Lemma 3.7 — properties of Hasse derivatives
ProvedDerandLDT.lemma_3_7low-degree-testingp2o-gran-per-chapterp2o-plan-paperp2o-v1polynomialsresultants
Let be a field of characteristic and write for the first-order Hasse derivative on (Definition 3.6: is the coefficient of in , which is the formal derivative). Then:
- if is nonzero of degree at least one, then if and only if for some ;
- for all , ;
- for every and , .
These facts control when a restriction of a polynomial to a line is a -th power (Lemma 3.9) and when the -derivative of the interpolating polynomial is nonzero.
Formalization Note is Polynomial.hasseDeriv 1 and is Polynomial.expand F p g; the characteristic is given by CharP F p.
Differences from the source:
- encoding: a field of characteristic p → CharP F p for any natural number p (p = 0 is characteristic zero); reason: in characteristic zero both sides of item 1 are false for deg f ≥ 1, so the item holds there as well
Preamble
import Mathlib set_option autoImplicit false universe u
Formal statement
namespace DerandLDT
open Polynomial
/-- **Lemma 3.7** (properties of Hasse derivatives). Harsha, Kopparty, Kumar, Saptharishi, Sudan, *On derandomizing the line-point
low-degree test*, STOC 2027 submission manuscript (sha256 d57e7d6f…80df), Lemma 3.7, pp.11–12: *Let `F` be a
field of characteristic `p`. (1) If `f ∈ F[X]` is nonzero of degree at least one, its first order
Hasse derivative is identically zero iff `f(X) = g(X^p)` for some `g ∈ F[X]`. (2) For any two
polynomials, `∂_X(fg) = g·∂_X(f) + f·∂_X(g)`; by induction, `∂_X(fⁿ) = n fⁿ⁻¹ ∂_X(f)` for every
`n ∈ ℕ`.*
**Formalization Note.** `∂_X` is the first Hasse derivative `hasseDeriv 1` (Definition 3.6, p.11),
which equals the formal derivative. `g(X^p)` is `expand F p g`. The characteristic is any `p` with
`CharP F p`; in characteristic `0` both sides of (1) are false for `deg f ≥ 1`. For `n = 0` the
power rule reads `0 = 0`. -/
theorem lemma_3_7 (F : Type u) [Field F] (p : ℕ) [CharP F p] :
(∀ f : F[X], f ≠ 0 → 1 ≤ f.natDegree →
(hasseDeriv 1 f = 0 ↔ ∃ g : F[X], expand F p g = f)) ∧
(∀ f g : F[X], hasseDeriv 1 (f * g) = g * hasseDeriv 1 f + f * hasseDeriv 1 g) ∧
(∀ (f : F[X]) (n : ℕ), hasseDeriv 1 (f ^ n) = (n : F[X]) * f ^ (n - 1) * hasseDeriv 1 f) := by sorry
end DerandLDT
Source
Harsha–Kopparty–Kumar–Saptharishi–Sudan, On derandomizing the line-point low-degree test, STOC 2027 submission (manuscript sha256 d57e7d6f…80df), Lemma 3.7, pp.11–12
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.