11111
-
\(\varphi \ast 1= \operatorname{id}\)
-
\(\mu \ast 1 = \varepsilon\)
-
\(\mu \ast \operatorname{id} = \varphi\),证明
-
\(\operatorname{id} \ast \operatorname{id}=1\),用法 \((\varphi\times \operatorname{id}) \ast \operatorname{id} = \operatorname{id}^2\) (即对 \(\varphi(i)\times i\) 进行杜教筛)好像假了 -
若 \(f\) 是积性函数,要用线性筛预处理处理 \(h(n) = \sum\limits_{d \mid n} f(d)*\mu(\frac{n}{d})\)
因为 \(h\) 是积性函数,并且线性筛枚举的都是最小质因子,所以\[\begin{aligned} h(n) &= \prod h(p_i^{k_i}) \\ &= \prod\sum\limits_{t=0}^{k_i} f(p_i^t)*\mu(p_i^{k_i-t}) \\ &= \prod f(p_i^{k_i})*\mu(1)+f(p_i^{k_i-1})*\mu(p_i) \\ &= \prod f(p_i^{k_i}) - f(p_i^{k_i-1}) \end{aligned}\]哦这个好像没什么用……但有的时候那个减法可以提出公因式变成乘法,那就有用了