import OAI.NumberTheory.SingleFold.DescentMap namespace OAI namespace SingleFold.SquareSupport lemma squarefree_decomposition {x : ℚ} (hx : x ≠ 0) : ∃ (d : ℕ) (r : ℚ), 0 < d ∧ r ≠ 0 ∧ Squarefree d ∧ |x| = d*r^2 := by have hn : 0 < x.num.natAbs := Int.natAbs_pos.mpr (Rat.num_ne_zero.mpr hx) obtain ⟨d,k,hd,hk,he,hf⟩ := Nat.sq_mul_squarefree_of_pos (Nat.mul_pos hn x.den_pos) refine ⟨d,(k:ℚ)/x.den,hd,div_ne_zero (by exact_mod_cast hk.ne') (by exact_mod_cast x.den_ne_zero),hf,?_⟩ have he' : (k:ℚ)^2*d=x.num.natAbs*x.den := by exact_mod_cast he rw [Rat.abs_def, Rat.divInt_eq_div] push_cast field_simp have habs : |(x.num:ℚ)| = (x.num.natAbs:ℚ) := by simp rw [habs] nlinarith only [he'] lemma val_add_eq {p : ℕ} [Fact p.Prime] {x y : ℚ} (hx : x ≠ 0) (hy : y ≠ 0) (hval : padicValRat p x ≠ padicValRat p y) : padicValRat p (x+y) = min (padicValRat p x) (padicValRat p y) := by apply padicValRat.add_eq_min _ hx hy hval intro h apply hval have he : x = -y := by linarith simp [he] lemma even_val_triple {p : ℕ} [Fact p.Prime] {x y a b : ℚ} (hx : x ≠ 0) (hxa : x+a ≠ 0) (hxb : x+b ≠ 0) (ha : a ≠ 0) (hb : b ≠ 0) (hva : padicValRat p a = 0) (hvb : padicValRat p b = 0) (he : y^2=x*(x+a)*(x+b)) : Even (padicValRat p x) := by have hv := congrArg (padicValRat p) he rw [padicValRat.pow, padicValRat.mul (mul_ne_zero hx hxa) hxb, padicValRat.mul hx hxa] at hv by_cases hz : padicValRat p x=0 · simp [hz] have h₁ := val_add_eq (p := p) hx ha (by simpa [hva] using hz) have h₂ := val_add_eq (p := p) hx hb (by simpa [hvb] using hz) rw [hva] at h₁ rw [hvb] at h₂ rcases lt_or_gt_of_ne hz with hl | hl · rw [min_eq_left hl.le] at h₁ h₂ rw [h₁,h₂] at hv refine ⟨padicValRat p y - padicValRat p x, ?_⟩ omega · rw [min_eq_right hl.le] at h₁ h₂ rw [h₁,h₂] at hv exact ⟨padicValRat p y, by omega⟩ lemma squarefree_support {x r : ℚ} {d D : ℕ} (hd : 0 < d) (hr : r ≠ 0) (hf : Squarefree d) (he : |x| = d*r^2) (hD : D ≠ 0) (hv : ∀ p : ℕ, p.Prime → ¬p ∣ D → Even (padicValRat p x)) : d ∣ D := by apply (Nat.factorization_le_iff_dvd hd.ne' hD).mp intro p by_cases hp : p.Prime · by_cases hpd : p ∣ d · have hdd := Nat.factorization_eq_one_of_squarefree hf hp hpd have hpd' : p ∣ D := by by_contra hpD have : Fact p.Prime := ⟨hp⟩ have hxval := congrArg (padicValRat p) he have habs : padicValRat p |x| = padicValRat p x := by rcases le_or_gt 0 x with hs | hs · rw [abs_of_nonneg hs] · rw [abs_of_neg hs, padicValRat.neg] rw [habs, padicValRat.mul (by exact_mod_cast hd.ne') (pow_ne_zero _ hr), padicValRat.of_nat, padicValRat.pow, ← Nat.factorization_def d hp, hdd] at hxval obtain ⟨k,hk⟩ := hv p hp hpD omega rw [hdd] exact (hp.dvd_iff_one_le_factorization hD).mp hpd' · simp [Nat.factorization_eq_zero_of_not_dvd hpd] · simp [Nat.factorization_eq_zero_of_not_prime d hp] lemma supported_decomposition {x : ℚ} {D : ℕ} (hx : x ≠ 0) (hD : D ≠ 0) (hv : ∀ p : ℕ, p.Prime → ¬p ∣ D → Even (padicValRat p x)) : ∃ (d : ℕ) (r : ℚ), d ∣ D ∧ r ≠ 0 ∧ (x=d*r^2 ∨ x=-(d*r^2)) := by obtain ⟨d,r,hd,hr,hf,he⟩ := squarefree_decomposition hx refine ⟨d,r,squarefree_support hd hr hf he hD hv,hr,?_⟩ rcases le_or_gt 0 x with hs | hs · exact Or.inl (by simpa [abs_of_nonneg hs] using he) · exact Or.inr (by rw [abs_of_neg hs] at he; linarith) end SingleFold.SquareSupport namespace SingleFold.Descent open WeierstrassCurve WeierstrassCurve.Affine WeierstrassCurve.Affine.Point open SquareSupport lemma supported_zero (P : E.Point) : ∃ (d : ℕ) (r : ℚ), d ∣ 5 ∧ r ≠ 0 ∧ (alphaVal 0 P=d*r^2 ∨ alphaVal 0 P=-(d*r^2)) := by cases P with | zero => exact ⟨1,1,by norm_num,by norm_num,Or.inl (by norm_num [alphaVal])⟩ | some x y h => by_cases hx : x=0 · exact ⟨1,5,by norm_num,by norm_num,Or.inr (by norm_num [alphaVal,adjusted,hx])⟩ by_cases hx5 : x=5 · exact ⟨5,1,by norm_num,by norm_num,Or.inl (by norm_num [alphaVal,adjusted,hx5])⟩ by_cases hxm5 : x= -5 · exact ⟨5,1,by norm_num,by norm_num,Or.inr (by norm_num [alphaVal,adjusted,hxm5])⟩ simp only [alphaVal_some, sub_zero, adjusted, ite_eq_right hx] apply supported_decomposition hx (by norm_num) intro p hp hp5 have : Fact p.Prime := ⟨hp⟩ have hv : padicValRat p (5:ℚ)=0 := by rw [show (5:ℚ)=(5:ℕ) from rfl, padicValRat.of_nat, padicValNat.eq_zero_of_not_dvd hp5] rfl apply even_val_triple (a := -5) (b := 5) hx (by simpa only [sub_eq_add_neg] using sub_ne_zero.mpr hx5) (by intro hh; apply hxm5; linarith) (by norm_num) (by norm_num) (by simpa using hv) hv have he := equation h nlinarith only [he] lemma supported_five (P : E.Point) : ∃ (d : ℕ) (r : ℚ), d ∣ 10 ∧ r ≠ 0 ∧ (alphaVal 5 P=d*r^2 ∨ alphaVal 5 P=-(d*r^2)) := by cases P with | zero => exact ⟨1,1,by norm_num,by norm_num,Or.inl (by norm_num [alphaVal])⟩ | some x y h => by_cases hx : x=0 · exact ⟨5,1,by norm_num,by norm_num,Or.inr (by norm_num [alphaVal,adjusted,hx])⟩ by_cases hx5 : x=5 · exact ⟨2,5,by norm_num,by norm_num,Or.inl (by norm_num [alphaVal,adjusted,hx5])⟩ by_cases hxm5 : x= -5 · exact ⟨10,1,by norm_num,by norm_num,Or.inr (by norm_num [alphaVal,adjusted,hxm5])⟩ have hx5' : x-5 ≠ 0 := sub_ne_zero.mpr hx5 simp only [alphaVal_some, adjusted, ite_eq_right hx5'] apply supported_decomposition hx5' (by norm_num) intro p hp hp10 have : Fact p.Prime := ⟨hp⟩ have hv (k : ℕ) (hk : k ∣ 10) : padicValRat p (k:ℚ)=0 := by rw [padicValRat.of_nat, padicValNat.eq_zero_of_not_dvd (fun hh => hp10 (hh.trans hk))] rfl apply even_val_triple (a := 5) (b := 10) hx5' (by simpa using hx) (by intro hh; apply hxm5; linarith) (by norm_num) (by norm_num) (hv 5 (by norm_num)) (hv 10 (by norm_num)) have he := equation h nlinarith only [he] lemma classOf_mul_sq {d r : ℚ} (hd : d ≠ 0) (hr : r ≠ 0) : classOf (d*r^2)=classOf d := by rw [pow_two, classOf_mul hd (mul_ne_zero hr hr), classOf_square, mul_one] lemma alpha_zero_classes (P : E.Point) : classOf (alphaVal 0 P) = classOf 1 ∨ classOf (alphaVal 0 P) = classOf (-1) ∨ classOf (alphaVal 0 P) = classOf 5 ∨ classOf (alphaVal 0 P) = classOf (-5) := by obtain ⟨d,r,hd,hr,he | he⟩ := supported_zero P all_goals rcases (Nat.dvd_prime (by norm_num : Nat.Prime 5)).mp hd with rfl | rfl · left; rw [he]; exact classOf_mul_sq (by norm_num) hr · right; right; left; rw [he]; exact classOf_mul_sq (by norm_num) hr · right; left; rw [he, ← neg_mul]; exact classOf_mul_sq (by norm_num) hr · right; right; right; rw [he, ← neg_mul]; exact classOf_mul_sq (by norm_num) hr end SingleFold.Descent end OAI