A minimal complex differential equation descends to the series coefficient field
ProvedArithmeticE.minimal_operator_coefficient_descentLet be a field embedded in , and let . Suppose the image of in has a minimal polynomial differential equation of order . Then it has a minimal equation of order with coefficients in . One coefficient of its leading polynomial can moreover be normalized to one.
Choose a nonzero coefficient in the leading polynomial of , and a -linear functional taking to one. Apply coefficientwise to all coefficient polynomials of . Every formal coefficient equation is -linear, so the projected operator still annihilates . Its leading polynomial is nonzero by the chosen normalization. Mapping the equation back to gives the same order , and minimality excludes every smaller equation.
The proof does not require to be algebraic, or convergence or arithmetic bounds on . In the E-function application, it justifies working over the actual arithmetic coefficient field rather than assuming that a complex minimal operator already has arithmetic coefficients.
import Definitions.Def_beukersLiftingData open ArithmeticE
theorem ArithmeticE.minimal_operator_coefficient_descent {K : Type*} [Field K] [Algebra K ℂ] (f : PowerSeries K)
(p : ℕ → Polynomial ℂ) (n : ℕ)
(hm : MinimalEquation p n (f.map (algebraMap K ℂ))) :
∃ q : ℕ → Polynomial K,
MinimalEquation (fun k => (q k).map (algebraMap K ℂ)) n (f.map (algebraMap K ℂ)) ∧
∃ d, (q n).coeff d = 1 := by sorry