Measurability, reciprocal bounds, and ratio asymptotics of the Gamma function #
Real.Gamma is smooth away from the nonpositive integers and has a pole at each of them, so it is
neither continuous nor locally bounded on all of ℝ. It is nevertheless Borel measurable, because
the set of its singularities is countable.
Mathlib records the analytic side of this (Real.differentiableAt_Gamma,
Real.not_continuousAt_Gamma_neg_nat) but never draws the measurability conclusion. It is needed
as soon as a formula containing Real.Gamma s is integrated or measured in the variable s —
for instance for the normalizing constants of the Gamma and Beta densities, which is why their
family-specific measurability modules use this result.
The reciprocal (Real.Gamma s)⁻¹, on the other hand, is continuous on all of ℝ: Mathlib's
convention Real.Gamma (-n) = 0 makes it vanish at the poles, matching the zeros of the entire
function 1 / Γ. It is bounded on every half-line [a, ∞), which is what uniform estimates for
series with coefficients 1 / Γ(b + n), such as hypergeometric and Bessel series, need.
For a complex parameter the corresponding estimate is the factorial decay of 1 / Γ(s + n) on
the half-plane 1 ≤ Re s.
Finally, the rising factorial is recorded as a quotient of real Gamma values, and Mathlib's Euler
limit formula Real.GammaSeq_tendsto_Gamma is recast as the asymptotic
equivalence Γ(n + 1 + s) ~ n ^ s * n! along the natural numbers, valid for every real shift s.
This is the form in which polynomial decay of Gamma quotients such as Γ(n + a) / Γ(n + b) enters
summability and moment arguments for discrete laws with Beta-function masses.
Main results #
Real.measurable_Gamma—Real.Gammais measurable.Real.continuous_inv_Gamma—s ↦ (Real.Gamma s)⁻¹is continuous.TauCeti.continuous_Gamma_sum_div_prod— the finite Gamma quotient is jointly continuous on positive parameters.Real.exists_abs_inv_Gamma_le—(Real.Gamma s)⁻¹is bounded on[a, ∞).Real.Gamma_add_nat_div_Gamma_eq—Γ(x + n) / Γ xis the rising factorialx (x + 1) ⋯.Real.prod_range_add_mul_Gamma—s (s + 1) ⋯ (s + n - 1) Γ s = Γ(s + n)for positives.Real.Gamma_nat_add_half_eq_factorial—Γ(n + 1 / 2) = (2 n)! √π / (4 ^ n n!).Real.isEquivalent_Gamma_natCast_add_one_add—Γ(n + 1 + s) ~ n ^ s * n!asn → ∞.Real.tendsto_Gamma_natCast_add_div_Gamma_natCast_add_atTop—Γ(n + a) / Γ(n + b) → 0fora < b.Complex.norm_inv_Gamma_add_natCast_le—‖1 / Γ(s + k)‖ ≤ ‖1 / Γ(s)‖ / k!for1 ≤ Re s, the complex counterpart of the half-line bound used for series with coefficients1 / Γ(b + k).Complex.sum_range_neg_one_pow_div_factorial_mul_Gamma— the alternating sum∑_{i ≤ m} (−1)ⁱ / (i! Γ(a + m − i + 1)) = (−1)ᵐ / (m! (a + m) Γ(a))fora ∉ −ℕ.
The real Gamma function is Borel measurable on all of ℝ, including at its poles.
The reciprocal of the real Gamma function is continuous on all of ℝ, including at the poles
of Real.Gamma, where both it and the entire function 1 / Γ vanish.
For a finite family of positive parameters, the Gamma quotient
Γ(∑ i, a i) / ∏ i, Γ(a i) is jointly continuous.
The ascending Pochhammer symbol is given by the ratio of real Γ values; this is the real
counterpart of Complex.Gamma_add_nat_div_Gamma_eq.
The half-integer value Γ(n + 1 / 2) = (2 n)! √π / (4 ^ n n!), the factorial form of
Real.Gamma_nat_add_half, obtained from Legendre's duplication formula at n + 1 / 2.
Euler's limit formula in ratio form. For every real s, Γ(n + 1 + s) is asymptotic to
n ^ s * n! as the natural number n tends to infinity.
The Gamma quotients Γ(n + a) / Γ(n + b) tend to zero along the natural numbers when a < b,
since they decay like n ^ (a - b).
On the half-plane 1 ≤ Re s, the reciprocal Gamma function decays factorially along
s + k: ‖1 / Γ(s + k)‖ ≤ ‖1 / Γ(s)‖ / k!. Together with the boundedness of 1 / Γ on compact
sets, this gives locally uniform bounds for series with coefficients 1 / Γ(b + k) as functions
of a complex parameter b.
A finite alternating sum of reciprocal Gamma values: for a ∉ −ℕ,
∑_{i ≤ m} (−1)ⁱ / (i! Γ(a + m − i + 1)) = (−1)ᵐ / (m! (a + m) Γ(a)). This is the coefficient
identity behind the expansion of Tricomi's incomplete Gamma function in powers of −z
(DLMF 8.7.1).