Convolution roots generate every nonnegative real exponential power of the logarithm
Provedlevy_foundations_id_log_exp_powers_are_charFunFor an infinitely divisible law with a specified continuous normalised characteristic exponent ψ, all nonnegative real powers exp(t ψ) are probability characteristic functions. Use uniqueness of normalised convolution roots for rational t and the Lévy continuity theorem to pass to arbitrary nonnegative real times.
Preamble
import Mathlib.Probability.Independence.CharacteristicFunction import Definitions.Def_levy_foundations_isInfinitelyDivisible open MeasureTheory ProbabilityTheory
Formal statement
/-- Positive real exponential powers of the normalised logarithm of an
infinitely divisible characteristic function are again characteristic functions. -/
theorem levy_foundations_id_log_exp_powers_are_charFun
{μ : Measure ℝ} [IsProbabilityMeasure μ]
(hμ : LevyFoundations.isInfinitelyDivisible μ)
(ψ : ℝ → ℂ) (hcont : Continuous ψ) (hzero : ψ 0 = 0)
(hchar : ∀ ξ : ℝ, charFun μ ξ = Complex.exp (ψ ξ)) :
∀ t : ℝ, 0 ≤ t →
∃ ν : Measure ℝ, IsProbabilityMeasure ν ∧
∀ ξ : ℝ, charFun ν ξ = Complex.exp ((t : ℂ) * ψ ξ) := by sorry