n + 1) = Qk (j).
j=1

But we can also write Qk (j) as a degree k polynomial Qk (j) = ck j k + ck−1 j k−1 + . . . + c0 where
the coefficients are rational. Thus,
n X
X k
Qk+1 (n + 1) = ci j i
j=1 i=0
k
X n
X
= ci ji
i=0 j=1
k−1 n
!
X X
= ci Pj (n) + ck ji
i=0 j=1
n k−1
!
X Qk+1 (n + 1) X ci
ji = − Pj (n) .
ck ck
j=1 i=0

It is easy to check that the RHS of the final line is indeed a polynomial of degree k + 1. So by
induction, we are done.

Remark. Another similar proof by induction considers the sum nj=1 j k+1 − (j − 1)k+1 .
P

In other words, the sum of a polynomial at spaced intervals creates a new polynomial with degree
one higher.

Lemma 3.1.1. For any polynomial Q(x) of degree k and constants a, b, there is some polynomial
P of degree k + 1 such that
Xn
P (n) = Q(aj + b).
j=1

Remark. In a sense, this is the reverse of finite differences. Instead of lowering the degree by one,
it raises it.

7 of 16
2024 Winter Camp Real Polynomials Victor Rong

3.2 Chebyshev Polynomials

Theorem (Chebyshev Polynomials). For n ∈ N, there is a unique polynomial Tn (x) such that
Tn (cos θ) = cos(nθ) for any θ. Furthermore, Tn (x) is a degree n integer polynomial with leading
coefficient 2n−1 .

Proof. By the trigonometric identity cos(nθ) + cos((n − 2)θ) = 2 cos θ cos((n − 1)θ), we have

Tn (x) = 2xTn−1 (x) − Tn−2 (x),

and T0 (x) = 1, T1 (x) = x. The fact that Tn (x) is a degree n integer polynomial with leading
coefficient 2n−1 follows in a straightforward manner from this recurrence.

Remark. The roots of the Chebyshev polynomials, cos 2k−1



2n π , are good guesses for solution sets
in [0, 1]n .

4 The Geometry of Polynomials


Polynomials are very well-behaved, being both continuous and differentiable. These properties
allow us to deduce information about the roots of a polynomial even if we do not explicitly know
their locations.

4.1 Continuity

Theorem (Intermediate Value Theorem). If real-valued function f is continuous over [a, b], then
it takes on any given value between f (a) and f (b) at some point within the interval.

Lemma 4.1.1. If P (x) is a polynomial with real coefficients and has odd degree, it has at least
one real root.

Proof. Without loss of generality, consider P (x) monic. As x → ∞, P (x) → ∞ and as x → −∞,
P (x) → −∞. So P (x) must be 0 for some x.

Remark. This also follows from Lemma 1.0.1.

Example 5 (Minimality of Chebyshev Polynomials). Let P (x) be a monic polynomial of degree


1
≤ n. Then maxx∈[−1,1] P (x) ≥ 2n−1 , with equality when P (x) = T2nn−1
(x)
.

Proof. Assume for the sake of contradiction that we have monic polynomial P (x) of degree n such
1
that maxx∈[−1,1] P (x) < 2n−1 . Then define Q(x) = T2nn−1
(x)
−P (x). Note that Q(x) has degree ≤ n−1.
2kπ
For x = cos n , 0 ≤ 2k ≤ n, we have

1
Q(x) = − P (x) > 0.
2n−1

8 of 16
2024 Winter Camp Real Polynomials Victor Rong

For x = cos (2k+1)π


n , 0 ≤ 2k + 1 ≤ n, we have

1
Q(x) = − − P (x) < 0.
2n−1
By the Intermediate Value Theorem, Q must have n roots. But Q(x) has degree ≤ n − 1 and
Q(x) ̸= 0, so contradiction.
Tn (x)
Remark. In fact, equality holds iff P (x) = 2n−1
.

4.2 Derivatives
In the previous example, it worked out well that Q(x) was either strictly positive or negative at the
locations we chose. What if we have a polynomial P such that P (−1) = 1, P (0) = 0, and P (1) = 1.
What can we say about the number of roots (counting multiplicity) that P has in (−1, 1)? Edge
cases like these can be handled by using the derivative. In particular, the derivative is effective at
analyzing repeated roots.

Definition. For a differentiable function f (x), its derivative f ′ (x) denotes the slope of the tangent
to f at x.

Remark. The optima of a differentiable function correspond to the zeros of its derivative (but its
derivative may have zeros that are not optima).

Lemma 4.2.1 (Derivative Rules). We have the following rules for taking derivatives:
• Constant Rule: The derivative of a constant is 0.
• Product by a Constant Rule: (cf (x))′ = cf ′ (x)
• Sum Rule: (f (x) + g(x))′ = f ′ (x) + g ′ (x)
• Product Rule: (f (x)g(x))′ = f ′ (x)g(x) + f (x)g ′ (x)
• Power Rule: (xn )′ = nxn−1
• Chain Rule: (f (g(x)))′ = f ′ (g(x))g ′ (x)

We won’t go over any rigorous proof or background for these, but it may be helpful to generally
know where they come from. Without going into details, the recipe for derivatives is to consider
f (x+h)−f (x)
h and look for the terms independent of h. The handwavy justification is that h is
infinitesimally small and so any terms with h go to 0. For example, let’s look at the power rule
when n is a natural number. By the binomial formula, we have
Pn n n−i i 
(x + h)n − xn i=0 i x h − xn
=
h h
nxn−1 h + n2 xn−2 h2 + . . . + hn
=
  h
n n−2
= nxn−1 + x h + . . . + hn−1
2
= nxn−1 + O(h).

9 of 16
2024 Winter Camp Real Polynomials Victor Rong

Example 6. What is the derivative of (x4 + 3x + 1)2 ?

Proof. Let f (x) = x2 and g(x) = x4 + 3x + 1. The power rule tells us that f ′ (x) = 2x and
g ′ (x) = 4x3 + 3. So by the chain rule, we have

(f (g(x))′ = f ′ (g(x))g ′ (x) = 2(x4 + 3x + 1)(4x3 + 3).

Alternatively, we could have expanded (x4 + 3x + 1)2 to individual powers and applied the power
rule to obtain the same result, although it would not be factorized for free.

Lemma 4.2.2 (Derivative of a Polynomial). Let non-constant polynomial P (x) = cn xn +


cn−1 xn−1 + . . . + c0 have roots r1 , r2 , . . . , rn . Then we have two formulas for its derivative:
n−1
X

P (x) = (i + 1)ci+1 xi
i=0

and
n Y n
!

X X 1
P (x) = cn (x − rj ) = P (x) .
x − ri
i=1 j̸=i i=1

Note in particular that the degree of P ′ (x) is always one less than the degree of P (x) for P
non-constant.

Lemma 4.2.3 (Repeated Roots). Let polynomial P (x) have root r with multiplicity k ≥ 1.
Then r is a root of P ′ (x) with multiplicity k − 1 (or if k = 1, r is not a root of P ′ (x)).

Proof. As P (x) has root r with multiplicity k, we can write P (x) = (x−r)k Q(x) for some polynomial
Q(x) which does not have r as a root. Then by the product rule,

P ′ (x) = k(x − r)k−1 Q(x) + (x − r)k Q′ (x) = (x − r)k−1 kQ(x) + (x − r)Q′ (x) .


Note that plugging in x = r to kQ(x) + (x − r)Q′ (x) gives kQ(r) ̸= 0, so we see that the multiplicity
of r in P ′ (x) is exactly k − 1.

The derivative of a polynomial can be used to analyze a polynomial’s repeated roots using the
above lemma. We will see how this is done in the following example.

Example 7 (Putnam 1956). The nonconstant polynomials P (z) and Q(z) with complex coef-
ficients have the same set of numbers for their zeros but possibly different multiplicities. The
same is true of the polynomials P (z) + 1 and Q(z) + 1. Prove that P (z) = Q(z).

Proof. Assume for the sake of contradiction that P ̸= Q. Let P have degree n and Q have degree
m. Without loss of generality, n ≥ m. Assume that P (z) has j distinct roots and P (z) + 1 has k
distinct roots. Note that the set of roots of P (z) is disjoint with the set of roots of P (z) + 1. Since

P (z) − Q(z) = (P (z) + 1) − (Q(z) + 1),

10 of 16
2024 Winter Camp Real Polynomials Victor Rong

note that P (z) − Q(z) has at least j + k roots. P (z) − Q(z) is not the zero polynomial and has
degree ≤ n, so j + k ≤ n. However, consider P ′ (z) = (P (z) + 1)′ . Lemma 4.2.3 tells us that P ′ (z)
shares at least n − j roots (counted with multiplicity) with P (z). It also tells us that P ′ (z) shares
at least n − k roots with P (z) + 1. Again, these two sets are disjoint. So P ′ (z) has at least 2n − j − k
roots. But deg P ′ = n − 1, so n − 1 ≥ 2n − j − k =⇒ j + k ≥ n + 1. So there is a contradiction.

Theorem (Gauss-Lucas). If P is a non-constant polynomial with complex coefficients, all zeroes


of P ′ belong to the convex hull of the set of zeros of P .

′ (x)
Proof. Let us write P (x) = c ni=1 (x − ri ). From Lemma 4.2.2, we have PP (x) = ni=1 x−r
1
Q P
i
. Let z
be a root of P ′ . If z is also a root of P , it is trivially within the convex hull. Otherwise, we have

P ′ (z)
0=
P (z)
n
X 1
=
z − ri
i=1
n
X z − ri
=
|z − ri |2
i=1
n n
X z X ri
=
|z − ri |2 |z − ri |2
i=1 i=1
n
X |z − ri |−2
z= ri Pn −2
.
i=1 j=1 |z − rj |

So z can indeed be written as a convex combination of the roots of P .

Example 8 (Putnam 2018). Let n be a positive integer, and let fn (z) = n + (n − 1)z + (n −
2)z 2 + · · · + z n−1 . Prove that fn has no roots in the closed unit disk {z ∈ C : |z| ≤ 1}.

Proof. Let gn (z) = z n +z n−1 +. . .+1. Note that fn (z) strongly resembles gn′ (z) except the coefficients
appear to be reversed. We can handle this reverse by considering fn ( z1 ). In particular, we have

1
gn′ (z) = nz n−1 + (n − 1)z n−2 + . . . + 1 = z n−1 fn ( ).
z
Note that z = 0 is clearly not a root of fn . So for z ̸= 0, z1 is a root of fn iff z is a root of gn′ .
By Gauss-Lucas, the roots of gn′ are contained in the convex hull of the roots of gn . But observe
that the roots of gn roots of unity, and so lie on the closed unit disk. Furthermore, gn has no
double-roots, so gn′ does not share any roots. Thus, all the roots of gn′ lie in the open unit disk,
and so all the roots of fn lie outside the closed unit disk.

Remark. This problem can also be solved without using Gauss-Lucas, but the above gives a deeper
reason for why it’s true.

11 of 16
2024 Winter Camp Real Polynomials Victor Rong

4.3 Interlacing
The location of roots of P also gives information on the location of roots of P ′ . This is called
interlacing, as the graphs of P and P ′ appear to be interlaced. To end off this handout, we will use
interlacing to prove Descartes’ Rule of Signs.

Theorem (Rolle’s Theorem). If real-valued function f is continuous over [a, b], differentiable
over (a, b), and f (a) = f (b), then there exists at least one c in (a, b) such that f ′ (c) = 0.

From Rolle’s Theorem and Lemma 4.2.3, the idea of interlacing follows.

Lemma 4.3.1 (Interlacing). If f (x) is a differentiable function in interval (a, b) and f (x) = 0
has k roots (counting multiplicity) in (a, b) then f ′ (x) = 0 has at least k − 1 roots in (a, b).

Theorem (Descartes’ Rule of Signs). Consider polynomial P (x) = ki=1 ci xpi with 0 ≤ p1 <
P
p2 + . . . + pk and non-zero real coefficients ci . Let V (P ) be the number of sign changes in the
coefficients of P , i.e. where ci ci+1 < 0. Let Z(P ) be the number of strictly positive roots of
P (x). Then Z(P ) ≤ V (P ) and V (P ) − Z(P ) is even.

Proof. We only need to consider p1 = 0 as dividing by xp1 does not change Z(P ) or V (P ). Let
us first prove that Z(P ) and V (P ) share the same parity. Note that P (0) = c0 and sgn (P (∞)) =
sgn(ck ). If c0 > 0 and ck < 0 for example, there are an odd number of positive roots and there an
odd number of sign changes from c0 to ck . Other cases are similar.

Now we will show that Z(P ) ≤ V (P ) by inducting over the degree. The base case is true for
constant polynomials. Assume that Descartes’ Rule of Signs is true for polynomials of degree
≤ deg(P ) − 1. As deg(P ′ ) ≤ deg(P ) − 1, we can assume the inductive hypothesis with P ′ . This
tells us that V (P ′ ) = Z(P ′ ) + 2s for some non-negative integer s. By Lemma 4.3.1 on interval
(0, ∞), we have Z(P ′ ) ≥ Z(P ) − 1. Additionally, we have that
k
X
P ′ (x) = ci pi xpi −1 ,
i=1

so V (P ′ ) will either remain the same or be one less than V (P ). Thus, we have

Z(P ) ≤ Z(P ′ ) + 1 ≤ V (P ′ ) − 2s + 1 ≤ V (P ) − 2s + 1 ≤ V (P ) + 1.

As Z(P ) and V (P ) are the same parity, we can conclude that Z(P ) ≤ V (P ). So we are done by
induction.

12 of 16
2024 Winter Camp Real Polynomials Victor Rong

A Problems
A1. (HMMT 2017) Let P (x), Q(x) be nonconstant polynomials with real coefficients. Prove that
if
⌊P (y)⌋ = ⌊Q(y)⌋
for all real numbers y, then P (x) = Q(x) for all real numbers x.
A2. (CMO 2013) Determine all polynomials P (x) with real coefficients such that

(x + 1)P (x − 1) − (x − 1)P (x)

is a constant polynomial.
A3. (Spain 2002) Find all polynomials P (t) of one variable that fulfill the following for all x, y ∈ R:

P (x2 − y 2 ) = P (x − y)P (x + y).

A4. (USAMO 2014) Let a, b, c, d be real numbers such that b − d ≥ 5 and all zeros x1 , x2 , x3 , and
x4 of the polynomial P (x) = x4 + ax3 + bx2 + cx + d are real. Find the smallest value the product
(x21 + 1)(x22 + 1)(x23 + 1)(x24 + 1) can take.
A5. (Unknown) A polynomial of p of degree n satisfies p(k) = 2k for all 0 ≤ k ≤ n. Find p(n + 1).
A6. (Putnam 2023) Let n be an even positive integer. Let p be a monic, real polynomial of degree
2n; that is to say, p(x) = x2n + a2n−1 x2n−1 + · · · + a1 x + a0 for some real coefficients a0 , . . . , a2n−1 .
Suppose that p(1/k) = k 2 for all integers k such that 1 ≤ |k| ≤ n. Find all other real numbers x
for which p(1/x) = x2 .
A7. (APMO 2009) Let a1 , a2 , a3 , a4 , a5 be real numbers satisfying the following equations:
a1 a2 a3 a4 a5 1
+ 2 + 2 + 2 + 2 = 2
k2 +1 k +2 k +3 k +4 k +5 k
a1 a2 a3 a4 a5
for k = 1, 2, 3, 4, 5. Find the value of 37 + 38 + 39 + 40 + 41 .
A8. (Serbia 2013) Let P (x) be a polynomial of degree 2012 with real coefficients satisfying the
condition
P (a)3 + P (b)3 + P (c)3 ≥ 3P (a)P (b)P (c),
for all real numbers a, b, c such that a + b + c = 0. Is it possible for P (x) to have exactly 2012
distinct real roots?
A9. (COMC 2014) A polynomial f (x) with real coefficients is said to be a sum of squares if there
are polynomials p1 (x), p2 (x), . . . , pn (x) with real coefficients for which

f (x) = p21 (x) + p22 (x) + . . . + p2n (x).

Suppose f (x) is a sum of squares. Prove there are polynomials u(x), v(x) with real coefficients
such that f (x) = u2 (x) + v 2 (x).
A10. (Putnam 2014) Show that for each positive integer n, all the roots of the polynomial
n
X
2k(n−k) xk
k=0

are real numbers.

13 of 16
2024 Winter Camp Real Polynomials Victor Rong

A11. (Unknown) Let P (x) be a polynomial of degree n with real coefficients such that |f (x)| ≤ 1
for all x ∈ [0, 1]. Show that |f (− n1 )| ≤ 2n+1 − 1.
A12. (Serbia 2023) Let p be a prime and P ∈ R[x] be a polynomial of degree less than p − 1 such
that |P (1)| = |P (2)| = . . . = |P (p)|. Prove that P is constant.
A13. (Putnam 2019) Given real numbers b0 , b1 , . . . , b2019 with b2019 ̸= 0, let z1 , z2 , . . . , z2019 be the
roots in the complex plane of the polynomial
2019
X
P (z) = bk z k .
k=0

Let µ = (|z1 | + · · · + |z2019 |)/2019 be the average of the distances from z1 , z2 , . . . , z2019 to the origin.
Determine the largest constant M such that µ ≥ M for all choices of b0 , b1 , . . . , b2019 that satisfy

1 ≤ b0 < b1 < b2 < · · · < b2019 ≤ 2019.

B Problems
B1. (Serbia 2018) For any positive integer n, define

cn = min |z1 · 12018 + z2 · 22018 + ... + zn · n2018 |.


(z1 ,z2 ,...,zn )∈{−1,1}n

Is the sequence (cn )n∈Z+ bounded?


B2. (CMO 2018) Find all polynomials p(x) with real coefficients that have the following property:
There exists a polynomial q(x) with real coefficients such that

p(1) + p(2) + p(3) + . . . + p(n) = p(n)q(n)

for all positive integers n.


B3. (Olympic Revenge 2020) Let n be a positive integer and a1 , a2 , . . . , an non-zero real numbers.
What is the least number of non-zero coefficients that the polynomial P (x) = (x−a1 )(x−a2 ) · · · (x−
an ) can have?
B4. (Kürschák 2016) If p, q ∈ R[x] satisfy p(p(x)) = q(x)2 , does it follow that p(x) = r(x)2 for
some r ∈ R[x]?
B5. (ISL 2019) Let x1 , x2 , . . . , xn be different real numbers. Prove that
X Y 1 − xi xj 
0, if n is even;
=
xi − xj 1, if n is odd.
1⩽i⩽n j̸=i

B6. (China 1995) A and B play the following game with a polynomial of degree at least 4:

x2n + □x2n−1 + □x2n−2 + . . . + □x + 1 = 0

A and B take turns to fill in one of the blanks with a real number until all the blanks are filled
up. If the resulting polynomial has no real roots, A wins. Otherwise, B wins. If A begins, which
player has a winning strategy?

14 of 16
2024 Winter Camp Real Polynomials Victor Rong

B7. (RMM 2018) Determine whether there exist non-constant polynomials P (x) and Q(x) with
real coefficients satisfying
P (x)10 + P (x)9 = Q(x)21 + Q(x)20 .

B8. (ELMOSL 2012) Prove that any polynomial of the form 1 + an xn + an+1 xn+1 + · · · + ak xk
(k ≥ n) has at least n − 2 non-real roots (counting multiplicity), where the ai (n ≤ i ≤ k) are real
and ak ̸= 0.
B9. (IMO 2016) The equation

(x − 1)(x − 2) · · · (x − 2016) = (x − 1)(x − 2) · · · (x − 2016)

is written on the board, with 2016 linear factors on each side. What is the least possible value of
k for which it is possible to erase exactly k of these 4032 linear factors so that at least one factor
remains on each side and the resulting equation has no real solutions?
B10. (ISL 2020) A magician intends to perform the following trick. She announces a positive
integer n, along with 2n real numbers x1 < · · · < x2n , to the audience. A member of the audience
then secretly chooses a polynomial P (x) of degree n with real coefficients, computes the 2n values
P (x1 ), . . . , P (x2n ), and writes down these 2n values on the blackboard in non-decreasing order.
After that the magician announces the secret polynomial to the audience. Can the magician find
a strategy to perform such a trick?
B11. (Russia 1997) Let a1 , a2 , . . . , am be nonzero real numbers satisfying

1k · a1 + 2k · a2 + · · · + mk · am = 0

for each k = 0, 1, . . . , n. Prove that the sequence a1 , a2 , . . . , am changes sign at least n + 1 times.

C Problems
C1. (USAMO 2002) Prove that any monic polynomial of degree n with real coefficients is the
average of two monic polynomials of degree n with n real roots.
C2. (Kürschák 2017) Do there exist polynomials p(x) and q(x) with real coefficients such that
p3 (x) − q 2 (x) is linear but not constant?
C3. (ISL 2019) A polynomial P (x, y, z) in three variables with real coefficients satisfies the iden-
tities
P (x, y, z) = P (x, y, xy − z) = P (x, zx − y, z) = P (yz − x, y, z).
Prove that there exists a polynomial F (t) in one variable such that

P (x, y, z) = F (x2 + y 2 + z 2 − xyz).

C4. (USATSTST 2017) Consider solutions to the equation

f (x)
x2 − cx + 1 = ,
g(x)

where f and g are polynomials with nonnegative real coefficients. For each c > 0, determine the
minimum possible degree of f , or show that no such f, g exist.

15 of 16
2024 Winter Camp Real Polynomials Victor Rong

C5. (ISL 2014) Consider all polynomials P (x) with real coefficients that have the following prop-
erty: for any two real numbers x and y one has

|y 2 − P (x)| ≤ 2|x| if and only if |x2 − P (y)| ≤ 2|y|.

Determine all possible values of P (0).


C6. (ISL 2013) Let m ̸= 0 be an integer. Find all polynomials P (x) with real coefficients such
that
(x3 − mx2 + 1)P (x + 1) + (x3 + mx2 + 1)P (x − 1) = 2(x3 − mx + 1)P (x)
for all real number x.
C7. (Putnam 2022) Let n be a positive integer. Determine, in terms of n, the largest integer m with
the following property: There exist real numbers x1 , . . . , x2n with −1 < x1 < x2 < . . . < x2n < 1
such that the sum of the lengths of the n intervals

[x2k−1
1 , x22k−1 ], [x2k−1
3 , x2k−1
4 ], . . . , [x2k−1 2k−1
2n−1 , x2n ]

is equal to 1 for all integers k with 1 ≤ k ≤ m.


C8. (USATST 2017) Let P, Q ∈ R[x] be relatively prime nonconstant polynomials. Show that
there can be at most three real numbers λ such that P + λQ is the square of a polynomial.
C9. (ELMOSL 2011) Let n > 1 be an integer and a, b, c be three complex numbers such that
a + b + c = 0 and an + bn + cn = 0. Prove that two of a, b, c have the same magnitude.
C10. (China 2018) Suppose the real number λ ∈ (0, 1) , and let n be a positive integer. Prove that
the modulus of all the roots of the polynomial
n  
X n k(n−k) k
f (x) = λ x
k
k=0

are 1.
C11. (ISL 2015) Let n be a fixed integer with n ≥ 2. We say that two polynomials P and Q with
real coefficients are block-similar if for each i ∈ {1, 2, . . . , n} the sequences

P (2015i), P (2015i − 1), . . . , P (2015i − 2014) and


Q(2015i), Q(2015i − 1), . . . , Q(2015i − 2014)

are permutations of each other.


(a) Prove that there exist distinct block-similar polynomials of degree n + 1.
(b) Prove that there do not exist distinct block-similar polynomials of degree n.

16 of 16

You might also like