In Combinatorics the Rota Way Kung gives a proof of Phillip Hall's Theorem of chains;
If $P$ a poset and $(x,y)$ an interval on P then $\mu(x,y)=-c_1+c_2-c_3...$
Here $\mu$ is the Möbius function, $\zeta=\mu^{-1}$, $c_i$ counts the number of lenghth-i chains with x and y as endpoints (so $x < p_1 <...p_{i-1}<y$). The proof given is;
''Let $\eta=\zeta-\delta$ then we have $\mu=(\eta+\delta)^{-1}=*=\delta-\eta+\eta^2-\eta^3...$'' and we are done
So I buy everything except *. It looks like $\frac{1}{1-r}=\sum_0^{\infty}r^n$ but I quite connect them precisely.
EDIT;
Similar to; Geometric series of an operator
But how could I guarantee that $\eta$ is nilpotent? There are no restrictions on $P$..it could be (viscously) infinite (like $P=\mathbb{R}$). Surely the series dosn't converge for real intervals?