Abstract
Let S be a finite, fixed set of primes. In this paper, we show that the set of integers c which have at least two representations as a difference between a factorial and an S-unit is finite and effectively computable. In particular, we find all integers that can be written in at least two ways as a difference of a factorial and an S-unit associated with the set of primes \(\{2,3,5,7\}\).
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
In the 1930s, Pillai (see [20, 21]) conjectured that for any given integer \(c\ge 1\), the number of positive integer solutions (a, b, x, y), with \(x, y\ge 2\) to the equation
where c is a fixed positive integer is finite. This conjecture is still open for all \( c\ne 1 \). The case \( c=1 \), the so-called Catalan’s conjecture was proved by Mihăilescu (see [19]). Pillai studied the case that a and b are fixed. Let us note that this is an extension of the work of Herschfeld (see [15, 16]), who had already studied the particular case that \((a,b)=(2,3)\). Since then, numerous variations of the Pillai equation have been studied. Some recent results of such variations of Pillai’s problem involving Fibonacci numbers, Tribonacci numbers, Pell numbers, the k-generalized Fibonacci numbers and other generalized linearly recurrent sequences, have been studied, for example, in [2,3,4,5,6,7,8,9, 13, 14].
In [12], Gúzman Sánchez and the second author studied the problem of representing a term of a binary linear recurrent sequence \(\{u_n\}\) as a linear combination of a factorial and an S-unit, in particular they studied the Diophantine equation
where s is some S-unit for a fixed set of primes S. In fact, they proved that this equation has only finitely many solutions, which are all effectively computable. This problem has been recently revisited in [17] where the case of Cullen or Woodall numbers has been considered.
In this paper, we study the hybrid problem of representing integers as difference between a factorial and an S-unit. Therefore let S be a finite set of primes. Then the set \(\mathbb {Z}_S\) of S-units consists of all integers whose prime divisors lie all within the set S. We consider for a fixed integer c the Diophantine equation
with \((s,n)\in \mathbb {Z}_S\times \mathbb {N}^+\) (we denote by \(\mathbb {N}^+\) the set of positive integers). In particular, we are interested in finding those integers c admitting at least two solutions \((s_1,n_1)\) and \((s_2,n_2)\) to (2). This is a variation of Pillai’s equation (1).
In order to find all integers c such that (2) has at least two solutions we consider the Diophantine equation
with \((s_1,s_2,n_1,n_2)\in \mathbb {Z}_S\times \mathbb {Z}_S \times \mathbb {N}^+ \times \mathbb {N}^+\). We call a solution to (3) with \(s_1=s_2\) and \(n_1=n_2\) trivial. In order to avoid trivial solutions we assume from now on that \(n_2>n_1\).
Let us also introduce the following notation to avoid technical difficulties stating our results. We write for any real number \(x\ge 1\)
Our main result is that there are only finitely many non-trivial solutions to (3) all of which are effectively computable. In particular, we prove the following theorem:
Theorem 1
Let \(S=\{p_1,\ldots ,p_k\}\) be a finite set of primes with \(p_1< \cdots <p_k\) and \(\mathbb {Z}_S\) the corresponding set of S-units and choose \(P=\max \{5,p_k\}\). Assume that \((s_1,s_2,n_1,n_2)\) is a solution to (3) with \(n_2>n_1\), and
and let \(d:=\max \{a_i,b_i\,\, 1\le i\le k \}\). Then all non-trivial solutions to Diophantine equation (3) satisfy
where \(C_d(k,P),C_{n_1}(k,P)\) and \(C_{n_2}(k,P)\) are effectively computable.
Let us note that the special case that \(k=1\) and \(p_1=2\) has been resolved by Elsholtz et.al. [11, Theorem 7]. An explicit description how to compute \(C_d(k,P)\), \(C_{n_1}(k,P)\) and \(C_{n_2}(k,P)\) is given by Lemma 7.
Although the bound for d implied by Theorem 1 is quite large and a brute force search is in practice unfeasible it is possible to use approximation lattices to reduce these huge bounds for a concrete choice of the set of primes S. Taking \(S=\{2,3,5,7\}\) we can prove the following numerical result.
Theorem 2
Let \(S=\{2,3,5,7\}\) and \(\mathbb {Z}_S\) be its corresponding set of S-units. Then, there exist exactly 1080 integers c such that
where \(s_1,s_2\in \mathbb {Z}_S\) and \(n_2>n_1>0\) are integers. In all instances we have
and \(n_2\le 22\) and \(d\le 28\).
In particular, the Diophantine equation (2) has for \(c\ne 0\) at most six solutions. Moreover, (2) has
-
exactly ten solutions if and only if \(c=0\),
-
exactly six solutions if and only if \(c=48\),
-
exactly five solutions if and only if \(c=6,8,30,4680,5760\),
-
exactly four solutions if and only if
$$\begin{aligned} c= & {} 1, 3, 4, 15, 24, 26, 80, 120, 144, 360, 480, 624,\\{} & {} 840, 1680, 45360, 47880, 241920 \end{aligned}$$ -
and provided that \(c<0\) it has exactly three solutions if and only if
$$\begin{aligned} c=-20,- 630, -4480, -226800, -1170505728000. \end{aligned}$$
We refrain from listing all integers c and all solutions to (3) in the case that \(S=\{2,3,5,7\}\).
The rest of the paper is organized as follows. In Sect. 2, we state lower bounds for linear forms in complex and p-adic logarithms which are important for our argument. Moreover, we explain how approximation lattices in combination with LLL-latices can reduce our initial bounds. In Sect. 3, we give the proof of Theorem 1. In Sect. 4, we show how to reduce the huge initial bounds for \(n_1,n_2\) and d given by Theorem 1 to small bounds by using approximation lattices. A tricky computer search allows us to find all solutions to (3). This computer search is explained in detail in Sect. 5.
2 Results from Diophantine approximation
To prove our main result Theorem 1, we use several times a Baker-type lower bound for a non-zero linear form in logarithms of algebraic numbers. There are many such bounds in the literature like that of Baker and Wüstholz from [1]. In this paper we use the result of Matveev [18], which is one of our main tools.
We start by recalling some basic notions from height theory. Let \( \gamma \) be an algebraic number of degree d with minimal primitive polynomial over the integers
where the leading coefficient \( a_{0} \) is positive and the \( \gamma ^{(i)} \)’s are the conjugates of \( \gamma \). Then the logarithmic height of \( \gamma \) is given by
In particular, if \( \gamma = p/q \) is a rational number with \( \gcd (p,q) = 1 \) and \( q\ge 1 \), then \( h(\gamma ) = \log \left( \max \{|p|, q\}\right) \). The following are some of the properties of the logarithmic height function \( h(\cdot ) \), which will be used in the next sections of this paper without further reference:
Theorem 3
(Matveev) Let \(\gamma _1,\ldots ,\gamma _t\) be positive real algebraic numbers in a real algebraic number field \(\mathbb {K}\) of degree D, \(e_1,\ldots ,e_t\) be nonzero integers, and assume that
is nonzero. Then,
where
and
A p-adic analogue of Matveev’s theorem is due to K. Yu [23]. Although Yu’s result holds for arbitrary number fields we only need the following rational version of Yu’s result. Therefore, let us denote by \(\nu _p(\cdot )\) the standard p-adic valuation on the rational field \(\mathbb {Q}\).
Theorem 4
(Yu) Let \(\gamma _1,\ldots ,\gamma _t\) be nonzero rational numbers and \(b_1,\ldots ,b_n\) nonzero integers, and assume that
is nonzero. Then we have
where \(B\ge 3\) is a number such that \(|e_i|\le B\) for \(1\le i\le t\) and
The following lemma is also useful which is Lemma 7 in [12].
Lemma 1
(Gúzman Sánchez, Luca) If \(m\geqslant 1\), \(Y>(4m^2)^m\) and \(Y>x/(\log x)^m\), then
Let \(\mathcal {L}\subseteq \mathbb {R}^k\) be a k-dimensional lattice with LLL-reduced basis \(v_1,\ldots ,v_k\) and denote by B be the matrix with columns \(v_1,\ldots , v_k\). Moreover, we denote by \(v^*_1,\ldots ,v^*_k\) the orthogonal basis of \(\mathbb {R}^k\) which we obtain by applying the Gram-Schmidt process to the basis \(v_1,\ldots ,v_k\). In particular, we have that
Further, let us define
where \(\Vert \cdot \Vert \) denotes the Euclidean norm on \(\mathbb {R}^k\). It is well known, that by applying the LLL-algorithm it is possible to give in polynomial time a lower bound \(c_1\) for \(l(\mathcal {L},y)\) (see e.g. [22, Sect. 5.4]):
Lemma 2
Let \(y\in \mathbb {R}^k\), \(z=B^{-1}y\). Furthermore we define
-
If \(y\not \in \mathcal {L}\) let \(i_0\) be the largest index such that \(z_{i_0}\ne 0\) and put \(\sigma =\{z_{i_0}\}\), where \(\{\cdot \}\) denotes the distance to the nearest integer.
-
If \(y\in \mathcal {L}\) we put \(\sigma =1\).
Finally let
Then, we have
In some of our applications we are given real numbers \(\eta _0,\eta _1,\ldots ,\eta _k\) linearly independent over \(\mathbb {Q}\) and two positive constants \({\tilde{c}}_3,{\tilde{c}}_4\) such that
where the integers \(x_i\) are bounded by \(|x_i| \le X_i\), for given upper bounds \(X_i\) with \(1\le i\le k\). We write \(X_0=\max _{1 \le i \le k}\{X_i\}\). The basic idea in such a situation, due to de Weger [10], is to approximate the linear form (6) by an approximation lattice. Namely, we consider the lattice \(\mathcal {L}\) generated by the columns of the matrix
where \({\tilde{C}}\) is a large constant usually of the size of about \(X_0^k\). Moreover, we denote by \(\lfloor x\rfloor \) the largest integer \(\le x\). Let us assume that we have an LLL-reduced basis \(b_1,\ldots ,b_k\) of \(\mathcal {L}\) and that we have a lower bound \(l(\mathcal {L},y)\ge {\tilde{c}}_1\) with \(y=(0,0,\ldots ,-\lfloor {C\eta _0}\rfloor )\). Note that \({\tilde{c}}_1\) can be computed by using the results of Lemma 2. Then we have with these notations the following lemma (e.g. see [22, Lemma VI.1]):
Lemma 3
Assume that \(S=\sum _{i=1}^{k-1}X_i^2\) and \(T=\frac{1+\sum _{i=1}^k{X_i}}{2}\). If \({\tilde{c}}_1^2 \ge T^2+S\), then inequality (6) implies that we have either \(x_1=x_2=\cdots =x_{k-1}=0\) and \(x_k=-\frac{\lfloor {{\tilde{C}}\eta _0}\rfloor }{\lfloor {{\tilde{C}}\eta _k}\rfloor }\) or
In our applications we also have to apply these techniques in a p-adic setting. Therefore, let us introduce p-adic logarithms and p-adic approximation lattices.
For a prime number p denote by \(\mathbb {Q}_p\) the field of p-adic numbers with the standard p-adic valuation \(\nu _p\). Note that \(\mathbb {Q}_p\) is complete with respect to the p-adic norm \(|x|_p:=p^{-\nu _p(x)}\) and there exists a function \(\log _p x\), the so-called p-adic logarithm, defined on all \(\mathbb {Q}_p\) such that \(\log _p(xy) = \log _p x + \log _p y\) and for every \(\xi \in \mathbb {Q}_p\) with \(|\xi - 1|_p < p^{-1/(p-1)}\) the p-adic logarithm is defined by
Moreover, we have
or equivalently
provided that \(|\xi - 1|_p < p^{-1/(p-1)}\). Note that in the case that p is an odd prime \(|\xi - 1|_p < p^{-1/(p-1)}\) is equivalent to \(\nu _p(\xi -1)\ge 1\) and \(|\xi - 1|_2 < 2^{-1/(2-1)}\) is equivalent to \(\nu _2(\xi -1)\ge 2\).
In our application we want to find an upper bound B such that
Provided that \(B\ge 1\) if p is odd and \(B\ge 2\) if \(p=2\), this is equivalent to finding an upper bound B such that
holds for all \(|x_i|\le X_0\), with \((x_1,\ldots ,x_t)\ne (0,\ldots ,0)\), where the bound \(X_0\) is given. In particular, we want to choose B as small as possible. Indeed we are interested not only to bound one linear form in p-adic logarithms, but we want to bound several linear forms of p-adic logarithms simultaneously with different primes p.
Let \(k,t>0\) be fixed integers and assume that for each \(1\le i \le k\) we are given a prime \(p_i\) and \(\eta _{0,i},\ldots ,\eta _{t,i} \in \mathbb {Z}_{p_i}\). Assume that for each \(1\le i \le k\) we are given a positive integer \(u_i\) and additionally assume that \(u_i\ge 2\) in case that \(p_i=2\). Let us denote by \({\bar{\eta }}_{j,i}\) the smallest non-negative integer such that
We consider the \(k+t\) dimensional approximation lattice \(\mathcal {L}\) spanned by the columns of the matrix
and the vector \(\vec y=(0,\ldots ,0,-{\bar{\eta }}_{0,1},\ldots ,-{\bar{\eta }}_{0,k})^T\). Assume that we have a lower bound \(c_1\) such that \(l(\mathcal {L},y)\ge c_1\). Then the following lemma can be applied (see [22, Lemma VI.3] for the case that \(k=1\)):
Lemma 4
Let us assume that \(c_1>X_0 \sqrt{t}\), then the only solution to the system
of inequalities with \(|x_i|\le X_0\), is \(x_1=\cdots =x_t=0\).
Proof
We follow the ideas of the case \(k=1\) proof given in [22, Lemma VI.3]. Assume that the system of inequalities holds for \(x_1,\ldots ,x_t\). Then we have by the definition of \({\bar{\eta }}_{i,j}\) the system of modular equations:
Therefore, the numbers
are all integers and
By the definition of \(\vec y\) we have
If \(\vec x \ne \vec y\), then we have
which is a contradiction to our initial assumption that \(c_1>X_0 \sqrt{t}\). Hence, we have \(\vec x = \vec y\), which implies \(x_1=\cdots =x_t=0\). \(\square \)
3 Proof of Theorem 1
In order to avoid trivial solutions we assume that \(s_1\ne s_2\) and \(n_2>n_1\). We rewrite Eq. (3) as
We put \(S=\{p_1,\ldots ,p_k\}\) and consider the set
of S-units. Let us write
Let \(P=\max \{5,p_k\}\) and \(d=\max _{i=1,\ldots ,k} \{a_i,b_i\}\). Furthermore, let p be the first prime such that \(p\not \in S\). Due to Bertrand’s postulate we have \(p<2P\). Let us also note that in any case we can choose \(p\ne 2\); i.e., \(p\ge 3\).
Let us recall that for every integer \(n\in \mathbb {N}\) we have, provided that \(n\ge p\)
We will also need the finer estimate
where \(s_p(n)\) denotes the sum of digits of the p-adic digit expansion of n. Also, we will frequently use the following obvious estimates:
We divide the proof of our main theorem into three steps:
3.1 Bounding \(n_1\) in terms of d
Let us assume that \(n_1\ge p\), since otherwise we have \(n_1<2P\) and Lemma 5 below will hold trivially. In view of Theorem 1 we may also assume that \(d\ge 3\).
We consider Eq. (3) and obtain by considering p-adic valuations the following inequality
where the last inequality follows from (9). On the other-hand, we have that
Now, we are entitled to apply Theorem 4 to obtain an upper bound for (13).
In our application, we take \(t=k\) and put
Let
Then we have \(\Lambda _1\ne 0\) since by our hypothesis \(s_1\ne s_2\). Thus, we have with \(B=d\) the inequality
where
Since \(h(p_i)= \log p_i\) and \(p_i<p<2P\), then we can take \(H_i=\log (2P)\) for \(i=1,\ldots ,k\). Furthermore, inequalities (12) and (14) yield
First, let us note that \(\log (ke^5)=5+\log k\le 6\log ^+k\). Moreover let us note that the function \(f(x)=\frac{x}{\log x}\) is strictly increasing for \(x\ge e\). Therefore, replacing \(\left( \frac{p}{\log p}\right) ^2\) by \(4\left( \frac{P}{\log 2P}\right) ^2\) we obtain
We record this as a lemma:
Lemma 5
All solutions to Eq. (3) satisfy \(n_1\le C_1(k,P)\log d\) with
3.2 Bounding \(n_2\) in terms of k
Let \(q\in S\) be the smallest prime such that
We can assume that \(\nu _q(n_2)>\nu _q(n_1)\), since \(\nu _q(n_2)=\nu _q(n_1)\) would imply that
Let us assume for the moment that \(d\le v_q(n_1!)\). This would imply that
By an application of Lemma 1 we obtain
Thus, we may assume that \(d>v_q(n_1!)\). We distinguish now between four cases:
- Case 1a::
-
\(d=\nu _q(s_1)\) and \(\nu _q(s_2+n_1!)< d\);
- Case 1b::
-
\(d=\nu _q(s_1)\) and \(\nu _q(s_2+n_1!)\ge d\);
- Case 2a::
-
\(d=\nu _q(s_2)\) and \(\nu _q(s_1-n_1!)< d\);
- Case 2b::
-
\(d=\nu _q(s_2)\) and \(\nu _q(s_1-n_1!)\ge d\).
Let us note that the Cases 1a and 2a respectively 1b and 2b can be treated similarly. Thus, we will give details only for the cases 1a and 1b.
3.2.1 Case 1a: \(\nu _q(s_2+n_1!)< d\)
In this case we have that
Furthermore, inequality (9) gives that
Now, let us apply Theorem 4 to
with \(t=k+1\) and
Let us note that since \(p_1^{b_1} \ldots p_k^{b_k}(-n_1!)^{-1}<0\), we have \(\Lambda _2\ne 0.\) Thus, with \(B=d\) we get
However, with \(H_i=\log P,\) for \(i=1,\ldots ,k\) we get
since
and \(p_i,q<P\). Furthermore, we choose \(H_{k+1}=n_1^2\) since we have
due to (10). Altogether, we get
Therefore, a combination of the inequalities (15), (16) and Lemma 5 yields the following bound for \(n_2\):
where
3.2.2 Case 2a: \(\nu _q(s_1-n_1!)< d\)
This case is almost identical to Case 1a. However, instead of \(\Lambda _2\) we consider
An application of Yu’s theorem (Theorem 4) yields the same bound for \(n_2\) as in the Case 1a.
3.2.3 Case 1b: \(\nu _q(s_2+n_1!)\ge d\)
In this case, we have that
and by using inequality (16), we obtain similarly as in (17) the upper bound
But, we also have by Eq. (3) that
Moreover, we also have
where we used the fact that \(n!\ge \left( \frac{n}{e}\right) ^n\) on the right-side of the above inequality. Hence, we have
Using the above inequality and (19), we get
Note that \(n_2\le (n_2-1)\log \left( \frac{n_2-1}{e}\right) \) holds for all \(n_2\ge 11\). Therefore, we have
3.2.4 Case 2b: \(\nu _q(n_1!-s_1)\ge d\)
Since we have the same upper bound for \(\nu _q(n_1!-s_1)\) as for \(\nu _q(n_1!+s_2)\) this case can be resolved as Case 1b and we get the same upper bound (21) for \(n_2\).
We summarise what we have proved so far in the following lemma.
Lemma 6
All solutions of the Eq. (3) satisfy
with \(C_3(k,P)=C_2(k,P) k \log P\), where \(C_2(k,P)\) is given by (18).
3.3 Bounding d in terms of k
We rewrite Eq. (3) as
Let us assume for the moment that there is a \(q\in S\) such that \(\nu _q(s_1)=d\). Then, we have that
Now, we use Matveev’s theorem (Theorem 3) to obtain a lower bound for the left-hand side of the above inequality. Therefore, we put
We apply Theorem 3 with \(t=k\), \((\gamma _1,\ldots , \gamma _k):=(p_1,\ldots , p_k)\), \(e_i:=b_i-a_i\) and \(B=d\). As before, we can choose \(A_i=\log P\ge h(p_i)\) and \(D=1\). Therefore, we have that
Comparing this last inequality with (22), we get
Assume next, that we are in the case that there is a \(q\in S\) such that \(\nu _q(s_2)=d\). We consider
instead and obtain by the same argument inequality (23).
Let us now distinguish between the two following cases:
- Case 1::
-
\(n_2 \log n_2 \le d/3\) and
- Case 2::
-
\(n_2 \log n_2 > d/3\).
3.3.1 Case 1: \(n_2\log n_2\le d/3\)
In this case, we have
and together with (23) we get by assuming that \(d>3\) the inequality
Applying Lemma 1 on (23) with the data \(m=1\), \(Y:= 1.4 \times 30^{k+3} \times k^{4.5} (\log P)^k,\) and \(x= d\) leads to
3.3.2 Case 2: \(n_2\log n_2 > d/3\)
Now, we have \(n_2\log n_2> d/3\) and applying Lemma 6, we get
Therefore, we have due to Lemma 1
In any case we have
Then, from Lemmas 5, 6 and inequality (26), we obtain the following bound on \(n_1\), \(n_2\) and d which prove Theorem 1.
Lemma 7
Let us define:
Then all non-trivial solutions to Eq. (3) satisfy
4 Reduction of the bounds
Now, we consider the case where \(S=\{2,3,5,7\}\). Using the bounds given in Lemma 7 we obtain the following bounds for \(n_1,n_2\) and d which we record in the following lemma:
Lemma 8
Let \((s_1,s_2,n_1,n_2)\) be a non-trivial solution to (3), where \(S=\{2,3,5,7\}\). Then we have
We want to reduce these huge bounds to much smaller bounds. We do this in three reduction steps:
Step 1: We show that \(n_1\le 2210\):
We consider the main Eq. (3) and rewrite it to
Let us assume for the moment that \(n_1\ge 2p_i\) for some prime \(p_i>7\). Then the \(p_i\)-adic absolute value on the right hand side is at most \(p_i^{-2}\). Therefore, we obtain by the properties of the p-adic logarithm the inequality:
Let us consider the p-adic approximation lattice from (8), with
Moreover, we choose \({\bar{\eta }}_{i,t}\equiv \log _{p_t} {q_i} \mod p_t^{u_t}\), where \(q_1=2\), \(q_2=3\), \(q_3=5\) and \(q_4=7\). With \(X_0\le 1.17 \times 10^{84}\) we obtain \(l(\mathcal {L},y)>5.43\times 10^{94}>2 X_0\), where \(y=(0,\ldots ,0)^T\). That is either
or at least one of the inequalities \(\nu _{p_i}(n_1!)< u_i\) holds due to Lemma 4.
Let us note that due to (10) the inequalities
imply \(n_1\le 2210\), \(n_1\le 2209\), \(n_1\le 2209\) and \(n_1\le 2203\), respectively.
Step 2: Let \(q\in S=\{2,3,5,7\}\) such that \(v_q(s_1)=d\) or \(v_q(s_2)=d\) holds. We show that
-
\(n_2\le 3413\) if \(q=2\);
-
\(n_2\le 3791\) if \(q=3\);
-
\(n_2\le 4364\) if \(q=5\);
-
\(n_2\le 4871\) if \(q=7\).
Let us assume that \(q=2\). We start by showing that in this case we have \(\nu _2(\pm s_i/n_1!-1)\le 1199\) for \(i=1,2\). Assume for the moment that \(n_1\le 10\). Then \(s_i/n_1!\) is a rational number, such that the only prime factors appearing in the numerator or denominator are 2, 3, 5 or 7. We want to find an upper bound for \(v_2(\pm s_i/n_1!-1)\). Let us assume for the moment that \(\nu _2(\pm s_i/n_1!-1)\ge 2\), then we have due to properties of the 2-adic logarithm
where \(s_i/n_1!=3^{a'_2}5^{a'_3}7^{a'_4}\). Note that if in the prime factorization of \(s_i/n_1!\) would appear a 2, we would have \(\nu _2(\pm s_i/n_1!-1)=0\), which contradicts our assumption. We consider now the approximation lattice (8), with \(t=1\), \(p_1=2\), \(u_1=1200\) and \(\eta _{i,1}=\log _2 3\), \(\eta _{i,2}=\log _2 5\) and \(\eta _{i,3}=\log _2 7\). With
and \(y=(0,0,0,0)^T\) the inequality \(l(\mathcal {L},y)>\sqrt{3} X_0\) is satisfied and an application of Lemma 4 yields either \(s_i=n_1!\) or \(\nu _2(\pm s_i/n_1!-1)\le 1199\). Note that \(s_1=n_1!\) implies \(s_2=n_2!\), that is \(n_2!\) has only the prime factors 2, 3, 5 and 7, which implies \(n_2\le 10\). Furthermore \(s_2=n_1!\) implies that \(2s_2-s_1=n_2!\). Now, we can proceed as in Step 1 but consider the inequality
instead of inequality (27). Hence, we obtain similarly as in Step 1 that \(n_2\le 2210\) by an application of Lemma 4.
Assume next that we are in the case that \(n_1>10\). We consider the approximation lattice (8), with \(t=1\), \(p_1=2\), \(u_1=1200\) and \(\eta _{i,1}=\log _2 3\), \(\eta _{i,2}=\log _2 5\), \(\eta _{i,3}=\log _2 7\) and \(X_0\le 1.17 \times 10^{83}\). For each \(11\le n_1 \le 2210\) we consider the inequality
and an application of Lemma 4 with \(y=(0,0,0,\log _2 (n_1!) \mod 2^{1298})^T\) shows that this inequality has at most a solution with \(a_2=a_3=a_4=0\). Computing \(|\log _2 (n_1!)|_2\) for all instances, we indeed have no solution to inequality (28).
Therefore we have shown that \(\nu _2(\pm s_i/n_1!-1)\le 1199\) holds for \(i=1,2\). This bound implies
According to the various cases that we considered in Subsection 3.2 we have either
or
In the latter case we obtain due to (10) the upper bound \(n_2\le 3413\).
Let us assume for the moment that \(d\le 3405\) holds. We consider inequality (20) and obtain in our case the inequality
which yields \(n_2\le 2646\).
Thus, we obtain that \(n_2\le 3413\), provided that \(q=2\).
Similar computations for \(q=3\), \(q=5\) and \(q=7\) show that \(n_2\le 3791\), \(n_2\le 4364\) and \(n_2\le 4871\) holds, respectively.
Step 3: We show that
-
\(d\le 36088\) if \(q=2\);
-
\(d\le 25586\) if \(q=3\);
-
\(d\le 20425\) if \(q=5\);
-
\(d\le 19091\) if \(q=7\).
We assume that \(q=2\) and reconsider inequality (22) and aim to apply Lemma 3. In particular, we consider the inequality
where \(|x_i|=|b_i-a_i|\le X_0:=1.17\times 10^{83}\). Thus, we consider the approximation lattice
with \(C=10^{380}\) and \(y=(0,0,0,0)^T\). Moreover, we put \({\tilde{c}}_3=3413!\) and \({\tilde{c}}_4=\log 2\). Thus, an application of Lemma 2 with Lemma 3 yields \(d\le 36088.\)
In the cases that \(q=3\), \(q=5\) and \(q=7\) we obtain similar bounds for d.
Thus, we have found after the three steps the new much smaller upper bounds
We can repeat the previously described reduction steps with these bounds inductively and obtain after three more applications of Steps 1–3:
Lemma 9
Let \((s_1,s_2,n_1,n_2)\) be a non-trivial solution to (3), where \(S=\{2,3,5,7\}\). Then we have \(n_1\le 90\) and
-
\(n_2\le 186\), \(d\le 931\) if \(q=2\);
-
\(n_2\le 182\), \(d\le 722\) if \(q=3\);
-
\(n_2\le 204\), \(d\le 565\) if \(q=5\);
-
\(n_2\le 230\), \(d\le 539\) if \(q=7\).
Let us consider the equation
Let \(p\not \in S\) be a prime and assume that
holds for some integer \(u_p\). Then by taking p-adic logarithms we obtain
where \(|x_i|=|b_i-a_i|\le 931\). Taking p-adic logarithms we obtain
where we write \(\eta _{q,p}=\log _p q \mod p^{u_p}\).
We consider the primes \(p=11,13,17\) and \(p=19\) and choose \(u_{11}=4\), \(u_{13}=4\), \(u_{17}=3\) and \(u_{19}=2\). Considering inequality (29) modulo \(p^{u_p}\) for \(p=11,13,17\) and 19 we obtain the modular linear system
Let us note that a solution of this linear system does not yield a solution to (3). But on the contrary every solution to (3) with \(\nu _p(n_1!)\ge u_p\) yields a solution to the linear system.
To find all solutions to (30) with \(|x_i|\le 931\) we compute two lists. For \(-931\le x_1,x_2\le 931\) we compute the list List1 consisting of the sixtuples
and the list List2 consisting of the sixtuples
If for two entries, one from List1 and one from List2, the first four entries coincide then we have found a solution \((x_1,x_2,x_3,x_4)\) to (30). Constructing these lists and comparing them takes a few minutes on a usual PC. We found 747 possible solutions, but only the trivial solution \(x_1=x_2=x_3=x_4=0\) satisfies
for \(p=11,13,17\) and 19.
Therefore, we have proved that there exists no solution with \(v_{p}(n_1!)\ge u_p\). That is there exists no solution with \(n_1\ge 52\). Therefore we may assume that \(n_1\le 51\). With this new bound for \(n_1\) we can perform the reduction Steps 2 and 3 and obtain \(n_2\le 188\) and \(d\le 655\).
We reconsider the Diophantine inequality (29) for \(p=11,13,17\) and 19 with \(u_{11}=4\), \(u_{13}=3\), \(u_{17}=2\) and \(u_{19}=2\). We get a similar linear system of modular equations as in (30) and find all solutions by computing again two long lists. There are 40669 possible solutions but only the trivial solution \(x_1=x_2=x_3=x_4=0\) satisfies
for \(p=11,13,17\) and 19 and additionally
Therefore, we may conclude that \(v_{p}(n_1!)\ge u_p\) does not hold; i.e., we have \(n_1\le 43\). The reduction Steps 2 and 3 yield now \(n_2\le 174\) and \(d\le 574\).
Applying a third time the same approach with \(u_{11}=3\), \(u_{13}=3\), \(u_{17}=2\) and \(u_{19}=2\) leads to the following lemma:
Lemma 10
Let \((s_1,s_2,n_1,n_2)\) be a non-trivial solution to (3), with \(S=\{2,3,5,7\}\). Then we have
For some \(p\not \in S\) we consider again the inequality
for some integer \(u_p\). That is we consider the modular equation
Instead of using p-adic logarithms to resolve this inequality we try to find solutions directly.
That is we choose \(p=11,13,17\) and 19 together with \(u_{11}=3 \), \(u_{13}=2 \), \(u_{17}=2 \) and \(u_{19}=1\). Similar as in the p-adic logarithms approach we compute two lists. For \(-548\le x_1,x_2\le 548\) we compute the list List1 consisting of the sixtuples
and the list List2 consisting of the sixtuples
We compare these lists and if we have found a sixtuple form List1 and a sixtuple form List2 for which the first four entries coincide, then we have found a solution \((x_1,x_2,x_3,x_4)\) to our system of modular equations (32). Indeed we find 3163 solutions, but none of them, except the trivial solution \(x_1=x_2=x_3=x_4=0\), satisfies
Therefore, we conclude that \(\nu _p(n_1!)<u_p\) for some \(p\in \{11,13,17,19\}\). Thus, we conclude \(n_1\le 33\).
Performing our reduction Steps 2 and 3 with the new upper bound for \(n_1\) we obtain
Next, we aim to prove that \(n_1\le 22\). Therefore, we consider inequality (31) with \(p=11\), \(p=13\), \(p=17\) and \(p=19\) again. But, this time we employ the multiplicative modular approach from above with \(u_{11}=2\), \(u_{13}=1 \), \(u_{17}=1 \) and \(u_{19}=1\) instead. Again we dismiss those solutions which do not satisfy
This leaves us with a list of 256777 possible solutions \((x_1,x_2,x_3,x_4)\). Let us call this list Solutions. We have to find a way to show that these solutions do not yield a solution to the main Eq. (3). Therefore, we perform for each \(n_1\) and \(n_2\) with \(23\le n_1 \le 33\) and \(n_1<n_2\le 146\) and for each solution \((x_1,x_2,x_3,x_4)\) from the list Solutions the following computations (this takes about 1 h on a usual desktop PC):
-
(1)
We compute \(M=n_2!-n_1!\).
-
(2)
We compute for each solution \((x_1,x_2,x_3,x_4)\) from the list Solutions the numerator N of \(2^{x_1}3^{x_2}5^{x_3}7^{x_4}-1\).
-
(3)
If \(M\not \equiv 0 \mod N\), then we discard the solution.
Besides the trivial solution, we are left with four possible candidates:
We want to exclude the candidate \((n_1,n_2,x_1,x_2,x_3,x_4) =(29,63,32,0,-14,2)\). Therefore, we note
and since \(v_3(2^{32}5^{-14}7^2-1)=2\) we must have
since otherwise the p-adic valuation would not coincide on both sides for \(p=2,3,5\) and 7. But, a numerical computation shows that with \(n_2=63\) and \(n_1=29\) this equation does not hold; i.e., this candidate does not yield a solution to (3). By a similar argument the other three candidates can be discarded. Hence, we have \(n_1\le 22\).
With this smaller bound for \(n_1\) we once again carry out the reduction Steps 2 and 3 and obtain
More precisely we have the following lemma:
Lemma 11
Let \((s_1,s_2,n_1,n_2)\) be a non-trivial solution to (3), with \(S=\{2,3,5,7\}\). Then we have \(n_1\le 22\) and
-
in the case that \(q=2\), we have \(n_2\le 106\) and \(d\le 419\);
-
in the case that \(q=3\), we have \(n_2\le 83\) and \(d\le 288\);
-
in the case that \(q=5\), we have \(n_2\le 114\) and \(d\le 285\);
-
in the case that \(q=7\), we have \(n_2\le 139\) and \(d\le 298\).
Next, we aim to show that \(n_2\le 48\). Therefore, we consider the four cases discussed in Sect. 3.2:
- Case 1a::
-
\(d=\nu _q(s_1)\) and \(\nu _q(s_2+n_1!)< d\);
- Case 1b::
-
\(d=\nu _q(s_1)\) and \(\nu _q(s_2+n_1!)\ge d\);
- Case 2a::
-
\(d=\nu _q(s_2)\) and \(\nu _q(s_1-n_1!)< d\);
- Case 2b::
-
\(d=\nu _q(s_2)\) and \(\nu _q(s_1-n_1!)\ge d\).
Let us remind the reader that q is chosen such that \(\max \{\nu _q(s_1),\nu _q(s_2)\}=d\). Let us assume for the moment that \(q=7\). We want to show that \(\nu _7(s_2(-n_1!)^{-1}-1)\le 6\) respectively \(\nu _7(s_1(n_1!)^{-1}-1)\le 6\). In the Case 1a (and similarly in the Case 2a) this leads, as it has been shown in Sect. 3.2, to
Therefore, we obtain \(n_2\le 62\).
In the Case 1b (and similarly in the Case 2b) we would obtain \(d\le 6\) which yields
This implies \(n_2\le 17\).
To show that indeed \(\nu _7(s_2(-n_1!)^{-1}-1)\le 6\) holds we proceed as follows. First, we note that \(\nu _7((s_2(-n_1!)^{-1}-1)>0\) if and only if \(\nu _7(n_1!)=\nu _7(s_2)\); i.e., the powers of 7 cancel each other out. Similar as above we compute two lists, where List1 consists of all triples of the form
with \(0\le x_1,x_2\le 298\) and List2 consists of all triples of the form
with \(0\le x_3\le 298\) and \(1\le n_1\le 22\), where \((n_1!)_7\) denotes the 7-free part of the factorial of \(n_1\); i.e., \((n_1!)_7=\frac{n_1!}{7^{\nu _7(n_1!)}}\). Entries from List1 and List2 whose first entry coincide yield a solution to
In particular, we find 865 solutions \((n_1,x_1,x_2,x_3)\). For each solution we compute
and find that
which implies \(n_2\le 90\) or \(d\le 13\). But since
implies \(n_2\le 90\), we may assume that \(n_2\le 90\) holds in Cases 1a and 1b. A similar computation concerning \(\nu _7(s_1(n_1!)^{-1}-1)\) we obtain that \(n_2\le 90\) also holds in the Cases 2a and 2b. We apply reduction Step 3 with this new lower bound for \(n_2\) and find \(d\le 175\).
Once again we compute the lists List1 and List2 consisting of the triples
and
By comparing the lists we find a list Solutions of 58277 solutions to
For all \(49\le n_2\le 90\) we check for all solutions \((n_1,x_1,x_2,x_3)\) whether the number
has only 2, 3, 5 or 7 as prime divisors; i.e., yields a solution to (3). However, none of the solutions from Solutions yields a solution to (3). That is \(n_2\le 48\) or \(\nu _7(n_1!-s_1)\le 3\). Note that the later case yields
and therefore also \(n_2\le 48\) in the Cases 1a and 1b. Similarly, we can show that in the Cases 2a and 2b also \(n_2\le 48\) holds; i.e., in any case we have \(n_2\le 48\). Moreover, by an application of the reduction Step 3 we obtain \(d\le 84\) in the case that \(q=7\).
And in general for \(q=2,~q=3\) and \(q=5\) we proceed similarly. However in the case that \(q=2\) the Cases 1b and 2b yield larger upper bounds for \(n_2\) but much smaller upper bounds for d. Therefore, we obtain:
Lemma 12
Let \((s_1,s_2,n_1,n_2)\) be a non-trivial solution to (3), with \(S=\{2,3,5,7\}\). Then we have \(n_1\le 22\) and
-
in the case that \(q=2\), we have \(n_2\le 33\) and \(d\le 156\) (Cases 1a and 2a) or \(n_2\le 61\) and \(d\le 36\) (Cases 1b and 2b);
-
in the case that \(q=3\), we have \(n_2\le 35\) and \(d\le 105\);
-
in the case that \(q=5\), we have \(n_2\le 39\) and \(d\le 81\);
-
in the case that \(q=7\), we have \(n_2\le 48\) and \(d\le 84\).
Next for all \(0\le x_1,x_2,x_3,x_4\le 36\):
-
We compute \(s_2=2^{x_1}3^{x_2}5^{x_3}7^{x_4}\).
-
We find the smallest zero of \(\Gamma (x)-s_2=0\) numerically and obtain the upper bound \(n_2\le x\) since \((n_2-1)!=\Gamma (n_2)\) and the \(\Gamma \) function is striclty increasing for \(x\ge 0\).
-
If \(x<34\) we continue to the next quadruple \((x_1,x_2,x_3,x_4)\).
-
For all \(34\le n_2\le x\) and all \(1\le n_1\le 22\)
-
we compute \(M=s_2-n_2!+n_1!\).
-
If \(M<0\) we discard this possible solution.
-
We compute \(v_p=\nu _p(M)\) for \(p=2,3,5,7\) and if \(M=2^{v_2}3^{v_3}5^{v_5}7^{v_7}\) we have found a solution to Diophantine equation (3).
-
However, this computation yields no solution. Thus, we have proved so far
Proposition 1
Let \((s_1,s_2,n_1,n_2)\) be a non-trivial solution to (3), with \(S=\{2,3,5,7\}\). Then we have \(n_1\le 22\) and
-
in the case that \(q=2\), we have \(n_2\le 33\) and \(d\le 156\);
-
in the case that \(q=3\), we have \(n_2\le 35\) and \(d\le 105\);
-
in the case that \(q=5\), we have \(n_2\le 39\) and \(d\le 81\);
-
in the case that \(q=7\), we have \(n_2\le 48\) and \(d\le 84\).
5 Final computations
The bounds for \(n_1\) and \(n_2\) are now small enough to perform a brute force computer search. However, before we perform this computer search we once again apply the reduction Step 3 with the bounds for \(n_1\) and \(n_2\) given by Lemma 12.
The reduction Step 3 is to reconsider inequality (22) which is
An application of Lemma 3 yields
with \({\tilde{c}}_4=1\), \({\tilde{c}}_3=n_2!\). With the choice of \(C=10^{11}\) we have \({\tilde{c}}_1^2 \ge T^2+S\) and therefore obtain
We will divide our brute force search into four segments:
-
\(n_2\le 33\);
-
\(n_2=34,35\);
-
\(36\le n_2\le 39\);
-
\(40 \le n_2 \le 48\).
For \(1\le n_2\le 33\) we proceed as follows:
-
We compute \(B=\log (10^{11} n_2!)-\log 69.52\).
-
For all \(1\le n_1 < \min \{n_2,23\}\) we compute \(B_7=\frac{B}{\log 7}\).
-
For all \(0\le x_4 \le B_7\) we compute \(B_5=\frac{B-x_4\log 7}{\log 5}\).
-
For all \(0\le x_3 \le B_5\) we compute \(B_3=\frac{B-x_4\log 7-x_3\log 5}{\log 3}\).
-
For all \(0\le x_2 \le B_3\) we compute \(B_2=\frac{B-x_4\log 7-x_3\log 5-x_2\log 3}{\log 2}\).
-
For all \(0\le x_1 \le B_2\) we compute \(M=n_2!-n_1!+2^{x_1}3^{x_2}5^{x_3}7^{x_4}\) and we compute
-
\(v_2=\nu _2(M)\);
-
\(v_3=\nu _3(M)\);
-
\(v_5=\nu _5(M)\);
-
\(v_7=\nu _7(M)\).
If \(M=2^{v_2}3^{v_3}5^{v_5}7^{v_7}\) we have found a solution
$$\begin{aligned} (n_1,n_2,x_1,x_2,x_3,x_4,v_2,v_3,v_5,v_7) \end{aligned}$$to the Diophantine equation (3) and append it to our list SolutionsMain of solutions to the Diophantine equation (3).
-
In the case that \(n_2=34\) or \(n_2=35\) the case that \(q=2\) is impossible. To use this fact efficiently we distinguish now between the two cases \(\nu _q(s_1)=d\) and \(\nu _q(s_2)=d\). In the case that \(\nu _q(s_1)=d\) holds we proceed as before. However in the case that \(\nu _q(s_2)=d\) we consider
and we obtain by the same arguments that
holds. Therefore, we proceed in the case that \(n_2=34\) or \(n_2=35\) as follows:
-
We compute \(B=\log (10^{11} n_2!)-\log 69.52\).
-
For all \(1\le n_1 \le 22\) we compute \(B_7=\frac{B}{\log 7}\).
-
For all \(0\le x_4 \le B_7\) we compute \(B_5=\frac{B-x_4\log 7}{\log 5}\).
-
For all \(0\le x_3 \le B_5\) we compute \(B_3=\frac{B-x_4\log 7-x_3\log 5}{\log 3}\).
-
For all \(0\le x_2 \le B_3\) we compute \(B_2=\frac{B-x_4\log 7-x_3\log 5-x_2\log 3}{\log 2}\).
-
For all \(0\le x_1 \le \min \{B_2,\max \{x_2,x_3,x_4\}\}\) we compute
-
\(M_1=n_2!-n_1!+2^{x_1}3^{x_2}5^{x_3}7^{x_4}\) (in case that \(d=\nu _q(s_1)\));
-
\(M_2=n_1!-n_2!+2^{x_1}3^{x_2}5^{x_3}7^{x_4}\) (in case that \(d=\nu _q(s_2)\));
-
\(v_2=\nu _2(M_1)\) and \(v'_2=\nu _2(M_2)\);
-
\(v_3=\nu _3(M_1)\) and \(v'_3=\nu _3(M_2)\);
-
\(v_5=\nu _5(M_1)\) and \(v'_5=\nu _5(M_2)\);
-
\(v_7=\nu _7(M_1)\) and \(v'_7=\nu _7(M_2)\).
If \(M_1=2^{v_2}3^{v_3}5^{v_5}7^{v_7}\) we have found a solution with \(M_1=s_1\) and we have a solution
$$\begin{aligned} (n_1,n_2,x_1,x_2,x_3,x_4,v_2,v_3,v_5,v_7) \end{aligned}$$to the Diophantine equation (3) and we append this solution to our list SolutionsMain of solutions to the Diophantine equation (3). Similarly, if \(M_2=2^{v'_2}3^{v'_3}5^{v'_5}7^{v'_7}\) we have found a solution with \(M_2=s_2\) and we have a solution
$$\begin{aligned} (n_1,n_2,v'_2,v'_3,v'_5,v'_7,x_1,x_2,x_3,x_4) \end{aligned}$$to the Diophantine equation (3) which we append to our list of solutions.
-
In the case that \(36\le n_2\le 39\), the case that \(q=2\) and \(q=3\) are impossible; i.e. we proceed similarly as in the case that \(n_2=34\) or \(n_2=35\):
-
We compute \(B=\log (10^{11} n_2!)-\log 69.52\).
-
For all \(1\le n_1 \le 22\) we compute \(B_7=\frac{B}{\log 7}\).
-
For all \(0\le x_4 \le B_7\) we compute \(B_5=\frac{B-x_4\log 7}{\log 5}\).
-
For all \(0\le x_3 \le B_5\) we compute \(B_3=\frac{B-x_4\log 7-x_3\log 5}{\log 3}\).
-
For all \(0\le x_2 \le \min \{B_3,\max \{x_3,x_4\}\}\) we compute
$$\begin{aligned} B_2=\frac{B-x_4\log 7-x_3\log 5-x_2\log 3}{\log 2}. \end{aligned}$$ -
For all \(0\le x_1 \le \min \{B_2,\max \{x_3,x_4\}\}\) we compute
-
\(M_1=n_2!-n_1!+2^{x_1}3^{x_2}5^{x_3}7^{x_4}\) (in case that \(d=\nu _q(s_1)\));
-
\(M_2=n_1!-n_2!+2^{x_1}3^{x_2}5^{x_3}7^{x_4}\) (in case that \(d=\nu _q(s_2)\));
-
\(v_2=\nu _2(M_1)\) and \(v'_2=\nu _2(M_2)\);
-
\(v_3=\nu _3(M_1)\) and \(v'_3=\nu _3(M_2)\);
-
\(v_5=\nu _5(M_1)\) and \(v'_5=\nu _5(M_2)\);
-
\(v_7=\nu _7(M_1)\) and \(v'_7=\nu _7(M_2)\).
If \(M_1=2^{v_2}3^{v_3}5^{v_5}7^{v_7}\) we have found a solution with \(M_1=s_1\) and we have a solution
$$\begin{aligned} (n_1,n_2,x_1,x_2,x_3,x_4,v_2,v_3,v_5,v_7) \end{aligned}$$to the Diophantine equation (3) and append it to our list SolutionsMain of solutions to the Diophantine equation (3). Similarly, if \(M_2=2^{v'_2}3^{v'_3}5^{v'_5}7^{v'_7}\) we have found a solution with \(M_2=s_2\) and we have a solution
$$\begin{aligned} (n_1,n_2,v'_2,v'_3,v'_5,v'_7,x_1,x_2,x_3,x_4) \end{aligned}$$to the Diophantine equation (3) which we append to our list of solutions.
-
In the case that \(40\le n_2\le 48\), we have that \(q=7\) and we proceed as follows:
-
We compute \(B=\log (10^{11} n_2!)-\log 69.52\).
-
For all \(1\le n_1 \le 22\) we compute \(B_7=\frac{B}{\log 7}\).
-
For all \(0\le x_4 \le B_7\) we compute \(B_5=\frac{B-x_4\log 7}{\log 5}\).
-
For all \(0\le x_3 \le \min \{B_5,x_4\}\) we compute \(B_3=\frac{B-x_4\log 7-x_3\log 5}{\log 3}\).
-
For all \(0\le x_2 \le \min \{B_3,x_4\}\) we compute \(B_2=\frac{B-x_4\log 7-x_3\log 5-x_2\log 3}{\log 2}\).
-
For all \(0\le x_1 \le \min \{B_2,x_4\}\) we compute
-
\(M_1=n_2!-n_1!+2^{x_1}3^{x_2}5^{x_3}7^{x_4}\) (in case that \(d=\nu _q(s_1)\));
-
\(M_2=n_1!-n_2!+2^{x_1}3^{x_2}5^{x_3}7^{x_4}\) (in case that \(d=\nu _q(s_2)\));
-
\(v_2=\nu _2(M_1)\) and \(v'_2=\nu _2(M_2)\);
-
\(v_3=\nu _3(M_1)\) and \(v'_3=\nu _3(M_2)\);
-
\(v_5=\nu _5(M_1)\) and \(v'_5=\nu _5(M_2)\);
-
\(v_7=\nu _7(M_1)\) and \(v'_7=\nu _7(M_2)\).
If \(M_1=2^{v_2}3^{v_3}5^{v_5}7^{v_7}\) we have found a solution with \(M_1=s_1\) and we have a solution
$$\begin{aligned} (n_1,n_2,x_1,x_2,x_3,x_4,v_2,v_3,v_5,v_7) \end{aligned}$$to the Diophantine equation (3) and append it to our list SolutionsMain of solutions to the Diophantine equation (3). Similarly, if \(M_2=2^{v'_2}3^{v'_3}5^{v'_5}7^{v'_7}\) we have found a solution with \(M_2=s_2\) and we have a solution
$$\begin{aligned} (n_1,n_2,v'_2,v'_3,v'_5,v'_7,x_1,x_2,x_3,x_4) \end{aligned}$$to the Diophantine equation (3) which we append to our list of solutions.
-
It takes about 150 min on a usual desktop PC to perform this computer search and obtain all 1430 solutions to Diophantine equation (3). For each solution
in our list SolutionsMain we compute the quantity
and sort our list SolutionsMain according to these values c. Therefore, we easily find all 1030 values of c such that the Diophantine equation (2) has at least two solutions. Note that the Diophantine equation (2) has exactly k solutions if the value c appears exactly \(\left( {\begin{array}{c}k\\ 2\end{array}}\right) \) times in our list SolutionsMain. Thus a search for values of c that appear exactly \(\left( {\begin{array}{c}k\\ 2\end{array}}\right) \) times leads to the lists of values of c such that Diophantine equation (2) has ten, six, five, four or three solutions, respectively.
Data Availability
This manuscript has no associated data.
References
Baker, A., Wüstholz, G.: Logarithmic forms and group varieties. J. Reine Angew. Math. 442, 19–62 (1993)
Bravo, J.J., Luca, F., Yazán, K.: On Pillai’s problem with Tribonacci numbers and powers of 2. Bull. Korean Math. Soc. 54(3), 1069–1080 (2017)
Chim, K.C., Pink, I., Ziegler, V.: On a variant of Pillai’s problem. Int. J. Number Theory 13(7), 1711–1727 (2017)
Chim, K.C., Pink, I., Ziegler, V.: On a variant of Pillai’s problem II. J. Number Theory 183, 269–290 (2018)
Ddamulira, M.: On the problem of Pillai with tribonacci numbers and powers of 3. J. Integer Seq. 22(5), Art. 19.5.6, 14 (2019)
Ddamulira, M.: On a problem of Pillai with Fibonacci numbers and powers of 3. Bol. Soc. Mat. Mex. (3) 26(2), 263–277 (2020)
Ddamulira, M., Luca, F.: On the problem of Pillai with \(k\)-generalized Fibonacci numbers and powers of 3. Int. J. Number Theory 16(7), 1643–1666 (2020)
Ddamulira, M., Luca, F., Rakotomalala, M.: On a problem of Pillai with Fibonacci numbers and powers of 2. Proc. Indian Acad. Sci. Math. Sci. 127(3), 411–421 (2017)
Ddamulira, M., Gómez, C.A., Luca, F.: On a problem of Pillai with \(k\)-generalized Fibonacci numbers and powers of 2. Monatsh. Math. 187(4), 635–664 (2018)
de Weger, B.M.M.: Solving exponential Diophantine equations using lattice basis reduction algorithms. J. Number Theory 26(3), 325–367 (1987)
Elsholtz, C., Luca, F., Planitzer, S.: Romanov type problems. Ramanujan J. 47(2), 267–289 (2018)
Gúzman Sanchez, S., Luca, F.: Linear combinations of factorials and \(S\)-units in a binary recurrence sequence. Ann. Math. Qué. 38(2), 169–188 (2014)
Hernane, M.O., Luca, F., Rihane, S.E., Togbé, A.: On Pillai’s problem with Pell numbers and powers of 2. Hardy-Ramanujan J. 41, 22–31 (2018)
Hernández, S.H., Luca, F., Rivera, L.M.: On Pillai’s problem with the Fibonacci and Pell sequences. Bol. Soc. Mat. Mex. (3) 25(3), 495–507 (2019)
Herschfeld, A.: The equation \(2^x- 3^y = d\). Bull. Am. Math. Soc. 41, 631 (1935)
Herschfeld, A.: The equation \(2^x-3^y=d\). Bull. Am. Math. Soc. 42, 231–234 (1936)
Luca, F., Noubissie, A.: Linear combinations of factorial and \(s\)-unit in a ternary recurrence sequence with a double root. Period. Math. Hung. 86, 422–441 (2022)
Matveev, E.M.: An explicit lower bound for a homogeneous rational linear form in logarithms of algebraic numbers. II. Izv. Ross. Akad. Nauk Ser. Mat. 64(6), 125–180 (2000)
Mihăilescu, P.: Primary cyclotomic units and a proof of Catalan’s conjecture. J. Reine Angew. Math. 572, 167–195 (2004)
Pillai, S.S.: On \(a^x-b^y=c\). J. Indian Math. Soc. (N.S.) 2, 119–122 (1936)
Pillai, S.S.: A correction to the paper “On \(a^x - b^y=c\)’’. J. Indian Math. Soc. New Ser. 2, 215 (1937)
Smart, N.P.: The Algorithmic Resolution of Diophantine Equations. London Mathematical Society Student Texts, vol. 41. Cambridge University Press, Cambridge (1998)
Yu, K.: \(p\)-adic logarithmic forms and group varieties. II. Acta Arith. 89(4), 337–378 (1999)
Funding
Open access funding provided by Paris Lodron University of Salzburg.
Author information
Authors and Affiliations
Contributions
All authors contributed equally to the manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare no 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
Faye, B., Luca, F. & Ziegler, V. On a variant of Pillai’s problem with factorials and S-units. Ramanujan J 63, 773–802 (2024). https://doi.org/10.1007/s11139-023-00787-1
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11139-023-00787-1