1 Erratum to: Period Math Hung (2015) 71:11–23 https://doi.org/10.1007/s10998-014-0079-3

2 Replacement of [3, Proposition 4.3]

We first illustrate a counter-example for [3, Proposition 4.3]. Take \(m=273 = 3 \times 7 \times 13\). Using the notation in [3, Proposition 4.3], we have \(d=3\) and \(d^{\prime }=1\). That is, the homomorphism \(\phi _m\) defined there is surjective. However, for any positive integer a coprime to m, \(C_m(a)\) is divisible by 3, because \(6\mid \lambda (m)\) and then \(9\mid a^{\lambda (m)}-1\). This leads to a contradiction.

Proposition 4.3 and its proof in [3] should be replaced by Proposition 1.1 below. Fortunately, this does not affect other results and arguments in [3], although Proposition 4.3 in [3] was quoted several times there.

Assume that positive integer m has the prime factorization \(m=p_{1}^{r_{1}}\cdots p_{k}^{r_{k}}\). In [1, Proposition 4.4], the Euler quotient has been used to define a homomorphism from \((\mathbb {Z}/m^{2}\mathbb {Z})^{*}\) to \((\mathbb {Z}/m\mathbb {Z},+)\), whose image is \(d\mathbb {Z}/m\mathbb {Z}\), where

$$\begin{aligned} d=\prod _{i=1}^{k}d_{i} \quad \text {and} \quad d_{i}=\left\{ \begin{array}{ll} \gcd (p_{i}^{r_{i}}, 2\varphi (m)/\varphi (p_i^{r_i})) &{} \text {if}\, p_{i}=2\ \text {and}\ r_{i}\ge 2,\\ \gcd (p_{i}^{r_{i}}, \varphi (m)/\varphi (p_i^{r_i})) &{} \text {otherwise}.\\ \end{array} \right. \end{aligned}$$
(1.1)

In fact, the above \(d,d_i\) are equivalent to those \(d,d_i\) defined in [3], respectively.

By [3, Proposition 2.2 (2)], the Carmichael quotient \(C_{m}(x)\) induces a homomorphism

$$\begin{aligned} \phi _m:(\mathbb {Z}/m^{2}\mathbb {Z})^{*}\rightarrow (\mathbb {Z}/m\mathbb {Z},+), x\mapsto C_m(x), \end{aligned}$$

where \(C_m(x) = (x^{\lambda (m)}-1)/m\) and \(\lambda (m)\) is the Carmichael function.

Proposition 1.1

Let \(m=p_{1}^{r_{1}}\cdots p_{k}^{r_{k}}\) be the prime factorization of \(m \ge 2\). For \(1\le i\le k\), put

$$\begin{aligned} d^{\prime }_{i}=\left\{ \begin{array}{ll} \gcd (p_{i}^{r_{i}}, 2\lambda (m)/\lambda (p_i^{r_i})) &{}\quad \text {if}\, p_{i}=2\ \text {and}\ r_{i}= 2,\\ \gcd (p_{i}^{r_{i}}, \lambda (m)/\lambda (p_i^{r_i})) &{}\quad \text {otherwise}. \end{array} \right. \nonumber \end{aligned}$$

Let \(d^{\prime }=\prod _{i=1}^{k}d^{\prime }_{i}\). Then the image of the homomorphism \(\phi _m\) is \(d^{\prime }\mathbb {Z}/m\mathbb {Z}\).

Proof

We show the desired result case by case.

(I) First we prove the result for the case \(k=1\), that is \(m=p^r\), where p is a prime and r is a positive integer.

Suppose that \(p=2\). If \(r=2\), then \(C_m(3)=2\), and for any positive integer n we have \(C_m(2n+1)= n(n+1)\), which is even, so the image of \(\phi _m\) is \(2{\mathbb Z}/m{\mathbb Z}\). On the other hand, if \(r = 1\) or \(r \ge 3\), since \(C_2(3)=1\) and \(C_8(3)=1\), by using [3, Proposition 2.8] we see that \(C_m(3)\) is an odd integer, so the image of \(\phi _m\) is \({\mathbb Z}/m{\mathbb Z}\).

Now, assume that \(p>2\). Note that \(C_p(p+1) \equiv -1 \pmod p\), by [3, Proposition 2.8] we have \(C_m(p+1) \equiv -1 \pmod p\), which implies that \(p \not \mid C_m(p+1)\). Thus, there exists a positive integer n such that \(n C_m(p+1) \equiv 1 \pmod m\). Then, by [3, Proposition 2.2 (1)] we deduce that \(C_m((p+1)^n) \equiv 1 \pmod m\). So, the image of \(\phi _m\) is \({\mathbb Z}/m{\mathbb Z}\).

(II) To complete the proof, we prove the result when \(k \ge 2\).

For simplicity, denote \(m_i = m/p_i^{r_i}\) and \(n_i = \lambda (m) / \lambda (p_i^{r_i})\) for each \(1\le i \le k\), and then let \(m_i^{\prime }\) be an integer such that \(m_i^{2}m_i^{\prime } \equiv 1 \pmod {p_i^{r_i}}\). By [3, Proposition 2.7], we have

$$\begin{aligned} C_m(a) \equiv \sum _{i=1}^{k} m_i m_i^{\prime } n_i C_{p_i^{r_i}}(a) \pmod {m}. \end{aligned}$$
(1.2)

So, for each \(1\le i \le k\), \(C_m(a) \equiv m_i m_i^{\prime } n_i C_{p_i^{r_i}}(a) \pmod {p_i^{r_i}}\). If \(p_i=2\) and \(r_i=2\), note that for any odd integer \(a>1\), \(C_4(a)\) is even, then we see that \(d_i^{\prime } \mid n_i C_{p_i^{r_i}}(a)\), and thus \(d_i^{\prime } \mid C_m(a)\). Otherwise if \(p_i>2\) or \(r_i \ne 2\), then \(d_i^{\prime } \mid n_i\), and so \(d_i^{\prime } \mid C_m(a)\). Hence, we have \(d^{\prime } \mid C_m(a)\) for any integer a coprime to m.

Let \(b = \gcd (m, m_1m_1^{\prime }n_1, \ldots , m_km_k^{\prime }n_k)\). Then, there exist integers \(X_1,\ldots , X_k\) such that

$$\begin{aligned} b \equiv \sum _{i=1}^{k} m_i m_i^{\prime } n_i X_i \pmod {m}. \end{aligned}$$
(1.3)

If we denote \(b_i = \gcd (p_i^{r_i}, m_i m_i^{\prime }n_i)\) for each \(1 \le i \le k\), then \(b = \prod _{i=1}^{k} b_i\); here, we remark that \(b_i = \gcd (p_i^{r_i}, n_i)\). It is easy to see that for each \(1 \le i \le k\), if \(p_i>2\) or \(r_i \ne 2\), we have \(d_i^{\prime } = b_i\). Further, when \(p_i=2\) and \(r_i=2\), \(d_i^{\prime } = 2b_i\) if \(8 \not \mid \lambda (2p_1\ldots p_k)\), and \(d_i^{\prime } = b_i\) otherwise.

We now have three cases for m:

  1. (i)

    There exists \(1\le j \le k\) such that \(p_j=2, r_j=2\) and

    $$\begin{aligned} 8 \not \mid \lambda (2p_1\ldots p_k). \end{aligned}$$
  2. (ii)

    There exists \(1\le j \le k\) such that \(p_j=2, r_j=2\) and

    $$\begin{aligned} 8 \mid \lambda (2p_1\ldots p_k). \end{aligned}$$
  3. (iii)

    All the other cases.

Clearly, in Cases (ii) and (iii) we have \(d^{\prime }=b\), and in Case (i) \(d^{\prime } = 2b\).

According to (I), there exist integers \(a_i\) with \(p_i \not \mid a_i\) for \(1 \le i \le k\) defined by

$$\begin{aligned} C_{p_i^{r_i}}(a_{i}) \equiv \left\{ \begin{array}{ll} 2X_i &{} \text {in Case (i)}, \\ X_i &{} \text {in Case (iii)}, \\ &{} \qquad \qquad \qquad \qquad \qquad \pmod {p_i^{r_i}} \\ X_i &{} \text {in Case (ii)}\ \text {and}\ i\ne j, \\ 0 &{} \text {in Case (ii)}\ \text {and}\ i=j. \end{array} \right. \end{aligned}$$

By the Chinese Remainder Theorem, we can choose a positive integer a such that \(a \equiv a_i \pmod {p_i^{2r_i}}\). So, by [3, Proposition 2.2 (2)] we have \(C_{p_i^{r_i}}(a) \equiv C_{p_i^{r_i}}(a_i) \pmod {p_i^{r_i}}\). Then, combining with (1.3) and the relation between b and \(d^{\prime }\), we obtain \(m_im_i^{\prime }n_i C_{p_i^{r_i}}(a) \equiv d^{\prime } \pmod {p_i^{r_i}}\) for each \(1\le i \le k\) in all the three cases. Finally, using (1.2) we have \(C_m(a) \equiv d^{\prime } \pmod {m}\), which completes the proof. \(\square \)

Comparing (1.1) with Proposition 1.1, we have \(d^{\prime } \mid d\). Moreover, by [3, Proposition 2.1] we get

$$\begin{aligned} \frac{\varphi (m)}{\lambda (m)}d^{\prime }{\mathbb Z}/m{\mathbb Z}= d{\mathbb Z}/m{\mathbb Z}, \end{aligned}$$

which implies that \(\gcd (\frac{\varphi (m)}{\lambda (m)}d^{\prime },m)=d\).

3 Another error

We take this opportunity to correct another error. In the proof of [3, Lemma 3.4], the last identity “\(\equiv \ell n^{-1} 2^{r-2}\)” may be not true, and it should be deleted. Because by using \(n^{2^{r-2}} \equiv 1 \pmod {2^r}\), we only know that \((n^{2^{r-2}}+1)/2\) is an odd integer, which may be not congruent to 1 modulo \(2^r\). Clearly, this error does not change the result there.