Negative binomial series 2026-10-06
For a nonnegative integer and , this specialization of the binomial series sums the normalizing coefficients of a negative binomial distribution.
Let and . Decomposing at the first return and using the Markov property gives the renewal equation
For , define the probability generating functions and . Absolute convergence, or nonnegative summation, justifies the convolution identity . The binomial series gives
Solving the renewal identity gives the first-return generating function of the simple symmetric random walk
As consistency checks, as , confirming recurrence, while . By monotone convergence of , this last limit is , confirming null recurrent states.
Put and . Then
The binomial series gives
so
Since , a second binomial series expansion yields
Thus the constants in the stated Big O notation expansion are
An analytic branch of a square root determined by the branch is
If and are two such branches, their ratio is analytic and satisfies . Since a domain is connected and takes values in the discrete set , is constant. Thus throughout or throughout .
For , the principal square root is
On , one may instead take
The respective removed half-axes are their branch cuts.
On , define
It is analytic there, its square is , and
so it is the required branch. Substitution gives, for ,
The binomial series yields
hence the first three terms of the Laurent series are
Since
its residue at zero is . Under the change of variable , the two orientation reversals cancel, and the residue theorem gives
For each nonzero period lattice point , the binomial series gives
Pairing with cancels every odd power. Hence the Laurent series of the Weierstrass elliptic function is
which is the stated coefficient formula.