本文结论来自于数学装逼神器——高端三角恒等变换——Aries
收集到这么多挺不容易的,一直没有那么大的耐心去收集整理,这里仅作为阅读随笔。
(sin)
证明使用了正弦的降幂公式,取偶数次的,转化为(cos)的求和再化简(我的证明)
[sin^{2r}x=frac{1}{2^{2r-1}}sum_{k=0}^{r-1}(-1)^{r+k}inom{2r}{k}cos(2r-2k)x+frac{1}{4^r}inom{2r}{r}
]
还有余弦的基本等差求和公式
[sum_{m=1}^n cos(2mx)=frac{sin(2n+1)x}{2sin x}-frac{1}{2}
]
得到
[sum_{m=1}^ncosleft[(2r-2k)frac{mpi}{2n+1}
ight]=frac{sinfrac{(2r-2k)pi}{2}}{2sinfrac{(2r-2k)pi}{2(2n+1)}}-frac{1}{2}=-frac{1}{2}\
sum_{m=1}^{n-1}cosleft[(2r-2k)frac{mpi}{2n+1}
ight]=frac{sinfrac{(2r-2k)(2n-1)pi}{4n}}{2sinfrac{(2r-2k)pi}{4n}}-frac{1}{2}=frac{1}{2}left((-1)^{r-k+1}-1
ight)]
这些都是套用繁琐的公式,没有技术含量,但是计算量较大,化简得到
[sum_{m=1}^n sin^{2r}frac{mpi}{2n+1}=frac{1}{4^r}left[sum_{k=0}^{r-1}(-1)^{r+k-1}inom{2r}{k}+inom{2r}{r}n
ight]\
sum_{m=1}^{n-1}sin^{2r} frac{mpi}{2n}=frac{1}{4^r}left[sum_{k=0}^{r-1}(-1)^{r+k}inom{2r}{k}left((-1)^{r-k+1}-1
ight)+inom{2r}{r}(n-1)
ight]]
然后利用组合数化简(看起来还可)
[sum_{k=0}^{r-1}(-1)^{r+k+1} inom{2r}{k}= frac{1}{2} inom{2r}{r},sum_{k=0}^{r-1}inom{2r}{k}=frac{1}{2}4^r-frac{1}{2}inom{2r}{r}
]
化简得到
[sum_{m=1}^n sin^{2r}frac{mpi}{2n+1}=frac{1}{4^r}inom{2r}{r}left(n+frac{1}{2}
ight),sum_{m=1}^{n-1}sin^{2r}frac{mpi}{2n}=frac{1}{4^r}inom{2r}{r}n-frac{1}{2}
]
(cos)
这组和上面的是一样的,没必要写过程了,而且使用区间再现也可以直接通过正弦推导得到
[sum_{m=1}^n cos^{2r}frac{mpi}{2n+1}=frac{1}{4^r}inom{2r}{r}left(n+frac{1}{2}
ight)-frac{1}{2},sum_{m=1}^{n-1}cos^{2r}frac{mpi}{2n}=frac{1}{4^r}inom{2r}{r}n-frac{1}{2}
]
( an)
证明用到的,这就是牛顿公式递推,根与系数的关系
[若sum_{k=0}^n a_k x^{n-k}=prod_{k=1}^n(x+b_k),S_k=sum_{m=1}^n b_m^k,则S_1=a_1,S_j=sum_{k=1}^{j-1}(-1)^{k+1}a_kS_{j-k}+(-1)^{j+1}ja_j ]
[sum_{k=0}^ninom{2n+1}{2k}x^{n-k}x^{n-k}=prod_{k=1}^nleft(x+ an^2 frac{kpi}{2n+1}
ight)
]
[egin{align}S_k:&=sum_{m=1}^n an^{2k}frac{mpi}{2n+1}\
S_1&=2n^2+n\S_j&=sum_{k=1}^{j-1}(-1)^{k+1}inom{2n+1}{2k}S_{j-k}+(-1)^{j+1}jinom{2n+1}{2j}end{align}]
类似地
[egin{align}S_k:&=sum_{m=1}^{n-1} an^{2k}frac{mpi}{2n}\
S_1&=frac{2}{3}n^2-n+frac{1}{3}\S_j&=frac{1}{2n}left[sum_{k=1}^{j-1}(-1)^{k+1}inom{2n}{2k+1}S_{j-k}+(-1)^{j+1}jinom{2n}{2j+1}
ight]end{align}]
(cot)
余切和正切类似证明
[egin{align}S_k:&=sum_{m=1}^ncot^{2k}frac{mpi}{2n+1}\
S_1&=frac{2}{3}n^2-frac{1}{3}n\S_j&=frac{1}{2n+1}left[sum_{k=1}^{j-1}(-1)^{k+1}inom{2n+1}{2k+1}S_{j-k}+(-1)^{j+1}jinom{2n+1}{2j+1}
ight]end{align}]
这里地公式和上面不太相同,是组合数不一样了
[sum_{m=1}^{n-1}cot^{2k}frac{mpi}{2n}=sum_{m=1}^{n-1} an^{2k}frac{mpi}{2n}
]
(sec)和(csc)
这可以通过上面两个推导得到
乘积
比较根与系数的常数项得到
[egin{align}prod_{m=1}^n sinfrac{mpi}{2n+1}&=frac{sqrt{2n+1}}{2^n}\
prod_{m=1}^n cosfrac{mpi}{2n+1}&=frac{1}{2^n}\
prod_{m=1}^n anfrac{mpi}{2n+1}&=sqrt{2n+1}\
prod_{m=1}^n cotfrac{mpi}{2n+1}&=frac{1}{sqrt{2n+1}}\
prod_{m=1}^n secfrac{mpi}{2n+1}&=2^n\
prod_{m=1}^n cscfrac{mpi}{2n+1}&=frac{2^n}{sqrt{2n+1}}end{align}]
[egin{align}prod_{m=1}^{n-1}sinfrac{mpi}{2n}&=prod_{m=1}^{n-1}cosfrac{mpi}{2n}=frac{sqrt{n}}{2^{n-1}}\
prod_{m=1}^{n-1} anfrac{mpi}{2n}&=prod_{m=1}^{n-1}cotfrac{mpi}{2n}=1\
prod_{m=1}^{n-1}secfrac{mpi}{2n}&=prod_{m=1}^{n-1}cscfrac{mpi}{2n}=frac{2^{n-1}}{sqrt{n}}end{align}]