大家国庆快乐！
题一
设 n ≥ 0 , α ∈ C \ { 0 , 1 , … , n − 1 } n\ge 0, \alpha \in \mathbb C \backslash \{0, 1, \dots, n-1\} n≥0,α∈C\{0,1,…,n−1}，求算
∑ 0 ≤ k ≤ n ( n k ) ( α k ) \sum_{0\le k\le n} \frac{\binom n k}{\binom \alpha k} 0≤k≤n∑(kα)(kn)
我们首先用 Gamma 函数做一下变形
∑ k ( n k ) ( α k ) = ∑ k Γ ( n + 1 ) Γ ( n − k + 1 ) Γ ( k + 1 ) Γ ( α + 1 ) Γ ( α − k + 1 ) Γ ( k + 1 ) = Γ ( n + 1 ) Γ ( α + 1 ) ∑ k Γ ( α − k + 1 ) Γ ( n − k + 1 ) = ( α n ) − 1 ∑ k ( α − k n − k ) \begin{aligned} \sum_k \frac{\binom n k}{\binom \alpha k} &= \sum_k \frac{\frac{\Gamma(n+1)}{\Gamma(n-k+1)\Gamma(k+1)}}{\frac{\Gamma(\alpha+1)}{\Gamma(\alpha-k+1)\Gamma(k+1)}}\\ &= \frac{\Gamma(n+1)}{\Gamma(\alpha+1)} \sum_k \frac{\Gamma(\alpha-k+1)}{\Gamma(n-k+1)}\\ &= \binom{\alpha}{n}^{-1} \sum_k \binom{\alpha -k}{n-k} \end{aligned} k∑(kα)(kn)=k∑Γ(α−k+1)Γ(k+1)Γ(α+1)Γ(n−k+1)Γ(k+1)Γ(n+1)=Γ(α+1)Γ(n+1)k∑Γ(n−k+1)Γ(α−k+1)=(nα)−1k∑(n−kα−k)
接下来就是常规推导了
= ( α n ) − 1 ∑ k [ x n − k ] ( 1 + x ) α − k = ( α n ) − 1 ∑ k [ x n ] ( 1 + x ) α − k x k = ( α n ) − 1 [ x n ] ( 1 + x ) α ∑ k x k ( 1 + x ) k = ( α n ) − 1 [ x n ] ( 1 + x ) α 1 1 − x 1 + x = ( α n ) − 1 [ x n ] ( 1 + x ) α + 1 = ( α n ) − 1 ( α + 1 n ) = α + 1 α + 1 − n \begin{aligned} &= \binom{\alpha}{n}^{-1} \sum_k [x^{n-k}] (1+x)^{\alpha-k}\\ &= \binom{\alpha}{n}^{-1} \sum_k [x^n] (1+x)^{\alpha-k}x^k\\ &= \binom{\alpha}{n}^{-1} [x^n] (1+x)^{\alpha} \sum_k \frac{x^k}{(1+x)^k}\\ &= \binom{\alpha}{n}^{-1} [x^n] (1+x)^{\alpha} \frac{1}{1-\frac{x}{1+x}}\\ &= \binom{\alpha}{n}^{-1} [x^n] (1+x)^{\alpha+1}\\ &= \binom{\alpha}{n}^{-1} \binom{\alpha+1}n\\ &= \frac{\alpha+1}{\alpha+1-n} \end{aligned} =(nα)−1k∑[xn−k](1+x)α−k=(nα)−1k∑[xn](1+x)α−kxk=(nα)−1[xn](1+x)αk∑(1+x)kxk=(nα)−1[xn](1+x)α1−1+xx1=(nα)−1[xn](1+x)α+1=(nα)−1(nα+1)=α+1−nα+1
另解
这个式子其实还等价于一个恒等式，我们将 α \alpha α 在无穷处展开，先把等式两边减一变成
∑ k n k + 1 ‾ α k + 1 ‾ = n α + ( 1 − n ) \sum_{k} \frac{n^{\underline {k+1}}}{\alpha^{\underline {k+1}}} = \frac n{\alpha + (1 - n)} k∑αk+1nk+1=α+(1−n)n
注意
1 α ( α − 1 ) ⋯ ( α − k ) = ∑ j ≥ 0 { j k } α − j − 1 \frac1{\alpha(\alpha - 1) \cdots (\alpha - k)} = \sum_{j\ge 0} {j\brace k} \alpha^{-j-1} α(α−1)⋯(α−k)1=j≥0∑{kj}α−j−1
而
n α + ( 1 − n ) = ∑ j ≥ 0 n ( n − 1 ) j α − j − 1 \frac{n}{\alpha + (1-n)} = \sum_{j\ge 0} n(n-1)^j \alpha^{-j-1} α+(1−n)n=j≥0∑n(n−1)jα−j−1
因此
[ α − j − 1 ] ∑ k n k + 1 ‾ α k + 1 ‾ = [ α − j − 1 ] n α + ( 1 − n ) ∑ k n k + 1 ‾ { j k } = n ( n − 1 ) j ∑ k ( n − 1 k ) k ! { j k } = ( n − 1 ) j \begin{aligned} [\alpha^{-j-1}]\sum_{k} \frac{n^{\underline {k+1}}}{\alpha^{\underline {k+1}}} &= [\alpha^{-j-1}] \frac n{\alpha + (1 - n)}\\ \sum_k n^{\underline {k+1}} {j \brace k} &= n(n-1)^j\\ \sum_k \binom{n-1}k k!{j\brace k} &= (n-1)^j \end{aligned} [α−j−1]k∑αk+1nk+1k∑nk+1{kj}k∑(kn−1)k!{kj}=[α−j−1]α+(1−n)n=n(n−1)j=(n−1)j
上面的变形均可逆，因此我们通过斯特林降幂公式也是可以得到这个结论的。
题二
设 n , m ≥ 0 n,m\ge 0 n,m≥0，求算
∑ k ( n + k 2 k ) ( 2 k k ) ( − 1 ) k k + 1 + m \sum_{k} \binom {n+k}{2k} \binom{2k}k \frac{(-1)^k}{k+1+m} k∑(2kn+k)(k2k)k+1+m(−1)k
这个题如果我们硬刚 ( 2 k k ) = [ x k ] ( 1 − 4 x ) − 1 2 \binom{2k}k = [x^k](1-4x)^{-\frac 12} (k2k)=[xk](1−4x)−21 的话非常吃力，我们考虑变形为
( n + k 2 k ) ( 2 k k ) = ( n + k k ) ( n k ) \binom{n+k}{2k} \binom{2k}k = \binom{n+k}k \binom n k (2kn+k)(k2k)=(kn+k)(kn)
那么就有
∑ k ( n + k 2 k ) ( 2 k k ) ( − 1 ) k k + 1 + m = ∑ k ( n + k k ) ( n k ) ( − 1 ) k k + 1 + m = ∫ 0 1 ∑ k ( n + k k ) ( n k ) ( − t ) k t m d t = ∫ 0 1 ∑ k [ x 0 ] ( 1 + x ) n ( 1 + x x ) k ( n k ) ( − t ) k t m d t = ∫ 0 1 [ x 0 ] ( 1 + x ) n ( 1 − 1 + x x t ) n t m d t = ∫ 0 1 [ x n ] ( 1 + x ) n ( t + x ( 1 − t ) ) n t m d t = ∑ k ( n k ) [ x n − k ] ∫ 0 1 ( − t + x ( 1 − t ) ) n t m d t = ∑ k ( n k ) 2 ∫ 0 1 ( − 1 ) k ( 1 − t ) n − k t m + k d t \begin{aligned} &\quad \sum_{k} \binom {n+k}{2k} \binom{2k}k \frac{(-1)^k}{k+1+m}\\ & = \sum_{k} \binom {n+k}{k} \binom{n}k \frac{(-1)^k}{k+1+m}\\ &= \int_0^1 \sum_{k} \binom {n+k}{k} \binom{n}k (-t)^k t^m \,\mathrm{d}t\\ &= \int_0^1 \sum_{k} [x^0](1+x)^n \left(\frac{1+x}x\right)^k \binom{n}k (-t)^k t^m \,\mathrm{d}t\\ &= \int_0^1[x^0](1+x)^n \left(1-\frac{1+x}xt\right)^n t^m \,\mathrm{d}t\\ &= \int_0^1[x^n](1+x)^n (t+x(1-t))^n t^m \,\mathrm{d}t\\ &= \sum_k \binom n k [x^{n-k}]\int_0^1 (-t+x(1-t))^n t^m \,\mathrm{d}t\\ &= \sum_k \binom n k ^2 \int_0^1 (-1)^k(1-t)^{n-k} t^{m+k} \,\mathrm{d}t\\ \end{aligned} k∑(2kn+k)(k2k)k+1+m(−1)k=k∑(kn+k)(kn)k+1+m(−1)k=∫01k∑(kn+k)(kn)(−t)ktmdt=∫01k∑[x0](1+x)n(x1+x)k(kn)(−t)ktmdt=∫01[x0](1+x)n(1−x1+xt)ntmdt=∫01[xn](1+x)n(t+x(1−t))ntmdt=k∑(kn)[xn−k]∫01(−t+x(1−t))ntmdt=k∑(kn)2∫01(−1)k(1−t)n−ktm+kdt
我们要熟悉 Beta 积分 ∫ 0 1 ( 1 − t ) a t b d t = a ! b ! ( a + b + 1 ) ! \int_0^1 (1-t)^a t^b \,\mathrm{d}t=\frac{a!b!}{(a+b+1)!} ∫01(1−t)atbdt=(a+b+1)!a!b!（事实上这由分部积分是很容易证明的），接下来我们继续：
∑ k ( − 1 ) k ( n k ) 2 ( n − k ) ! ( m + k ) ! ( n + m + 1 ) ! = 1 n + m + 1 ( n + m n ) − 1 ∑ k ( − 1 ) k ( m + k m ) ( n k ) = 1 n + m + 1 ( n + m n ) − 1 ∑ k ( n k ) [ x 0 ] ( 1 + x ) m ( − 1 + x x ) k = 1 n + m + 1 ( n + m n ) − 1 [ x 0 ] ( 1 + x ) m ( 1 − 1 + x x ) n = 1 n + m + 1 ( n + m n ) − 1 [ x n ] ( 1 + x ) m = 1 n + m + 1 ( n + m n ) − 1 ( m n ) \begin{aligned} &\quad\sum_k (-1)^k\binom n k^2 \frac{(n-k)!(m+k)!}{(n+m+1)!}\\ &= \frac 1{n+m+1}\binom{n+m}n^{-1} \sum_k(-1)^k \binom{m+k}m\binom nk\\ &= \frac 1{n+m+1}\binom{n+m}n^{-1} \sum_k \binom nk[x^0] (1+x)^m \left(-\frac{1+x}{x}\right)^k\\ &= \frac 1{n+m+1}\binom{n+m}n^{-1} [x^0] (1+x)^m \left(1-\frac{1+x}{x}\right)^n\\ &= \frac 1{n+m+1}\binom{n+m}n^{-1} [x^n] (1+x)^m\\ &= \frac 1{n+m+1}\binom{n+m}n^{-1} \binom mn \end{aligned} k∑(−1)k(kn)2(n+m+1)!(n−k)!(m+k)!=n+m+11(nn+m)−1k∑(−1)k(mm+k)(kn)=n+m+11(nn+m)−1k∑(kn)[x0](1+x)m(−x1+x)k=n+m+11(nn+m)−1[x0](1+x)m(1−x1+x)n=n+m+11(nn+m)−1[xn](1+x)m=n+m+11(nn+m)−1(nm)
题三
设 n , m , k ≥ 0 n,m,k\ge0 n,m,k≥0，求
∑ i ( − 1 ) i ( m + n m + i ) ( n + k n + i ) ( k + m k + i ) \sum_i (-1)^i \binom{m+n}{m+i}\binom {n+k}{n+i} \binom{k+m}{k+i} i∑(−1)i(m+im+n)(n+in+k)(k+ik+m)
为了用生成函数算这个东西，这里我们不妨介绍多元拉格朗日反演公式的一个直接推论——MacMahon 主定理。
MacMahon 主定理（MacMahon Master Theorem） 令 A = [ a i , j ] n × n , X = diag ( x 1 , … , x n ) \mathbf A=[a_{i,j}]_{n\times n}, \mathbf X = \operatorname{diag} (x_1, \dots, x_n) A=[ai,j]n×n,X=diag(x1,…,xn)，则有
[ x k ] ∏ i = 1 n ( a i , 1 x 1 + ⋯ + a i , n x n ) k i = [ x k ] ∣ I − X A ∣ − 1 [\mathbf x^{\mathbf k}] \prod_{i=1}^n (a_{i,1}x_1 + \cdots + a_{i,n}x_n)^{k_i} = [\mathbf x^\mathbf k] \vert \mathbf {I-XA} \vert^{-1} [xk]i=1∏n(ai,1x1+⋯+ai,nxn)ki=[xk]∣I−XA∣−1
其中 k = ( k 1 , … , k n ) ≥ 0 \mathbf k =(k_1, \dots, k_n) \ge \mathbf 0 k=(k1,…,kn)≥0。
我们考虑和式
(
1
−
x
z
)
m
+
n
(
1
−
y
x
)
n
+
k
(
1
−
z
y
)
k
+
m
=
∑
a
,
b
,
c
(
−
1
)
a
+
b
+
c
(
m
+
n
a
)
(
n
+
k
b
)
(
k
+
m
c
)
x
a
−
b
y
b
−
c
z
c
−
a
\left(1-\frac xz\right)^{m+n} \left(1-\frac yx\right)^{n+k} \left(1-\frac zy\right)^{k+m} = \sum_{a,b,c} (-1)^{a+b+c} \binom{m+n}a \binom{n+k}b \binom{k+m}c x^{a-b} y^{b-c}z^{c-a}
(1−zx)m+n(1−xy)n+k(1−yz)k+m=a,b,c∑(−1)a+b+c(am+n)(bn+k)(ck+m)xa−byb−czc−a
考虑其中
(
m
+
n
m
+
i
)
(
n
+
k
n
+
i
)
(
k
+
m
k
+
i
)
\binom{m+n}{m+i}\binom {n+k}{n+i} \binom{k+m}{k+i}
(m+im+n)(n+in+k)(k+ik+m) 的项，是
(
−
1
)
m
+
n
+
k
∑
i
(
−
1
)
i
(
m
+
n
m
+
i
)
(
n
+
k
n
+
i
)
(
k
+
m
k
+
i
)
x
m
−
n
y
n
−
k
z
k
−
m
(-1)^{m+n+k}\sum_i (-1)^i \binom{m+n}{m+i}\binom {n+k}{n+i} \binom{k+m}{k+i} x^{m-n}y^{n-k}z^{k-m}
(−1)m+n+ki∑(−1)i(m+im+n)(n+in+k)(k+ik+m)xm−nyn−kzk−m
我们只需提取
(
−
1
)
m
+
n
+
k
[
x
m
−
n
y
n
−
k
z
k
−
m
]
(
1
−
x
z
)
m
+
n
(
1
−
y
x
)
n
+
k
(
1
−
z
y
)
k
+
m
(-1)^{m+n+k}[x^{m-n}y^{n-k}z^{k-m}]\left(1-\frac xz\right)^{m+n}\left(1-\frac yx\right)^{n+k}\left(1-\frac zy\right)^{k+m}
(−1)m+n+k[xm−nyn−kzk−m](1−zx)m+n(1−xy)n+k(1−yz)k+m 即可。
即提取
(
−
1
)
m
+
n
+
k
[
x
m
+
k
y
n
+
m
z
k
+
n
]
(
z
−
x
)
m
+
n
(
x
−
y
)
n
+
k
(
y
−
z
)
k
+
m
(-1)^{m+n+k}[x^{m+k}y^{n+m}z^{k+n}](z-x)^{m+n}(x-y)^{n+k}(y-z)^{k+m}
(−1)m+n+k[xm+kyn+mzk+n](z−x)m+n(x−y)n+k(y−z)k+m
整理一下，就是
(
−
1
)
m
+
n
+
k
[
x
k
+
m
y
m
+
n
z
n
+
k
]
(
y
−
z
)
k
+
m
(
z
−
x
)
m
+
n
(
x
−
y
)
n
+
k
(-1)^{m+n+k} [x^{k+m}y^{m+n}z^{n+k}](y-z)^{k+m}(z-x)^{m+n}(x-y)^{n+k}
(−1)m+n+k[xk+mym+nzn+k](y−z)k+m(z−x)m+n(x−y)n+k
符合 MacMahon 主定理的形式，说明它等于
(
−
1
)
m
+
n
+
k
[
x
k
+
m
y
m
+
n
z
n
+
k
]
∣
1
−
x
x
y
1
−
y
−
z
z
1
∣
−
1
(-1)^{m+n+k} [x^{k+m}y^{m+n}z^{n+k}] \left | \begin{matrix} 1 & -x & x\\ y & 1 & -y\\ -z & z & 1 \end{matrix} \right|^{-1}
(−1)m+n+k[xk+mym+nzn+k]∣∣∣∣∣∣1y−z−x1zx−y1∣∣∣∣∣∣−1
化简可得
(
−
1
)
m
+
n
+
k
[
x
k
+
m
y
m
+
n
z
n
+
k
]
(
1
+
x
y
+
y
z
+
z
x
)
−
1
(-1)^{m+n+k}[x^{k+m}y^{m+n}z^{n+k}] (1+xy+yz+zx)^{-1}
(−1)m+n+k[xk+mym+nzn+k](1+xy+yz+zx)−1
注意到假设取到
(
x
y
)
a
(
y
z
)
b
(
z
x
)
c
(xy)^a (yz)^b(zx)^c
(xy)a(yz)b(zx)c，只有唯一解
a
=
m
,
b
=
n
,
c
=
k
a=m, b=n, c=k
a=m,b=n,c=k。因此答案就是
(
−
1
)
m
+
n
+
k
[
u
m
v
n
w
k
]
(
1
+
u
+
v
+
w
)
−
1
=
(
m
+
n
+
k
m
,
n
,
k
)
(-1)^{m+n+k} [u^mv^nw^k] (1+u+v+w)^{-1} = \binom{m+n+k}{m,n,k}
(−1)m+n+k[umvnwk](1+u+v+w)−1=(m,n,km+n+k)
因此我们有
∑
i
(
−
1
)
i
(
m
+
n
m
+
i
)
(
n
+
k
n
+
i
)
(
k
+
m
k
+
i
)
=
(
m
+
n
+
k
m
,
n
,
k
)
\sum_i (-1)^i \binom{m+n}{m+i}\binom {n+k}{n+i} \binom{k+m}{k+i}=\binom{m+n+k}{m,n,k}
i∑(−1)i(m+im+n)(n+in+k)(k+ik+m)=(m,n,km+n+k)
Dixon’s Identity 是它的一个特例：
Dixon’s Identity
∑ k ( − 1 ) k ( 2 n k ) 3 = ( − 1 ) n ( 3 n n , n , n ) \sum_k (-1)^k \binom{2n}k ^3 = (-1)^n \binom{3n}{n,n,n} k∑(−1)k(k2n)3=(−1)n(n,n,n3n)
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
