Abstract
Let \( (P_n)_{n\ge 0}\) be the sequence of Perrin numbers defined by ternary relation \( P_0=3 \), \( P_1=0 \), \( P_2=2 \), and \( P_{n+3}=P_{n+1}+P_n \) for all \( n\ge 0 \). In this paper, we use Baker’s theory for nonzero linear forms in logarithms of algebraic numbers and the reduction procedure involving the theory of continued fractions, to explicitly determine all Perrin numbers that are concatenations of two repeated digit numbers.
Similar content being viewed by others
1 Introduction
Let \((P_n)_{n \ge 0}\) be the sequence of Perrin numbers, given by the ternary recurrence relation
The first few terms of this sequence are
A repdigit (in base 10) is a non-negative integer N that has only one distinct digit. That is, the decimal expansion of N takes the form
for some non-negative integers d and \(\ell \) with \(1 \le d \le 9\) and \(\ell \ge 1\). This paper is an addition to the growing literature around the study of Diophantine properties of certain linear recurrence sequences. More specifically, our paper focuses on a Diophantine equation involving the Perrin numbers and repdigits. This is a variation on a theme on the analogous problem for the Padovan numbers, a program developed in [5, 7].
In [7], the authors found all repdigits that can be written as a sum of two Padovan numbers. This result was later extended by the third author to repdigits that are a sum of three Padovan numbers in [4]. In another direction, in [5], Ddamulira considered all Padovan numbers that can be written as a concatenation of two repdigits and showed that the largest such number is \(Pad(21) = 200\). More formally, it was shown that if Pad(n) is a solution of the Diophantine equation \(Pad(n) = \overline{\underbrace{d_1\cdots d_1}_{\ell \text { times}} \underbrace{d_2\cdots d_2}_{m\text { times}}}\), then
The Padovan numbers and Perrin numbers share many similar properties. In particular, they have the same recurrence relation, the difference being that the Padovan numbers are initialized via \(Pad(0) = 0\) and \(Pad(1) = Pad(2) = 1\). This means that the two sequences also have the same characteristic equation.
Despite the similarities, the two sequences also have some stark differences. For instance, the Perrin numbers satisfy the remarkable divisibility property that if n is prime, then n divides \(P_n\). One can easily confirm that this does not hold for the Padovan numbers.
Inspired by the second author’s result in [5], we study and completely solve the Diophantine equation
where \(d_1 , d_2 \in \{0, 1, 2, 3,\ldots , 9 \}\), \( d_1>0 \), \(\ell , m \ge 1\), and \(n \ge 0\).
Our main result is the following.
Theorem 1.1
The only Perrin numbers which are concatenations of two repdigits are
2 Preliminary results
In this section, we collect some facts about Perrin numbers and other preliminary lemmas that are crucial to our main argument.
2.1 Some properties of the Perrin numbers
Recall that the characteristic equation of the Perrin sequence is given by \(\phi (x) := x^3 - x - 1=0\), with zeros \(\alpha \), \(\beta \) and \(\gamma = {\overline{\beta }}\) given by
where
For all \(n \ge 0\), Binet’s formula for the Perrin sequence tells us that the nth Perrin number is given by
Numerically, the following estimates hold for the quantities \(\{\alpha , \beta , \gamma \}\):
It follows that the complex conjugate roots \(\beta \) and \(\gamma \) only have a minor contribution to the right-hand side of Eq. (2.1). More specifically, let
The following estimate also holds:
Lemma 2.1
Let \(n \ge 2\) be a positive integer. Then
Lemma 2.1 follows from a simple inductive argument, and the fact that \(\alpha ^3 = \alpha + 1\), from the characteristic polynomial \(\phi \).
Let \({\mathbb {K}} := {\mathbb {Q}}(\alpha , \beta )\) be the splitting field of the polynomial \(\phi \) over \({\mathbb {Q}}\). Then, \([{\mathbb {K}}: {\mathbb {Q}}] = 6\) and \([{\mathbb {Q}}(\alpha ): {\mathbb {Q}}] = 3\). We note that the Galois group of \({\mathbb {K}}/{\mathbb {Q}}\) is given by
We therefore identify the automorphisms of \({\mathcal {G}}\) with the permutation group of the zeroes of \(\phi \). We highlight the permutation \((\alpha \beta )\), corresponding to the automorphism \(\sigma : \alpha \mapsto \beta , \beta \mapsto \alpha , \gamma \mapsto \gamma \), which we use later to obtain a contradiction on the size of the absolute value of a certain bound.
2.2 Linear forms in logarithms
Our approach follows the standard procedure of obtaining bounds for certain linear forms in (nonzero) logarithms. The upper bounds are obtained via a manipulation of the associated Binet’s formula for the given sequence. For the lower bounds, we need the celebrated Baker’s theorem on linear forms in logarithms. Before stating the result, we need the definition of the (logarithmic) Weil height of an algebraic number.
Let \(\eta \) be an algebraic number of degree d with minimal polynomial
where the leading coefficient \(a_0\) is positive and the \(\eta _j\)’s are the conjugates of \(\eta \). The logarithmic height of \(\eta \) is given by
Note that, if \(\eta = \frac{p}{q} \in {\mathbb {Q}}\) is a reduced rational number with \(q > 0\), then the above definition reduces to
We list some well-known properties of the height function below, which we shall subsequently use without reference
We quote the version of Baker’s theorem proved by Bugeaud et al. [1, Theorem 9.4].
Theorem 2.2
[1] Let \(\eta _1, \ldots , \eta _t\) be positive real algebraic numbers in a real algebraic number field \({\mathbb {K}} \subset {\mathbb {R}}\) of degree D. Let \(b_1, \ldots , b_t\) be nonzero integers, such that
Then
where
and
2.3 Reduction procedure
The bounds on the variables obtained via Baker’s theorem are usually too large for any computational purposes. To get further refinements, we use the Baker–Davenport reduction procedure. The variant we apply here is the one due to Dujella and Pethő [6, Lemma 5a]. For a real number r, we denote by \(\parallel r \parallel \) the quantity \(\min \{|r - n| : n \in {\mathbb {Z}} \}\), the distance from r to the nearest integer.
Lemma 2.3
[6] Let \(\kappa \ne 0, A, B\) and \(\mu \) be real numbers, such that \(A > 0\) and \(B > 1\). Let \(M > 1\) be a positive integer and suppose that \(\frac{p}{q}\) is a convergent of the continued fraction expansion of \(\kappa \) with \(q > 6M\). Let
If \(\varepsilon > 0\), then there is no solution of the inequality
in positive integers m, n, k with
Lemma 2.3 cannot be applied when \(\mu =0\) (since then \(\varepsilon <0\)). In this case, we use the following criterion due to Legendre, a well-known result from the theory of Diophantine approximation. For further details, we refer the reader to the books of Cohen [2, 3].
Lemma 2.4
[2, 3] Let \(\kappa \) be real number and x, y integers, such that
Then, \(x/y=p_k/q_k\) is a convergent of \(\kappa \). Furthermore, let M and N be a non-negative integers, such that \( q_N> M \). Then, putting \( a(M):=\max \{a_{i}: i=0, 1, 2, \ldots , N\} \), the inequality
holds for all pairs (x, y) of positive integers with \( 0<y<M \).
We will also need the following lemma by Gúzman Sánchez and Luca [8, Lemma 7]:
Lemma 2.5
[8] Let \(r \ge 1\) and \(H > 0\) be such that \(H > (4r^2)^r\) and \(H > L/(\log L)^r\). Then
3 Proof of the main result
3.1 The low range
We used a computer program in Mathematica to check all the solutions of the Diophantine Eq. (1.1) for the parameters \(d_1, d_2 \in \{0, 1,2,3, \ldots , 9 \}\), \( d_1>0 \), \(1 \le \ell , m \), and \(1 \le n \le 500\). We only found the solutions listed in Theorem 1.1. Henceforth, we assume \(n > 500\).
3.2 The initial bound on n
We note that Eq. (1.1) can be rewritten as
The next lemma relates the sizes of n and \(\ell + m\).
Lemma 3.1
All solutions of (3.1) satisfy
Proof
Recall that \(\alpha ^{n-2} \le P_n \le \alpha ^{n+1}\). We note that
Taking the logarithm on both sides, we get
Hence, \(n \log \alpha < (\ell + m) \log 10 + 1\). The lower bound follows via the same technique, upon noting that \(10^{\ell + m -1} < P_n \le \alpha ^{n+1}\). \(\square \)
We proceed to examine (3.1) in two different steps as follows.
Step 1. From Eqs. (2.1) and (3.1), we have
Hence
Thus, we have
where we used the fact that \(n > 500\). Dividing both sides by \(d_1 \times 10^{\ell + m}\), we get
We let
We shall proceed to compare this upper bound on \(|\Gamma _1|\) with the lower bound we deduce from Theorem 2.2. Note that \(\Gamma _1 \ne 0\), since this would imply that \( \alpha ^n = \frac{10^{\ell + m} \times d_1}{9}\). If this is the case, then applying the automorphism \(\sigma \) on both sides of the preceding equation and taking absolute values, we have that
which is false. We thus have \(\Gamma _1 \ne 0\).
With a view towards applying Theorem 2.2, we define the following parameters:
Note that, by Lemma 3.1, we have that \(\ell + m < n\). Thus, we take \(B = n\). We note that \({\mathbb {K}}: = {\mathbb {Q}}(\eta _1, \eta _2, \eta _3) = {\mathbb {Q}}(\alpha )\). Hence, \(D: = [{\mathbb {K}}: {\mathbb {Q}}] = 3\).
We note that
We also have that \(h(\eta _2) = h(\alpha ) = \frac{\log \alpha }{3}\) and \(h(\eta _3) = \log 10\). Hence, we let
Thus, we deduce via Theorem 2.2 that
Comparing the last inequality obtained above with (3.2), we get
Hence
Step 2. We rewrite Eq. (3.1) as
That is
Hence
Dividing throughout by \(9 \alpha ^n\), we have
We put
As before, we have that \(\Gamma _2 \ne 0\), because this would imply that
which in turn implies that
which is false. In preparation towards applying Theorem 2.2, we define the following parameters:
To determine what \(A_1\) will be, we need to find the maximum of the quantities \(h(\eta _1)\) and \(|\log \eta _1|\).
We note that
where, in the last inequality above, we used (3.4). On the other hand, we also have
where, in the second last inequality, we used Eq. (3.4). We note that \(D \cdot h(\eta _1) > |\log \eta _1|\).
Thus, we put \(A_1: = 2.46 \times 10^{13} (1 + \log n)\). We take \(A_2: = \log \alpha \) and \(A_3: = 3 \log 10\), as defined in Step 1. Similarly, we take \(B: = n\).
Theorem 2.2 then tells us that
Comparing the last inequality with (3.5), we obtain
Thus, we can conclude that
With the notation of Lemma 2.5, we let \(r: = 2\), \(L: = n\) and \(H: = 2.6 \times 10^{27}\) and notice that these data meet the conditions of the lemma. Applying the lemma, we obtain
After a simplification, we obtain the bound
Lemma 3.1 then implies that
The following lemma summarizes what we have proved so far:
Lemma 3.2
All solutions to the Diophantine Eq. (1.1) satisfy
3.3 The reduction procedure
We note that the bounds from Lemma 3.2 are too large for computational purposes. However, with the help of Lemma 2.3, they can be considerably sharpened. The rest of this section is dedicated towards this goal. We proceed as in [5].
Using Eq. (3.3), we define the quantity \(\Lambda _1\) as
Equation (3.2) can thus be rewritten as
If \(\ell \ge 2\), then the above inequality is bounded above by \(\frac{1}{2}\). Recall that if x and y are real numbers, such that \(|e^x - 1|< y\), then \(x < 2y\). We therefore conclude that \(|\Lambda _1|< \frac{92}{10^\ell }\). Equivalently
Dividing throughout by \(\log \alpha \), we get
Towards applying Lemma 2.3, we define the following quantities:
The continued fraction expansion of \(\tau \) is given by
We take \(M: = 5.5 \times 10^{30}\), which, by Lemma 3.2, is an upper bound for \(\ell + m\). A computer assisted computation of the convergents of \(\tau \) returns the convergent
as the first one for which the denominator \(q = q_{70} > 3.3 \times 10^{31} = 6M\). Maintaining the notation of Lemma 2.3, the smallest (positive) value of \(\epsilon \), corresponding to \(d_1 =5\) is chosen as \(\epsilon = 0.154964 < \Vert \mu q \Vert - M \Vert \tau q \Vert \). We deduce that
For the case \( d_1=9 \), we have that \( \mu (d_1)=0 \). In this case, we apply Lemma 2.4. The inequality (3.7) can be rewritten as
because \( \ell +m<5.5\times 10^{30}:=M \). It follows from Lemma 2.4 that \( \frac{n}{\ell +m} \) is a convergent of \( \kappa :=\frac{\log 10}{\log \alpha } \). So \( \frac{n}{\ell +m} \) is of the form \( p_k/q_k \) for some \( k=0, 1, 2, \ldots , 70 \). Thus
Since \( a(M)=\max \{a_{k}: k=0, 1, 2, \ldots , 70\}=49 \), we get that
Thus, \( \ell \le 34 \) in both cases. In the case \( \ell <2 \), we have that \( \ell<2<35 \). Thus, \( \ell \le 34 \) holds in all cases.
Proceeding, recall that \(d_1, d_2\in \{1, \ldots , 9 \}\). We now have that \(1 \le \ell \le 34\). We define
We rewrite inequality (3.5) as
Recall that \(n > 500\); therefore, \(\frac{4}{\alpha ^n} < \frac{1}{2}\). Hence, \(|\Lambda _2| < \frac{8}{\alpha ^n}\). Equivalently
Dividing both sides by \(\log \alpha \), we have that
Again, we apply Lemma 2.3 with the quantities
We take the same \( \kappa \) and its convergent \( p/q=p_{70}/q_{70} \) as before. Since \( m<l+m<5.5\times 10^{30} \), we choose \( M:=5.5\times 10^{30} \) as the upper bound on m. With the help of Mathematica, we get that \( \varepsilon >0.00044 \), and thus
Therefore, we have that \( n\le 294 \). This contradicts our assumption that \( n>500 \). Hence, Theorem 1.1 is proved. \(\square \)
References
Bugeaud, Y.; Mignotte, M.; Siksek, S.: Classical and modular approaches to exponential Diophantine equations I. Fibonacci and Lucas perfect powers. Ann. Math. (2) 163(2), 969–1018 (2006)
Cohen, H.: A Course in Computational Algebraic Number Theory. Springer, New York (1993)
Cohen, H.: Number Theory. Volume I: Tools and Diophantine Equations. Springer, New York (2007)
Ddamulira, M.: Repdigits as sums of three Padovan numbers. Bol. Soc. Mat. Mex. (3) 26(3), 247–261 (2020)
Ddamulira, M.: Padovan numbers that are concatenations of two distinct repdigits. Math. Slovaca 71(2), 275–284 (2021)
Dujella, A.; Pethő, A.: A generalization of a theorem of Baker and Davenport. Quart. J. Math. (2) 49(195), 291–306 (1998)
Garcíá Lomelí, A.C.; Hernández Hernández, S.: Repdigits as sums of two Padovan numbers. J. Integer Seq. 22, Art. No. 19.2.3 (2019)
Gúzman Sánchez, S.; Luca, F.: Linear combinations of factorials and \(s\)-units in a binary recurrence sequence. Ann. Math. Qué. 38(2), 169–188 (2014)
Luca, F.: Fibonacci and Lucas numbers with only one distinct digit. Portug. Math. 57(2), 243–254 (2000)
Acknowledgements
The authors thank the referee and the editor for the useful comments and suggestions that have greatly improved on the quality of presentation of the current paper.
Funding
The authors acknowledge the financial support for publication preparations costs extended from NORHED-II project “Mathematics for Sustainable Development (MATH4SD), 2021–2026” at Makerere University in collaboration with the University of Dar es Salaam and University of Bergen in Norway.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors have not disclosed any competing interests.
Additional information
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
Batte, H., Chalebgwa, T.P. & Ddamulira, M. Perrin numbers that are concatenations of two repdigits. Arab. J. Math. 11, 469–478 (2022). https://doi.org/10.1007/s40065-022-00388-8
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s40065-022-00388-8