A generalized Montgomery model can be rescaled by a square: if `r² = B`
ProvedCryptography.IsogenySIDH.genMont_untwistaether-catalogcryptography
A generalized Montgomery model can be rescaled by a square: if r² = B
then B y² = … becomes the untwisted equation for r y. (This reuses the
catalog's radicalNormalize.)
theorem Cryptography.IsogenySIDH.genMont_untwist{A B r : K} {P : K × K} (hr : r ^ 2 = B)
(hP : genMont B A P) : OnMontgomery A (radicalNormalize r P) := by sorry
Formalization Note Transplanted verbatim from the Aether Catalog source Cryptography/IsogenySIDH/RadicalMontgomeryFormula.lean; the statement is byte-identical to the source declaration, elaborated with autoImplicit disabled in the platform environment.
Preamble
-- Thm stub generated from Cryptography/IsogenySIDH/RadicalMontgomeryFormula.lean
import Mathlib
import Definitions.Def_Cryptography_IsogenySIDH_DeepRadicalMontgomery
import Definitions.Def_Cryptography_IsogenySIDH_RadicalMontgomery
import Definitions.Def_Cryptography_IsogenySIDH_RadicalMontgomeryFormula
/-
# The radical 2-isogeny formula on Montgomery curves
`Catalog/Cryptography/IsogenySIDH/RadicalMontgomery.lean` verified the *affine
quotient* of a Montgomery curve `E_A : y² = x³ + A x² + x` by its rational
two-torsion point `(0,0)`: the image satisfies the (non-Montgomery) equation
`Y² = X³ + A X² - 4 X - 4 A`. That is only half of a radical-isogeny step. The
genuinely *radical* half is the renormalisation of the image back into Montgomery
form, which requires extracting a square root.
This file supplies that missing half and therefore closes the loop:
* `genMont B A P` is the generalized (twisted) Montgomery equation
`B y² = x³ + A x² + x`. Allowing the twist coefficient `B` makes the
renormalisation *rational in a single radical* `α = √(A+2)`, which is exactly
the shape used by radical-isogeny algorithms.
* `radTwoParam A α = (A+6)/(2α)` is the **radical 2-isogeny parameter formula**.
* `radTwoIso α (x,y) = ((x-1)²/(2αx), y(x²-1)/x²)` is the associated explicit
rational map, and `radTwoIso_mem` is its correctness theorem.
* `radical_is_four_torsion_ordinate` explains *where the radical comes from*:
`α` is precisely the `y`-coordinate of an affine point of `E_A` lying above
`(0,0)` under duplication, i.e. of a point of order four. This is the
structural reason radical isogenies avoid square-root extraction at run time.
* `mont_normalisation_unique` shows the two sign choices `±α` exhaust all
Montgomery renormalisations, and `radTwoParam_neg` identifies the second one
with the quadratic twist.
* `radChain` iterates the step and `radChain_mem` verifies an entire chain by
induction, which is the algorithmic content of a radical-isogeny walk.
Everything is stated over an arbitrary field, so it applies verbatim to the
quadratic finite fields `𝔽_{p²}` on which supersingular isogeny cryptography
takes place.
-/
open Cryptography.IsogenySIDH
open Cryptography.IsogenySIDH
variable {K : Type*} [Field K]
/-! ## Generalized (twisted) Montgomery models -/Formal statement
theorem Cryptography.IsogenySIDH.genMont_untwist{A B r : K} {P : K × K} (hr : r ^ 2 = B)
(hP : genMont B A P) : OnMontgomery A (radicalNormalize r P) := by sorrySource