Proc. Indian Acad. Sci. (Math. Sci.

) (2019) 129:81
https://doi.org/10.1007/s12044-019-0524-6

On solutions of the diophantine equation Fn − Fm = 3a

BAHAR DEMİRTÜRK BİTİM1,∗ and REFİK KESKİN2

1 Department of Fundamental Sciences, Izmir Bakırçay University, Izmir, Turkey
2 Department of Mathematics, Sakarya University, Adapazarı, Turkey
*Corresponding author.
E-mail: [email protected]; [email protected]

MS received 19 March 2018; revised 17 April 2019; accepted 3 May 2019

Abstract. In this paper, we find non-negative (n, m, a) integer solutions of the

diophantine equation Fn − Fm = 3a , where Fn and Fm are Fibonacci numbers. For
proving our theorem, we use lower bounds in linear forms.

Keywords. Diophantine equation; lower bounds; logarithmic method.

2000 Mathematics Subject Classification. 11B39, 11J86, 11D61.

1. Introduction
The Fibonacci sequence (Fn ) is given by the recurrence relation

Fn = Fn−1 + Fn−2 , n ≥ 2

with initial conditions F0 = 0, F1 = 1. The Lucas sequence (L n ) is given by the recurrence

relation

L n = L n−1 + L n−2 , n ≥ 2

with initial conditions L 0 = 2, L 1 = 1. Fn and L n are called the n-th terms of Fibonacci
and Lucas sequences, respectively. The Binet formulas for these sequences are given by

αn − β n
Fn = and L n = αn + β n ,
α−β
√ √
1+ 5 1− 5
where α = and β = . It is easy to see that L n = Fn−1 + Fn+1 and
2 2
5Fn = L n−1 + L n+1 . Many properties of these recurrence sequences are given in [6,12].
In [2], Bravo and Luca determined all non-negative solutions (n, m, a) of the diophantine
equation Fn + Fm = 2a with n ≥ m. Then in [10], Pink and Zeigler considered a more
general form u n + u m = wp1z 1 · · · psz s in non-negative integers n, m, z 1 , . . . , z s , where
(u n )n≥0 is a binary nondegenerate recurrence sequence, p1 , . . . , ps are the distinct primes

© Indian Academy of Sciences

0123456789().: V,-vol
 81 Page 2 of 10 Proc. Indian Acad. Sci. (Math. Sci.) (2019) 129:81

and w is a non-zero integer with pi  w for all 1 ≤ i ≤ s. After noticing Pink and Zeigler’s
more general diophantine equation for sums of terms of recurrence sequence (u n )n≥0 , Şiar
and Keskin [11] proved that all non-negative integer solutions of the diophantine equation
Fn − Fm = 2a are given by

(n, m, a) ∈ {(1, 0, 0), (2, 0, 0), (3, 0, 1), (6, 0, 3), (3, 1, 0), (4, 1, 1),
(5, 1, 2), (3, 2, 0)}

and

(n, m, a) ∈ {(4, 3, 0), (4, 2, 1), (5, 2, 2), (9, 3, 5), (5, 4, 1), (7, 5, 3),
(8, 5, 4), (8, 7, 3)}.

From this point of [11], we consider the solutions of diophantine equation of the form
Fn − Fm = 3a in non-negative integers. Moreover, Erduvan et al. [5] showed that the
solutions of the equation Fn − Fm = 5a is given by

F1 − F0 = F2 − F0 = F3 − F2 = F3 − F1 = 50

and

F5 − F0 = F4 − F3 = F6 − F4 = F7 − F6 = 5,

in nonnegative integers m, n and a.

Now we can give the logarithmic height definition from [8].

DEFINITION 1

Let α be an algebraic number of degree d and


d
f (x) = ai x d−i ∈ Z[x]
i=0

be the minimal polynomial of α with a0 > 0 and gcd(a0 , . . . , ad ) = 1. The logarithmic

height of α is given by
 
1 
d
(i)
h(α) = log |a0 | + log max{|α |, 1} ,
d
i=1

where α (i) ’s are the conjugates of α.

Some known properties of the logarithmic heights are as follows:

h(α ± β) ≤ h(α) + h(β) + log 2, (1.1)
±1
h(αβ ) ≤ h(α) + h(β), (1.2)
h(α k ) = |k| h(α). (1.3)
Proc. Indian Acad. Sci. (Math. Sci.) (2019) 129:81 Page 3 of 10 81

Now we can give the following lemmas from [7], that will be useful in the proof of
Theorem 1.

Lemma 1. If n ≡ m (mod 2), then


F(n−m)/2 L (n+m)/2 , n ≡ m (mod 4)
Fn − Fm =
F(n+m)/2 L (n−m)/2 , n ≡ m + 2 (mod 4).

Lemma 2. Let L n = 3s · y b for some integers n ≥ 1, y ≥ 1, b ≥ 2 and s ≥ 0. The

solutions of this equation are given by n ∈ {1, 2, 3} .

The following lemma was an interesting problem and was completely proved by
Bugeaud et al. [3]. This lemma will be used in the proof of Theorem 1.

Lemma 3. The only perfect powers in the Fibonacci sequence are F0 = 0, F1 = F2 =

1, F6 = 8 and F12 = 144.

In [1], Baker gave an effective lower bound for a non-zero expression of the form
c1 log α1 + · · · + cn log αn , where αi are the algebraic numbers and ci are the integers
for all 1 ≤ i ≤ n. We will use the reduction method of Baker–Davenport in the proof of
Theorem 1.

2. Preliminaries
Before proving the main theorem, we shall state a useful inequality associated with
Fibonacci sequence.

Lemma 4. Let n ≥ 1. Then α n−2 ≤ Fn ≤ α n−1 .

Since the proof of the following lemma is given firstly in [9] and then in [3], we will
omit its proof.

Lemma 5 [3,9]. Let L be a number field of degree D and α1 , α2 , . . . , αn be non-zero

elements of L, and let b1 , b2 , . . . , bn be rational integers such that

 = a1b1 · · · anbn − 1

and

B = max {|b1 | , . . . , |bn |} .

Let h denote the absolute logarithmic height and A1 , . . . , An be real numbers with

A j ≥ max{Dh(α j ), | log α j |, 0.16} for all 1 ≤ j ≤ n.

If  = 0, then

log || > −3 · 30n+4 (n + 1)5.5 D 2 (1 + log D)(1 + log n B)A1 · · · An .

 81 Page 4 of 10 Proc. Indian Acad. Sci. (Math. Sci.) (2019) 129:81

Furthermore, if L is real, then

log || > −1.4 · 30n+3 n 4.5 D 2 (1 + log D)(1 + log B)A1 · · · An .

We give the following lemma from [4].

Lemma 6. Let M be a positive integer and p/q be a convergent of the continued fraction
of the irrational number γ such that q > 6M. Let A, B, μ be some real numbers with
A > 0 and B > 1. Let ε := μq − M γ q , where · denotes the distance from the
nearest integer. If ε > 0, then there exist no solutions of the inequality

0 < |uγ − v + μ| < AB −w ,

in positive integers u, v and w with

log(Aq/ε)
u≤M and w ≥ .
log B

3. Main theorem
Theorem 1. Let n, m and a be non-negative integers with n > m. Then all solutions of
the equation

Fn − Fm = 3a (3.1)

are given by

(n, m, a) ∈ {(1, 0, 0), (2, 0, 0), (4, 0, 1), (3, 1, 0), (3, 2, 0), (4, 3, 0), (5, 3, 1),
(6, 5, 1), (11, 6, 4)}.

Proof. If m = 0, then from Lemma 3, we get (n, m, a) = (1, 0, 0), (2, 0, 0), (4, 0, 1). Let
1 ≤ m < n ≤ 100. Then the solutions of (3.1) are (n, m, a) ∈ {(3, 1, 0), (3, 2, 0), (4, 3, 0),
(5, 3, 1), (6, 5, 1), (11, 6, 4)}. Thus, from now on, we will assume that n > 100. If n−m =
1, then we get 3a = Fn − Fm = Fm−1 . This implies that m − 1 = 1, 2, 4, by Lemma 3. So
m = 2, 3, 5. If n − m = 2, then 3a = Fn − Fm = Fm+1 . This implies that m + 1 = 2, 4, by
Lemma 3. So m = 1, 3. But these solutions are given in the theorem for 1 ≤ m < n ≤ 100.
Therefore, we may suppose n − m ≥ 3. Since 3a = Fn − Fm < Fn ≤ α n−1 < 3n−1 by
Lemma 4, we get a < n.
Now recalling Fn − Fm = 3a , we get

αn βn
√ − 3a = √ + Fm . (3.2)
5 5

Taking absolute value of (3.2), we have

 n 
α 
 √ − 3a  ≤ |β| |β|n
n 1
  √ + Fm < √ + α m < + α m , (3.3)
5 5 5 2
Proc. Indian Acad. Sci. (Math. Sci.) (2019) 129:81 Page 5 of 10 81

|β|n 1
 0.6 < 1 and √ < . Dividing both sides of (3.3) by
where we take into account |β| =
5 2
αn
√ , we obtain
5

√ √ 1 −m
|1 − 3a · α −n · 5| < 5 · α m−n · α +1 ,
2

which implies that


−n
√ √ 1 3√ 4
|1 − 3 · α
a
· 5| < 5 · α m−n · +1 = 5 · α m−n < n−m . (3.4)
2 2 α
√
Take the parameters t := 3, γ1 = 3, γ2 = α, γ3 = 5, b1 = a, b2 = −n and b3 = 1 in
Lemma√ 5. We also notice that D = 2 and γ1 , γ2 , γ3 are positive real numbers and belong
to Q( 5). √
Now it is necessary
√ to show that  = 1 − 3a · α −n · 5 is non-zero. Assume that
 = 1 − 3a · α −n · 5 = 0. Then we get α 2n = 5 · 3a , which is impossible since α 2n ∈
/ Z.
log α √ √
Thus we have h(γ1 ) = log 3, h(γ2 ) = h(α) = and h(γ3 ) = h( 5) = log 5.
2
Therefore, we can choose

A1 := 2.2 ≥ max{2 log 3, | log γ1 |, 0.16} = 2.1972, A2 := 0.5 , A3 := 1.7.

Now by considering a < n, we can take

B := max{|a|, | − n|, 1} = n.

According to Lemma 5, we get

√
|1 − 3a · α −n · 5| > exp{−1.4 × 306 × 34.5 × 4 × (1 + log 2)(1 + log n)
× 2.2 × 0.5 × 1.7}. (3.5)

If we take logarithms in equality (3.4) and combine the calculation of the right-hand side
of (3.5), we get

(n − m) log α < 18.139 × 1011 (1 + log n) + log 4. (3.6)

Using the fact that 1 + log n < 2 log n for all n ≥ 3, we have

(n − m) log α < 3.63 × 1012 log n. (3.7)

Moreover, we can obtain a second linear form by using equation (3.1) as follows:
 n   n 
α  β β m  |β|n + |β|m
 √ (1 − α m−n ) − 3a  =  √ −√ ≤ √ < 0.445. (3.8)
  
5 5 5 5
 81 Page 6 of 10 Proc. Indian Acad. Sci. (Math. Sci.) (2019) 129:81

αn
Now dividing both sides of (3.8) by √ (1 − α m−n ), we get
5
√
−n
√ m−n −1 0.445 × 5
|1 − 3 · α
a
· 5 · (1 − α ) |≤ n .
α (1 − α m−n )

1 1 1
Since α m−n = < < 0.67, it is obvious that < 2.71. Thus it follows
α n−m α 1 − α m−n
that

√ 2.71
|1 − 3a · α −n · 5 · (1 − α m−n )−1 | < n . (3.9)
α
√
Let  = 3a · α −n · 5 · (1 − α m−n )−1 − 1. Now, in order to apply Lemma 5 again, we
take parameters t := 3 and
√
γ1 = 3, γ2 = α, γ3 = 5 · (1 − α m−n )−1 , b1 = a, b2 = −n, b3 = 1.
√
As before, we have D := √ [Q( 5) : Q] = 2. It is obvious that  = 0. Because, if  = 0,
√ α − α = 3 5. Taking the conjugate of this equation, we have β − β =
we get n m a n m

−3 5. Addition of these two conjugate equations gives L n − L m = 0, which contradicts

a

the fact that n > m.

Hence the left-hand side of (3.9) is non-zero. As before, A1 = 2.2, A2 = 0.5 and B = n
in Lemma 5.
Now we calculate A3 . For all n − m > 3, we have | log γ3 | < 1. Taking the identities
(1.1), (1.2) and (1.3) into account, we get
√
h(γ3 ) = h( 5 · (1 − α m−n )−1 )
√
≤ h( 5) + h(1) + h(α m−n ) + log 2
√
≤ log 5 + |m − n| h(α) + log 2
√ log α
= log 5 + (n − m) + log 2
2
log α √
= (n − m) + log 2 5.
2

Thus, it follows that

A3 ≥ max{2h(γ3 ), | log γ3 |, 0.16} ≥ log 20 + (n − m) log α.

Hence, we can take A3 = log 20 + (n − m) log α.

By Lemma 5, we obtain a lower bound for the left-hand side of (3.9) as

3.03
> || > exp{−1.4 × 306 × 34.5 × 4 × (1 + log 2)(1 + log n)
αn
×2.2 × 0.5 × [log 20 + (n − m) log α]}.
Proc. Indian Acad. Sci. (Math. Sci.) (2019) 129:81 Page 7 of 10 81

Taking logarithms on both sides of this inequality and considering the fact that 1 + log n <
2 log n for all n > 1, we get

n log α < 2.134 × 1012 log n × log 20 + (n − m) log α . (3.10)

By a fast calculation with Mathematica, we obtain n < 7.09616 × 1028 .

If (n, m, a) is a positive integer solution of equation (3.1) with n > m, then we have
a < n < 7.09616 × 1028 . We have obtained an upper bound for n and now we will reduce
this bound to a size that can be easily dealt with. For doing this, we will use Lemma 4
again. √
Let z 1 := a log 3 − n log α + log 5. By considering equation (3.4), we get

  4
1 − e z 1  < .
α n−m

Thus, by using equation (3.1) and the Binet formula, we obtain

αn βn
√ = Fn + √ > Fn − 1 > Fn − Fm = 3a .
5 5

√ 4
Hence z 1 = log(3a 5/α n ) < 0. It is obvious that < 0.945 for all n − m ≥ 3.
α n−m
Therefore we get e|z 1 | < 18.2 and therefore, it follows that

73
0 < |z 1 | < e|z 1 | − 1 ≤ e|z 1 | |1 − ez 1 | < .
α n−m

Thus it can be seen that

√ 73
0 < |a log 3 − n log α + log 5| < . (3.11)
α n−m

Dividing both sides of the inequality (3.11) by log α, we get

 √ 
 log 3 log 5 
 73
0 < a −n+ < · α −(n−m) . (3.12)
 log α log α  log α

log 3
Now considering Lemma 6, we have the irrational γ = with
log α
√
log 5 73
μ= , A= , B = α, w = n − m.
log α log α

Also, we know that a < n < 7.09616 × 1028 . So it follows that M := 7.09616 × 1028 ,
according to Lemma 6 and q > 6M is the denominator of a convergent of the continued
fraction of γ such that ε = μq − M γ q > 0. Considering the denominator of the 61-st
 81 Page 8 of 10 Proc. Indian Acad. Sci. (Math. Sci.) (2019) 129:81

log 3
convergence of , we have q = 10.52×1029 . By some calculations with Mathematica,
log α
we obtain ε = 0.154453.
According to Lemma 6, we know that there is no solution of the inequality (3.12) for
log(Aq/ε)
the values n − m with n − m ≥ . Therefore, it follows that inequality (3.12)
log B
has no solutions for n − m ≥ 157.972. This means that a bound for n − m is n − m ≤ 157.
Considering this fact in inequality (3.10), we get n < 1.29184 × 1016 .
Let us work on (3.9) for finding an upper bound on n. Now take
√
z 2 := a log 3 − n log α + log( 5(1 − α m−n )−1 ).

Thus (3.9) implies that

2.71
|1 − ez 2 | < .
αn

2.71 1 2.71
It is obvious that < . If z 2 > 0, then 0 < z 2 < ez 2 − 1 < n . If z 2 < 0, then
α n 2 α
2.71 1 1
|1 − e | = 1 − e < n < . Thus, we get 1 − < e , so that e|z 2 | < 2. Therefore,
z 2 z 2 z 2
α 2 2
we have

2.71
0 < |z 2 | < e|z 2 | − 1 ≤ e|z 2 | · |1 − ez 2 | < 2 × .
αn

Thus it follows that

 √ 
 log 3 log 5 · (1 − α m−n )−1 
 5.42
0 < a −n+ < · α −n . (3.13)
 log α log α  log α

Now considering Lemma 6, we obtain

√
log 3 log 5 · (1 − α m−n )−1 5.42
γ = , μ= , A= , B = α, w = n.
log α log α log α

It is obvious that γ is irrational. Also, 3 ≤ n − m ≤ 157. Firstly, calculate the denominator

q of continued fraction of γ . Since M = 1.29184 × 1016 , we must choose the 39-th
denominator q = 48, 9 × 1016 such that q > 6M = 7.75107 × 1016 . Hence by applying
Lemma 6 to (3.13) with 3 ≤ n − m ≤ 157 except for n − m = 4 or 8, we obtain

ε = μq40 − M γ q40 ≥ 0.492868

by a fast computation with Mathematica. Furthermore, according to Lemma 6, we know

log(Aq/ε)
that there is no solution of the inequality (3.13) for values n with n ≥ =
log B
91.1453. Thus, an upper bound for n must be n ≤ 91. This contradicts our assumption
that n > 100.
Proc. Indian Acad. Sci. (Math. Sci.) (2019) 129:81 Page 9 of 10 81

Finally, we consider the cases n − m = 4 or 8. According to Lemma 1, when n ≡ m

(mod 4), we have Fn − Fm = F(n−m)/2 L (n+m)/2 . Thus, it follows that Fn − Fm =
F2 L m+2 = L m+2 for n − m = 4. This gives L m+2 = 3a . According to Lemma 2,
the possible values for m + 2 are 1, 2, 3. Since 4  3a and m is a non-negative integer,
we have that m + 2 = 3 and m + 2 = 1. If m + 2 = 2, then we get L 2 = 3a . Thus,
it follows that (n, m, a) = (4, 0, 1) is a solution to (3.1). Moreover, the case n − m = 8
gives Fn − Fm = F4 L m+4 by Lemma 1 . Hence we get L m+4 = 3a−1 , which implies that
m + 4 = 1, 2, 3 by Lemma 2. This is impossible since m is a non-negative integer. 

4. Conclusion
In [7], it is shown that if n ≡ m (mod 2), then all the solutions of the equation

Fn − Fm = y p , p ≥ 2, y ≥ 1 (4.1)

satisfy max{n, m} ≤ 36. Then the authors conjectured that all the solutions of equation
(4.1) are
F1 − F0 = 1, F2 − F0 = 1, F3 − F1 = 1, F3 − F2 = 1, F4 − F3 = 1,
F5 − F1 = 22 , F5 − F2 = 22 , F6 − F4 = 5, F7 − F5 = 23 ,
F7 − F6 = 5, F8 − F5 = 24 , F8 − F7 = 23 , F9 − F3 = 25 ,
F11 − F6 = 92 , F13 − F6 = 152 , F13 − F11 = 122 , F14 − F9 = 73 ,
F14 − F13 = 122 , F15 − F9 = 242 .
Consequently, it is true that the above conjecture is valid for y = 2, 3 by our result and
the results in [5,11]. It is reasonable to conjecture that if Fn − Fm = pa for some prime
p and positive integer a, then p = 2, 3, 5, 7.

References
[1] Baker A, Linear forms in the logarithms of algebraic numbers, Matematika 11(3) (1967) 155–
166
[2] Bravo J J and Luca F, On the diophantine equation Fn + Fm = 2a , Quaestiones Mathemati-
cae 39(3) (2015) 391–400
[3] Bugeaud Y, Mignotte M and Siksek S, Classical and modular approaches to exponential Dio-
phantine equations I: Fibonacci and Lucas perfect powers, Ann. Math. 163(3) (2006) 969–1018
[4] Dujella A and Pethö A, A generalization of a theorem of Baker and Davenport, Q. J. Math.
Oxf. Ser. (2) 49(3) (1998) 291–306
[5] Erduvan F, Duman M G and Keskin R, Nonnegative integer solutions of the equation Fn −Fm =
5a , Turk. J. Math. 43 (2019) 1115–1123
[6] Koshy T, Fibonacci and Lucas Numbers with Applications (2001) (New York: Wiley-
Interscience Publ.)
[7] Luca F and Patel V, On perfect powers that are sums of two Fibonacci numbers, J. Number
Theory 189 (2018) 90–96
[8] Luca F, Effective Methods for Diophantine Equations (2009) (Mexico: Universidad Nacional
Autonoma de Mexico)
[9] Matveev E M, An explicit lower bound for a homogenous rational linear form in the logarithms
of algebraic numbers II, Izv. Ross. Akad. Nauk Ser. Mat. 64(6) (2000) 1217–1269
 81 Page 10 of 10 Proc. Indian Acad. Sci. (Math. Sci.) (2019) 129:81

[10] Pink I and Ziegler V, Effective resolution of diophantine equations of the form u n + u m =
z
wp1z 1 · · · ps s , Monatsh Math. 185 (2018) 103–131
[11] Şiar Z and Keskin R, On the Diophantine equation Fn − Fm = 2a , Colloq. Math. (in press)
[12] Vajda S, Fibonacci and Lucas Numbers and the Golden Section (1989) (England: Ellis Horwood
Limited Publ.)

Communicating Editor: Sanoli Gun