欧拉筛法与积性函数

线性筛

欧拉筛本质上是在对每一个数找到它最小的质因数，然后把它筛除，复杂度 $O(n)$ 。
本文中所有的 $p$ 都表示一个质数。

PROOF：
From the code below we have that $p_j \mid a$.
Let $n =ap_k\ (k > j，p_k > p_j)$.
$\because p_j \mid a$，记 $a = bp_j$.
$\therefore n = (bp_j)p_k = p_j(bp_k)$.
$\therefore p_j \mid n\ (p_k >p_j，bp_k > p_j)$
Q.E.D.

#include <cstdio>
#define R register
const int MaxN = 1000000
bool vis[MaxN];
int P[MaxN], tot;
int main()
{
    for(R int i = 2; i <= MaxN; i++)
    {
        if(!vis[i]) P[++tot] = i, phi[i] = P[tot] - 1;
        for(R int j = 1; j <= tot && P[j] * i <= MaxN; j++)
        {
            vis[i * P[j]] = 1;
            if(i % P[j] == 0) break;
        }
    }
    return 0;
}

---

欧拉函数 $\varphi$

$1 ... n$ 中与 $n$ 互质的整数个数。
欧拉函数：

$$\varphi(n)=n\prod_{p \mid n}(1 - \frac{1}{p})$$

Euler Theorem：$a^{\varphi(n)}\equiv 1 (mod\ n)$, if $(a,n) = 1$.

Theorem 1：$\sum_{d \mid n} \varphi(d) = n$.

PROOF
Let $n = p_1^{c_1}p_2^{c_2}...p_m^{c_m}$.
$\therefore \sum_{d \mid n} \varphi(d) = \sum_{d \mid n} \varphi(p_1^{k_1}p_2^{k_2}...p_m^{k_m}),(0 \le k_i \le c_i)$.
$\because \varphi(p_i^{c_i}p_j^{c_j}) = \varphi(p_i^{c_i})\varphi(p_j^{c_j})$.
$\therefore \sum_{d \mid n} \varphi(d) = \sum_{d \mid n} \varphi(p_1^{k_1})\varphi(p_2^{k_2})...\varphi(p_m^{k_m})=\prod_{i=1}^m\sum_{k=0}^{c_i}\varphi(p_i^k)$.
$\because \varphi(n) = n \prod_{p \mid n} (1 - \frac{1}{p})$.
$\therefore \sum_{d \mid n} \varphi(d) = \prod_{i=1}^m\sum_{k=0}^{c_i}p_i^{k-1}(p_i-1)$.
$\therefore \sum_{d \mid n} \varphi(d) =\prod_{i=1}^m[(p_i-1)(p^0+p^1+...+p^{c_i})+1]=\prod_{i=1}^mp_i^{c_i}=n$.
Q.E.D.

欧拉筛解决欧拉函数，$O(n)$。
由前文可知，欧拉筛可以让我们快速的找出每一个数的最小质因数。
由于欧拉函数是积性函数，所以我们只需要求出 $\varphi(p^k)$ ，即有 $\varphi(n) = \varphi(n) \varphi(\frac{n}{p^k})$

PROOF：
$\because \varphi(n) = n \prod_{p \mid n} (1 - \frac{1}{p})$.
$\therefore \varphi(p^k) = (p - 1)p^k$.
Let $n = ap_j$，while $p_j$ is the minimun prime factor of $n$.
We conclude that

$$\varphi(n) = \begin{cases} \varphi(a) \times (p_j-1)\ ,\ p_j \nmid a, \\ \varphi(a) \times p_j\ ,\ p_j \mid a,\\ p_j - 1\ ,\ n \mathrm{\ is\ a\ prime}.\end{cases}$$
Q.E.D.

#include <cstdio>
#define R register
bool vis[1000010];
int P[1000010], tot, phi[1000010];
int main()
{
    for(R int i = 2; i <= 1000000; i++)
    {
        if(!vis[i]) P[++tot] = i, phi[i] = P[tot] - 1;
        for(R int j = 1; j <= tot && P[j] * i <= 1000000; j++)
        {
            vis[i * P[j]] = 1;
            if(i % P[j] == 0) 
            {
                phi[i * P[j]] = phi[i] * P[j];
                break;
            }
            else phi[i * P[j]] = phi[i] * (P[j] - 1);
        }
    }
    return 0;
}

---

莫比乌斯反演 $\mu$

欧拉筛解决莫比乌斯函数，$O(n)$。
莫比乌斯函数：

$$\mu(n) = \begin {cases} 1\ ,\ n = 1 \\ (-1)^k\ ,\ n = p_1p_2...p_k \\ 0 \ ,\ \mathrm{others}\end{cases}$$
莫比乌斯函数可以用于解决形如： $$g(n) = \sum_{d \mid n} f(d)$$ $$f(n) = \sum_{d|n} \mu(d) g(\frac{n}{d})$$
同理，莫比乌斯函数也是一个积性函数，所以我们只需快速地求出 $\mu(p^k)$ 即可。

PROOF：
From the defination of Möbius inversion formula we have that iff $k = 1$，$\mu(p^k) = -1$，otherwise $k=0$.
We conclude that

$$\mu(n) = \begin{cases} -\mu(a)\ ,\ p_j \nmid a, \\0\ ,\ p_j \mid a,\\ -1\ ,\ n \mathrm{\ is\ a\ prime}.\end{cases}$$
Q.E.D.

#include <cstdio>
#define R register
bool vis[1000010];
int P[1000010], tot, mu[1000010];
int main()
{
    mu[1] = 1;
    for(R int i = 2; i <= 1000000; i++)
    {
        if(!vis[i]) P[++tot] = i, mu[i] = -1;
        for(R int j = 1; j <= tot && P[j] * i <= 1000000; j++)
        {
            vis[i * P[j]] = 1;
            if(i % P[j] == 0) 
            {
                mu[i * P[j]] = 0;
                break;
            }
            else mu[i * P[j]] = -mu[i];
        }
    }
    return 0;
}

---

参考文献：
1.贾志鹏，《线性筛法与积性函数》。