\begin{eqnarray*} \arctan1&=&\frac{\pi}{4}\\ \arctan x&\stackrel{?}{=}&\sum_{k=0}^\infty a_kx^k\qquad\mbox{Taylor-Series}\end{eqnarray*}
We could calculate \(f^n(0)\) but it may be simpler in the following very instructive way: We assume that \(|x|\lt1\) \begin{eqnarray*} \arctan x&=&\int_0^x\frac{1}{1+t^2}dt\;\leftarrow\,\mbox{note: } [\arctan x]'=\frac{1}{1+x^2}\\ &=&\int_0^x\sum_{n=0}^\infty(-1)^n(t^2)^n dt\;\leftarrow\,\mbox{note:} \frac{1}{1-a}=\sum_{n=0}^\infty a^n,\;|a|\lt1\\ &=&\sum_{n=0}^\infty (-1)^n\int_0^x t^{2n} dt =\sum_{n=0}^\infty (-1)^n\frac{x^{2n+1}}{2n+1} \end{eqnarray*}
Thus: \begin{eqnarray*}\arctan x&=&x-\frac{x^3}{3}+\frac{x^5}{5}-\frac{x^7}{7}+\ldots\;\;;\mbox{also convergent for $x=1$ (can be shown)}\\ x&=&1\rightarrow\;\arctan1=\frac{\pi}{4}=1-\frac{1}{3}+\frac{1}{5}-\frac{1}{7}+\ldots\\ &\rightarrow&\mbox{very slow converging series: e.g.\ $10^{10}$ terms for {\bf10} precise digits of $\pi$!!}\end{eqnarray*}
Much better: To exploit the properties of the arctan function \begin{eqnarray*}\frac{\pi}{4}&=&4\arctan\frac{1}{5}-\arctan\frac{1}{239}\\ &=&\frac{4}{5}\sum_{k=0}^\infty\frac{(-1)^k}{2k+1}\left(\frac{1}{5}\right)^{2k}-\frac{1}{239}\sum_{k=0}^\infty\frac{(-1)^k}{2k+1}\left(\frac{1}{239}\right)^{2k}\\ &\sim&\mbox{10 terms for $\gt10$ precise digits of $\pi$!!}\;\;\left(\frac{1}{5}\right)^{20}\sim10^{-14} \end{eqnarray*}
It remains to prove that
\[\arctan 1 = 4\arctan\frac{1}{5}-\arctan\frac{1}{239}.\] |
This we show in three steps by each time applying the addition theorem for the arctan function
\[\arctan x+\arctan y=\arctan\frac{x+y}{1-xy}.\] |
\[\arctan 1 + \arctan \frac{1}{239}=\arctan \frac{1+\frac{1}{239}}{1-\frac{1}{239}}=\arctan\frac{120}{119}.\] |
\[\arctan \frac{1}{5}+\arctan \frac{1}{5} =\arctan\frac{\frac{2}{5}}{1-\frac{1}{25}}=\arctan\frac{5}{12}.\] |
\[\arctan \frac{5}{12}+\arctan \frac{5}{12} =\arctan\frac{\frac{10}{12}}{1-\frac{25}{144}}=\arctan\frac{10*20}{144-25}=\arctan\frac{120}{119}.\] |
Combining these equation we easily get the desired result.
Taylor series and error estimation
© J. Carstensen (Math for MS)