Points reducing to the singular point #
Let v be a valuation on a field F, with valuation ring O, maximal ideal 𝔪 and residue
field k, and let W be a Weierstrass curve over F with an integral model W_O over O. When
the reduced curve W_k is singular, the points of W(F) with nonsingular reduction form the
subgroup E₀(F) (WeierstrassCurve.Affine.nonsingularReduction). When F is complete for a
discrete valuation with perfect residue field and the equation is minimal, the index
[E(F) : E₀(F)] is the local index c_v that Tate's algorithm computes. This file gives
conditions on the equation under which no point of W(F) reduces to the singular point, so that
E₀(F) = W(F) and the index is one.
The input is a computation at an integral solution (x₀, y₀) of W_O. Translating it to the
origin by x ↦ x + x₀, y ↦ y + y₀ gives an equation with a₆ = 0, whose a₃ and -a₄ are
the values at (x₀, y₀) of the two partial derivatives of the Weierstrass polynomial, and whose
discriminant is that of W. For such an equation
Δ = b₂² (a₄² + a₁ a₃ a₄ − a₂ a₃²) − 8 (2 a₄ + a₁ a₃)³ − 27 a₃⁴ + 9 b₂ (2 a₄ + a₁ a₃) a₃²,
so Δ lies in the square of any ideal containing both partial values
(WeierstrassCurve.Affine.Δ_mem_sq_of_equation). If (x₀, y₀) reduces to
the singular point of W_k, both partial values lie in 𝔪, hence Δ ∈ 𝔪². If moreover the
singular point is a nonsplit node, the tangent form a₄² + a₁ a₃ a₄ − a₂ a₃² is anisotropic
modulo 𝔪, so its value is a unit times the square of a₃ or of a₄; the remaining terms are of
higher order, so Δ is a unit times a square. This gives three criteria for E₀(F) = W(F):
Δ ∉ 𝔪²; for a discrete valuation with normalized additive valuationordthis isord(Δ) ≤ 1, the reduction symbolsI₀andI₁;- the reduction is a nonsplit node and
Δis not a unit times a square; for a discrete valuation this is nonsplit multiplicative reduction of typeIₙwithn = ord(Δ)odd; - the reduction is singular at the origin and
a₆ ∉ 𝔪²; on such an equation withb₂ ∈ 𝔪this is Tate's test for the reduction symbolII.
In each case Tate's algorithm reports the local index 1 (Silverman, Advanced Topics, IV.9,
Table 4.1), and the theorems here show that E₀(F) = W(F) there, for an arbitrary valuation and
without completeness.
When the reduction is singular at the origin, that is a₃, a₄, a₆ ∈ 𝔪, the points outside E₀(F)
are the affine points (x, y) with x, y ∈ 𝔪, since the singular point of W_k is unique.
Main results #
WeierstrassCurve.Δ_mem_maximalIdeal_sq_of_isSingular: over a local ring, a solution reducing to a singular point forcesΔ ∈ 𝔪².WeierstrassCurve.exists_associated_sq_Δ_of_isSingular: over a valuation ring, a solution reducing to a nonsplit node forcesΔto be a unit times a square.WeierstrassCurve.mem_maximalIdeal_of_isSingular: over a local ring, if the reduction is singular at the origin, a point reducing to a singular point has coordinates in𝔪.WeierstrassCurve.a₆_mem_maximalIdeal_sq_of_isSingular: over a local ring, if the reduction is singular at the origin, a solution reducing to it forcesa₆ ∈ 𝔪².WeierstrassCurve.Affine.nonsingularReduction_eq_top_of_forall_not_isSingular: if no integral solution reduces to a singular point, thenE₀(F) = W(F).WeierstrassCurve.Affine.nonsingularReduction_eq_top_of_forall_mem_maximalIdeal_not_equation: if the reduction is singular at the origin and no integral solution has both coordinates in𝔪, thenE₀(F) = W(F).WeierstrassCurve.Affine.nonsingularReduction_eq_top_of_Δ_notMem,WeierstrassCurve.Affine.nonsingularReduction_eq_top_of_not_splitsandWeierstrassCurve.Affine.nonsingularReduction_eq_top_of_a₆_notMem: the three criteria above.WeierstrassCurve.Affine.some_notMem_nonsingularReduction_iff: when the reduction is singular at the origin, an affine point has singular reduction exactly when both of its coordinates lie in the maximal ideal, andWeierstrassCurve.Affine.exists_eq_some_of_notMem_nonsingularReductionwrites such a point with integral coordinates.WeierstrassCurve.Affine.some_coe_notMem_nonsingularReduction_iff: the same criterion for points with coordinates in the valuation ring.WeierstrassCurve.Affine.exists_chord_of_add_notMem_nonsingularReduction: the identities satisfied by the chord through two points whose sum has singular reduction.WeierstrassCurve.Affine.two_nsmul_notMem_of_slope_mem: twice a point with singular reduction whose tangent slope lies in the maximal ideal has singular reduction.
References #
- J. Silverman, The Arithmetic of Elliptic Curves, VII.2 and VII.6.
- J. H. Silverman, Advanced Topics in the Arithmetic of Elliptic Curves, GTM 151, IV.9.
- J. Tate, Algorithm for determining the type of a singular fibre in an elliptic pencil, in Modular Functions of One Variable IV, LNM 476 (1975), 33–52.
A solution reducing to a singular point forces Δ ∈ 𝔪². Over a local ring, if an integral
solution (x, y) of the Weierstrass equation reduces to a singular point of the reduced curve,
then the discriminant lies in the square of the maximal ideal.
A point reducing to the singular point at the origin has coordinates in the maximal
ideal. Over a local ring, suppose that a₃, a₄ and a₆ lie in the maximal ideal, so that the
reduced curve is singular at the origin. If (x, y) reduces to a singular point of the reduced
curve, that point is the origin, since the reduced curve has only one singular point, so x and y
lie in the maximal ideal.
A solution reducing to the singular point at the origin forces a₆ ∈ 𝔪². Over a local
ring, suppose that a₃, a₄ and a₆ lie in the maximal ideal, so that the reduced curve is
singular at the origin. If an integral solution (x, y) reduces to a singular point, that point is
the origin, so x and y lie in the maximal ideal, and then a₆ lies in its square.
Solutions reducing to a nonsplit node #
A solution reducing to a nonsplit node forces Δ to be a unit times a square. Over a
valuation ring, suppose that c₄ is a unit and that the node polynomial
(WeierstrassCurve.nodePolynomial) does not split over the residue field, so that a singular point
of the reduced curve is a node whose tangent directions are not rational. If an integral solution
(x, y) reduces to a singular point, then Δ is associated to the square of an element; for a
discrete valuation with normalized additive valuation ord, ord(Δ) is even.
Criteria for every point to have nonsingular reduction #
If no integral solution reduces to a singular point, every point has nonsingular
reduction: E₀(F) = W(F). Points whose x-coordinate has a pole reduce to the point at
infinity, and every other affine point is an integral solution of the integral model.
If Δ ∉ 𝔪², every point has nonsingular reduction: E₀(F) = W(F). For a discrete
valuation with normalized additive valuation ord the hypothesis is ord(Δ) ≤ 1, which on a
minimal equation is the reduction symbol I₀ or I₁.
If the reduction is singular at the origin and no integral solution lies in 𝔪 × 𝔪, every
point has nonsingular reduction: E₀(F) = W(F). Here a₃, a₄ and a₆ of the integral model
lie in the maximal ideal, so that the singular point of the reduced curve is the origin, and the
points reducing to it are the integral solutions with both coordinates in the maximal ideal.
At a nonsplit node with Δ not a unit times a square, every point has nonsingular
reduction: E₀(F) = W(F). Here the node polynomial (WeierstrassCurve.nodePolynomial) of the
integral model does not split over the residue field (so c₄ is a unit, since the node polynomial
is constant otherwise), so the singular point of the reduced curve, if any, is a node without
rational tangent directions, and the discriminant of the integral model is not associated to a
square. For a discrete valuation, on a minimal
equation, this is nonsplit multiplicative reduction of type Iₙ with n odd.
If the reduction is singular at the origin and a₆ ∉ 𝔪², every point has nonsingular
reduction: E₀(F) = W(F). Here a₃, a₄ and a₆ of the integral model lie in the maximal
ideal, so that the reduced curve is singular at the origin. For a discrete valuation, on an
equation with b₂ ∈ 𝔪 as well, these are the conditions at which Tate's algorithm stops with
the symbol II.
Points with singular reduction #
When the reduction is singular at the origin, the points with singular reduction are the
affine points with both coordinates in the maximal ideal. Here a₃, a₄ and a₆ of the
integral model lie in the maximal ideal, and (x, y) lies outside E₀(F) exactly when v x < 1
and v y < 1.
When the reduction is singular at the origin, a point with integral coordinates has singular
reduction exactly when both coordinates lie in the maximal ideal. This is
WeierstrassCurve.Affine.some_notMem_nonsingularReduction_iff for coordinates in the valuation
ring.
When the reduction is singular at the origin, a point with singular reduction is an affine
point whose coordinates lie in the maximal ideal of the valuation ring. Here a₃, a₄ and a₆
of the integral model lie in the maximal ideal.
The chord through two points whose sum has singular reduction. Let (x₁, y₁) and
(x₂, y₂) be points of W(F) with integral coordinates, at a reduction singular at the origin, and
suppose that their sum has singular reduction. Then the sum is an affine point (x₃, y₃) with
x₃, y₃ ∈ 𝔪, and the slope ℓ = W.slope x₁ x₂ y₁ y₂ of the line through the two points is
integral, with
ℓ² + a₁ ℓ − a₂ = x₁ + x₂ + x₃,y₁ − y₂ = ℓ (x₁ − x₂),a₄ − a₃ ℓ − (2 ℓ + a₁) (y₁ − ℓ x₁) = x₁ x₂ + x₁ x₃ + x₂ x₃, andx₁ x₂ x₃ = (y₁ − ℓ x₁)² + a₃ (y₁ − ℓ x₁) − a₆.
All but the second identity compare coefficients in the Weierstrass cubic restricted to the line
y = ℓ (x − x₁) + y₁, whose roots are x₁, x₂ and x₃. The coefficients are those of the
integral model.
Twice a point whose tangent slope lies in 𝔪 has singular reduction. Here the reduction is
singular at the origin, with a₃ ∈ 𝔪 and a₂ ∈ 𝔪 (a₄ and a₆ then lie in 𝔪 too, by the
slope and curve equations). If (x, y) has integral coordinates in 𝔪 and the
tangent there has slope ℓ ∈ 𝔪, that is ℓ (2 y + a₁ x + a₃) = 3 x² + 2 a₂ x + a₄ − a₁ y with
2 y + a₁ x + a₃ ≠ 0, then both coordinates of twice the point lie in 𝔪.
The chord from (ϖ X₁, ϖ² Y₁), with X₁ a non-multiple of ϖ, to (ϖ² X₂, ϖ² Y₂) is not
vertical and its slope is a multiple of ϖ in the valuation ring: it is
ϖ (Y₁ − Y₂) / (X₁ − ϖ X₂), and X₁ − ϖ X₂ is a unit.