Chapter 3: The Algebraic Description of Conic Sections
Treating ellipses, hyperbolas, and parabolas uniformly as second-degree curves
From cones to algebra
Apollonius of ancient Greece (3rd century BC) treated the circle, ellipse, hyperbola, and parabola uniformly as sections of a cone cut by a plane. Changing the tilt of the cut yields different curves, as follows:
- Parallel to the base → circle
- Tilted, cutting both sides → ellipse
- Parallel to a generator line → parabola
- Parallel to the axis (cutting both nappes of the cone) → hyperbola (with two branches)
After Descartes introduced the coordinate plane in the 17th century, it became clear that all of these can be written as the zero set of a two-variable second-degree polynomial:
$$ax^2 + bxy + cy^2 + dx + ey + g = 0$$
That is, the correspondence "a geometric section of a cone" = "an algebraic second-degree curve" is the first great achievement of algebraic geometry.
Canonical form
By a suitable coordinate transformation (translation + rotation), a second-degree curve can be rewritten in a concise form called its canonical form. The result is one of the following three kinds (apart from degenerate cases):
Ellipse (centered at the origin, axes along the coordinate axes)
$$\dfrac{x^2}{a^2} + \dfrac{y^2}{b^2} = 1 \quad (a, b > 0)$$
If $a = b$ it is a circle. If $a > b$ it is an ellipse with major axis along the $x$-axis and minor axis along the $y$-axis.
Parabola
$$y^2 = 4px \quad (p > 0)$$
Or the form $y = ax^2$ (this just swaps the roles of $x$ and $y$). A symmetric figure defined by one focus and one directrix.
Hyperbola
$$\dfrac{x^2}{a^2} - \dfrac{y^2}{b^2} = 1 \quad (a, b > 0)$$
It has two branches. It approaches the asymptotes $y = \pm \dfrac{b}{a} x$.
Classification by the discriminant
Which of ellipse, hyperbola, or parabola the general form $ax^2 + bxy + cy^2 + dx + ey + g = 0$ corresponds to is determined by the sign of the discriminant
$$\Delta = b^2 - 4ac$$
as follows:
| Discriminant | Curve | Geometric feature |
|---|---|---|
| $\Delta < 0$ | Elliptic |
A bounded, closed figure (ellipse, circle) |
| $\Delta = 0$ | Parabolic |
A single curve extending to infinity |
| $\Delta > 0$ | Hyperbolic |
Has two branches |
This naming is the same as the classification of second-order linear PDEs (elliptic, parabolic, hyperbolic), using the same discriminant. This is no coincidence: it comes from the fact that the characteristic equation of a PDE is a second-degree curve.
Worked checks
(1) $x^2 + y^2 = 1$ (circle): $a = c = 1, b = 0$, $\Delta = 0 - 4 = -4 < 0$ → elliptic ✓
(2) $y - x^2 = 0$ (parabola $y = x^2$): the second-degree term is only $-x^2$, so $a = -1, b = c = 0$, $\Delta = 0 - 0 = 0$ → parabolic ✓
(3) $x^2 - y^2 = 1$ (hyperbola): $a = 1, c = -1, b = 0$, $\Delta = 0 + 4 = 4 > 0$ → hyperbolic ✓
(4) $xy = 1$ (rectangular hyperbola): $a = c = 0, b = 1$, $\Delta = 1 > 0$ → hyperbolic ✓
Degenerate conics
The "zero set of a second-degree equation" is not necessarily an ellipse, hyperbola, or parabola in the ordinary sense. Degenerate forms:
- Union of two lines: $x^2 - y^2 = 0$ ⇔ $(x - y)(x + y) = 0$ → the union of the two lines $y = x, y = -x$. The discriminant is $\Delta = 0 - 0 = 0$ but it is not a parabola
- A single (double) line: $x^2 = 0$ ⇔ the $y$-axis drawn "doubly"
- A single point: $x^2 + y^2 = 0$ → the only real solution is $(0, 0)$. The discriminant is $\Delta = -4 < 0$ ("elliptic"), but it is a single point as a limit of an ellipse
- The empty set: $x^2 + y^2 + 1 = 0$ → no real solutions
In the context of classical algebraic geometry, these degenerate forms arise naturally as "limiting cases of ellipses, hyperbolas, and parabolas". Moreover, passing to projective geometry changes how they look; for example, the difference between "two parallel lines" and "two intersecting lines" disappears. Details are covered at the Intermediate level.
Five points determine a conic
A second-degree curve $ax^2 + bxy + cy^2 + dx + ey + g = 0$ has six coefficients $a, b, c, d, e, g$, but scaling everything by a constant gives the same curve, so the number of degrees of freedom is 5. Therefore, specifying five points in general position determines (in general) a unique conic through them.
This reduces to a problem in linear algebra: the condition of passing through five points $(x_i, y_i)$ ($i = 1, \ldots, 5$) is the system of five linear equations in the coefficients
$$a x_i^2 + b x_i y_i + c y_i^2 + d x_i + e y_i + g = 0 \quad (i = 1, \ldots, 5)$$
Since this is five equations in six unknowns, the solution space is in general one-dimensional (a single solution up to scaling).
This fact that "five points determine a conic" is applied, for example, in Pascal's theorem in projective geometry and in Steiner's construction problems.
Next step
So far we can describe first-degree curves (lines) and second-degree curves (conics) in the language of polynomials. That is the end of the Intro level. At the Intermediate level we proceed to:
- The multivariable case (affine algebraic varieties in three or more dimensions)
- Introduction of projective space and projective varieties
- The correspondence between sets of polynomials and ideals (Hilbert's basis theorem)
- The Zariski topology and irreducible decomposition
The Basic-level article Classical Algebraic Geometry can also be read as a continuation of this article.
Frequently Asked Questions
Is the discriminant $b^2 - 4ac$ the same as the one in the quadratic formula?
The form is the same but the meaning differs. The discriminant of a quadratic equation $at^2 + bt + c = 0$ determines the "number of real solutions". The discriminant of a conic $ax^2 + bxy + cy^2 + \cdots = 0$ is determined by the second-degree terms (the principal part of the equation) alone, and decides the "type" of the curve. They coincide because the same algebraic structure appears in the standardization of a quadratic form by a linear transformation.
$xy = 1$ (a rectangular hyperbola) is classified as hyperbolic by $b^2 - 4ac = 1$, yet it does not appear among the canonical forms. Why?
Rotating $xy = 1$ by 45 degrees gives the canonical form $\dfrac{x^2 - y^2}{2} = 1$ (the canonical form of a hyperbola). That is, by rotating the coordinate system one can make $b = 0$. The discriminant is invariant under rotation, so the "type" does not change before or after the rotation.
Are degenerate conics also objects of algebraic geometry?
Yes. Indeed, one of the central themes of classical algebraic geometry is the study of "singularities of curves (including degeneracies)". Degenerate conics (a union of two lines, etc.) arise as limits of smooth conics and sit on the boundary of "the moduli space of a family of curves". This is explored in depth at the Intermediate and Advanced levels.