2QWKH&ODVVLFDO'LRSKDQWLQH(TXDWLRQ
x4
+
y4
+
kx2y2
=
z2

Miroslav
Stoenchev1,a)
and
Venelin
Todorov2,3,b)
1
Department
of
Mathematics,
Technical
University,
Sofia,
Bulgaria
2
Institute
of
Mathematics
and
Informatics,
Bulgarian
Academy
of
Sciences,
Sofia,
Bulgaria
3
Institute
of
Information
and
Communication
Technologies,
Bulgarian
Academy
of
Sciences,
Sofia,
Bulgaria
a)
Corresponding
author:
mrs@tu-sofia.bg
b)
[email protected],
[email protected]

Abstract.
The
purpose
of
this
article
is
to
describe
parametrically
all
nontrivial
solutions
of
the
diophantine
equation
in
the
title.

One
parameter
family
of
elliptic
curves
is
naturally
associated
with
the
equation,
for
which
family
we
apply
”complete
2-descend”

algorithm
to
obtain
a
parametric
description
of
all
possible
values
of
parameter
k
∈
Z,
for
which
nontrivial
solutions
exist.
The

article
is
a
natural
continuation
of
[1].

,1752'8&7,21
We
consider
the
diophantine
equation
x4
+
y4
+
kx2 y2
=
z2 ,
where
k
∈
Z
is
a
parameter.
Our
aim
is
to
determine
the

integers
k
for
which
the
equation
has
a
solution
in
positive
integers
(x,
y,
z)
and
to
describe
parametrically
all
solutions.

Each
diophantine
equation
can
be
considered
as
an
equation
that
defines
affine
variety
in
the
corresponding
affine

space,
or
after
homogenization,
projective
variety
in
the
corresponding
projective
space.
Thus
the
initial
equation

defines
affine
surface
in
a
three
dimensional
affine
space
over
algebraic
closure
of
Q,
denoted
by

T k
=
{(x,
y,
z)
∈
A3 (Q)
|
x4
+
y4
+
kx2 y2
=
z2 }
or
T k
:
x4
+
y4
+
kx2 y2
=
z2 ,
and
for
the
corresponding
projective
case,
after
homogenization
with
introducing
a
new
variable
t,
the
notation
is:

T k
=
{[x,
y,
z,
t]
∈
P3 (Q)
|
x4
+
y4
+
kx2 y2
=
z2 t2 }
or
T k
:
x4
+
y4
+
kx2 y2
=
z2 t2
For
convenience,
we
will
use
affine
equations,
but
with
the
comprehension
that
we
work
with
projective
varieties

(curves
and
surfaces).
The
difference
is
at
the
points
at
infinity,
given
by
intersection
of
projective
variety
with
the

hyperplane
at
infinity
H∞
:
t
=
0.
Basic
objects
of
consideration
are
smooth
projective
curves
and
surfaces,
and
the
main
apparatus
is
related
to
algebraic

and
analytic
invariants
of
elliptic
curves.

%$6,&'(),1,7,216
In
this
section
are
given
definitions
of
affine
and
projective
spaces,
elliptic
curves
over
an
arbitrary
field,
and
the

structure
preserving
maps
between
elliptic
curves.
The
following
definitions
are
necessary
([4],[8],[9],[13]).

Definition
1
Affine
n-space
over
Q
is
the
set
An (Q)
=
{(x1 ,
x2 ,
.
.
.
,
xn )
|
xi
∈
Q}.

The
zero
point
of
An
is
OAn
=
(0,
.
.
.
,
0),
and
if
A,
B
are
sets
then
A
−
B
means
the
set-theoretical
subtraction.
Definition
2
Projective
n-space
over
Q,
denoted
by
Pn ,
is
the
quotient
space
(An+1 (Q)
−
OAn+1
)/
∼,
where
the

factorization
by
∼
means
that
the
points
(x0 ,
.
.
.
,
xn ),
(y0 ,
.
.
.
,
yn )
∈
An+1 (Q)
−
OAn+1
are
equivalent,
if
there
exists
∗ ∗
λ
∈
Q ,
such
that
y0
=
λx0 ,
.
.
.
,
yn
=
λxn .
An
equivalence
class
{(λx0 ,
.
.
.
,
λxn )
|
λ
∈
Q }
is
denoted
by
[x0 ,
.
.
.
,
xn ],

and
the
individual
x0 ,
.
.
.
,
xn
are
called
homogeneous
coordinates
for
the
corresponding
point
of
Pn .

Applications of Mathematics in Engineering and Economics (AMEE’20)

AIP Conf. Proc. 2333, 110002-1–110002-14; https://doi.org/10.1063/5.0042739
Published by AIP Publishing. 978-0-7354-4077-7/$30.00

110002-1
Thus, the projective space consists of lines through the origin in affine space, with one dimension higher.

Definition 3 Elliptic curve over Q is a smooth projective curve with affine equation

y2 = x3 + ax2 + bx + c, (1)

where a, b, c ∈ Q. In general, elliptic curve E over field k is denoted by E/k.

The smoothness condition is equivalent to the condition that the polynomial x3 + ax2 + bx + c has distinct roots. The
unique point at infinity that lies on the elliptic curve is denoted by O = [0, 1, 0]. The discriminant of E/k : y2 = f (x)
given by (1) is defined as ΔE = 16Δ f = 16(−4a3 c + a2 b2 + 18abc − 4b3 − 27c2 ).
Let E/Q be an elliptic curve given by equation (1). Therefore E ⊂ P2 (Q) consists of the points P = (x, y)
satisfying the equation (1), together with the point at infinity O = [0, 1, 0]. Let l ⊂ P2 (Q) be a line, then by Bezout’s
theorem, the number of points of intersection for l ∩ E, taken with multiplicities, is exactly 3, say P, Q, R (need not be
distinct). The definition of composition law ⊕ on elliptic curve E is as follows:

Definition 4 The composition law E × E −→ E (P, Q)

−→ −R, is denoted by P ⊕ Q := −R, where the map
E −→ E P = (x, y)
−→ −P = (x, −y) is an orthogonal symmetry with respect to the coordinate axis.

Remark 1 The composition law is in fact a group law, i.e. makes E into an abelian group, with O = [0, 1, 0] as
neutral element for the group operation, and each element P has inverse −P. By the definition above, it follows that
three points on E have zero sum, if and only if they lie on the same line.

As notation : E = E(Q) = {(x, y) ∈ A2 (Q) | y2 = x3 + ax2 + bx + c} ∪ {O}, and for every subfield k ⊂ Q denote by E(k)
the set of k-rational points on E:

E(k) = {(x, y) ∈ A2 (k) | y2 = x3 + ax2 + bx + c} ∪ {O}. (2)

For elliptic curve E/k, the set E(k) is a group, E(k)  E(Q), in particular let k = Q:
Definition 5 The group E(Q) is called the Mordell-Weil group of rational points on E.
Elliptic curves have an algebraic structure as abelian groups and a geometric structure as smooth projective
curves. The structure preserving maps between elliptic curves are called isogenies. Let k be a field and E/k be an
elliptic curve, given by equation f (x, y, z) = x3 + ax2 z + bxz2 + cz3 − y2 z = 0.

Definition 6 T he f unction f ield k(E) of elliptic curve E/k consists of rational functions gh , where
1) g, h ∈ k[x, y, z] are homogeneous polynomials of the same degree,
2) h  ( f ), i.e. h is not divisible by f ,
3) gh11 and gh22 are considered equivalent whenever g1 h2 − g2 h1 ∈ ( f ).

Definition 7 Let E1 /k and E2 /k be elliptic curves. A rational map ϕ : E1 −→ E2 is a projective triple ϕ =

[ϕ1 , ϕ2 , ϕ3 ] ∈ P2 (k(E1 )), such that for every point P ∈ E1 (k), where ϕ1 (P), ϕ2 (P), ϕ3 (P) are defined, are not all zero
and the projective point [ϕ1 (P), ϕ2 (P), ϕ3 (P)] lies in E2 (k). The map ϕ is regular at P if there exists λ ∈ k(E1 )∗ , such
that λϕ1 , λϕ2 , λϕ3 are defined at P and are not all zero at P. Everywhere regular rational map is called a morphism.

Remark 2 Every rational map between elliptic curves is a morphism and every morphism between smooth pro-
jective curves is either constant or surjective.

110002-2
 Let E1 /k and E2 /k be elliptic curves.
Definition 8 An isogeny ϕ : E1 −→ E2 is a surjective morphism of curves that induces a group homomorphism
E1 (k)
−→
E2 (k).
The
elliptic
curves
E1
and
E2
are
then
said
to
be
isogenous.

Example
1
For
m
∈
N
denote
by
[m]P
:=
P⊕ P⊕·
·
·⊕ P
(m
-
times
addition).
The
map
[m]
:
E
−→
E
P

−→
[m]P

E[m].
The
elements
of
E[m]
are
called
m-torsion
points
of
E.
For
E/k
with

is
an
isogeny.
Denote
its
kernel
by
char
k
=
0
holds
that
E[m]

Z/mZ
Z/mZ.

Remark
3
Let
ϕ
:
E1
−→
E2
be
an
isogeny.
Then
there
exists
a
unique
isogeny
ϕ̃
:
E2
−→
E1
satisfying

ϕ̃
◦
ϕ
=
[m]
and
ϕ
◦
ϕ̃
=
[m],
for
appropriate
positive
integer
m.
The
isogeny
ϕ̃
is
called
dual
isogeny
for
ϕ,
and
the

integer
m
is
called
degree
of
ϕ.

%$6,&7+(25(06

7KH6WUXFWXUHRI0RUGHOO:HLO*URXS
Let
E/Q
be
an
elliptic
curve.
Theorem 1 The Mordell-Weil group E(Q) is finitely generated and abelian.

Theorem 2 Every finitely generated abelian group A is a direct sum of a free subgroup and a torsion subgroup,
i.e. A = A f ree Atorsion  Zr Atorsion , where the integer r ≥ 0 is called rank of A and is denoted by rank A = r.

Remark 4 From the theorems above it follows that E(Q)  Zr E(Q)tor .

7KHRUHPVIRU7RUVLRQV
The
torsion
group
E(Q)tor
is
finite
and
effectively
computable
by
algorithms
as
Lutz-Nagell
theorem,
the
reduction

theorem
and
the
general
theorem
of
Mazur.
The
necessary
definitions
for
Q p ,
Z p ,
F p
and
the
reduction
map
modulo

p
are
given
in
the
Appendix.

Theorem
3
(Lutz-Nagell)
Let
E
:
y2
=
x3
+
ax2
+
bx
+
c
be
an
elliptic
curve
with
integer
coefficients
and
P
=
(x,
y)

be
a
torsion
point
of
E.
Then
x
and
y
are
integers,
and
either
y
=
0
or
y2
is
a
divisor
of
the
discriminant
Δ
of
polynomial

f
(x)
=
x3
+
ax2
+
bx
+
c.
(Δ f
=
−4a3 c
+
a2 b2
+
18abc
−
4b3
−
27c2 )

Theorem
4
(Reduction)
Let
p
be
a
prime
number,
m
be
a
positive
integer
not
divisible
by
p,
and
E/Q p
be
an

elliptic
curve.
If
the
reduction
modulo
p
E/Q p
−→
Ẽ/F p
gives
a
nonsingular
curve
Ẽ/F p ,
then
the
reduction
map

E(Q p )[m]
−→
Ẽ(F p )
is
an
injective
homomorphism
of
groups.
Theorem 5 (Mazur) Let E/Q be an elliptic curve. Then the torsion group E(Q)tor is isomorphic to one of the
following fifteen groups:
Z/nZ, 1 ≤ n ≤ 10 or n = 12,
Z/2Z ⊕ Z/2nZ, 1 ≤ n ≤ 4.

7KHRUHPVIRUWKH5DQN
In
general,
an
effective
algorithm
for
determining
the
rank
of
every
elliptic
curve
over
Q,
for
finite
amount
of
time,
is

not
known.
In
the
case,
when
all
two-torsions
for
E/Q
are
rational,
i.e.
E[2]
=
E(Q)[2],
one
rank
searching
algorithm

is
the
Complete
2-Descent
([8]),
formulated
as
two
theorems
with
common
assumptions:

Theorem 6 (Complete 2-Descent) Let E/Q : y2 = (x − e1 )(x − e2 )(x − e3 ), e1 , e2 , e3 ∈ Q be an elliptic curve. Let
S be a finite set of primes, including 2, ∞ and all primes that divide the discriminant of E.
Let Q(S , 2) = {b ∈ Q∗ /(Q∗ )2 | ord p (b) ≡ 0 mod 2 ∀p  S }. Then there is injective group homomorphism

E(Q)/2E(Q) −→ Q(S , 2) × Q(S , 2),

110002-3
defined by
⎧
⎪
⎪
⎪ (x − e1 , x − e2 ) , x  e1 , e2
⎪
⎪
⎪
⎪
⎪
⎪  e −e 
⎪
⎪
⎪ e1 −e2 , e1 − e2 , x = e1
1 3
⎪
⎪
⎨
P = (x, y)
−→ ⎪
⎪  
⎪
⎪
⎪ e2 −e3
⎪
⎪
⎪ e2 − e1 , e2 −e1 , x = e2
⎪
⎪
⎪
⎪
⎪
⎪
⎩ (1, 1), P=O

Theorem 7 (Complete 2-Descent) Let (b1 , b2 ) ∈ Q(S , 2) × Q(S , 2) be a pair that is not an image of any of the
points O, (e1 , 0), (e2 , 0), (e3 , 0). Then (b1 , b2 ) is the image of a point P = (x, y) ∈ E(Q)/2E(Q) if and only if, the
equations
b1 z21 − b2 z22 = e2 − e1 , b1 z21 − b1 b2 z23 = e3 − e1 ,
have a solution (z1 , z2 , z3 ) ∈ Q∗ × Q∗ × Q∗ . If such a solution exists, then

P = (x, y) = (b1 z21 + e1 , b1 b2 z1 z2 z3 ).

Remark 5 Complete 2-Descent algorithm has a geometric interpretation: the two equations in theorem 7 define
quadric surfaces in P3 , which intersect in smooth quartic curve in P3 , called a homogeneous space for E/Q.
In the case when E[2]  E(Q)[2], the general algorithm for determining the rank of an elliptic curve uses Selmer and
Shafarevich-Tate groups ([6],[8]).

Theorem 8 The rank is invariant under isogeny maps, hence isogenous elliptic curves have the same rank.

5(68/76

7KH0DLQ,GHD
Let Ek , Ek and Hk be the curves with affine equations:

Ek : Y 2 = X(X − k + 2)(X − k − 2),

Ek : y2 = x (x2 + kx + 1),

Hk : v2 = u4 + ku2 + 1.
Determining the solutions of the initial diophantine equation means to determine integral points on surface T k :
x4 + y4 + kx2 y2 − z2 = 0. Finding integral points on quartic surface T k is equivalent to finding rational points on quartic
curve Hk , which is contained in T k . The map T k −→ Hk (x, y, z)
−→ (u, v), defined by u = yx , v = yz2 , transforms the
equation x4 + y4 + kx2 y2 = z2 to v2 = u4 + ku2 + 1. For k  ±2 the family of curves Hk is nonsingular and contains
a rational point, for example (0, ±1), therefore for fixed k  ±2, Hk is birational equivalent to an elliptic curve Ek ,
hence the map

Hk −→ Ek (u, v)
−→ (X, Y) defined by X = 2u2 + 2v + k, Y = 2u(2u2 + 2v + k),

is an isomorphism, with inverse

Y X − k Y
2
Ek −→ Hk (X, Y)
−→ (u, v), defined by u = , v= − .
2X 2 2X
 
Y 2 Y(k2 −4−X 2 )
The map Ek −→ Ek (X, Y)
−→ (x , y ), defined by x = 2X , y = 8X 2
is an isogeny,
 2
y y (1 − x2 )
and the dual isogeny Ek  
−→ Ek (x , y )
−→ (X, Y), is defined by X =  , Y= .
x x2

110002-4
Let S k = {(x, y, z) ∈ N3 | x4 + y4 + kx2 y2 = z2 , gcd(x, y, z) = 1, xy > 1} be the set of nontrivial solutions. Using the
map T k −→ Hk , the isomorphism Hk −→ Ek and the isogenies Ek −→ Ek −→ Ek , we obtain for the cardinality of S k :
 2 √
x = x x = y x
|S k | ≥ 1 ⇔ y = ± yx . z2 ⇔
y  2
y = y, z = y√yx ⇔ rank Ek (Q) ≥ 1,
y    
(x, y, z) ∈ S k (x , y ) ∈ Ek (Q) − Ek (Q)tor

where Ek (Q) and Ek (Q)tor are respectively Mordell-Weil group of rational points and its torsion subgroup for Ek .
Consequently, the initial equation has a solution in S k , if and only if the rank of Ek (Q) is at least 1 (as noted in [2]):

{k ∈ Z | card S k ≥ 1} = {k ∈ Z | rank Ek (Q) ≥ 1}.

The rational torsion points of Ek generate only the trivial solutions x = y = 1, with k in the form k = n2 − 2, which are
not included in S k . In what follows, the above statements are formulated as lemmas with their proofs.

Lemma 1 The rational torsion points of Ek generate only the trivial solutions x = y = 1, with k in the form
k = n − 2, which are not included in S k .
2

Proof 1 Let P = (x , y ) ∈ Ek and denote [m]P = (xm


, ym ), where (x , y ) = (x1 , y1 ). Then by the group law of Ek ,
(using duplication formula [8]) one obtains

(x2 − 1)2 (x2 − 1)(x4 + 2kx3 + 6x2 + 2kx + 1)

x2 = 
, y2 = (3)
(2y ) 2 (2y )3
Assume that P ∈ Ek (Q)[m], i.e. P ∈ Ek (Q) with [m]P = O. Then, by theorem 3, it follows that xm

and ym are integers.
By theorem 5, it follows 1 ≤ m ≤ 12 and m  11.

case 1: m = 2. Then [2]P = O ⇐⇒ P = −P ⇐⇒ (x , y ) = (x , −y ) ⇐⇒ y = 0. Consequently x = 0 or

x2 + kx + 1 = 0 which is equivalent to x = 0 or k2 − 4 is a perfect square, i.e. k = ±2. Therefore the only point of
order 2 is P = (0, 0).

case 2: m = 4. So [4]P = O ⇐⇒ [2]P = −[2]P ⇐⇒ (x2 , y2 ) = (x2 , −y2 ) ⇐⇒ y2 = 0. Consequently x2 = 0 and by
(3) we obtain x2 − 1 = 0, i.e. x = ±1. Therefore every rational torsion of order 4 must be of the type: P(1, n) with k
necessarily in the form k = n2 − 2, n  0, 2, or P(−1, n) with k necessarily in the form k = n2 + 2, n  0. Therefore
there are no rational torsion points of order 4, if k  n2 ± 2.

case 3: m = 8. So [8]P = O ⇐⇒ [4]P = −[4]P ⇐⇒ (x4 , y4 ) = (x4 , −y4 ) ⇐⇒ y4 = 0. Consequently [4]P is
a two-torsion point and [2]P must be a four-torsion. Then by cases 1 and 2, it follows that, x4 = y4 = 0 and
2
−1)2
x2 = ±1, y2 = n, k = n2 ∓ 2. We obtain [4]P = (0, 0), [2]P = (±1, n) and by (3), it follows that 4(x(x
3 +kx2 +x ) = ±1

which have no solutions in integers x . Therefore rational torsion points of order 8 do not exist.
2
−1) 2
case 4: m = 3. Then [3]P = O ⇐⇒ [2]P = −P ⇐⇒ (x2 , y2 ) = (x , −y ) ⇐⇒ x2 = x . Consequently 4(x(x3 +kx2 +x ) = ±x


which has no solutions in integers x . Therefore rational torsion points of order 3 do not exist which means that
Ek (Q)tors has no subgroups of order 3. Consequently rational torsion points of order 6, 9 and 12 do not exist.
−1) 2 2
case 5: m = 5. So [5]P = O ⇐⇒ [4]P = −P ⇐⇒ (x4 , y4 ) = (x , −y ) ⇐⇒ x4 = x . Let us denote f (x) = 4(x(x3 +kx 2 +x) .
       
Then x2 = f (x ) and x4 = f (x2 ) = f ( f (x )). Consequently f ( f (x )) = x which has no solutions in integers x . Indeed,
P(x)
let us write the rational function f ( f (x)) − x as a quotient of two polynomials: f ( f (x)) − x = Q(x) . Then P(0) = 1,
 
therefore from P(x ) = 0 it follows that x = ±1, which is impossible. Therefore rational torsion points of order 5 do
not exist, which means that Ek (Q)tors has no subgroups of order 5. The latter implies that rational torsion points of
order 10 do not exist.

case 6: m = 7. Then [7]P = O ⇐⇒ [8]P = P ⇐⇒ (x8 , y8 ) = (x , y ) ⇐⇒ x8 = x . We obtain the equation
x1 = f ( f ( f (x ))) which has no solutions in integers x with the same arguments as in case 5. Therefore rational

110002-5
torsion points of order 7 do not exist.

Z/4Z, k = n2 ± 2
Summarizing the results: Ek (Q)tors 
Z/2Z, k  n2 ± 2

Lemma 2 The rational torsion points of Ek generate only the trivial solutions x = y = 1, with k in the form
k = n2 − 2 which are not included in S k .

Proof 2 Similar to the proof of lemma 1. Some considerations may be reduced by using theorem 4. Since Ek and
Ek are isogenous under isogeny Ek −→ Ek with kernel of order 2 (kernel = {O, (0, 0)}), then odd torsion subgroups of
that curves are isomorphic (and hence trivial by lemma 1), but even torsion subgroups are not the same:

Z/2Z ⊕ Z/4Z, k = n2 − 2
Ek (Q)tors  , i.e.
Z/2Z ⊕ Z/2Z, k  n2 − 2


{O, (0, 0), (n2 − 4, 0), (n2 , 0), (n2 ± 2n, 2(n2 ± 2n))}, k = n2 − 2
Ek (Q)tors =
{O, (0, 0), (k − 2, 0), (k + 2, 0)}, k  n2 − 2
with  = ±1.

Lemma 3 {k ∈ Z | card S k ≥ 1} = {k ∈ Z | rank Ek (Q) ≥ 1}.

Proof 3 From a direct application of lemma 1 and the calculations above: if (x, y, z) ∈ S k , then the point
P(x , y ) ∈ Ek with coordinates x = (x/y)2 , y = ± xz/y3 is rational and by lemma 1 that point is nontorsion, since
x = (x/y)2  0, ±1. Thus rank Ek (Q) ≥ 1.

Let assume that rank Ek (Q) ≥ 1 and let P(x , y ) ∈ Ek is a rational nontorsion point. Let x = x1 /x2 , y =
y1 /y2 , gcd(x1 , x2 ) = gcd(y1 , y2 ) = 1 and let substitute in the equation of Ek :

x23 y21 = y22 x1 (x12 + kx1 x2 + x22 ), (4)

so x23 = ±y22 , therefore x2 = ±s2 , y2 = s3 and y21 = ±x1 (x12 ± kx1 s2 + s4 ). From gcd(x1 , s) = 1, one obtains gcd(x1 , x12 ±
kx1 s2 + s4 ) = 1. Then x1 = t2, y1 = tw and consequently w2 = ±(t4 ± kt2 s2 + s4 ). Then every nontorsion point on
Ek (Q) has the type (x , y ) = ± st 2 , st . sw2 with t2  s2 and ts  0. If the sign is +, then (t, s, w) ∈ S k . Otherwise, if
2

 2 
P = (x , y ) = − st 2 , st . sw2 , then [2]P = (x2 , y2 ) such that

 2
t4 − s4 (t4 − s4 )[(t4 + s4 )2 + 4t4 s4 − 2kt2 s2 (t4 + s4 )]
x2 = , y2 = . (5)
2stw (2stw)3
 |t4 −s4 | 
2stw |w4 −(k2 −4)t4 s4 |
Therefore d , d , d2
∈ S k , where gcd(t4 − s4 , 2stw) = d.

Remark 6 In the case k = ±2 the surface T k is degenerate: T −2 consists of two hyperbolic paraboloids, since
T −2 : (x2 − y2 + z)(x2 − y2 − z) = 0; T +2 consists of two elliptic paraboloids: T +2 : (x2 + y2 + z)(x2 + y2 − z) = 0. The
corresponding
positive
integral
solutions
are
(a,
b,
|a2
−
b2|),
(a,
b,
a2
+
b2),
a

b.

&21&/86,21
In
this
subsection
is
given
a
complete
parametric
description
of
the
non-trivial
solutions,
an
example
of
higher
rank

curves
in
the
elliptic
family
Ek ,
motivation
and
a
description
via
matrix
equations
of
theorem
9.
Our
main
result
is
the

following.

110002-6
Theorem 9 The equation x4 + y4 + kx2 y2 = z2 has a non trivial solution, i.e. solution in S k , if and only if k satisfies
at least one of the following systems:

x1 , x2 , y1 , y2 − odd y1 , y2 − odd x1 , x2 , y1 , y2 − odd
βy22 − αy21 = x12 βy22 − αy21 = 4x12 βy22 − αy21 = 2x12
δy2 − γy2 = x2 δy2 − γy2 = 4x2 δy2 − γy2 = 2x2
2 1 2 2 1 2 2 1 2
αδ − βγ = 1 αδ − βγ = 4 αδ − βγ = 4
k = 4αδ − 2 k = αδ − 2 k = αδ − 2,

where α, β, γ, δ, x1 , x2 , y1 , y2 are nonzero integers satisfying the conditions gcd(x1 , x2 ) = gcd(y1 , y2 ) =

gcd(x1 x2 , y1 y2 ) = 1, x1 x2 y1 y2 > 1, (with equality x1 x2 y1 y2 = 1 only possible for the second system of equations).
Solutions of the equation for the corresponding three cases are:

x = x1 x2 x = 2x1 x2 x = x1 x2
y = y y y = y1 y2
y = y y
1 2 1 2
βδy4 −αγy4 βδy4 −αγy4
4

z = βδy2 − αγy1
4
z= 2 1
z = 2 1
4 4

Proof 4 Ek and Ek are isogenous, then by theorem 8 it follows that rank Ek = rank Ek . Lemma 3 gives card S k ≥ 1
if and only if rank Ek ≥ 1. For Ek we may apply complete 2 descent algorithm, since all two-torsions of Ek are rational,
i.e. Ek [2] = Ek (Q)[2] = {O, (0, 0), (k − 2, 0), (k + 2, 0)}.
Assume that k has the form k = αδ − 2 = βγ + 2,  = 1 or  = 4, with (α, β, γ, δ, x1 , x2 , y1 , y2 ) as in the theorem.
Then, one could check directly that the triple (x, y, z) corresponding to k (as in the theorem) is a nontrivial solution,
and by lemma 3 that solution comes from a rational nontorsion point of E. Thus rank Ek ≥ 1.
Assume that rank Ek ≥ 1. We are going to prove that k has the form k = αδ − 2 = βγ + 2. Applying theorems 6 and 7
to Ek , with e1 = 0, e2 = k − 2, e3 = k + 2, we obtain the following: there exist a square-free integers b1 , b2 ∈ Q(S , 2),
such that the system of equations

b1 z21 − b2 z22 = k − 2, b1 z21 − b1 b2 z23 = k + 2, (6)

has a solution (z1 , z2 , z3 ) ∈ Q∗ ×Q∗ ×Q∗ and P = (b1 z21 , b1 b2 z1 z2 z3 ) is a nontorsion point. The condition b1 , b2 ∈ Q(S , 2)
is equivalent to:

• 16(k2 − 4) = ΔEk ≡ 0 mod bi for i = 1, 2

• p2  b1 , p2  b2 for all primes p.
The proof will be accomplished in several steps, formulated belows as lemmas.

Lemma 4 System (6) with the condition gcd(b1 , b2 ) = 1 is equivalent to the union of the following four systems:

b1 Z12 ± Z22 = (k − 2)Z 2 , b1 Z12 ± b1 Z32 = (k + 2)Z 2 ; (7)

b1 Z12 ± 2Z22 = (k − 2)Z 2 , b1 Z12 ± 2b1 Z32 = (k + 2)Z 2 ; (8)

where Z, Z1 , Z2 , Z3 are positive integers, satisfying gcd(Z, Zi ) = gcd(Z, b1 ) = 1, i = 1, 2, 3 and k + 2 ≡ 0 mod b1 .

Lemma 4 states that b2 = ±1 or b2 = ±2, and if zi = ZZiii , gcd(Zi , Zii ) = 1, i = 1, 2, 3, then Z11 = Z22 = Z33 (we set
them equal to Z). Now we prove that statement for all four cases. System (6) is equivalent to

b1 Z12 Z22
2
− b2 Z22 Z11
2
= (k − 2)Z11
2 2
Z22 , b1 Z12 Z33
2
− b1 b2 Z32 Z11
2
= (k + 2)Z11
2 2
Z33 . (9)

From the first equation one obtains b1 Z22 2

≡ 0 mod Z11
2 2
and b2 Z11 ≡ 0 mod Z222
. Then by the square-free property
of b1 , b2 , it follows that Z11 and Z22 have the same set of prime divisors and the same powers for each prime in
the product decomposition, so Z11 = Z22 . From the second equation of (9), one could imply that Z11 = Z33 . Thus
Z11 = Z22 = Z33 (= Z) and dividing by Z 2 in (9), we obtain:

b1 Z12 − b2 Z22 = (k − 2)Z 2 , b1 Z12 − b1 b2 Z32 = (k + 2)Z 2 . (10)

110002-7
If gcd(b1 , Z) = d1 , then by the first equation of (10), we will get b2 Z22 ≡ 0 mod d1 . Thus Z22 ≡ 0 mod d1 and
gcd(Z, Z2 ) ≡ 0 mod d1 . Since gcd(Z, Z2 ) = 1, then d1 = 1 and by the second equation of (10), we will have that
k + 2 ≡ 0 mod b1 . Subtracting the equations of (10), we obtain b2 (Z22 − b1 Z32 ) = 4Z 2 . Therefore 4 ≡ 0 mod b2 , since
gcd(b1 , b2 ) = 1 and gcd(Z, b2 ) = 1. This means that b2 = ±1 or b2 = ±2 which completes the proof of lemma 4.

A generalization of lemma 4 is:

Lemma 5 Let gcd(b1 , b2 ) = d. System (6) is equivalent to the union of the following four systems:
d(eZ12 ± Z22 ) = (k − 2)Z 2 , e(dZ12 ± Z32 ) = (k + 2)Z 2 ; (11)

d(eZ12 ± 2Z22 ) = (k − 2)Z 2 , e(dZ12 ± 2Z32 ) = (k + 2)Z 2 ; (12)

where Z, Z1 , Z2 , Z3 are positive integers, satisfying gcd(Z, Zi ) = gcd(Z, e) = gcd(e, d)
= 1, i = 1, 2, 3 and k + 2 ≡ 0 mod e, k − 2 ≡ 0 mod d, and p2  de for all primes p.
If b1 = de, b2 = d f , then gcd(d, e) = gcd(d, f ) = gcd(e, f ) = 1 and d, e, f are square-free. As in lemma 4, we
obtain Z11 = Z22 = Z and (6) is equivalent to:
b1 Z12 − b2 Z22 = (k − 2)Z 2 , b1 Z12 Z33
2
− b1 b2 Z32 Z 2 = (k + 2)Z33
2 2
Z . (13)
From the second equation of (13), one obtains b1 Z33 2
≡ 0 mod Z 2 and b1 b2 Z 2 ≡ 0 mod Z33 2
. Then by the square-free
property of b1 , b2 , there exists divisor d1 of d such that Z33 = d1 Z. Thus (13) is equivalent to
 2
d
b1 Z12 − b2 Z22 = (k − 2)Z 2 , b1 Z12 − e f Z32 = (k + 2)Z 2 . (14)
d1
Subtracting the equations of (14), we obtain
 2
d
b2 Z22 − ef Z32 = 4Z 2 , (15)
d1

Therefore 4Z 2 ≡ 0 mod dd1 and 4Z 2 ≡ 0 mod f . By the second equation of (14), gcd( f, Z) = gcd(d/d1 , Z) = 1 which
implies that the square-free integers d/d1 and f are divisors of 4. Hence d/d1 = 1 or 2, f = ±1 or ±2. There are
four cases:
• (d/d1 , f ) = (1, ±1). Then (14) is equivalent to (11),
• (d/d1 , f ) = (1, ±2). Then (14) is equivalent to (12),
• (d/d1 , f ) = (2, ±1). Then (14) is equivalent to a subsystem of (11),
• (d/d1 , f ) = (2, ±2). Then (14) is equivalent to a subsystem of (12).
It is straightforward that gcd(e, Z) = gcd(d, Z) = 1 which implies k + 2 ≡ 0 mod e and k − 2 ≡ 0 mod d. This
completes the proof of lemma 5.

By using lemma 5, equation (11) gives us

k−2 2 k+2 2
eZ12 − Z = ∓Z22 , dZ12 − Z = ∓Z32 . (16)
d e
There are three cases:

case 1: Z1 , Z2 , Z3 -even, Z-odd. Then k ≡ 2 mod 4 and Zi = 2Zi , i = 1, 2, 3 and Zi are odd. If the sign in (16) is minus,
then we have  
k−2 k+2
(α, β, γ, δ) = e, , d, and (x1 , x2 , y1 , y2 ) = (Z2 , Z3 , Z1 , Z).
4d 4e
Otherwise, if the sign is plus, then we have
 
k+2 k−2
(α, β, γ, δ) = , d, , e and (x1 , x2 , y1 , y2 ) = (Z3 , Z2 , Z, Z1 ).
4e 4d

110002-8
Therefore k = 4αδ − 2 = 4βγ + 2 and we obtain the first system of theorem 9.

In case 1, it is impossible to have Z = Z1 = Z2 = Z3 = 1 since the point P = (b1 z21 , b1 b2 z1 z2 z3 ) is a torsion.
Indeed, by lemma 2 and the calculations above
2
z1 = z2 = 2, z3 = , k = (2d ± 2)2 − 2, b1 = d2 ± d, b2 = ±d,
d
P = (4(d2 ± d), 8(d2 ± d)) ∈ Ek (Q)tor .
case 2: Z and Z1 are odd, Z2 and Z3 are even. Hence Z2 = 2Z2 , Z3 = 2Z3 and gcd(Z2 , Z3 ) = 1. If the sign in (16) is
minus, then we have
 
k−2 k+2
(α, β, γ, δ) = e, , d, and (x1 , x2 , y1 , y2 ) = (Z2 , Z3 , Z1 , Z).
d e

Otherwise, if the sign is plus, then we have

 
k+2 k−2
(α, β, γ, δ) = , d, , e and (x1 , x2 , y1 , y2 ) = (Z3 , Z2 , Z, Z1 ).
e d

It is possible to have Z = Z1 = Z2 = Z3 = 1

since the point P = (b1 z21 , b1 b2 z1 z2 z3 ) in this case is not a torsion. Indeed, by lemma 2 and the calculations above, we
obtain
2
z1 = 1, z2 = 2, z3 = , k = d2 ± 5d − 2, b1 = d2 ± d, b2 = ±d,
d
P = (d2 ± d, 4(d2 ± d))  Ek (Q)tor .
Therefore k = αδ − 2 = βγ + 2 and we obtain the second system of theorem 9.

By using lemma 5, equation (12), we obtain

k−2 2 k+2 2
eZ12 − Z = ∓2Z22 , dZ12 − Z = ∓2Z32 . (17)
d e
case 3: k ≡ 1 mod 2 and Z, Zi are odd. Hence d ≡ e ≡ 1 mod 2. If the sign in (17) is minus, then we have
 
k−2 k+2
(α, β, γ, δ) = e, , d, and (x1 , x2 , y1 , y2 ) = (Z2 , Z3 , Z1 , Z).
d e

Otherwise, if the sign is plus, we have

 
k+2 k−2
(α, β, γ, δ) = , d, , e and (x1 , x2 , y1 , y2 ) = (Z3 , Z2 , Z, Z1 ).
e d

Consequently k = αδ − 2 = βγ + 2 and we obtain the third system of theorem 9.

For case 3, it is impossible to have Z = Z1 = Z2 = Z3 = 1 since the point P = (b1 z21 , b1 b2 z1 z2 z3 ) is a torsion.
Indeed, by lemma 2 and the calculations above
1
z1 = z2 = 1, z3 = , k = (d ± 2)2 − 2, b1 = d2 ± 2d, b2 = ±2d,
d
P = (d2 ± 2d, 2(d2 ± 2d)) ∈ Ek (Q)tor .
All the other possibilities for the parity of Z and Zi lead to the same results, as already obtained above, which completes
the proof of the theorem.

110002-9
Remark 7 Theorem 9 allows compact description by matrix equations. Let M2 (Z) be the set of 2 × 2 matrices
with integral elements and for convenience, let us introduce the following definition: two 2-dimensional vectors X and
Y are called perfect pair if their coordinates are integral, perfect squares and pairwise coprime, with product greater
than 1. Then theorem 9 states that all values of k come from such matrix g ∈ M2 (Z), with nonzero elements, for which
the equation
g ◦ Y = X (18)
has a solution (X, Y) which is a perfect pair. In correspondence with the notations above, let
  ⎛ 2 ⎞ ⎛ 2 ⎞
⎜⎜⎜ x1 ⎟⎟⎟ ⎜⎜ y1 ⎟⎟⎟
g=
−α β
, X = ⎜⎜⎜⎝ ⎜ ⎟⎟⎟ , Y = ⎜⎜⎜⎜⎜
⎟ ⎟⎟⎟ , ε ∈ {1, 2, 4}.
⎟⎠ (19)
−γ δ ⎠ ⎝ 2
x22 y2

There are three cases that correspond to the three systems of equations in the theorem:

case 1: det(g) = −1, ε = 1, k = 4αδ − 2,

case 2: det(g) = −4, ε = 4, k = αδ − 2,
case 3: det(g) = −4, ε = 2, k = αδ − 2.

Case 1 gives all solutions (x, y, z), with x, y odd, z even, and k even.
Case 2 gives all (x, y, z), with x even, y, z odd, and k can be even or odd.
Case 3 gives all solutions (x, y, z), with x, y, z odd, and k odd.

Remark 8 To every generator P of the Mordell-Weil group Ek (Q), corresponds a unique matrix gP ∈ M2 (Z) given
by following short non-exact sequence:

0 −→ Ek (Q)/2Ek (Q) −→ Q(S , 2) × Q(S , 2) −→ M2 (Z)

 
−α β
P = (x, y)
−→ (b1 , b2 )
−→ gP = .
−γ δ
The middle map is an injective group homomorphism, defined in theorem 6 where S = {p − prime | ΔEk ≡ 0 mod p} ∪
{±1}. The third map is defined as follows. Let d = gcd(b1 , b2 ). Set ρ = 4, when k ≡ 2 mod 4 and x = r/s, r-even,
s-odd; and ρ = 1 otherwise. There are two cases that correspond to sign(b2 ) = −1 and sign(b2 ) = +1:
⎛ b1 ⎞ ⎛ (k+2)d ⎞
⎜⎜⎜ − d k−2
ρd
⎟⎟⎟ ⎜⎜⎜ − ρb1 d ⎟⎟⎟
⎜ ⎟⎟⎟ ⎜ ⎟⎟⎟
gP = ⎜⎜⎜⎜⎜ ⎟⎟⎟ , gP = ⎜⎜⎜⎜⎜ ⎟⎟ .
⎝ (k+2)d ⎠ ⎝ b1 ⎟ ⎠
−d ρb1 − k−2
ρd d

Example 2 Using remark 7 and the Magma software ([12]), we obtain higher rank curves in the elliptic family
Ek , with solutions of the general equation that correspond to generators of the Mordell-Weil group for these curves.
Consider the matrix gi ∈ M2 (Z)
   
−2 102 −2 4182
g1 = , g2 = ,
−287 14639 −7 14639
   
−29278 1394 −2 6
g3 = , g4 = .
−21 1 −4879 14639
The following equalities are satisfied, i.e. solutions of the equation (18) gi ◦ Y = 4X:
 2   2     2 
7 1 5032 1
g1 ◦ =4 , g2 ◦ =4 ,
12 122 112 82
       
52 372 192 12
g3 ◦ =4 , g4 ◦ =4 .
232 12 112 502

110002-10
Since det(gi ) = −4,  = 4, then case 2 of theorem 9 states that k = αi δi − 2 = 29 276 and the solutions of the general
equation are (xi , yi , zi ), such that

(x1 , y1 , z1 ) = (24, 7, 28 751), (x2 , y2 , z2 ) = (16, 5533, 34 156 471),

(x3 , y3 , z3 ) = (74, 115, 1 456 151), (x4 , y4 , z4 ) = (100, 209, 3 576 319).
By remark 8, gi corresponds to a generator Pi ∈ E29 276 (Q) for i = 1, 2, 3, 4. Then (xi , yi , zi ) correspond to generators
of E29 276 (Q) and rank E29 276 (Q) ≥ 4. It can be shown that rank E29 276 (Q) = 4 (see Table 1).

Example 3 Another rank four elliptic curve from the elliptic family Ek is given by k = 70 808. As in example 2,
the calculation of matrix generators g1 , g2 , g3 , g4 corresponding to generators of E70 808 (Q) is as follows: consider the
matrix gi ∈ M2 (Z) :    
 −6 70810  −73 11801
g1 = , g2 = ,
−1 11801 −6 970
   
 −14162 70806  −2 70806
g3 = , g4 = .
−1 5 −1 35405
The following equalities are satisfied, i.e. solutions of the equation (18) gi ◦ Y = 4X:
    2    2   2 
−6 70810 1012 49 −73 11801 89 2
= 4 , = 4 .
1 11801 12 202 −6 970 72 12
         
−14162 70806 1992 792 −2 70806 1792 412
=4 , =4 .
−1 5 892 12 −1 35405 12 292
Since det(gi ) = −4,  = 4, then case 2 of theorem 9 states that k = αi δi − 2 = 70 808 and the solutions of the general
equation are (xi , yi , zi ), such that

(x1 , y1 , z1 ) = (1960, 101, 52 816 601), (x2 , y2 , z2 ) = (4, 623, 768 353),

(x3 , y3 , z3 ) = (158, 17711, 808 004 167), (x4 , y4 , z4 ) = (2378, 179, 113 408 767).
By remark 8, gi corresponds to a generator Pi ∈ E70 808 (Q) for i = 1, 2, 3, 4. Then (xi , yi , zi ) correspond to generators
of E70 808 (Q) and rank E70 808 (Q) ≥ 4. It can be shown that rank E70 808 (Q) = 4 (Table 1).

7$%/(+LJKHUUDQNFXUYHVLQWKHHOOLSWLFIDPLO\
Ek

k x y z rank Ek generators

29 276 24 7 28 751 4 g1
16 5 533 34 156 471 g2
74 115 1 456 151 g3
100 209 3 576 319 g4

70 808 1960 101 52 816 601 4 g1

4 623 768 353 g2
158 17 711 808 004 167 g3
2378 179 113 408 767 g4

110002-11
Example 4 There are infinitely many integers k for which the main considered equation has a solution (x, y, z) in
distinct odd prime numbers. It is a necessary and sufficient condition that k is in the form

k = ±2 + n(2p2 ± 2q2 + np2 q2 ), (20)

where p, q and |p2 ± q2 + np2 q2 | are primes, for some n ∈ Z.

Proof: When p and q are distinct odd primes, the arithmetic progression {p2 ± q2 + np2 q2 }n∈Z contains in-
finitely many primes, by the Dirichlet theorem. When k has the form (20), then a solution in prime numbers
is
(x, y, z) = (p, q, |p2 ± q2 + np2 q2 |).
If (x, y, z) = (p, q, r) is a solution in primes, then by theorem 9, case 3 of remark 7 and the following identities
  2   2 
−(2 + np2 ) 2p2 + 2q2 + np2 q2 q p
=2 ,
−n 2 + nq2 1 1
    
−n 2 + nq2 q2 1
=2 ,
−(np2 − 2) 2p − 2q2 + np2 q2
2 1 p2
it follows that r = |p2 ± q2 + np2 q2 | for some n ∈ Z and k has the form (20).

Remark 9 The motivation for theorem 9 is the following observation concerning the possible values of k:
 
z2 − x4 − y4 1 (z − y2 )(z + y2 )
k= = 2 −x ,
2
(21)
x 2 y2 y x2

where (x, y, z) ∈ S k and we may assume that y is odd since gcd(x, y, z) = 1. There are two cases:

case 1: z ≡ 0 mod 2. Then x is odd and gcd(z − y2 , z + y2 ) = 1. From (21), one could see that (z − y2 )(z + y2 )/x2 is an
integer. Therefore there exist odd integers x1 , x2 , t1 , t2 such that:

z − y2 = t1 x12 , z + y2 = t2 x22 , x = x1 x2 , gcd(t1 x1 , t2 x2 ) = 1. (22)

We obtain k = (t1 t2 − x12 x22 )/y2 and 2y2 = t2 x22 − t1 x12 . Consequently,

t1 t2 ≡ x12 x22 mod y2 , t1 x12 ≡ t2 x22 mod y2 , (23)

where gcd(y, t1 t2 x1 x2 ) = 1 and t1 t22 ≡ t2 (x1 x2 )2 ≡ x12 (t2 x22 ) ≡ t1 x14 mod y2 . Thus (t2 − x12 )(t2 + x12 ) ≡ 0 mod y2 and
similarly (t1 − x22 )(t1 + x22 ) ≡ 0 mod y2 . In addition, there exist nonzero even integers A, B, C, D such that:

t2 − x12 = Ay21 , t2 + x12 = By22 , t1 − x22 = Cy21 , t1 + x22 = Dy22 , y = y1 y2 , (24)

where gcd(y1 , y2 ) = 1. Solving the equations (24), we obtain

t1 = (Cy21 + Dy22 )/2, t2 = (Ay21 + By22 )/2, (25)

x12 = (By22 − Ay21 )/2, x22 = (Dy22 − Cy21 )/2. (26)

Finally, k = (t1 t2 − = (AD + BC)/2 and 2y =
x12 x22 )/y2 2
− = (AD − BC)y /2. Therefore AD = k + 2 and
t2 x22 t1 x12 2

BC = k − 2. Let (α, β, γ, δ) = (A/2, B/2, C/2, D/2). Then case 1 is equivalent to the existence of nonzero integers
α, β, γ, δ, such that:
x12 = βy22 − αy21 , x22 = δy22 − γy21 , αδ = (k + 2)/4, βγ = (k − 2)/4. (27)
Hence, we obtain the first system for k in theorem 9.

110002-12
 2 2
case 2: z ≡ 1 mod 2. Then gcd( z−y z+y
2 , 2 ) = 1. There are two cases for the parity of x.

case 2.1: x ≡ 0 mod 2. Then x = 2x1 x2 , gcd(x1 , x2 ) = 1. Similarly as in case 1, we obtain

z − y2 = 2t1 x12 , z + y2 = 2t2 x22 , gcd(t1 x1 , t2 x2 ) = 1. (28)

There exist nonzero integers A, B, C, D such that:

t2 − 2x12 = Ay21 , t2 + 2x12 = By22 , t1 − 2x22 = Cy21 , t1 + 2x22 = Dy22 , y = y1 y2 , (29)

where gcd(y1 , y2 ) = 1, A ≡ B mod 4 and C ≡ D mod 4, A − C ≡ 1 mod 2.

t1 = (Cy21 + Dy22 )/2, t2 = (Ay21 + By22 )/2, (30)

x12 = (By22 − Ay21 )/4, x22 = (Dy22 − Cy21 )/4. (31)

Finally, k = (t1 t2 − = (AD + BC)/2 and y =
4x12 x22 )/y2 − 2
= (AD − BC)y /4. Therefore AD = k + 2 and
t2 x22 t1 x12 2

BC = k − 2. Let (α, β, γ, δ) = (A, B, C, D). Then case 2.1 is equivalent to the existence of nonzero integers α, β, γ, δ,
such that:
(2x1 )2 = βy22 − αy21 , (2x2 )2 = δy22 − γy21 , αδ = k + 2, βγ = k − 2 (32)
Therefore, we obtain the second system for k in theorem 9.

case 2.2: x ≡ 1 mod 2. Then x = x1 x2 , gcd(x1 , x2 ) = 1. Similarly to the previous cases, we obtain

z − y2 = 2t1 x12 , z + y2 = 2t2 x22 , gcd(t1 x1 , t2 x2 ) = 1, t1 − t2 ≡ 1 mod 2. (33)

There exist nonzero integers A, B, C, D such that:

2t2 − x12 = Ay21 , 2t2 + x12 = By22 , 2t1 − x22 = Cy21 , 2t1 + x22 = Dy22 , y = y1 y2 , (34)

where gcd(y1 , y2 ) = 1, A + B ≡ C + D ≡ A + C ≡ B + D ≡ 0 mod 4. Therefore:

t1 = (Cy21 + Dy22 )/4, t2 = (Ay21 + By22 )/4, (35)

x12 = (By22 − Ay21 )/2, x22 = (Dy22 − Cy21 )/2. (36)

Finally, k = (4t1 t2 − x12 x22 )/y2 = (AD + BC)/2 and y2 = t2 x22 − t1 x12 = (AD − BC)y2 /4. Consequently, AD = k + 2
and BC = k − 2. Let (α, β, γ, δ) = (A, B, C, D). Then case 2.2 is equivalent to the existence of nonzero odd integers
α, β, γ, δ, such that:
2x12 = βy22 − αy21 , 2x22 = δy22 − γy21 , αδ = k + 2, βγ = k − 2 (37)
Hence,
we
obtain
the
third
system
for
k
in
theorem
9
which
completes
the
survey.1.

$33(1',&(6
In
this
section
are
given
definitions
of
Z p ,
Q p
and
the
reduction
map
modulo
p.
Let
p
be
a
prime
number,
E/Q
be
an

elliptic
curve
over
Q,
and
F p
=
Z/pZ
be
a
finite
field
with
p
elements.
Let
us
denote
by

∞
Z/pn Z, ρn : Z/pn+1 Z −→ Z/pn Z a + pn+1 Z
−→ a + pn Z
n=1

the direct product of the rings Z/pn Z, n = 1, 2, ... and the cannonical projections.
1 Important results for considered diophantine problem are obtained in [2],[3],[5],[7],[10],[11]

110002-13
Definition 9 The ring Z p of p-adic integers is defined by

∞
Z p = {(xn ) ∈ Z/pn Z | ρn (xn+1 ) = xn },
n=1

with operations given by componentwise addition and multiplication.

Remark 10 The map Z −→ Z p n

−→ (n+ pZ, n+ p2 Z, . . . , n+ pn Z, . . .) is inclusion, therefore Z can be considered
as subring of Z p .

Definition 10 The field Q p of the p-adic rational numbers is defined as a field of fractions for Z p .

Remark 11 The map Q −→ Q p a

−→ a is inclusion, thus Q can be considered as subfield of Q p .

Remark 12 The map Z p −→ F p a

−→ a + pZ is surjective ring homomorphism with kernel pZ p . Therefore

Z p /pZ p  F p .

Remark 13 E/Q : y2 = x3 + ax2 + bx + c can be transformed to an elliptic curve with integral coefficients: let
u be the least common multiplier of the denominators of the coefficients a, b, c. The change (x, y)
−→ (X, Y) defined
by X = u2 x, Y = u3 y transforms the equation to the equation Y 2 = X 3 + au2 X 2 + bu4 X + cu6 which has integral
coefficients.

Remark 14 Let E/Q p be an elliptic curve over Q p . With a modification of the above remark, we may assume that
E/Q p has coefficients in Z p . By reduction of the coefficients of E modulo pZ p , we obtain a curve Ẽ with coefficients
in Z p /pZ p  F p . The map E/Q p −→ Ẽ/F p is called reduction.

Acknowledgement
Miroslav Stoenchev is supported by the Bulgarian National Science Fund under Young Scientists Project KP-06
M32/2-17.12.2019. Venelin Todorov is supported by the NSP ”Information and Communication Technologies for a
Single Digital Market in Science, Education and Security (ICT in SES)”, contract No DO1-205/23.11.2018 and by
the BNSF under Project DN 12/5-2017.

REFERENCES
[1] S. Apostolov, M. Stoenchev, V. Todorov, One parameter family of elliptic curves and the equation x4 + y4 +
kx2 y2 = z4 , Studies in Computational Intelligence, Springer, (2020)
[2] A. Bremner, J. Jones, On the equation x4 + y4 + mx2 y2 = z2 , Journal of Number Theory 50, 286-298 (1995).
[3] E. Brown, x4 +y4 +mx2 y2 = z2 : Some cases with only trivial solutions - and a solution Euler missed, Glasgow
Math. J.31 (1989) 297-307.
[4] J.W.S. Cassels, Lectures on elliptic curves, Cambridge University Press, 1991.
[5] L. Euler, De casibus quibus formulam x4 + mxxyy + y4 ad quadratum reducere licet, Mem.acad.sci. St.
Petersbourg 7 (1815/16, 1820), 10-22; Opera Omnia, ser. I, V, 35-47, Geneva, 1944.
[6] V. A. Kolyvagin, On the Mordell-Weil Group and the Shafarevich-Tate Group of Modular Elliptic Curves ,
Proceedings of the International Congress of Mathematicians, Kyoto, Japan, 1990 ,pp. 429-436.
[7] H. C. Pocklington, Some diophantine impossibilities, Proc. Cambridge Phil. Soc. 17 (1914), 108-121.
[8] J. H. Silverman, The arithmetic of elliptic curves, Springer Verlag, New York/Berlin, 1986.
[9] J. H. Silverman, J. Tate, Rational points on elliptic curves, Springer Verlag, New York, 1992.
[10] T. N. Sinha, A class of quartic diophantine equations with only trivial solutions, Amer.J.Math. 100 (1978),
585-590.
[11] M. Z. Zhang, On the diophantine equation x4 + kx2 y2 + y4 = z2 , Sichuan Daxue Xuebao 2 (1983), 24-31.
[12] http://magma.maths.usyd.edu.au/calc/
[13] https://math.mit.edu/classes/18.783/2017/lectures.html

110002-14