International Mathematical Forum, Vol. 8, 2013, no.

20, 957 - 965

HIKARI Ltd, www.m-hikari.com

On the Exponential Diophantine Equation

a2x + axby + b2y = cz
Nobuhiro Terai

Division of Information System Design, Ashikaga Institute of Technology

268-1 Omae, Ashikaga, Tochigi 326–8558, Japan
terai@ashitech.ac.jp

Kei Takakuwa

Department of Mathematics, Faculty of Science, Gakushuin University

Mejiro, Toshima-ku, Tokyo 171–8588, Japan
takakuwa@math.gakushuin.ac.jp

Abstract
Let a, b, c be fixed positive integers satisfying a2 + ab + b2 = c with
gcd(a, b) = 1. We show that the Diophantine equation a2x +ax by +b2y =
cz has only the positive integer solution (x, y, z) = (1, 1, 1) under some
conditions. The proof is based on elementary methods and Cohn’s ones
concerning the Diophantine equation x2 + C = y n .
Mathematics Subject Classification: 11D61

Keywords: exponential Diophantine equation, integer solution, Eisenstein

numbers

1 Introduction.
Let a, b, c be primitive Pythagorean numbers, i.e., relatively prime positive
integers satisfying a2 + b2 = c2 . Jeśmanowicz [J] conjectured that if a, b, c are
primitive Pythagorean numbers, then the Diophantine equation
ax + by = cz
has only the positive integer solution (x, y, z) = (2, 2, 2). Although Jeśmanowicz’
Conjecture holds for many special Pythagorean numbers, this remains un-
solved in general. (See Terai[Te1] and [Te2] for analogues to Jeśmanowicz’
conjecture.)
958 N. Terai and K. Takakuwa

On the other hand, we call a, b, c Eisenstein numbers if a, b, c are positive

integers satisfying a2 + ab+ b2 = c2 . As another analogue to Jeśmanowicz’ con-
jecture concerning Pythagorean numbers, in previous papers [TT] and [Te3],
we proposed the following:
Conjecture 1. Let a, b, c be fixed positive integers satisfying a2 + ab + b2 = c2
with gcd(a, b) = 1. Then the Diophantine equation
a2x + ax by + b2y = cz (1)
has only the positive integer solution (x, y, z) = (1, 1, 2).
In this paper, as an analogue to Conjecture 1, we also propose the following:
Conjecture 2. Let a, b, c be fixed positive integers satisfying a2 + ab + b2 = c
with gcd(a, b) = 1. Then equation (1) has only the positive integer solution
(x, y, z) = (1, 1, 1).
Using the methods of Cohn[C1] and [C2] concerning the Diophantine equa-
tion x2 +C = y n as well as the ones in [TT], we show that the above Conjecture
holds under some conditions. In fact, we establish the following:

Theorem 1. Let a, b, c be fixed positive integers satisfying

a = ps , b = 2q t , c = a2 + ab + b2 (2)
with c ≡ 0 (mod 3), where p, q are distinct odd primes and s, t are positive
integers. Sppose that q t < 10000. Then equation (1) has only the positive
integer solution (x, y, z) = (1, 1, 1).
Theorem 2. Let a, b, c be fixed positive integers satisfying
a = m, b = 2, c = a2 + ab + b2 (3)
with c ≡ 0 (mod 3), where m is odd ≥ 3. Then equation (1) has only the
positive integer solution (x, y, z) = (1, 1, 1).

2 Lemmas.
We use the following Lemmas to show Theorems 1,2.
Lemma 1 ([TT]). Eisenstein numbers a, b, c with gcd(a, b) = 1 and a − b ≡
1 (mod 3) are given as follows:
a = u2 − v 2 , b = v(2u + v), c = u2 + uv + v 2 ,
where u, v are positive integers such that gcd(u, v) = 1, u > v and u ≡
v (mod 3).
On the exponential Diophantine equation a2x + ax by + b2y = cz 959

Lemma 2 (Nagell[N1]). The Diophantine equation

x2 + x + 1 = y n

has only the positive integer solution (x, y, n) = (18, 7, 3) with n ≥ 2.

Lemma 3 (Nagell[N2]). The Diophantine equation

3x2 + 1 = y n

has no positive integer solutions x, y, n with y odd and n odd ≥ 3.

Lemma 4 (Ljunggren[L]). The Diophantine equation

3x2 + 1 = 4y n

has no positive integer solutions x, y, n with y > 1 and n ≥ 3.

Lemma 5 (Brown[B1], [B2]). The Diophantine equation

x2 + 3m = y n

has only the positive integer solution (x, y, m, n) = (10, 7, 5, 3) with

gcd(x, y) = 1, m odd and n ≥ 3.

Lemma 6 (Cohn[C1]). Let C be a positive integer with C = cd2 , c square-

free and c ≡ 7 (mod 8). If p is an odd prime and

x2 + C = y p

for coprime integers x, y, then either

(a) there exist integers a, b with b | d, y = a2 + cb2 and
√ √
±x + d −c = (a + b −c)p ;or
(b) c ≡ 3 (mod 8), p = 3 and there exist odd integers A, B with B | d,
 √ 3
√ A + B −c
y = (A2 + cB 2 )/4, ±x + d −c = ;or
2
√
(c) p | h, the class number of the imaginary quadratic field Q( −c);or
(d) C = 3A2 ± 8, p = 3, x = A3 ± 3A;or
(e) C = 48D 6, p = 3, x = 4D 3 .
960 N. Terai and K. Takakuwa

3 The Diophantine equation X 2 + 3D 2 = Y n.

When y = 1, equation (1) can be written as

(ax + D)2 + 3D 2 = cz ,
where b = 2q t = 2D. Hence we need the following Proposition to show Theo-
rem 1.
Proposition 1. Let D = q t < 10000, where q is an odd prime and t is a
positive integer. Then the Diophantine equation
X 2 + 3D 2 = Y n (4)
has only the positive integer solutions (X, Y, D, n) = (10, 7, 9, 3), (118, 7, 31, 5),
(254, 7, 503, 7), (716, 19, 809, 5), (28724, 19, 4789, 7) with gcd(X, Y ) = 1 and
n ≥ 3.

Proof. When D = 3t , it follows from Lemma 5 that equation (4) has only
the positive integer solution (X, Y, D, n) = (10, 7, 9, 3) with gcd(X, Y ) = 1 and
n ≥ 3.
Next consider the case D = q t < 10000 with q > 3. If X is odd ,then
taking (4) modulo 8 implies that 4 ≡ X 2 + 3D 2 = Y n ≡ 0 (mod 8), which is
impossible. Hence X is even and so n is odd. We may suppose that n is an
odd prime, say n = l.
If l = 3, then Lemma 6 (b) shows that there exist odd integers u, v such
that
 √ 3
√ u + v −3
±X + D −3 =
2
with v | D and Y = (u2 + 3v 2 )/4. Equating the imaginary parts yields
8D = 3v(u2 − v 2 ),
which is impossible, because D = q t with q > 3.
Now suppose that l > 3. Then in view of Lemma 6 (a), there exist integers
u, v such that
√ √
±X + D −3 = (u + v −3)l
with Y = u2 +3v 2. We observe that u is even, since Y and v are odd. Equating
the imaginary parts yields
(l−1)/2  
 l
D=v ul−(2j+1) v 2j (−3)j . (5)
j=0
2j + 1
On the exponential Diophantine equation a2x + ax by + b2y = cz 961

Since D = q t , we have v = ±q k with 0 ≤ k ≤ t. We now have to distinguish

three cases: (i) k = 0, (ii) 0 < k < t, (iii) k = t.
(i) k = 0. Then v = ±1 and so
(l−1)/2  
 l
ul−(2j+1) (−3)j ± D = 0. (6)
j=0
2j + 1

Note that l | (D±1). Indeed, taking (6) modulo l implies that D ≡ ±(−3)(l−1)/2
≡ ±1 (mod l). Hence if D is given, then l is bounded and (6) is a equa-
tion in u of degree l − 1. When D = q t < 10000, we can easily check,
by a computer, that (6) has only the positive integer solutions (D, l, u) =
(31, 5, 2), (503, 7, 2), (809, 5, 4), (4789, 7, 4) with u even. Therefore equation
(4) has only the solutions listed above.
(ii) 0 < k < t. Then
(l−1)/2  
 l
t−k
q =± ul−(2j+1) (−3q 2k )j .
j=0
2j + 1

Taking the above equation modulo q implies that l = q, and so t − k = 1.

The minus sign can be rejected. Indeed, otherwise q ≡ −quq−1 (mod q 2 ), so
1 ≡ −uq−1 ≡ −1 (mod q), which is impossible. Consequently
(q−1)/2  
 q
q= uq−(2j+1) (−3q 2k )j . (7)
j=0
2j + 1

Since u is even, we have q ≡ 1 (mod 4). Suppose now that 2m (q − 1) with
m ≥ 2. We observe that the general term in the right-hand side of (7) has the
following form:
q(q − 1)  u q−2j−1 q − 2
q−2j−1
·2 · (−3q 2k )j .
(2j + 1)(q − 2j − 1) 2 2j
When 0 ≤ j < (q − 1)/2, the above term is divisible by 2m+1 , because the
inequality q − 2j − 1 < 2q−2j−1 holds. Therefore it follows from (7) that
q ≡ (−3q 2k )(q−1)/2 ≡ 3(q−1)/2 q k(q−1) (mod 2m+1 ).
m−1
Using induction on m, we easily see that 32 ≡ 1 (mod 2m+1 ) for all m ≥ 2.
Since q ≡ 2m + 1 (mod 2m+1 ), we have q 2 ≡ 1 (mod 2m+1 ). Hence we conclude
that q ≡ 1 (mod 2m+1 ), which is impossible.
(iii) k = t. Then v = ±q t = ±D. Hence equation (5) leads to
αl − β l
= ±1,
α−β
962 N. Terai and K. Takakuwa

√ √
where α = u ± D −3 and β = u ∓ D −3 . In view of Theorem of Cohn[C2]
by means of [BHV], we see that the above equation has no solutions.

4 Proof of Theorem 1.
Suppose that our assumptions are all satisfied. Let (x, y, z) be a solution
of (1). From (2), we see that c ≡ 3 (mod 4).
We distinguish two cases: (i) y = 1 , (ii) y ≥ 2.
(i) y = 1. Then equation (1) can be written as

(ax + D)2 + 3D 2 = cz . (8)

Taking equation (8) modulo 4 implies that z is odd. Hence it follows from
Proposition 1 that equation (8) has no solutions x, z with z ≥ 3 under our
assumptions. We therefore conclude that equation (1) has only the positive
integer solution (x, y, z) = (1, 1, 1).
(ii) y ≥ 2. Then it follows from (1) that z is even, say z = 2Z. By Lemma
1, we have
ax = U 2 − V 2 , by = V (2U + V ), cZ = U 2 + UV + V 2 , (E1 )
or
ax = V (2U + V ), by = U 2 − V 2 , cZ = U 2 + UV + V 2 , (E2 )
where U, V are positive integers such that gcd(U, V ) = 1, U > V and U ≡
V (mod 3).
First consider (E1 ). Since a is a power of an odd prime and gcd(U, V ) = 1,
we have ax = U + V, U − V = 1 and so

3(2V + 1)2 + 1 = 4cZ .

Now Lemma 4 implies that Z = 1 or 2. We show that neither Z = 1 nor Z = 2

occurs.
When Z = 2, then it follows from Lemma 1 that

U, V = h2 − k 2 , k(2h + k) ; c = h2 + hk + k 2 ,

where h, k are positive integers such that gcd(h, k) = 1, h > k and h ≡

k (mod 3). Thus we obtain

U + V = h(2k + h) = ax = psx .

This implies that 2k + h = psx and h = 1, which contradicts 1 ≤ k < h.

On the exponential Diophantine equation a2x + ax by + b2y = cz 963

When Z = 1, we first suppose that a < b. Then c = a2 + ab + b2 < 3b2 < b3

from 3 < b. From (1), we have

b2y < a2x + ax by + b2y = c2 < b6 .

Since y ≥ 2, we have y = 2. It follows from (1) and (2) that

ax b2 ≡ 2ab3 (mod a2 ),

so x = 1. Then a2 + ab2 + b4 = c2 = (a2 + ab + b2 )2 , which is impossible.

Suppose that a > b. Then c = a2 + ab + b2 < 3a2 ≤ a3 from 3 ≤ a. From
(1), we have

a2x < a2x + ax by + b2y = c2 < a6 ,

so x = 1, 2. If x = 2, then it follows from (1) and (2) that

a4 ≡ a4 + 2a3 b (mod b2 ),

which is impossible. Hence we obtain x = 1. Then since U = V + 1 and

a = 2V + 1, we have

3V 2 + 3V + 1 = U 2 + UV + V 2 = c = a2 + ab + b2 > a2 = (2V + 1)2 = 4V 2 + 4V + 1,

which is impossible.
Next consider (E2 ). Since a is a power of an odd prime and gcd(U, V ) = 1,
we have V = 1, 2U + V = ax and so

U 2 + U + 1 = cZ .

Now Lemma 2 implies that Z = 1. As above, we have x = 1. Then since

V = 1 and a = 2U + 1, we have

U 2 + U + 1 = c = a2 + ab + b2 > a2 = (2U + 1)2 = 4U 2 + 4U + 1,

which is impossible.

5 Proof of Theorem 2.
Suppose that our assumptions are all satisfied. Let (x, y, z) be a solution
of (1). From (3), we see that c ≡ 3 (mod 4).
We distinguish two cases: (i) y = 1, (ii) y ≥ 2.
(i) y = 1. Then equation (1) can be written as

(ax + 1)2 + 3 = cz . (9)

964 N. Terai and K. Takakuwa

Taking equation (9) modulo 4 implies that z is odd. Hence it follows from
Lemma 5 that equation (9) has no solutions x, z with z ≥ 3 under our as-
sumptions. We therefore conclude that equation (1) has only the positive
integer solution (x, y, z) = (1, 1, 1).
(ii) y ≥ 2. Then it follows from (1) that z is even, say z = 2Z. By Lemma
1, we have
ax = U 2 − V 2 , 2y = V (2U + V ), cZ = U 2 + UV + V 2 , (E1 )
or
ax = V (2U + V ), 2y = U 2 − V 2 , cZ = U 2 + UV + V 2 , (E2 )
where U, V are positive integers such that gcd(U, V ) = 1, U > V and U ≡
V (mod 3).
First consider (E1 ). Then we have V = 2 and 2U + V = 2y−1 and so
(U + 1)2 + 3 = cZ .
Now Lemma 5 implies that Z = 1. We show that the case Z = 1 does not
occur.
If Z = 1, then
c = U 2 + 2U + 4 = m2 + 2m + 4,
so U = m. From ax = U 2 − V 2 , we have
mx = m2 − 4,
which is impossible, since m is odd ≥ 3.
Next consider (E2 ). Then we have U + V = 2y−1 and U − V = 2 and so
3(V + 1)2 + 1 = cZ .
Now Lemma 3 implies that Z = 1, 2. We show that neither Z = 1 nor Z = 2
occurs.
If Z = 1, then taking the above equation modulo 4 yields
c = cZ ≡ 1 (mod 4),
which is impossible, since c ≡ 3 (mod 4).
If Z = 2, then it follows from Lemma 1 that
U + V = r(2s + r), c = r 2 + rs + s2 ,
where r, s are positive integers such that gcd(r, s) = 1, r > s and r ≡ s (mod 3).
Since r(2s + r) = 2y−1 , we have r = 2, s = 2y−3 − 1. Then c = s2 + 2s + 4 =
m2 + 2m + 4 and so s = m. Also U = 2y−2 + 1 = 2s + 3 = 2m + 3 and
V = U − 2 = 2m + 1. In vew of ax = V (2U + V ), we have
mx = 12m2 + 20m + 7,
which is impossible, since m is odd ≥ 3.
On the exponential Diophantine equation a2x + ax by + b2y = cz 965

References
[BHV] Yu. Bilu, G. Hanrot and P. M. Voutier, Existence of primitive divisors
of Lucas and Lehmer numbers, J. Reine Angew. Math. 539(2001), 75–122.

[B1] E. Brown, Diophantine equations of the form x2 + D = y n , J. reine.

angew. Math. 274/275(1975), 385–389.

[B2] E. Brown, Diophantine equations of the form ax2 + Db2 = y n , J. reine.

angew. Math. 291(1977), 118–127.

[C1] J.H.E. Cohn, The diophantine equation x2 + C = y n , Acta Arith.

65(1993), 367-381.

[C2] J.H.E. Cohn, The diophantine equation x2 + C = y n , II, Acta Arith.

109(2003), 205–206.

[J] L. Jeśmanowicz, Some remarks on Pythagorean numbers (in Polish),

Wiadom. Mat. 1(1955/1956), 196–202.

[L] W. Ljunggren, Über die Gleichungen 1+Dx2 = 2y n und 1+Dx2 = 4y n , Det

Kgl. Norske Videnskabers Selskab, Forhandl. Bd. 15, Nr. 30, Trondheim
1942.

[N1] T. Nagell, Des équations indéterminées x2 +x+1 = y n et x2 +x+1 = 3y n ,

Norsk. Mat. Forenings Skrifter No. 2(1921), 12–14.

[N2] T. Nagell, Contributions to the theory of a category of Diophantine equa-

tions of the second degree with two unkowns, Nova Acta Soc. Sci. Upsal,
Ser IV, (2), 16(1955), 1–38.

[Te1] N. Terai, The Diophantine equation x2 + q m = pn , Acta Arith. 63(1993),

351–358.

[Te2] N. Terai, Applications of a lower bound for linear forms in two logarithms
to exponential Diophantine equations, Acta Aith. 90(1999), 17–35.

[Te3] N. Terai, A remark on a conjecture concerning Eisenstein numbers,

Compt. Rend. Math. Rep. Acad. Sci. Canada 22(2000), 105–110.

[TT] N. Terai and K. Takakuwa, On a Diophantine equation concerning Eisen-

stein numbers, Tokyo J. Math. 24(2001), 429–439.

Received: March 5, 2013