Forgot password?
 Create new account
View 130|Reply 2

[函数] $-\log(\log(1-x))$在$1-e$的各阶导数的渐近估计

[Copy link]

3151

Threads

8498

Posts

610K

Credits

Credits
66208
QQ

Show all posts

hbghlyj Posted at 2022-11-8 04:27:09 |Read mode
Last edited by hbghlyj at 2023-5-30 15:14:00$f(x)=-\log\log(1-x)$在$x=1-e$的$n$阶导数的$e^n$倍是正整数:
  1. Table[Derivative[n][Function[x,-Log[Log[1-x]]]][1-E]E^n,{n,1,10}]
Copy the Code

在Mathematica中, 逐阶计算导数不如计算级数展开速度快. 上述代码可以改进为
  1. Series[-Log[Log[1-x]],{x,1-E,10}][[3]]Exp[Range[10]]Factorial[Range[10]]
Copy the Code

计算结果
\[1,2,7,35,228,1834,17582,195866,2487832,35499576\]
如何得到它的渐近估计呢? 可以先通过奇点算出级数收敛半径, 再使用 Cauchy-Hadamard formula
$x=0$是$-\log(\log(1-x))$的奇点[因为$\log1=0$是外层$\log$的奇点]
所以$f(x)$在$x=1-e$的级数收敛半径是$e-1$
所以$\frac{f^{(n)}(1-e)}{n!}\sim\frac1{(e-1)^n}$
所以$f^{(n)}(1-e)\sim\frac{n!}{(e-1)^n}$
所以$$a_n=e^nf^{(n)}(1-e)\sim n!\left(\frac{e}{e-1}\right)^n\tag1$$

3151

Threads

8498

Posts

610K

Credits

Credits
66208
QQ

Show all posts

 Author| hbghlyj Posted at 2022-11-8 04:30:01
\[0, 1, 2, 7, 35, 228, 1834, 17582, 195866, 2487832, 35499576\]
Expansion of e.g.f. $\log\left(1\over1+\log(1-x)\right)$.
A003713
$$a_0 = 0;\quad a_n = (n-1)! + \sum_{k=1}^{n-1} \binom{n-1}k(k-1)! a_{n-k}$$
$$\sum_{k=1}^n (k-1)!\left|\text{Stirling1}(n, k)\right|$$
$$a_n \sim (n-1)!\left(e\over e-1\right)^n\tag2$$

3151

Threads

8498

Posts

610K

Credits

Credits
66208
QQ

Show all posts

 Author| hbghlyj Posted at 2022-11-8 06:35:52
  1. series=Series[-Log[Log[1-x]],{x,1-E,50}][[3]];{a49,a50}=series[[49;;50]]Exp[{49,50}];
  2. {N[E/(E-1)],N[a50/a49]}
  3. {N[E/(E-1)],N[a50/a49 50/49]}
Copy the Code

由(1)式得到的估计是 1.58198, 1.55034
由(2)式得到的估计是 1.58198, 1.58198
可见, (2)式是比(1)式更精确的估计. 它是怎么得到的呢

手机版Mobile version|Leisure Math Forum

2025-4-21 14:09 GMT+8

Powered by Discuz!

× Quick Reply To Top Return to the list