On the $3$-adic valuation of the Narayana numbers

The arithmetic properties of the linear recurrence sequences are one of the topics in number theory that are extensively studied. One example concerns about the Fibonacci sequence $\{F_{n}\}_{n\geq 0}$ defined by $F_{n}=F_{n-1}+F_{n-2}$ for $n\geq 2$, with initial values $F_{0}=0$ and $F_{1}=1$. The first few terms of this sequence are

$$0,1,1,2,3,5,8,13,21,34,55,89,144,233,377,610,987,\ldots$$

For a prime $p$, the $p$-adic valuation (or the $p$-adic order) $\nu_{p}(r)$ is the exponent of the highest power of $p$ which divides $r$. The $p$-adic valuation of the Fibonacci numbers was fully characterized (see [6, 9, 10, 15, 17]). In particular, Lengyel [9] showed that for $n\geq 1$,

$$\nu_{2}(F_{n})=\begin{cases}0,&\text{ if }n\equiv 1,2\pmod{3};\\ 1,&\text{ if }n\equiv 3\pmod{6};\\ 3,&\text{ if }n\equiv 6\pmod{12};\\ \nu_{2}(n)+2,&\text{ if }n\equiv 0\pmod{12}\\ \end{cases}$$

using congruence properties involving $F_{n}$ (see [7, 14]). However, the behavior of the $p$-adic valuation of linear recurrence sequences of higher order is much less known. A particular case involves a well-known generalization of the Fibonacci numbers, the Tribonacci sequence $\{T_{n}\}_{n\geq 0}$ defined by $T_{n}=T_{n-1}+T_{n-2}+T_{n-3}$ for $n\geq 3$, with initial values $T_{0}=0$ and $T_{1}=T_{2}=1$. The first few terms of this sequence are

$$0,1,1,2,4,7,13,24,44,81,149,274,504,927,1705,\ldots$$

In 2014, Marques and Lengyel [13] proved that for $n\geq 1$,

$$\nu_{2}(T_{n})=\begin{cases}0,&\text{ if }n\equiv 1,2\pmod{4};\\ 1,&\text{ if }n\equiv 3,11\pmod{16};\\ 2,&\text{ if }n\equiv 4,8\pmod{16};\\ 3,&\text{ if }n\equiv 7\pmod{16};\\ \nu_{2}(n)-1,&\text{ if }n\equiv 0\pmod{16};\\ \nu_{2}(n+4)-1,&\text{ if }n\equiv 12\pmod{16};\\ \nu_{2}((n+1)(n+17))-3,&\text{ if }n\equiv 15\pmod{16}\end{cases}$$

and used the $2$-adic valuation of $T_{n}$ to show that $1,2$ and $24$ are the only Tribonacci numbers that are factorials. Since then, several authors have worked on the $2$-adic valuation of the generalized Fibonacci numbers (see [5, 11, 16, 18]). Another example is about the Tripell sequence $\{t_{n}\}_{n\geq 0}$ defined by $t_{n}=2t_{n-1}+t_{n-2}+t_{n-3}$ for $n\geq 3$, with initial values $t_{0}=0,t_{1}=1$ and $t_{2}=2$. The first few terms of this sequence are

$$0,1,2,5,13,33,84,214,545,1388,3535,9003,22929,58396,\ldots$$

In 2020, Bravo, Díaz, and Ramírez [4] proved that for $n\geq 1$,

$$\nu_{3}(t_{n})=\begin{cases}0,&\text{ if }n\equiv 1,2,3,4\pmod{6};\\ \nu_{3}(n),&\text{ if }n\equiv 0\pmod{6};\\ \nu_{3}(n+1),&\text{ if }n\equiv 5\pmod{6}\end{cases}$$

and applied the $3$-adic valuation of $t_{n}$ to show that $1$ and $2$ are the only Tripell numbers that are factorials.

Recall that the Narayana sequence $\{a_{n}\}_{n\geq 0}$ is defined by $a_{n}=a_{n-1}+a_{n-3}$ for $n\geq 3$, with initial values $a_{0}=0$ and $a_{1}=a_{2}=1$. The first few terms of this sequence are

$$0,1,1,1,2,3,4,6,9,13,19,28,41,60,88,129,189,277,\ldots$$

Some properties of Narayana numbers and its generalizations can be found in [1, 2, 3].

In this paper, we apply Zhou’s [19] method of constructing identities of linear recurrence sequences to deduce several congruence properties involving $a_{n}$ and to prove our main result, which fully describes the $3$-adic valuation of $a_{n}$.

One application of the $p$-adic valuation of linear recurrence sequences is to give upper bounds of the solutions of the Diophantine equations involving factorials and these sequences. In this paper, we use Theorem 1.1 to determine all Narayana numbers that are factorials, as shown by the following result.

In this section, we present some preliminary results that are used in the proofs of our main theorems. We begin with the bounds of the $p$-adic valuation of a factorial, which is a consequence of Legendre’s formula (see [8]).

We next have the exponential growth of the Narayana sequence $\{a_{n}\}$.

We finally have the following identity involving $a_{n}$, which plays a key role in the proofs of Theorem 1.1 and Theorem 1.2. We apply the method of constructing identities of linear recurrence sequences introduced by Zhou [19] to prove this identity.

We first present the following congruence property of the Narayana numbers.

We are now in a position to prove Theorem 1.1 by working on each case separately.

1. (a)

   Suppose that $n\equiv k\pmod{8}$ with $k\in\{1,2,3,4,6\}$. We note that the sequence $\{a_{n}\pmod{3}\}_{n\geq 0}$ is periodic with period $8$, so $a_{n}\equiv a_{k}\pmod{3}$. By routine calculation, we have $a_{k}\not\equiv 0\pmod{3}$ for all $k\in\{1,2,3,4,6\}$, so $\nu_{3}(a_{n})=0$.

2. (b)

   Suppose that $n\equiv k\pmod{24}$ with $k\in\{5,7,13,15\}$. We note that the sequence $\{a_{n}\pmod{9}\}_{n\geq 0}$ is periodic with period $24$, so $a_{n}\equiv a_{k}\pmod{9}$. By routine calculation, we have $a_{5}=3,a_{7}=6,a_{13}=60$ and $a_{15}=129$, all of which are divisible by $3$ but not by $9$. Thus, we have $\nu_{3}(a_{n})=1$.

3. (c)

   Suppose $n\equiv 8\pmod{24}$. Then $n=8s\cdot 3^{m}+8$ for some integers $m,s\geq 1$ with $3\nmid s$. Using the recurrence formula for $a_{n}$ and Proposition 3.1, we deduce that $a_{n}\equiv 9\pmod{3^{m+3}}$. Thus, we get $\nu_{3}(a_{n})=2$.

4. (d)

   Suppose $n\equiv 23\pmod{24}$. Then $n=8s\cdot 3^{m}-1$ for some integers $m,s\geq 1$ with $3\nmid s$, so that $\nu_{3}(n+1)=m$. Using the recurrence formula for $a_{n}$ and Proposition 3.1, we deduce that $a_{n}\equiv-s\cdot 3^{m+1}\pmod{3^{m+3}}$. Thus, we get $\nu_{3}(a_{n})=m+1=\nu_{3}(n+1)+1$.

5. (e)

   Suppose $n\equiv 21\pmod{24}$. Then $n=8s\cdot 3^{m}-3$ for some integers $m,s\geq 1$ with $3\nmid s$, so that $\nu_{3}(n+3)=m$. Using the recurrence formula for $a_{n}$ and Proposition 3.1, we deduce that $a_{n}\equiv(2s\cdot 3^{m+2}+s\cdot 3^{m+1})\pmod{3^{m+3}}$. Thus, we get $\nu_{3}(a_{n})=m+1=\nu_{3}(n+3)+1$.

6. (f)

   Suppose $n\equiv 0\pmod{24}$. Then $n=8s\cdot 3^{m}$ for some integers $m,s\geq 1$ with $3\nmid s$, so that $\nu_{3}(n)=m$. By Proposition 3.1, we have $a_{n}\equiv 2s\cdot 3^{m+2}\pmod{3^{m+3}}$. Thus, we get $\nu_{3}(a_{n})=m+2=\nu_{3}(n)+2$.

7. (g)

   Suppose $n\equiv 16\pmod{24}$. Then $n=8s\cdot 3^{m}-8$ for some integers $m,s\geq 1$ with $3\nmid s$, so that $\nu_{3}(n+8)=m$. Using the recurrence formula for $a_{n}$ and Proposition 3.1, we deduce that $a_{n}\equiv-2s\cdot 3^{m+2}\pmod{3^{m+3}}$. Thus, we get $\nu_{3}(a_{n})=m+2=\nu_{3}(n+8)+2$.

This completes the proof of Theorem 1.1.

We note that if $m\leq 5$, then the only solutions are the ones listed in Theorem 1.2, so we now suppose that $m\geq 6$. Applying Theorem 1.1 and Lemma 2.1 with $p=3$, we obtain

$\displaystyle\dfrac{m}{2}-\left\lfloor\dfrac{\log m}{\log 3}\right\rfloor-1$

$\displaystyle\leq\nu_{3}(a_{n})\leq\nu_{3}(n(n+1)(n+3)(n+8))+6\leq 4\nu_{3}(n+\delta)+6,$

for some $\delta\in\{0,1,3,8\}$. This implies that

$$3^{\lfloor m/8-\lfloor(\log m)/(\log 3)\rfloor/4-7/4\rfloor}\leq 3^{\nu_{3}(n+\delta)}\leq n+\delta\leq n+8$$

and taking logarithms leads to

(2)

$\displaystyle\left\lfloor\dfrac{m}{8}-\dfrac{1}{4}\left\lfloor\dfrac{\log m}{\log 3}\right\rfloor-\dfrac{7}{4}\right\rfloor\leq\dfrac{\log(n+8)}{\log 3}.$

From Lemma 2.2, we have $\alpha^{n-3}\leq a_{n}=m!<(m/2)^{m}$, so that $n<3+m\log(m/2)/\log\alpha$. Plugging this in eq. 2 yields

$$\left\lfloor\dfrac{m}{8}-\dfrac{1}{4}\left\lfloor\dfrac{\log m}{\log 3}\right\rfloor-\dfrac{7}{4}\right\rfloor\leq\dfrac{\log(11+m\log(m/2)/\log\alpha)}{\log 3}.$$

Thus, we deduce that $m\leq 68$ and $n\leq 630$. A simple computational search using Mathematica shows that there are no solutions in the range $m\in[6,68]$ and $n\in[1,630]$. This completes the proof of Theorem 1.2.