Journal de Théorie des Nombres de Bordeaux

YoungJu Choie

Nicolas Lichiardopol

Pieter Moree

Patrick Solé

On Robin’s criterion for the Riemann hypothesis

Amedeo Scremin

Diophantine inequalities with power sums

G. Griffith Elder

Jeffrey J. Hooper

On wild ramification in quaternion extensions

Henri Faure

Van der Corput sequences towards general (0,1)–sequences in base b

Hemar Godhino

Paulo H. A. Rodrigues

On ${p}$-adic zeros of systems of diagonal forms restricted by a congruence condition

Bjorn Poonen

Sieve methods for varieties over finite fields and arithmetic schemes

Jean Gillibert

Invariants de classes : propriétés fonctorielles et applications à l’étude du noyau

Mohamed Ayad

Florian Luca

Mohamed AYAD et Florian LUCA
Two divisors of (n2 + 1)/2 summing up to n+ 1
Tome 19, no 3 (2007), p. 561-566.
<http://jtnb.cedram.org/item?id=JTNB_2007__19_3_561_0>
© Université Bordeaux 1, 2007, tous droits réservés.
L’accès aux articles de la revue « Journal de Théorie des Nom-
bres de Bordeaux » (http://jtnb.cedram.org/), implique l’accord
avec les conditions générales d’utilisation (http://jtnb.cedram.
org/legal/). Toute reproduction en tout ou partie cet article sous
quelque forme que ce soit pour tout usage autre que l’utilisation à
fin strictement personnelle du copiste est constitutive d’une infrac-
tion pénale. Toute copie ou impression de ce fichier doit contenir la
présente mention de copyright.
cedram
Article mis en ligne dans le cadre du
Centre de diffusion des revues académiques de mathématiques
http://www.cedram.org/
Journal de Théorie des Nombres
de Bordeaux 19 (2007), 561-566
Two divisors of (n2 + 1)/2 summing up to n+ 1
par Mohamed AYAD et Florian LUCA
Résumé. Dans cette courte note, on donne une réponse affirma-
tive à une question d’Ayad posée dans [1].
Abstract. In this short note, we give an affirmative answer to
a question of Ayad from [1].
1. Main result
In [1], Mohamed Ayad asked to prove that there does not exist an odd
prime p and two positive divisors d1 and d2 of (p2 +1)/2 such that d1 +d2 =
p+ 1. In this note, we prove a bit more, namely:
Theorem 1.1. There does not exist an odd integer n > 1 and two positive
divisors d1 and d2 of (n2 + 1)/2 such that d1 + d2 = n+ 1.
The condition n > 1 cannot be dropped since for n = 1 we may take
d1 = d2 = 1.
Proof. Let n > 1 be an odd integer for which there exist two divisors d1
and d2 as in the statement of the theorem. Since n is odd, we get that
n2 ≡ 1 (mod 8), therefore (n2 + 1)/2 is odd. We show that d1 and d2 are
coprime. Indeed, if not there exists an odd prime q | gcd(d1, d2). Hence,
q | d1 + d2, therefore n ≡ −1 (mod q). Since also q | d1 | n2 + 1, we get
that n2 ≡ −1 (mod q). From the above two congruences we obtain that
(−1)2 ≡ −1 (mod q), so q | 2, which is impossible.
Since d1 and d2 divide (n2 + 1)/2 and are coprime, we get that d1d2 |
(n2 + 1)/2. Write d1d2 = (n2 + 1)/(2d). Then
(d1 − d2)2 = (d1 + d2)2 − 4d1d2 = (n+ 1)2 − 2
(
n2 + 1
d
)
= ((d− 2)n+ d)
2 + 4− 4d
d(d− 2)
.
Now notice that all divisors of n2 + 1 are congruent to 1 modulo 4. In
particular, this applies to d1, d2 and d. Hence, n = d1 +d2−1 is congruent
Manuscrit reçu le 26 septembre 2006.
562 Mohamed Ayad, Florian Luca
to 1 modulo 4 as well. It follows that
X =
∣∣∣∣(d− 2)n+ d4
∣∣∣∣ , Y = ∣∣∣∣d1 − d24
∣∣∣∣ , s = d− 14
are non-negative integers satisfying
(1.1) X2 −DY 2 = s,
with D = d(d − 2) = (4s)2 − 1. If s = 0, then X2 + Y 2 = 0, so Y = 0,
leading to d1 = d2. Since these two divisors are also coprime, we get that
d1 = d2 = 1, therefore n + 1 = d1 + d2 = 2, contradicting the fact that
n > 1. Hence, we may assume that s ≥ 1. The above Diophantine equation
(1.1) leads to ∣∣∣∣XY −√D
∣∣∣∣ = s
Y (X +
√
DY )
.
Note that X > Y
√
D, therefore X +
√
DY > 2
√
DY > 2sY , because√
D =
√
(4s)2 − 1 > s, therefore∣∣∣∣XY −√D
∣∣∣∣ < 12Y 2 .
It is a known criterion due to Legendre that the above inequality implies
that X/Y must be a convergent of
√
D. With λ = 4s, it is easily checked
that we have the following continued fraction expansion√
λ2 − 1 = [λ− 1, 1, {2(λ− 1), 1}],
where {. . .} emphasizes the period. Using the above continued fraction, one
checks easily that if pm/qm denotes the mth convergent to
√
λ2 − 1, then
p2m −Dq2m = −2λ+ 2 or 1, according to whether m is even or odd. Hence,
for our values of X and Y we should have that
X2 −DY 2 ∈ {−8s+ 2, 1},
and comparing it with equation (1.1) we get that the only chance is s = 1,
leading to d = 5 and D = 15. Hence, (X,Y ) is a solution of the Pell
equation
(1.2) X2 − 15Y 2 = 1.
The minimal solution of the above Pell equation is (X1, Y1) = (4, 1). Hence,
if we write (Xt, Yt) for the tth solution of Pell equation (1.2), we then get
that
Xt +
√
15Yt = (4 +
√
15)t holds for all t ≥ 1.
Using the above representation, one checks easily that Xt ≡ 1 (mod 3) for
all positive integers t. Thus, if Xt = s(n+1)−(n−1)/4 = (n+1)−(n−1)/4
for some positive integer n, we would then get that n = (4Xt − 5)/3, but
since 4Xt − 5 ≡ −1 (mod 3), we get that (4Xt − 5)/3 is never an integer
Two divisors of (n2 + 1)/2 summing up to n+ 1 563
for any positive integer t. Thus, there is no solution and the theorem is
completely proved. 
2. Applications
Next, we present a corollary to Theorem 1.1 which has already been
proved in [1] by a different method.
Corollary 2.1. Let p be an odd prime number. Let a and b be two distinct
complex numbers. Then any primitive of the polynomial
((x− a)(x− b))(p
2−1)/2
is indecomposable over C.
Proof. In order to prove the above corollary we will need, aside from Theo-
rem 1.1, to recall a result from [1]. Let f(x) be a polynomial with complex
coefficients of degree m. Let F be the set of the critical points of f(x) and
assume that F contains r elements. For any z ∈ F denote by νf (z) the
valency of z; i.e., νf (z) = mf ′(z) + 1, where mf ′(z) is the multiplicity of z
as a root of f ′. Assume that f = g ◦h, where g and h are polynomials with
complex coefficients of degree at least 2. Let k = deg h. It is then proved in
[1] that there exist distinct elements x1, . . . , xs in F , and positive divisors
d1, . . . , ds of νf (x1), . . . , νf (xs), respectively, such that 1 < di ≤ νf (xi) and
k − 1 =
∑i=s
i=1(di − 1).
We can now embark to the proof of Corollary 2.1. Let f(x) be a prim-
itive of ((x− a)(x− b))(p
2−1)/2. Suppose that there exits some non-trivial
decomposition of f(x) in the form f(x) = g(h(x)). Since deg f = p2, we get
that k = deg h = p. With the previous notations, we have F = {a, b}, and
νf (a) = νf (b) = (p2 + 1)/2. If s = 1, then k = p divides (p2 + 1)/2, which is
clearly impossible. Thus, s = 2 and there exist two positive divisors d1 and
d2 of (p2 + 1)/2 such that k− 1 = p− 1 = d1 − 1 + d2 − 1. However, this is
impossible by Theorem 1, which completes the proof of Corollary 1. 
Finally, we present a Diophantine application of Corollary 2.1. Let a < b
and c < d be integers, e be an integer and p and q be odd primes. Consider
the Diophantine equation
(2.1)
∫ x
0
((t− a)(t− b))(p
2−1)/2 dt−
∫ y
0
((s− c)(s− d))(q
2−1)/2 ds = e
in integer solutions (x, y). In some cases, it can have infinitely many integer
solutions (x, y). Indeed, suppose that p = q and that c − a = d − b = f .
564 Mohamed Ayad, Florian Luca
Then,∫ y
0
((s− c)(s− d))(q
2−1)/2 ds =
∫ y
0
((s− a− f)(s− b− f))(p
2−1)/2 ds
=
∫ y−f
−f
((t− a)(t− b))(p
2−1)/2 dt
=
∫ y−f
0
((t− a)(t− b))(p
2−1)/2 dt+ h,
where h =
∫ 0
−f ((t− a)(t− b))
(p2−1)/2 dt, so equation (2.1) is
F (x)− F (y − f) = e1,
where F (x) =
∫ x
0 ((t− a)(t− b))
(p2−1)/2 dt and e1 = e + h = e − F (−f).
If e1 = 0, then the above equation has infinitely many solutions (namely
x = y − f).
The next corollary shows that the above instance is the only one in which
the Diophantine equation (2.1) can have infinitely many integer solutions
(x, y).
Corollary 2.2. Let a < b, c < d, p ≤ q and e be fixed integers, where
p and q are odd primes. If the Diophantine equation (2.1) has infinitely
many integer solutions x, y, then p = q, c − a = d − b = f and e =∫−f
0 ((t− a)(t− b))
(p2−1)/2 dt.
Proof. Assume that p ≤ q and that the given equation has infinitely many
positive integer solutions x, y. Let F (x) =
∫ x
0 ((t− a)(t− b))
(p2−1)/2 dt and
G(x) =
∫ x
0 ((t− c)(t− d))
(q2−1)/2 dt + e. By Corollary 2.1, both F (x) and
G(x) are indecomposable. By the main finiteness criterion from [3], it fol-
lows that (F (x), G(x)) = (φ ◦ u ◦ κ(x), φ ◦ v ◦ `(x)), where φ(x) ∈ Q[X] is a
non-constant polynomial, κ(x) and `(x) are linear polynomials in Q[x] and
the pair of polynomials (u(x), v(x)) belongs to five kinds of standard pairs,
which are all listed both in [3] as well as on page 182 of [2]. Since for us F
and G are indecomposable, it follows that either φ is linear or both u and
v are linear polynomials.
Assume first that φ is linear. Note that for us the critical sets of F (x)
and G(x) are F = {a, b} and G = {c, d} respectively. Furthermore, the
two critical points in F have the same valency and the same is true for
G. In what follows, α and β are non-zero rational numbers, µ, ν, d are
positive integers, ρ is a non-negative integer and ν(x) ∈ Q[X] is a non-zero
polynomial which may be constant.
Since the standard pair of the first kind is (xd, αxρν(x)d), where 0 ≤ ρ <
d, gcd(ρ, d) = 1, ρ+deg ν(x) > 0 (or switched), it follows that in such pairs
Two divisors of (n2 + 1)/2 summing up to n+ 1 565
one of the polynomials has a critical set consisting of at most one element.
Hence, (F (x), G(x)) cannot be equal to (φ ◦ u ◦ κ(x), φ ◦ v ◦ `(x)) for some
linear polynomials φ(x), κ(x) and `(x) and a standard pair of the first kind
(u(x), v(x)).
Since the standard pair of the second kind is (x2, (αx2 + β)ν(x)2) (or
switched), it follows as in the previous case that in such pairs one of the
polynomials has a critical set consisting of at most one element, and the
same contradiction as in the previous case is obtained.
To introduce standard pairs of third and fourth kind, we need to intro-
duce the Dickson polynomials. Denote by Dµ(x, δ) the µth Dickson poly-
nomial defined by the functional equation
Dµ(z + δ/z, δ) = zµ + (δ/z)µ,
or explicitly by
Dµ(z, δ) =
bµ/2c∑
i=0
dµ,iz
µ−2i, where dµ,i =
µ
µ− i
(
µ− i
i
)
(−δ)i.
Then the standard pairs of third and fourth kind are the pairs of the
form (Dµ(x, αν), Dν(x, αµ)), where gcd(µ, ν) = 1, together with the pairs
(α−µ/2Dµ(x, α),−β−ν/2Dν(x, β)), where gcd(µ, ν) = 2, respectively. It is
known that the derivatives of the Dickson polynomials have only sim-
ple roots (a verification of this simple fact is done at the end of Section
6 in [4], for example). Since for us all roots of both F ′ and G′ are of
multiplicity larger than 1, we get that (F (x), G(x)) cannot be equal to
(φ ◦ u ◦ κ(x), φ ◦ v ◦ `(x)) for some linear polynomials φ(x), κ(x) and `(x)
and a standard pair of the third or fourth kind (u(x), v(x)).
Finally, a standard pair of the fifth kind is ((αx2 − 1)2, 3x4 − 4x3) (or
switched) which have degrees at most 4, whereas deg g = (q2 +1)/2 ≥ (32 +
1)/2 ≥ 5, so again (F (x), G(x)) cannot be equal to (φ◦u◦κ(x), φ◦v ◦ `(x))
for some linear polynomials φ(x), κ(x) and `(x) and a standard pair of the
fifth kind (u(x), v(x)).
Assume now that φ is not linear. Then degF = degG, therefore p = q.
Further, we may take u(x) = v(x) = x. Let κ(x) = κ1x+κ0 and `(x) = `1x+
`0. Identifying the leading coefficients in F (x) = φ(u((κ(x))) and G(x) =
φ(v(`(x))), respectively, we get that a0κ(p
2+1)/2
1 = a0`
(p2+1)/2
1 = 2/(p2 + 1),
where a0 6= 0 is the leading coefficient of φ(x). Hence, κ(p
2+1)/2
1 = `
(p2+1)/2
1
and since (p2 + 1)/2 is odd, we get that κ1 = `1. Thus,
G(x) = φ(κ1x+ `0) = φ(κ1(x+ (`0 − κ0)/κ1) + κ0) = F (x− f),
566 Mohamed Ayad, Florian Luca
where f = −(`0 − κ0)/κ1. Taking derivatives we get G′(x) = F ′(x− f), so
((x− c)(x− d))(p
2−1)/2 = ((x− f − a)(x− f − b))(p
2−1)/2 ;
hence, c− a = d− b = f . The statement about e is now immediate. 
Acknowledgements. We thank the referee for useful suggestions and for
a simplification of our original proof of Theorem 1.1. This work was done
in May of 2006, while both authors attended the conference on Congrès
International, Algèbre, Théorie des Nombres et leurs Applications at the
University Mohammed I, in Oujda, Morocco. They thank the organizers
for the opportunity of participating in this event. During the preparation
of this paper, F. L. was also supported in part by grants PAPIIT IN104505,
SEP-CONACyT 46755 and a Guggenheim Fellowship.
References
[1] M. Ayad, Critical points, critical values of a prime polynomial. Complex Var. Elliptic Equ.
51 (2006), 143–160.
[2] Yu. F. Bilu, B. Brindza, P. Kirschenhofer, A. Pintér and R. F. Tichy, Diophantine
equations and Bernoulli polynomials. With an appendix by A. Schinzel. Compositio Math.
131 (2002), 173–188.
[3] Yu. F. Bilu and R. F. Tichy, The Diophantine equation f(x) = g(y). Acta Arith. 95
(2000), 261–288.
[4] Y. Bugeaud and F. Luca, On Pillai’s Diophantine equation. New York J. Math. 12 (2006),
193–217.
Mohamed Ayad
Laboratoire de Mathématiques Pures et Appliquées
Université du Littoral
F-62228 Calais, France
E-mail: ayad@lmpa.univ-littoral.fr
Florian Luca
Florian Luca
Instituto de Matemáticas
Universidad Nacional Autonoma de México
C.P. 58089, Morelia, Michoacán, México
E-mail: fluca@matmor.unam.mx


Two divisors of $(n^2+1)/2$ summing up to $n+1$

András Biró

Characterizations of groups generated by Kronecker sets

Luis H. Gallardo

Olivier Rahavandrainy

Odd perfect polynomials over ${\mathbb{F}_2}$

Not a member yet

On Robin’s criterion for the Riemann hypothesis

Diophantine inequalities with power sums

On wild ramification in quaternion extensions

Van der Corput sequences towards general (0,1)–sequences in base b

On ${p}$ -adic zeros of systems of diagonal forms restricted by a congruence condition

Sieve methods for varieties over finite fields and arithmetic schemes

Invariants de classes : propriétés fonctorielles et applications à l’étude du noyau

Two divisors of $(n^2+1)/2$ summing up to $n+1$

Characterizations of groups generated by Kronecker sets

Odd perfect polynomials over ${\mathbb{F}_2}$

1,235

1,351