# 近似π的值 ## 數學上的定義 $\pi$的定義:圓周長與直徑之比值 數學上定義$\pi$的方式為使用積分 設單位圓$x^2+y^2=1$ 可得$y=\pm\sqrt{1-x^2}$ 取上半圓,即$y=f(x)=\sqrt{1-x^2}$ 可得其弧長與半徑之比值即為$\pi$ 故 $\pi=\displaystyle\int_{-1}^{1}\sqrt{1+(f'(x))^2}dx \\=\displaystyle\int_{-1}^{1}\dfrac{1}{\sqrt{1-x^2}}dx$ 利用視察法可輕易得出$\dfrac{1}{\sqrt{1-x^2}}$的其中一個反導函數為$\sin^{-1}x$ 故原式之值為$\sin^{-1}(1)-\sin^{-1}(-1)=\dfrac{\pi}{2}-(-\dfrac{\pi}{2})=\pi$ 而近似$\pi$之值最直覺的想法就是拿正多邊形去夾擠 有一半徑為$1$的圓$O$,內外接二正$n$邊形 則:內接正$n$邊形之周長$l_i$與外接正$n$邊形之周長$l_e$滿足$l_i\le2\pi\le l_e$ 利用此性質即可近似$\pi$的值 例: 各取$n=4,6,8$ $n=4$時 可得 $l_i=4\sqrt{2},l_e=8\implies 4\sqrt{2}\le2\pi\le8 \\\therefore2\sqrt{2}\approx2.828\le\pi\le4$ $n=6$時 可得 $l_i=6,l_e=4\sqrt{3}\implies 6\le2\pi\le4\sqrt{3} \\\therefore3\le\pi\le2\sqrt{3}\approx3.464$ $n=8$時 可得 $l_i=8\sqrt{2-2\cos{\dfrac{\pi}{4}}}=8\sqrt{2-\sqrt{2}} \\l_e=16\tan{\dfrac{\pi}{8}}=16\sqrt{\dfrac{1-\cos{\cfrac{\pi}{4}}}{1+\cos{\dfrac{\pi}{4}}}}=16\sqrt{2(1-\dfrac{\sqrt{2}}{2})^2}=16\sqrt{3-2\sqrt{2}} \\\therefore 3.0614\approx4\sqrt{2-\sqrt{2}}\le\pi\le8\sqrt{3-2\sqrt{2}}\approx3.3137$ 此時可求出$l_i$與$l_e$與$n$之關係,即可透過電腦運算近似$\pi$之值 設內接正$n$邊形之邊長為$x_i$ 以餘弦定理可知 $\cos{\dfrac{2\pi}{n}}=\dfrac{2-{x_i}^2}{2} \\\therefore l_i=nx_i=n\sqrt{2-2\cos{\dfrac{2\pi}{n}}},n\ge3$ 而外接正$n$邊形之邊長為$x_e$ 可得 $x_e=2\tan{\dfrac{\pi}{n}} \\\therefore l_e=nx_e=2n\tan{\dfrac{\pi}{n}},n\ge3$ 此時當$n\to\infty$時 有 $\displaystyle\lim_{n\to\infty}n\sqrt{2-2\cos{\dfrac{2\pi}{n}}}\le2\pi\le\lim_{n\to\infty}2n\tan{(\dfrac{\pi}{n})}$ 而第一個極限: 令$m=\dfrac{1}{n}\implies m\to{0^+}$ 原式變為 $\displaystyle\lim_{m\to{0^+}}\dfrac{\sqrt{2}\sqrt{1-\cos{2\pi m}}}{m}=\lim_{m\to{0^+}}2\dfrac{\sin{m\pi}}{m}=2\pi$ 第二個極限: 原式為 $\displaystyle\lim_{m\to{0^+}}2\dfrac{\tan{n\pi}}{n}=2\pi$ ## 近似方法 ### 面積 證明圓面積為$\pi r^2$,其中$r$為圓半徑 $pf:$ 設$C:x^2+y^2=r^2$,則滿足$x^2+y^2\le r^2$之點集合所形成之面積$A$為 $A=4\displaystyle\int_{0}^{r}\sqrt{r^2-x^2}dx$ 令$x=r\sin{u}\implies dx=r\cos{u}du$ $A=4\displaystyle\int_{0}^{\frac{\pi}{2}}r^2\cos^2{u}du \\=4r^2\displaystyle\int_{0}^{\frac{\pi}{2}}\cos^2{u}du \\=4r^2\displaystyle\int_{0}^{\frac{\pi}{2}}\dfrac{1+\cos{2u}}{2}du \\=4r^2(\dfrac{u}{2}+\dfrac{u}{4}\sin{2u}\Bigg|_{0}^{\frac{\pi}{2}}) \\=4r^2[(\dfrac{\pi}{4})-(0)]=r^2\pi$ ### 利用上和與下和近似 考慮一半徑為$2$,以原點為圓心之$\dfrac{1}{4}$圓$C$ 則$C:\begin{cases}x^2+y^2=4\\x\ge0\\y\ge0\end{cases}$ 其面積為$\displaystyle\int_{0}^{2}\sqrt{4-x^2}dx=\dfrac{1}{4}\times\pi\times2^2=\pi$ 函數$f(x)=\sqrt{4-x^2}$ 函數在$[0,2]$上的上和為 $U_n=\displaystyle\sum_{k=1}^{n}\dfrac{2}{n}\sqrt{4-(\dfrac{2(k-1)}{n})^2}$ 在$[0,2]$的上的下和為 $L_n=\displaystyle\sum_{k=1}^{n-1}\dfrac{2}{n}\sqrt{4-(\dfrac{2k}{n})^2}$ 故可得$L_n\le\pi\le U_n$ 此時可使用程式計算,當$n$取值越大,$U_n-L_n,U_n-\pi,\pi-L_n$會越小且大於$0$ ```python from math import sqrt def upper_sum(n): sum = 0 for i in range(n): k=i+1 delta_x = 2/n f_x = sqrt(4-(2*(k-1)/n)*(2*(k-1)/n)) sum += f_x*delta_x return sum def lower_sum(n): sum = 0 for i in range(n-1): k=i+1 delta_x = 2/n f_x = sqrt(4-(2*(k)/n)*(2*(k-1)/n)) sum += f_x*delta_x return sum n = 4 while n < 1000000: u_n = upper_sum(n) l_n = lower_sum(n) print(f'upper sum is {u_n} ,lower sum is {l_n} for n = {n}') n += 100 ``` 執行部分結果:  ### 利用梯形近似 考慮相同函數,若使用梯形近似其面積 $A\approx T_n=\displaystyle\sum_{k=1}^{n}\dfrac{2}{n}(\dfrac{f(\dfrac{2(k-1)}{n})+f(\dfrac{2k}{n})}{2}) \\=\displaystyle\sum_{k=1}^{n}\dfrac{\sqrt{4-(\dfrac{2(k-1)}{n})^2}+\sqrt{4-(\dfrac{2k}{n})^2}}{n}=2\displaystyle\sum_{k=1}^{n}\dfrac{\sqrt{1-(\dfrac{k-1}{n})^2}+\sqrt{1-(\dfrac{k}{n})^2}}{n} \\=2\displaystyle\sum_{k=1}^{n}\dfrac{\sqrt{n^2-(k-1)^2}+\sqrt{n^2-k^2}}{n^2}$ 利用程式 ```python def Trapezoidal_sum(n): sum = 0 for i in range(n): k = i+1 sum += sqrt(n*n-(k-1)*(k-1))+sqrt(n*n-k*k) sum*=2 sum/=n*n return sum n=4 while n< 10000: t_n = Trapezoidal_sum(n) print(f'{t_n} for n = {n}') ``` 執行部分結果:  可以發現當$n=99504$時,上和為$3.1416127158\dots$,下和為$3.1415925372\dots$ 而梯形分割則為$3.1415926161\dots$ 又真實的$\pi=3.1415926535\dots$ 將差距$\delta$取至小數點後10位,之後無條件捨去 上和與$\pi$之差距絕對值$\delta_u=|\pi-U_{99504}|=0.0000200623=2.00623\times10^{-5}$ 下和與$\pi$之差距絕對值$\delta_l=|\pi-U_{99504}|=0.0000001163=1.163\times10^{-7}$ 梯形分割法與$\pi$之差距絕對值$\delta_{t}=|\pi-T_{99504}|=0.0000000374=3.74\times10^{-8}$ 誤差百分比為 $E_u=\dfrac{\delta_u}{3.1415926535}\times100\%\approx6.386028\times10^{-4}\% \\E_l=\dfrac{\delta_l}{3.1415926535}\times100\%\approx3.701943\times10^{-6}\% \\E_t=\dfrac{\delta_t}{3.1415926535}\times100\%\approx 1.190478\times10^{-6}\%$ 可見梯形分割法較上和下和逼近法準確
×
Sign in
Email
Password
Forgot password
or
By clicking below, you agree to our
terms of service
.
Sign in via Facebook
Sign in via Twitter
Sign in via GitHub
Sign in via Dropbox
Sign in with Wallet
Wallet (
)
Connect another wallet
New to HackMD?
Sign up