Abstract
The well-known Fibonacci sequence has several generalizations, among them, the k-generalized Fibonacci sequence denoted by \(F^{(k)}\). The first k terms of this generalization are \(0, \ldots , 0, 1\) and each one afterward corresponds to the sum of the preceding k terms. For the Fibonacci sequence the formula \(F_{n+1}^2 - F_{n-1}^2 = F_{2n}\) holds for every \(n \ge 1\). In this paper, we study the above identity on the k-generalized Fibonacci sequence terms, completing the work done by Bensella et al. (On the exponential Diophantine equation \((F_{m+1}^{(k)})^x - (F_{m-1}^{(k)})^x = F_n^{(k)}\), 2022. arxiv:2205.13168).
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Let \(\{F_n\}_{n\ge 0}\) be the classical Fibonacci sequence given by \(F_0 = 0, \,\, F_1 = 1\) and \(F_{n+2} = F_{n+1}+F_{n}\) for all \(n\ge 0\). A familiar identity related to these numbers is given by the formula
which holds for all integers n if we extend the Fibonacci sequence to all integer indices using its recurrence formula.
Patel and Chaves [18] investigated an analogue of (1) in higher powers, namely the Diophantine equation
obtaining the following result.
Theorem 1
The Eq. (2) on (n, m, x) has only the non-negative integral solutions (n, 2n, 2), (1, 1, x), (1, 2, x), (n, 0, 0).
Let \(k\ge 2\) be an integer. A generalization of the Fibonacci sequence, called the k-generalized Fibonacci sequence, denoted by \(F^{(k)}:=\{F_{n}^{(k)}\}_{n\ge -(k-2)}\), is given by the linear recurrence
with the initial conditions \(F_{-(k-2)}^{(k)} = F_{-(k-3)}^{(k)} = \cdots = F_{0}^{(k)} = 0~ ~{{\text {and}}} ~~F_{1}^{(k)} = 1\). We refer to each \(F_{n}^{(k)}\) as the nth k-generalized Fibonacci number. Note that for \(k = 2\), we have \(F_n^{(2)} = F_n\), the usual nth Fibonacci number.
Recently, Bensella et al. [1] studied an analogue of Diophantine equation (2) in k-generalized Fibonacci numbers, showing that
has no positive integral solutions (k, n, m, x) with \(x \ge 2\) under the restriction that \(3 \le k \le \min \{n, \log x\}\). In this paper, we revisit the above Diophantine equation, remove the restriction from [1] and find parametric families of solutions. Namely, we establish the following result.
Main Theorem The only solutions (k, n, m, x) of the Diophantine equation (4) with \(k\ge 3\), \(n\ge 4\), \(m\ge 2\) and \(x\ge 2\) are
Solutions with at least one small index or exponent (so, which do not satisfy the inequalities from the hypothesis of our Main Theorem) are called trivial and appear in Theorem 3.
Before we move into details, let us do a brief description of the method. First, we use lower bounds for linear forms in logarithms of algebraic numbers to get a polynomial bound for n, m and x in terms of k. When k is small, we use a variation of a result due to Dujella and Pethő on continued fractions to lower the bounds into a computationally feasible range. When k is large, we use the fact that the dominant root of the k-generalized Fibonacci sequence is exponentially close to 2 and substitute it into our calculations, this way we get a simpler linear form in logarithms which allows us to bound k and complete our calculations.
The calculations were done with Mathematica and the running time was about one hour on 25 computers.
2 Preliminary Results
2.1 On k-Fibonacci Numbers
The k-generalized Fibonacci sequence and its terms have an extensive list of properties. Here, we only present some of them, namely those strictly necessary to address our problem. For more related data, we invite the reader to consult [1, 10], where one finds some background for this paper, as well as [2, 3, 5, 7, 9, 11, 13, 16, 17, 19].
Lemma 1
Let \(\alpha _{1}, \alpha _2, \ldots ,\alpha _{k}\) be all the zeroes of the characteristic polynomial \(\Psi _k(z) = z^k-z^{k-1}-\cdots -z-1\) for the sequence \(F^{(k)}\).
- (i):
-
\(\Psi _k\) is irreducible over \({\mathbb {Q}}[z]\) with one real zero outside the unit circle, named \(\alpha :=\alpha _1\) (sometimes denoted by \(\alpha (k)\) to emphasize its dependence on k), which satisfies \(2(1-2^{-k})<\alpha <2\).
- (ii):
-
The first \(k+1\) non-zero terms in \(F^{(k)}\) are powers of two, namely
$$\begin{aligned} F_{1}^{(k)}=1\quad \text {and}\quad F_{n}^{(k)}=2^{n-2} \quad \text {for all} \quad 2 \le n \le k+1. \end{aligned}$$Also, \(F_{k+2}^{(k)}=2^{k}-1\) and, moreover,
$$\begin{aligned} F_n^{(k)}<2^{n-2} \quad \text {for all}\quad n\ge k+2. \end{aligned}$$(5) - (iii):
-
Let \(f_k(z):= (z-1)/(2+(k+1)(z-2))\). For all \(n\ge 1\) and \(k\ge 2\),
$$\begin{aligned} F_n^{(k)} = \sum _{i=1}^{k}f_k(\alpha _{i}){\alpha _{i}}^{n-1} = f_k(\alpha ){\alpha }^{n-1}+e_k(n) \end{aligned}$$(6)with \(|e_k(n)|< 1/2\).
- (iv):
-
For all \(k\ge 2\) and \(i = 2,\ldots , k\)
$$\begin{aligned} f_k(\alpha )\in [1/2, 3/4]\qquad \textrm{and} \qquad |f_k(\alpha _i)|<1. \end{aligned}$$(7)Thus, \(f_k(\alpha )\) is not an algebraic integer, for any \(k \ge 2 \).
- (v):
-
For all \(k \ge 19\),
$$\begin{aligned} f_k(\alpha )> \dfrac{1}{2} + \dfrac{k-1}{2^{k+2}} \qquad \textrm{and} \qquad f_k(\alpha )\alpha > 1+\dfrac{k}{2^{k+2}}. \end{aligned}$$ - (vi):
-
For all \(n\ge 1\) and \(k\ge 2\),
$$\begin{aligned} \alpha ^{n-2}\le F_n^{(k)}\le \alpha ^{n-1}. \end{aligned}$$(8) - (vii):
-
The sequences \((F_{n}^{(k)})_{n\ge 1}, ~ (F_{n}^{(k)})_{k\ge 2}\) and \((\alpha (k))_{k\ge 2}\) are non decreasing.
- (viii):
-
For all \(n \ge 3\) and \(k \ge 3\), we have \(F_{n-1}^{(k)}/F_{n+1}^{(k)} \le 3/7\).
Since our equation involves the powers \((F_{r}^{(k)})^x\) for \(r=n\pm 1\), we need the following lemma.
Lemma 2
Let \(k\ge 2\), x, r positive integers. Then
Proof
By Binet’s formula (6) and inequalities (7), it follows that
Then, if we take
we get
where we have used the fact that (7) implies \(|e_k(r)|\le 1/2\le f_k(\alpha )\). \(\square \)
The following identity is due to Cooper and Howard (see [6]).
Lemma 3
For \(k\ge 2\) and \(n\ge k+2\), we have
where
with the classical convention that \({\displaystyle {\left( {\begin{array}{c}a\\ b\end{array}}\right) =0}}\) if either \(a<b\) or if one of a or b is negative and \(\lfloor x \rfloor \) denotes the greatest integer less than or equal to x.
The following lemmas are consequences of the previous one and will be essential to effectively solve the Diophantine equation (4).
Lemma 4
If \(k+2\le r < 2^{ck}\) for some \( c\in (0,1) \), then the following estimates hold:
- (i):
-
\( F_{r}^{(k)} = 2^{r-2}\left( 1-\frac{r-k}{2^{k+1}}+\frac{f(r,k)}{2^{2k+3}}+\zeta (r,k)\right) \), with \(|\zeta (r,k)|<\frac{4\, r^3}{2^{3k+3}}, \) where \(f(r,k):= \delta (r-2k+1)(r-2k-2)\) and
$$\begin{aligned} \delta := {\left\{ \begin{array}{ll} 0, &{}\quad \textrm{if} ~~ r \le 2k+2,\\ 1, &{}\quad \textrm{if} ~~r > 2k+2. \end{array}\right. } \end{aligned}$$Even more,
- (ii):
-
\({\displaystyle {F_{r}^{(k)} = 2^{r-2}\left( 1+\zeta '\right) }}\), with \({\displaystyle {|\zeta '|<\frac{2r}{2^{k}}<\frac{2}{2^{k(1-c)}}}}\).
- (iii):
-
\({\displaystyle {F_{r}^{(k)} = 2^{r-2}\left( 1-\dfrac{r-k}{2^{k+1}}+\zeta ^{''}\right) }}\), with \({\displaystyle {|\zeta ''|< \frac{4r^2}{2^{2k+2}} < \frac{1}{2^{2k(1-c)}}}}\).
Proof
By Cooper and Howard’s identity in Lemma 3, we can write
where \(C_{r,1} = r-k, ~C_{r,2} = (r-2k+1)(r-2k-2)/2\) and
Since \( r < 2^{ck} \) with \( c\in (0,1) \), then \( e^{r/2^{k+1}}<2 \). Thus, we get
which corresponds to (i). Further, since we have
we can conclude (ii) and (iii). \(\square \)
As a consequence of the previous lemma, we have the following result.
Lemma 5
For integers \(x\ge 1\), \(k\ge 2\), \(i\in \{-1,1\}\), \(n+i \ge k+2\) and \(\max \{n+i, x\} < 2^{ck}\) for some \( c\in (0,1/2) \), the estimate
holds with
Proof
By item (iii) of Lemma 4 with \(r=n+i\) and \(c\in (0,1/2)\), we have
Hence,
We now use the binomial theorem to analyze
We put \(A_i: = -(n+i-k)/2^{k+1}\), so \(|A_i| < (n+i)/2^{k+1}\). Since \(n+i<2^{ck}\), we get that
Further, let
Thus,
In the above, we used that
Furthermore, since the above number is smaller than \(\frac{1.5}{2^{(1-2c)k+1}}<\frac{3}{4}\) it follows that the sum of the geometrical progression is at most
In addition to this, we used for the last inequality that \(n\ge 3\), \(i\le 1\), so
\(\square \)
Finally, we present some relations between the variables in the Eq. (4).
Lemma 6
Let (n, m, x, k) be a solution of the Diophantine equation (4) with \(n, m \ge k+1, ~x\ge 1\) and \(k \ge 2\) and assume that
Then, one of the following holds:
- (i):
-
\(n-1\equiv -1, ~0, 1 \) or \(2 \pmod {k+1}\). Thus, \(m \equiv 1 ~ { or} ~ 2 \pmod {k+1}\).
- (ii):
-
Letting \(r_1\) be the residue of n by division with \(k+1\), with \(3 \le r_1 \le k-1\), we have \((r_1-2) x \le 2\log _2(m+k) + k-1\).
Proof
Let us take
By relation (3), we have the identity \(F_{n}^{(k)}=2F_{n-1}^{(k)}-F_{n-k-1}^{(k)}\) for all \(n\ge k+1\). Reducing modulo 2 we get that \( F_n^{(k)}\equiv F_{n-k-1}^{(k)}\pmod 2\), so \(F^{(k)}\) is periodic modulo 2 with period \(k+1\). Since \(F_1^{(k)}=F_2^{(k)}=1\) and \(F_j^{(k)}=2^{j-2}\equiv 0\pmod 2\) for \(j=3,\ldots ,k+1\) it follows that \(F_n^{(k)}\) is even except when \(n\equiv 1,2\pmod {k+1}\). So, assuming \(F_{n+1}^{(k)}\) and \(F_{n-1}^{(k)}\) are both even, we get that \(n-1\not \equiv -1, 0, 1, 2, \pmod {k+1}\)). We write \(n-1=(k+1)s_1+r_1\) with \(r_1\in \{3,\ldots ,k-1\}\). Then \(n-1+k=(k+1)(s_1+1)+r_1-1\) and \(\lfloor (n-1+k)/(k+1)\rfloor -1 =s_1\). We recall that by Cooper and Howard’s formula (Lemma 3),
Note that, the exponent of 2 in the last term above is \(n-(k+1)s_1-3 = r_1-2\). Thus, \(2^{r_1-2}\mid F_{n-1}^{(k)}\). Similarly, since \(n+1=(k+1)s_1+r_1+2\), we have that \((n+1)+k=(k+1)(s_1+1)+r_1+1\), so, we have that \(2^{r_1}\mid F_{n+1}^{(k)}\). Hence,
We now use Cooper and Howard’s formula
where \(s=\lfloor (m+k)/(k+1)\rfloor (\ge 2)\). Write \(m+k=s(k+1)+r\), where \(0\le r\le k\). Then the last term of the above sum has exponent of 2 equal to
Therefore, we have
so
with some integer M. We proceed to calculate an upper bound for \(\nu _2(C_{m,s-1})\). We have
Then,
where \(\mid \) denote the symbol of divisibility.
Thus, we have
In the first inequality we have used Kummer’s inequality (see [14]), which states that the 2-adic valuationFootnote 1 of \(\left( {\begin{array}{c}v\\ w\end{array}}\right) \), is equal to the number of carries in the sum between w and \(v - w\) in base 2, in particular
as the above sum contains at most \(\lfloor \log _2 v\rfloor \) nonzero terms and each nonzero term is equal to 1.
Since
we get that \(\nu _2(C_{m,s-1}) \le 2\log _2\left( m+k\right) .\) Thus, if \(k >2 \log _2\left( m+k\right) \), then
Since \(\nu _2(F_m^{(k)})\ge (r_1-2)x\) according to (9) and \(r\le k\), we get (ii). \(\square \)
3 Tools
3.1 Linear Forms in Logarithms
Since we use a Baker-type lower bound for a nonzero linear form in logarithms of algebraic numbers, in this section we present the necessary concepts and results.
Let \(\eta \) be an algebraic number of degree d over \({\mathbb {Q}}\) with minimal primitive polynomial over the integers \(m(z):= a_0 \prod _{i=1}^{d}(z-\eta _{i}) \in {\mathbb {Z}}[z], \) where the leading coefficient \(a_0\) is positive and the \(\eta _{i}\)’s are the conjugates of \(\eta \). The logarithmic height of \(\eta \) is given by
In particular, if \(\eta =p/q\) is a rational number with \(\gcd (p,q)=1\) and \(q>0\), we then have \(h(\eta )=\log \max \{|p|,q\}\).
For example, since \(\Psi _k(z)\) is the minimal primitive polynomial of \(\alpha \), we have \({\mathbb {Q}}(\alpha )= {\mathbb {Q}}(f_k(\alpha ))\). Besides, by (7), we have that \(|f_k(\alpha _{i})| \le 1\) for all \(i = 1,\ldots ,k\) and \(k \ge 2\). Thus,
See [5] for further details concerning the proof of (10).
The following properties of \(h(\cdot )\) will be used in the following sections:
Our main tool is a lower bound for a non-zero linear form in logarithms of algebraic numbers due to Matveev [15]:
Theorem 2
(Matveev’s theorem) Let \({\mathbb {K}}\) be a number field of degree D over \({\mathbb {Q}},\,\,\) \(\gamma _1, \ldots , \gamma _t\) be positive real numbers of \({\mathbb {K}}\), and \(b_1, \ldots , b_t\) rational integers. Put
Let \(A_i \ge \max \{Dh(\gamma _i), |\log \gamma _i|, 0.16\}\) for \(i = 1, \ldots , t.\) If \(\Lambda \not = 0\), then
3.2 Analytical Arguments
Note that, for \(\gamma _1,\ldots ,\gamma _t\) real algebraic numbers,
are such that \(\Lambda =e^{\Gamma }-1\). Therefore, it is a straight-forward exercise to show that \(|\Gamma |<(1-c)^{-1}|\Lambda |\), when \(|\Lambda | < c\), for all constant c in (0, 1). We use this argument several times without mentioning it.
On the other hand, in some specific parts of our work, we need the following analytic result, which correspond to Lemma 7 from [12].
Lemma 7
If \(r \ge 1\) and \(T>(4r^2)^r\), then
3.3 Reduction by Continued Fractions
The application of the results from the previous subsections give us some large bounds on the integers variables of the Diophantine equation (4). Since these bounds are very large, we use some results from the theory of continued fractions and geometry of numbers to reduce them. Here, we present these reduction techniques.
For the treatment of homogeneous linear forms in two integer variables we use a classical theorem of Legendre.
Lemma 8
Let M be a positive integer and \(P_1/Q_1,P_2/Q_2,\ldots \) the convergents of the continued fraction \([a_0,a_1,\ldots ]\) for \(\tau \). Let N be a positive integer such that \(M<Q_{N+1}\). If \(a_M:=\max \left\{ a_\ell :0\le \ell \le N+1\right\} \), then the inequality
holds for all pairs (u, v) of integers with \(0<u<M\).
Next, we present a slight variation of the result of Dujella and Pethő (Lemma 5a in [8]). For a real number X, we use
to denote the distance from X to its nearest integer.
Lemma 9
Let M and Q be positive integers such that \(Q>6M\), and \(A,B,\tau ,\mu \) be real numbers with \(A>0\) and \(B>1\). Let \(\varepsilon :=||\mu Q||-M||\tau Q||\). If \(\varepsilon >0\), then there is no solution to the inequality
in positive integers u, v and w with
In practical applications, Q is the denominator of a continued fraction convergent for \(\tau \).
4 Some Considerations
Recall that we are working with integers \(k\ge 2,\) \(n\ge 1,\) \(m \ge 2\) and \(x\ge 1\). First, we present some trivial solutions of Eq. (4).
Theorem 3
The trivial solutions (k, n, m, x) of Diophantine equation (4) are
and
Proof
When \(k=2\), according to Theorem 1, the Diophantine equation (2) only has the solutions \((n,m,x)=(n,n,1)\) or (n, 2n, 2), for all \(n\ge 1\) or (1, 2, x) for all \(x\ge 1\).
Now, let us consider some particular values of n:
-
If \(n=1\), then (4) corresponds to \(F_m^{(k)} = 1\), which has the solution \(m=2\) for all \(k\ge 2\) and \(x\ge 1\).
-
If \(n=2\), then (4) corresponds to \(F_m^{(k)} = 2^{x}-1\), which has only the solutions \((m,x)=(2,1)\) and \((k+2,k)\) for all \(k\ge 3\) by the results from [4].
-
If \(n=3\), then, when \(k\ge 3\), we have that Eq. (4) corresponds to \(F_m^{(k)}=2^{2x}-1,\) which has only the solutions \((k,m,x)=(x,2x+2,x)\) for all \(x\ge 2\), again by the results from [4].
Hence, from now on we may assume that \(k\ge 3 \) and \(n\ge 4\). Note that by inequalities (8), we have that
and
Thus, we can conclude that
So, if we take \(x=1\), the previous inequality implies that \(m = n\) or \(n+1\), which do not provide solutions for Eq. (4). \(\square \)
The previous arguments reduce our problem to find the solutions \((k,n,m,x)\) of the Diophantine equation
with \(k\ge 3,\) \(n\ge 4,\) \(m\ge 2\) and \(x\ge 2\).
4.1 Non-zero Linear Forms in Logarithms
In this section, we show that the linear forms in logarithms of algebraic numbers, to which we apply Theorem 2, are non-zero. Let us take
\({\mathbb {K}}={\mathbb {Q}}(\alpha )\) and \(D=[{\mathbb {K}}:{\mathbb {Q}}]=k\). With the previous notation, we prove the following result.
Lemma 10
\( \Lambda _{i}\ne 0 \) for all \( i \in \{1, 2, 3, 4, 5\}\).
Proof
For \( i = 1, 2, 3, 4\) and 5, if \(\Lambda _i =0\), then
and
All these cases imply that \(f_k (\alpha )\) would be an algebraic integer since \(\alpha \) is a unit in \({\mathcal {O}}_{\mathbb {K}}\), which contradicts (iv) from Lemma 1. \(\square \)
5 The Case \(n \le k\)
By (ii) in Lemma 1, the equation is
Since \(n\ge 4\) and
it is clear that \(F_m^{(k)}\) cannot be a power of 2. Thus, by (ii) of Lemma 1, \(m\ge k+2\). Using inequalities (5) and (8) in (12), we have
which gives
We next establish some additional relations between the variables in the Diophantine equation (12).
Lemma 11
Let \(4\le n\le k\). If (k, n, m, x) is a solution of (12) with \(x\ge 2\), then
and
Proof
The Binet formula (6) implies that
We apply Theorem 2 on inequality (16) with \(t:=3\), \({\mathbb {K}}:= {\mathbb {Q}}(\alpha )\), \(D:=k\), \(B:= m\) and
We have
We conclude that
as we wanted to prove.
Returning to Eq. (12), we also include \(2^{(n-3)x}\) into the above linear form in logarithms, getting
We apply again Theorem 2 with \(t:=4\), the same \({{\mathbb {K}}}\), D and B, and
The conclusion of Theorem 2 together with the above inequality (17) yield
where we have used inequality (14). Hence, by Lemma 7 with \((y, r):=(m,2)\) and \(T:= 2.2 \times 10^{25} k^9 (\log k)^4\), we have
which corresponds to inequality (15). \(\square \)
5.1 When \(k \le 600\)
We start by looking for an upper bound on x. So, we take
Since \(x \ge 2\), by inequality (16) we have that
So, if we set
we then get
For each \(k\in [4,600]\), we consider \(M:= 3.4 \times 10^{29} k^9 (\log k)^6\), which is an upper bound to \(m-1\), according to inequality (15). A computer search shows that
Hence, by Lemma 9, we can conclude that \(x \le 595\).
Now that we have bounded x, let us fix it in [2, 595] and consider
Using inequality (17) in its logarithmic form, we obtain a similar inequality to (18), namely
where we put
Therefore, for \(k \in [4,600]\) and \(x \in [2,595]\), we apply Lemma 9 to inequality (19) using \(M:= 3.4 \times 10^{29} k^9 (\log k)^6\). With computational support, we obtain
Thus, by Lemma 9, we have that \(m \le 1520\).
In summary, for \(n\le k\), the integer solutions (k, n, m, x) of (12) must satisfy \(k \in [4, 600]\), \(x \in [2, 595]\), \(m \in [k+2, 1520]\) and, by (13)
However, a computational search in the above range for solutions of the Diophantine equation (12) gave us only those that we indicated in the statement of the Main Theorem (the first family for \(\ell \in [2,8]\) and the second family for \(\ell \in [3,9]\) and odd).
5.2 When \(k > 600\)
Given that
by equality (12) and Lemma 4, (ii) with \(c=1/2\) for \(r=m\), we get
Thus, dividing by \(2^{m-2}\), we get
By (13), we have \((n-1)x - (m-2) \le 0\). If \((n-1)x - (m-2) \le -1\), then in the above inequality the left-hand side is larger than 1/2, which contradicts the fact that \(k > 600\). So, \(m-2 =(n-1)x\).
Now, by equality (12) and Lemma 4, (iii) with \(c=1/2\) for \(r=m\), we have
Assume that the left-hand side is non-zero. Then
If \(2x\le k+1\), the left-hand side above is \(\ge 1/2^{k+1}\). If \(2x\ge k+2\), then
Hence, the inequality
holds in all cases. Thus, we get
which leads to
However, this implies \(k<400\), a contradiction. In summary, we have
Now, if \(m > 2k+2\) then by Lemma 4, (i) with \(c=1/2\) for \(r=m\), we obtain
which lead us to a contradiction on \(k>600\). Hence, \(m \le 2k+2\). Thus, by Lemma 3, our equation becomes
We get \(2^{m-k-3}(m-k)=2^{(n-3)x}\). But also \(m-k=2^{k+1-2x}\). Since \(m\le 2k+2\), we have that
so
However,
showing that \(n\le 5\). Hence, \(n\in \{4,5\}\). Further, let \(\ell :=k+1-2x\). Then \(m=k+2^{\ell }\) so \(\ell \ge 1\). Next, we have
Finally, since also \(m-k=2^{(n-3)x-(m-k-3)}\), we get that
This gives
Since \(m\le 2k+2\), we have \(2^{\ell }=m-k\le k+2\). When \(n=4\), we get \(2^{\ell }\le k+2=2^{\ell +1}+3\ell -5\), and the above inequality holds when \(\ell \ge 1\). When \(n=5\), we get \(2^{\ell }\le k+2=2^{\ell }+2\ell -2\), and again the above inequality holds for all \(\ell \ge 1\). So, we have
which give us the parametric families of solutions
indicated in the statement of the Main Theorem. In the first case \(\ell \ge 9\) (to insure \(k>600\)), while in the second case \(\ell \ge 10\) (again, to insure \(k>600\)) and \(\ell \) must be odd to insure that x is an integer.
6 The Case \(n>k\)
Here, as before, we need to establish some relations between the variables in our equation. The following result gives us an inequality for x in terms of k and n.
Lemma 12
Let (k, n, m, x) be an integral solution of (4) with \(n>k\ge 3\) and \(x\ge 2\), then
Proof
Equation (4) can be rewritten as
Dividing both sides of equation (21) by \((F_{n+1}^{(k)})^{x}\) and taking absolute values, we get
where we used Lemma 1, (viii).
We apply Theorem 2 with the parameters \(t:=3\),
and \({\mathbb {K}}, D, B\) as for \(\Lambda _1\).
Now, Theorem 2 combined with inequality (22) yields
where we used the fact that \(m < nx+2\), which follows from (11).
We next extract from (23) an upper bound for x depending on n and k. Multiplying by n both sides of the inequality (23) we obtain
Taking \(y:= nx\) and \(T:=1.2 \times 10^{14} n^2 k^4(\log k)^2\), by Lemma 7 and the fact that \(n > k\),
It remains to divide by n both sides of the previous inequality. \(\square \)
We now work under the assumption that \(n>700\) to find an upper bound for n, m and x in terms of k only.
Lemma 13
Let (k, n, m, x) be an integral solution of (4) with \(n >\max \{k,700\}\). Then the following inequalities
hold.
Proof
Given that \(n>k\), from (20), we have that
Thus, for \(i = \pm 1\),
Then, by Lemma 2, we can write
We now use (26) to rewrite the Eq. (4) as
Dividing both sides of the previous inequality by \(f_k(\alpha )^x \alpha ^{nx}\), we conclude that
where we have used the fact that \(1-\alpha ^{-2x} < 1/2\), \(f_k(\alpha ) \alpha ^{n} > \alpha ^{n-2}\) and \((n-2)x+1 \ge 0.8n\) for all \(n>700\), \(k\ge 3\) and \(x\ge 2\). Hence,
with \(\kappa := \min \{0.8n, 2x\}\).
We apply again Theorem 2 with the parameters \(t:=2\),
and \({\mathbb {K}}\) and D as before. Moreover, we can take \(B:=x\), since \( | m-1-nx | \le x\) by inequality (11).
The conclusion of Theorem 2 and the inequality (28) yield, after taking logarithms, the following upper bound for \(\kappa \):
If \(\kappa = 0.8n\), then from (29),
and using the inequality (25), we obtain that
since \(n > 700\). Hence, we apply Lemma 7 with \(T:=1.5\times 10^{1}k^3 (\log k)^2\) and \((y,r):= (n,1)\) to obtain an upper bound on n depending only on k. Further, inserting the resulting bound on n in terms of k in (20) and using the inequality (11), we have that
If \(\kappa = 2x\), then by (29) and Lemma 7 with \(T:= 4.5\times 10^{9}k^3 (\log k)^2\) and \((y,r):= (x,1)\), we get
Furthermore, given that \(x \le 0.45n\) we have by Lemma 2, that for \(i = \pm 1\) \(x/\alpha ^{n+i-1}< 0.45n/\alpha ^{n-2} < 1/\alpha ^{0.98n}\) since \(n > 700\). Thus,
We return to the inequality (27) and dividing both sides by \(f_k(\alpha )\alpha ^{m-1}\), we obtain
where we used the inequalities:
valid for \(n>700,\) \(x \ge 2\) and \(k \ge 3\). In conclusion, we have shown that
Here, we apply again Theorem 2 with the parameters \(t:=3\),
and \({\mathbb {K}}\) and D as before. Moreover, again we can take \(B:=x\). Combining the conclusion of Theorem 2 with inequality (32), we get
By (31), we have \(x<2.2\times 10^{11}k^3(\log k)^3\), therefore
since \(k\ge 3\). Hence, returning to inequality (33) and taking into account that \(m<nx+2\), we have in summary
Comparing inequalities (30) and (34), we get that
as we wanted to show. \(\square \)
The inequalities in Lemma 13 were obtained under the assumptions that \(n>700\). However, when \( n \le 700 \) the inequalities (11) and (25) yield smaller upper bounds for x and m in terms of k.
6.1 When \(k \le 700\)
Here, we prove the following result where we establish some computational ranges to search for the integral solutions of equation (4).
Lemma 14
Let (k, n, m, x) be an integral solution of Diophantine equation (4) with \(n>k\), \(k\le 700 \) and \(x\ge 2\). Then \(m\in [M_0,M_1]\) with
Furthermore, if \(n>700\), then \(n\le 1810\) and \(x\le 1260\), otherwise \(x\le 1150\).
Proof
Note that the range for m is given by inequality (13). Now, let us start assuming \(n>700\), which allows us to use the inequalities of Lemma 13 in order to obtain upper bounds on n, x and m. Taking into account the inequality (28), we take
Then, using the analytic argument of Sect. 3.2, we get
Dividing the above inequality by \((x-1)\log \alpha \), we obtain
Now, we need to distinguish two cases:
-
Case \(m=1+nx\). Here, the inequality (36) correspond to
$$\begin{aligned} \left| \dfrac{\log ( f_k(\alpha )^{-1})}{\log \alpha } \right| < \dfrac{10}{\alpha ^{\kappa }(x-1)}. \end{aligned}$$A quick computational search shows that the left-hand side of the previous inequality is greater than 0.7 for all \(k \in [3,700]\). Thus, since \(x\ge 2\), we get
$$\begin{aligned} 0.7<\frac{10}{\alpha ^{\kappa }}. \end{aligned}$$(37)Now, since \(\kappa =\min \{0.8n,2x\}\), if we assume that \(\kappa = 0.8n,\) then, by inequality (37), we get \(n\le 6\), a contradiction with our assumption about n. Therefore, we have \(\kappa = 2x\), which together with inequality (37) implies \(x=2\). Now, by inequality (32) with \(m=1+nx\) and using \(k\in [3,700]\), we get
$$\begin{aligned} 0.5<|\log (f_k(\alpha )) + \log \left( 1-\alpha ^{-4}\right) |<4/\alpha ^{0.38n}, \end{aligned}$$where the left-hand side was found using computations. However, this inequality implies \(n\le 11\), again a contradiction.
-
Case \(m\ne 1+nx\): Here we apply Lemma 8 to inequality (36) using \(k \in [3, 700]\). To do it, let us take \(\tau _k:=\log (f_k(\alpha )^{-1})/\log \alpha \). So, by inequality (24), we look for the integer \(t_k\) such that
$$\begin{aligned} Q_{t_k}^{(k)}> 1.8\times 10^{30}k^{7}(\log k)^{6} > x-1, \end{aligned}$$and take \(a_M:= \max \{a_i^{(k)} \,: \, 0\le i \le t_k,~3\le k \le 700\}.\) Then, by Lemma 8, we have that
$$\begin{aligned} \begin{aligned} \left| \tau _k - \dfrac{nx-(m-1)}{x-1}\right| > \dfrac{1}{(a_M+2)(x-1)^2}. \end{aligned} \end{aligned}$$(38)Hence, combining the inequalities (36) and (38), and taking into account that \(a_M+2 < 1.1\times 10^{208}\) (confirmed by computations), we obtain
$$\begin{aligned} \alpha ^{\kappa }<1.1\times 10^{209}x. \end{aligned}$$If \(\kappa = 0.8n,\) since \(n>700\ge k\), by inequality (25) we have
$$\begin{aligned} \alpha ^{0.8n} < 8.6\times 10^{224}n^5(\log n)^3, \end{aligned}$$which implies
$$\begin{aligned} n \le 1010. \end{aligned}$$(39)Thus, let us consider
$$\begin{aligned} \Gamma _3: = \log f_k(\alpha )+(m-1)\log \alpha -x \log F_{n+1}^{(k)}, \end{aligned}$$(40)with \(k \in [3,700]\) and \(n \in [701,1010]\). Note that, by inequality (22), we have
$$\begin{aligned} |\Gamma _3|< \dfrac{4}{2.3^x}. \end{aligned}$$Dividing both sides by \(\log F_{n+1}^{(k)}\), we get
$$\begin{aligned} \left| (m-1)\left( \frac{\log \alpha }{\log F_{n+1}^{(k)}}\right) -x + \frac{\log f_k(\alpha )}{\log F_{n+1}^{(k)}}\right| < \frac{3}{2.3^x} \end{aligned}$$(41)where we used that \(\log F_{n+1}^{(k)} \ge \log F_5^{(3)} =\log 7\). In order to apply Lemma 9, we take
$$\begin{aligned} \gamma _{k,n}:=\dfrac{\log \alpha }{\log F_{n+1}^{(k)}},\quad \quad \mu _{k,n}:= \dfrac{\log f_k(\alpha )}{\log F_{n+1}^{(k)}} \quad \quad A:= 3 \quad \text { and } \quad B:= 2.3, \end{aligned}$$for \(k\in [3,700]\) and \(n\in [701,1010]\) with \(M:= 3.7\times 10^{33} k^{7}(\log k)^{6}\), thanks to inequalities (13) and (24). We obtain
$$\begin{aligned} \max _{k\in [3,700], ~ n \in [701,1010]}\left\{ \lfloor \log \left( A Q^{(k, n)}/\varepsilon _{k, n}\right) /\log B\rfloor \right\} \le 1260, \end{aligned}$$which, by Lemma 9, implies
$$\begin{aligned} x \le 1260. \end{aligned}$$(42)Now, if \(\kappa = 2x\), then \( \alpha ^{2x} < 1.1\times 10^{209}x,\) which implies
$$\begin{aligned} x \le 360. \end{aligned}$$(43)So, we go back to inequality (32) and take
$$\begin{aligned} \Gamma _5:=(x-1)\log (f_k(\alpha )) + (nx-(m-1))\log \alpha + \log \left( 1-\alpha ^{-2x}\right) . \end{aligned}$$Since \(|\Gamma _5| < 4/\alpha ^{0.38n}\), dividing by \(\log \alpha \), we obtain
$$\begin{aligned} \begin{aligned} \left| (x-1)\dfrac{\log ( f_k(\alpha ))}{\log \alpha } + \frac{\log \left( 1-\alpha ^{-2x}\right) }{\log \alpha } - (m-1-nx) \right|&< \dfrac{7}{\alpha ^{0.38n}}. \end{aligned} \end{aligned}$$We take
$$\begin{aligned} \tau _{k, x}:=(x-1)\dfrac{\log ( f_k(\alpha ))}{\log \alpha } + \frac{\log \left( 1-\alpha ^{-2x}\right) }{\log \alpha }. \end{aligned}$$We have that
$$\begin{aligned} \min _{k\in [3,700],~x\in [2,360]}\left\| \tau _{k, x} \right\|< |\tau _{k, x} - (m-1-nx)| < \dfrac{7}{\alpha ^{0.38n}}. \end{aligned}$$Computationally, we found that the minimum on the left-hand of the previous inequality is at least \(7\times 10^{-207}\). Therefore, we get
$$\begin{aligned} n \le 1810. \end{aligned}$$(44)
To sum up, if \(n>700\), then by inequalities (39), (42), (43) and (44), all the positive integral solutions (k, n, m, x) of Eq. (4) satisfy \(n\le 1810\) and \(x \le 1260\).
Finally, let us consider \(n\le 700\). Since in this section we are working with \(k\le 700\) and \(n>k\), it is clear that \(k\le 699\). So, let us use \(\Gamma _3\) as we defined it in (40) to proceed as we did with (41). This time we take \(k \in [3,699]\) and \(n \in [k+1,700]\) with \(M:= 2.6\times 10^{33} k^{7}(\log k)^{6}\), which is given by inequalities (13) and (24). We get
which, by Lemma 9, implies \(x \le 1150\). \(\square \)
In conclusion, our problem is now reduced to a computational search for integral solutions of the Diophantine equation (4) in the ranges indicated by Lemma 14; i.e., in the ranges \(k\in [3,700]\), \(m\in [M_0,M_1]\) (with the limits given by (35)),
or
A computer search using Lemma 6 allow us to conclude that there are no integral solutions for Eq. (4) in these ranges.
6.2 When \(k>700\)
From now on, we assume that \(k>700\). We show that there are no such solutions. We have, from (24), that
We recall that \(n \ge k+1\), so \(m>k+1\) according to (11). By item (iii) of Lemma 4 (for m with \(c:= 0.48\)) and Lemma 5 (for \(n+i\) with \(i\in \{\pm 1\}\) and \(c:=0.24\)), we conclude that
where \(\delta _{1} = 1\) for all \(n \ge k+1\) and
Now, let us take \(M=\max \{(n-1)x,m-2\}\) and \(N=\min \{(n-1)x,m-2\}\), so, we get
In the above, we used that \(x(n+i-k)< x(n-1)<m<2^{0.48k}\) for \(i\in \{\pm 1\}\), where the second inequality follows from (11), in addition to \(k > 700\). After dividing by \(2^M\) becomes
If \(M>N\), then the left-hand side is at least 1/4, so \(2^{0.52k}<16\), a contradiction since \(k>700\). Thus, \(M=N\), or, equivalently, \((n-1)x=m-2\). We next get
or
But
Thus,
and we get
or \(2^{0.04k}<16\), a contradiction with \(k>700\). Thus, we showed that our Diophantine equation has no solution in the range \(n>k>700\).
Data Availability
The data mentioned in the paper (computer codes) are available from the first author.
Notes
\(\nu _2(m)\) is the exponent of 2 in the factorizatión of \(m \ne 0\) in primes factors.
References
Bensella, H., Patel, B.K., Behloul, D.: On the exponential Diophantine equation \((F_{m+1}^{(k)})^x - (F_{m-1}^{(k)})^x = F_n^{(k)}\) (2022). arxiv:2205.13168
Bravo, J., Luca, F.: Powers of two in generalized Fibonacci sequences. Rev. Colomb. Mat. 46, 67–79 (2012)
Bravo, J., Luca, F.: On a conjecture about repdigits in \(k\)-generalized Fibonacci sequences. Publ. Math. Debr. 82, 623–639 (2013)
Bravo, J.J., Gómez, C.A.: Mersenne \(k\)-Fibonacci numbers. Glas. Mat. 51(71), 307–319 (2016)
Bravo, J.J., Gómez, C.A., Luca, F.: Powers of two as sums of two \(k\)-Fibonacci numbers. Miskolc Math. Notes 17(1), 85–100 (2016)
Cooper, C., Howard, F.T.: Some identities for \(r\)-Fibonacci numbers. Fibonacci Q. 49, 231–243 (2011)
Dresden, G.P., Du, Z.: A simplified Binet formula for \(k\)-generalized Fibonacci numbers. J. Integer Seq. 17, Article 14.4.7 (2014)
Dujella, A., Pethő, A.: A generalization of a theorem of Baker and Davenport. Q. J. Math. Oxf. Ser. 49(195), 291–306 (1998)
Everest, G., Van der Poorten, A., Shparlinski, I., Ward, T.: Recurrence Sequences, Mathematical Surveys and Monographs, vol. 104. American Mathematical Society, Providence (2003)
Gómez, C.A., Luca, F.: An exponential Diophantine equation related to the sum of powers of two consecutive \(k\)-generalized Fibonacci numbers. Colloq. Math. 137(2), 171–188 (2014)
Gómez, C.A., Luca, F.: Multiplicative independence in \(k\)-generalized Fibonacci sequences. Lith. Math. J. 56(49), 503–517 (2016)
Guzmán, S., Luca, F.: Linear combinations on factorials and \(S\)-units in a binary recurrence sequence. Ann. Math. Québec 38(2), 169–188 (2014)
Hua, L. K., Wang, Y. : Applications of Number Theory to Numerical Analysis, Translated from Chinese. Springer, Berlin; Kexue Chubanshe (Science Press), Beijing (1981)
Kummer, E.E.: Über die Ergänzungssätze zu den allgemeinen Reciprocitätsgesetzen. J. Reine Angew. Math. 1852(44), 93–164 (1851). https://doi.org/10.1515/crll.1852.44.93
Matveev, E.M.: An explicit lower bound for a homogeneous rational linear form in the logarithms of algebraic numbers. Izv. Math. 64, 1217–1269 (2000)
Miles, E.P., Jr.: Generalized Fibonacci numbers and associated matrices. Am. Math. Mon. 67, 745–752 (1960)
Miller, M.D.: Mathematical notes: on generalized Fibonacci numbers. Am. Math. Mon. 78, 1108–1109 (1971)
Patel, B.K., Chaves, A.P.: On the exponential Diophantine equation \(F_{n+1}^x-F_{n-1}^x=F_m\), Mediterr. J. Math. 18(5), Paper No, 187, 11 pp (2021)
Wolfram, D.A.: Solving generalized Fibonacci recurrences. Fibonacci Q. 36, 129–145 (1998)
Acknowledgements
The authors thank the anonymous referee for useful comments. F. L. worked on this paper during extended visits at the Max Planck Institute for Software Systems in Saarbrücken, Germany in 2022 and the Stellenbosch Institute for Advanced Studies in Stellenbosch, South Africa in 2023. This author thanks these institutions for their hospitality and support. All authors thank the Department of Mathematics at the Universidad del Valle for the Cluster time used to perform calculations, and especially the Computer Center Jurgen Tischer for their advice on parallelisation of the algorithm used.
Funding
Open Access funding provided by Colombia Consortium. C. A. Gomez was supported in part by Project 71327 (Universidad del Valle). J. C. Gomez was supported in part by Universidad Santiago de Cali. F. L. was partially supported by project 2022-064-NUM GANDA from CoE-MaSS at the University of the Witwatersrand.
Author information
Authors and Affiliations
Contributions
All authors contributed to this research. They all read and approved the final version.
Corresponding author
Ethics declarations
Conflict of Interest
The authors have no relevant financial or non-financial interests to disclose.
Additional information
To the memory of Fred Howard.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Gómez, C.A., Gómez, J.C. & Luca, F. The Complete Solution of the Diophantine Equation \(\left( F_{n+1}^{(k)}\right) ^x - \left( F_{n-1}^{(k)}\right) ^x = F_{m}^{(k)}\). Mediterr. J. Math. 21, 13 (2024). https://doi.org/10.1007/s00009-023-02529-5
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00009-023-02529-5
Keywords
- k-generalized Fibonacci numbers
- lower bounds for nonzero linear forms in logarithms of algebraic numbers
- effective solution for exponential Diophantine equation
- method of reduction by continued fractions