I've been working through Silverman and Tate's book Rational Points on Elliptic Curves. They use conic equations as an introduction to singular/nonsingular curves. I've reproduced the problem with my comments below, then my questions follow.

Let $C$ be the conic given by the equation $$F(x,y) = ax^2 + bxy + cy^2 + dx + ey + f = 0$$ and let $\delta$ be the determinant $$\det \begin{bmatrix} 2a & b & d \\ b & 2c & e \\ d & e & 2f \\ \end{bmatrix}$$

(a) Show that if $\delta \neq 0$ then $C$ has no singular points.

Answered here.

(b) Conversely, show that if $\delta = 0$ and $b^2 - 4ac \neq 0$ then there is a unique singular point on $C$.

Given these two conditions we can classify the curve as two intersecting lines or a single point. So there is one singular point on the curve. This is from wikipedia.

(c) Let $L$ be the line $y = \alpha x + \beta$ with $\alpha \neq 0$. Show that the intersection of $L$ and $C$ consists of either zero, one, or two points.

Substitute the equation of the line into the equation of the conic. Then we get a quadratic equation $(a + \alpha b + \alpha^2 c)x^2 + (\beta b + 2\alpha\beta c + d + \alpha e)x + (c\beta^2 + \beta e + f) = 0$. This equation has zero, one, or two solutions.

(d) Determine the conditions on the coefficients which ensure that the intersection $L \cap C$ consists of exactly one point. What is the geometric significance of these conditions?

Let $(a + \alpha b + \alpha^2 c)x^2 + (\beta b + 2\alpha\beta c + d + \alpha e)x + (c\beta^2 + \beta e + f) = a'x^2 + b'x + c'$ There is exactly one solution if the determinant $b'^2 - 4a'c'$ of the quadratic equation is 0. Geometrically, the intersections collide: the line must be tangent.

My questions:

(b) Can anyone provide a reference other than Wikipedia for this? I've searched a lot but can't find anything that actually gives a derivation.

(c,d) What is the condition for zero, one, two roots? The field is not specified so I assume these roots could be complex. Then

  • we have zero roots if $a' = b' = 0$
  • we have one root if $a' = 0, b' \neq 0$
  • we have two roots if $a' \neq, b' \neq 0$

But the two roots could be one root of multiplicity two. Is that correct? Or should I only be thinking about the third case (over real numbers), then considering the possible roots, with complex roots left out.

(d) How do I flesh out a proof of this fact? (i.e. How would I prove that $b'^2 - 4a'c'$ implies the line is tangent?)

  • $\begingroup$ The line is tangent to a curve if infinitesimal order of distance between them near an intersection point is more than infinitesimal order of distance to that point $\endgroup$ – Alexey Burdin Apr 26 '15 at 0:40
  • $\begingroup$ (a) is not answered as you say; for, Silverman does not expect an answer using "projective conic": it is an exercise 1.2 in the book; no tools have been mentioned, let alone developed. How did you answer (a)---I should like to know---in a more elementary way? $\endgroup$ – Wulfgang Mar 5 '17 at 5:09
  • $\begingroup$ You're right that the other answer I linked to is quite a bit more complex than is necessary. I ended up answering: $\delta \neq 0$ implies that the matrix is invertible, i.e. that the only solution to $Ax = 0$ is $x = 0$. But there is no point $(0,0,0)$ in projective space. So $C$ has no singular points. $\endgroup$ – positivepeter Mar 6 '17 at 22:04


Let $Q(x,y) = ax^2 + 2hxy + by^2 + 2gx + 2fy + c$

$Q_x(x,y) = 2ax + 2hy + 2g$ and $Q_y(x,y) = 2hx + 2by+2f$

Solving $Q_x = 0$ and $Q_y = 0$, $$\begin{bmatrix}a &h\\ h & b \end{bmatrix}\begin{bmatrix}x\\ y \end{bmatrix} = \begin{bmatrix}-g\\ -f \end{bmatrix}$$

Given $ab - h^2 \ne 0$, there is a unique solution for $Q_x = 0, Q_y = 0$ and $Q$ has a unique centre. Let that centre be $(u ,v)$. Shifting the centre of $Q$ to origin with $x = X + u$ and $y = Y + v$ gives $$R(X, Y) = aX^2 + 2hXY + bY^2 + \dfrac{\Delta}{\gamma}$$

where $\Delta$ is determinant of matrix of $Q$ and $\gamma = ab - h^2$. Given $\Delta = 0$, origin lies on $R(X,Y)$. $$0 = R(0,0) = Q(u, v)$$

Therefore, there exist only one point $(u, v)$ for $Q_x = 0, Q_y = 0$ and $Q = 0$. Hence $(u, v)$ is unique singular point of $Q$.


Let $L(t) = (u + tX, b + tY)$, putting this in $Q = 0$ give $pt^2 + qt + r = 0$ where

$$\begin{cases}p = aX^2 + 2hXY + bY^2 \\ q = X Q_x(u, v) + Y Q_y (u, v) \\ r = Q(u,v) \end{cases}$$

Hence $L$ intersects $Q$ at one point, two points or never provided $p, q, r$ all are not zero.

d): I don't understand this part. $L$ is tangent to $Q$ when $q^2 - 4pr = 0$ but finding the condition on coefficients of $Q$ and $L$ is a lot of dull calculations (at least with my method).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.