The power series of a function, if it exits, is unique, i.e. there is at most one power series of the form $\sum_{n=0}^{\infty} c_n (z-a)^n$ which converges to a given function within a circle of convergence centered at $a$. We call this a power series “expanded around $a$”.
Note: This theorem is an open invitation to collect a bag of cute tricks. It doesn't matter how you find a series for a function, once you have it, it is the series. The rest of these theorems should be in your bag of cute tricks. Scientists use these theorems, rather that the method in § {Power Series} to find series expansions whenever possible.
Example: \begin{equation} \sin z = \sum_{n=0}^{\infty} {(-1)^n z^{2n+1}\over (2n+1)!} \qquad\qquad\forall z \end{equation} \begin{eqnarray} \cos z = {d\over dz}\sin z &=& \sum_{n=0}^{\infty} {(-1)^n (2n+1) z^{2n}\over (2n+1)!} \nonumber\\ &=& \sum_{n=0}^{\infty} {(-1)^n z^{2n}\over (2n)!} \qquad\qquad\forall z \end{eqnarray}
Example: \begin{equation} {1\over 1+z} = 1 - z + z^2 - z^3 + \dots = \sum_{n=0}^{\infty} (-z)^n \qquad\qquad |z|<1 \end{equation}
\begin{eqnarray} {1\over 1+\sin z} &=& 1 - \sum_{n=0}^{\infty} {(-1)^n z^{2n+1}\over(2n+1)!} + \left( \sum_{n=0}^{\infty} {(-1)^n z^{2n+1}\over (2n+1)!} \right)^2 \nonumber\\ && \qquad - \left( \sum_{n=0}^{\infty} {(-1)^n z^{2n+1}\over (2n+1)!} \right)^3 + \dots \\ &=& 1 - z + z^2 + \left({1\over 3!}-1\right)\, z^3 + \dots \qquad |\sin z|<1 \nonumber \end{eqnarray} What happens if you try this same trick to find a power series for $1/(1+\cos z)$? Why?
Example: \begin{equation} z-{\pi\over 2} = z-{\pi\over 2}\qquad\qquad\forall z \end{equation} Note: This is a very short power series with just two non-zero terms. \begin{equation} \cos(z) = -\sin (z-{\pi\over 2}) = \sum_{n=0}^{\infty} {(-1)^{n+1} (z-{\pi\over 2})^{2n+1}\over(2n+1)!} \qquad\qquad\forall z \end{equation}
Note: Starting with a power series for $\sin(z)$ expanded around $z=0$, we have obtained a power series for $\cos(z)$ expanded around $z={\pi\over 2}$.
Example: \begin{eqnarray} {2\over 1-z^2} &=& {1\over 1+z} + {1\over 1-z} \nonumber\\ &=& (1 - z + z^2 -z^3 + \dots) + (1 + z + z^2 + z^3 + \dots) \nonumber\\ &=& 2 (1 + z^2 + z^4 + \dots) \qquad\qquad |z|<1 \end{eqnarray} Compare this to the result you would get using the previous theorem. Which method is faster?
Example: \begin{eqnarray} {\sin z\over 1+z} &=& \left( z - {z^3\over 3!} + {z^5\over 5!} + \dots \right) \left( 1 - z + z^2 - z^3 + z^4 - z^5 \right) \nonumber\\ &=& z - z^2 + \left(-{1\over 3!}+1\right) z^3 + \left({1\over 3!}-1\right) z^4 \nonumber\\ && \qquad + \left({1\over 5!}-{1\over 3!}+1\right)z^5 +\dots \qquad\qquad |z|<1 \end{eqnarray}
Compare this series to the series for the function $1-{1\over 1+\sin(z)}$ (see the first example in Theorem 2.) What can you conclude about the wisdom of assuming two series are the same if their first three terms are identical?
Try the previous example $\sin z/(1+z)$ using synthetic division, instead. Is this method easier or harder? Imagine what you would do if the denominator were a power series with an infinite number of non-zero terms.
Example: Expand $\sin z$ around $z=\pi$. \begin{eqnarray} \sin z &=& \sin[(z-\pi)+\pi] \nonumber\\ &=& \sin(z-\pi)\cos\pi +\cos(z-\pi)\sin\pi \nonumber\\ &=& -\sin(z-\pi) \nonumber\\ &=& -\sum_{n=0}^{\infty} {(-1)^n (z-\pi)^{2n+1}\over (2n+1)!} \qquad\qquad\forall z \end{eqnarray}