Gaussian-sech reciprocity
ProvedRamanujanLostNotebook.PartIV.eq_18_2_23Let with . Then . Differs from the printed source: the print gives only the product condition and does not say ; positivity is made explicit as a domain convention for the printed real square roots (ours; not a numerically discovered correction).
Discrepancy from the printed source. Book, p. 368, (18.2.23): the display has the condition on alpha*beta and the real square roots sqrt(alpha), sqrt(beta), but does not print alpha > 0, beta > 0. The Lean statement assumes 0 < alpha and 0 < beta. This is an explicit domain convention for the printed real square roots, not a numerically discovered correction; it keeps Real.sqrt of a negative number out of the statement. The audit's independent 42-digit check agrees with ours at alpha = 0.8, 3.1, 0.05. Convention chosen by us.
import Mathlib open MeasureTheory
namespace RamanujanLostNotebook.PartIV
theorem eq_18_2_23 (α β : ℝ) (hα : 0 < α) (hβ : 0 < β) (hαβ : α * β = Real.pi) :
Real.sqrt α * (∫ x in Set.Ioi (0 : ℝ),
Real.exp (-(α * x) ^ 2) / Real.cosh (Real.pi * x))
= Real.sqrt β * (∫ x in Set.Ioi (0 : ℝ),
Real.exp (-(β * x) ^ 2) / Real.cosh (Real.pi * x)) := by sorry
end RamanujanLostNotebook.PartIV