Simultaneous torsion points in a Weierstrass family of elliptic curves. Myrto Mavraki University of British Columbia University of Michigan, December 5, 2015 Motivation Theorem (Lang, 1965. Ihara-Serre-Tate) Let C be an irreducible complex plane curve which contains infinitely many points with both coordinates roots of unity. Then the curve C is given by an equation of the form x n y m − ω = 0, for n, m ∈ Z and a root of unity ω. Motivation Theorem (Lang, 1965. Ihara-Serre-Tate) Let C be an irreducible complex plane curve which contains infinitely many points with both coordinates roots of unity. Then the curve C is given by an equation of the form x n y m − ω = 0, for n, m ∈ Z and a root of unity ω. Let µ denote the set of roots of unity. Motivation Theorem (Lang, 1965. Ihara-Serre-Tate) Let C be an irreducible complex plane curve which contains infinitely many points with both coordinates roots of unity. Then the curve C is given by an equation of the form x n y m − ω = 0, for n, m ∈ Z and a root of unity ω. Let µ denote the set of roots of unity. (ξ, ζ) ∈ µ × µ ↔ torsion points of G2m ↔ special points Motivation Theorem (Lang, 1965. Ihara-Serre-Tate) Let C be an irreducible complex plane curve which contains infinitely many points with both coordinates roots of unity. Then the curve C is given by an equation of the form x n y m − ω = 0, for n, m ∈ Z and a root of unity ω. Let µ denote the set of roots of unity. (ξ, ζ) ∈ µ × µ ↔ torsion points of G2m ↔ Curves ↔ translates of algebraic subgroups ↔ x ny m − ω = 0 of G2m by a torsion point special points special curves Motivation Theorem (Lang, 1965. Ihara-Serre-Tate) Let C be an irreducible complex plane curve which contains infinitely many points with both coordinates roots of unity. Then the curve C is given by an equation of the form x n y m − ω = 0, for n, m ∈ Z and a root of unity ω. Let µ denote the set of roots of unity. (ξ, ζ) ∈ µ × µ ↔ torsion points of G2m ↔ Curves ↔ translates of algebraic subgroups ↔ x ny m − ω = 0 of G2m by a torsion point special points special curves Philosophical restatement If a curve has an infinite (Zariski dense) set of special points, then it is a special curve. Dynamical reformulation The map φ : P1 → P1 given by [x : y ] 7→ [x 2 : y 2 ] induces a map φ × φ : P1 × P1 → P1 × P1 as follows. (P, Q) 7→ (φ(P), φ(Q)). Dynamical reformulation The map φ : P1 → P1 given by [x : y ] 7→ [x 2 : y 2 ] induces a map φ × φ : P1 × P1 → P1 × P1 as follows. (P, Q) 7→ (φ(P), φ(Q)). e×µ e where µ e := µ ∪ {0, ∞}. Prepφ×φ (P1 × P1 ) = µ Dynamical reformulation The map φ : P1 → P1 given by [x : y ] 7→ [x 2 : y 2 ] induces a map φ × φ : P1 × P1 → P1 × P1 as follows. (P, Q) 7→ (φ(P), φ(Q)). e×µ e where µ e := µ ∪ {0, ∞}. Prepφ×φ (P1 × P1 ) = µ Special points ↔ (P, Q) ∈ Prepφ×φ . Dynamical reformulation The map φ : P1 → P1 given by [x : y ] 7→ [x 2 : y 2 ] induces a map φ × φ : P1 × P1 → P1 × P1 as follows. (P, Q) 7→ (φ(P), φ(Q)). e×µ e where µ e := µ ∪ {0, ∞}. Prepφ×φ (P1 × P1 ) = µ Special points Special curves ↔ ↔ (P, Q) ∈ Prepφ×φ . preperiodic curves under the action of φ × φ. Dynamical reformulation The map φ : P1 → P1 given by [x : y ] 7→ [x 2 : y 2 ] induces a map φ × φ : P1 × P1 → P1 × P1 as follows. (P, Q) 7→ (φ(P), φ(Q)). e×µ e where µ e := µ ∪ {0, ∞}. Prepφ×φ (P1 × P1 ) = µ Special points Special curves ↔ ↔ (P, Q) ∈ Prepφ×φ . preperiodic curves under the action of φ × φ. Reformulation of Lang’s theorem. A curve C ⊂ P1 × P1 contains infinitely many (a Zariski dense set) elements in Prepφ×φ if and only if C is a preperiodic curve under the action of φ × φ. Questions that arise Question What if φ : P1 → P1 is a Lattès map corresponding to an elliptic curve E ? For example, φ could be the Lattès map, corresponding to multiplication by 2 on E . When does a curve C ⊂ P1 × P1 contain infinitely many points in Prepφ×φ ? Questions that arise Question What if φ : P1 → P1 is a Lattès map corresponding to an elliptic curve E ? For example, φ could be the Lattès map, corresponding to multiplication by 2 on E . When does a curve C ⊂ P1 × P1 contain infinitely many points in Prepφ×φ ? Special points ↔ points in Prepφ×φ ↔ (P, Q) ∈ (E × E )tors . Questions that arise Question What if φ : P1 → P1 is a Lattès map corresponding to an elliptic curve E ? For example, φ could be the Lattès map, corresponding to multiplication by 2 on E . When does a curve C ⊂ P1 × P1 contain infinitely many points in Prepφ×φ ? Special points ↔ points in Prepφ×φ ↔ (P, Q) ∈ (E × E )tors . Theorem (Laurent, McQuillan, Raynaud. Manin-Mumford conjecture) If an irreducible subvariety of a semiabelian variety contains a Zariski dense set of torsion points, then it is a translate of an algebraic subgroup by a torsion point. Questions that arise Question What if φ : P1 → P1 is a Lattès map corresponding to an elliptic curve E ? For example, φ could be the Lattès map, corresponding to multiplication by 2 on E . When does a curve C ⊂ P1 × P1 contain infinitely many points in Prepφ×φ ? Special points ↔ points in Prepφ×φ ↔ (P, Q) ∈ (E × E )tors . Theorem (Laurent, McQuillan, Raynaud. Manin-Mumford conjecture) If an irreducible subvariety of a semiabelian variety contains a Zariski dense set of torsion points, then it is a translate of an algebraic subgroup by a torsion point. Lang’s result is the case A = G2m . Our question is the case A = E × E. Pink-Zilber conjecture, an example Pink-Zilber conjecture, an example Consider the family of elliptic curves Eλ : y 2 = x(x − 1)(x − λ), λ ∈ C \ {0, 1}. and the two sections Pλ = (2, p p 2(2 − λ)), Qλ = (3, 6(3 − λ)). Pink-Zilber conjecture, an example Consider the family of elliptic curves Eλ : y 2 = x(x − 1)(x − λ), λ ∈ C \ {0, 1}. and the two sections Pλ = (2, p p 2(2 − λ)), Qλ = (3, 6(3 − λ)). Let T (2) = {λ ∈ C \ {0, 1} | Pλ ∈ (Eλ )tors }; similarly T (3). Pink-Zilber conjecture, an example Consider the family of elliptic curves Eλ : y 2 = x(x − 1)(x − λ), λ ∈ C \ {0, 1}. and the two sections Pλ = (2, p p 2(2 − λ)), Qλ = (3, 6(3 − λ)). Let T (2) = {λ ∈ C \ {0, 1} | Pλ ∈ (Eλ )tors }; similarly T (3). Remark 1 There are infinitely many λ ∈ T (2) (similarly for T (3)). Pink-Zilber conjecture, an example Consider the family of elliptic curves Eλ : y 2 = x(x − 1)(x − λ), λ ∈ C \ {0, 1}. and the two sections Pλ = (2, p p 2(2 − λ)), Qλ = (3, 6(3 − λ)). Let T (2) = {λ ∈ C \ {0, 1} | Pλ ∈ (Eλ )tors }; similarly T (3). Remark 1 There are infinitely many λ ∈ T (2) (similarly for T (3)). 2 Pλ ∈ Eλ is not identically torsion. Same for Qλ . Pink-Zilber conjecture, an example Consider the family of elliptic curves Eλ : y 2 = x(x − 1)(x − λ), λ ∈ C \ {0, 1}. and the two sections Pλ = (2, p p 2(2 − λ)), Qλ = (3, 6(3 − λ)). Let T (2) = {λ ∈ C \ {0, 1} | Pλ ∈ (Eλ )tors }; similarly T (3). Remark 1 There are infinitely many λ ∈ T (2) (similarly for T (3)). 2 Pλ ∈ Eλ is not identically torsion. Same for Qλ . 3 {Pλ }{λ∈C} and {Qλ }{λ∈C} are linearly independent over Z: there are no (n, m) ∈ Z \ {(0, 0)} such that [n]Pλ + [m]Qλ = O. Masser and Zannier’s theorem Question (Masser) Can T (2) ∩ T (3) be infinite? Masser and Zannier’s theorem Question (Masser) Can T (2) ∩ T (3) be infinite? Theorem (Masser-Zannier, 2010) There are at most finitely many λ ∈ C such that Pλ and Qλ are simultaneously torsion points for Eλ . Masser and Zannier’s theorem Question (Masser) Can T (2) ∩ T (3) be infinite? Theorem (Masser-Zannier, 2010) There are at most finitely many λ ∈ C such that Pλ and Qλ are simultaneously torsion points for Eλ . T (2) (respectively T (3)) is infinite Masser and Zannier’s theorem Question (Masser) Can T (2) ∩ T (3) be infinite? Theorem (Masser-Zannier, 2010) There are at most finitely many λ ∈ C such that Pλ and Qλ are simultaneously torsion points for Eλ . T (2) (respectively T (3)) is infinite but sparse; Masser and Zannier’s theorem Question (Masser) Can T (2) ∩ T (3) be infinite? Theorem (Masser-Zannier, 2010) There are at most finitely many λ ∈ C such that Pλ and Qλ are simultaneously torsion points for Eλ . T (2) (respectively T (3)) is infinite but sparse; think bounded Weil height. Masser and Zannier’s theorem Question (Masser) Can T (2) ∩ T (3) be infinite? Theorem (Masser-Zannier, 2010) There are at most finitely many λ ∈ C such that Pλ and Qλ are simultaneously torsion points for Eλ . T (2) (respectively T (3)) is infinite but sparse; think bounded Weil height. Double sparseness =⇒ finiteness. General theorem In fact a much more general statement is true. General theorem In fact a much more general statement is true. There is only one special case when there are infinitely many torsion parameters in the intersection. General theorem In fact a much more general statement is true. There is only one special case when there are infinitely many torsion parameters in the intersection. Theorem (Masser-Zannier, 2012) Let Pt = (a(t), ∗), Qt = (b(t), ∗) ∈ Et (C(t)). There are infinitely many t = λ ∈ C \ {0, 1} such that (Pλ , Qλ ) ∈ (Eλ × Eλ )tors if and only if there are (n, m) ∈ Z \ {(0, 0)}, such that [n]Pt + [m]Qt = O. General theorem In fact a much more general statement is true. There is only one special case when there are infinitely many torsion parameters in the intersection. Theorem (Masser-Zannier, 2012) Let Pt = (a(t), ∗), Qt = (b(t), ∗) ∈ Et (C(t)). There are infinitely many t = λ ∈ C \ {0, 1} such that (Pλ , Qλ ) ∈ (Eλ × Eλ )tors if and only if there are (n, m) ∈ Z \ {(0, 0)}, such that [n]Pt + [m]Qt = O. Masser-Zannier’s proof involves: • Model theory, results of Pila, Pila and Wilkie. • The geometry of an elliptic curve, uniformization map. Connection If Et = E is a constant family of elliptic curves, then Masser and Zannier’s theorem says The curve C = {(Pt , Qt ) : t ∈ C} contains infinitely many torsion points of E × E if and only if there are (n, m) ∈ Z \ {(0, 0)} such that [n]Pt + [m]Qt = O. Connection If Et = E is a constant family of elliptic curves, then Masser and Zannier’s theorem says The curve C = {(Pt , Qt ) : t ∈ C} contains infinitely many torsion points of E × E if and only if there are (n, m) ∈ Z \ {(0, 0)} such that [n]Pt + [m]Qt = O. This is a special case of the Manin-Mumford conjecture. Connection If Et = E is a constant family of elliptic curves, then Masser and Zannier’s theorem says The curve C = {(Pt , Qt ) : t ∈ C} contains infinitely many torsion points of E × E if and only if there are (n, m) ∈ Z \ {(0, 0)} such that [n]Pt + [m]Qt = O. This is a special case of the Manin-Mumford conjecture. Lang’s theorem also implied. Let a(t), b(t) ∈ C(t). The curve C = {(a(t), b(t)) : t ∈ C} contains infinitely many torsion points of G2m if and only if there are (n, m) ∈ Z \ {(0, 0)} such that a(t)n b(t)m = 1. Connection If Et = E is a constant family of elliptic curves, then Masser and Zannier’s theorem says The curve C = {(Pt , Qt ) : t ∈ C} contains infinitely many torsion points of E × E if and only if there are (n, m) ∈ Z \ {(0, 0)} such that [n]Pt + [m]Qt = O. This is a special case of the Manin-Mumford conjecture. Lang’s theorem also implied. Let a(t), b(t) ∈ C(t). The curve C = {(a(t), b(t)) : t ∈ C} contains infinitely many torsion points of G2m if and only if there are (n, m) ∈ Z \ {(0, 0)} such that a(t)n b(t)m = 1. Hence, Masser’s question can be viewed as a varying Manin-Mumford. Stoll’s result Stoll’s result 2 and 3 are different modulo 2. Stoll’s result 2 and 3 are different modulo 2. Fix an embedding ι : Q ,→ Q2 , and consider the reduction map ρ : P1 (Q2 ) → P1 (F2 ). Stoll’s result 2 and 3 are different modulo 2. Fix an embedding ι : Q ,→ Q2 , and consider the reduction map ρ : P1 (Q2 ) → P1 (F2 ). For a ∈ Q \ {0, 1}, write T (a) = {λ ∈ C \ {0, 1} : (a, p a(a − 1)(a − λ)) ∈ (Eλ )tors }. Stoll’s result 2 and 3 are different modulo 2. Fix an embedding ι : Q ,→ Q2 , and consider the reduction map ρ : P1 (Q2 ) → P1 (F2 ). For a ∈ Q \ {0, 1}, write T (a) = {λ ∈ C \ {0, 1} : (a, p a(a − 1)(a − λ)) ∈ (Eλ )tors }. Theorem (Stoll, 2014) Let α, β ∈ Q \ {0, 1} such that ρ(α) 6= ρ(β). Then T (α) ∩ T (β) ⊂ {α, β}. Stoll’s approach Examples 1 T (2) ∩ T (3) ⊂ {2, 3} and one can check T (2) ∩ T (3) = ∅. 2 For a primitive 3rd root of unity ω, T (ω) ∩ T (ω 2 ) ⊂ {ω, ω 2 } and one can check T (ω) ∩ T (ω 2 ) = {ω, ω 2 }. Stoll’s approach Examples 1 T (2) ∩ T (3) ⊂ {2, 3} and one can check T (2) ∩ T (3) = ∅. 2 For a primitive 3rd root of unity ω, T (ω) ∩ T (ω 2 ) ⊂ {ω, ω 2 } and one can check T (ω) ∩ T (ω 2 ) = {ω, ω 2 }. ψn (λ, x) := the n−th reduced division polynomial of Eλ Stoll’s approach Examples 1 T (2) ∩ T (3) ⊂ {2, 3} and one can check T (2) ∩ T (3) = ∅. 2 For a primitive 3rd root of unity ω, T (ω) ∩ T (ω 2 ) ⊂ {ω, ω 2 } and one can check T (ω) ∩ T (ω 2 ) = {ω, ω 2 }. ψn (λ, x) := the n−th reduced division polynomial of Eλ Its roots are the x coordinates of the points of Eλ whose order divides n and is > 2. Stoll’s approach Examples 1 T (2) ∩ T (3) ⊂ {2, 3} and one can check T (2) ∩ T (3) = ∅. 2 For a primitive 3rd root of unity ω, T (ω) ∩ T (ω 2 ) ⊂ {ω, ω 2 } and one can check T (ω) ∩ T (ω 2 ) = {ω, ω 2 }. ψn (λ, x) := the n−th reduced division polynomial of Eλ Its roots are the x coordinates of the points of Eλ whose order divides n and is > 2. Inductively, Stoll shows that ψn (λ, x) ≡ 2e(n) (λ − x 2 ) mod 2e(n)+1 Z[λ, x]. Stoll’s approach Examples 1 T (2) ∩ T (3) ⊂ {2, 3} and one can check T (2) ∩ T (3) = ∅. 2 For a primitive 3rd root of unity ω, T (ω) ∩ T (ω 2 ) ⊂ {ω, ω 2 } and one can check T (ω) ∩ T (ω 2 ) = {ω, ω 2 }. ψn (λ, x) := the n−th reduced division polynomial of Eλ Its roots are the x coordinates of the points of Eλ whose order divides n and is > 2. Inductively, Stoll shows that ψn (λ, x) ≡ 2e(n) (λ − x 2 ) mod 2e(n)+1 Z[λ, x]. λ ∈ T (α) \ {α} =⇒ ψn (λ, α) = 0 =⇒ ρ(λ) = ρ(α2 ). Dynamical reformulation of Masser’s question Consider the Lattès map induced by multiplication by 2 on Eλ (x 2 − λ)2 . 4x(x − 1)(x − λ) p Here fλ (a) is the x−coordinate of [2](a, a(a − 1)(a − λ)) ∈ Eλ . fλ (x) = Dynamical reformulation of Masser’s question Consider the Lattès map induced by multiplication by 2 on Eλ (x 2 − λ)2 . 4x(x − 1)(x − λ) p Here fλ (a) is the x−coordinate of [2](a, a(a − 1)(a − λ)) ∈ Eλ . fλ (x) = (a, p a(a − 1)(a − λ)) ∈ (Eλ )tors ⇐⇒ a ∈ Prepfλ Dynamical reformulation of Masser’s question Consider the Lattès map induced by multiplication by 2 on Eλ (x 2 − λ)2 . 4x(x − 1)(x − λ) p Here fλ (a) is the x−coordinate of [2](a, a(a − 1)(a − λ)) ∈ Eλ . fλ (x) = (a, p a(a − 1)(a − λ)) ∈ (Eλ )tors ⇐⇒ a ∈ Prepfλ Question (Reformulation) Are there infinitely many λ ∈ C such that 2, 3 ∈ Prepfλ ? Another approach and more questions DeMarco, Wang and Ye re-proved Masser and Zannier’s result using the equidistribution theorem of Baker-Rumely. Another approach and more questions DeMarco, Wang and Ye re-proved Masser and Zannier’s result using the equidistribution theorem of Baker-Rumely. The approach adopted by them was introduced by Baker and DeMarco for studying simultaneous preperiodic points for other 1-parameter families of rational maps, such as t 7→ z d + t. Recovering Stoll’s result (M.,2015) Let | · | := | · |2 be the 2−adic absolute value such that |2| = 12 . Recovering Stoll’s result (M.,2015) Let | · | := | · |2 be the 2−adic absolute value such that |2| = 12 . Consider a ∈ C2 \ {0, 1} with |a| ≤ 1, e.g. a = 2, 3. Recovering Stoll’s result (M.,2015) Let | · | := | · |2 be the 2−adic absolute value such that |2| = 12 . Consider a ∈ C2 \ {0, 1} with |a| ≤ 1, e.g. a = 2, 3. Analyse |fλn (a)|. Recovering Stoll’s result (M.,2015) Let | · | := | · |2 be the 2−adic absolute value such that |2| = 12 . Consider a ∈ C2 \ {0, 1} with |a| ≤ 1, e.g. a = 2, 3. Analyse |fλn (a)|. • If λ ∈ T (a) \ {a}, then |λ| ≤ 1. Recovering Stoll’s result (M.,2015) Let | · | := | · |2 be the 2−adic absolute value such that |2| = 12 . Consider a ∈ C2 \ {0, 1} with |a| ≤ 1, e.g. a = 2, 3. Analyse |fλn (a)|. • If λ ∈ T (a) \ {a}, then |λ| ≤ 1. Assume to the contrary that |λ| > 1. Recovering Stoll’s result (M.,2015) Let | · | := | · |2 be the 2−adic absolute value such that |2| = 12 . Consider a ∈ C2 \ {0, 1} with |a| ≤ 1, e.g. a = 2, 3. Analyse |fλn (a)|. • If λ ∈ T (a) \ {a}, then |λ| ≤ 1. Assume to the contrary that |λ| > 1. Then |fλ (a)| = 4|λ|2 |a||a−1|||λ| ≥ 4|λ|. Recovering Stoll’s result (M.,2015) Let | · | := | · |2 be the 2−adic absolute value such that |2| = 12 . Consider a ∈ C2 \ {0, 1} with |a| ≤ 1, e.g. a = 2, 3. Analyse |fλn (a)|. • If λ ∈ T (a) \ {a}, then |λ| ≤ 1. Assume to the contrary that |λ| > 1. 4|λ|2 |a||a−1|||λ| ≥ 4|λ|. |fλn+1 (a)| = 4|fλn (a)|, Then |fλ (a)| = Inductively, for n ≥ 1. Recovering Stoll’s result (M.,2015) Let | · | := | · |2 be the 2−adic absolute value such that |2| = 12 . Consider a ∈ C2 \ {0, 1} with |a| ≤ 1, e.g. a = 2, 3. Analyse |fλn (a)|. • If λ ∈ T (a) \ {a}, then |λ| ≤ 1. Assume to the contrary that |λ| > 1. 4|λ|2 |a||a−1|||λ| ≥ 4|λ|. |fλn+1 (a)| = 4|fλn (a)|, Then |fλ (a)| = Inductively, for n ≥ 1. This contradicts the fact that λ ∈ T (a) and fλ (a) 6= ∞. Recovering Stoll’s result (M.,2015) • If λ ∈ T (a) \ {a}, then |λ − a2 | < 1. Recovering Stoll’s result (M.,2015) • If λ ∈ T (a) \ {a}, then |λ − a2 | < 1. Assume to the contrary that |λ − a2 | ≥ 1. Recovering Stoll’s result (M.,2015) • If λ ∈ T (a) \ {a}, then |λ − a2 | < 1. Assume to the contrary that |λ − a2 | ≥ 1. Since |λ| ≤ 1, we get |fλ (a)| = 4|a2 −λ|2 |a||a−1||a−λ| ≥ 4. Recovering Stoll’s result (M.,2015) • If λ ∈ T (a) \ {a}, then |λ − a2 | < 1. Assume to the contrary that |λ − a2 | ≥ 1. Since |λ| ≤ 1, we get |fλ (a)| = 4|a2 −λ|2 |a||a−1||a−λ| Inductively, |fλn+1 (a)| = 4|fλn (a)|. ≥ 4. Recovering Stoll’s result (M.,2015) • If λ ∈ T (a) \ {a}, then |λ − a2 | < 1. Assume to the contrary that |λ − a2 | ≥ 1. Since |λ| ≤ 1, we get |fλ (a)| = 4|a2 −λ|2 |a||a−1||a−λ| Inductively, |fλn+1 (a)| = 4|fλn (a)|. Contradiction. ≥ 4. A Weierstrass family Eλ : y 2 = x 3 + λ, λ ∈ C \ {0} A Weierstrass family Eλ : y 2 = x 3 + λ, λ ∈ C \ {0} p T (a) = {λ ∈ C \ {0} : (a, a3 + λ) ∈ (Eλ )tors } A Weierstrass family Eλ : y 2 = x 3 + λ, λ ∈ C \ {0} p T (a) = {λ ∈ C \ {0} : (a, a3 + λ) ∈ (Eλ )tors } The Lattès map induced by multiplication by 2 in Eλ is fλ (x) = x 4 − 8λx . 4(x 3 + λ) A Weierstrass family Eλ : y 2 = x 3 + λ, λ ∈ C \ {0} p T (a) = {λ ∈ C \ {0} : (a, a3 + λ) ∈ (Eλ )tors } The Lattès map induced by multiplication by 2 in Eλ is fλ (x) = x 4 − 8λx . 4(x 3 + λ) Let a ∈ C2 . We study the 2−adic absolute value of elements in T (a), by looking at {|fλn (a)|2 }n∈N . More impossible intersections Theorem (M.,2015) If α, β ∈ Q \ {0} are such that α β ∈ / {−2, − 12 }, then T (α) ∩ T (β) = ∅. More impossible intersections Theorem (M.,2015) If α, β ∈ Q \ {0} are such that α β ∈ / {−2, − 12 }, then T (α) ∩ T (β) = ∅. Moreover, for all a ∈ Q \ {0} we have T (a) ∩ T (−2a) = {−a3 }. More impossible intersections Theorem (M.,2015) If α, β ∈ Q \ {0} are such that α β ∈ / {−2, − 12 }, then T (α) ∩ T (β) = ∅. Moreover, for all a ∈ Q \ {0} we have T (a) ∩ T (−2a) = {−a3 }. Theorem (M.,2015) If α, β ∈ Q2 are such that gcd(6, e(Q2 (α, β)|Q2 )) = 1, |α| = 6 |β| α3 1 and β 3 ∈ / {−8, − 8 }, then we have that T (α) ∩ T (β) = ∅. Moreover, for all a ∈ C2 \ {0} we have T (a) ∩ T (−2a) = {−a3 }. A trichotomy and more impossible intersections Theorem (M.,2015) n o α3 3 Let λ ∈ T (α). Then either λ ∈ −α , 8 or m m |λ| ∈ {4|α|3 , 41−(1/4) |α|3 , 22+(1/4) |α|3 : m ∈ N≥1 }. Moreover, exactly one of the following is true. 1 |λ| = 4|α|3 ⇐⇒ 0, ∞ ∈ / Ofλ (α). 2 |λ| = 41−(1/4) |α|3 for some m ∈ N≥1 , or λ = −α3 ⇐⇒ ∞ ∈ Ofλ (α). 3 |λ| = 22+(1/4) |α|3 for some m ∈ N≥1 , or λ = 0 ∈ Ofλ (α). m m α3 8 ⇐⇒ The end Thank you for your attention!