MATH 431-2018 PROBLEM SET 6

CONICS AND QUADRICS

OCTOBER 30, 2018

1. Conics
While lines in P2 are defined by linear (first-order) equations, conics
are defined by second-order equations. Such equations are defined by
homogeneous quadratic polynomials, that is, polynomial equations of
the form
(1) aX 2 + bXY + cY 2 + dXZ + eY Z + f Z 2 = 0
in homogeneous coordinates [X : Y : Z].
Exercise 1.1. Why must this equation be homogeneous?
While conics in affine geometry fall into three types: ellipses, parabo-
las and hyperbolas, these are all equivalent projectively. That is, when
we supplement A2 by a (projective) line of ideal points and don’t think
of ideal points as anything special, they are all equivalent.
For example, the circle defined by x2 + y 2 = 1 is defined in the
standard affine patch [x : y : 1] by the homogeneous equation
X 2 + Y 2 − Z 2 = 0,
the parabola y = x2 by
X2 − Y Z = 0
and the hyperbola xy = 1 by
XY − Z 2 = 0.
Exercise 1.2. Express the hyperbola x2 − y 2 = 1 in homogeneous co-
ordinates and find a projective coordinate change to make it look like
the preceding hyperbola.
We can write the equation (1) in vector form, using the symmetric
matrix
 
a b/2 d/2
(2) M :=  b/2 c e/2
d/2 e/2 f
1
2 OCTOBER 30, 2018

Consider a vector variable

 
X
v :=  Y  ;
Z

then the homogeneous quadratic equation (1) becomes:

(3) v† Mv = 0

We shall assume that matrix M is invertible, that is, it has nonzero

determinant. Even then, it might not define an interesting curve in P2 :

Exercise 1.3. Suppose that M is the 3×3 identity matrix. What subset
of P2 does it define? Find other invertible symmetric 3 × 3 matrices
defining the same subset.

At the other extreme, if M is the zero matrix, then every vector v

solves (3), so that doesn’t define a curve either.
Although we won’t prove it here, a 3 × 3 symmetric matrix is always
diagonalizable, and has 3 (real) eigenvalues. It’s invertible if and only
if 0 is not an eigenvalue. Thus each of the three eigenvalues is either
positive or negative. The condition that M define a subset of P2 is
that it has at least one positive eigenvalue and at least one negative
eigenvalue. In other words it either has one positive and two negative
eigenvalues, or it has two positive and one negative eigenvalue.
These two different cases aren’t really all that different:

Exercise 1.4. Show that if M and M0 are two matrices which are
projectively equivalent, that is, M0 = λM for some scalar λ 6= 0, then
they define the same subset of P2 . How are the eigenvalues of M and
M0 related?

A given conic can look quite different when viewed in different affine
patches, as you saw in the earlier homework assignment:
 CONICS AND QUADRICS 3

Consider the four affine patches:

A
A2 −−−
→ P2
1

(η, z) 7−−−→ [1 : η : z]

A
A2 −−−
→ P2
2

(ξ, ζ) 7−−−→ [ξ : 1 : ζ]

A
A2 −−−
→ P2
3

(x, y) 7−−−→ [x : y : 1]

A
A2 −−−
→ P2
4

(u, v) 7−−−→ [u + 1 : u − v : u + v]

Recall the parabola

P := {(x, y) ∈ A2 | y = x2 }

and its closure C of A3 (P ) ⊂ P2 . Then we saw that

C = {[X : Y : Z] | X 2 = Y Z}.

Recall the calculation. The two sets of coordinates x, y (affine) and

X, Y, Z (homogeneous) are related by conversion formulas:

x = X/Z
y = Y /Z

so that P is defined by

Y X2
(4) = 2
Z Z
so multiplying by Z 2 gives Y Z = X 2 . This has the effect of adding the
(single) ideal point where Z = 0, since implicit in (4) is the assumption
that Z 6= 0. (This ideal point is [0 : 1 : 0], or any [0 : Y : 0], as long as
Y 6= 0. They are all equal.)
4 OCTOBER 30, 2018

Here are the defining covectors for the four affine patches:
 
φ1 = 1 0 0
 
φ2 = 0 1 0
 
φ3 = 0 0 1
 
φ4 = 1 −1/2 −1/2
That is, the patch Ai corresponds to the affine plane φ−1 3
i (1) ⊂ R . The
scale factor is the nonzero scalar λ such that if v ∈ R3 is a nonzero
vector then λv lies on the viewing plane.
(1) Compute scale factors and viewing planes for these four affine
patches.
(2) Find the sections of the cone X 2 + Y 2 − Z 2 = 0 in these affine
patches.
(3) Relate the four sets of coordinates (η, z), (ξ, ζ), (x, y), (u, v) and
identify the sets where they can be related.
(4) Now find an affine chart with affine coordinates (p, q) so that C
is the circle p2 + q 2 = 1.
(5) Look at the cone C defined by X 2 + Y 2 − Z 2 = 0 in R3 . Rotate
the affine plane Z = 1 around the line X = Z − 1 = 0 by
angle θ, obtaining a plane Pθ . Then the conic section Pθ ∩ C
can be either an ellipse, a parabola, a hyperbola, or a pair of
intersecting lines.
(a) When θ 6= ±π/2, this plane has the equation Z = 1 +
tan(θ)X.
(b) When θ = ±π/2, this plane has the equation X = 0.
(c) Let t = tan(θ). Then Pθ corresponds to the affine chart
A
(x, y) 7−−→ [x : y : 1 + tan(θ)x]
(d) Suppose that |θ| < π/4, so that |t| < 1. In this affine chart
the conic section is the ellipse
 2
t 1
x− 2
+ y 2 = (1 − t2 )−4
1−t 1 − t2
which has eccentricity
√
p
cos(2θ)
1 − t2 = .
cos(θ)
In particular it is a circle exactly when θ = 0.
(e) When |θ| = π/4 (so that |t| = 1), this conic section is a
parabola.
 CONICS AND QUADRICS 5

(f) When π/4 < |θ| < π/2 (so that 1 < |t| < ∞), this conic
section is a hyperbola.
(g) When |θ| = π/2 the cone intersects the plane in a pair of
intersecting lines (a degenerate hyperbola).
The three types of conics are easily characterized by their relation
to the line-at-infinity. Let C ⊂ P2 be a conic, and let ` ⊂ P2 be a line.
Then A2 := P2 \ ` is an affine patch.
• C ∩ A2 is an ellipse ⇐⇒ C ∩ ` = ∅;
• C ∩ A2 is a parabola ⇐⇒ C ∩ ` is a single point;
• C ∩ A2 is a hyperbola ⇐⇒ C ∩ ` consists of two points.

2. Quadric surfaces
2.1. Three Types of Unruled Quadrics. Define the ellipsoid, el-
liptic paraboloid, and two-sheeted hyperboloid:
n   x 2  y 2  z 2 o
Ea,b,c := (x, y, z) ∈ R3  + + =1

a b c
n   x 2  y 2 o
Pa,b := (x, y, z) ∈ R3  + =z

a b
n   x 2  y 2  z 2 o
Ha,b,c := (x, y, z) ∈ R3  + − = −1

a b c
where a, b, c 6= 0 (they are usually taken to be positive). Visualize these
surfaces.
In projective space P3 define the quadric

Q := [X : Y : Z : W ] ∈ P3  X 2 + Y 2 + Z 2 = W 2 .


(1) In the usual affine patch (x, y, z) 7−→ [x : y : z : 1], find the
ideal points of Ea,b,c , Pa,b , Ha,b,c .
(2) Find three affine patches A such that Ea,b,c , Pa,b and Ha,b,c are
each A−1 (Q). (Hint: use the formulas a2 − b2 = (a − b)(a + b)
and 4ab = (a + b)2 − (a − b)2 .)
(3) (Bonus problem) Prove or disprove: The affine patch Ea,b,c −→
Q is a homeomorphism. (Recall that a homeomorphism is a con-
tinuous bijection whose inverse is continuous. In other words,
it is a mapping which preserves the topology, the way the points
are “organized” into a space. It can stretch, squeeze and oth-
erwise distort the geometry, but it can’t tear, collapse or break
the space. Being continuous means preserving the underlying
“topological fabric.”)
(4) (Bonus problem) Find a homeomorphism of Ea,b,c −→ S 2 where
S 2 is the 2-dimensional sphere.
6 OCTOBER 30, 2018

2.2. Surfaces of revolution and cylindrical coordinates. A sur-

face of revolution is a surface obtained by rotating a plane curve about
a straight line in that plane. A simple example is revolving a line about
a parallel line to obtain a cylinder. To fix notation, let (x0 , y0 ) ∈ R2
be a point in the xy-plane. To obtain a cylinder, rotate the line `(1,0)
about the z-axis `(0,0) : rotation through  angle angle θ takes `1,0 to the
line `(cos(θ),sin(θ)) and the cylinder equals x2 + y 2 = 1 .
(5) Express the ellipsoid E1,1,c and the paraboloid P1,1 as surfaces
of revolution. 
(6) How is the cone x2 + y 2 = z 2 a surface of revolution?
(7) Define cylindrical coordinates (r, θ, z) on E3 by
p
r := x2 + y 2
θ := tan−1 (y/x)
z := z
and
x := r cos(θ)
y := r sin(θ)
z := z.
Express the above surfaces in cylindrical coordinates.
(8) (Bonus problem) Prove or disprove: Every surface defined im-
plicitly by an equation f (r, z) = 0 is a surface of revolution
about the z-axis.
2.3. Ruled Quadrics. Sometimes quadrics contain straight lines. Then
the quadric is said to be ruled. In that case the quadric corresponds to
the surface in projective space:
Q0 := {[X : Y : Z : W ] ∈ P3 | X 2 + Y 2 = Z 2 + W 2 }
A simple example is the hyperbolic paraboloid or saddle:
S 0 := {(x, y, z) | xy = z}
The intersection of S 0 with the xy-plane z = 0 is defined by xy = z = 0,
which decomposes as the union of two lines: the y-axis x = z = 0 and
the x-axis y = z = 0.
(9) For any point p0 = (x0 , y0 , x0 y0 ), find two lines through the
point and lying on S. (Hint compute the tangent plane to S at
p0 . In the preceding example, what is the relation between S,
the origin (0, 0, 0) and the xy-plane?)
(10) Find an affine patch A such that S = A−1 (Q0 ).
 CONICS AND QUADRICS 7

(11) Find a set of affine coordinates (u, v, w) so that S is given by

the equation
w = u2 − v 2 .
Another example is the one-sheeted hyperboloid:
H 0 := {(x, y, z) ∈ R3 | x2 + y 2 − z 2 = 1}
which is a surface of revolution in several different ways. ls In the usual
affine patch
A(x, y, z) = [x : y : z : 1]
(with viewing hyperplane W = 1), H 0 = A−1 (Q0 ). Notice that it is
invariant under the one-parameter group of rotations about the z-axis:
   
cos(θ) − sin(θ) 0 0 −θ 0
 sin(θ) cos(θ) 0 = exp θ 0 0
0 0 1 0 0 0
H 0 is obtained by revolving the hyperbola
y2 − z2 − 1 = x = 0
around the z-axis.
(12) Write H 0 in cylindrical coordinates.
(13) Find the ideal points of Q0 in the affine patch A.
(14) For each θ ∈ R representing an angle (that is, only defined
modulo 2π),
    
cos(θ) cos(θ) − sin(θ) 0 0
±
`θ := sin(θ) + R sin(θ) cos(θ) 0
     ±1
0 0 0 1 1
determines two lines (from different choices ±) which lie on H 0 .
(15) H 0 is also the surface of revolution obtained by revolving any
one of these lines about the z-axis.
These families of lines are called rulings and this quadric is ruled in
two different ways.
Similarly, this quadric is a surface of revolution in two different ways.
Try making a model of H 0 out of string and two flat circular (or ellip-
tical) rings.

2.4. Topology of a ruled quadric. The projective surface Q0 is ac-

tually a torus, a surface homeomorphic to a a bagel, doughnut, or inner
tube. This can be seen as follows. Write the equation defining Q0 in
the form
X2 + Y 2 = Z2 + W 2
8 OCTOBER 30, 2018

and note that this quantity is positive. (Being a sum of squares it is

always nonnegative, and if it is zero, then X 2 + Y 2 = Z 2 + W 2 = 0,
which implies X = Y = Z = W = 0, a contradiction.) By scaling
the homogeneous coordinates, we can assume that X 2 + Y 2 = 1 and
Z 2 + W 2 = 1, and equation defines a pair of circles (one in the X, Y -
plane and the other in the Z, W -plane).
Here is an explicit formula. Let θ, φ ∈ R represent angles (so they
are only defined modulo 2π). Write
Xθ := cos(θ)X − sin(θ)Y Yθ := sin(θ)X + cos(θ)Y
Zφ := cos(φ)Z − sin(φ)W Wφ := sin(φ)Z + cos(φ)W
and note that
Xθ2 + Yθ2 = X 2 + Y 2 Zφ2 + Wφ2 = Z 2 + W 2
This will enable us to understand the topology.
Let T ⊂ R4 denote the subset defined by
X 2 + Y 2 = Z 2 + W 2 = 1.
Prove or disprove the following statements.
(16) (Bonus problem) The map
S 1 × S 1 −→ T ⊂ R4
 
cos(θ)
 sin(θ) 
(θ, φ) 7−→ 
cos(φ)

sin(φ)
is a homeomorphism (a topological equivalence).
(17) (Bonus problem) The map
S 1 × S 1 −→ Q0 ⊂ P3
 
(θ, φ) 7−→ cos(θ) : sin(θ) : cos(φ) : sin(φ)
is a homeomorphism (Hint: Look at what happens to (π, π).)
(18) (Bonus problem) Q0 is homeomorphic to S 1 × S 1 .