[BZOJ2154]Crash的数字表格（莫比乌斯反演）

最新推荐文章于 2019-09-03 14:17:17 发布

xyz32768

最新推荐文章于 2019-09-03 14:17:17 发布

阅读量134

点赞数

分类专栏： BZOJ、UOJ、LOJ 文章标签：数论 Mobius反演线性筛

本文链接：https://blog.csdn.net/xyz32768/article/details/81358525

版权

BZOJ、UOJ、LOJ 专栏收录该内容

341 篇文章 1 订阅

订阅专栏

Address

https://www.lydsy.com/JudgeOnline/problem.php?id=2154

Solution

显然是莫比乌斯反演。

\sum i = 1 N \sum j = 1 M lcm (i, j)

$\sum_{i=1}^N\sum_{j=1}^M\text{lcm}(i,j)$

= \sum i = 1 N \sum j = 1 M i j gcd ( i , j )

$=\sum_{i=1}^N\sum_{j=1}^M\frac{ij}{\gcd(i,j)}$

\sum d \sum i = 1 N \sum j = 1 M i j d [gcd (i, j) = d]

$\sum_d\sum_{i=1}^N\sum_{j=1}^M\frac{ij}d[\gcd(i,j)=d]$

= \sum d \sum i = 1 ⌊ N d ⌋ \sum j = 1 ⌊ M d ⌋ d i j [gcd (i, j) = 1]

$=\sum_d\sum_{i=1}^{\lfloor\frac Nd\rfloor}\sum_{j=1}^{\lfloor\frac Md\rfloor}dij[\gcd(i,j)=1]$

= \sum d d \sum i = 1 ⌊ N d ⌋ \sum j = 1 ⌊ M d ⌋ i j \sum k | gcd (i, j) μ (k)

$=\sum_dd\sum_{i=1}^{\lfloor\frac Nd\rfloor}\sum_{j=1}^{\lfloor\frac Md\rfloor}ij\sum_{k|\gcd(i,j)}\mu(k)$

= \sum d d \sum k \sum i = 1, k | i ⌊ N d ⌋ \sum j = 1, k | j ⌊ M d ⌋ i j μ (k)

$=\sum_dd\sum_k\sum_{i=1,k|i}^{\lfloor\frac Nd\rfloor}\sum_{j=1,k|j}^{\lfloor\frac Md\rfloor}ij\mu(k)$

= \sum d d \sum k k 2 S (⌊ N d k ⌋) S (⌊ M d k ⌋) μ (k)

$=\sum_dd\sum_kk^2S(\lfloor\frac N{dk}\rfloor)S(\lfloor\frac M{dk}\rfloor)\mu(k)$
其中

S(x)=x(x+1)2 S ( x ) = x ( x + 1 ) 2 $S(x)=\frac{x(x+1)}2$ 。
发现下取整内的分母是乘积的形式，还是不好做。
所以设

T=dk T = d k $T=dk$ ：

\sum T \sum k | T T k k 2 S (⌊ N T ⌋) S (⌊ M T ⌋) μ (k)

$\sum_T\sum_{k|T}\frac Tkk^2S(\lfloor\frac NT\rfloor)S(\lfloor\frac MT\rfloor)\mu(k)$

= \sum T T S (⌊ N T ⌋) S (⌊ M T ⌋) \sum k | T k μ (k)

$=\sum_TTS(\lfloor\frac NT\rfloor)S(\lfloor\frac MT\rfloor)\sum_{k|T}k\mu(k)$
再考虑求

f(T)=∑k|Tkμ(k) f ( T ) = ∑ k | T k μ ( k ) $f(T)=\sum_{k|T}k\mu(k)$ 。
对于任意的

gcd(S,T)=1 gcd ( S , T ) = 1 $\gcd(S,T)=1$ ，都有：

f (S T) = \sum k | S T k μ (k)

$f(ST)=\sum_{k|ST}k\mu(k)$

\sum d | S \sum k | T d k μ (d k)

$\sum_{d|S}\sum_{k|T}dk\mu(dk)$

= (\sum d | S μ (d)) (\sum k | T μ (k))

$=(\sum_{d|S}\mu(d))(\sum_{k|T}\mu(k))$

= f (S) f (T)

$=f(S)f(T)$
我们得到：

f f $f$ 是积性函数。可以线性筛求。

f (1) = 1

$f(1)=1$

\forall n 为 质 数 且 k \geq 1, f (n k) = 1 - n

$\forall n为质数且k\ge 1,f(n^k)=1-n$

\forall m 为 质 数 且 m | n, f (n m) = f (n)

$\forall m为质数且m|n,f(nm)=f(n)$

\forall gcd (n, m) = 1, f (n m) = f (n) f (m)

$\forall\gcd(n,m)=1,f(nm)=f(n)f(m)$

Code

#include <cmath>
#include <cstdio>
#include <cstring>
#include <iostream>
#include <algorithm>
#define For(i, a, b) for (i = a; i <= b; i++)
#define Step(i, a, b, x) for (i = a; i <= b; i += x)
using namespace std;
const int MaxN = 1e7, N = 1e7 + 5, ZZQ = 20101009;
int n, m, tot, pri[N], zzq[N], ans;
bool mark[N];
void sieve() {
    int i, j;
    mark[0] = mark[zzq[1] = 1] = 1;
    For (i, 2, MaxN) {
        if (!mark[i]) zzq[pri[++tot] = i] = 1 - i + ZZQ;
        For (j, 1, tot) {
            if (1ll * i * pri[j] > MaxN) break;
            mark[i * pri[j]] = 1;
            if (i % pri[j] == 0) {
                zzq[i * pri[j]] = zzq[i];
                break;
            }
            else zzq[i * pri[j]] = 1ll * zzq[i] * zzq[pri[j]] % ZZQ;
        }
    }
}
int S(int n) {
    return (1ll * n * (n + 1) >> 1) % ZZQ;
}
int main() {
    int i;
    cin >> n >> m;
    sieve();
    For (i, 1, min(n, m))
        ans = (ans + 1ll * i * S(n / i) % ZZQ * S(m / i)
        % ZZQ * zzq[i] % ZZQ) % ZZQ;
    cout << ans << endl;
    return 0;
}