Dirichlet series

A Dirichlet series is a series of the form

∑n=1∞anns.

Example: an=1 gives the Riemann zeta function ζ(s).

Generalized definition

The classical form is the special case λn=ln⁡n of the generalized Dirichlet series (also called Dirichlet series in the sense of Riesz):

∑n=1∞ane−λns,λn↗∞ strictly increasing,

since n−s=e−sln⁡n. Any such series is a Dirichlet series; the choice λn only sets the "spacing" of the exponentials.

Made by Claudio, to be reviewed

Example: the geometric series as a Dirichlet series. With λn=n, an=1 and s=t:

∑n=1∞e−nt=e−t1−e−t=1et−1.

Converges for all t>0 (arithmetic-progression exponents dominate the ln⁡n of ζ, improving the abscissa from σ>1 to σ>0). Its Mellin transform is the bridge to the classical case:

Γ(s)ζ(s)=∫0∞ts−1et−1dt=∑n=1∞∫0∞ts−1e−ntdt=Γ(s)∑n=1∞1ns.

So ∑e−nt and ∑n−s are Mellin-dual faces of the same object. See Fourier, Laplace and Dirichlet relationship for the unified "comb" picture.

Convergence

For a fixed coefficient sequence the convergence region is a half-plane (not a disc, unlike Laurent series):

  • Abscissa of convergence: σc=inf{σ:∑ann−σ converges}.
  • Abscissa of absolute convergence: σa=inf{σ:∑|an|n−σ converges}, with σc≤σa≤σc+1.
  • In every half-plane Re(s)≥σc+ε the convergence is uniform, hence the sum is holomorphic there. This is the standard route to the analytic continuation of ζ.
  • Partial sums obey the Riemann--Stieltjes integral identity
∑n≤xann−s=A(x)xs+s∫1xA(t)ts+1dt,A(t)=∑n≤tan,

which recasts convergence questions in terms of the growth of A(t).

See absolute convergence and series of functions.

Examples

  • an=1: ζ(s)=∑n−s (Riemann zeta function).
  • an=(−1)n+1: η(s)=∑(−1)n+1n−s, the alternating zeta, related to ζ by η(s)=(1−21−s)ζ(s).
  • an a Dirichlet character: L-functions L(s,χ), the building blocks of the Dirichlet L-series.
  • an=μ(n) (Möbius): ∑μ(n)n−s=1/ζ(s).
  • an=Λ(n)/ln⁡n: ∑Λ(n)n−s=−ζ′(s)/ζ(s), the logarithmic derivative governing the primes.

Euler product

If an is multiplicative (amn=aman for coprime m,n) and apk grows mildly, the series factorizes over primes:

∑n=1∞anns=∏p(1+apps+ap2p2s+⋯).

For an=1 this is Euler's product ζ(s)=∏p(1−p−s)−1, the analytic bridge to the primes: via Perron's formula it recovers the partial sums ∑n≤xan, and the zeros/poles of the Euler product control π(x) and the Chebyshev function. The relation with the logarithmic integral and the location of the zeros of ζ is the content of the prime number theorem and the Riemann hypothesis.

Relation to Fourier series

A Dirichlet series is the Mellin transform of the integer Dirac comb ∑nanδ(x−n). Summary: same "basis change" picture as the Laplace/Fourier transforms, but the transform variable enters multiplicatively. See Fourier, Laplace and Dirichlet relationship.

Related: Riemann zeta function, L-functions, Mellin transform, Perron's formula, prime counting function, Chebyshev function, Fourier series expansion, absolute convergence.