Junior problems

J679. Let n be a positive integer. Find a representation of N = 68n + 64n+1 + 1 as the sum of the squares
of three distinct positive integers.

Proposed by Adrian Andreescu, University of Texas at Dallas, USA

Solution by Raja Oktovin Parhasian Damanik, Canberra ACT, Australia

Note that

68n + 64n+1 + 1 = 68n + 2 · 64n + 1 + 4 · 64n

= (64n + 1)2 + (2 · 62n )2
= (64n + 1 − 2 · 62n )2 + 2 · (64n + 1) · (2 · 62n )
= (64n − 2 · 62n + 1)2 + 4 · 66n + 4 · 62n

2
= (62n − 1)2 + (2 · 63n )2 + (2 · 6n )2 .

Note that (62n − 1)2 is an odd positive integer and 2 · 63n > 2 · 6n are even numbers, hence we have just
written 68n + 64n+1 + 1 as sum of the squares of three distinct positive integers.

Also solved by Ángel Plaza, Universidad de Las Palmas de Gran Canaria, Spain; Marin Chirciu, Colegiul
National Zinca Golescu Piteşti, Romania; Sundaresh H. R., Shivamogga, Karnataka, India; Theo Koupelis,
Clark College, Washington, USA; Israel Castillo Pilco, Universidad Nacional de Ingenierı́a, Peru; Ganghun
Kim, The Lawrenceville School, NJ, USA; Kayla Song, Hopkins School, CT, USA; Polyahedra, Polk State
College, FL, USA; Prasanna Pawar, AMMOC, Bengaluru, India; Suyeong Hahn, Choate Rosemary Hall,
Connecticut, USA; William Han, Horace Mann School, Bronx, NY, USA; Yoonwoo Lee, Academic Affiliation
Seoul International School, South Korea.

Mathematical Reflections 6 (2024) 1

J680. Let ABC be a triangle with incenter I and inradius r. On altitude AD point F is taken such that
AF = r. Let E be the midpoint of BC. Prove that E, I, F are collinear.

Proposed by Mihaela Berindeanu, Bucharest, Romania

Solution 1 by Kousik Sett, India

Without loss of generality (WLOG), we assume that AC is the largest side of triangle ABC. Let the
incircle touch side BD at point X. We have BE = CE = a/2, BX = (c + a − b)/2, and BD = c cos B =
(c2 + a2 − b2 )/(2a).

Also,
a c2 + a2 − b2 (b + c)(b − c)
DE = BE − BD = − = ,
2 2a 2a
and
a c+a−b b−c
XE = BE − BX = − = .
2 2 2
So,
DE b+c
= .
XE a
Let AD = h, and let ∆ denotes the area of triangle ABC. Since AF = IX = r, we have

FD AD − AF h 2∆/a 2s b+c
= = −1= −1= −1= .
IX IX r ∆/s a a
Hence,
DE FD IX FD
= , i.e., = ,
XE IX XE DE
and since ∠IXE = ∠F DE = 90◦ , we have triangles IXE and F DE are similar.
Therefore, ∠XEI = ∠DEF and since E, X, D are collinear, we must have E, I, F are collinear.

Solution 2 by Miguel Amengual Covas, Cala Figuera, Mallorca, Spain

Let a, b, c be the lengths of the sides of △ABC opposite A, B, C, semiperimeter s and area ∆.
If b = c, then F , I, E all lie on AD.
Suppose b > c. Let AI (extended) intersect BC at K. Then K divides BC in the ratio of the sides about
A, i.e.,
ac
BK =
b+c
and we have
AF AF r r a a
= = 2∆ = 2rs = = ,
FD AD − AF a −r a −r
2s − a b+c

Mathematical Reflections 6 (2024) 2

 a a c2 +a2 −b2 2
− c · cos B −

DE BE − BD 2 2 2a b+c
= = a ac = a(b−c)
=
EK BE − BK 2 − b+c a
2(b+c)

and   ac
KI by the angle bisector theorem BK a
= = = b+c = .
IA applied to △ABK at B AB c b+c
Therefore  2
AF DE KI a b+c a
· · = · · = 1,
F D EK IA b+c a b+c
whence F , I, E are collinear, by the converse of the celebrated Menelaus’s theorem applied to the triad of
points F IE on the sides of triangle DAK.

∗
∗
B D K E C

Also solved by Cristian Muth, UMFST Highschool in G. E. Palade University of Medicine, Pharmacy,
Science, and Technology, Romania; Raja Oktovin Parhasian Damanik, Canberra ACT, Australia; Theo
Koupelis, Clark College, Washington, USA; Benjamin Cha, Choate Rosemary Hall, Wallingford, CT, USA;
Daniel Pascuas, Barcelona, Spain; Diar Gashi, University of Primorska, Slovenia; Polyahedra, Polk State
College, FL, USA.

Mathematical Reflections 6 (2024) 3

J681. Solve in positive real numbers the system of equations
5xy + y + 5 xy + 1 37
+ =
xy + 1 xy + y + 1 6
3yz + z + 3 yz + z + 1 17
+ =
yz + 1 yz + 2z + 1 4
2zx + x + 2 zx + 3x + 1 10
+ = .
zx + 1 zx + 4x + 1 3

Proposed by Alessandro Ventullo, Milan, Italy

Solution 1 by the author

The given equations can be written as
y y 1
5+ +1− = 6+
xy + 1 xy + y + 1 6
z z 1
3+ +1− = 4+
yz + 1 yz + 2z + 1 4
x x 1
2+ +1− = 3+ ,
zx + 1 zx + 4x + 1 3
i.e.
1 1 1
− =
x + 1/y x + 1/y + 1 6
1 1 1
− =
y + 1/z y + 1/z + 2 4
1 1 1
− = .
z + 1/x z + 1/x + 4 3
1 1 1
Set a = x + , b = y + , c = z + . We get
y z x
1 1 1
− =
a a+1 6
1 1 1
− =
b b+2 4
1 1 1
− = .
c c+4 3
Since a, b, c > 0, solving each equation, we get a = b = c = 2. It follows that
1
x+ = 2
y
1
y+ = 2
z
1
z+ = 2.
x
Adding these three equations, we get
2 2 2
√ √
  
1 √ 1 1
x− √ + y−√ + z−√ = 0,
x y z

so x = y = z = 1. It is easy to see that this is indeed a solution.

Mathematical Reflections 6 (2024) 4

Solution 2 by Raja Oktovin Parhasian Damanik, Canberra ACT, Australia
By calling (1), (2) and (3) the equations of the system, note that
y 37 y 1 1 1
(1) ⇒ 5 + +1= + ⇒ = +
xy + 1 6 xy + y + 1 xy + 1 6y xy + y + 1
z 17 z 1 1 1
(2) ⇒ 3 + +1= + ⇒ = +
yz + 1 4 yz + 2z + 1 yz + 1 4z yz + 2z + 1
x 10 x 1 1 1
(3) ⇒ 2 + +1= + ⇒ = + .
zx + 1 3 zx + 4x + 1 zx + 1 3x zx + 4x + 1
By the AM-HM inequality,
1 1 1 1 9
= + + ≥
xy + 1 6y 2(xy + y + 1) 2(xy + y + 1) 4xy + 10y + 4
1 1 1 4
= + ≥
yz + 1 4z yz + 2z + 1 yz + 6z + 1
1 1 1 1 9
= + + ≥
zx + 1 6x 6x zx + 4x + 1 zx + 16x + 1
Rearranging,

4xy + 10y + 4 ≥ 9xy + 9 ⇔ 2y ≥ xy + 1

yz + 6z + 1 ≥ 4yz + 4 ⇔ 2z ≥ yz + 1
zx + 16x + 1 ≥ 9zx + 9 ⇔ 2x ≥ zx + 1.

But by multiplying the three inequalities and by using the AM-GM inequality, we have
√ √ √
8xyz = 2x · 2y · 2y ≥ (zx + 1) · (xy + 1) · (yz + 1) ≥ 2 zx · 2 xy · 2 yz = 8xyz.

Hence, the equality must hold, and this happens if and only if xy = yz = zx = 1, i.e. x = y = z = 1. It is
easy to check that (x, y, z) = (1, 1, 1) is indeed a solution to our system.

Solution 3 by Diar Gashi, University of Primorska, Slovenia

We notice that the equations have the following form

nab + b + n ab + (t − 1)b + 1 (n + 1)2 + 1

+ =
ab + 1 ab + tb + 1 n+1
4
for t = n−1 , where n = 5, 3 and 2 respectively. And this can be simplified further as

nab + b + n ab + (t − 1)b + 1 (n + 1)2 + 1

0= −n+ −1− + (n + 1)
ab + 1 ab + tb + 1 n+1
b b 1
= − −
ab + 1 ab + tb + 1 n + 1
tb2 1
= −
(ab + 1)(ab + tb + 1) n + 1
(ab + 1)(ab + tb + 1) − tb2 (n + 1)
=−
(ab + 1)(ab + tb + 1)(n + 1)
(ab + 1)2 + tb(ab + 1) − t(n + 1)b2
=−
(ab + 1)(ab + tb + 1)(n + 1)

The denominator is nonzero, so

p
2 2 −tb ± t2 b2 + 4t(n + 1)b2
0 = (ab + 1) + tb(ab + 1) − t(n + 1)b =⇒ ab + 1 =
2

Mathematical Reflections 6 (2024) 5

 q
4 16 16(n+1)
− n−1 ± (n−1)2
+ n−1

−1 ± n

=⇒ ab + 1 = b=2 b.
2 n−1
Since ab + 1 > 0 =⇒ ab + 1 = 2b. So, by AM-GM and this we get that

b2 + 1 ≥ 2b = ab + 1 =⇒ b2 ≥ ab =⇒ b ≥ a.

Hence, x ≥ y ≥ z ≥ x =⇒ x = y = z = 1 .

Also solved by Ángel Plaza, Universidad de Las Palmas de Gran Canaria, Spain; Arkady Alt, San Jose,
California, USA; G. C. Greubel, Newport News, VA, USA; Nicuşor Zlota, “Traian Vuia” Technical College,
Focşani, Romania; Paolo Perfetti, Università degli studi di Tor Vergata Roma, Italy; Sundaresh H. R.,
Shivamogga, Karnataka, India; Theo Koupelis, Clark College, Washington, USA; Brian Bradie, Christo-
pher Newport University, Newport News, VA, USA; Ganghun Kim, The Lawrenceville School, NJ, USA;
Polyahedra, Polk State College, FL, USA; William Han, Horace Mann School, Bronx, NY, USA.

Mathematical Reflections 6 (2024) 6

J682. Let a and b be positive real numbers. Prove that

a b2
  2 
a b 9ab
+ 2 2
+ + 2 ≥ 13.
b a b a a − ab + b2

Proposed by An Zhenping, Xianyang National University, China

Solution 1 by Arkady Alt, San Jose, California, USA

a b
Let t := + ≥ 2. Then
b a
a b2
 2
a3 b3
 
a b 9ab 9
+ 2 2
+ + 2 − 13 = + + − 11
b a b a a − ab + b2 b3 a3 a
b + b
a −1
9
= t3 − 3t + − 11
t−1
t2 + 3t + 5 (t − 2)2


= ≥ 0.
t−1
Equality holds if and only if a = b.

Solution 2 by Daniel Pascuas, Barcelona, Spain

a
Let t = and write the inequality to be proved as
b
  
1 2 1 9t
t+ 2 t + + 2 ≥ 13 (t > 0). (∗)
t t t −t+1

Since
(t3 + 1)2 13t2 − 22t + 13
  
1 1 9t
t+ 2 t2 + = , 13 − = ,
t t t3 t2 −t+1 t2 − t + 1
and t2 − t + 1 > 0 (because t2 + 1 ≥ 2t > t), (∗) can be written as

(t3 + 1)2 (t2 − t + 1) ≥ t3 (13t2 − 22t + 13) (t > 0),

that is,
t8 − t7 + t6 − 11t5 + 20t4 − 11t3 + t2 − t + 1 ≥ 0 (t > 0).
And this inequality holds because the polynomial

P (t) = t8 − t7 + t6 − 11t5 + 20t4 − 11t3 + t2 − t + 1

factorizes as P (t) = (t − 1)4 (t4 + 3t3 + 7t2 + 3t + 1). Equality holds if and only if t = 1, i.e. when a = b.

Solution 3 by Polyahedra, Polk State College, FL, USA

The inequality is equivalent to

(a3 + b3 )2 9ab
0 ≤ −4+ 2 −9
a3 b3 a − ab + b2

2
a3 − b3 9(a − b)2
= 3 3
− 2
a b a − ab + b2

a2 + ab + b2
 
2 2 2 9
= (a − b) a + ab + b − 4 ,
a3 b3 a + a2 b2 + b4

which is clearly true since a2 + ab + b2 a4 + a2 b2 + b4 ≥ 3ab · 3a2 b2 = 9a3 b3 .

Mathematical Reflections 6 (2024) 7

Solution 4 by Ioan Viorel Codreanu, Satulung, Maramures, Romania
The inequality is equivalent to
(a3 + b3 )2 9ab(a + b)
+ ≥ 13.
a3 b3 a3 + b3
Using the AM-GM Inequality, we get

(a3 + b3 )2 9ab (a3 + b3 )2 (a3 + b3 )2 ab(a + b) ab(a + b)

3 3
+ 2 = + . . . + + 3 + ... + 3
a b a − ab + b2 3
| 4a b
3
{z
3 3
4a b } a +b3 a + b3
| {z }
4 times 9 times
s
(a3 + b3 )8 a9 b9 (a + b)9
≥ 13 13 4 12 12 ·
4 a b (a3 + b3 )9
s
(a + b)9
= 13 13 8 3 3 3 .
2 a b (a + b3 )

It is enough to prove that

(a + b)9 ≥ 28 a3 b3 (a3 + b3 ).
Let s = a + b, p = ab. The last inequality becomes

s9 ≥ 256p3 s(s2 − 3p),

equivalent to
s8 + 256 · 3p4 ≥ 256p3 s2 .
Using the AM-GM Inequality, we get
p
s8 + 256 · 3p4 = s8 + 256p4 + 256p4 + 256p4 ≥ 4 4 2563 s8 p12 = 256p3 s2 .

Also solved by Ángel Plaza, Universidad de Las Palmas de Gran Canaria, Spain; Arshia Rahimi, Malard,
Iran; Daniel Văcaru, National Economic College “Maria Teiuleanu”, Pites, ti, Romania; Henry Ricardo,
Westchester Area Math Circle, New York, USA; Marin Chirciu, Colegiul National Zinca Golescu Piteşti,
Romania; Nicuşor Zlota, “Traian Vuia” Technical College, Focşani, Romania; Paolo Perfetti, Università
degli studi di Tor Vergata Roma, Italy; Raja Oktovin Parhasian Damanik, Canberra ACT, Australia; Sun-
daresh H. R., Shivamogga, Karnataka, India; Theo Koupelis, Clark College, Washington, USA; Anderson
Torres, Sao Paulo, Brazil; Brian Bradie, Christopher Newport University, Newport News, VA, USA; Diar
Gashi, University of Primorska, Slovenia; Israel Castillo Pilco, Universidad Nacional de Ingenierı́a, Peru;
Ganghun Kim, The Lawrenceville School, NJ, USA; Kayla Song, Hopkins School, CT, USA; Suyeong Hahn,
Choate Rosemary Hall, Connecticut, USA; William Han, Horace Mann School, Bronx, NY, USA; Yoonwoo
Lee, Academic Affiliation Seoul International School, South Korea.

Mathematical Reflections 6 (2024) 8

J683. Let a0 be a real number and let an+1 = a2n + 16 . Prove that an+2 > a2n−1 for all positive integers n.

Proposed by Titu Andreescu, University of Texas at Dallas, USA

Solution 1 by the author

We have
1
an = a2n−1 +
6
1
an+1 = a2n +
6
1
an+2 = a2n+1 +
6
for all positive integers n. Summing up and completing squares yields

1 2 1 2
   
an+2 = a2n−1 + an − + an+1 − .
2 2
 2
1 1 1 1
Thus an+2 ≥ a2n−1 . We cannot have equality, as that would imply an = an+1 = , yielding = + ,
2 2 2 6
a contradiction. Hence the conclusion.

Solution 2 by Raja Oktovin Parhasian Damanik, Canberra ACT, Australia

Note that by definition of the sequence and AM-GM inequality, for every n ≥ 1,

1 2 1
   
1 1 2 7 1 1 1
an+2 = a2n+1 + =2
an + 4
+ = an + an + 4
= an + + a2n +
6 6 6 3 36 9 3 12
 2  
2 1 1 1 1 1 1 1
≥ a2n + a2n + = a2n + = a2n−1 + + = a4n−1 + + a2n−1
3 3 12 12 6 12 9 3
2 2 1 2
≥ an−1 + an−1
3 3
2
= an−1

The equalities in both AM-GM inequality applications will occur simultaneously iff a2n = a2n−1 = 13 , which
is impossible since a2n = (a2n−1 + 16 )2 = ( 13 + 16 )2 = 41 ̸= 13 . So the inequality an+2 > a2n−1 must be strict.

Also solved by Ángel Plaza, Universidad de Las Palmas de Gran Canaria, Spain; Arkady Alt, San Jose,
California, USA; Nicuşor Zlota, “Traian Vuia” Technical College, Focşani, Romania; Paolo Perfetti, Uni-
versità degli studi di Tor Vergata Roma, Italy; Theo Koupelis, Clark College, Washington, USA; Anderson
Torres, Sao Paulo, Brazil; Brian Bradie, Christopher Newport University, Newport News, VA, USA; Daniel
Pascuas, Barcelona, Spain; Polyahedra, Polk State College, FL, USA; Suyeong Hahn, Choate Rosemary
Hall, Connecticut, USA; William Han, Horace Mann School, Bronx, NY, USA.

Mathematical Reflections 6 (2024) 9

J684. Prove that there are no rational numbers x, y such that

2x2 + 12y 2 = 1.

Proposed by Dorin Andrica and George Cătălin Ţurcaş, Cluj-Napoca, Romania

Solution 1 by Benjamin Cha, Choate Rosemary Hall, Wallingford,

√ CT, USA
It is easy
√ to see that assuming x = 0 leads to y = ±1/ 12, so y is irrational. Similarly, y = 0 gives
x = ±1/ 2, also irrational.
Suppose that there exist nonzero integers p, q, r, s with gcd(p, q) = 1, gcd(r, s) = 1, with x = p/q, y = r/s
and 2x2 + 12y 2 = 1.
Then, the initial equation can be written as

12q 2 r2 = s2 (q 2 − 2p2 ).
Denoting by vp (n) the p-adic valuation of an integer n, we see that v3 (12q 2 r2 ) is necessarily odd, since
12 = 22 · 31 and q 2 r2 is a perfect square. Since v3 (s2 ) is even (possibly 0), it follows that v3 (q 2 − 2p2 ) is
necessarily odd.
In particular, q 2 − 2p2 must be divisible by 3. But both p2 and q 2 are congruent to 0 or 1 modulo 3. The
only way for which q 2 − 2p2 ≡ 0 (mod 3) is when both p and q are divisible by 3. But this contradicts the
fact that gcd(p, q) = 1.
We reached a contradiction and so the equation 2x2 + 12y 2 = 1 has no rational solutions.

Solution 2 by Polyahedra, Polk State College, USA

If there are such rational numbers x, y, then there is a least common denominator d such that x = m/d and
y = n/d. Therefore, 2m2 + 12n2 = d2 , so d must be even, say, d = 2a. Then m2 + 6n2 = 2a2 , so m is even,
say, m = 2b. Then 2b2 + 3n2 = a2 . If n is even, then both m/d and n/d can be reduced, contradicting the
fact that d is the least common denominator. Therefore, n and a must both be odd. Then a2 − 3n2 ≡ −2
(mod 8) while 2b2 ≡ 0, 2 (mod 8), a contradiction again.

Also solved by Ángel Plaza, Universidad de Las Palmas de Gran Canaria, Spain; Bui Linh Phuc An, Hung
Vuong High School for the gifted, Binh Duong, Vietnam; Daniel Văcaru, National Economic College “Maria
Teiuleanu”, Pites, ti, Romania; Raja Oktovin Parhasian Damanik, Canberra ACT, Australia; Theo Koupelis,
Clark College, Washington, USA; Daniel Pascuas, Barcelona, Spain; Kayla Song, Hopkins School, CT,
USA; Prasanna Pawar, AMMOC, Bengaluru, India; Suyeong Hahn, Choate Rosemary Hall, Connecticut,
USA.

Mathematical Reflections 6 (2024) 10

 Senior problems

S679. Find all positive integers n for which there is a positive integer k such that (n − 1)! + 1 = (k − 16)2
and (n + 1)! + 9 = (k + 16)2 .

Proposed by Adrian Andreescu, University of Texas at Dallas, USA

Solution 1 by the author

Clearly, k is odd. We see that (n − 1)! + 1 is not a perfect square for n = 1, 2, 3, 4, while for n = 5 (and
k = 11), 4! + 1 = (11 − 16)2 and 6! + 9 = (11 + 16)2 . Suppose now that n ≥ 6. We have

(n − 1)! + (n + 1)! + 10 = (k − 16)2 + (k + 16)2 ,

implying
(n − 1)! + (n + 1)! = 2k 2 + 502.
Because k is odd, k 2 = 8s + 1 for some positive integer s. The left-hand side is divisible by 16, while the
right-hand side, which is equal to 8(2s + 63), is only divisible by 8, a contradiction. In conclusion, n = 5 is
the only solution.

Solution 2 by Theo Koupelis, Clark College, Washington, USA

Subtracting the two given equations we get

(n − 1)! [n(n + 1) − 1] + 8 = 64k. (∗)

If n ≥ 7, then n − 1 ≥ 6, and thus (n − 1)! is a multiple of 16. From (∗) we get that this is not possible
because 16 ∤ 8. Also, k is a positive integer, and therefore, from the second of the given equations we get
(n + 1)! ≥ (1 + 16)2 − 9 = 280, and thus n ≥ 5, because 5! = 120. Therefore, we only need examine the
cases n = 5 and n = 6. For n = 5 we get (k − 16)2 = 25 = 52 and (k + 16)2 = 729 = 272 , from which we get
k = 11. For n = 6 we get (k − 16)2 = 121 = 112 and (k + 16)2 = 5049, which is not the square of an integer.
Thus, (n, k) = (5, 11) is the only acceptable solution.

Also solved by G. C. Greubel, Newport News, VA, USA; Nicuşor Zlota, “Traian Vuia” Technical College,
Focşani, Romania; Sundaresh H. R., Shivamogga, Karnataka, India; Daniel Pascuas, Barcelona, Spain;
Diar Gashi, University of Primorska, Slovenia; Ganghun Kim, The Lawrenceville School, NJ, USA; William
Han, Horace Mann School, Bronx, NY, USA.

Mathematical Reflections 6 (2024) 11

S680. Let x, y, z be positive real numbers. Prove that
s
√ √ √ 16(x + y + z) 2(xy + yz + zx)
x+y+ y+z+ z+x≤ + .
3 x+y+z

Proposed by Marius Stănean, Zalău, Romania

Solution by the author √

√ √
There is an acute triangle ABC with sides a = y + z, b = z + x, c = x + y. The inequality becomes
r
8(a2 + b2 + c2 ) 2(a2 b2 + b2 c2 + c2 a2 ) − (a4 + b4 + c4 )
a+b+c≤ + .
3 a2 + b2 + c2
We can rewrite this as
s
5(a4 + b4 + c4 ) + 22(a2 b2 + b2 c2 + c2 a2 )
a+b+c≤ ,
3(a2 + b2 + c2 )

so that after squaring we get

3(a + b + c)2 (a2 + b2 + c2 ) ≤ 5(a4 + b4 + c4 ) + 22(a2 b2 + b2 c2 + c2 a2 ).

We prove that this inequality is valid in any triangle, not only in an acute triangle. Using the Ravi’s
substitutions i.e. a = v + w, b = w + u, c = u + v, u, v, w > 0, the inequality can be rewritten as follows
X X
24(u + v + w)2 (u2 + v 2 + w2 + uv + vw + wu) ≤ 5 (u + v)4 + 22 (v + w)2 (w + u)2 ,
cyc cyc

or after expanding and reducing terms, it becomes

u4 + v 4 + w4 + uvw(u + v + w) ≥ u3 (v + w) + v 3 (w + u) + w3 (u + v),

or
u2 (u − v)(u − w) + v 2 (v − w)(v − u) + w2 (w − u)(w − v) ≥ 0,
which is true by Schur’s Inequality. The equality holds when x = y = z.

Also solved by Arkady Alt, San Jose, California, USA; Nicuşor Zlota, “Traian Vuia” Technical College,
Focşani, Romania; Paolo Perfetti, Università degli studi di Tor Vergata Roma, Italy; Theo Koupelis, Clark
College, Washington, USA.

Mathematical Reflections 6 (2024) 12

S681. Show that there are no integers x and y satisfying

x2 + y 1013 = 20242025 .

Proposed by Vlad Crişan, Switzerland

Solution by the author

Given that we are asked to show there are no integer solutions to the given equation, it is natural to wonder
if we can find a positive integer n such that the given equation has no solutions modulo n.
We will now motivate how to choose a suitable value for n. For the given equation to not have solutions
modulo n, we need x2 + y 1013 to not cover all possible residues modulo n. Recall that if n is a positive
integer and a is an integer coprime to n, then by Euler’s theorem one has aϕ(n) ≡ 1 (mod n), where ϕ(n)
stands for Euler’s totient function.
Looking at the exponent 1013, a natural choice for n would be values for which 1013 is a divisor of ϕ(n), as
then y 1013 would cover a restricted set of values modulo n.
Notice that 1013 is prime, so we can’t have ϕ(n) = 1013. However, 2 · 1013 = 2026 and 2027 is prime, thus
ϕ(2027) = 2026. All this motivates trying n = 2027.
We will now show that indeed the equation has no solutions modulo 2027, and thus no integer solutions.
From Euler’s theorem, we know that y 2026 is either 0 or 1 modulo 2027, and thus y 1013 is one of {−1, 0, 1}
modulo 2027. Also, since 20242026 ≡ 1 (mod 2027) (again by Euler), it follows that 20242025 is the inverse
of 2024 modulo 2027. Since 2024 ≡ −3 (mod 2027) and 2027 ≡ 2 (mod 3), it follows that 2 · 2027 ≡ 1
(mod 3), which gives us −3 · 1351 ≡ 1 (mod 2027), and thus 20242025 ≡ 1351 (mod 2027).
It follows that 20242025 − y 1013 can only be one of {1350, 1351, 1352} modulo 2027. We need to show that
none of the values in the set {1350, 1351, 1352} is a quadratic residue modulo 2027. Notice that 3·1351 ≡ −1
(mod 2027). From the quadratic reciprocity law, we have that 3 is a quadratic residue modulo 2027, while
−1 is not a quadratic residue modulo 2027, since 2027 ≡ 3 (mod 4). It follows that 1351 is not a quadratic
residue modulo 2027. Since 3 · 1350 ≡ −4 (mod 2027), and −4 is not a quadratic residue modulo 2027
(since −1 isn’t), it follows that 1350 is also not a quadratic residue modulo 2027. Finally, since 3 · 1352 ≡ 2
(mod 2027) and we saw that 3 is a quadratic residue modulo 2027, it follows that 1352 is a quadratic residue
if and only if 2 is. But 2 is not a quadratic residue, since 2027 is not ±1 modulo 8. This concludes our
proof.

Mathematical Reflections 6 (2024) 13

S682. Let a, b, c be positive real numbers. Prove that

a2 b c a
+ + ≥ + 1.
2bc a + b a + c b+c

Proposed by Nguyen Viet Hung, Hanoi University of Science, Vietnam

Solution by Arkady Alt, San Jose, California, USA

By Cauchy-Schwarz inequality

a2 b c a2 b2 c2
+ + = + +
2bc a + b a + c 2bc ab + b2 ca + c2
(a + b + c)2
≥
2bc + ab + b2 + ca + c2
(a + b + c)2
=
(b + c) (a + b + c)
a+b+c
=
b+c
a
= + 1.
b+c
Equality occurs if and only if a = b = c.

Also solved by Daniel Văcaru, National Economic College “Maria Teiuleanu”, Pites, ti, Romania; Ioan Viorel
Codreanu, Satulung, Maramures, Romania; Marin Chirciu, Colegiul National Zinca Golescu, Piteşti, Ro-
mania; Nicuşor Zlota, “Traian Vuia” Technical College, Focşani, Romania; Paolo Perfetti, Università degli
studi di Tor Vergata Roma, Italy; Sundaresh H. R., Shivamogga, Karnataka, India; Theo Koupelis, Clark
College, Washington, USA; Daniel Pascuas, Barcelona, Spain; Diar Gashi, University of Primorska, Slove-
nia; Israel Castillo Pilco, Universidad Nacional de Ingenierı́a, Peru.

Mathematical Reflections 6 (2024) 14

S683. Let OA, OB, OC be rays in the plane such that OB 2 = 2OA · OC and

AB 2 BC 2
+ = 3(OA + OC).
OA OC
Find ∠AOC.

Proposed by Titu Andreescu, University of Texas at Dallas, USA

Solution by Kousik Sett, India

Join A, B and B, C. Let ∠AOB = α and ∠BOC = β. Then ∠AOC = α + β.

Applying the cosine law on triangles AOB and BOC, we obtain

OA2 + OB 2 − AB 2 OC 2 + OB 2 − BC 2
cos α = and cos β = .
2 · OA · OB 2 · OC · OB

AB 2 BC 2
Substituting OB 2 = 2OA · OC and using + = 3(OA + OC), we have
OA OC

AB 2 BC 2
 
OA OC OC OA 1
cos α + cos β = + + + − +
2 · OB OB 2 · OB OB 2 · OB OA OC
3(OA + OC) 3(OA + OC)
= − = 0.
2OB 2OB
Hence,
cos α = − cos β = cos(π − β) ⇒ α + β = π = 180◦ .
Therefore,
∠AOC = 180◦ .

Remark. Observe that the two given conditions, when combined, can be written as

AB 2 · OC + BC 2 · OA = (OC + OA)(OB 2 + OC · OA),

which is the Stewart’s Theorem for triangle ABC.

Also solved by Marin Chirciu, Colegiul National Zinca Golescu, Piteşti, Romania; Theo Koupelis, Clark
College, Washington, USA.

Mathematical Reflections 6 (2024) 15

S684. Let O and P be points in the plane. Prove that there is a circle Γ of center O, whose radius is an
integer and there are points A, B, C on Γ such that
−−→ −→ −−→ −−→
OP = OA + OB + OC.

Proposed by Dorin Andrica and George Cătălin Ţurcaş, Cluj-Napoca, România

Solution 1 by Kousik Sett, India

If points A, B, and C on Γ then O is the circumcenter of triangle ABC. Let H be the orthocenter of triangle
ABC and D be the midpoint of BC. It is well-known that AH = 2OD.

−−→ −−→ −−→

In triangle OBC, we have OB + OC = 2OD. Therefore,
−→ −−→ −−→ −→ −−→ −→ −−→ −−→
OA + OB + OC = OA + 2OD = OA + AH = OH.
−−→ −→ −−→ −−→ −−→
From the given condition, we have OP = OA + OB + OC = OH. So, P = H.
−→ −−→ −−→
Let |OA| = |OB| = |OC| = R. It is known that ∠BOC = 2A, ∠COA = 2B, and ∠AOB = 2C.
Therefore,
−−→ −−→ −−→ −−→ −→ −−→ −−→ −→ −−→ −−→
|OP |2 = (OP )2 =OP · OP = (OA + OB + OC) · (OA + OB + OC)
−−→ −−→ −−→ −→ −→ −−→
=3R2 + 2(OB · OC + OC · OA + OA · OB)
=3R2 + 2R2 (cos 2A + cos 2B + cos 2C)
=3R2 + 2R2 (1 − 2 sin2 A + 1 − 2 sin2 B + 1 − 2 sin2 C)
=9R2 − (a2 + b2 + c2 )

where BC = a = 2R sin A, CA = b = 2R sin B, and AB = c = 2R sin C.

Therefore,
a2 + b2 + c2 + x2 = 9R2 ,
where |OP | = x.
Now, we demonstrate that for some integer values of R, the above equation admits acceptable solutions.
We know that R = abc/(4∆), where ∆ denotes the area of triangle ABC. To find such an R, we need to
find suitable triplets (a, b, c) for which the area of the triangle is an integer.
Let us choose, a = 13k, b = 14k, c = 15k, where k is a positive integer. Then using Heron’s formula, we
obtain ∆ = 84k 2 . Hence, R = 65k/8. So, we need to choose k = 8m where m ∈ N, where N denotes the set of
Natural numbers. Therefore, for m = 1, i.e., for k = 8, we obtain a solution (a, b, c, R) = (104, 112, 120, 65).
Clearly, there exist infinitely many solutions since we can choose any integer value for m.

Mathematical Reflections 6 (2024) 16

Other solutions can be found by choosing different triplets such as (a, b, c) = (4k, 13k, 15k), (9k, 10k, 17k),
and so on.

Solution 2 by Theo Koupelis, Clark College, Washington, USA

Let d = OP > 0 in some units of distance. Consider the circle Γ of center O and radius r, such that r is any
positive integer ≥ ⌈d⌉. Let C be the point of intersection of the line OP with Γ, such that P is on the segment
OC. Let A, B be points on Γ, symmetrically positioned about the line OP, such that ∠COA = ∠COB = θ,
where π2 ≤ θ < 2π3 . Then, in such a configuration, we get

−−→ −→ −−→ −−→ d−r

OP = OA + OB + OC ⇐⇒ d = 2r cos θ + r ⇐⇒ cos θ = .
2r
1 d−r
Such an angle always exists under the chosen conditions because − < ≤ 0.
2 2r

Mathematical Reflections 6 (2024) 17

 Undergraduate problems

U679. Evaluate π
1 − ln cos x
Z
2
dx.
0 x2 + (1 − ln cos x)2

Proposed by Paolo Perfetti, Università degli studi ”Tor Vergata”, Roma, Italy

Solution by Theo Koupelis, Clark College, Washington, USA

For |x| < π2 we have

eix + e−ix 1 + ei2x

 
−1+ix −1+ix
1 − ln cos x − ix = − ln(e cos x) = − ln e · = − ln
2 2e

Thus,
π Z π Z π
1 − ln cos x
Z
2 2 1 1 2 1
I:= 2 2
dx = − ℜ 1+ei2x dx = − ℜ 1+ei2x dx
0 x + (1 − ln cos x) 0 ln 2 − π ln
Z π Z π 2e 2 2e
1 1 1 1
= − PV ℜ 1+eix dx = − ℜPV 1+eix
dx.
4 −π ln 4 −π ln
2e 2e

As a contour of integration we choose the counterclockwise contour that consists of a circle Cϵ of radius 1,
centered at the origin, and a semicircular indent γϵ around −1. We choose the branch cut to be the negative
real axis. Let z = eix , and
1
f (z) = .
z ln 1+z
2e
Then
Z π−ϵ I Z 
1 1 1 dz
I = − ℜ lim dx = − ℑ lim −
z ln 1+z
1+eix
4 ϵ→0+ −π+ϵ ln 2 4 ϵ→0+ Cϵ γϵ 2e
 
1 1 −1 π
= − ℑ (2πi · Resz=0 f (z) − 0) = − ℑ 2πi · = ,
4 4 ln(2e) 2 ln(2e)

where we used the result that for z = −1 + ϵ · eiθ we have

ϵieiθ dθ
Z Z
dz
lim = lim iθ = 0.
ϵ→0+ z ln 1+z
2e
ϵ→0+ (−1 + ϵeiθ ) ln ϵe2e

Also solved by Prasanna Pawar, AMMOC, Bengaluru, India.

Mathematical Reflections 6 (2024) 18

U680. Prove that 18 is the largest positive constant k such that
1 1 1 3
+ 2 + 2 ≥
ab2 + k bc + k ca + k 1+k
for all a ≥ b ≥ c ≥ 0 such that a + b + c = 3.

Proposed by Vasile Cı̂rtoaje, Oil-Gas University of Ploieşti, România

Solution by the author

3
Setting a = b = and c = 0, the inequality leads to k ≤ 18. We will further show that the inequality is
2
true for k = 18. Let us denote p = a + b + c, q = ab + bc + ca, r = abc and

A = a2 b + b2 c + c2 a, B = ab2 + bc2 + ca2 .

Since p = 3, we have q ≤ p2 /3 = 3 and r ≤ p3 /27 = 1. By expanding, the inequality can be restated as

follows:
3k 2 ≥ 3r3 + (2k − 1)rA + k(k − 2)B,
6k 2 ≥ 6r3 + [(2k − 1)r + k(k − 2)](A + B) + [(2k − 1)r − k(k − 2)](A − B).
Since (2k − 1)r − k(k − 2) ≤ 2k − 1 − k(k − 2) = −k 2 + 4k − 1 < 0 and A − B = (a − b)(b − c)(a − c) ≥ 0, it
suffices to show that
6k 2 ≥ 6r3 + [(2k − 1)r + k(k − 2)](A + B),
i.e.
6k 2 ≥ 6r3 + [(2k − 1)r + k(k − 2)](pq − 3r),
648 ≥ 2r3 + (35r + 288)(q − r).
Case 1: 0 ≤ q ≤ 9/4. It suffices to show that
 
3 9
648 ≥ 2r + (35r + 288) −r ,
4

which is equivalent to
r(837 + 140r − 8r2 ) ≥ 0.
Case 2: 9/4 ≤ q ≤ 3. Let z = q/3 ∈ [3/4, 1]. For fixed z, we need to show that 648 ≥ f (r), where
f (r) = 2r3 + (35r + 288)(3z − r). By the fourth degree Schur’s inequality, we have

(p2 − q)(4q − p2 ) (3 − z)(4z − 3)

r≤ = := r0 .
6p 2

Since f ′ (r) = 6r2 − 70r + 105z − 288 ≤ 6 − 70r + 105 − 288 < 0, f (r) is decreasing. So, we only need to
show that 648 ≥ f (r0 ), i.e.
648 − 864z ≥ 2r03 − 35r02 − (288 − 105z)r0 .
Since
−432r0
648 − 864z = −216(4z − 3) = ,
3−z
we need to show that
−432
≥ 2r02 − 35r0 − (288 − 105z),
3−z
that is
−864
≥ 16z 4 − 120z 3 + 437z 2 − 585z − 180,
3−z
16z 5 − 168z 4 + 797z 3 − 1896z 2 + 1575z − 324 ≥ 0,

Mathematical Reflections 6 (2024) 19

 (z − 1)g(z) ≥ 0,
where
g(z) = 16z 4 − 152z 3 + 645z 2 − 1251z + 324.
The inequality holds if g(z) ≤ 0. Indeed,

g(z) < 645z 2 − 1251z + 324 = −645z(1 − z) − 151(4z − 3) − 2z − 129 < 0.

3
For k = 18, the equality occurs when a = b = c = 1, and also when a = b = and c = 0.
2
Also solved by Nicuşor Zlota, “Traian Vuia” Technical College, Focşani, Romania; Theo Koupelis, Clark
College, Washington, USA.

Mathematical Reflections 6 (2024) 20

U681. Let a and b be integers. prove that
n−1
Y 
2 2 2kπ 2kπ
a + b + 1 − 2a cos − 2b sin
n n
k=1

is an integer which is a sum of two perfect squares.

Proposed by Dorin Andrica, Babeş-Bolyai University, Cluj-Napoca, România

Solution by Brian Bradie, Christopher Newport University, Newport News, VA, USA
Let z = a + bi and ωk = exp(i · 2kπ/n). Then

2kπ 2 2kπ 2
   
(z − ωk )(z − ωk ) = a − cos + b − sin
n n
2kπ 2kπ
= a2 + b2 + 1 − 2a cos − 2b sin .
n n
It follows that
n−1
Y  n−1
2kπ 2kπ Y
a2 + b2 + 1 − 2a cos − 2b sin = (z − ωk )(z − ωk )
n n
k=1 k=1
Qn
k=1 (z − ωk )(z − ωk )
=
(z − 1)(z − 1)
(z n − 1)(z n − 1)
=
(z − 1)(z − 1)
= pn (z)pn (z)
= (Re(pn (z)))2 + (Im(pn (z)))2 ,

where pn (z) = 1 + z + z 2 + · · · + z n−1 . Because a and b are integers, both the real part and the imaginary
part of pn (z) are also integers. Thus,
n−1
Y 
2 2 2kπ 2kπ
a + b + 1 − 2a cos − 2b sin
n n
k=1

is an integer which is a sum of two perfect squares.

Also solved by Theo Koupelis, Clark College, Washington, USA; Benjamin Cha, Choate Rosemary Hall,
Wallingford, CT, USA; Daniel Pascuas, Barcelona, Spain.

Mathematical Reflections 6 (2024) 21

U682. Evaluate Z
sin x cos x
√ dx.
5 + sin 4x

Proposed by Mihaela Berindeanu, Bucharest, Romania

Solution 1 by Mingcan Fan, Huizhou University, Huizhou, China

Using substitution x = π8 + 2t , i.e., t = 2x − π4 , we have
Z Z Z
sin x cos x sin 2x 1 sin t + cos t
√ dx = √ dx = √ √ dt
5 + sin 4x 2 5 + sin 4x 4 2 5 + cos 2t
Z Z !
1 d(sin t) d(cos t)
= √ p dt − √ dt
4 2 6 − 2 sin2 t 4 + 2 cos2 t
√
1 sin t 1 cos t + 2 + cos2 t
= arcsin √ − ln √ +C
8 3 8 2
sin 2x − π4

 r 
1 1  π 2
 π
= arcsin √ − ln cos 2x − + 2 + cos 2x − + C.
8 3 8 4 4

where the following formula

Z Z
dx p
√ = sec udu = ln | tan u + sec u| + C = ln |x + 1 + x2 | + C
1 + x2
is used by using substitution x = tan u.

Solution 2 by Diar Gashi, University of Primorska, Slovenia

By Trigonometry identities,
Z Z
sin(x) cos(x) sin(2x)
I := p dx = p dx
5 + sin(4x) 2 4 + sin(2x) + cos(2x)2 + 2 sin(2x) cos(2x)
2
Z
1 sin(2x)
= p dx
2 4 + (sin(2x) + cos(2x))2
Z
1 sin(2x)
= q √ dx
2 π 2
4 + ( 2 sin(2x + 4 ))
Z
1 sin(2x)
= √ dx.
2 + (sin(2x + π4 ))2
p
2 2
π
Substituting u = 2x + 4 =⇒ du = 2dx and therefore
Z √1 (sin(u) − cos(u))
sin(u − π4 ) du sin(u) − cos(u)
Z Z
1 1 2 1
I= √ p
2
= √ p
2
du = p du.
2 2 2 + sin (u) 2 4 2 2 + sin (u) 8 2 + sin2 (u)
√ √
Below, substituting − cos(u) = 3t =⇒ sin(u)du = 3dt, we see that
Z Z Z √ Z
sin(u) sin(u) 3dt dt
I1 := p du = p du = √ √ = √
2 2
3 − cos (u) 3 1−t 2 1 − t2
2 + sin (u)
= arcsin(t) + C1
 
− cos(u)
= arcsin √ + C1
3

Mathematical Reflections 6 (2024) 22

 √ √
Also, substituting below sin(u) = 2t =⇒ cos(u) = 2dt, we get
Z Z √ Z  
cos(u) 2dt dt −1 −1 sin(u)
I2 := p du = √ = √ = sinh (t) + C1 = sinh √ + C2
2 + sin2 (u) 2 + 2t2 1 + t2 2

Thus, in total we get

    
1 − cos(u) −1 sin(u)
I= arcsin √ − sinh √ + C.
8 3 2

cos(2x + π4 ) sin(2x + π4 )
    
1 −1
=⇒ I = − arcsin √ + sinh √ + C.
8 3 2

Also solved by Arkady Alt, San Jose, California, USA; G. C. Greubel, Newport News, VA, USA; Marin
Chirciu, Colegiul National Zinca Golescu, Piteşti, Romania; Nicuşor Zlota, “Traian Vuia” Technical College,
Focşani, Romania; Paolo Perfetti, Università degli studi di Tor Vergata Roma, Italy; Theo Koupelis, Clark
College, Washington, USA; Anderson Torres, Sao Paulo, Brazil.

Mathematical Reflections 6 (2024) 23

U683. Let f : R → R be a function that has a primitive and let F : R → R be such a primitive. Prove that
the following are equivalent:

(i) (x − y)(f (x) − f (y)) ≤ (x − y)2 , for all x, y ∈ R.

   
1 1 1
(ii) F x + − 2F (x) + F x − ≤ 2 , for all x, y ∈ R and all n ∈ N.
n n n

Proposed by Dan-Ştefan Marinescu, Romania

Solution by Daniel Pascuas, Barcelona, Spain

First, note that (i) can be written as
f (x) − f (y) ≤ x − y, for all x, y ∈ R such that x ≥ y,
or equivalently
(i’) y − f (y) ≤ x − f (x), for all x, y ∈ R such that x ≥ y.
2
Since F is a primitive of f , (i’) says that the derivative of the function G(x) = x2 − F (x) is nondecreasing.
It follows that (i) means that G is a convex function. On the other hand, it is clear that (ii) can be written
as
(ii’) 0 ≤ G(x + n1 ) − 2G(x) + G(x − n1 ), for all x ∈ R and all n ∈ N.
Now let us prove the equivalence of (i) and (ii).

(i) ⇒ (ii) By (i), G is a convex function, so Jensen’s inequality gives that

G(x + h) + G(x − h)
G(x) ≤ (x ∈ R, h > 0),
2
which can be written as

F (x + h) − 2F (x) + F (x − h) ≤ h2 (x ∈ R, h > 0).

Putting h = n1 , n ∈ N, in this inequality we obtain (ii).

(ii) ⇒ (i) We want to prove that (ii’) implies that G is convex. To do that we need the following
2
Theorem. A function g : R → R is convex if and only if it is continuous and D g(x) ≥ 0, for all x ∈ R,
2
where D g denotes the upper second symmetric derivative of g defined by

2 g(x + h) − 2g(x) + g(x − h)

D g(x) := lim sup (x ∈ R).
h→0+ h2

(See Theorem 1.4.7 in C. P. Niculescu and L.-P. Persson, Convex Functions and Their Applications - A
contemporary approach, Second Edition, CMS Books in Mathematics, Springer, 2018.)
Since we already know that G is continuous (because it is differentiable), we only have to show that (ii’)
2
implies that D G(x) ≥ 0, for all x ∈ R, and that is easy:

2 G(x + n1 ) − 2G(x) + G(x − n1 )

D G(x) ≥ lim sup 1 ≥0 (x ∈ R).
n→∞ n2

Also solved by Ángel Plaza, Universidad de Las Palmas de Gran Canaria, Spain; Theo Koupelis, Clark
College, Washington, USA.

Mathematical Reflections 6 (2024) 24

U684. Find all pairs (a, b) of relatively prime positive integers a and b with a > b, and such that n2 divides
an − bn for only finitely many positive integers n.

Proposed by Titu Andreescu, USA and Marian Tetiva, Romania

Solution by the authors

We show that n2 divides an − bn (whenever a and b are relatively prime positive integers for which a > b) for
only finitely many n if and only a − b = 1, so the required pairs are those of the form (x + 1, x) for positive
integer x.
For the ”if” part, assume that a − b = 1, and let n > 1 be a positive integer such that n2 divides an − bn
(clearly, the divisibility n2 | an − bn holds for n = 1), and let p be the smallest prime divisor of n. Note that
a − b = 1 implies that a and b have different parities, so n is odd — and, consequently, p is odd, too. We
have an ≡ bn mod p and, also, ap−1 ≡ bp−1 (≡ 1) mod p, according to Fermat’s little theorem. (Note that
neither a, nor b is divisible by p, since if any of them was so, the other would be so, too, via p | an − bn —
which is not possible as long as a and b are relatively prime.) Consequently (with gcd(x, y) denoting the
greatest common divisor of the integers x and y), we get

agcd(n,p−1) ≡ bgcd(n,p−1) mod p,

that is (because, n being odd and p being its smallest prime divisor, we have gcd(n, p − 1) = 1), a ≡ b mod p.
This is evidently absurd, so (in the case of a − b = 1) the divisibility of an − bn by n2 only holds for n = 1.
For the ”only if” part we further show that, whenever a and b are relatively prime positive integers such
that a − b ≥ 2, we can find infinitely many positive integers n such that n2 divides an − bn . In order to do
that we use one case of lifting the exponent lemma, whose proof we will skip (as it can already be considered
well-known), and one more helping result. Here and further vp (N ) denotes the exponent of the prime p in
the prime factorization of the nonzero integer N .
Lemma 1. (Lifting the exponent lemma) Let a and b be distinct integers, and let p be an odd prime that
divides a − b, but it does not divide any of a and b. Then we have

vp (an − bn ) = vp (a − b) + vp (n).

Note that if a and b are relatively prime, we do not need to assume that p does not divide a, or b. And here
is the second result, followed by its proof.
Lemma 2. Let a > b be relatively prime positive integers, and let p be an odd prime that divides a − b.
Then we can find an odd prime q such that q divides ap − bp , but it does not divide a − b. In particular, we
have q ̸= p.
Proof. By lifting the exponent lemma, we have vp (ap − bp ) = vp (a − b) + 1, and vπ (ap − bp ) = vπ (a − b) for
any prime π ̸= p that divides a − b. (Note that neither p, nor π divides a, or b, since if it did, a and b would
not be relatively prime anymore. Also, observe that, for π = 2, the equality v2 (ap − bp ) = v2 (a − b) follows
from the fact that ap−1 + ap−2 b + · · · + abp−2 + bp−1 is odd as long as a and b do not share the same parity;
and they don’t — see the final of the proof.) We can also write this equalities as
 p
a − bp
 p
a − bp
 
vp = 1 and vπ = 0.
a−b a−b

These observations imply that

ap − bp
= pt,
a−b
where t is a positive integer not divisible by any prime divisor of a − b (including p).
Since p ≥ 3, b > 0, and p | a − b, we also have

ap − bp
pt = = ap−1 + ap−2 b + · · · + abp−2 + bp−1 > ap−1 ≥ a > a − b ≥ p
a−b

Mathematical Reflections 6 (2024) 25

that is, t > 1. But then t surely has a prime divisor q, which is clearly different from all divisors of a − b.
Also, since
ap − bp
= ap−1 + ap−2 b + · · · + abp−2 + bp−1
a−b
is odd whenever a and b do not have the same parity (which is the case here, since they are relatively prime),
q is also odd, and the lemma is proved.
Now let a and b be relatively prime positive integers with a − b ≥ 2, and consider first that a − b has an
odd prime divisor p1 . Then, by lifting the exponent lemma, ap1 − bp1 is divisible by p21 , while, by lemma 2,
ap1 − bp1 has an odd prime divisor p2 ̸= p1 (actually distinct from any prime divisor of a − b). Again, lifting
the exponent lemma tells us that ap1 p2 − bp1 p2 is divisible by p22 ; but ap1 p2 − bp1 p2 is divisible by ap1 − ap2 , so
it is divisible by p21 too, consequently p21 p22 divides ap1 p2 − bp1 p2 . So far we found two distinct solutions for
n2 | an − bn , namely n = p1 and n = p1 p2 . Applying again lemma 2 for ap1 , bp1 , and p2 instead of a, b, and
p respectively we get the existence of a prime p3 that does not divide ap1 − bp1 (so it differs from p1 ), it is
different from p2 , and divides ap1 p2 −bp1 p2 . By lifting the exponent lemma once again we have that p23 divides
ap1 p2 p3 − bp1 p2 p3 ; but p21 p22 also divides ap1 p2 p3 − bp1 p2 p3 (because it divides ap1 p2 − bp1 p2 , which, on its turn,
is a divisor of ap1 p2 p3 − bp1 p2 p3 ), so for n = p1 p2 p3 we have again n2 | an − bn . Clearly, p1 < p1 p2 < p1 p2 p3 ,
thus we have by now three solutions for n2 | an − bn . Moreover, we can continue this procedure (in a similar
manner) in order to get an infinite sequence p1 , p2 , . . . of distinct odd primes such that each n = p1 · · · pj
(j = 1, 2, . . .) verifies n2 | an − bn — meaning that we achieved our goal in this case.
It remains to prove the same (that n2 | an − bn for infinitely many n) when a − b ≥ 2 has no odd prime
divisor. This means that a and b are now relatively prime positive integers such that a − b = 2k for some
positive integer k. Notice that a and b are necessarily odd, so we have a2 ≡ b2 ≡ 1 mod 4, yielding that
a2 − b2 (and, also, any a2m − b2m for positive integer m) is divisible by 4. If a + b has an odd prime divisor p1 ,
then a2 − b2 is divisible by p1 , thus we can apply the same procedure as above to a2 and b2 in the place of a
and b respectively in order to get infinitely many distinct odd primes p1 , p2 , . . . such that (p1 · · · pj )2 divides
(a2 )p1 ···pj − (b2 )p1 ···pj for all j = 1, 2, . . .. But this shows that for infinitely many n (namely n = 2p1 · · · pj
with j = 1, 2, . . .) we have n2 | an − bn , as desired.
Finally, assume that a − b = 2k for some k ≥ 1 and a + b has no odd prime divisor, that is a + b = 2l for
some positive integer l. Clearly l > k, so k ≥ 1 implies l ≥ 2. Then a = 2l−1 + 2k−1 , b = 2l−1 − 2k−1 , and
hence
a4 − b4 = (a − b)(a + b)(a2 + b2 ) = 2k · 2l · (22l−1 + 22k−1 ) = 23k+l−1 (22l−2k + 1),
implying
v2 (a4 − b4 ) = 3k + l − 1 ≥ 3 + 2 − 1 = 4.
This shows that a4 − b4 (and any a4m − b4m with positive integer m) is divisible by 16 and, also, that
a4 − b4 has an odd prime divisor p1 (because its factor 22l−2k + 1 is odd and greater than 1). Consequently,
proceeding analogously as above for a4 and b4 instead of a and b, we obtain a sequence of distinct primes
p1 , p2 , . . . such that every (p1 · · · pj )2 divides (a4 )p1 ···pj − (b4 )p1 ···pj (j = 1, 2, . . .). But this means that n2
divides an − bn for every n = 4p1 · · · pj (j = 1, 2, . . .), and our proof ends here.

Mathematical Reflections 6 (2024) 26

 Olympiad problems

O679. Let P be a point inside a circle Γ of center O. Line OP intersects circle Γ in points A and Q. A variable
chord through P intersects Γ in B and C. Prove that if OP ≤ P Q then ∠BAC ≥ π/3.

Proposed by Mircea Becheanu, Canada

Solution 1 by the author

The solution is based on the following well known fact: in a given circle, the longest chord cuts the longest
circular arc. When we speak about an arc cut by a chord we mean the smallest arc. We also mention the
following result:
Lemma. Let AQ be a diameter in a circle and P a point on AQ. The shortest chord through P is that one
which is perpendicular to the diameter AQ.
Proof of the Lemma. Assume that P is between O and Q. Let KL be the perpendicular in P to AQ and
let M N be a chord through P such that M lies on the shortest arc cut by KL, between K and Q. The
triangles KM P and N LP have equal angles, so they are similar. Therefore we have
KM MP KP
= = .
LN PL PN
Because P L < P N and KP = P L we have
KM KP
= < 1,
LN PN
hence KM < LN. It follows that arc(KM ) < arc(LN ) giving that arc(KQL) < arc(M QN ) and KL < M N .
Back to the problem, assume for beginning that OP = P Q and let BC the chord through P , perpendicular
to AQ. Then the triangle ABC is equilateral and ∠BAC = π/3. If B ′ C ′ is another chord through P we
have B ′ C ′ > BC and then ∠B ′ AC ′ > ∠BAC = π/3.
Assume that OP < P Q. Using the same notations like in the previous argument we have ∠BAC > π/3 and
similarly ∠B ′ AC ′ > π/3.

Solution 2 by Kousik Sett, India

Join A, B; A, C; O, B; O, C; Q, B; and Q, C. Let OP = x, P Q = y, BC = a, CA = b, and AB = c. Then
OQ = x + y = OA = OB = OC, AP = 2x + y and AQ = 2(x + y).
Since AQ is the diameter of ⊙(ABC), we have ∠ABQ = ∠ACQ = 90◦ . Therefore,

AB 2 + BQ2 = AC 2 + CQ2 = AQ2 i.e., c2 + BQ2 = b2 + CQ2 = 4(x + y)2 .

Since BQ > 0 and CQ > 0, we have 4(x + y)2 > c2 and 4(x + y)2 > b2 . So,

4(x + y)2 − bc > 0. (1)

Moreover,
BQ2 = 4(x + y)2 − c2 and CQ2 = 4(x + y)2 − b2 , (2)
which yields
BQ2 · CQ2 = 16(x + y)4 − (b2 + c2 ) · 4(x + y)2 + b2 c2 .
Since b2 + c2 ≥ 2bc, we have
BQ2 · CQ2 ≤ (4(x + y)2 − bc)2 .
By (1), it follows that
BQ · CQ ≤ 4(x + y)2 − bc. (3)

Mathematical Reflections 6 (2024) 27

The similarity of triangles ABP and CQP yields
AB AP c 2x + y
= ⇒ = .
CQ CP CQ CP
The similarity of triangles ACP and BQP yields
AC CP b CP
= ⇒ = .
BQ PQ BQ y
The above two relations give
 
bc 2x + y y
= ⇒ BQ · CQ = bc · . (4)
BQ · CQ y 2x + y

From (3) and (4), we have  

2x + y
bc ≤ (4(x + y)2 − bc),
y
which implies
bc ≤ 2(x + y)(2x + y). (5)
Applying Ptolemy’s Theorem on cyclic quadrilateral ABQC, we get

c · CQ + b · BQ = a · AQ.

By squaring and using (2) and (4), and simplifying, we obtain

c2 · CQ2 + b2 · BQ2 + 2bc · BQ · CQ = a2 · AQ2

 
2 2 2 2 2 2 y
⇒ c (AQ − b ) + b (AQ − c ) + 2bc · bc = a2 · AQ2
2x + y
   
2 2 2 2 2 2 2 2 y 2 2 2x
⇒ AQ (b + c − a ) = 2b c − 2b c = 2b c
2x + y 2x + y
2 2 2
  
b +c −a 2x 1
⇒ = 2bc .
bc 2x + y 4(x + y)2

Using (5), the above equation can be written as

b2 + c2 − a2
  
2x 1 2x
≤ 4(x + y)(2x + y) = .
bc 2x + y 4(x + y)2 x+y

Mathematical Reflections 6 (2024) 28

 2x
Since OP ≤ P Q, i.e., x ≤ y, we have ≤ 1, which yields
x+y

b2 + c2 − a2
≤ 1.
bc
Therefore, we have
b2 + c2 − a2 1
≤ , i.e, cos ∠BAC ≤ cos π/3,
2bc 2
which gives ∠BAC ≥ π/3 and we are done!
Observe that, the equality holds when b = c and x = y. These two conditions together imply that triangle
ABC is equilateral. Therefore, the equality holds only when triangle ABC is equilateral.

Also solved by Theo Koupelis, Clark College, Washington, USA.

Mathematical Reflections 6 (2024) 29

O680. Let a ≥ b ≥ 1 ≥ c ≥ 0 such that ab + bc + ca = 3. Prove that
1 1 1 3
+ 2 + 2 ≤ .
3a2 + 5 3b + 5 3c + 5 8

Proposed by Vasile Cı̂rtoaje, Oil-Gas Universty of Ploieşti, Romania

Solution by the author

(a + b)2
Denote x = and p = ab. From (a−1)(b−1) ≥ 0, we obtain a+b ≤ ab+1, hence (a+b)2 ≤ (ab+1)2 .
4
Therefore,
(p + 1)2
1≤p≤x≤ .
4
Since
1 1 3(a2 + b2 ) + 10 2(6x − 3p + 5)
2
+ 2 = 2 2 =
3a + 5 3b + 5 9a b + 15(a2 + b2 ) + 25 60x + 9p2 − 30p + 25
and
(3 − ab)2 (3 − p)2
c2 = = ,
(a + b)2 4x
the desired inequality can be written as F ≤ 0, where
6x − 3p + 5 2x 3
F = + − .
60x + 9p2 − 30p + 25 20x + 3p2 − 18p + 27 16

Moreover, for fixed p, the inequality F ≤ 0 is equivalent to f (x) ≤ 0, where

f (x) = −240x2 + B(p)x + C(p).

Since f (x) is concave and x ∈ [p, (p + 1)2 /4], it suffices to prove that F ≥ 0 for x = p and x = (p + 1)2 /4.
For x = p, we have

1 2p 3 27(1 − p)3
F = + 2 − = ≤ 0,
3p + 5 3p + 2p + 27 16 16(3p + 5)(3p2 + 2p + 27)

and for x = (p + 1)2 /4, we have

3p2 + 13 p2 + 2p + 1 −3(p − 1)4

16F = + − 3 = ≤ 0.
3p2 + 5 p2 − p + 4 (3p2 + 5)(p2 − p + 4)

The proof is completed. The equality occurs for a = b = c = 1.

Also solved by Nicuşor Zlota, “Traian Vuia” Technical College, Focşani, Romania; Paolo Perfetti, Università
degli studi di Tor Vergata Roma, Italy; Theo Koupelis, Clark College, Washington, USA.

Mathematical Reflections 6 (2024) 30

O681. Let x, y, z be positive real numbers. Prove that
s
1 1 1 9(x + y + z)
√ +√ +√ ≤ .
x+y y+z z+x 2(xy + yz + zx)

Proposed by Marius Stănean, Zalău, Romania

Solution 1 by Alessandro Ventullo, Milan, Italy

1 1 1
Since 3(a2 + b2 + c2 ) ≥ (a + b + c)2 for any real numbers a, b, c, setting a = √
,b= √ ,c= √ ,
x+y y+z z+x
we get
   2
1 1 1 1 1 1
3 + + ≥ √ +√ +√ .
x+y y+z z+x x+y y+z z+x
Therefore it is enough to prove that
s s  
9(x + y + z) 1 1 1
≥ 3 + + .
2(xy + yz + zx) x+y y+z z+x
By simplifying, this is equivalent to
3 x3 y + xy 3 + y 3 z + yz 3 + z 3 x + zx3 − 2xyz(x + y + z)


≥ 0,
2(xy + yz + zx)(x + y)(y + z)(z + x)
which is true because by the AM-GM Inequality we have
1X 3
(x y + x3 z + xy 3 + xz 3 − 4x2 yz) ≥ 0.
2 cyc

Equality holds if and only if x = y = z.

Solution 2 by Theo Koupelis, Clark College, Washington, USA

The desired inequality is equivalent to
√ Y√ p Xp
3 x+y+z x + y ≥ 2(xy + yz + zx (x + y)(y + z).
c c
√
The function f (t) =
t is clearly concave for t > 0, and thus, using Jensen’s inequality we get
rP r
Xp
c (x + y)(y + z) x2 + y 2 + z 2
(x + y)(y + z) ≤ 3 = 3 xy + yz + zx + .
c
3 3

Therefore, it suffices to show that

r
√ Y√ p x2 + y 2 + z 2
x+y+z x + y ≥ 2(xy + yz + zx) xy + yz + zx + .
c
3

Squaring and simplifying we get

x3 y + xy 3 + y 3 z + yz 3 + z 3 x + zx3 ≥ 2xyz(x + y + z),
or equivalently
1X 3
(x y + x3 z + xy 3 + xz 3 − 4x2 yz) ≥ 0,
2 c
which is obvious by AM-GM. Equality occurs when x = y = z.

Also solved by Arkady Alt, San Jose, California, USA; Nicuşor Zlota, “Traian Vuia” Technical College,
Focşani, Romania; Paolo Perfetti, Università degli studi di Tor Vergata Roma, Italy; Titu Zvonaru, Comăneşti,
Romania.

Mathematical Reflections 6 (2024) 31

O682. Let a, b, c be positive real numbers. Prove that

a2 + b2 + c2 a+b+c
≥√ √ √ .
ab + bc + ca ab + bc + ca

Proposed by Nguyen Viet Hung, Hanoi University of Science, Vietnam

Solution 1 by the author

By GM-HM and Cauchy-Schwarz inequality we have
X√ X 2ab X 2(ab)2 2(ab + bc + ca)2
ab ≥ = ≥ .
cyc cyc
a+b cyc
ab(a + b) ab(a + b) + bc(b + c) + ca(c + a)

Therefore
a+b+c (a + b + c)[ab(a + b) + bc(b + c) + ca(c + a)]
√ √ √ ≤ .
ab + bc + ca 2(ab + bc + ca)2
Now we need to show that
(a + b + c)[ab(a + b) + bc(b + c) + ca(c + a)] a2 + b2 + c2
2
≤ .
2(ab + bc + ca) ab + bc + ca
This is equivalent to

(a + b + c)[ab(a + b) + bc(b + c) + ca(c + a)] ≤ 2(ab + bc + ca)(a2 + b2 + c2 ),

X X
ab(a + b)2 ≤ 2ab(a2 + b2 ),
cyc cyc
X
ab(a − b)2 ≥ 0.
cyc

The last inequality is clearly true and the proof is completed.

Solution 2 by Theo Koupelis, Clark College, Washington, USA

Let a = x2 , b = y 2 , c = z 2 , where x, y, z are positive real numbers. The desired inequality is equivalent to

(x4 + y 4 + z 4 )(xy + yz + zx) ≥ (x2 y 2 + y 2 z 2 + z 2 x2 )(x2 + y 2 + z 2 ),

or after expanding and simplifying,

X
xyz(x3 + y 3 + z 3 − 3xyz) + (x5 y + xy 5 − x4 y 2 − x2 y 4 ) ≥ 0,
c

or X
xyz(x3 + y 3 + z 3 − 3xyz) + xy(x − y)2 (x2 + xy + y 2 ) ≥ 0,
c
which is obvious by AM-GM. Equality occurs when a = b = c. If we were to allow at most one of the
variables a, b, c to be zero, equality would also occur at (a, b, c) = (a, a, 0) and cyclic permutations.

Solution 3 by Daniel
√ Pascuas, Barcelona, Spain
√ √
Let x = a, y = b, z = c and write the inequality as

x4 + y 4 + z 4 x2 + y 2 + z 2
≥ ,
x2 y 2 + y 2 z 2 + z 2 x2 xy + yz + zx
or equivalently
(x2 y 2 + y 2 z 2 + z 2 x2 )(x2 + y 2 + z 2 ) ≤ (x4 + y 4 + z 4 )(xy + yz + zx). (∗)

Mathematical Reflections 6 (2024) 32

Now (x2 y 2 + y 2 z 2 + z 2 x2 )(x2 + y 2 + z 2 ) = A + 3x2 y 2 z 2 and (x4 + y 4 + z 4 )(xy + yz + zx) = B + C,
where A = x4 (y 2 + z 2 ) + y 4 (x2 + z 2 ) + z 4 (x2 + y 2 ), B = x5 (y + z) + y 5 (x + z) + z 5 (x + y) and
4 4 4
C = x yz + xy z + xyz . But, since
3 1 3 1 3 1 3 1 3 1 3 1
A = (x5 y) 4 (xy 5 ) 4 + (x5 z) 4 (xz 5 ) 4 + (y 5 x) 4 (yx5 ) 4 + (y 5 z) 4 (yz 5 ) 4 + (z 5 x) 4 (zx5 ) 4 + (z 5 y) 4 (zy 5 ) 4 ,
4
Hölder’s inequality with exponents p = 3 and q = 4 gives that A ≤ B, while the AM-GM inequality shows
that 3x2 y 2 z 2 ≤ C, and hence (∗) holds.

Also solved by Ángel Plaza, Universidad de Las Palmas de Gran Canaria, Spain; Arkady Alt, San Jose, Cal-
ifornia, USA; Nicuşor Zlota, “Traian Vuia” Technical College, Focşani, Romania; Marin Chirciu, Colegiul
National Zinca Golescu, Piteşti, Romania; Nicuşor Zlota, “Traian Vuia” Technical College, Focşani, Ro-
mania; Paolo Perfetti, Università degli studi di Tor Vergata Roma, Italy; Israel Castillo Pilco, Universidad
Nacional de Ingenierı́a, Peru.

Mathematical Reflections 6 (2024) 33

O683. Let a, b, c be positive real numbers such that a + b + c = 1. Prove that
b+c c+a a+b
√ +√ +√ ≥ 2.
2
bc + 2a + 2a 2
ca + 2b + 2b ab + 2c2 + 2c

Proposed by An Zhenping, Xianyang National University, China

Solution 1 by Ángel Plaza, Universidad de Las Palmas de Gran Canaria, Spain

b+c 2
 
By the AM-GM inequality bc ≤ , and so
2

b+c b+c 2(b + c)

√ ≥q =p
bc + 2a2 + 2a b+c 2


+ 2a2 + 2a (b + c)2 + 8a2 + 8a
2

and because b + c = 1 − a, it follows that

b+c 2 − 2a 2 − 2a
√ ≥√ = .
2
bc + 2a + 2a 2
9a + 6a + 1 3a + 1

Then, applying the same argument to the left-hand side terms of the proposed inequality, it is enough to
prove that
1−a 1−b 1−c
+ + ≥ 1.
3a + 1 3b + 1 3c + 1
1−x 18
Function f (x) = is convex because f ′′ (x) = > 0. Therefore, by Jensen’s inequality
3x + 1 (3x + 1)2

1−a 1−b 1−c 1 − 1/3

+ + ≥3· = 1.
3a + 1 3b + 1 3c + 1 3(1/3) + 1
1
Equality occurs if and only if a = b = c = .
3
Solution 2 by Theo Koupelis, Clark College, Washington, USA
Using the condition a + b + c = 1 and AM-GM, we get
X b+c X b+c X 2(1 − a)
√ ≥ q = p
c bc + 2a2 + 2a c
(b+c)2
+ 2a(a + 1) c (1 − a)2 + 8a(a + 1)
4
X 2(1 − a) X 2 − 3a X
= ≥ 2· =4−2 a = 2.
c
3a + 1 c
3 c

In the above we used the tangent-line trick to get

1−x 2 − 3x
≥ ⇐⇒ (3x − 1)2 ≥ 0,
3x + 1 3
which is obvious for x > 0. Equality occurs when a = b = c = 1/3.

Solution 3 by Arkady Alt, San Jose, California, USA

Noting that
bc + 2a2 + 2a = bc + 2a2 + 2a (a + b + c) = (2a + c) (2a + b)
and
((2a + c) + (2a + b))2 (4a + b + c)2
(2a + c) (2a + b) ≤ =
4 4

Mathematical Reflections 6 (2024) 34

we obtain
X b+c X b+c
√ ≥2 .
cyc bc + 2a2 + 2a cyc
4a + b + c

Thus it remains to prove that

X b+c
≥ 1.
cyc
4a + b + c

By Cauchy-Schwarz inequality,
X b+c X (b + c)2
=
cyc
4a + b + c cyc
(b + c) (4a + (b + c))
!2
P
(b + c)
cyc
≥ P
(b + c) (4a + (b + c))
cyc
!2
P
4 a
cyc
= !2
P P
2 a +6 ab
cyc cyc
2
= P
1 + 3 ab
cyc

and !2
X X
3 ab ≤ a = 1.
cyc cyc

Then
X b+c 2 2
≥ P ≥ = 1.
cyc
4a + b + c 1 + 3 ab 1+1
cyc

Also solved by Cristian Muth, UMFST Highschool in G. E. Palade University of Medicine, Pharmacy,
Science, and Technology, Romania; Marin Chirciu, Colegiul National Zinca Golescu, Piteşti, Romania;
Nicuşor Zlota, “Traian Vuia” Technical College, Focşani, Romania; Paolo Perfetti, Università degli studi
di Tor Vergata Roma, Italy; Titu Zvonaru, Comăneşti, Romania; Daniel Pascuas, Barcelona, Spain; Israel
Castillo Pilco, Universidad Nacional de Ingenierı́a, Peru.

Mathematical Reflections 6 (2024) 35

O684. A sequence (an )n≥1 of positive integers is called admissible if for all n > 2 the divisibility

1 + a1 + · · · + an−1 | (1 + a1 )(1 + a2 ) · · · (1 + an−1 ) − 1

holds precisely when n is a prime number. Let f be a nonconstant polynomial with integer coefficients.

(a) Can the sequence (an )n≥1 defined by an = f (n) be admissible?

(b) Can the sequence (an )n≥1 defined by an = f (2n ) be admissible?

Proposed by Gabriel Dospinescu, Lyon, France

Solution by the author

We refer the reader to the previous proposed problem for part b), which shows that taking f (X) = X gives
an admissible sequence an = 2n . Let us focus on part a) and let us suppose that an = f (n) is admissible.
By a standard application of the binomial formula, we see that there is a polynomial g ∈ Q[X] such that
g(n) = 1 + f (1) + · · · + f (n − 1) for all n ≥ 1. Note that since f is not constant, g has degree at least 2.
Let N > 1 be a positive integer such that h := N g has integer coefficients and let A = 1 + f . Then A, h are
nonconstant, have integer coefficients and h(p) | N A(1)A(2) · · · A(p − 1) − N for all odd primes p.
First, we claim that h(0) ̸= 0. Indeed, otherwise p divides h(p) = 1 + f (1) + · · · + f (p − 1) for all primes p
and so p | g(p) for all primes p > N . By a classical result p also divides f (0) + f (1) + · · · + f (p − 1) for all
large enough primes (for instance all p > deg f ), so that necessarily f (0) = 1 and A(0) = 2. On the other
Q theorem we can find a prime p > N for which the congruence A(x) ≡ 0 (mod p) is solvable.
hand, by Schur’s
Since p | N p−1k=1 A(k) − N , p cannot divide any of A(1), · · · , A(p − 1) and so necessarily p | A(0) = 2, a
contradiction. This shows that h(0) ̸= 0.
Since A and h are nonconstant, a classical but nontrivial result ensures that there are infinitely many primes
p for which both congruences h(x) ≡ 0 (mod p) and A(y) ≡ 0 (mod p) are solvable. Pick a large prime
p > N which does not divide h(0) and for which p | h(x) for some x. Necessarily p is prime Q to x, so by
Dirichlet’s theorem there is a prime q > p of the form q = x + pk. We have p | h(q) | N q−1 k=1 A(k) − N .
But this is impossible since by assumption there is k ∈ {1, 2, · · · , q − 1} with p | A(k), and thus p | N ,
contradicting the inequality p > N .

Mathematical Reflections 6 (2024) 36