Characteristic `3`: backtracking at `j = 0`.
ProvedCryptography.IsogenySIDH.backtracking_at_zero_char_threeaether-catalogcryptography
Characteristic 3: backtracking at j = 0. In characteristic 3 the
parameter A = 0 satisfies btNum 0 = 132 = 0 and btDen 0 = 24 = 0, so the
radical walk returns to its starting j-invariant after two steps, and that
j-invariant is 0.
theorem Cryptography.IsogenySIDH.backtracking_at_zero_char_three{α : K} (h3 : (3 : K) = 0) (hα : α ≠ 0)
(hsq : α ^ 2 = (0 : K) + 2) :
jQuot (radTwoParam (0 : K) α) = jMont (0 : K) ∧ jMont (0 : K) = 0 := by sorry
Formalization Note Transplanted verbatim from the Aether Catalog source Cryptography/IsogenySIDH/BacktrackingCharacteristic.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/BacktrackingCharacteristic.lean
import Mathlib
import Definitions.Def_Cryptography_IsogenySIDH_ModularTwoIsogeny
import Definitions.Def_Cryptography_IsogenySIDH_RadicalMontgomeryFormula
import Definitions.Def_Cryptography_IsogenySIDH_RadicalNonBacktracking
/-
# The exceptional `j`-invariant `0` is a characteristic phenomenon
`RadicalNonBacktracking.lean` proved that, over a field containing no primitive
cube root of unity, a radical 2-isogeny walk can return to its starting
`j`-invariant after two steps only when
`j ∈ {0, -3375, 287496}`,
and it exhibited backtracking at `-3375` and at `287496`. The value `0` was
left as a *possible* exception: it enters the classification only through the
degenerate case `btDen A = 0` of the cube-root branch, and the previous cycle's
`FUTURE_DIRECTIONS.md` conjectured (Conjecture 2) that it too is attained.
This file settles that question, and the answer is a **correction** of the
conjecture:
* `btNum_btDen_no_common_root` — the two comparison polynomials
`btNum A = A² + 60A + 132` and `btDen A = 4(A²-3)(A-2)` have **no common
root** in any field whose characteristic avoids `2, 3, 5, 11`. The proof is
an explicit elimination: a common root forces `A = 2` (excluded, since
`btNum 2 = 2⁸`) or `A² = 3` together with `15(4A+9) = 0`, whence
`16·3 = (4A)² = 81`, i.e. `33 = 0`.
* `radical_two_step_nonbacktracking_sharp` — consequently the exceptional set
shrinks: away from characteristic `2, 3, 5, 11`, and over a field with no
primitive cube root of unity, two-step backtracking forces
`j ∈ {-3375, 287496}`. The value `0` is **not** exceptional.
* `radChain_two_step_nonbacktracking_sharp` — the walk-level version.
* Sharpness in the excluded characteristics: `backtracking_at_zero_char_three`
and `backtracking_at_sqrt_three_char_five` show that in characteristic `3`
(at `A = 0`) and in characteristic `5` (at `A² = 3`) backtracking at `j = 0`
really does occur, so the hypotheses `(3 : K) ≠ 0` and `(5 : K) ≠ 0` cannot be
dropped; `backtracking_at_six_char_eleven` shows that in characteristic `11`
the `j = 0` locus `A² = 3` collapses onto the principal branch `A = 6`, which
is why `11` also has to be excluded from the elimination even though it
produces no new exceptional value.
* `char_five_backtracking_exists` makes the characteristic-`5` counterexample
unconditional by producing it over `AlgebraicClosure (ZMod 5)`.
So the final statement of the two-step non-backtracking theorem is: outside of
characteristics `2, 3, 5, 11`, and outside of fields containing a primitive cube
root of unity, **exactly two** `j`-invariants can backtrack, namely the CM
values `-3375` (discriminant `-7`) and `287496` (discriminant `-16`).
-/
set_option maxHeartbeats 1000000
open Cryptography.IsogenySIDH
variable {K : Type*} [Field K]
/-! ## Eliminating the degenerate branch -/
/-! ## The sharpened classification -/
/-! ## Sharpness in the excluded characteristics -/Formal statement
theorem Cryptography.IsogenySIDH.backtracking_at_zero_char_three{α : K} (h3 : (3 : K) = 0) (hα : α ≠ 0)
(hsq : α ^ 2 = (0 : K) + 2) :
jQuot (radTwoParam (0 : K) α) = jMont (0 : K) ∧ jMont (0 : K) = 0 := by sorrySource