-- | -- Module: Math.NumberTheory.Zeta.Dirichlet -- Copyright: (c) 2018 Alexandre Rodrigues Baldé -- Licence: MIT -- Maintainer: Alexandre Rodrigues Baldé -- -- Dirichlet beta-function. {-# LANGUAGE ScopedTypeVariables #-} {-# OPTIONS_HADDOCK hide #-} module Math.NumberTheory.Zeta.Dirichlet ( betas , betasEven , betasOdd ) where import Data.ExactPi import Data.List (zipWith4) import Data.Ratio ((%)) import Math.NumberTheory.Recurrences (euler, factorial) import Math.NumberTheory.Zeta.Hurwitz (zetaHurwitz) import Math.NumberTheory.Zeta.Utils (intertwine, skipOdds) -- | Infinite sequence of exact values of Dirichlet beta-function at odd arguments, starting with @β(1)@. -- -- >>> approximateValue (betasOdd !! 25) :: Double -- 0.9999999999999987 -- >>> import Data.Number.Fixed -- >>> approximateValue (betasOdd !! 25) :: Fixed Prec50 -- 0.99999999999999999999999960726927497384196726751694 betasOdd :: [ExactPi] betasOdd = zipWith Exact [1, 3 ..] $ zipWith4 (\sgn denom eul twos -> sgn * (eul % (twos * denom))) (cycle [1, -1]) (skipOdds factorial) (skipOdds euler) (iterate (4 *) 4) -- | Infinite sequence of approximate values of the Dirichlet @β@ function at -- positive even integer arguments, starting with @β(0)@. betasEven :: forall a. (Floating a, Ord a) => a -> [a] betasEven eps = (1 / 2) : hurwitz where hurwitz :: [a] hurwitz = zipWith3 (\quarter threeQuarters four -> (quarter - threeQuarters) / four) (tail . skipOdds $ zetaHurwitz eps 0.25) (tail . skipOdds $ zetaHurwitz eps 0.75) (iterate (16 *) 16) -- | Infinite sequence of approximate (up to given precision) -- values of Dirichlet beta-function at integer arguments, starting with @β(0)@. -- -- >>> take 5 (betas 1e-14) :: [Double] -- [0.5,0.7853981633974483,0.9159655941772189,0.9689461462593694,0.9889445517411051] betas :: (Floating a, Ord a) => a -> [a] betas eps = e : o : scanl1 f (intertwine es os) where e : es = betasEven eps o : os = map (getRationalLimit (\a b -> abs (a - b) < eps) . rationalApproximations) betasOdd -- Cap-and-floor to improve numerical stability: -- 1 > beta(n + 1) - 1 > (beta(n) - 1) / 2 -- A similar method is used in @Math.NumberTheory.Zeta.Riemann.zetas@. f x y = 1 `min` (y `max` (1 + (x - 1) / 2))