Junior problems

J685. Let a be a nonzero real number. Prove that

 
1 1 1
a3 − + a2
+ − 3 a − =3
a3 a2 a

if and only if    
5 1 3 1 1
a − 5 − 5 a − 3 + 11 a − = 1.
a a a

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

Solution by Daniel Pascuas, Barcelona, Spain

1
Let x = a − . Then
a
     
2 2 1 3 3 1 1 5 5 1 3 1 1
x = a + 2 − 2, x = a − 3 − 3 a − and x = a − 5 − 5 a − 3 + 10 a − ,
a a a a a a
so  
3 1 2 1 1
a − 3 +a + 2 −3 a− = x3 + x2 + 2,
a a a
and    
1 1 1
a5 − − 5 a3
− + 11 a − = x5 + x.
a5 a3 a
Therefore  
3 1 2 1 1
a − 3 +a + 2 −3 a− = 3 ⇐⇒ x3 + x2 − 1 = 0
a a a
and    
5 1 3 1 1
a − 5 − 5 a − 3 + 11 a − = 1 ⇐⇒ x5 + x − 1 = 0.
a a a
Since x5 + x − 1 = (x2 − x + 1)(x3 + x2 − 1) = (x − 12 )2 + 34 (x3 + x2 − 1) and x ∈ R, it is clear that

x5 + x − 1 = 0 is equivalent to x3 + x2 − 1 = 0. Hence the two identities in the statement are equivalent.

Also solved by Arkady Alt, San Jose, California, USA; Brian Bradie, Christopher Newport University,
Newport News, VA, USA; G. C. Greubel, Newport News, VA, USA; Ivan Hadinata, Department of Math-
ematics, Gadjah Mada University, Indonesia; Kousik Sett, India; Nicuşor Zlota, “Traian Vuia” Technical
College, Focşani, Romania; Paolo Perfetti, Università degli studi di Tor Vergata Roma, Italy; Raja Ok-
tovin Parhasian Damanik, Canberra ACT, Australia; Sundaresh H. R., Shivamogga, Karnataka, India;
Telemachus Baltsavias, Kerameies Junior High School, Kefallonia, Greece; Theo Koupelis, Clark College,
Washington, USA; Aaditya Bhaumik Sahu, AMMOC, India; Anderson Torres, Sao Paulo, Brazil; Nikolai
N. Mnev, AMMOC, Bengaluru, India; Polyahedra, Polk State College, FL, USA; Samhith Kadiyala, North
Carolina Cyber Academy, NC, USA; Srijan Sundar, Oxford, UK; Sumanyu Nandecha, AMMOC, India.

Mathematical Reflections 1 (2025) 1

J686. Let a and b be positive real numbers. Prove that
p p p
x2 − ax + a2 + x2 − bx + b2 ≥ a2 + ab + b2
for all real numbers x. When does equality occur?
Proposed by Mircea Becheanu, Canada

Solution 1 by Raja Oktovin Parhasian Damanik, CanberrapACT, Australia p √

If x ≤ 0, p
let x = −y with yp≥ 0. The inequality 2 2 2 2 2 2
p becomes y +pay + a + y + by + b ≥ a + ab + b .
Clearly, y 2 + ay + a2 + y 2 + by + b2 ≥ y 2 − ay + a2 + y 2 − by + b2 . So, it suffices to prove the
original inequality for positive real number x.
Let P A, P B, P C be rays with P A = a, P B = x, and P C = x such that ∠AP B = ∠BP C = 60◦ and
∠AP C = 120◦ .
By cosine law, we have
AB 2 = P A2 + P B 2 − 2 · P A · P B · cos 60◦ = a2 + x2 − ax
BC 2 = P B 2 + P C 2 − 2 · P B · P C · cos 60◦ = x2 + b2 − bx
CA2 = P A2 + P C 2 − 2 · P A · P C · cos 120◦ = a2 + b2 + ab.
√ √ √
By triangle inequality, we have AB +BC ≥ CA, and hence x2 − ax + a2 + x2 − bx + b2 ≥ a2 + ab + b2 .
The equality occurs when B lies on segment AC, and since P B bisects ∠ABC, we calculate P B using
[AP B] + [BP C] = [AP C], i.e.
1 √ 1 √ 1 √ ab
ax 3 + bx 3 = ab 3 ⇐⇒ x = .
4 4 4 a+b

Solution 2 by Daniel Pascuas, Barcelona, Spain

Let d be the Euclidean distance on R2 , i.e. d (x1 , y1 ), (x2 , y2 ) = (x1 − x2 )2 + (y1 − y2 )2 . Then our

p

inequality directly follows from the triangle inequality for d:

p p √ √

x2 − ax + a2 + x2 − bx + b2 = d ( a2 , a 2 3 ), (x, 0) + d (x, 0), ( 2b , − b 2 3 )


√ √
q
2 2
≥ d ( 2 , 2 ), ( 2 , − 2 ) = (a−b)
a a 3 b b 3
4 + 3(a+b) 4
p
= a2 + ab + b2 .
√ √
The equality holds if and only if the point (x, 0) belongs to the segment determined by ( a2 , a 2 3 ) and ( 2b , − b 2 3 ),
ab ab
which means that x = a+b . Thus the equality only occurs when x = a+b .
Solution 3 by Polyahedra, Polk State College, FL, USA
By squaring both sides we see that the inequality is equivalent to
p
2 (x2 − ax + a2 ) (x2 − bx + b2 ) ≥ ab + (a + b)x − 2x2 .
By the Cauchy-Schwarz inequality,
p p
(2x2 − 2ax + 2a2 ) (2x2 − 2bx + 2b2 ) = [a2 + (a − x)2 + x2 ] [b2 + x2 + (b − x)2 ]
≥ ab + (a − x)x + x(b − x) = ab + (a + b)x − 2x2 .
b x b−x ab
Equality holds if and only if a = a−x = x , which is equivalent to x = a+b .

Also solved by Arkady Alt, San Jose, California, USA; Daniel Văcaru, National Economic College “Maria
Teiuleanu”, Pites, ti, Romania; Ioan Viorel Codreanu, Satulung, Maramures, Romania; Ivan Hadinata, De-
partment of Mathematics, Gadjah Mada University, Indonesia; Kousik Sett, India; 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; Sundaresh H. R., Shivamogga,
Karnataka, India; Theo Koupelis, Clark College, Washington, USA; Aaditya Bhaumik Sahu, AMMOC, In-
dia; Anderson Torres, Sao Paulo, Brazil; Nihad Hashimov, AMMOC, Azerbaijan; Srijan Sundar, Oxford,
UK; Sumanyu Nandecha, AMMOC, India.

Mathematical Reflections 1 (2025) 2

J687. Let a, b, c be real numbers such that ab + bc + ca = 169 and

(a + 1)3 + (b + 1)3 + (c + 1)3 − 3abc = 1.

Find all possible values of a + b + c.

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

Solution by Ivan Hadinata, Department of Mathematics, Gadjah Mada University, Indonesia

Let x = a + b + c and y = ab + bc + ca = 169. Then

0 = (a + 1)3 + (b + 1)3 + (c + 1)3 − 3abc − 1

= (a3 + b3 + c3 − 3abc) + 3(a2 + b2 + c2 ) + 3(a + b + c) + 2
= x(x2 − 3y) + 3(x2 − 2y) + 3x + 2
= x3 + 3x2 − 504x − 1012
= (x + 2)(x + 23)(x − 22),

hence x = −2 or x = −23 or x = 22. Since a, b, c ∈ R, then x2 ≥ 3y = 507, and the only possibility is
x = −23. One example of (a, b, c) when x = −23 is
√ √
a = −8 + 7, b = −8 − 7, c = −7.

In conclusion, the only possible value of a + b + c is −23.

Editor’s Comment. Many readers failed to check the condition (a + b + c)2 ≥ 3(ab + bc + ca).
Raja Oktovin Parhasian Damanik found another triple (a, b, c) satisfying the conditions, arguing as follows.
Take a = −x, b = −y, c = −z, then x + y + z = 23 and√xy + yz + zx = 169. Take z = 5,√then x + y =√18
and xy = 79. This has solution, for example, x, y = 9 ± 2. Hence we can take a = −9 − 2, b = −9 + 2,
and c = −5.
Theo Koupelis noticed that there are infinitely many solutions that satisfy the given equations, arguing as
follows.
We have the system b + c = −23 − a and bc = 169 − a(b + c) = a2 + 23a + 169; thus, b, c are the solutions
of the quadratic t2 + (23 + a)t + (a2 + 23a + 169) = 0, whose
√ discriminant is √ ∆ = −(3a2 + 46a + 147). For
−23 − 2 22 −23 + 2 22
real solutions we must have ∆ ≥ 0; thus, for any ≤a≤ , the values of b and c are
p 3 3
−23 − a ± −(3a2 + 46a + 147)
given by .
2
Also solved by Arkady Alt, San Jose, California, USA; G. C. Greubel, Newport News, VA, USA; Ioan
Viorel Codreanu, Satulung, Maramures, Romania; Kalkin Gandhi, AMMOC Bengaluru, India; Marin Chir-
ciu, 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; Sundaresh H. R., Shivamogga, Karnataka, India; Telemachus Balt-
savias, Kerameies Junior High School, Kefallonia, Greece; Theo Koupelis, Clark College, Washington, USA;
Vishwesh Ravi Shrimali, Jaipur, Rajasthan, India; Aaditya Bhaumik Sahu, AMMOC, India; Daniel Pas-
cuas, Barcelona, Spain; Diar Gashi, University of Primorska, Slovenia; Madhav R Pillai, Delhi Public
School, Ranchi, Jharkhand, India; Medha Shrivastava, AMMOC,India; Nihad Hashimov, AMMOC, Azer-
baijan; Nikolai N. Mnev, AMMOC, Bengaluru, India; Polyahedra, Polk State College, FL, USA; Samhith
Kadiyala, North Carolina Cyber Academy, NC, USA; Srijan Sundar, Oxford, UK.

Mathematical Reflections 1 (2025) 3

J688. Prove that positive real numbers x, y, z are the lengths of the altitudes in a triangle if and only if
 
1 1 1 1 1 1
+ + > 2 max , , .
x y z x y z
Proposed by Dan-Ştefan Marinescu, Romania
Solution 1 by Raja Oktovin Parhasian Damanik, Canberra ACT, Australia
Suppose that a triangle has altitudes of lengths x, y, z which are respectively perpendicular to the sides
of length a, b, c. The area [ABC] satisfies 2[ABC] = ax = by = cz. By triangle inequality, we have that
a + b + c > 2max{a, b, c}. Hence,
 
[ABC] [ABC] [ABC] [ABC] [ABC] [ABC]
+ + > 2 max , ,
x y z x y z
 
1 1 1
= 2[ABC] max , ,
x y z
Hence,
 
1 1 1 1 1 1
+ + > 2 max , , .
x y z x y z
1 1 1
Now let x, y, z satisfy the inequality and hence , , are sides of a triangle; let A be the area of that
x y z
1 1 1
triangle. There exists a triangle whose sides are of length , , . The area of this triangle is
2Ax 2Ay 2Az
1
1 1 1 2·
·A = . Hence, the length of the altitude to side of the length will be 4A = x. Similarly,
4A2 4A 2Ax 1
2Ax
the other two altitudes will have length y and z, as desired.

Solution 2 by Daniel Pascuas, Barcelona, Spain

Let ABC be a triangle and, as usual, let a, b, c be the lengths of the sides BC, AC, AB, respectively. Let
x, y, z be the lengths of the altitudes of ABC corresponding to the sides BC, AC, AB, respectively. Then
xa = yb = zc = 2S, where S is the area of the triangle ABC. Therefore a = 2S 2S 2S
x , b = y , and c = z must
satisfy that a + b + c > 2 max{a, b, c}, which means that
 
1 1 1 1 1 1
+ + > 2 max , , .
x y z x y z
Conversely, assume that x, y, z are positive real numbers satisfying this inequality. Then
    
1 1 1 1 1 1 1 1 1 1 1 1
Λ := + + + − + − + − > 0.
x y z y z x x z y x y z
Thus a = x√2 Λ , b = y√2 Λ , c = z √2 Λ are positive real numbers which satisfy a + b + c > 2 max{a, b, c}, and so
they are the lengths of the sides BC, AC, AB, respectively, of a triangle ABC. The semiperimeter of this
triangle is p = 21 (a + b + c) = √1Λ ( x1 + y1 + z1 ), so Heron’s formula shows that its area is equal to
s    
p 1 1 1 1 1 1 1 1 1 1 1 1 1 1
p(p − a)(p − b)(p − c) = + + + − + − + − =√ .
Λ x y z y z x x z y x y z Λ
2 2 2
Hence the lengths of the altitudes corresponding to the sides BC, AC, AB are √
a Λ
= x, √
b Λ
= y, √
c Λ
= z,
respectively.

Also solved by Nicuşor Zlota, “Traian Vuia” Technical College, Focşani, Romania; Theo Koupelis, Clark
College, Washington, USA; Anderson Torres, Sao Paulo, Brazil; Diar Gashi, University of Primorska,
Slovenia; Medha Shrivastava, AMMOC,India; Polyahedra, Polk State College, FL, USA; Srijan Sundar,
Oxford, UK.

Mathematical Reflections 1 (2025) 4

J689. Find the least value of the function
2 1 2
f (x) = + 2 − 2 ,
x2 − 4x + 8 x − 6x + 10 x − 5x + 8
where x is any real number.

Proposed by An Zhenping, Xianyang Normal University, China

Solution 1 by Raja Oktovin Parhasian Damanik, Canberra ACT, Australia

We claim that f (x) ≥ 0 for all real numbers x.
We want to prove that
2 1 2
+ ≥ 2 .
x2 − 4x + 8 x2 − 6x + 10 x − 5x + 8
2 1 2
Note that if x = 2, then LHS = + = 1 and RHS = = 1, so the equality can occur.
4 2 2
To prove the inequality, we use AM-HM inequality:
1 1 4
1 2 + 2 ≥ 3 2 .
2x − 2x + 4 x − 6x + 10 2 x − 8x + 14

3 2 1
It suffices to prove that 2x2 − 10x + 16 ≥ x − 8x + 14. Rearranging, we have x2 − 2x + 2 ≥ 0, i.e.
2 2
x2 − 4x + 4 ≥ 0, i.e. (x − 2)2 ≥ 0 which is true.
Hence, the least value of f (x) is 0, and it is attained by x = 2.

Solution 2 by Daniel Pascuas, Barcelona, Spain

Just note that
1 2x
f (x) = − 2
x2 − 6x + 10 (x − 4x + 8)(x2 − 5x + 8)
(x2 − 4x + 8)(x2 − 5x + 8) − 2x(x2 − 6x + 10)
=
(x2 − 4x + 8)(x2 − 5x + 8)(x2 − 6x + 10)
x4 − 11x3 + 48x2 − 92x + 64
=
(x2 − 4x + 8)(x2 − 5x + 8)(x2 − 6x + 10)
(x − 2)2 (x2 − 7x + 16)
=
(x2 − 4x + 8)(x2 − 5x + 8)(x2 − 6x + 10)
(x − 2)2 (x − 72 )2 + 15


4
= ≥ 0 = f (2), for every x ∈ R,
(x − 2)2 + 4 (x − 52 )2 + 74 ((x − 3)2 + 1)

so the the least value of f (x) is 0.

Also solved by Arkady Alt, San Jose, California, USA; Henry Ricardo, Westchester Area Math Circle, New
York, USA; Kalkin Gandhi, AMMOC Bengaluru, India; 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; Aaditya Bhaumik Sahu,
AMMOC, India; Karma Pillai; Polyahedra, Polk State College, FL, USA; Sumanyu Nandecha, AMMOC,
India.

Mathematical Reflections 1 (2025) 5

J690. Let ABC be a triangle. The angle bisector of ∠CAB intersects BC in D and the circumcircle in X.
Let Y be the orthogonal projection of C on AX and Z the orthogonal projection of D on AB. The
line Y Z intersects BC in E. Prove that AE is perpendicular to BC.

Proposed by Mihaela Berindeanu, Bucharest, Romania

Solution 1 by Raja Oktovin Parhasian Damanik, Canberra ACT, Australia

Let ∠BAC, ∠ABC, ∠BCA are denoted by α, β, γ, respectively.
We want to prove that ∠AEC = ∠AY C = 90◦ , that is, we want to prove that AEY C is a cyclic quadrilateral.
α α
Note that ∠DAE = ∠DAB − ∠BAE = − (90◦ − β) = + β − 90◦ .
2 2 α
Also, note that ∠ECY = ∠DCY = ∠ACY − ∠ACB = 90◦ − − γ.
2
α α
But it is clear now that + β − 90◦ = 90◦ − − γ, i.e. α + β + γ = 180◦ , which is true for angles of a
2 2
triangle α, β, γ.
So, ∠Y AE = ∠DAE = ∠ECY , which implies that AEY C is a cyclic quadrilateral.

Mathematical Reflections 1 (2025) 6

Solution 2 by Diar Gashi, University of Primorska, Slovenia
Let E ′ be the orthogonal projection of A onto BC. Since ∠AE ′ C = 90 = ∠AY C, we have that AE ′ Y C is
cyclic. So,
∠AE ′ Y = 90 + ∠CE ′ Y = 90 + ∠CAY = 90 + ∠BAD.
Since ∠AZD = 90 = ∠AE ′ D, we have that AZE ′ D is cyclic. Thus,

∠AE ′ Z = ∠ADZ = 90 − ∠BAD.

Therefore ∠AE ′ Y + ∠AE ′ Z = 180 and this means that Z, E ′ and Y are collinear.
So, E = E ′ and therefore AE is indeed perpendicular to BC.

Mathematical Reflections 1 (2025) 7

Solution 3 by Kousik Sett, India
Let the circumcircle of triangle ACY intersect BC at E ′ . Then ∠AE ′ C = ∠AY C = 90◦ .
Also, ∠E ′ Y A = ∠E ′ CA = ∠BCA = ∠BXA, which implies E ′ Y ∥ BX.
Since ∠DE ′ A = ∠DZA = 90◦ , quadrilateral AZE ′ D is cyclic. So, ∠BE ′ Z = ∠ZAD = ∠BAX.

Since the angle bisector of ∠CAB intersects the circumcircle of triangle ABC at X, we infer that X is the
midpoint of arc BC not containing A. So, ∠XBC = ∠BCX = ∠BAX.
Therefore, we obtain ∠BE ′ Z = ∠XBC, which implies ZE ′ ∥ BX.
Since E ′ Y ∥ BX and ZE ′ ∥ BX, it follows that Z, E ′ , and Y are collinear, and E ′ = BC ∩ ZY . However,
it is given that E = BC ∩ ZY . Therefore, we conclude that E ′ = E and consequently, we have AE is
perpendicular to BC.

Also solved by Theo Koupelis, Clark College, Washington, USA; Karma Pillai; Polyahedra, Polk State
College, FL, USA; Prasanna Pawar, AMMOC, Bengaluru, India.

Mathematical Reflections 1 (2025) 8

 Senior problems

S685. Let a, b be positive real numbers. Prove that

a b a+b
2
+ 2
≥ .
(1 + b) (1 + a) 2(1 + ab)

Proposed by An Zhenping, Xianyang Normal University, China

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

By Cauchy-Schwarz Inequality, we have
 
b a 
(1 + b)2 ≤ + 1 (ab + 1) , (1 + a)2 ≤ + 1 (ab + 1) .
a b

Then
a b a b
2 + ≥ + a
(1 + a)2 b



(1 + b) a + 1 (ab + 1) b + 1 (ab + 1)
 2 2

1 a b
= +
ab + 1 a + b a + b
1
· a2 + b2


=
(ab + 1) (a + b)
1 (a + b)2
≥ ·
(ab + 1) (a + b) 2
a+b
= .
2 (1 + ab)

The equality holds if and only if a = b = 1.

Solution 2 by Ivan Hadinata, Department of Mathematics, Gadjah Mada University, Indonesia

By the AM-GM Inequality, a + b + a2 b + ab2 ≥ 4ab. Hence

2(a + b)(1 + ab) ≥ a + b + 4ab + a2 b + ab2 = a(1 + b)2 + b(1 + a)2 .

Then, by Cauchy-Schwarz Inequality we have

a b a2 b2 (a + b)2 (a + b)2 a+b

2
+ 2
= 2
+ 2
≥ 2 2
≥ = ,
(1 + b) (1 + a) a(1 + b) b(1 + a) a(1 + b) + b(1 + a) 2(a + b)(1 + ab) 2(1 + ab)

as wanted. The equality holds if and only if a = b = 1.

Also solved by Daniel Văcaru, National Economic College “Maria Teiuleanu”, Pites, ti, Romania; G. C.
Greubel, Newport News, VA, USA; Henry Ricardo, Westchester Area Math Circle, New York, USA; Ioan
Viorel Codreanu, Satulung, Maramures, Romania; Marin Chirciu, Colegiul National Zinca Golescu Piteşti,
Romania; Miguel Amengual Covas, Cala Figuera, Mallorca, Spain; Nicuşor Zlota, “Traian Vuia” Tech-
nical College, Focşani, Romania; Paolo Perfetti, Università degli studi di Tor Vergata Roma, Italy; Raja
Oktovin Parhasian Damanik, Canberra ACT, Australia; Sundaresh H. R., Shivamogga, Karnataka, India;
Theo Koupelis, Clark College, Washington, USA; Aaditya Bhaumik Sahu, AMMOC, India; Daniel Pascuas,
Barcelona, Spain; Diar Gashi, University of Primorska, Slovenia; Nihad Hashimov, AMMOC, Azerbaijan;
Prasanna Pawar, AMMOC, Bengaluru, India; Srijan Sundar, Oxford, UK; Sumanyu Nandecha, AMMOC,
India.

Mathematical Reflections 1 (2025) 9

S686. Let a, b, c be positive real numbers such that a + b + c = 3. Prove that
√
1 1 1 3 2 1
√ +√ +√ ≤ +√ .
2
b +c 2 2
c +a 2 2
a +b 2 ab + bc + ca 2

Proposed by Marius Stănean, Zalău, Romania

Solution by the author

By homogenizing the inequality becomes
1 1 1 a+b+c 3
p +p +p ≤ + .
2 2
2(b + c ) 2 2
2(c + a ) 2 2
2(a + b ) ab + bc + ca 2(a + b + c)

We have
X 1 X 1 (a + b + c)2 + ab + bc + ca
p ≤ = .
cyc 2(b2 + c2 ) cyc
b+c (a + b + c)(ab + bc + ca) − abc

Hence it suffices to prove that

(a + b + c)2 + ab + bc + ca a+b+c 3
≤ + .
(a + b + c)(ab + bc + ca) − abc ab + bc + ca 2(a + b + c)
Clearing denominators and expanding, the inequality becomes

(a + b + c)(ab + bc + ca)2 ≥ 2abc(a + b + c)2 + 3abc(ab + bc + ca).

But this follows from the AM-GM Inequality and (x + y + z)2 ≥ 3(xy + yz + zx). Indeed, we have

2(a + b + c)(ab + bc + ca)2 2(a + b + c)

= · (ab + bc + ca)2
3 3
2(a + b + c)
≥ · 3abc(a + b + c) = 2abc(a + b + c)2
3
and
(a + b + c)(ab + bc + ca)2 (a + b + c)(ab + bc + ca)
= · (ab + bc + ca)
3 3
≥3abc(ab + bc + ca).

Summing up these inequalities, we get the desired result.

Observation. This is stronger than problem 1 from 2024 Junior Balkan MO:
Let a, b, c be positive real numbers such that
1
a2 + b2 + c2 = .
4
Prove that √
1 1 1 2
√ +√ +√ ≤ .
2
b +c 2 2
c +a 2 2
a +b2 (a + b)(b + c)(c + a)
It can be shown, not hard, like
a+b+c 3 3(a + b + c) 4(a2 + b2 + c2 )
+ ≤ ≤ .
ab + bc + ca 2(a + b + c) 2(ab + bc + ca) (a + b)(b + c)(c + a)

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; Aaditya Bhaumik Sahu, AMMOC, India; Nguyen Thanh Nam, Tu Son Sec-
ondary School, Vietnam.

Mathematical Reflections 1 (2025) 10

S687. Solve in real numbers the system of equations:
s r r
x2 + 1 y2 + 1 z2 + 1
= x, 5 = 6y, 7 = 20z.
y2 + 1 z2 + 1 x2 + 1

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

Solution 1 by the author  π

Because x, y, z are positive, let x = tan u, y = tan v, z = tan w, where u, v, w ∈ 0, . Then
2
cos v sin u 5 cos w 6 sin v 7 cos u 20 sin w
= , = , = ,
cos u cos u cos v cos v cos w cos w
implying
cos v = sin u, 5 cos w = 6 sin v, 7 cos u = 20 sin w.
It follows that sin v = cos u and so 4(5 cos w) = 4(6 cos u) and, since 20 sin w = 7 cos u, we get

202 (cos2 w + sin2 w) = (242 + 72 ) cos2 u,

4 3 7 24 4 3
yielding cos u = . Then sin u = , sin w = , cos w = , sin v = , cos v = , hence
5 5 25 25 5 5
3 4 7
x = tan u = , y = tan v = , z = tan w = .
4 3 24

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

Clearly we have x, y, z > 0. Multiplying the three given equations we get 35 = 120xyz or

24xyz = 7. (1)

Squaring and simplifying the first of the given equations we get x2 + 1 = x2 (y 2 + 1) or

xy = 1. (2)

From (1) and (2) we get z = 7/24. Substituting into the second of the given equations, squaring and
simplifying we get
y2 + 1 16 4
25 · 2
= 36y 2 =⇒ y 2 = =⇒ y = .
(7/24) + 1 9 3
 
3 4 7
Therefore, (x, y, z) = , , is the only solution.
4 3 24

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; Raja Oktovin Parhasian
Damanik, Canberra ACT, Australia; Sundaresh H. R., Shivamogga, Karnataka, India; Daniel Pascuas,
Barcelona, Spain; Diar Gashi, University of Primorska, Slovenia; Madhav R Pillai, Delhi Public School,
Ranchi, Jharkhand, India; Nihad Hashimov, AMMOC, Azerbaijan; Nikolai N. Mnev, AMMOC, Bengaluru,
India; Srijan Sundar, Oxford, UK.

Mathematical Reflections 1 (2025) 11

S688. Prove that
ab2 + a + b2
a2 + ab + b
is an integer for infinitely many positive integers a and infinitely many positive integers b.

Proposed by Mircea Becheanu, Canada

Solution 1 by Theo Koupelis, Clark College, Washington, USA

For any positive integer a we set b = a[a(a − 1) + 1] = a3 − a2 + a, which is clearly a positive integer. Then

ab2 + a + b2 (a + 1)(a3 − a2 + a)2 + a a(a3 + a + 1)(a3 − a2 + 1)

= = = a3 − a2 + 1,
a2 + ab + b (a + 1)(a3 − a2 + a) + a2 a(a3 + a + 1)

which is clearly a positive integer.

Solution 2 by Daniel Pascuas, Barcelona, Spain

Just note that ab2 + a + b2 = (a2 + ab + b)(b − a + 1) + a3 − a2 + a − b so

ab2 + a + b2
= b − a + 1 is an integer if b = p(a), where p(x) = x3 − x2 + x.
a2 + ab + b
Since p(a) is a positive integer, for any positive integer a, and p(x) is increasing on R (because p′ (x) =
3x2 − 2x + 1 = 3(x2 − 32 x + 13 ) = 3(x − 13 )2 + 23 > 0, for any x ∈ R), the values of p(x) on the positive integers
form an infinite set of positive integers. Therefore we conclude that

ab2 + a + b2
a2 + ab + b
is an integer for infinitely many positive integers a and infinitely many positive integers b.

Also solved by Anderson Torres, Sao Paulo, Brazil.

Mathematical Reflections 1 (2025) 12

S689. Let ABC be an acute triangle with incenter I, A1 , B1 , C1 the feet of the altitudes and A2 , B2 , C2 the
midpoints of the sides BC, AC, AB, respectively. The line A2 I intersects AA1 in A3 , the line B2 I
−−→ −−→ −−→ → −
intersects BB1 in B3 and the line C2 I intersects CC1 in C3 . Prove that if AA3 + BB3 + CC3 = 0
then ABC is equilateral.

Proposed by Mihaela Berindeanu, Bucharest, Romania

Solution by Kousik Sett, India

Let the altitudes AA1 , BB1 , and CC1 meet at H (the orthocenter of triangle ABC). Since A3 ∈ AA1 ,
−−→ −−→ −−→ → − −−→ −−→ −−→ → −
B3 ∈ BB1 , C3 ∈ CC1 , and AA3 + BB3 + CC3 = 0 , we infer that AH + BH + CH = 0 .

→
−
Let the position vectors of the vertices A, B, C be → −
α , β , and →
−
γ respectively. From right triangles AA1 B
and AA1 C, we have
tan C AA1 BA1 BA1
= · = .
tan B A1 C AA1 A1 C
Consequently, if the position vector of A is →
1
−
a , then
1

→
−
→
− (tan B) β + (tan C)→
−
γ
a1 = . (1)
tan B + tan C
Also, we have
tan C BA1 BA1 BA1 a tan C
= = = ⇒ BA1 = .
tan B + tan C BA1 + A1 C BC a tan B + tan C
Again, since ∠CBB1 = 90◦ − C and ∠B1 BA = 90◦ − A, we have
1
A1 H [A1 BH] 2 · BA1 · BH · sin CBB1 BA1 · cos C
= = 1 = .
HA [HBA] 2 · BH · BA · sin B1 BA c · cos A

Substituting the expressions for BA1 , and using the Sine law, we obtain
A1 H a tan C cos C sin A · sin C 1 tan A
= · = · = .
HA tan B + tan C c · cos A tan B + tan C sin C · cos A tan B + tan C
→
−
Consequently, if the position vector of H is h , then using (3), we obtain
→
−
→
− (tan A)→
−α + (tan B + tan C)→
−
a1 (tan A)→
−α + (tan B) β + (tan C)→
−
γ
h = = .
tan A + tan B + tan C tan A + tan B + tan C

Mathematical Reflections 1 (2025) 13

 −−→ −−→ −−→ → − →
− − →
− → − →
− − →
−
Since AH + BH + CH = 0 , we have ( h − →
α)+(h − β )+(h −→
γ ) = 0 , which implies
→
− − →
− →
−
(tan B)( β − →
α ) + (tan C)(→
−
γ −→
−
α ) + (tan A)(→
−
α − β ) + (tan C)(→−
γ − β)
→
− − →
−
+ (tan A)(→
−
α −→−
γ ) + (tan B)( β − →
γ)= 0,

which yields
→
− →
−
(2 tan A − tan B − tan C)→
−
α + (2 tan B − tan C − tan A) β + (2 tan C − tan A − tan B)→
−
γ = 0.
→
−
Since →
−
α , β , and →
−
γ are non-collinear vector, we must have

2 tan A − tan B − tan C = 0, 2 tan B − tan C − tan A = 0, and 2 tan C − tan A − tan B = 0.

Therefore,
3 tan A = tan A + tan B + tan C = 3 tan B = 3 tan C,
and since A, B, and C are acute angles, we conclude that A = B = C. Hence the triangle must be
equilateral.

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

Mathematical Reflections 1 (2025) 14

S690. Let
2a 2b 2c
x= , y= , z= ,
b+c c+a a+b
where a, b, c are positive real numbers such that at most one of them is less that 1 and abc = 1. Prove
that
1 1 1
+ + + 3 ≥ 2(x + y + z).
x y z

Proposed by Vasile Cı̂rtoaje, Oil-Gas University, Ploieşti, Romania

Solution by the author

Assume that a ≥ b ≥ 1 ≥ c and denote p = a + b + c and q = ab + bc + ca. By the AM-GM inequality, we
have
p ≥ 3(abc)1/3 = 3.
In addition, from (a − 1)(b − 1)(c − 1) ≤ 0, we get

q ≥ p.

Write the inequality as follows:

Xb+c X 8a
+6≥ ,
cyc
a cyc
b+c
X 8 X
bc(b + c) + 6 ≥ a(a + b)(a + c),
cyc
(b + c)(c + a)(a + b) cyc

8(p3 − 2pq + 3)
pq + 3 ≥ .
pq − 1
So, for fixed p, we need to show that f (q) ≥ 0, where

f (q) = p2 q 2 + 18pq − 8p3 − 27.

Since f (q) is increasing, we have

f (q) ≥ f (p) = p4 − 8p3 + 18p2 − 27 = (p + 1)(p3 − 9p2 + 27p − 27) = (p + 1)(p − 3)3 ≥ 0.

The equality occurs for a = b = c = 1.

Also solved by Arkady Alt, San Jose, California, USA; Marin Chirciu, Colegiul National Zinca Golescu
Piteşti, Romania; 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.

Mathematical Reflections 1 (2025) 15

 Undergraduate problems

U685. Let {an } be the sequence defined by a1 = 2 and an = 2a2n−1 − 1, n = 2, 3, . . . . Evaluate

2n−1 (an−1 + 1)(an−2 + 1) · · · (a1 + 1)

lim .
n→∞ an

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

Solution by 1 the author

From an = 2a2n−1 − 1 we obtain

an − 1 = 2(a2n−1 − 1) = 2(an−1 − 1)(an−1 + 1).

This gives us
an − 1
= 2(an−1 + 1).
an−1 − 1
Similarly
an−1 − 1
= 2(an−2 + 1),
an−2 − 1
... ... ...,
a2 − 1
= 2(a1 + 1).
a1 − 1
Multiplying these relations we get
an − 1
2n−1 (an−1 + 1)(an−2 + 1) · · · (a1 + 1) = = an − 1.
a1 − 1
Therefore
2n−1 (an−1 + 1)(an−2 + 1) · · · (a1 + 1) 1
=1− .
an an
It’s easy to check that an > 1 forall n. Furthermore

an − an−1 = 2a2n−1 − an−1 − 1 = (an−1 − 1)(2an−1 + 1) > 0 ∀ n.

Hence {an } is a increasing sequence. Now if {an } has a finite upper bound, then there exits ℓ = lim an
n→∞
where ℓ is finite. Since an > a1 = 2 forall n then ℓ ≥ 2. From an = 2a2n−1 − 1 we take n → ∞ to get
ℓ = 2ℓ2 − 1. This yields (ℓ − 1)(2ℓ + 1) = 0 or ℓ = 1 which is a contradiction. So lim an = ∞. From here
n→∞
we obtain
2n−1 (an−1 + 1)(an−2 + 1) · · · (a1 + 1)
 
1
lim = lim 1 − = 1.
n→∞ an n→∞ an

Solution by 2 Seán M. Stewart, King Abdullah University of Science and Technology, Saudi Arabia
Denote the limit to be found by L. We show that L = 1.
We first show that an → ∞ as n → ∞. To do this, an explicit expression for an will first be found, before
taking the appropriate limit. Let  
1 1
a1 = x+ , x > 1.
2 x
√
Since a1 = 2, on solving for x, for x > 1 we find x = 2 + 3. Next

1 2
    
1 1 1
a2 = 2a21 − 1 = 2 x+ −1= x2 + 2 .
2 x 2 x

Mathematical Reflections 1 (2025) 16

By induction on n, it can be readily shown that
 
1 2n−1 1
an = x + 2n−1 ,
2 x
√
where n is a positive integer and x = 2 + 3. From this explicit expression we readily see that

an → ∞ as n → ∞.

Now consider the limit. Using the given recurrence relation for the sequence {an }n≥1 , namely a1 = 2 and

an = 2a2n−1 − 1, n = 2, 3, . . .

allows the limit to be rewritten as

2n−1 (an−1 + 1)(an−2 + 1) · · · (a1 + 1)
L = lim
n→∞ an
2n−1 · 2an−2 · 2an−3 · · · 2a21 · 3
2 2
= lim
n→∞ an
2n−1 12 (2an−2 )2 · 12 (2an−3 )2 · 12 (2a1 )2 · 3
= lim
n→∞ an
2n−1 (2an−2 ) (2an−3 )2 · · · (2a1 )2 · 3
2
= lim
n→∞ 2n−2 an
n−2
1 Y
= 6 lim (2ak )2 .
n→∞ an
k=1

Now
2(a2k+1 − 1) 2(a2k+1 − 1) a2k+1 − 1
(2ak )2 = 2(2a2k ) = 2(ak+1 + 1) = = = .
ak+1 − 1 (2a2k − 1) − 1 a2k − 1
Returning to the limit and making use of the above result we have
n−2
1 Y a2k+1 − 1
L = 6 lim
n→∞ an a2k − 1
k=1
1 a2n−1 − 1
= 6 lim · 2
n→∞ an a1 − 1
a 2 −1
= 2 lim n−1
n→∞ an
a 2 −1
= 2 lim n−1
n→∞ 2a2
n−1 − 1
1
1− a2n−1
= 2 lim 1
n→∞ 2− a2n−1
1
=2· = 1,
2
since an → ∞ as n → ∞, as announced.

Solution by 3 Brian
√ Bradie, Christopher Newport University, Newport News, VA, USA
Let x = ln(2 + 3). Then
√ √ √
2 + 3 + 2+1√3 2+ 3+2− 3
cosh x = = = 2,
2 2

Mathematical Reflections 1 (2025) 17

so a1 = cosh x. Next, a2 = 2 cosh2 x − 1 = cosh(2x) and by a straightforward induction argument, an =
cosh(2n−1 x) for all n ≥ 1. From here, we find

j−1 x j−1 x
 j−2 j−2
2
2(aj + 1) = e2 + 2 + e−2 = e2 x + e−2 x

and
n−1
Y 2
j−2 x j−2 x
2n−1 (an−1 + 1)(an−2 + 1) · · · (a1 + 1) = e2 + e−2
j=1
 2
n−1
2X
n−1 −2j+1)x/2
=  e(2 
j=1
n−1 −1)x/2 n−1 −1)x/2
!2
e(2 − e(−2
=
1 − e−x
n−2 n−2
!2
e2 x − e−2 x
=
ex/2 − e−x/2
sinh2 (2n−2 x)
= .
sinh2 (x/2)

Note that √ √ √
1√
2+ 3−2+ 2+ 3 3+2− 3 1
2
sinh (x/2) = = = ,
4 4 2
so
sinh2 (2n−2 x)
= 2 sinh2 (2n−2 x) = 2(cosh2 (2n−2 x) − 1)
sinh2 (x/2)
= (2 cosh2 (2n−2 x) − 1) − 1 = cosh(2n−1 x) − 1.

Finally,
2n−1 (an−1 + 1)(an−2 + 1) · · · (a1 + 1) cosh(2n−1 x) − 1
lim = lim = 1.
n→∞ an n→∞ cosh(2n−1 x)

Also solved by Arkady Alt, San Jose, California, USA; Devis Alvarado, UNAH y UPNFM, Tegucigalpa,
Honduras; G. C. Greubel, Newport News, VA, USA; Joshua Pité, Cambridge Rindge and Latin School,
MA, USA; Juan José Granier, Cencosud, Chile; 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; Diar Gashi, University of Primorska, Slovenia.

Mathematical Reflections 1 (2025) 18

U686. Let f : [0, 1] −→ R be a convex function such that f (0) = 0. Prove that
Z 1 2 Z 1
3
f (x)dx ≤ f (x)2 dx.
0 4 0

Proposed by Robert Rogozsan, Baia Mare, Romania

Solution by Daniel Pascuas, Barcelona, Spain

The inequality at the statement is equivalent to
Z 1
√ Z 1 1
3 2
f (x) dx ≤ f (x)2 dx . (1)
0 2 0

We are going to prove that

Z 1
√ Z 1 1
3 2
2
f (x) dx ≤ f (x) dx , (2)
0 2 0
R1
which clearly implies (1) whenever 0 f (x) dx ≥ 0. However we will show that (1) fails for many convex
functions f : [0, 1] → R such that f (0) = 0. By the way, note that the inequality (2) is sharp since it
becomes an equality for f (x) = x.
Let us prove (2). Since f is convex and f (0) = 0, we have that


f (tx) = f tx + (1 − t)0 ≤ tf (x) + (1 − t)f (0) = tf (x) (t, x ∈ [0, 1]),

so Z x Z 1 Z 1
1
f (t) dt = x f (tx) dt ≤ xf (x) t dt = xf (x) (x ∈ [0, 1]).
0 0 0 2
By integrating this inequality and applying Fubini’s theorem, we get that

1 1
Z Z 1 Z x  Z 1 Z 1 
xf (x) dx ≥ f (t) dt dx = f (t) dx dt
2 0 0 0 0 t
Z 1 Z 1 Z 1
= (1 − t)f (t) dt = f (x) dx − xf (x) dx,
0 0 0

and therefore Z 1 Z 1
3
f (x) dx ≤ xf (x) dx.
0 2 0
Note that we may apply Fubini’s theorem because the convexity of f on [0, 1] implies that f is bounded
and continuous on (0, 1) (see, for example, Proposition 1.1.3 in C. P. Niculescu and L.-P. Persson, Convex
Functions and Their Applications - A contemporary approach, Second Edition, CMS Books in Mathematics,
Springer, 2018). Hence Cauchy-Schwarz inequality gives (2):
Z 1 Z 1 Z 1  1 Z 1 1 √ Z 1 1
3 3 2
2
2
2 3 2
2
f (x) dx ≤ xf (x) dx ≤ x dx f (x) dx = f (x) dx .
0 2 0 2 0 0 2 0

Finally, for any c > 21 , consider fc (x) = (x − c)2 − c2 = x2 − 2cx, which is a convex function on [0, 1] and
satisfies fc (0) = 0. Then
Z 1 Z 1 Z 1
1 2 1 4c2
fc (x) dx = − c and fc (x) dx = (x4 − 4cx3 + 4c2 x2 ) dx = − c + ,
0 3 0 0 5 3
so 1 √ Z 1 1
Z 1 
1 3 3c 2 3 2
fc (x) dx = c − > − + c2 = 2
fc (x) dx ,
0 3 20 4 2 0

Mathematical Reflections 1 (2025) 19

because (c − 13 )2 − ( 20
3
− 3c
4 + c2 ) = c
12 − 7
180 = 15c−7
180 > 0, since c > 12 . Hence (1) fails for fc .

Editor’s Comment. Paolo Perfetti provided the following counterexample for f . Let f (x) be the convex
function equal to −xn for 0 ≤ x ≤ 1/n, f (x) = −1 for 1/n ≤ x ≤ 1−1/n, f (x) = (x−1)n for 1−1/n ≤ x ≤ 1.
2
1
1 2 3 1 4 2
Z   Z  
2 3
f (x)dx = 1− ̸≤ (f (x)) dx = 1−
0 n 4 0 4 3n

for any n ≥ 2.

Also solved by Artem Borisov; G. C. Greubel, Newport News, VA, USA; Joshua Pité, Cambridge Rindge and
Latin School, MA, 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 1 (2025) 20

U687. Prove that
1
√
3
√ Z 1√ √ Z 1 √ √
1+x+ 31−x 3
1+x− 31−x 3
1+x+ 31−x
Z
4
√ dx < √ dx < √ dx.
4 x 0 x 0 x
5

Proposed by Alessandro Ventullo, Milan, Italy

Solution 1 by the author

1 8 4 5 8 1
Since + log 2 < and − log 2 < , it is sufficient to prove that
6 9 5 6 9 4
Z 1 √
3
√ Z 1√ √ Z 5 − 8 log 2 √ √
1+x+ 31−x 3
1+x− 31−x 6 9
3
1+x+ 31−x
√ dx ≤ √ dx ≤ √ dx.
1
+ 8 log 2 x 0 x 0 x
6 9

First, we observe that the last two improper integrals converge because
√3
√ 2
1+x− 31−x 3x
√ ∼√ →0 as x → 0
x x
and √ √
3
1+x+ 31−x 2
√ ∼√
x x
and the integral of the last function converges as x → 0. In order to conclude, it is sufficient to use Steffensen’s
Inequality: if f and g are two integrable functions on [a, b] such that f is decreasing and 0 ≤ g(x) ≤ 1, then
Z b Z b Z a+λ
f (x) dx ≤ f (x)g(x) dx ≤ f (x) dx,
b−λ a a
Z b
where λ = g(x) dx.
a√
3
√ √ √
1+x+ 31−x 3
1+x− 31−x
By taking f (x) = √ and g(x) = √3
√ , it is easy to check that these functions
x 1+x+ 31−x
are integrable on [0, 1], f (x) is decreasing and 0 ≤ g(x) ≤ 1 for all x ∈ [0, 1]. Moreover, from Problem U531
proposed by the author we know that
Z 1√3
√
1+x− 31−x 5 8
λ= √
3
√ 3
dx = − log 2,
0 1+x+ 1−x 6 9
and the conclusion follows.

Solution 2 by Daniel Pascuas, Barcelona, Spain

√
By applying the substitution t = x we may write the inequality to be proved as
1
Z 1 p p Z 1 p p Z
2 p p
3 2 3 2

3 2 3 2

3 3

1 + t + 1 − t dt < 1 + t − 1 − t dt < 1 + t2 + 1 − t2 dt. (∗)
√2 0 0
5

Taking into account the binomial expansion

∞
√ 1 X
3
1 + x = (1 + x) 3 = cn xn (−1 < x < 1),
n=0
1

where cn = 3
n , we get that
∞ ∞
√ √ X √ √ X
3
1+x− 3
1−x=2 c2k+1 x2k+1 and 3
1+x+ 3
1−x=2 c2k x2k ,
k=0 k=0

Mathematical Reflections 1 (2025) 21

for 0 < x < 1, and, in particular,
p p ∞
X ∞
X
3 3
1 + t2 − 1 − t2 = 2 c2k+1 t4k+2 = 2
3 +2 c2k+1 t4k+2 and
k=0 k=1
p p ∞
X ∞
X
3 3
1 + t2 + 1 − t2 = 2 c2k t4k = 2 + 2 c2k t4k , for 0 < t < 1.
k=0 k=1

But cn = 13 ( 31 − 1) · · · ( 13 − (n − 1)) n!
1
, for n ≥ 1, so c2k < 0 < c2k+1 , for k ≥ 1, and therefore
p
3
p
3
p
3
p
3
2
1 + t2 + 1 − t2 < 2 and 3 < 1 + t2 − 1 − t2 , for 0 < t < 1.

Hence we obtain the first inequality in (∗):

Z 1   Z 1 p
p
3 2
p
3 2

2 2 3
p
3

1 + t + 1 − t dt < 2 1 − √ < < 1 + t2 − 1 − t2 dt.
2
√ 5 3 0
5

The second inequality in (∗) can be written as

1
Z 1 p p Z
2 p
3 3
3
1 + t2 − 1 − t2 dt < 2 1 − t2 dt,
1
2
0

which holds because

Z 1 p Z 1 q q
2
p 
3 3
1 + (t + 21 )2 − 3 1 − (t + 12 )2 dt

3
2 2
1 + t − 1 − t dt =
1
2
0

and q q p
3
3
1 + (t + 12 )2 − 3 1 − (t + 12 )2 < 2 1 − t2 (0 < t < 21 ),
√ q q
since the function f (t) = 2 3 1 − t2 + 3 1 − (t + 12 )2 − 3 1 + (t + 12 )2 satisfies that:

• f is decreasing on [0, 21 ]: Just note that

2
− 2
− 2
f ′ (t) = 43 t(1 − t2 )− 3 − 23 (t + 12 ) 1 − (t + 21 )2 3 − 32 (t + 12 ) 1 + (t + 12 )2 3

− 2
− 2
< 13 4t − 2(t + 21 ) 1 − (t + 12 )2 3 = 13 (2t − 1) 1 − (t + 12 )2 3 ≤ 0,

for any t ∈ [0, 21 ].

q √ √ √
• f ( 21 ) = 2 3 34 − 3 2 = 3 6 − 3 2 > 0.

Also solved by Joshua Pité, Cambridge Rindge and Latin School, MA, USA; Paolo Perfetti, Università degli
studi di Tor Vergata Roma, Italy; Theo Koupelis, Clark College, Washington, USA.

Mathematical Reflections 1 (2025) 22

U688. Evaluate n
1
xn ln(1 + x)
 Z
lim 1 + dx .
n→∞ 0 1 + xn

Proposed by Mihaela Berindeanu, Bucharest, Romania

Solution 1 by Seán M. Stewart, King Abdullah University of Science and Technology, Saudi Arabia
Denote the limit to be evaluated by L. We show that
2
L = elog (2)
.

Write the limit as

L = lim (1 + In )n ,
n→∞

where
1
xn log(1 + x)
Z
In = dx.
0 1 + xn
Setting x = t1/n in the integral gives
1 1/n
+ t1/n )
Z
1 t log(1
In = dt.
n 0 1+t

Note that as n → ∞, for t ∈ (0, 1) we see that t1/n → 1. So for n → ∞, by the dominated convergence
theorem we have
1 1 1 · log(1 + 1) log(2) 1 dt
Z Z
1
In → dt = = log2 (2).
n 0 1+t n 0 1+t n
So for the limit, by the dominated convergence theorem we have
n
log2 (2)

2
L = lim 1 + = elog (2) ,
n→∞ n

as announced. Here the well known result of

 x n
ex = lim 1+ ,
x→∞ n
with x = log2 (2) has been used.

Solution 2 by Daniel Pascuas, Barcelona, Spain

First, we integrate by parts to obtain that
Z 1 n
1 1
Z
x ln(1 + x)
n
dx = x ln(1 + x) d ln(1 + xn )
0 1 + x n 0
 1 Z 1 
1 n n


= x ln(1 + x) ln(1 + x ) − ln(1 + x ) d x ln(1 + x)
n 0 0
Z 1 Z 1
x ln(1 + xn )
 
1 2 n
= (ln 2) − ln(1 + x ) ln(1 + x) dx − dx .
n 0 0 1+x

Since 0 ≤ ln(1 + t) ≤ t, for any t ≥ 0, we have that

1 1
x ln(1 + xn )
Z Z
1
0≤ xn+1 dx =
dx ≤ and
0 1+x 0 n + 2
Z 1 Z 1
n ln 2
0≤ ln(1 + x ) ln(1 + x) dx ≤ (ln 2) xn dx = .
0 0 n+1

Mathematical Reflections 1 (2025) 23

It follows that
1
xn ln(1 + x) (log 2)2
Z  
1
dx = +O 2 ,
0 1 + xn n n
so, taking into account that ln(1 + t) = t + O(t2 ), as t → 0, we deduce that
1 n Z 1 n
xn ln(1 + x)
 Z   
x ln(1 + x)
1+ dx = exp n ln 1 + dx
0 1 + xn 0 1 + xn
(ln 2)2
   
1
= exp n ln 1 + +O
n n2
  
1
= exp (ln 2)2 + O .
n

Therefore n
1
xn ln(1 + x)
 Z
2
lim 1 + dx = e(ln 2) = 2ln 2 .
n→∞ 0 1 + xn

Also solved by 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; Theo Koupelis, Clark
College, Washington, USA; Srijan Sundar, Oxford, UK.

Mathematical Reflections 1 (2025) 24

U689. Evaluate the integral:

∞
arctan x · ln2 x
Z
dx.
0 x2 − x + 1

Proposed by Vasile Mircea Popa, Sibiu, Romania

Solution 1 by Seán M. Stewart, King Abdullah University of Science and Technology, Saudi Arabia
Denote the integral to be evaluated by I. We show that

5π 4
I= √ .
81 3
1
We begin by enforcing a substitution of x 7→ x in the integral. This gives
∞ arctan( x1 ) log2 (x)
Z
I= dx.
0 1 − x + x2

Since
1

π
arctan x = − arctan(x), x > 0,
2
allows the integral to be rewritten as

π ∞ log2 (x)
Z ∞
arctan(x) log2 (x) π ∞ log2 (x)
Z Z
I= dx − dx = dx − I,
2 0 x2 − x + 1 0 x2 − x + 1 2 0 x2 − x + 1
or
∞
log2 (x)
Z
π
I= dx.
4 0 x2 − x + 1
Since
1 + x3 = (1 + x)(x2 − x + 1),
the above integral for I can be rewritten as

π ∞ (1 + x) log2 (x)
Z ∞ Z ∞
log2 (x) x log2 (x)
Z 
π
I= dx = dx + dx . (2)
4 0 1 + x3 4 0 1 + x3 0 1 + x3

To find the values of the two integrals appearing in (2), we consider the parametric integral
Z ∞ a−1
x
J(a, b) = dx, 0 < a < b.
0 1 + xb

Enforcing a substitution of u = xb produces

a
∞
u b −1
Z
1
J(a, b) = du.
b 0 1+u

The integral that has appeared can be expressed in terms of a beta function. Here
1 a a 1 a  a π
J(a, b) = B ,1 − = Γ Γ 1− = .
b b b b b b b sin( aπ
b )

Here the connection between the beta and gamma function of

Γ(m)Γ(n)
B(m, n) = ,
Γ(m + n)

Mathematical Reflections 1 (2025) 25

has been used together with Euler’s reflexion formula for the gamma function of
π
Γ(x)Γ(1 − x) = .
sin(πx)
Thus
∂2J ∞
xa−1 log2 (x) π 3 (cos( 2aπ
b ) + 3)
Z
= dx = .
∂α2 0 1+x b 2b sin3 ( aπ
3
b )
Returning to the integral in (2) we see that
( )
π π 3 (cos( 2π
3 ) + 3) π 3 (cos( 4π
3 ) + 3) 5π 4
I = π4 J(1, 3) + J(2, 3) =

3 π + = √ ,
4 3
2 · 3 sin ( 3 ) 2 · 3 sin3 ( 2π
3
3 ) 81 3

as desired.

Solution 2 by Joshua Pité, Cambridge Rindge and Latin School, MA, USA
x ln2 x
The function f (x) = arctan
x2 −x+1
is continuous over (0, ∞).
Near 0 we have f (x) ∼ x(ln x) , and near ∞, f (x) = o x−3/2 . Hence the given integral converges. Denote
+ 2


R∞ x·ln2 x
I = 0 arctan
x2 −x+1
dx.
Using the substitution x = u1 we have

− arctan u1 (ln u)2 arctan u1 (ln u)2

Z 0
Z ∞

I= du = du.
∞ 1 − u + u2 0 u2 − u + 1

For u > 0, we know the identity arctan u1 = π2 − arctan u.

Thus  
Z ∞ 2 Z ∞ π − arctan u (ln u)2
arctan u (ln u) 2
I +I = 2−u+1
du + du
0 u 0 u2 − u + 1
 
Z ∞ arctan u + π − arctan u (ln u)2
π ∞ (ln u)2
Z
2
= du = du.
0 u2 − u + 1 2 0 u2 − u + 1
That is
∞
(ln u)2
Z
π π
I= du =: J.
4 0 u2 − u + 1 4
R1 ln2 x
R∞ ln2 x
R1 2
Writing J = 0 x2 −x+1dx + 1 dx and u = 1/x again, we obtain that J = 2 0 x2ln−x+1
x2 −x+1
x
dx.
1 P n
We can write, for all x ∈ R, with |x| < 1, x2 −x+1 = n≥0 an x , where the coefficients an are

1
 n ≡ 0, 1 (mod 6)
an = −1 n ≡ 3, 4 (mod 6)

0 otherwise


so that
1
ln2 x 1
Z X Z
J =2 dx = 2 an xn ln2 x dx.
0 x2 − x + 1 0
n≥0

By integration by parts, with u = ln2 x, dv = xn dx, we get

Z 1 Z 1 n+1 Z 1
n 2 2 x −2
x ln x dx = − ln x dx = xn ln x dx
0 n+1 0 x n+1 0
With one more integration by parts,
Z 1
2
xn ln2 x dx = .
0 (n + 1)3

Mathematical Reflections 1 (2025) 26

 " #
X 1 1 1 1
J =4 + − −
(6m + 1)3 (6m + 2)3 (6m + 4)3 (6m + 5)3
m≥0
" #
4 X 1 1 1 1
= 3 + − −
6 (m + 1/6)3 (m + 1/3)3 (m + 2/3)3 (m + 5/6)3
m≥0

Recall that the Hurwitz Zeta function is ζ(s, a) = ∞ 1

P
n=0 (n+a)s for ℜ(s) > 1 and a > 0.
We can then write
 
π
I = 3 ζ(3, 1/6) + ζ(3, 1/3) − ζ(3, 2/3) − ζ(3, 5/6) .
6
d n+1
The polygamma function is ψ (n) (x) = dx n+1 ln Γ(x), where Γ(x) is the Gamma function. We’ll use ψ
(2) (x),

which is related to ζ(3, x) by ψ (2) (x) = −2ζ(3, x).

We can use a reflection formula for the polygamma function:

d2
∀x ∈ (0, 1), ψ (2) (1 − x) − ψ (2) (x) = π cot(πx) = 2π 3 cot(πx) csc2 (πx).
dx2
5
√
Hence ψ (2) − ψ (2) 16 = 2π 3 cot π6 csc2 π
= 8 3π 3 .





6 √ 6
ψ (2) 4 (2) 2 = 8 3 π 3 .



Similarly 6 −ψ 6 9
After simplification, we conclude that
∞
√
arctan x · ln2 x
Z
5 3 4
dx = π ≈ 3.47155.
0 x2 − x + 1 243

Solution 3 by Daniel Pascuas, Barcelona, Spain

The substitution y = 1/x yields that
∞
arctan x · ln2 x ( π2 − arctan y) ln2 y
∞
Z Z
dx = dy
0 x2 − x + 1 0 1 − y + y2
π ∞
Z ∞
ln2 x arctan x · ln2 x
Z
= dx − dx,
2 0 x2 − x + 1 0 x2 − x + 1
so
∞ ∞
arctan x · ln2 x ln2 x 5π 4
Z Z
π
2
dx = dx = √ ,
0 x −x+1 4 0 x2 − x + 1 81 3
because
∞
ln2 x 20π 3
Z
dx = √ . (∗)
0 x2 − x + 1 81 3
We give two different proofs of (∗). The first one is just a standard evaluation of our integral by means of
the residues theorem. In the second proof we apply the monotone convergence theorem to write our integral
as the sum of a series which is evaluated by using the classical simple fraction expansion of the cotangent
(that is also a consequence of the residues theorem).
R∞ 2
Proof 1 of (∗). Note that 0 x2ln−x+1
x
dx = F ′′ (0), where F is the holomorphic function on the strip | Re z| < 1
defined by Z ∞
xz dx
F (z) = (| Re z| < 1).
0 x2 − x + 1
Now we are going to compute the integral F (z) by applying the residue theorem to the meromorphic function
z
wz
fz (w) = w2 −w+1 = (w+1)w
w3 +1
and the keyhole contour γδ,r,R , which we are going to define in a while. Here wz
denotes the branch of the z-power in the domain Ω = C \ [0, ∞) (where [0, ∞) = {x ∈ R : x ≥ 0}, as usual)

Mathematical Reflections 1 (2025) 27

 π
whose value at w = −1 equals to eiπ . Thus fz is a meromorphic function on Ω whose only poles are e±i 3
(simple poles).
For − π2 < δ < π2 and 0 < r < R, let αδ,r,R be the line segment which goes from reiδ to Reiδ .
For 0 < δ < π2 and ϱ > 0, let βδ,ϱ be the arc of the circle centered at the origin with radius ϱ which goes
from ϱeiδ to ϱe−iδ counterclockwise. Moreover, if α is a path on the complex plane, then α−1 denotes the
path α traversed in the opposite direction. For 0 < δ < π2 and 0 < r < 1 < R, γδ,r,R is the concatenation of
−1 −1 π
αδ,r,R , βδ,R , α−δ,r,R , and βδ,r . Note that γδ,r,R is a closed rectifiable curve in Ω \ {e±i 3 }. Then the residue
theorem shows that
Z  
i π3 −i π3



fz (w) dw = 2πi Res fz (w); w = e + Res fz (w); w = e .
γδ,r,R

Since π
π wz π
Res fz (w); w = ei 3 = lim π (w − ei 3 )fz (w) = lim π 1
π = √ e3
iz
and
i3 i 3 w − e−i 3 i 3
w→e w→e
π
π wz 5π
Res fz (w); w = e−i 3 = lim π (w − e−i 3 )fz (w) = lim π i
1
π = − √ e 3
i 3
iz
,
−i 3 −i 3 w − e 3
w→e w→e

we have that Z
π 5π
iz iz 2π πiz − 2π 2π
fz (w) dw = 2π
√
3
(e 3 −e 3 )= √
3
e (e 3 iz −e 3
iz
).
γδ,r,R

Therefore
Z Z Z Z Z 
2π πiz − 2π 2π
√
3
e (e 3 iz −e 3
iz
)= fz (w) dw = + − − fz (w) dw.
γδ,r,R αδ,r,R βδ,R α−δ,r,R βδ,r

Then let δ ↘ 0 to get that

R 2π 2π
xz dx Rz eizθ iReiθ dθ rz eizθ ireiθ dθ
Z Z Z
2π πiz − 2π 2π
√ e (e 3 iz −e 3
iz
) = (1 − e 2πiz
) + − .
3
r x2 − x + 1 0 R2 e2iθ − Reiθ + 1 0 r2 e2iθ − reiθ + 1
But
2π
Rz eizθ iReiθ dθ R1+Re z e2π| Re z|
Z
≤ 2π −→ 0, as R ↗ ∞, and
0 R2 e2iθ − Reiθ + 1 R2 − R − 1
2π
rz eizθ ireiθ dθ r1+Re z e2π| Re z|
Z
≤ 2π −→ 0, as r ↘ 0, for | Re z| < 1.
0 r2 e2iθ − reiθ + 1 1 − r − r2
It follows that
2π 2π
∞
xz dx 2π eiπz (e− 3 iz − e 3 iz
2π sin( 2π
3 z) 4π sinc( 2π
3 z)
Z
)
F (z) = = √ =√ = √ (| Re z| < 1),
0 x2 − x + 1 3 (1 − e2πiz ) 3 sin(πz) 3 3 sinc(πz)
where
∞
sin z X (−1)k 2k
sinc z = = z (z ∈ C),
z (2k + 1)!
k=0
which is an entire function without zeros on the strip | Re z| < 1. As a consequence,

4π d2 sinc( 2π 4π d 2π sinc′ ( 2π z) sinc(πz) − π sinc( 2π z) sinc′ (πz)

   
′′ 3 z) 3 3 3
F (z) = √ = √ ,
3 3 dz 2 sinc(πz) 3 3 dz sinc2 (πz)
3 3
for −1 < Re z < 1, and therefore F ′′ (0) = − 27√ sinc′′ (0) =
20π 20π
√ , since sinc(0) = 1, sinc′ (0) = 0, and
3 81 3
sinc′′ (0) = − 31 . Hence we finally obtain (∗).
Proof 2 of (∗). The substitution y = 1/x yields that
Z ∞
ln2 x ln2 y ln2 x
Z 1 Z 1
dx = dy = dx,
1 x2 − x + 1 0 1−y+y
2 2
0 x −x+1

Mathematical Reflections 1 (2025) 28

so
∞ ∞
ln2 x 1
ln2 x 1
ln2 x 1
(1 + x) ln2 x
Z Z Z Z Z
dx = + dx = 2 dx = 2 dx.
0 x2 − x + 1 0 1 x2 − x + 1 0 x2 − x + 1 0 1 + x3

Since
∞ ∞ ∞
1 X
3 n
X
3 2n 3 2n+1

X
= (−x ) = (x ) − (x ) = (1 − x3 )x6n (0 < x < 1),
1 + x3
n=0 n=0 n=0

the monotone convergence theorem shows that

∞ ∞
ln2 x 1
(1 + x) ln2 x 1
Z Z X Z
dx = 2 dx = 2 (1 + x)(1 − x3 )x6n ln2 x dx
0 x2 − x + 1 0 1 + x3
n=0 0
∞
X

=2 I6n + I6n+1 − I6n+3 − I6n+4 ,
n=0

where Z 1
Im = xm ln2 x dx (m = 0, 1, 2, . . . ).
0
We evaluate Im by integration by parts:
Z 1 h Z 1 
1 2 m+1 1 m+1 2
i1
m
Im = ln x dx = x ln x − 2 x ln x dx
m+1 0 m+1 0 0
Z 1 h i1 Z 1 
2 m+1 2 m+1 m 2
=− ln x dx =− x ln x − x dx = .
(m + 1)2 0 (m + 1)2 0 0 (m + 1)3

Therefore
∞ ∞
ln2 x
Z X

dx = 2 I 6n + I 6n+1 − I 6n+3 − I 6n+4
0 x2 − x + 1
n=0
∞  
X 1 1 1 1
=4 + − −
(6n + 1)3 (6n + 2)3 (6n + 4)3 (6n + 5)3
n=0
∞  
X 1 1 1 1
=4 + + +
(6n + 1)3 (6n + 2)3 (2 − 6(n + 1))3 (1 − 6(n + 1))3
n=0
∞   ∞  
X 1 1 1 X 1 1
=4 + = + .
n=−∞
(6n + 1)3 (6n + 2)3 54 n=−∞ (n + 16 )3 (n + 13 )3

Now we evaluate this series by a nice complex variable argument. Recall the classical simple fraction
expansion of the cotangent:
∞
X 1
π cot(πz) = (z ∈ C \ Z),
n=−∞
z+n

where the series converges uniformly on any compact subset of C \ Z. It follows that
∞ ∞
1 X d2 π d2
 
X 1 1 cos(πz)
= = cot(πz) = π 3 (z ∈ C \ Z).
n=−∞
(z + n)3 2
2 n=−∞ dz z + n 2 dz 2 sin3 (πx)

Hence we finally obtain (∗):

∞
ln2 x π 3 cos( π6 ) cos( π3 ) 2π 3 √ 20π 3
Z    
1
dx = + = 3 + √ = √ .
0 x2 − x + 1 54 sin3 ( π6 ) sin3 ( π3 ) 27 3 3 81 3

Mathematical Reflections 1 (2025) 29

Editor’s Comment. In order to calculate the integral

π ∞ ln2 u
Z
I= du,
4 0 u2 − u + 1

Henry Ricardo concluded immediately by using Formula 4.261.2 in Table of Integrals, Series, and Products
(Sixth Edition) by I. S. Gradshteyn and I. M. Ryzhik, which states that
∞
ln2 u 20π 3
Z
du = √ ,
0 u2 − u + 1 81 3
√
giving I = 5π 4 /81 3.

Also solved by Brian Bradie, Christopher Newport University, Newport News, VA, USA; Devis Alvarado,
UNAH y UPNFM, Tegucigalpa, Honduras; G. C. Greubel, Newport News, VA, USA; Henry Ricardo, Westch-
ester Area Math Circle, New York, 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; Prasanna Pawar, AMMOC, Bengaluru, India; Srijan Sundar, Oxford, UK.

Mathematical Reflections 1 (2025) 30

U690. Find all the complex numbers z such that

(Γ(z + 1))3 + (z − 1)3 (Γ(z − 1))3 = (z + 1)3 + 3 (Γ(z))3 ,

where Γ(z) is the Gamma function.

Proposed by Alessandro Ventullo, Milan, Italy

Solution 1 by the author

The given equality can be rewritten as

(Γ(z + 1))3 − (z + 1)3 (Γ(z))3 + (z − 1)3 (Γ(z − 1))3 = 3 (Γ(z))3 .

If we set a = Γ(z + 1), b = −(z + 1)Γ(z) and c = (z − 1)Γ(z − 1), the given equality becomes

a3 + b3 + c3 = 3(Γ(z))3 .

Now, observe that for every complex number z we have Γ(z + 1) = zΓ(z), so

a + b + c = 0,

and from the identity a3 + b3 + c3 − 3abc = (a + b + c)(a2 + b2 + c2 − ab − bc − ca), we get

3abc = 3 (Γ(z))3 ,

i.e.
−(z 2 − 1)Γ(z + 1)Γ(z)Γ(z − 1) = (Γ(z))3 .
By using once again Γ(z + 1) = zΓ(z), we get

−(z 2 − 1)z (Γ(z))2 Γ(z − 1) = (Γ(z))3

and since Γ(z) ̸= 0 for all complex numbers z, then

−(z 2 − 1)zΓ(z − 1) = Γ(z),

i.e.
−(z + 1)z = 1.
√
2 −1 ± i 3
Therefore, z + z + 1 = 0, which gives z = .
2
Solution 2 by Brian Bradie, Christopher Newport University, Newport News, VA, USA
Because
Γ(z + 1) = zΓ(z) and (z − 1)Γ(z − 1) = Γ(z),
the given equation is equivalent to

z 3 (Γ(z))3 + (Γ(z))3 = ((z + 1)3 + 3)(Γ(z))3 ,

i.e.
3(Γ(z))3 (z 2 + z + 1) = 0.
√
1 3
Since there is no z ∈ C for which Γ(z) = 0, it follows that z2 + z + 1 = 0, i.e. z = − ± i .
2 2
Also solved by 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; Seán M. Stewart, King
Abdullah University of Science and Technology, Saudi Arabia; Theo Koupelis, Clark College, Washington,
USA; Daniel Pascuas, Barcelona, Spain; Diar Gashi, University of Primorska, Slovenia; Srijan Sundar,
Oxford, UK.

Mathematical Reflections 1 (2025) 31

 Olympiad problems

O685. Let a, b, c be positive real numbers such that a + b + c = 1. Prove that

ab bc ca
+ + + 3(ab + bc + ca) ≥ 2.
c a b

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

Solution 1 by Srijan Sundar, Oxford, UK

By the AM-GM Inequality we have
  X b c
ab bc ca
(a + b + c) + + = ab + bc + ca + a2 +
c a b cyc
c b
≥ ab + bc + ca + 2(a2 + b2 + c2 ).

Hence, since a + b + c = 1, we get

ab bc ca
+ + + 3(ab + bc + ca) ≥ 2(a + b + c)2 = 2.
c a b
1
The equality holds if and only if a = b = c = .
3
Solution 2 by Theo Koupelis, Clark College, Washington, USA
The desired inequality is equivalent to

a2 b2 + b2 c2 + c2 a2 + 3abc(ab + bc + ca) ≥ 2abc.

Homogenizing the desired inequality, it suffices to show that

(a + b + c)(a2 b2 + b2 c2 + c2 a2 ) + 3abc(ab + bc + ca) ≥ 2abc(a + b + c)2 .

Expanding and simplifying we get

a3 (b − c)2 + b3 (c − a)2 + c3 (a − b)2 ≥ 0,

1
which is obvious. The equality occurs when a = b = c = .
3
Also solved by Hoang-Bao Le, Dublin, Ireland; 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; Sundaresh H. R., Shivamogga, Karnataka, India; Anderson Torres,
Sao Paulo, Brazil; Daniel Pascuas, Barcelona, Spain; Nguyen Thanh Nam, Tu Son Secondary School,
Vietnam.

Mathematical Reflections 1 (2025) 32

O686. Let a, b, c, d be positive real numbers such that
X X
2 a+ ab + 3 = abcd.
cyc sym

Prove that √ √ √ √ p
4 4 4 4 4
bcd + cda + dab + abc ≤ 27(1 + a)(1 + b)(1 + c)(1 + d)

Proposed by An Zhenping, Xianyang Normal University, China

Solution by Theo Koupelis, Clark College, Washington, USA

The given condition is equivalent to
1 1 1 1
+ + + = 1.
a+1 b+1 c+1 d+1
 
1 1 1 1
Let (x, y, z, w) = , , , , where x, y, z, w ∈ (0, 1) and x + y + z + w = 1. Thus, the
a+1 b+1 c+1 d+1
desired inequality is equivalent to
s r
X 4 (1 − y)(1 − z)(1 − w) 27
≤ 4 ,
yzw xyzw
or Xp
4
3x(1 − y)(1 − z)(1 − w) ≤ 3.
Using AM-GM we get
p
2 + 4x = 3x + (1 − y) + (1 − z) + (1 − w) ≥ 4 4 3x(1 − y)(1 − z)(1 − w)

with similar expressions for 2 + 4y, 2 + 4z, 2 + 4z. Adding these expressions we get
Xp
4 8 + 4(x + y + z + w)
3x(1 − y)(1 − z)(1 − w) ≤ = 3,
4
which completes the proof.

Also solved by Nicuşor Zlota, “Traian Vuia” Technical College, Focşani, Romania.

Mathematical Reflections 1 (2025) 33

O687. Let x, y, z be real numbers such that xy + yz + zx = 1 and x + y + z > 0. Prove that
√ √ p
xy z 2 + 1 + yz x2 + 1 + zx y 2 + 1 2
≤ .
x+y+z 3
Proposed by Marius Stănean, Zalău, Romania

Solution 1 by the author

The inequality can be rewritten as
p p p
xy (y + z)(z + x) + yz (x + y)(z + x) + zx (y + z)(x + y) 2
≤ .
(x + y + z)(xy + yz + zx) 3
We have two cases: x + y, y + z, z + x ≤ 0 or x + y, y + z, z + x ≥ 0. In first case with the substitutions x →
−x, y → −y, z → −z, we are in the second case. Hence, we will prove the inequality for x+y, y +z, z +x ≥ 0.
It is obvious that only one of the numbers x, y, z can be negative. Also, we have
√ √
q p
x + z + y + z = x + z + y + z + 2 (x + z)(y + z)
q p
= x + y + 2z + 2 z 2 + 1
√
> x + y.
√ √ √
Therefore, there is a triangle ABC with sides a = y + z, b = z + x, c = x + y. The inequality becomes
X X
3 ab(b2 + c2 − a2 )(c2 + a2 − b2 ) ≤ (a2 + b2 + c2 ) (b2 + c2 − a2 )(c2 + a2 − b2 ),
cyc cyc
or !
X X X X X X X
3 3 3 4 2 2 2 4
6 a b + 6abc a −3 ab a ≤ a 2 a b − a ,
cyc cyc cyc cyc cyc cyc cyc
or ! !
X X X X X X X X
3 3 3 2 2 2 2 2 4
6 a b + 6abc a −6 ab a b ≤ a −3 ab 2 a b − a ,
cyc cyc cyc cyc cyc cyc cyc cyc
or ! ! !
X X X X X X X
6abc a 2 ab − a2 + a2 − 3 ab 2 a2 b2 − a4 − 36a2 b2 c2 ≥ 0.
cyc cyc cyc cyc cyc cyc cyc
Using the known identities in a triangle
abc = 4Rrs, ab + bc + ca = s2 + r2 + 4Rr, a2 + b2 + c2 = 2(s2 − r2 − 4Rr),
after performing the calculations, it becomes
48Rrs2 (4r2 + 16Rr) − 16r2 s2 (s2 + 5r2 + 20Rr) − 36(16R2 r2 s2 ) ≥ 0,
or
s2 ≤ 12R2 − 8Rr − 5r2 .
Since by Gerretsen’s Inequality, we have
s2 ≤ 4R2 + 4Rr + 3r2 ,
it suffices to show that
4R2 + 4Rr + 3r2 ≤ 12R2 − 8Rr − 5r2 .
This factors as
(R − 2r)(2R + r) ≥ 0,
which is true by Euler’s Inequality. The equality holds when x = y = z.

Also solved by Arkady Alt, San Jose, California, USA; Marin Chirciu, Colegiul National Zinca Golescu
Piteşti, Romania; Nicuşor Zlota, “Traian Vuia” Technical College, Focşani, Romania; Theo Koupelis, Clark
College, Washington, USA; Srijan Sundar, Oxford, UK.

Mathematical Reflections 1 (2025) 34

O688. Inside the square ABCD the points E, F are taken such that∡F DE = ∡EBF = 45◦ It is given that
area F CD+area DEA+area EF B = a2 . Find the length of the side AB.

Proposed by Mihaela Berindeanu, Bucharest, Romania

Solution by Kousik Sett, India

Let lines parallel to AD, passing through E, F meet AB, DC at G, I and H,J, respectively. Similarly, let
lines parallel to AB, passing through E, F meet AD, BC at K, M and L, N , respectively.
Let P = IJ ∩ KL, Q = GH ∩ M N , and let x be the side length of the square ABCD such that x =
h1 + h2 + h3 = k1 + k2 + k3 , where h1 = HQ, h2 = QE, h3 = EG, and k1 = M Q, k2 = QF , k3 = F N ,
respectively.
Assume that ∠ABE = α, ∠CBF = β, ∠CDF = γ, and ∠ADE = δ. Since ∠F DE = ∠EBF = 45◦ , we
infer that α + β = 45◦ and γ + δ = 45◦ .
Lines GH, IJ, KL, and M N divide the square ABCD into nine rectangles as shown at the figure below.
Let [XY Z · · · W ] denotes the area of the polygon [XY Z · · · W ].

From right triangles BGE and BN F , we get

h3 k3
tan α = and tan β = .
x − k1 x − h1
Since α + β = 45◦ , we have
tan α + tan β + tan α · tan β = 1.
Substituting the expressions for tan α, tan β, and simplifying, we obtain

x2 − x(h1 + k1 + h3 + k3 )x + h1 h3 + k1 k3 + h1 k1 − h3 k3 = 0. (3)

From right triangles DJF and DKE, we get

h1 k1
tan γ = and tan δ = .
x − k3 x − h3
Since γ + δ = 45◦ , we have
tan γ + tan δ + tan γ · tan δ = 1.
Substituting the expressions for tan γ, tan δ, and simplifying, we obtain

x2 − x(h1 + k1 + h3 + k3 )x + h1 h3 + k1 k3 − h1 k1 + h3 k3 = 0. (4)

Mathematical Reflections 1 (2025) 35

From (3) and (4), we obtain

h1 k1 = h3 k3 , i.e., [M QHD] = [IBLP ]. (5)

Observe that

[F CD] + [DEA] + [EF B]

1
= ([M N CD] + [AGHD]) + [GBN Q] − [GBE] − [BN F ] − [EF Q]
2
1 1
= ([M N CD] + [AGHD]) + [GBN Q] − ([GBLE] + [IBN F ] + [EP F Q])
2 2
1 1
= ([M N CD] + [AGHD]) + [GBN Q] − ([GBN Q] + [IBLP ])
2 2
1 1
= ([M QHD] + [QN CH] + [M QHD] + [AGQM ]) + ([GBN Q] − [IBLP ])
2 2
1 1
= ([M QHD] + [QN CH] + [AGQM ] + [GBN Q]) + ([M QHD] − [IBLP ])
2 2
1
= ([ABCD]). [by (5)]
2
√
Since [F CD] + [DEA] + [EF B] = a2 , we 2 2 2
√ have [ABCD] = 2a , i.e., x = 2a . So, x = 2a.
Therefore, the length of the side AB = 2a.

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

Mathematical Reflections 1 (2025) 36

 5
O689. Prove that is the minimum positive value of the constant k such that
2
r r r r
b+c c+a a+b 2
+ + ≥3
ka + b + c kb + c + a kc + a + b k+2
holds for any nonnegative real numbers a, b, c with a + b + c > 0.

Proposed by Vasile Cı̂rtoaje, Oil-Gas University, Ploieşti, Romania

Solution 1 by the author r

2 5 5
For b = c = 0, the inequality becomes 2 ≥ 3 , which is equivalent to k ≥ . To show that is the
k+2 2 2
minimum value of the constant k, we need to prove the inequality
√
r r r
b+c c+a a+b
+ + ≥ 2.
5a + 2b + 2c 5b + 2c + 2a 5c + 2a + 2b
Let
p = a + b + c, q = ab + bc + ca, r = abc.
Applying the Cauchy-Schwarz inequality twice, we have
[ (b + c)]2
r
4(a + b + c)2
P
X b+c
≥P √ ≥ pP
5a + 2b + 2c (b + c)3/2 5a + 2b + 2c [ (b + c)] [ (b + c)2 (5a + b + c)]
P
s
4p2 8p3
=p = 3
.
4p − pq + 9r
P 3 P
2p [4 a + 11 ab(a + b) + 30abc]
So, it suffices to show that
8p3
≥ 2,
4p3 − pq + 9r
which is equivalent to the known inequality pq ≥ 9r.
The proof is completed. The equality occurs for a = b = c, and also for a = b = 0 or b = c = 0 or c = a = 0.

Solution 2 by Paolo Perfetti, Università degli studi di Tor Vergata Roma, Italy
Let b = c = 0. The inequality becomes
r
2
2≥3 ⇐⇒ k ≥ 5/2
k+2
By taking k = 5/2 and using Hölder’s inequality we get
r r !3
X 2(b + c) X 2(b + c) X X
(2b + 2c)2 (5a + 2b + 2c) ≥ (2b + 2c)
cyc
5a + 2b + 2c cyc
5a + 2b + 2c cyc cyc

The inequality is equivalent to

!3
X X 2
(2b + 2c) ≥ (2b + 2c)2 (5a + 2b + 2c) · 9 5
cyc cyc 2 +2

Upon simplifying this last inequality is

a2 b + ab2 + b2 c + bc2 + c2 a + ca2 ≥ 6abc
which is evidently true by the AM-GM Inequality. The equality occurs when a = b = c or when a = b = 0
or b = c = 0 or c = a = 0.

Also solved by Nicuşor Zlota, “Traian Vuia” Technical College, Focşani, Romania; Theo Koupelis, Clark
College, Washington, USA.

Mathematical Reflections 1 (2025) 37

O690. Prove that there is an infinite sequence S of primes such that for any finite subset A ⊂ S the sum of
elements of A is not a perfect power.

Proposed by Mohammadreza Ghelichkhani, Tehran, Iran

Solution by the author

We construct S inductively. First, we add 2 to S. Now, assume that at some stage S = {p1 , p2 , . . . , pt }, and
let the set of all possible sums of elements from its subsets be A = {a1 , a2 , . . . , an }, where n = 2t .
Let qi be the i-th prime number that does not divide any element in A. We choose the next prime number,
pt+1 , such that for each 1 ≤ i ≤ n,
pt+1 + ai ≡ qi (mod qi2 ).
It is clear that the solution to this congruence is coprime to qi for each i. Therefore, by Dirichlet’s theorem,
there exists such a prime number.

Also solved by Juan José Granier, Cencosud, Chile.

Mathematical Reflections 1 (2025) 38