Hostname: page-component-848d4c4894-m9kch Total loading time: 0 Render date: 2024-05-27T07:47:27.892Z Has data issue: false hasContentIssue false


Published online by Cambridge University Press:  12 May 2023

Department of Economics, Università degli Studi dell’Insubria, via Monte Generoso 71, 21100 Varese, Italy
School of Mathematics, University of the Witwatersrand, Private Bag 3, Wits 2050, South Africa e-mail:
Rights & Permissions [Opens in a new window]


Let $\varphi $ be Euler’s function and fix an integer $k\ge 0$. We show that for every initial value $x_1\ge 1$, the sequence of positive integers $(x_n)_{n\ge 1}$ defined by $x_{n+1}=\varphi (x_n)+k$ for all $n\ge 1$ is eventually periodic. Similarly, for all initial values $x_1,x_2\ge 1$, the sequence of positive integers $(x_n)_{n\ge 1}$ defined by $x_{n+2}=\varphi (x_{n+1})+\varphi (x_n)+k$ for all $n\ge 1$ is eventually periodic, provided that k is even.

Research Article
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (, which permits unrestricted re-use, distribution, and reproduction in any medium, provided the original work is properly cited.
© The Author(s), 2023. Published by Cambridge University Press on behalf of Australian Mathematical Publishing Association Inc.

1 Introduction and main results

Let $\mathbb {N}$ be the set of positive integers and fix an arithmetic function $f: \mathbb {N}^d\to \mathbb {N}$ for some $d \in \mathbb {N}$ . Let $(x_n)_{n\ge 1}$ be a sequence of positive integers which satisfies the recurrence

(1.1) $$ \begin{align} x_{n+d}=f(x_n,\ldots,x_{n+d-2},x_{n+d-1}) \quad \text{for all }n \in \mathbb{N}, \end{align} $$

with starting values $x_1,\ldots ,x_d\in \mathbb {N}$ . In the case where $d=1$ , the sequence $(x_n)_{n\ge 1}$ is simply the orbit of $x_1$ with respect to f. The aim of this note is to study whether certain recurrence sequences $(x_n)_{n\ge 1}$ of the type (1.1) are eventually periodic independent of their starting values, that is, for all $x_1,\ldots ,x_d \in \mathbb {N}$ , there exists $T \in \mathbb {N}$ such that $x_n=x_{n+T}$ for all sufficiently large n.

We will frequently use the basic observation that a recurrence sequence $(x_n)_{n\ge 1}$ , as in (1.1), is eventually periodic if and only if it is bounded (see [Reference Everest, van der Poorten, Shparlinski and Ward5, page 45]).

We start with a simple result for functions f which are not too large.

Proposition 1.1. Let $f: \mathbb {N}^d \to \mathbb {N}$ be an arithmetical function, with $d \in \mathbb {N}$ , and suppose that there exists $C \in \mathbb {N}$ such that

(1.2) $$ \begin{align} f(n_1,\ldots,n_d) < \max\{n_1,\ldots,n_d\} \end{align} $$

for all $n_1,\ldots ,n_d \in \mathbb {N}$ with $n_i \ge C$ for some $i\in \{1,\ldots ,d\}$ . Let $(x_n)_{n\ge 1}$ be a recurrence sequence as in (1.1), with starting values $x_1,\ldots ,x_d \in \mathbb {N}$ . Then

(1.3) $$ \begin{align} \limsup_{n\to \infty} x_n \le \max \{f(n_1,\ldots,n_d): n_1,\ldots,n_d\le C-1\}. \end{align} $$

In particular, $(x_n)_{n\ge 1}$ is bounded above and hence eventually periodic.

Special instances of Proposition 1.1 in the one-dimensional case $d=1$ have been previously obtained in the literature. For example, Porges [Reference Porges15] considered the case where $f(n)$ is the sum of squares of the digits of n (see also [Reference Isaacs7, Reference Lorenzini, Melistas, Suresh, Suwama and Wang11, Reference Stewart19]). Note that (1.2) holds if $f(n_1,\ldots ,n_d)=o(N)$ as $N=\max \{n_1,\ldots ,n_d\}\to \infty $ .

Of course, there exist other functions f which do not satisfy (1.2) and such that every sequence $(x_n)_{n\ge 1}$ , as in (1.1), is eventually periodic: as a trivial example, one can consider $f(n_1,\ldots ,n_d):=\max \{n_1,\ldots ,n_d\}$ for all $(n_1,\ldots ,n_d) \in \mathbb {N}^d$ . At the opposite extreme, if f is slightly bigger (for instance, $f(n_1,\ldots ,n_d):=\max \{n_1,\ldots ,n_d\}+1$ ), then there are no eventually periodic sequences $(x_n)_{n\ge 1}$ as in (1.1). This is the starting point for this work, which motivates the heuristic: if a function f satisfies (1.2) ‘on average’, then every sequence $(x_{n})_{n\ge 1}$ as in (1.1) should be eventually periodic, independent of its starting values.

We are going to confirm the above heuristic in two cases which involve (shifted iterates of) Euler’s function $\varphi $ (recall that $\varphi (n)$ is the number of integers in $\{1,\ldots ,n\}$ which are coprime with n). Our first main result is the following theorem.

Theorem 1.2. Fix an integer $k\ge 0$ and let $(x_n)_{n\ge 1}$ be the recurrence sequence defined by

$$ \begin{align*} x_{n+1}=\varphi(x_n)+k \quad \text{for all }n \in \mathbb{N}, \end{align*} $$

with starting value $x_1 \in \mathbb {N}$ . Then,

(1.4) $$ \begin{align} \sup_{n \in \mathbb{N}} x_n \le \max\{x_1,k^4\}+(k+1)^2. \end{align} $$

In particular, $(x_n)_{n\ge 1}$ is eventually periodic.

Note that the dependence of the upper bound (1.4) on $x_1$ cannot be removed: indeed, if $k:=x_1-\varphi (x_1)$ for some $x_1\in \mathbb {N}$ , then the sequence $(x_{n})_{n\ge 1}$ is constantly equal to $x_1$ .

The trivial case $k=0$ in Theorem 1.2 has been already considered in the literature from different viewpoints (and, of course, it follows by Proposition 1.1 since $\varphi (n)\le n-1$ for all $n\ge 2$ ). Indeed, given a starting value $x_1 \in \mathbb {N}$ , then $x_{n+1}=\varphi ^{(n)}(x_1)$ for all $n \in \mathbb {N}$ , where $\varphi ^{(m)}$ is the m-fold iteration of $\varphi $ . For instance, Pillai [Reference Pillai12] showed that

$$ \begin{align*} \bigg\lfloor \frac{\log x_1-\log 2}{\log 3}\bigg\rfloor+1 \le N(x_1) \le \bigg\lfloor \frac{\log x_1}{\log 2}\bigg\rfloor+1 \quad \text{for all }x_1 \in \mathbb{N}, \end{align*} $$

where $N(x_1)$ is the minimal integer n for which $x_n=1$ (see also [Reference Shapiro17]) and it has been conjectured by Erdős et al. [Reference Erdős, Granville, Pomerance, Spiro, Berndt, Diamond, Halberstam and Hildebrand4] that $N(x_1) \sim \alpha \log x_1$ as $x_1\to \infty $ , for some $\alpha \in \mathbb {R}$ . It is known that the understanding of the multiplicative structure of $\varphi $ and its iterates is, in some sense, equivalent to the study of the behaviour of the integers of the form $p-1$ , where p is a prime. See also [Reference Lamzouri8, Reference Loomis and Luca10, Reference Pollack14, Reference Shparlinski18] for related work.

However, if $k\ge 1$ , then the function $f(n):=\varphi (n)+k$ does not satisfy (1.2): indeed, $\varphi (p)=p-1$ for all primes p, and hence $f(p)\ge p$ . However, it is well known that

(1.5) $$ \begin{align} \frac{1}{n}\sum_{i=1}^n \varphi(i)=\frac{3}{\pi^2}n+O(\log n) \quad \text{as }n\to \infty \end{align} $$

(see, for example, [Reference De Koninck and Luca2, Theorem 8.6]). Hence, approximating roughly $f(n)$ with $cn+k$ , where $c=3/\pi ^2 \in (0,1)$ , we expect that (1.2) holds ‘on average’, which is the heuristic behind Theorem 1.2.

Our second main result is the following theorem.

Theorem 1.3. Fix an even integer $k\ge 0$ and let $(x_n)_{n\ge 1}$ be the recurrence sequence defined by

$$ \begin{align*} x_{n+2}=\varphi(x_{n+1})+\varphi(x_n)+k \quad \text{for all }n \in \mathbb{N}, \end{align*} $$

with starting values $x_1,x_2 \in \mathbb {N}$ . Then,

(1.6) $$ \begin{align} \sup_{n\in \mathbb{N}} x_n \le 4^{X^{3^{k+1}}}, \quad \text{where } X:=\frac{3x_1+5x_2+7k}{2}. \end{align} $$

In particular, $(x_n)_{n\ge 1}$ is eventually periodic.

The heuristic supporting Theorem 1.3 is similar: thanks to (1.5), the value $f(n,m):=\varphi (n)+\varphi (m)+k$ can be roughly bounded above by $2c\max \{n,m\}+k$ , which is definitely smaller than $\max \{n,m\}$ since $2c=6/\pi ^2 <1$ .

We end with an open question to check if d is sufficiently large, then there exist starting values $x_1,\ldots ,x_d \in \mathbb {N}$ such that the sequence $(x_n)_{n\ge 1}$ defined as in (1.1) with

$$ \begin{align*} f(n_1,\ldots,n_d)=\varphi(n_1)+\cdots+\varphi(n_d) \end{align*} $$

is not eventually periodic. In a sense, this is related to the open question known as Lehmer’s totient problem [Reference Lehmer9], which asks about the existence of a composite $q\ge 2$ such that $\varphi (q)$ divides $q-1$ : indeed, if $r:=(q-1)/\varphi (q)$ , $d=r$ and $x_1=\cdots =x_d=q$ , then the sequence $(x_n)_{n\ge 1}$ would be constant.

Lastly, as suggested by the referee, it would be interesting to check whether the upper estimates of our main results are, in some sense, sufficiently sharp and whether some nontrivial lower bounds on the size of $(x_n)_{n\ge 1}$ could be obtained. We do not have an answer to these questions. However, we suspect that they are quite difficult. Indeed, in the setting of Theorem 1.2, one could try to obtain a cycle with period $(x_1,\ldots ,x_T)$ made by all primes in arithmetic progression but the last one (so that $x_1=x_{T+1}=\varphi (x_T)+k$ ). A possible interpretation may amount to asking whether such examples exist with comparable sizes of T and k.

2 Proofs

Proof of Proposition 1.1.

Let $\mathcal {Q}$ be the set of d-uples $(n_1,\ldots ,n_d) \in \mathbb {N}^d$ with $n_i\le C-1$ for all $i \in \{1,\ldots ,d\}$ . Suppose that $(n_1,\ldots ,n_d) \in \mathbb {N}\setminus \mathcal {Q}$ . Hence, we can pick the largest index $i \in \{1,\ldots ,d\}$ such that $n_i=\max \{n_1,\ldots ,n_d\}$ . In particular, $n_i\ge C$ . We claim that there exists $m \in \mathbb {N}$ such that $(n_{m+1}, \ldots , n_{m+d}) \in \mathcal {Q}$ . In addition, if m is the least such integer, then $\max \{n_j,\ldots ,n_{j+d-1}\}$ is decreasing for $j \in \{1,\ldots ,m\}$ . To this aim, suppose for the sake of contradiction that the claim does not hold. By the standing hypothesis (1.2), we get $n_{d+1}=f(n_1,\ldots ,n_d)< n_{i}$ . Repeating this reasoning, we obtain $n_{d+j}< \max \{n_j,\ldots ,n_{d+j-1}\}= n_{i}$ for all $j\in \{1,\ldots ,i\}$ . Hence, $\max \{n_{i+1},\ldots ,n_{i+d}\}\le n_{i}~-~1$ . Proceeding similarly, it follows that

$$ \begin{align*} \max\{n_{i+(k-1)d+1},\ldots,n_{i+kd}\}\le n_{i}-k \quad \text{for all }k\in \mathbb{N}. \end{align*} $$

However, if $k=n_{i}+1-C$ , then $(n_{i+(k-1)d+1},\ldots ,n_{i+kd}) \in \mathcal {Q}$ , which proves the claim.

To complete the proof, fix starting values $x_1,\ldots ,x_d \in \mathbb {N}$ . By the above claim and the finiteness of $\mathcal {Q}$ , it follows that the sequence $(x_n)_{n\ge 1}$ is bounded above by the constant

$$ \begin{align*} \max(\{x_1,\ldots,x_d\}\cup \{f(n_1,\ldots,n_d): (n_1,\ldots,n_d) \in \mathcal{Q}\}) \end{align*} $$

and that the upper limit in (1.3) holds.

Proof of Theorem 1.2.

First, let us suppose $k \le 1$ and fix a starting value $x_1 \in \mathbb {N}$ . Then,

$$ \begin{align*} x_{n+1}=\varphi(x_n)+k \le \max\{1,x_n-1\}+1=\max\{2,x_n\} \end{align*} $$

for all $n \in \mathbb {N}$ , with the consequence that $x_n \le \max \{x_1,2\}$ for all $n \in \mathbb {N}$ .

Suppose hereafter that $k \ge 2$ . Note that for all $n,m \in \mathbb {N}$ ,

(2.1) $$ \begin{align} \begin{split} x_{n+m}&\le \max\{x_{n+m-1}-1,1\}+k \\ &= \max\{x_{n+m-1}+k-1,k+1\} \le \max\{x_n+m(k-1),k+1\}. \end{split} \end{align} $$

Let us suppose for the sake of contradiction that $(x_n)_{n\ge 1}$ is not bounded above. Hence, there exists a minimal $r_1 \in \mathbb {N}$ such that $x_{r_1} \ge k^4$ (in particular, $x_{r_1}>4$ ).

Claim 2.1. There exists $i \in \{1,\ldots ,k\}$ such that $x_{r_1+i}<x_{r_1}$ .

Proof. Since $\varphi (n) \le n-\sqrt {n}$ whenever n is composite (by the fact that there exists a divisor of n which is at most $\sqrt {n}$ ), it follows that, if $x_{{r_1}+i-1}$ is composite for some $i \in \{1,\ldots ,k\}$ , then

$$ \begin{align*} x_{r_1+i}=\varphi(x_{r_1+i-1})+k \le x_{r_1+i-1}-\sqrt{x_{r_1+i-1}}+k. \end{align*} $$

Considering that the map $x\mapsto x-\sqrt {x}$ is increasing on $(4,\infty )$ and using (2.1), we obtain

$$ \begin{align*} \begin{split} x_{r_1+i} &\le x_{r_1}+(i-1)(k-1)-\sqrt{x_{r_1}+(i-1)(k-1)}+k\\ &\le x_{r_1}+(k-1)^2-\sqrt{k^4}+k < x_{r_1}. \end{split} \end{align*} $$

To conclude, we show that there exists some $i \in \{1,\ldots ,k\}$ for which $x_{{r_1}+i-1}$ is composite. Indeed, in the opposite case, these $x_{{r_1}+i-1}$ terms are all primes (and greater than k), and hence

$$ \begin{align*} x_{r_1+i-1}=x_{r_1}+(i-1)(k-1) \equiv x_{r_1}-(i-1) \bmod{k} \end{align*} $$

for all $i \in \{1,\ldots ,k\}$ . This is impossible, because there would exist $i \in \{1,\ldots ,k\}$ such that k divides $x_{{r_1}+i-1}$ .

At this point, (2.1) and Claim 2.1 imply that there exists a minimal $i_1 \in \{1,\ldots ,k\}$ such that $x_{r_1+i_1}< x_{r_1}$ ; and hence

(2.2) $$ \begin{align} \max\{x_1,\ldots,x_{r_1+i_1-1}\} \le x_{r_1}+(i_1-1)(k-1) < x_{r_1}+k^2. \end{align} $$

With the same reasoning, we can construct recursively sequences of positive integers $(r_n)$ and $(i_n)$ such that for all $n \in {\mathbb {N}}$ :

  1. (i) $r_{n+1}$ is the minimal integer such that $r_{n+1}\ge r_n+i_n$ and $x_{r_{n+1}}\ge k^4$ ;

  2. (ii) $i_{n+1}$ is the minimal integer in $\{1,\ldots ,k\}$ such that $x_{r_{n+1}+i_{n+1}}<x_{r_n}$ ; and hence,

    (2.3) $$ \begin{align} \max\{x_{r_n+i_n},\ldots,x_{r_{n+1}+i_{n+1}-1}\} \le x_{r_{n+1}}+(i_{n+1}-1)(k-1) < x_{r_{n+1}}+k^2. \end{align} $$

Lastly, note that

(2.4) $$ \begin{align} x_{r_{n}}=\varphi(x_{r_n-1})+k\le x_{r_n-1}+k \le k^4+k\le x_{r_1}+k \end{align} $$

for all $n \in \mathbb {N}$ . Using (2.2), (2.3) and (2.4), we conclude that

$$ \begin{align*} x_n < x_{r_1}+k^2+k\le \max\{x_1,k^4+k\}+k^2+k<\max\{x_1,k^4\}+(k+1)^2 \end{align*} $$

for all $n \in \mathbb {N}$ . This proves (1.4), concluding the proof.

Remark 2.2. A sketch of a shorter proof that the sequence $(x_n)_{n\ge 1}$ in Theorem 1.2 is eventually periodic goes as follows. Set $y_n:=\varphi (x_n)$ for all $n \in \mathbb {N}$ and note that $y_n\le C n$ with $C:=\max \{x_1,k\}$ . Hence, $\{y_1,\ldots ,y_n\}$ is contained in $V_n:=\varphi (\mathbb {N})\cap [1,Cn]$ . By a classical result of Pillai [Reference Pillai13] (see also [Reference Ford6] and references therein), we have $|V_n|=o(n)$ as $n\to \infty $ , and hence there exist distinct $i,j \in \mathbb {N}$ with $y_i=y_j$ . This implies that $x_{i+1}=x_{j+1}$ ; therefore, $(x_n)_{n\ge 1}$ is eventually periodic. However, this does not lead to an effective upper bound as in (1.4).

In the proof of Theorem 1.3, we will need also the effective version of the third Mertens’ theorem given by Rosser and Schoenfeld [Reference Rosser and Schoenfeld16] in 1962 (see also [Reference Axler1, Reference Dusart3]). As usual, hereafter, we reserve the letter p for primes.

Proposition 2.3. Let $\gamma :=\lim _n(\sum _{i\le n}1/i-\log n)=0.57721\ldots $ be the Euler– Mascheroni constant. Then the following inequality holds for all $x\ge 2{:}$

$$ \begin{align*} \frac{e^{-\gamma}}{\log x} \bigg(1- \frac{1}{\log^2 x} \bigg) < \prod_{p\le x} \bigg( 1 - \frac{1}{p} \bigg) < \dfrac{e^{-\gamma}}{\log x} \bigg(1+ \frac{1}{2\log^2 x} \bigg). \end{align*} $$

Proof. See [Reference Rosser and Schoenfeld16, Theorem 7 and its Corollary].

Corollary 2.4. If $x\ge 6$ , then

$$ \begin{align*} \prod_ {x<p\le x^3} \bigg( 1 - \frac{1}{p} \bigg)<\frac{1}{2}. \end{align*} $$

Proof. Thanks to Proposition 2.3, for each $r \ge 2$ , there exists $c_r \in (-1,\tfrac 12)$ such that

$$ \begin{align*} \prod_ {p\le r} \bigg( 1 - \frac{1}{p} \bigg) =\frac{e^{-\gamma}}{\log x} \bigg(1+ \frac{c_r}{\log^2 x} \bigg). \end{align*} $$

At this point, fix $x\ge 6$ . It follows that

$$ \begin{align*} \begin{split} \prod_ {x<p\le x^3} \bigg( 1 - \frac{1}{p} \bigg) &=\frac {\prod_{p \leq x^3} ( 1 - {1}/{p} )} {\prod_{p \leq x} ( 1 - {1}/{p} )}\\ &=\frac{e^{-\gamma}/(\log x^3)}{e^{-\gamma}/\log x}\cdot \frac{(1+c_{x^3}/(\log x^3)^2)}{1+c_x/(\log x)^2}\\ & < \frac{1}{3}\cdot \frac{1+0.5/ (9\log^2 x)}{1-1/(\log^2 x)} < \frac{1}{2}. \end{split} \end{align*} $$

Indeed, the last inequality is equivalent to

$$ \begin{align*} \frac{2}{3} \bigg(1+\frac{1}{18 \log^2 x}\bigg)< 1-\frac{1}{\log^2 x}, \end{align*} $$

which holds if and only if

$$ \begin{align*} x> e^{\sqrt{3(1+{1}/{27})}}. \end{align*} $$

The conclusion follows since the value of the right-hand side above is smaller than $6$ .

Proof of Theorem 1.3.

If $\max \{x_1,x_2\} \le 2$ and $k=0$ , then $x_n=2$ for all $n\ge 3$ , so the claimed (1.6) holds since $4^{X^{3^{k+1}}}\ge 4> x_n$ for all $n \in \mathbb {N}$ .

Suppose now that $\max \{x_1,x_2\} \ge 3$ or $k \ge 1$ , and note that ${X= \tfrac 12(3x_1+5x_2+7k)\ge 6}$ . Then, $\min \{x_3,x_4\}\ge 3$ ; and hence, $x_n$ is even for all $n\ge 5$ . In addition, since $\max \{x_1,\ldots ,x_6\}\le 2X$ and

$$ \begin{align*} 4^{X^{3^{k+1}}} \ge 4^X=2^{2X}>2X, \end{align*} $$

it follows that the claimed inequality holds for all $n\le 6$ .

Let $(p_n)_{n\ge 1}$ be the increasing enumeration of the primes greater than X. Since $\prod _{i=1}^n(1-{1}/{p_i})$ converges to zero as $n\to \infty $ by Proposition 2.3, one can find integers $1 =: r_0 < r_1 < \cdots < r_{k+1}$ such that

(2.5) $$ \begin{align} \prod_{i=r_j}^{r_{j+1} - 1} \bigg(1 - \frac{1}{p_i}\bigg) < \frac{1}{2} <\prod_{i=r_j}^{r_{j+1} - 2} \bigg(1 - \frac{1}{p_i}\bigg) \quad \text{for all }j \in \{0, 1, \ldots, k\}. \end{align} $$

Define also

$$ \begin{align*} q_j:=\prod_{i=r_j}^{r_{j+1} - 1}p_i \quad \text{for all }j \in \{0, 1, \ldots, k\}. \end{align*} $$

Since $\{q_0,q_1,\ldots ,q_k\}$ are pairwise coprime, the Chinese remainder theorem yields the existence of some $y \in \mathbb {N}$ such that $y\equiv j\bmod {q_j}$ for all $j \in \{0,1,\ldots ,k\}$ . In particular,

(2.6) $$ \begin{align} y \ge q_0 \ge p_{r_0}>X. \end{align} $$

Let us suppose for the sake of contradiction that $(x_n)_{n\ge 1}$ is not bounded. Then, there exists a minimal $v \in \mathbb {N}$ such that $x_v \ge 2y$ . Since $\max \{x_1,\ldots ,x_6\} \le 2X$ , it follows by (2.9) that $v\ge 7$ . In particular, $x_{v-1}$ and $x_{v-2}$ are even.

Claim 2.5. $\max \{\varphi (x_{v-1}),\varphi (x_{v-2})\} < y-k$ .

Proof. If $x_{v-1}<2y-2k$ then, by the fact that $x_{v-1}$ is even, $\varphi (x_{v-1}) \le \tfrac 12x_{v-1}<y-k$ . Otherwise, recalling that v is the minimal integer such that $x_v \ge 2y$ , then $2y-2k \le x_{v-1}<2y$ . In addition, $x_{v-1}$ is even, so $x_{v-1}=2y-2j$ for some $j \in \{1,\ldots ,k\}$ . It follows that

$$ \begin{align*} \varphi(x_{v-1})=x_{v-1} \, {\prod_{p\;\! \mid \;\! x_{v-1}}} \bigg(1 - \frac{1}{p}\bigg) = (y-j) {\prod_{p\;\! \mid \;\! x_{v-1}}}^{\!\!\prime} \bigg(1 - \frac{1}{p}\bigg), \end{align*} $$

where the last product is extended over the odd prime divisors of $x_{v-1}$ . Since, by construction, we have $y\equiv j\bmod {q_j}$ , we obtain by (2.8) that

$$ \begin{align*} \varphi(x_{v-1}) \le (y-j) \prod_{i=r_j}^{r_{j+1} - 1} \bigg(1 - \frac{1}{p_i}\bigg)< \frac{y-j}{2} < y-k. \end{align*} $$

Note that the last inequality holds because $y>2k$ , thanks to (2.9).

The same argument can be repeated for $x_{v-2}$ .

We conclude by Claim 2.5 that

$$ \begin{align*} 2y \le x_v = \varphi(x_{v-1}) + \varphi(x_{v-2}) + k < 2(y-k)+k \le 2y, \end{align*} $$

which is contradiction. It follows that $x_n<2y$ for all $n \in \mathbb {N}$ .

To complete the proof, it will be enough to show that $2y\le 4^{X^{3^{k+1}}}$ . For this, define $X_n:=X^{3^n}$ for all $n\ge 0$ and note that, thanks to Corollary 2.4, we have

$$ \begin{align*} \prod_{X_n<p\le X_n^3}\bigg(1-\frac{1}{p}\bigg)<\frac{1}{2} \quad \text{for all }n\ge 0. \end{align*} $$

By the definition of $r_j$ , it follows that

$$ \begin{align*} r_j \le X_j \quad \text{for all }j \in \{0,1,\ldots,k+1\}. \end{align*} $$

Lastly, since $\prod _{p\le x}p<4^x$ for all $x\ge 1$ (see, for example, [Reference De Koninck and Luca2, Lemma 2.8]), we conclude that

$$ \begin{align*} 2y \le 2\prod_{j=0}^k q_j \le \prod_{i=1}^{r_{k+1}}p_i \le \prod_{p\le X_{k+1}}p\le 4^{X_{k+1}}. \end{align*} $$

Therefore, $x_n \le 4^{X_{k+1}}$ for all $n \in \mathbb {N}$ .


The authors are thankful to the anonymous referee for a careful reading of the manuscript and for suggesting the open questions stated at the end of Section 1.


Axler, C., ‘New estimates for some functions defined over primes’, Integers 18 (2018), Article no. A52, 21 pages.Google Scholar
De Koninck, J.-M. and Luca, F., Analytic Number Theory: Exploring the Anatomy of Integers, Graduate Studies in Mathematics, 134 (American Mathematical Society, Providence, RI, 2012).CrossRefGoogle Scholar
Dusart, P., ‘Estimates of $\psi, \theta$ for large values of $x$ without the Riemann hypothesis’, Math. Comp. 85(298) (2016), 875888.CrossRefGoogle Scholar
Erdős, P., Granville, A., Pomerance, C. and Spiro, C., ‘On the normal behavior of the iterates of some arithmetic functions’, in: Analytic Number Theory (Allerton Park, IL, 1989) (eds. Berndt, B. C., Diamond, H. G., Halberstam, H. and Hildebrand, A.), Progress in Mathematics, 85 (Birkhäuser, Boston, MA, 1990), 165204.CrossRefGoogle Scholar
Everest, G., van der Poorten, A., Shparlinski, I. and Ward, T., Recurrence Sequences, Mathematical Surveys and Monographs, 104 (American Mathematical Society, Providence, RI, 2003).CrossRefGoogle Scholar
Ford, K., ‘The distribution of totients’, Ramanujan J. 2 (1998), 67151.CrossRefGoogle Scholar
Isaacs, R., ‘Iterates of fractional order’, Canad. J. Math. 2 (1950), 409416.CrossRefGoogle Scholar
Lamzouri, Y., ‘Smooth values of the iterates of the Euler phi-function’, Canad. J. Math. 59(1) (2007), 127147.CrossRefGoogle Scholar
Lehmer, D. H., ‘On Euler’s totient function’, Bull. Amer. Math. Soc. (N.S.) 38(10) (1932), 745751.CrossRefGoogle Scholar
Loomis, P. and Luca, F., ‘On totient abundant numbers’, Integers 8 (2008), Article no. A6, 7 pages.Google Scholar
Lorenzini, D., Melistas, M., Suresh, A., Suwama, M. and Wang, H., ‘Integer dynamics’, Int. J. Number Theory 18(2) (2022), 397415.CrossRefGoogle Scholar
Pillai, S. S., ‘On a function connected with $\phi (n)$ ’, Bull. Amer. Math. Soc. (N.S.) 35(6) (1929), 837841.CrossRefGoogle Scholar
Pillai, S. S., ‘On some functions connected with $\phi (n)$ ’, Bull. Amer. Math. Soc. (N.S.) 35(6) (1929), 832836.CrossRefGoogle Scholar
Pollack, P., ‘Two remarks on iterates of Euler’s totient function’, Arch. Math. (Basel) 97(5) (2011), 443452.CrossRefGoogle Scholar
Porges, A., ‘A set of eight numbers’, Amer. Math. Monthly 52 (1945), 379382.CrossRefGoogle Scholar
Rosser, J. B. and Schoenfeld, L., ‘Approximate formulas for some functions of prime numbers’, Illinois J. Math. 6 (1962), 6494.CrossRefGoogle Scholar
Shapiro, H., ‘An arithmetic function arising from the $\phi$ function’, Amer. Math. Monthly 50 (1943), 1830.Google Scholar
Shparlinski, I. E., ‘On the sum of iterations of the Euler function’, J. Integer Seq. 9(1) (2006), Article no. 06.1.6.Google Scholar
Stewart, B. M., ‘Sums of functions of digits’, Canad. J. Math. 12 (1960), 374389.CrossRefGoogle Scholar