M231: Exercise sheet 1

1. Let V be a vector space over a field F.

(a) Let v, w ∈ V \ {0}. Show that [v] = [w] if and only if w = λv, for some λ ∈ F× .
(b) Show that the relation ∼ defined on V \ {0} by v1 ∼ v2 if and only if v2 = λv1 ,for
some λ ∈ F× , is indeed an equivalence relation.
(c) Let V \ {0}/ ∼ be the set of equivalence classes and define a map Φ : P(V ) →
V \ {0}/ ∼ be letting Φ([v]) be the equivalence class of v. Show that Φ is a
well-defined bijection.
2. Proposition 1.1 states: Let U1 , U2 ≤ V then

dim(U1 + U2 ) = dim U1 + dim U2 − dim(U1 ∩ U2 ).

Here is a slick way to prove the proposition: make the Cartesian product U1 × U2
into a vector space in the usual way: thus (u1 , u2 ) + (v1 , v2 ) = (u1 + v1 , u2 + v2 ),
λ(u1 , u2 ) = (λu1 , λu2 ). Define a map α : U1 × U2 → U1 + U2 by α(u1 , u2 ) = u1 + u2 .
(i) Show that α is a linear transformation.
(ii) Find an isomorphism U1 ∩ U2 ∼ = ker α.
(iii) Apply The Rank-Nullity Theorem to prove Proposition 1.1.
3. Let P(V ) be a projective space of dimension at least 2 (thus dim V ≥ 3). Prove:
(a) Distinct projective lines in P(V ) intersect in at most one point. (Very easy!)
(b) Distinct projective lines in P(V ) intersect if and only if they lie in some (unique)
projective plane.
(c) Give an example of two non-intersecting projective lines in P(R4 ).
4. Let P(V ) be a projective space and X1 , X2 ≤ P(V ) projective subspaces of comple-
mentary dimension: dim X1 + dim X2 = dim P(V ).
Prove that the Xi intersect.
What does this tell us when dim P(V ) = 3?
Maybe more challenging
5. Let F = Z/2Z be the field with two elements.
(a) How many points are there in P(F2 )?
(b) How many points are there in the plane P(F3 ) and how many lines?
(c) How many lines go through a point of P(F3 )?
(d) Repeat the exercise with F = Z/3Z, the field with three elements.
Are any patterns emerging?

February 26, 2013

Home page: http://go.bath.ac.uk/ma30231

M231: Exercise sheet 1—Solutions

1. (a) If [v] = [w] then w ∈ [v] which means there is λ ∈ F× with w = λv. Conversely, if w = λv,
then for any µ ∈ F× , µw = (µλ)v ∈ [v] so that [w] ⊂ [v]. Applying the same argument
starting from v = λ−1 w gives [v] ⊂ [w] so that [v] = [w].
(b) Reflexive: v ∼ v since v = 1v.
Symmetric: v1 ∼ v2 means v2 = λv1 , some λ ∈ F, if and only if v1 = λ−1 v2 so that
v2 ∼ v1 .
Transitive: if v1 ∼ v2 and v2 ∼ v3 , we have v2 = λv1 and v3 = µv2 whence v3 = (µλ)v1
and v1 ∼ v3 .
(c) First we show that Φ is well-defined: if [v] = [w], by part (a), we have that v ∼ w so
that v and w define the same equivalence class. Moreover, if Φ([v]) = Φ([w]) then v ∼ w
so, again by part (a), [v] = [w] and Φ is injective. Finally, the equivalence class of any
v ∈ V \ {0} is Φ([v]) so Φ surjects also.
2. (i) We must show first that α((u1 , u2 ) + (v1 , v2 )) = α(u1 , u2 ) + α(v1 , v2 ). But (u1 , u2 ) +
(v1 , v2 ) = (u1 + v1 , u2 + v2 ) so

α((u1 , u2 ) + (v1 , v2 )) = α(u1 + v1 , u2 + v2 )

= (u1 + u2 ) + (v1 + v2 ) = α(u1 , u2 ) + α(v1 , v2 ).

Now we must show that α(λ(u1 , u2 )) = λα(u1 , u2 ) which is similar (and easier).
(ii) Note that (u1 , u2 ) is in the kernel of α if and only if u1 +u2 = 0 or, equivalently, u1 = −u2 .
In this case, we must have u1 , u2 ∈ U1 ∩U2 . We can therefore define a map U1 ∩U2 → ker α
by u 7→ (u, −u) and one readily checks that this is a linear isomorphism.
(iii) We are now almost done: clearly Im α = U1 + U2 while dim(U1 × U2 ) = dim U1 + dim U2
and, by (ii), dim α = dim(U1 ∩ U2 ). The rank-nullity theorem now bakes the cake!
3. (a) Theorem 1.5 says there is a unique line through two distinct points. Thus distinct lines
can intersect in at most one point.
(b) Let Li = P(Ui ), i = 1, 2, be distinct lines in P(V ) and consider the join L1 L2 . The
dimension formula gives:

dim(L1 L2 ) = dim L1 + dim L2 − dim(L1 ∩ L2 ) = 2 − dim(L1 ∩ L2 )

and dim(L1 ∪ L2 ) < 1 since L1 6= L2 . Thus L1 and L2 intersect if and only if dim(L1 ∩
L2 ) = 0 if and only if L1 L2 is a projective plane. In particular, if the Li intersect, they
lie in the plane L1 L2 . The converse is the second part of Theorem 1.5 which follows
at once from our analysis: if P(V ) is a projective plane, then dim(L1 L2 ) ≤ 2 so that
1 > dim(L1 ∩ L2 ) ≥ 0.
Finally, for uniqueness, any projective plane P that contains L1 and L2 must also contain
L1 L2 whence P = L1 L2 since both have dimension 2.
(c) We just need linear subspaces Ui ≤ R4 with U1 ∩ U2 = {0}. For example, take U1 =
{(λ0 , λ1 , 0, 0) : λ0 , λ1 ∈ R} and U2 = {(0, 0, λ2 , λ3 ) : λ2 , λ3 ∈ R}.
4. We argue as in Theorem 1.5 and wheel out the dimension formula: let dim P(V ) = n and
dim X1 = k so that dim X2 = n − k. Then

n = dim P(V ) ≥ dim(X1 X2 ) = dim X1 + dim X2 − dim(X1 ∩ X2 )

= k + (n − k) − dim(X1 ∩ X2 )

and rearranging this gives dim(X1 ∩ X2 ) ≥ 0 so that the Xi intersect.

When n = 3, the only interesting possibility is k = 1, 2 and the result asserts that any
projective line must intersect any projective plane. It is easy to see that this intersection
point is unique dim(X1 ∩ X2 ) = 0 or the line lies in the plane dim(X1 ∩ X2 ) = 1.
5. (a) The same argument as that given in lectures establishes that, for any field F, P(F2 ) ∼ =
F t {∞} so that |P(F2 )| = |F| + 1. When F = Z/2Z, this yields |P(F2 )| = 2 + 1 = 3.
Alternatively, view P(F2 ) as F2 \ {0}/ ∼ and observe that all equivalence classes have
the same size which is |F× | because, for any v ∈ F2 \ {0}, λ 7→ λv is a bijection from
F× to the equivalence class of v. Since equivalence classes partition, we have |P(F2 )| =
|F2 \ {0}|/|F× | = (22 − 1)/(2 − 1) = 3.
(b) Either argue that P(F3 ) ∼ = F2 t F t {∞} so that |P(F3 )| = 22 + 2 + 1 = 7 or that
×
|P(F )| = |F \ {0}|/|F | = (23 − 1)/(2 − 1) = 7.
3 3

As for the lines in P(F3 ), a line is determined by any pair of distinct points lying on it.
There are 7 × 6/2 such pairs and any particular line is equally well determined by any of
the 3 pairs of the 3 points on it. Thus there are 7 = 7 × 6/(2 × 3) lines in the plane P(F2 ).
(c) Fix a point, then there are 6 pairs that include that point and each determines a line
which is counted twice over: once for each of the other two points on the line. Thus there
are 6/2 = 3 lines through each point.
(d) The same arguments apply for the field with 3 elements: |P(F2 )| = 3 + 1 = 4 = (32 −
1)/(3 − 1). Similarly, |P(F3 )| = 13. Meanwhile, each line has 4 points and is determined
by any of the 6 pairs of point that lie thereon. In all, there are 13 × 12/2 = 13 × 6 pairs
of points in P(F3 ) and thus 13 lines. Finally, there are 12 pairs that include a given point
and each determines a line through that point which is counted 3 times over, giving 4
lines through a given point.
This suggests two facts:
1. Projective planes contain the same number of points as lines.
2. The number of points on a line in a projective plane is the same as the number of lines
through a point of a projective plane.
All of this is true, as we shall see later. Meanwhile, I invite you to prove these facts now for
an arbitrary finite field.
M231: Exercise sheet 2

Lines and planes

1. Let A, B, C be three points in a projective plane with homogeneous coordinates
[a0 , a1 , a2 ], [b0 , b1 , b2 ] and [c0 , c1 , c2 ], respectively. Show that A, B, C are collinear1 if
and only if the determinant
a0 a1 a2
b0 b1 b2 = 0.
c0 c1 c2
Deduce that the equation of the line BC is

λ0 λ1 λ2
b0 b1 b2 = 0.
c0 c1 c2

2. Find the vertices of the triangle in P(R3 ) whose sides are given by the points [λ0 , λ1 , λ2 ]
satisfying the following homogeneous equations:
(a) λ0 = 0.
(b) λ0 + λ1 + λ2 = 0.
(c) 3λ0 − 4λ1 + 5λ2 = 0.
3. Consider the following three lines in R3 and the corresponding lines in P(R4 ):
(a) {(x1 , x2 , x3 ) : x1 + x2 = 1 and x2 + x3 = 1}.
(b) {(x1 , x2 , x3 ) : x1 = 0 and x3 = 0}.
(c) {(x1 , x2 , x3 ) : x1 − x3 = 1 and x2 + x3 = 0}.
Which of these lines meet in R3 and which meet in P(R4 )? Find the coordinates in
R3 and/or the homogeneous coordinates in P(R4 ) of the intersection points.
4. Show that the following lines in P(R4 ) are coplanar (that is, they lie in a common
projective plane) and find the vertices of the triangle they form:
(a) {[λ0 , λ1 , λ2 , λ3 ] : λ0 + λ1 = 0 and λ2 + λ3 = 0}.
(b) {[λ0 , λ1 , λ2 , λ3 ] : λ0 + λ2 = 0 and λ1 + λ3 = 0}.
(c) {[λ0 , λ1 , λ2 , λ3 ] : λ0 + λ3 = 0 and λ1 + λ2 = 0}.
(Recall that lines in P(Rn+1 ) are coplanar if and only if they intersect.)
Projections and direct sums
5. Let V1 , V2 ≤ V be subspaces of the vector space V . Recall that V is the direct sum of
V1 and V2 if
(i) V = V1 + V2 ;
(ii) V1 ∩ V2 = {0}.
In this case, we write V = V1 ⊕ V2 .
1
That is, they lie on the same line.

Home page: http://go.bath.ac.uk/ma30231

 (a) Show that V = V1 ⊕ V2 if and only if any v ∈ V can be uniquely written as
v = v1 + v2 with vi ∈ Vi .
(b) Use this to define a map P1 : V → V1 by P1 v = v1 . Show that P1 is a linear map.
(c) Prove that P12 = P1 ; ker P1 = V2 and Im P1 = V1 .
(d) Conversely, if P : V → V is a linear map with P 2 = P , set V1 = Im P , V2 = ker P
and show that V = V1 ⊕ V2 and, with v = v1 + v2 as above, P v = v1 , for all v ∈ V .
6. Let X1 , X2 ≤ P(V ) be projective subspaces. Show that the join X1 X2 is the union of
all the lines in P(V ) containing a point of X1 and a point of X2 :
[
X1 X2 = A1 A2 .
Ai ∈Xi

Further questions
7. Let L1 and L2 be non-intersecting lines in P(R4 ) and A a point on neither line. Prove
that there is a unique line through A that intersects both L1 and L2 .
8. Recall that two affine lines {x + u(y − x) : u ∈ F} and {x0 + u(y 0 − x0 ) : u ∈ F} in Fn
are parallel if and only if there is λ ∈ F× with y 0 − x0 = λ(y − x).
(a) Show that two affine lines are parallel if and only if the projective lines containing
them meet on the hyperplane at infinity.
(b) Introduce a relation on affine lines by `1 ∼ `2 if and only if `1 and `2 are parallel.
Show that this is an equivalence relation.
(c) Show that there is a bijection between the equivalence classes of affine lines and
the hyperplane at infinity.

February 28, 2013

M231: Exercise sheet 2—Solutions

1. Let our plane be P(V ) and v0 , v1 , v2 be the basis of V against which

P homogeneous
P coordinates
P
are computed. Now A, B, C lie in a line L = P(U ) if and only if i ai vi , i bi vi and i ci vi
are linearly dependent and that is equivalent to the vanishing of the first determinant.
Now a point with homogeneous coordinates [λ0 , λ1 , λ2 ] lies on the line BC if and only if it is
collinear with B and C which yields the stated equation of the line.
2. Let us begin by finding the intersection of the first two lines. Thus we seek [λ0 , λ1 , λ2 ] solving
the equations λ0 = 0 and λ0 + λ1 + λ2 = 0. This is easy enough to do in our heads: we need
λ0 = 0 and λ1 = −λ2 so the solution is [0, λ, −λ] = [0, 1, −1].
For the intersection of first and third lines, we solve λ0 = 0 and 3λ0 − 4λ1 + 5λ2 = 0 so that
4λ1 = 5λ2 and our solution is [0, 5, 4].
Finally, for the intersection of the second and third lines, we need a solution of λ0 +λ1 +λ2 = 0
and 3λ0 − 4λ1 + 5λ2 = 0 and you must do whatever you do to solve linear equations (Gaussian
elimination? Ask a computer? Subtract 3 times the first equation from the second?). In any
case, you should arrive at the solution [−9, 2, 7].
3. Let us work with homogeneous coordinates from the start: thus, we remember that xi =
λi /λ0 , substitute in and multiply through by λ0 to get homogeneous linear equations for the
homogeneous coordinates of intersection points.
Now the intersection of the first and second lines must satisfy

λ0 − λ1 − λ2 =0
λ0 − λ2 − λ3 =0
λ1 =0
λ3 =0

which we can do in our heads to see that λ0 = λ2 and the intersection has homogeneous
coordinates [1, 0, 1, 0]. Since λ0 6= 0, this lies in R3 with inhomogeneous coordinates (0, 1, 0).
Similarly, the intersection of the second and third lines must satisfy:

λ1 =0
λ3 =0
λ0 − λ1 + λ3 =0
λ2 + λ3 =0

which quickly leads to the unique solution λi = 0, for 1 ≤ i ≤ 3. This means that the lines do
not intersect.
Finally, the intersection of the first and third lines solves:

λ0 − λ1 − λ2 =0
λ0 − λ2 − λ3 =0
λ0 − λ1 + λ3 =0
λ2 + λ3 =0

and we do whatever we must to see that λ0 = 0, λ1 = −λ2 = λ3 so that the lines intersect at
[0, 1, −1, 1] which is in P(R4 ) \ R3 since λ0 = 0.
4. The lines will be coplanar so long as they intersect pair-wise as distinct pairs of intersecting
lines lie in a unique plane. So let us seek the intersection points.
For the intersection of the first and second lines we need λ0 = −λ1 , λ2 = −λ3 , λ0 = −λ2 ,
λ1 = −λ3 yielding λ0 = −λ1 = −λ2 = λ3 so that the intersection lies at [1, −1, −1, 1].
Similarly, the intersection of the first and third lines is at [1, −1, 1, −1] and that of the second
and third at [1, 1, −1, −1].
5. (a) Suppose that V = V1 ⊕ V2 and let v ∈ V . Since V = V1 + V2 , there are vi ∈ Vi such
that v = v1 + v2 . For uniqueness, suppose we also have vi0 ∈ Vi with v = v10 + v20 . Then
 v1 + v2 = v10 + v20 yielding v1 − v10 = v20 − v2 . But the left side lies in V1 and the right in
V2 so that both lie in V1 ∩ V2 = {0}. Thus vi = vi0 , i = 1, 2 and we are done.
For the converse, suppose that each v ∈ V has a unique expression v = v1 + v2 as a sum
of elements of the Vi . That there is such an expression at all yields V = V1 + V2 . Now
let v ∈ V1 ∩ V2 . Since 0 ∈ V1 ∩ V2 , we can write v = 0 + v = v + 0 with the first summand
in V1 and the second in V2 . Uniqueness now forces 0 = v so that V1 ∩ V2 = {0}.
(b) Let v, w ∈ V and λ ∈ F. We must show that P1 (λv + w) = λP1 v + P1 w. Writing
v = v1 + v2 , w = w1 + w2 with vi , wi ∈ Vi , we have λv + w = λ(v1 + v2 ) + (w1 + w2 ) =
(λv1 + w1 ) + (λv2 + w2 ) with the latter summands in V1 and V2 respectively. Thus

P1 (λv + w) = λv1 + w1 = λP1 v + P1 w.

(c) For v ∈ V1 , the unique expression as a sum of elements of the Vi is v = v + 0 so that

P1 v = v. Thus, for arbitrary v = v1 + v2 , P12 v = P1 v1 = v1 = P1 v giving P12 = P1 .
v = v1 + v2 ∈ ker P1 if and only if v1 = 0 if and only if v = v2 ∈ V2 .
Finally, by construction Im P1 ≤ V1 while, for v1 ∈ V1 , we have P1 v1 = v1 and so equality.
(d) Let I : V → V be the identity map and note that 0 = P 2 − P = P (I − P ) so that
Im(I − P ) ≤ ker P . Now I = P + (I − P ) so that, for any v ∈ V , v = P v + (I − P )v with
the first summand in V1 and the second in V2 . Thus V = V1 + V2 .
Now let v ∈ ker P ∩ Im P . Thus v = P w, for some w ∈ V , but 0 = P v = P 2 w = P w = v.
Thus V1 ∩ V2 = {0} and we conclude that V = V1 ⊕ V2 .
Finally, we have already seen that the (now unique) way to write v ∈ V as a sum
v = v1 + v2 , vi ∈ Vi , is v = P v + (I − P )v so P v = v1 as required.
S
6. First, for any Ai ∈ Xi , we have A1 , A2 ≤ X1 X2 so that A1 A2 ≤ X1 X2 also. Thus Ai ∈Xi A1 A2 ⊂
X1 X2 .
For the converse, suppose that Xi = P(Ui ), for U1 , U2 ≤ V and let [v] ∈ X1 X2 . Thus
v ∈ U1 + U2 and we may write v = u1 + u2 with ui ∈ Ui . If both of the ui are non-zero, set
Ai = [ui ] and then [v] ∈ A1 A2 . Otherwise, [v] ∈ Xi , for some i, so take Ai = [v] and choose
any Aj ∈ Xj , j 6= i to get [v] ∈ A1 A2 .
7. The dimension formula rapidly yields dim(L1 L2 ) = 3 so that L1 L2 = P(R4 ). Thus, by question
6, for any point A ≤ P(R4 ), there is a line L through A intersecting both L1 and L2 .
For uniqueness, if L0 was another such line then L and L0 lie in a plane (since they intersect)
and that plane contains the intersections L∩L1 , L0 ∩L1 (necessarily distinct otherwise L = L0 )
and so the unique line joining these which is L1 . Similarly, L2 lies in this plane and so, by
Theorem 1.5, L1 and L2 intersect: a contradiction.
8. (a) In lectures, we saw that the affine line through x and y is contained in a projective
line L which intersects the hyperplane at infinity in [0, y − x]. By definition, the affine
lines through x, y and x0 , y 0 are parallel if and only if [y − x] = [y 0 − x0 ] if and only if
[0, y − x] = [0, y 0 − x0 ].
(b) Let `i , i = 1, 2, 3 be affine lines and xi 6= yi ∈ `i . Then y1 − x1 = 1(y1 − x1 ) so that
`1 ∼ `1 and ∼ is reflexive. If `1 ∼ `2 , then (y2 − x2 ) = λ(y1 − x1 ), for some λ ∈ F× ,
and then (y1 − x1 ) = λ−1 (y2 − x2 ) so that `2 ∼ `1 and ∼ is symmetric. If, additionally,
`2 ∼ `3 , we have (y3 − x3 ) = µ(y2 − x2 ) yielding (y3 − x3 ) = µλ(y1 − x1 ) so that `3 ∼ `1
and ∼ is transitive.
(c) Define a map on equivalence classes by sending the equivalence class of ` to the point
where the corresponding projective line intersects the hyperplane at infinity. By part (a),
this is well-defined and injective. For surjectivity, let A be a point on the hyperplane
at infinity, B any point in the affine part of P(Fn+1 ) and set ` = AB ∩ Fn . Since AB
intersects the hyperplane at infinity in A, we are done.
 M231: Exercise sheet 3

1. Here is a perspective picture of train tracks

in a desert. The sleepers on the track are
parallel and 1 metre apart. I have drawn the
first two sleepers. Explain how to accurately
draw the next sleeper.
2. Show that three points in a projective space
of dimension at least 2 lie on a projective
plane which is unique unless the points are collinear.
3. Let P, P 0 be distinct projective planes in a 3-dimensional projective space. Show that
P ∩ P 0 is a projective line.
Points in general position
4. Prove that the points in the projective plane P(F3 ) with homogeneous coordinates [1, 0, 0],
[0, 1, 0], [0, 0, 1], [1, 1, 1] are in general position in P(F3 ).
5. Let A0 , . . . , An+1 be n + 2 points in an n-dimensional projective space P(V ).
(a) Show that A0 , . . . , An+1 are in general position if and only if no n + 1 of them lie
in a proper projective subspace.
(b) Now let τ : P(V ) → P(W ) be a bijective projective transformation. Show that
if A0 , . . . , An+1 are in general position in P(V ) then τ (A0 ), . . . , τ (An+1 ) are in
general position in P(W ).
(c) Suppose that A0 , . . . , An+1 are in general position in P(V ) and τ1 , τ2 : P(V ) →
P(U ) are projective transformations that agree on the Ai : τ1 Ai = τ2 Ai , for i =
0, . . . , n + 1. Show that τ1 = τ2 .
Around Desargues’ Theorem
6. In the setting of Desargues’ Theorem, explain carefully why corresponding sides of the
two triangles meet.
7. Here is a cheap proof of Desargues’ Theorem in the case when dim P(V ) ≥ 3 and the
triangles ABC and A0 B 0 C 0 lie in different planes p and p0 .
(a) Prove that P, A, A0 , B, B 0 , C, C 0 lie in a 3-dimensional projective subspace (so that
we can restrict attention to that subspace and assume dim P(V ) = 3).
(b) Show that the intersections Q, R, S all lie in P ∩ P 0 which is a line by question 3.
8. The converse of Desargues’ Theorem is also true: let ABC and A0 B 0 C 0 be two triangles
with the property that corresponding pairs of sides meet in collinear points. Then the
lines AA0 , BB 0 and CC 0 are concurrent unless one of the triangles is collinear (in which
case the result may be false).
Prove this! (We shall see a conceptual proof later in lectures).
9. Desargues’ Theorem provides a configuration of 10 lines and 10 points with 3 of the points
on each line and 3 of the lines through each point.
In fact, this configuration is highly symmetrical: one can take any of the 10 points as
the starting point of perspective, find two triangles whose vertices lie on the three lines
through that point and whose sides intersect at the remaining three points. (Thus we get
ten instances of Desargues’ Theorem for the price of one!)

Home page: http://go.bath.ac.uk/ma30231

(a) Convince yourself of the truth of this assertion by staring at the picture of Desar-
gues’ Theorem.
In particular, what are the two triangles in perspective from Q?
(b) Must any configuration of 10 points and 10 lines, with 3 lines through each point
and 3 points on each line, come from Desargues’ Theorem? Hint: ask the internet!

March 21, 2013

M231: Exercise sheet 3—Solutions

1. Because the sleepers are equidistant and parallel, it follows that the diagonals between suc-
cessive sleepers are parallel and so, in the picture, these diagonals must meet in the line at
infinity. Here is the picture that tells the whole story:

P Q

C C0
B B0

A A0

In more detail, with points labelled as in the picture, the diagonal AB 0 must intersect the
diagonal BC 0 at some point Q on the line at infinity. But we find Q as the intersection of
AB 0 with the line at infinity and then C 0 as the intersection of BQ with the train track A0 B 0 .
Thus we have constructed C 0 and we build C similarly and join them with a line segment to
get the next sleeper.
2. Let the three points be A, B, C. These all lie on the join of the lines AB and AC. Unless
our points are collinear, these lines are distinct and intersect at A only so that their join P
has dimension 1 + 1 − 0 = 2 and so is a plane. If A, B, C lie in some other plane P 0 then
AB, AC ≤ P 0 so their join P ≤ P 0 and we must have equality since both P and P 0 are
2-dimensional.
If A, B, C lie on a line L and O is any point not on that line then the join OL has dimension
0 + 1 − (−1) = 2 and so is a plane containing A, B, C.
3. The argument for this should be familiar by now:

3 = dim P(V ) ≥ dim(P P 0 ) = 2 + 2 − dim(P ∩ P 0 )

so that dim(P ∩ P 0 ) ≥ 1. Moreover dim(P ∩ P 0 ) < dim P = 2 otherwise P ≤ P 0 so that

dim(P ∩ P 0 ) = 1 as required.
4. We must see that each three out of the four vectors (1, 0, 0), (0, 1, 0), (0, 0, 1) and (1, 1, 1) are
linearly independent. One fast way to do this is to check that the four determinants

1 0 0 1 0 0 1 0 0 0 1 0
0 1 0 , 0 1 0 , 0 0 1 , 0 0 1
0 0 1 1 1 1 1 1 1 1 1 1

are non-zero which is the work of seconds.

5. (a) n + 1 vectors v1 , . . . , vn+1 in an n + 1-dimensional vector space V are linearly independent
if and only if they span V if and only if they lie in no proper subspace of V .
(b) Suppose not, then n + 1 of the τ (Ai ) lie in a proper projective subspace X whence the
same Ai lie in the proper subspace τ −1 X ⊂ P(V ): a contradiction.
 P
(c) We may suppose that Ai = [vi ], i = 0, . . . , n + 1, where i vi = 0. Let τ1 , τ2 be induced
by injective linear maps T1 , T2 : V → U . We are told that [T1 vi ] = [T2 vi ], for all i, so
that there are λi ∈ F× , i = 0, . . . , n + 1, with T2 vi = λi T1 vi , for all i. Now,
X X X
0= T2 vi = λi T1 vi = T1 ( λi vi )
i i i
P
and, since T1 injects, we deduce that i λi vi = 0. Thus both the vi and the λi vi are
representative vectors for the Ai which sum to zero and so, by the uniqueness part of
Lemma 1.7, differ by a common scale. Thus all λi = λ, say, and T2 and λT1 agree on the
basis v1 , . . . , vn+1 and so everywhere. We conclude that τ1 = τ2 .
6. We prove that the lines BC and B 0 C 0 meet. We know that BB 0 and CC 0 meet at P and so
lie in a projective plane. Thus BC and B 0 C 0 lie in that plane and so meet.
7. (a) P, A, A, B, B 0 all lie in the plane π containing the lines AA0 and BB 0 . The remaining
points lie on the line CC 0 which intersects this plane exactly at the point P (otherwise
CC 0 lies in the plane which then contains both triangles contrary to assumption). The
join of this line and plane has dimension 2 + 1 − 0 = 3.
(b) Consider Q = BC ∩ B 0 C 0 : both B and C lie in the plane p whence the line BC lies in p
also. Similarly, B 0 C 0 lies in p0 so that Q = BC ∩ B 0 C 0 lies in p ∩ p0 . The same argument
shows that R and S also lie on the line p ∩ p0 .
8. Here is a beautiful argument I learned from Gemma Down: set P := BB 0 ∩ CC 0 as before set
Q = BC ∩ B 0 C 0 , R = AC ∩ A0 C 0 and S = AB ∩ A0 B 0 . We observe that Q, R, S are distinct
otherwise the one of the triangles will be collinear.
Now for the clever bit: observe that the triangles SBB 0 and RCC 0 are in perspective from Q!
So apply Desargues’ Theorem to see that SB ∩ RC = A, SB 0 ∩ RC 0 = A0 and BB 0 ∩ CC 0 = P
are collinear. Thus AA0 goes through P and so is concurrent with BB 0 and CC 0 .
9. (a) The triangles in perspective from Q are RCC 0 and SBB 0 with corresponding sides inter-
secting at A, A0 and P .
(b) There are configurations that do not come from Desargues’ Theorem. Here is an example1 :

1
From the wikipedia article on Desargues’ Theorem, doh!
M231: Exercise sheet 4

On projections
1. Let L1 and L2 be distinct projective lines in a projective plane that intersect at A.
Let τ : L1 → L2 be a projective transformation such that τ A = A.
Show that τ is a projection from some point of the plane.
Hint: Choose B, C on L1 distinct from A and let B 0 = τ B, C 0 = τ C. If τ was a
projection, where would its centre have to lie?
2. Let L1 and L2 be distinct projective lines in a projective plane and τ : L1 → L2
a projective transformation. We are going to show that τ is a composition of two
projections.
Let A, B, C be distinct points on L1 and let A0 = τ A, B 0 = τ B and C 0 = τ C. Without
loss of generality, assume that neither A or A0 are in L1 ∩ L2 .
Set P = AB 0 ∩ A0 B, Q = AC 0 ∩ A0 C and let L0 be the line P Q.
Let τ1 : L1 → L0 be the projection with centre A0 and τ2 : L0 → L2 be the projection
with centre A.
(a) Prove that τ = τ2 ◦ τ1 .
Hint: What does τ2 ◦ τ1 do to A, B, C?
(b) What happens to L0 if we replace A and A0 by B and B 0 ?
Fun with cross-ratios
3. Let A, B, C, D be distinct points in P(F2 ) with inhomogeneous coordinates a, b, c, d ∈
F ∪ {∞}. How does the formula for the cross-ratio look when a = ∞? Be careful to
justify your answer.
4. What is the meaning of the sign of the cross-ratio of four points in P(R2 )?
5. Let A, B, C, D be distinct points on a projective line. Prove:
(a) (A, B; C, D) = (B, A; D, C) = (C, D; A, B) = (D, C; B, A).
(b) (A, B; C, D) = 1/(A, B; D, C).
(c) Express (A, C; B, D) in terms of (A, B; C, D).
6. Let A, B, C, D be distinct points on a projective line. We say that A, B and C, D are
harmonically separated if (A, B; C, D) = −1.
(a) Prove that a, b ∈ F are harmonically separated by c, ∞ if and only if c is the
midpoint of the segment ab.
(b) Prove that A, B and C, D are harmonically separated if and only if (A, B; C, D) =
(A, B; D, C).
More substantial questions
7. Let A, B, C, D be four points in general position in a projective plane. Set P =
AD ∩ BC, Q = AC ∩ BD, R = AB ∩ CD. This configuration of four points and the
six lines joining them is called a complete quadrangle. (Draw a picture!)
Now let S = QR ∩ AD and T = QR ∩ BC. Show that Q, R and S, T are harmonically
separated.
Can you see two other sets of harmonically separated points in this picture?
8. Prove Brianchon’s Theorem: Let the sides AB 0 , B 0 C, CA0 , A0 B, BC 0 , C 0 A of a
hexagon pass alternately through two (different) points P and Q in a projective plane.
Then the lines joining opposite vertices AA0 , BB 0 , CC 0 are concurrent.
Hints: Either choose a basis so that P, C, Q, C 0 have homogeneous coordinates [1, 0, 0],
[0, 1, 0], [0, 0, 1], [1, 1, 1] and compute homogeneous coordinates of everything in sight
or set α = P , β = C 0 , γ = B and then α0 = Q, β 0 = B 0 , γ 0 = C and think Pappus
thoughts (this is much faster).

March 21, 2013

M231: Exercise sheet 4—Solutions

1. Let P = BB 0 ∩ CC 0 and contemplate the projection τ1 : L1 → L2 with centre P . By

construction, we have τ1 (B) = B 0 and τ1 (C) = C 0 but, also, τ1 (A) = A. Thus τ and τ1 agree
on three distinct points of L1 (thus three points in general position) and so agree everywhere:
τ = τ1 so that τ is a projection. Here is the picture:

C
L2 B
A
B0
C0

L1

2. (a) Let R = AA0 ∩ L0 (see the picture below). Then we have τ1 (A) = R, τ1 (B) = P and
τ1 (C) = Q. Moreover, τ2 (R) = A0 , τ2 (P ) = B 0 and τ2 (C) = C 0 . Thus τ and τ2 ◦ τ1 agree
on the three points A, B, C in general position and so everywhere.

C
B
A

L1
Q
P
R
L2
A0 B0 C0
0
L

(b) If we use B and B 0 as the centres of the projections, then L0 gets replaced by the line
through BC 0 ∩ B 0 C and BA0 ∩ B 0 A which is exactly L0 again by Pappus’ theorem!
In fact, no matter what points A, B, C are start off with, we always get the same line L0
(exercise!) which is called the axis of τ .
3. There are two ways to do this. First the quick and dirty computation: write a = 1/0 in the
usual formula for cross-ratio and multiply top and bottom by 0:
(a − c)(b − d) (1/0 − c)(b − d) (1 − 0c)(b − d)
(A, B; C, D) = = = = (b − d)/(b − c).
(a − d)(b − c) (1/0 − d)(b − c) (1 − 0d)(b − c)
A rigorous argument uses the determinant formula with homogeneous coordinates: we have
A = [0, 1], B = [1, b], C = [1, c], D = [1, d] so that
0 1 1 b
1 c 1 d
(A, B; C, D) = = (b − d)/(b − c)
0 1 1 b
1 d 1 c

(which, among other things, justifies the use of the quick and dirty approach).
4. Choose a basis of R2 so that our four points A, B, C, D have inhomogeneous coordinates
a, b, c, d ∈ R and
(a − c)/(a − d)
(A, B; C, D) = .
(b − c)/(b − d)
Now (A, B; C, D) > 0 if and only if (a − c)/(a − d) and (b − c)/(b − d) have the same sign.
Moreover, (a − c)/(a − d) < 0 if and only if a − c and a − d have different signs if and only if
a lies between c and d. Thus (A, B; C, D) < 0 if and only if one of a, b lies between c, d and
the other doesn’t or, equivalently, the line segments AB and CD overlap.
5. We use the determinant formula when we can so as not to have to make special arguments
about ∞.
(a) These are easy:

det(A, C) det(B, D) det(B, D) det(A, C)

(A, B; C, D) = = = (B, A; D, C).
det(A, D) det(B, C) det(B, C) det(A, D)

Now observe that det(A, C) = − det(C, A) and so on giving:

det(A, C) det(B, D) det(C, A) det(D, B)

(A, B; C, D) = = = (C, D; A, B).
det(A, D) det(B, C) det(D, A) det(C, B)

Now apply the first identity to (C, D; A, B) to get (C, D; A, B) = (D, C; B, A).
(b) We have:
det(A, D) det(B, C)
(A, B; D, C) = = 1/(A, B; C, D).
det(A, C) det(B, D)
(c) Here we will make a quick and dirty argument with inhomogeneous coordinates. Let
A, B, C, D have inhomogeneous coordinates a, b, c, d. Then

(a − b)(c − d) (a − c + c − b)(c − d)
(A, C; B, D) = =
(a − d)(c − b) (a − d)(c − b)
(a − c)(c − d) c−d
= +
(a − d)(c − b) a − d
(a − c)(c − b + b − d) c−d
= +
(a − d)(c − b) a−d
a−c (a − c)(b − d) c−d
= + +
a − d (a − d)(c − b) a − d
= 1 − (A, B; C, D).

There is a determinant approach to this which is the start of a long and beautiful story
but not as elementary. I may expand on this later.
6. (a) We compute (a, b; c, ∞) using the ∞ = 1/0 trick:

(a − c)(b − 1/0) a−c

(a, b; c, ∞) = = .
(a − 1/0)(b − c) b−c

Now (a − c)/(b − c) = −1 if and only if c = (a + b)/2,which is what we wanted.

(b) From question 5(b), we know that (A, B; D, C) = 1/(A, B; C, D) so that these two cross-
ratios agree if and only if (A, B; C, D)2 = 1 yielding (A, B; C, D) = ±1. However, the
cross-ratio of four distinct points is never 1 (why?) and so we are done.
7. We begin by drawing the picture:
 P

A M B Q
R T
S
C
N

Here I have added the line P R and the points M = P R ∩ AB and N = P R ∩ DC because
they will be useful.
Project from QR to AB with centre P to get (Q, R; S, T ) = (Q, M ; A, B). Now project from
QR to DC with centre R to get (Q, M ; A, B) = (Q, N ; C, D) (this is the cunning bit: we have
swopped the order of the “A, B” pair). Finally, project from DC back to QR with centre P
again to get (Q, N ; C, D) = (Q, R; T, S).
Thus (Q, R; S, T ) = (Q, R; T, S) so that Q, R and S, T are harmonically separated by ques-
tion 6.
Similarly P, R and M, N are harmonically separated as are P, Q and the intersections of P Q
with AC and BD (not shown in the picture).
8. The situation is drawn in Figure 1. This result seems related to Pappus’ Theorem and indeed
it is. I will give three proofs in decreasing order of length and increasing order of cleverness.
For the first proof, we argue as in Pappus’ Theorem by making a good choice of basis and
then computing everything in sight with homogeneous coordinates. So convince yourself that
P, C, Q, C 0 are in general position and then choose a basis as in the hint. Our plan is to compute
the homogeneous coordinates for everything in sight and so get our hands on equations for
the three lines AA0 , BB 0 and CC 0 .
First we take the easy case: C and C 0 both lie on the line with equation λ0 − λ2 = 0 and so,
by uniqueness of the line through C and C 0 , CC 0 has equation λ0 − λ2 = 0.
Similarly, QC 0 has equation λ0 − λ1 = 0 while P C has equation λ2 = 0. Now A lies on
QC 0 and so has homogeneous coordinates [a, a, 1], some a ∈ R, while A0 lies on P C and so
A0 = [1, a0 , 0], for some a0 ∈ R. The line AA0 therefore has an equation µ0 λ0 + µ1 λ1 + µ2 λ2 = 0
where
aµ0 + aµ1 + µ2 = 0
µ0 + a 0 µ1 =0
which we solve to conclude that AA0 has equation −a0 λ0 + λ1 + a(a0 − 1)λ2 = 0.
Again, B is on P C 0 and B 0 is on QC and we find that B = [1, b, b], B 0 = [0, b0 , 1], for b, b0 ∈ R,
so that BB 0 has equation b(b0 − 1)λ0 + λ1 − b0 λ2 .
Now notice that P, A, B 0 are collinear as are Q, B, A0 so that the following determinants vanish:

1 0 0 0 0 1
a a 1 , 1 a0 0 ,
0 b0 1 1 b b

from which we learn that a = b0 and b = a0 .

We can now bake the cake and see that the three target lines are concurrent. For this, we
 P

A0
B
A
C0
Q

B0

Figure 1: Brianchon’s Theorem

need a point [λ0 , λ1 , λ2 ], with λi not all zero, solving the three equations for all three lines:
  
1 0 −1 λ0
 −a0 1 a(a0 − 1) λ1  = 0
b(b0 − 1) 1 −b0 λ2

and that happens exactly when the determinant

1 0 −1
−a0 1 a(a0 − 1)
b(b0 − 1) 1 −b0

vanishes. However, we readily compute that determinant to be b(b0 − 1) + a0 − (a(a0 − 1) + b0 )

which is indeed zero since a = b0 and b = a0 . Phew!
Our second proof reduces the whole affair to an appeal to Pappus’ Theorem by a careful
relabelling of the points. Set α = P , β = C 0 and γ = B and then α0 = Q, β 0 = B 0 and γ 0 = C
to get two collinear triples of points.
Observe that αβ 0 ∩ α0 β = A, αγ 0 ∩ α0 γ = A0 and Pappus says that A, A0 are collinear with
R := βγ 0 ∩ β 0 γ = CC 0 ∩ BB 0 . Otherwise said, AA0 , BB 0 and CC 0 are concurrent at R!.
For a third proof, observe that Brianchon’s Theorem is simply Pappus’ Theorem in the dual
projective plane reinterpreted as a statement in our original plane via the Principle of Duality!
M231: Exercise sheet 5

1. Let v1 , . . . , vn be a basis of a vector space V and define v1∗ , . . . , vn∗ in V ∗ by

(
∗ 1 if i = j;
vi (vj ) =
0 otherwise.

and extending by linearity.

(a) If v = ni=1 λi vi , what is vj∗ (v)?
P

(b) Prove that v1∗ , . . . , vn∗ is a basis of V ∗ .

2. Let U1 , U2 ≤ V with annihilators U1◦ , U2◦ ≤ V ∗ . Prove:
(a) (U1 + U2 )◦ = U1◦ ∩ U2◦ .
(b) (U1 ∩ U2 )◦ = U1◦ + U2◦ ;
3. Let V be a 3-dimensional vector space with basis v0 , v1 , v2 and suppose that P ∈ P(V )
has homogeneous coordinates [λ0 , λ1 , λ2 ].
(a) Describe the lines in P(V ) corresponding by duality to the points [v0∗ ], [v1∗ ], [v2∗ ] in
P(V ∗ ).
(b) Suppose that P ∈ P(V ) has homogeneous coordinates [λ0 , λ1 , λ2 ]. Show that the
corresponding line P ◦ in P(V ∗ ) is given by {[µ0 , µ1 , µ2 ] ∈ P(V ∗ ) : λ0 µ0 + λ1 µ1 +
λ2 µ2 = 0}, where the homogeneous coordinates of P(V ∗ ) are with respect to the
dual basis of the vi .
(c) Let A, B, C ∈ P(V ) have homogeneous coordinates [2, 1, 0], [0, 1, 1], [−1, 1, 2] re-
spectively with respect to the vi .
Compute the homogeneous coordinates with respect to the dual basis of the points
in P(V ∗ ) corresponding to the lines AB, AC, BC.
4. Recall the cross product on F3 : if a = (a0 , a1 , a2 ) and b = (b0 , b1 , b2 ) then

e0 e1 e2
a × b = (a1 b2 − a2 b1 , a2 b0 − a0 b2 , a0 b1 − a1 b0 ) = a0 a1 a2 ,
b0 b1 b2

where e0 , e1 , e2 are the usual basis of F3 .

Now let P(V ) be a projective plane and use dual bases of V and V ∗ to compute
homogeneous coordinates.
(a) Let A, B ≤ P(V ) have homogeneous coordinates [a] = [a0 , a1 , a2 ], [b] = [b0 , b1 , b2 ].
Show that the line AB has homogeneous coordinates [a × b].
(b) Let L1 , L2 ≤ P(V ) be lines whose homogeneous coordinates are [µ] = [µ0 , µ1 , µ2 ]
and [ν] = [ν0 , ν1 , ν2 ]. Show that the intersection L1 ∩ L2 has homogeneous coor-
dinates [µ × ν].
Hint: Either argue directly using determinants or use (a) and the Principle of
Duality!
5. A complete quadrangle consists of four points in a plane in general position along with
the six lines through pairs of these points.
(a) Draw one!

Home page: http://go.bath.ac.uk/ma30231

 (b) What is the dual configuration? That is: what configuration of points and lines
in a plane P(V ) corresponds to a complete quadrangle in P(V ∗ )?
6. Let P(V ) be a projective line. In P(V ), points are the same as hyperplanes so the
duality isomorphism gives a bijection P(V ) → P(V ∗ ).
Show that this map is a projective transformation.
7. Let L1 , L2 , L3 , L4 be four concurrent lines in a projective plane P(V ) and let l1 , l2 , l3 , l4
be the corresponding collinear points in P(V ∗ ).
We now have two notions of cross-ratio of the four lines: the li are four points on a
projective line so we can compute (l1 , l2 ; l3 , l4 ). In lectures, we defined (L1 , L2 ; L3 , L4 )
to be (A1 , A2 ; A3 , A4 ) where Ai = Li ∩ L for L some (any!) line not concurrent with
the Li .
Prove that these notions coincide: (l1 , l2 ; l3 , l4 ) = (L1 , L2 ; L3 , L4 ).

May 20, 2013

M231: Exercise sheet 5—Solutions

1. (a) This is very easy:

n
X n
X
vj∗ (v) = vj∗ ( λi v i ) = λi vj∗ (vi ) = λj .
i=1 i=1
Pn
(b) Linear independence: ifP i=1 λi vi∗ = 0, evaluate on vj to get λj = 0.
n
Span: if f ∈ V ∗ , f = i=1 f (vi )vi∗ : indeed, both sides evaluate to f (vj ) on vj and so
agree on a basis of V and, therefore, everywhere.
2. (a) Let f ∈ (U1 + U2 )◦ . Then, in particular f vanishes on U1 and U2 and so lies in U1◦ ∩ U2◦ .
Thus (U1 + U2 )◦ ≤ U1◦ ∩ U2◦ .
Conversely, if f ∈ U1◦ ∩ U2◦ , then f vanishes on U1 and U2 and so, by linearity, on U1 + U2 .
So f ∈ (U1 + U2 )◦ and U1◦ ∩ U2◦ ≤ (U1 + U2 )◦ .
(b) If f ∈ U1◦ + U2◦ then we write f = f1 + f2 with fi ∈ Ui◦ . Then f1 vanishes on U1 and so,
in particular, on U1 ∩ U2 and similarly, f2 vanishes on U1 ∩ U2 . Thus both fi ∈ (U1 ∩ U2 )◦
whence f = f1 + f2 ∈ (U1 ∩ U2 )◦ . This gives U1◦ + U2◦ ≤ (U1 ∩ U2 )◦ . For equality, we
show that both subspaces have the same dimension: first we have:

dim(U1 ∩ U2 )◦ = dim V − dim(U1 ∩ U2 ).

On the other hand,

dim U1◦ + U2◦ = dim U1◦ + dim U2◦ − dim(U1◦ ∩ U2◦ )

= dim U1◦ + dim U2◦ − dim(U1 + U2 )◦ ,

by part (b)

= (dim V − dim U1 ) + (dim V − dim U2 ) − (dim V − dim(U1 + U2 ))

= dim V − dim U1 − dim U2 + dim(U1 + U2 ) = dim V − dim(U1 ∩ U2 ).

3. (a) The line corresponding to [v0∗ ] is P(ker v0∗ ). Since v0∗ (v1 ) = v0∗ (v2 ) = 0, this is the line
through [v1 ] and [v2 ]. More geometrically, the [vj∗ ] correspond to the sides of the triangle
whose vertices are the [vi ] with [vi∗ ] corresponding to the side opposite [vi ].

[v0 ]
[v1∗ ]
[v2∗ ]
[v1 ] [v2 ]
[v0∗ ]

vector v then P ◦ = {[f ] ∈ P(V ∗ ) : f (v)

P
(b) If P has representative = 0}. Now v = λi vi and
writing f ∈ V ∗ as µj vj∗ we readily compute that f (v) = i λi µi whence the result.
P P

(c) Using the first part, the line AB has homogeneous coordinates [µ0 , µ1 , µ2 ] satisfying
2µ0 + µ1 = 0 (since the line goes through A) and µ1 + µ2 = 0 (since the line goes
through B). Solve these linear equations to see that AB has homogeneous coordinates
[µ0 , −2µ0 , 2µ0 ] = [1, −2, 2].
Similarly, AC = [2, −4, 3] and BC = [1, −1, 1].
4. (a) We already know that the equation of the line AB is

λ0 λ1 λ2
a0 a1 a2 = 0
b0 b1 b2

and expanding this out we get µ0 λ0 + µ1 λ1 + µ2 λ2 = 0 for µ = a × b.

5. We give two arguments. First, a direct argument via determinants: let a = (a0 , a1 , a2 ) = µ × ν
and A = [a]. We want to see that A lies on L1 and L2 or, equivalently, that

0 = µ0 a0 + µ1 a1 + µ2 a2
0 = ν0 a0 + ν1 a1 + ν2 a2

. However, the right hand sides of these equations are, respectively,

µ0 µ1 µ2 ν0 ν1 ν2
µ0 µ1 µ2 , µ0 µ1 µ2 ,
ν0 ν1 ν2 ν0 ν1 ν2

which both vanish.

For a second argument, simply apply the Principle of Duality to part (a): For A = L1 ∩ L2 ,
A◦ is the line through the points L◦i in P(V ∗ ) with homogeneous coordinates [µ] and [ν]. From
part (a), we know that this line is given by an equation with coefficients µ × ν so that, by
question 3(b), A has homogeneous coordinates [µ × ν] also.
6. (a) A complete quadrangle:

B
A

(b) The complete quadrangle consists of 4 points and 6 lines with two of the four points on
each line. Thus the dual configuration consists of 4 lines and 6 points with no three of the
lines concurrent and each point lying on two of the lines. Otherwise said, we have 4 lines
along with the six points of intersection of pairs of these lines. This is called a complete
quadrilateral. Here is a picture:

B
A F
E
C

7. The duality isomorphism sends [v] ≤ P(V ) to [f ] ≤ P(V ∗ ) where f (v) = 0. Choose a basis
v0 , v1 of V and so a dual basis v0∗ , v1∗ of V ∗ . Observe that if v = λ0 v0 + λ1 v1 then (λ1 v0∗ −
λ0 v1∗ )(v) = 0 so that the duality isomorphism is induced by the linear isomorphism λ0 v0 +
λ1 v1 7→ λ1 v0∗ − λ0 v1∗ and so is a projective transformation.
8. Suppose that the Li are concurrent at A so that A◦ ≤ P(V ∗ ) is the corresponding line.
Consider the map τ : A◦ → L given by letting τ (l) be the intersection of the line corresponding
to l with L. We then have τ (li ) = Ai and we will be done if we can show that τ is a projective
transformation.
For this, let A1 = [v1 ], A2 = [v2 ], vi ∈ V , l1 = [f1 ], l2 = [f2 ], fi ∈ V ∗ , so that fi (vi ) = 0
since Ai lies on Li , i = 1, 2. Moreover, adjust the scalings so that f1 (v2 ) = f2 (v1 ) = 1. Now
observe that (λ1 f1 + λ2 f2 )(λ1 v1 − λ2 v2 ) = 0 so that, just as in question 6, using f1 , f2 and
v1 , v2 as bases, τ ([λ1 , λ2 ]) = [λ1 , −λ2 ]. Thus τ is induced by a linear isomorphism and so is a
projective transformation.
M231: Exercise sheet 7

1. (Very easy) Suppose that three collinear points lie on a quadric. Show that the line
lies in the quadric also.
2. Find all of the conics which pass through [1, 0, 0], [1, 1, 0] and [1, 0, 1] and are tangent
to the line λ1 + λ2 = 0 at [1, 0, 0].
3. Here is another useful normal form and parametrisation for a non-singular conic: Let
P, Q, R be distinct points on a non-singular conic C in a projective plane P(V ).
(a) Show that there is a basis of V with respect to which P, Q, R have homogeneous
coordinates [1, 0, 0], [0, 1, 0], [0, 0, 1] respectively and C is defined by the quadratic
form λ0 λ1 + λ0 λ2 + λ1 λ2 .
(b) Using this basis and considering lines from P intersecting the line L defined
by λ0 = 0, show that any point of C has homogeneous coordinates [0, 0, 1] or
[−t, 1 + t, t(1 + t)], for t ∈ F.

Polars
4. Let C be a non-singular conic and L a line. Show that the polar lines to X ∈ L are
all concurrent at some point Y .
Hint: What is the relationship between Y and L?
5. Let C be a non-singular real conic and A a point not on C. Here is a construction of
A⊥ that works even when A is “inside” C. Choose points B0 , C0 on C so that A lies
on neither tangent line B0⊥ , C0⊥ and let B1 , C1 , be the points where the lines B0 A and
C0 A cut C again.
Show that A⊥ is the line through X := B0⊥ ∩ B1⊥ and Y := C0⊥ ∩ C1⊥ .
Draw a picture.
6. (a) Let S = {X1 , X2 } be a non-singular quadric on a projective line P(V ). Let Y ∈
/S
with polar Y ⊥ (which is also a point in P(V )).
Show that the pairs X1 , X2 and Y, Y ⊥ are harmonically separated.
(b) Now let C be a non-singular conic in a projective plane P(V ) and A ∈ / C. Let L
be a line though A which intersects C at two points B1 , B2 . Finally let A0 be the
intersection of L and the polar line to A.
Show that the pairs B1 , B2 and A, A0 are harmonically separated.
Further questions
7. Show that the tangent lines to a non-singular conic in P(V ) form a non-singular conic
in P(V ∗ ).
8. Prove Pascal’s Theorem: let A, B, C, A0 , B 0 , C 0 be six points on a non-singular conic
and set P = AB 0 ∩ A0 B, Q = AC 0 ∩ A0 C, R = BC 0 ∩ B 0 C. Then P , Q and R are
collinear.
Hint: Make full use of question 3 applied to the points A, B, C.
How is Pascal’s Theorem related to Pappus’ Theorem?

May 20, 2013

Home page: http://go.bath.ac.uk/ma30231

M231: Exercise sheet 7—Solutions

1. Let the line be L = P(U ) and the quadric S be defined by a quadratic form Q. If Q|U is
non-zero, then L ∩ S is a quadric in L and so consists of at most two points. However, L ∩ S
contains three points so that Q vanishes on U . Otherwise said L ⊂ S.
2. Any conic C is given by a quadratic form β00 λ20 + β11 λ21 + β22 λ22 + 2β01 λ0 λ1 + 2β02 λ0 λ2 +
2β12 λ1 λ2 . Now [1, 0, 0] ∈ C yields β00 = 0; [1, 1, 0] ∈ C then gives β11 + 2β01 = 0 while
[1, 0, 1] ∈ C gives β22 + 2β02 = 0. Thus our quadratic form has matrix
 
0 β01 β02
β01 −2β01 β12 
β02 β12 −2β02

with tangent line β01 λ1 + β02 λ2 at [1, 0, 0]. Since this line coincides with λ1 + λ2 = 0, we must
have β01 = β02 so our conic is of the form β01 (2λ0 λ1 +2λ0 λ2 −2λ21 −2λ22 )+β12 λ1 λ2 . Conversely,
any conic of this form contains [1, 0, 0], [1, 1, 0] and [1, 0, 1] and is tangent to λ1 + λ2 = 0 at
[1, 0, 0] so these conditions pick out a pencil of conics.
3. (a) First note that P, Q, R are not collinear (otherwise the line would intersect our non-
singular conic in more than two points). Thus the span of representative vectors v0 , v1 , v2
of P, Q, R is three-dimensional so that v0 , v1 , v2 are a basis for V . With
P respect to
this basis, P, Q, R have the desired homogeneous coordinates and, with i,j βij λi λj the
quadratic form defining C, we quickly get that β00 = β11 = β22 = 0. Replacing vi by ai vi
replaces βij by ai aj βij and we try and choose the ai to make all the new βij coincide,
i 6= j. Thus we must solve:

a0 a1 β01 = a0 a2 β02 = a1 a2 β12 .

For this, take a0 = 1 and then see that a1 β01 = a1 a2 β12 yields a2 = β01 /β12 and similarly
a1 = β02 /β12 (note that the βij 6= 0, for i 6= j, since C is non-singular). Now replace the
vi by ai vi to write the quadratic form as
β01 β02
(λ0 λ1 + λ0 λ2 + λ1 λ2 )
β12
and we are done.
(b) We follow the strategy in lectures: for each X on the line L, we look for the second
intersection of the conic with the line P X: we know that any point on the conic arises
this way. Now any point of L is of the form [0, 1, t], t ∈ F, or [0, 0, 1]. This last point is
already on the conic so we have nothing to do there. Otherwise points on the line P X
have representative vectors λ(1, 0, 0) + µ(0, 1, t) = (λ, µ, µt) and such a point lies on our
conic if λµ + λµt + µ2 t = 0. Now λµ + λµt + µ2 t = µ(λ(1 + t) + µt) so one solution of this
is µ = 0 (giving us P ∈ P X ∩ C) and we want the other solution so that λ(1 + t) + µt = 0.
So λ = −t and µ = (1 + t) and our point on the conic has representative vector:

−t(1, 0, 0) + (1 + t)(0, 1, t) = (−t, (1 + t), t(1 + t))

as required.
4. Set Y = L⊥ . Then X ≤ L if and only if Y ≤ X ⊥ . Thus all polar lines to X ≤ L are concurrent
at Y .
5. From lectures, we know that X ⊥ = B0 B1 and Y ⊥ = C0 C1 whence A = X ⊥ ∩ Y ⊥ yielding
A⊥ = XY .
Here is a picture:
 C1

B0
A
B1
C0

A⊥ Y

6. (a) Since X1 , X2 , Y are distinct points on a line P(V ) we can choose a basis v1 , v2 such that
X1 = [1, 0], X2 = [0, 1] and Y = [1, 1]. Otherwise said, Xi = [vi ] and Y = v1 + v2 . If B
is a symmetric bilinear form inducing S then B(vi , vi ) = 0 so that B(v1 + v2 , v1 − v2 ) =
B(v2 , v1 ) − B(v1 , v2 ) = 0 so that Y ⊥ = [v1 − v2 ]. Thus Y = [1, −1] and any easy
computation shows that (X1 , X2 ; Y, Y ⊥ ) = −1.
Here is a more conceptual argument: the map β : P(V ) → P(V ) given by X 7→ X ⊥
is a projective transformation (why?) which fixes the Xi and swops Y and Y ⊥ (since
Y ⊥⊥ = Y and Xi⊥ = Xi . Thus, since β preserves the cross-ratio, (X1 , X2 ; Y, Y ⊥ ) =
(X1 , X2 ; Y ⊥ , Y ) and then harmonic separation follows from a result on a previous sheet.
(b) This is immediate from the first part: let S = C ∩ L, let Xi = Bi , A = Y . Then the polar
Y ⊥ of Y with respect to S is A0 .
7. Let B be a non-degenerate symmetric bilinear form determining a conic C and β : V → V ∗
the associated isomorphism given by β(v)(w) = B(v, w). Then the tangent line to v ∈ C is
P(ker β(v)) so that the map C → P(V ∗ ) which sends X ∈ C to its tangent line is simply β|C .
So we must show that β(C) is a conic in P(V ∗ ). But [f ] ∈ β(C) if and only if β −1 (f ) ∈ C if
and only if B(β −1 (f ), β −1 (f )) = 0. Thus β(C) is the conic in P(V ) given by the symmetric
bilinear form B̂(f, g) = B(β −1 (f ), β −1 (g)) which is non-degenerate because B is.
8. Pappus’ Theorem is Pascal’s Theorem for a singular conic (a line-pair) and we shall use a
similar strategy to prove it.
Using question 3, we choose a basis with respect to which A = [1, 0, 0], B = [0, 1, 0] and
C = [0, 0, 1] and the conic is defined by λ0 λ1 +λ0 λ2 +λ1 λ2 = 0. In addition, from question 3(b),
we know that A0 = [−a, 1 + a, a(1 + a)], B 0 = [−b, 1 + b, b(1 + b)], C 0 = [−c, 1 + c, c(1 + c)], for
distinct a, b, c ∈ F. Since A0 , B 0 , C 0 are distinct from A, B, C, we note that a, b, c ∈ F \ 0, −1.
Now we get our hands on the homogeneous coordinates of P, Q, R. For P = [p0 , p1 , p2 ], we
note that A, P, B 0 and B, P, A0 are collinear so that
1 0 0 0 1 0
p0 p1 p2 = 0 = p0 p1 p2 ,
−b 1 + b b(1 + b) −a 1 + a a(1 + a)
yielding b(1 + b)p1 − (1 + b)p2 = 0 and a(1 + a)p0 + ap2 = 0. We can divide by a and 1 + b
so this boils down to bp1 = p2 and (1 + a)p0 = −p2 whence P = [−p2 /(1 + a), p2 /b, p2 ] =
[−b, 1 + a, b(1 + a)].
Similarly, Q = [−a, 1 + a, c(1 + a)] and R = [−b, 1 + b, b(1 + c)] and a brutal computation gives:
−b 1+a b(1 + a)
−a 1+a c(1 + a) = 0
−b 1+b b(1 + c)
so that P.Q, R are collinear.
M231: Exercise sheet 8

Subspaces in quadrics
1. Let S be a non-singular 2n-dimensional quadric in a complex projective space. Show
that S contains a projective subspace of dimension n.
Hint: Show that there is a basis v0 , . . . , v2n+1 with respect to which the quadric is
given by λ0 λ1 + λ2 λ3 + · · · + λ2n λ2n+1 = 0 and contemplate the span of the v2i .
2. Let S ⊂ P(R4 ) be the quadric given by µ0 µ1 = µ2 µ3 and contemplate the map
f : P(R2 ) × P(R2 ) → P(R4 ) given by

f ([s0 , s1 ], [t0 , t1 ]) = [s0 t0 , s1 t1 , s0 t1 , s1 t0 ].

Show that f is a bijection onto S.

3. Let S ⊂ P(R4 ) be a real, non-singular quadric surface. For any A ∈ S, we know that
A⊥ ∩ S is a line-pair or a point. Suppose that A⊥ ∩ S is a line-pair.
Show that, for all B ∈ S, B ⊥ ∩ S is a line-pair also.
Deduce that if A⊥ ∩ S is a point, so is each B ⊥ ∩ S.
Hint: Think about how S intersects the line A⊥ ∩ B ⊥ .
On pencils
4. Let A1 , A2 , A3 , A4 be four points in general position in a projective plane and let X
be a distinct fifth point.
(a) Show that there is a unique conic containing A1 , A2 , A3 , A4 , X.
(b) Show that conic is non-singular if and only if no three of these five points are
collinear.
5. Find all of the conics which pass through the points [1, 0, 0], [1, 0, 1], [1, 1, 0], and
[1, 1, 1] in P (R3 ). What are the singular conics in this pencil?
Hint: Answer the second part first!
6. Show that the set of conics tangent to the line λ0 − λ1 + λ2 = 0 at [1, 1, 0] and to the
line λ1 + λ2 = 0 at [1, 0, 0] is a pencil.
What are the singular conics in this pencil?
What is the base locus of the pencil?
Further questions
7. Let Q0 , Q1 be quadratic forms on a complex vector space V . Show that the corre-
sponding quadrics in P(V ) coincide if and only if Q1 = λQ0 , for some λ ∈ C \ {0}.
Hint: Deal with the case dim P(V ) = 1 first.
8. Let P(V ) be a complex projective line.
(a) Show that the space P(S 2 V ∗ ) of quadrics in P(V ) is a projective plane.
(b) Show that the set of singular quadrics is a non-singular conic in P(S 2 V ∗ ).
(c) Deduce the existence of a bijection from P(V ) to a non-singular conic.
[This is yet another proof of the result proved in lectures that a non-singular conic
is bijective with a projective line.]

April 25, 2013

Home page: http://go.bath.ac.uk/ma30231

M231: Exercise sheet 8—Solutions

1. Since the quadric is non-singular, there is a basis w0 , . . . , w2n+1 of the underlying 2n + 2-

dimensional vector space with respect to which the symmetric bilinear form B defining the
quadric is given by X X X
B( ai wi , bi wi ) = ai bi
√ √
Now set v2i = w2i − −1w2i+1 , v2i+1 = w2i + −1w2i+1 , 0 ≤ i ≤ n and note that B(v2i , v2j ) =
0 for 0 ≤ i, j ≤ n so that span{v2i } is an n-dimensional subspace of the quadric.
2. Let us prove injectivity first: if f ([s0 , s1 ], [t0 , t1 ]) = f ([a0 , a1 ], [b0 , b1 ]) then we have

[s0 t0 , s1 t1 , s0 t1 , s1 t0 ] = [a0 b0 , a1 b1 , a0 b1 , a1 b0 ]

and at least one of the homogeneous coordinates in non-zero, s1 t0 , say. Then a1 b0 6= 0 also
and we divide to get

[s0 /s1 , t1 /t0 , s0 t1 /s1 t0 , 1] = [a0 /a1 , b1 /b0 , a0 b1 /a1 b0 , 1]

so that, in particular, s0 /s1 = a0 /a1 and t1 /t0 = b1 /b0 yielding [s0 , s1 ] = [a0 , a1 ] and [t0 , t1 ] =
[b0 , b1 ]. This proves injectivity when s1 t0 6= 0. The remaining cases are dangerously similar
and reduce to the one we have already covered by a reordering of bases.
For surjectivity, let [µ0 , µ1 , µ2 , µ3 ] ∈ S so that µ0 µ1 = µ2 µ3 . Suppose that µ3 6= 0 (the other
cases are similar). Guided by the formulae above, contemplate

f ([µ0 , µ3 ], [µ3 , µ1 , ]) = [µ0 µ3 , µ3 µ1 , µ0 µ1 , µ23 ]

= [µ0 µ3 , µ3 µ1 , µ2 µ3 , µ23 ]
= [µ0 , µ1 , µ2 , µ3 ]

where to get from the first line to the second we used µ0 µ1 = µ2 µ3 . Thus f surjects.
3. Let B 6= A ∈ S,A⊥ ∩ S = L1 ∪ L2 and L = A⊥ ∩ B ⊥ which is a line. Since the Li and L
lie in the plane A⊥ , they must intersect at points that lie in A ∩ B ∩ S and there are two
possibilities. First, if L intersects the Li at separate points X1 , X2 , then the Xi are two points
in B ⊥ ∩ S so that the latter contains more than one point and so must be a line-pair. The
only other possibility is for the Li to intersect L at A. Then A ∈ B ⊥ ∩ S and now B ⊥ ∩ S
contains the two distinct points A, B and that again B ⊥ ∩ S is a line-pair.
We have shown that as soon as one tangent plane intersects S in a line-pair, all tangent planes
intersect S that way. So, as soon as one tangent plane intersects in a point, they all do.
4. We know from lectures that the conics through the Ai form a pencil {δ0 Q0 + δ1 Q1 } and that
the intersection of any two of these conics is exactly the Ai . With this in mind:
(a) If X = [v] is distinct from the Ai , at least one of Q0 (v), Q1 (v) 6= 0 (else X is in the
intersection of the corresponding C0 and C1 and so is one of the Ai ). Now X lies in
the conic with quadratic form Q1 (v)Q0 − Q0 (v)Q1 and this conic is unique since another
would intersect this one at X, forcing X to be one of the Ai .
(b) If the conic is non-singular then no three points on it can be collinear since no line
intersects such a conic in more than two points. Since A1 , . . . , A4 , X all lie on the conic,
no three are collinear.
Conversely, from lectures, we know that the singular conics in the pencil are the line pairs
through the Ai and if X lies on such a line-pair, then X is collinear with two of the Ai .
5. These four points are in general position so the conics passing through them form a pencil and
the line pairs through these points are the singular ones. So we compute the lines through
these points: the line through [1, 0, 0] and [1, 0, 1] is λ1 = 0 while that through [1, 1, 0] and
[1, 1, 1] is λ0 − λ1 = 0 so that the line-pair is λ1 (λ0 − λ1 ) = 0. Again, the line through [1, 0, 0]
and [1, 1, 0] is λ2 = 0 and that through [1, 0, 1] and [1, 1, 1] λ0 − λ2 so that the line-pair is
 given by λ2 (λ0 − λ2 ) = 0. We now know enough to describe the whole pencil: they are the
quadrics of the form
δ0 λ1 (λ0 − λ1 ) + δ1 λ2 (λ0 − λ2 ) = 0
and we readily check that the last singular quadratic is given at [δ0 , δ1 ] = [1, −1] by (λ1 −
λ2 )(λ0 − λ1 − λ2 ) = 0.
P
6. Write such a conic C as Q = i,j βij λi λj = 0. Then [1, 0, 0] ∈ C means β00 = 0 while
[1, 1, 0] ∈ C forces β11 + 2β01 = 0. Thus our quadratic form has matrix
 
0 β01 β02
P = β01 −2β01 β12 
β02 β12 β22

so that the tangent line at [1, 0, 0] is given by (1, 0, 0)P (λ0 , λ1 , λ2 )T = 0, that is, β01 λ1 +
β02 λ2 = 0. Since this must be the line λ1 + λ2 = 0, we get β01 = β02 . Similarly, the tangent
at [1, 1, 0] is given by β01 λ0 − β01 λ1 + (β01 + β12 )λ2 = 0. Since this is supposed to be the line
given by λ0 − λ1 + λ2 = 0 we conclude that β12 = 0. It follows that our conic is of the form

β01 (−2λ21 + 2λ0 λ1 + 2λ0 λ2 ) + β22 λ22 = 0,

for β01 , β22 not both zero. Conversely, any conic of this form is tangent at [1, 0, 0] and [1, 1, 0]
to the desired lines, as we can easily check. Thus the family of conics under consideration is
a pencil.
As to the singular quadrics in the pencil: at β01 = 0 we get the double line λ22 = 0. To find
the other, we contemplate the determinant for an arbitrary quadratic form in the family:

0 β01 β01 0 1 1
2 2
β01 −2β01 0 = β01 1 −2 0 = β01 (−2β01 + β22 ).
β01 0 β22 β01 0 β22

Thus there is one more singular conic when β22 = 2β01 : this has quadratic form 2λ22 − 2λ21 +
2λ0 λ1 + 2λ0 λ2 = (λ1 + λ2 )(λ0 − λ1 + λ2 ) which we recognise as the line-pair consisting of our
two tangent lines.
Finally the base locus is the intersection of our double line and line-pair and so consists of two
points which must be the original [1, 0, 0] and [1, 1, 0] that we started with.
7. First suppose that dim P(V ) = 1: in this case, a quadric either consists of two points [v] and
[w], say, or is a singular quadric [u]. In the first case, v, w are a basis of V and any symmetric
bilinear form giving rise to this quadric has B(v, v) = B(w, w) = 0 and B(v, w) 6= 0. These
form a 1-dimensional subspace in the space of such forms and we are done.
If the quadric is singular, extend u to a basis u0 . Then B(u, u) = B(u, u0 ) = 0 while B(u0 , u0 ) 6=
0 so again the symmetric bilinear forms giving this quadric form a 1-dimensional subspace.
Finally, let dim V be arbitrary and suppose that Q0 and Q1 are not scalar multiples of each
other. We show that the corresponding quadrics S0 , S1 are distinct. By hypothesis, there are
v, w ∈ V so that Q0 (v)Q1 (w) 6= Q0 (w)Q1 (v). Now intersect the Si with the line P(U ) through
[v] and [w]: we have that the Qi|U are not scalar multiples of each other so that, by the first
part, S1 ∩ L 6= S2 ∩ L, whence S1 6= S2 .
8. (a) dim S 2 V ∗ = (1 + 1)(1 + 2)/2 = 3 so that P(S 2 V ∗ ) is a projective plane.
(b) Use a basis v0 , v1 of V to identify a symmetric bilinear form with the symmetric matrix
2
βij = B(vi , vj ). Then the degenerate B are the zeros of det B = β00 β11 − β01 which is a
2 ∗
non-degenerate quadratic form on S V . Thus the degenerate quadrics constitute a conic
in the space of all quadrics on a projective line!
(c) The singular conics are precisely the double-points so that the map that sends a point [v]
to the singular conic {[v]} is a bijection.
M231: Exercise sheet 9

Simultaneous diagonalisation
1. Show that the quadratic forms below cannot be simultaneously diagonalised:
λ20 + λ21 − λ22 , λ20 + λ21 − λ1 λ2 .

2. Simultaneously diagonalise the quadratic forms

Q0 = λ21 + λ22 − λ0 λ1 − λ0 λ2
Q1 = 3λ21 + λ22 − 3λ0 λ1 − λ0 λ2 .

3. Find the pencil of conics through the points [1, 0, 0], [1, 1, 0] and tangent to the line
λ0 +2λ2 = 0 at [2, 1, −1]. Do these conics have simultaneously diagonalizable quadratic
forms?
4. Let Q0 , Q1 be quadratic forms defining nonsingular conics C0 and C1 . Suppose that
|C0 ∩ C1 | = 2 and that there are exactly 2 singular conics in the pencil containing C0
and C1 .
(a) Show that the singular conics must comprise one double line and one line-pair.
(b) Deduce that Q0 and Q1 are simultaneously diagonalisable.
(c) Show that the line-pair consists of the tangents to C0 (or C1 ) at the two points
of C0 ∩ C1 .
5. Consider the following quadratic forms in P(R3 ).
Q0 = λ21 + λ22 − λ0 λ1 − λ0 λ2
Q1 = λ21 − λ22 − λ0 λ1 − λ0 λ2
Q2 = λ21 + 2λ22 − λ0 λ1 − 2λ0 λ2
For each pair of these, find the singular conics in the pencil they form, and hence either
simultaneously diagonalize the quadratic forms, or show that it cannot be done.
Exterior algebra
6. Suppose v0 , v1 , v2 , v3 are linearly independent vectors in a vector space over R and let
θ ∈ R. Expand and simplify the following wedge products:
(a) (cos θ v0 + sin θ v1 ) ∧ (sin θ v0 − cos θ v1 )
(b) (v0 + v1 ) ∧ (v1 + v2 )
(c) (v0 + v1 ) ∧ (v2 + v3 )
(d) (v0 + v1 ) ∧ (v0 ∧ v1 − v1 ∧ v2 + v2 ∧ v0 )
(e) (v0 ∧ v1 + v2 ∧ v3 ) ∧ (v0 ∧ v2 − v1 ∧ v3 )
(f ) (v0 ∧ v1 + v2 ∧ v3 ) ∧ (v0 ∧ v1 − v2 ∧ v3 ).
7. Let a ∈ ∧k V and b ∈ ∧` V . Show that
a ∧ b = (−1)k` b ∧ a.
Deduce that if k is odd, a ∧ a = 0.

Home page: http://go.bath.ac.uk/ma30231

8. (Cartan’s Lemma) Let v0 , . . . , vk , w0 , . . . , wk be vectors in V with v0 , . . . , vk linearly
independent. Suppose that X
vi ∧ wi = 0.
i

Show that there is a symmetric (k + 1) × (k + 1) matrix (Aij ) such that

k
X
wi = Aij vj .
j=0

May 1, 2013
M231: Exercise sheet 9—Solutions

1. We examine the number of points in the intersection of the corresponding conics. If X =

[λ0 , λ1 , λ2 ] is such a point, we have λ20 + λ21 − λ22 = λ20 + λ21 − λ1 λ2 = 0 so that λ22 = λ1 λ2 .
This √gives two possibilities: first, if λ2 = 0, we have λ20 + λ21 = 0 giving two solutions X =
[1, ± −1, 0]. On the other hand, if λ1 = λ2 then λ20 = 0 and X = [0, λ1 , λ1 ] = [0, 1, 1].
Thus there are three points of intersection and, three being odd, we cannot simultaneously
diagonalise the two quadratic forms.
2. Step one is to seek some singular conics in the pencil. In the present case, we can do this by
eye: Q1 − Q0 = 2λ1 (λ1 − λ0 ) while Q1 − 3Q0 = 2λ2 (λ0 − λ2 ). Call the corresponding conics
(which are both line-pairs) C2 and C3 . The base-locus of the pencil is then C2 ∩ C3 which is
easy to compute being the intersections of these line-pairs: we get A1 = [1, 0, 0] as λ1 ∩ λ2 ,
A2 = [1, 0, 1] as λ1 ∩ (λ0 − λ2 ), A3 = [1, 1, 0] as λ2 ∩ (λ1 − λ0 ) and finally A4 = [1, 1, 1] as
(λ1 − λ0 ) ∩ (λ2 − λ0 ). Thus the base-locus is 4 points and we know we can simultaneously
diagonalise the quadratic forms in this pencil. To do so, find the last singular quadric C4 which
is the line-pair consisting of A1 A4 = λ1 − λ2 and A2 A3 = λ0 − λ1 − λ2 , that is, the quadratic
form is Q4 = (λ1 − λ2 )(λ0 − λ1 − λ2 ). Now find the vertices of the line-pairs: the vertex of C2
is X0 = [0, 0, 1], that of C2 is X1 = [0, 1, 0] and that of C4 is X2 = [2, 1, 1]. The corresponding
representative vectors are a diagonalising basis for the pencil while the lines joining them give
the dual basis which we compute: X0 X1 = λ0 ; X0 X2 = λ0 − 2λ1 and X1 X2 = λ0 − 2λ2 . Thus
we can find a0 , a1 , a2 and b0 , b1 , b2 such that

Q0 = a0 λ20 + a1 (λ0 − 2λ1 )2 + a2 (λ0 − 2λ2 )2

Q1 = b0 λ20 + b1 (λ0 − 2λ1 )2 + b2 (λ0 − 2λ2 )2 .

We compare coefficients of the λ2i to get a0 = −1/2, a1 = a2 = 1/4 while b0 = −1, b1 = 3/4,
b2 = 1/4.
P
3. Write an arbitrary conic C as i,j βij λi λj . Then [1, 0, 0] ∈ C tells us that β00 = 0; [1, 1, 0] ∈ C
then gives β11 + 2β01 = 0 and [2, 1, −1] ∈ C then yields 2β01 − 4β02 − 2β12 + β22 = 0. The
tangent line to C is given then by

(β01 − β02 )λ0 − β12 λ1 + (2β01 − 2β02 − β12 )λ2 = 0.

This coincides with the line given by λ0 + 2λ2 = 0 exactly when β12 = 0. We now have
β11 = −2β01 , β22 = 4β02 − 2β01 so that the pencil is

β01 (−2λ21 − 2λ22 + 2λ0 λ1 ) + β02 (4λ22 + 2λ0 λ2 ).

We note that taking β01 = 0 gives us a singular conic with quadratic form 2λ2 (λ0 + 2λ2 ), thus
a line pair. For the rest, we compute the determinant corresponding to an arbitrary conic in
the pencil to get 2β01 (β01 − β02 )2 so that the only other singular conic is given by β01 = β02
and this has quadratic form 2(−λ21 + λ22 + λ0 λ1 + λ0 λ2 ) = 2(λ1 + λ2 )(λ0 − λ1 + λ2 ). Finally,
we compute the base-locus of the pencil and discover that three of the lines which comprise
line-pairs are concurrent at [2, 1, −1] so that the singular conics intersect in only 3 points.
Consequently, we cannot simultaneously diagonalise the quadratic forms.
4. (a) A singular conic is either a double line or a line-pair. Two double lines intersect in a
single point so at least one of our singular conics must be a line-pair. If the other singular
conic was a line-pair also, it would intersect the first in either 4 points or 3 (one line of
one pair goes through the vertex of the other) or 1 (all four lines are concurrent) or the
two line-pairs have a line in common. In all these cases, there are not 2 points in the
base-locus so we must conclude that the second singular conic is a double line.
(b) There are f0 , f1 , f2 ∈ V ∗ such that the line-pair is given by f0 f1 = 0 and the double line
by f22 = 0. Moreover, the fi are linearly independent (otherwise the corresponding lines
 would be concurrent). Finally, set g0 = 12 (f0 + f1 ), g1 = 12 (f0 − f1 ) and g2 = f2 . Then
the line-pair is given by g02 − g12 = 0 while the double line is given by g22 = 0. Otherwise
said, the gi diagonalise the quadratic forms for the singular conics in the pencil and so
all conics in the pencil.
(c) To see that the line-pairs consist of tangents, one can proceed in two ways. Either use the
fact that all the conics are simultaneously diagonalisable and compute in a diagonalising
basis. However, a slicker argument1 is to note that the line pair intersects C0 in exactly
two points (the base-locus), one on each of the lines in the pair (if the base-locus lived
on just one of the lines, that line would have to coincide with the double line too!). This
means precisely that the lines are tangent to C0 at the points of the base-locus.
5. (a) Consider the pair Q0 , Q1 : we notice that Q0 − Q1 = 2λ22 gives a double line and that
Q1 = (λ1 + λ2 )(−λ0 + λ1 − λ2 ) is itself degenerate, giving a line-pair. The determinant of
δ0 Q1 + δ1 λ22 is δ02 δ1 /4 so we have found all the singular conics in the pencil. Since these
are a double line and a line-pair with two points of intersection at [1, 0, 0] and [1, 1, 0], we
know we can simultaneously diagonalise the pencil. Indeed, taking sums and differences
of λ1 + λ2 and −λ0 + λ1 − λ2 yields a diagonalising dual basis, 2λ1 − λ0 , 2λ2 + λ0 , λ2 .
Solving for the coefficients of the λ2i gives:
1 1
Q0 = 2λ22 + 4 (2λ1 − λ0 )2 − 4 (2λ2 − λ0 )2
1 1
Q1 = 4 (2λ1 − λ0 )2 − 4 (2λ2 − λ0 )2

and we are done.

(b) Now let us look at Q0 and Q2 . We note that Q2 − Q0 = λ2 (λ2 − λ0 ) and so gives a
line-pair with vertex [0, 1, 0]. Further 2Q0 − Q2 = λ1 (λ1 − λ0 ) gives another line-pair with
vertex [0, 0, 1]. Finally, a determinant computation, or finding the third line-pair through
the 4 points in the intersection of the first two line-pairs or plain alertness2 reveals that
the difference of these is the line-pair (λ2 − λ1 )(−λ0 + λ1 + λ2 ). We therefore have three
line-pairs as the singular conics in the pencil and so can simultaneously diagonalise. The
vertices of the line-pairs are [0, 1, 0], [0, 0, 1] and [2, 1, 1] and the lines through pairs of
these give the diagonalising dual basis λ0 , λ0 −2λ1 , λ0 −2λ2 . We now solve for coefficients
of λ2i to get:
Q0 = − 21 λ20 + 41 (λ0 − 2λ1 )2 + 14 (λ0 − 2λ2 )2
Q2 = − 34 λ20 + 41 (λ0 − 2λ1 )2 + 12 (λ0 − 2λ2 )2
(c) Finally, contemplate Q1 and Q2 . We already know that Q1 is degenerate and we quickly
notice that Q2 − Q1 = λ2 (3λ2 − λ0 ) is too. Less easy to spot is that 6Q1 − 5Q2 =
(λ1 − 4λ2 )(−λ0 + λ1 + 4λ2 ) is the third line-pair (I asked a machine to compute the
determinant of the pencil for this). Now we have 3 singular conics in the pencil, we know
we can diagonalise: the vertices of the line-pairs are [0, 1, 0], [2, 1, −1] and [8, 4, 1] and the
lines through pairs of these give the diagonalising dual basis λ0 − 2λ1 , λ0 + 2λ2 , λ0 − 8λ2 .
We check coefficients to get:
1 1
Q1 = 4 (λ0 − 2λ1 )2 − 4 (λ0 + 2λ2 )
2
1 3 1
Q2 = 4 (λ0 − 2λ1 )2 − 10 (λ0 + 2λ2 )
2
+ 20 (λ0 − 8λ2 )2

6. This is simply a matter of multiplying out and using the basic identities that v ∧ v = 0 so that
v ∧ w = −w ∧ v, for any vectors v, w. Thus:
(a) (cos θ v0 + sin θ v1 ) ∧ (sin θ v0 − cos θ v1 = − cos2 θ v0 ∧ v1 + sin2 θ v1 ∧ v0 = v1 ∧ v0 .
(b) This is v0 ∧ v1 + v0 ∧ v2 + v1 ∧ v2 .
(c) This is v0 ∧ v2 + v0 ∧ v3 + v1 ∧ v2 + v1 ∧ v3 .

1
Thanks, Amine!
2
I did the determinant computation!
 (d) Here v0 ∧ (v0 ∧ v1 − v1 ∧ v2 + v2 ∧ v0 ) = −v0 ∧ v1 ∧ v2 and v1 ∧ (v0 ∧ v1 − v1 ∧ v2 + v2 ∧ v0 ) =
v1 ∧ v2 ∧ v0 so that

(v0 + v1 ) ∧ (v0 ∧ v1 − v1 ∧ v2 + v2 ∧ v0 ) = −v0 ∧ v1 ∧ v2 + v1 ∧ v2 ∧ v0

= −v0 ∧ v1 ∧ v2 + v0 ∧ v1 ∧ v2 = 0.

(e) Each of the four products this expands into has a repeated appearance of some vi and so
vanishes. Thus (v0 ∧ v1 + v2 ∧ v3 ) ∧ (v0 ∧ v2 − v1 ∧ v3 ) = 0.
(f ) Multiplication of 2-vectors is commutative so that this is a difference of two squares:

(v0 ∧v1 +v2 ∧v3 )∧(v0 ∧v1 −v2 ∧v3 ) = (v0 ∧v1 )∧(v0 ∧v1 )−(v2 ∧v3 )∧(v2 ∧v3 ) = 0−0 = 0.

7. It suffices to prove this result for a, b of the form a = a1 ∧ · · · ∧ ak , b = b1 ∧ · · · ∧ b` , for

ai , bj ∈ V , since all elements of ∧k V and ∧` V are linear combinations of elements of this
type. Now it takes k swops to move a bj past all the ai and each swop changes the product
by a sign:

a1 ∧ · · · ∧ ak ∧ b1 ∧ · · · ∧ b` = (−1)k b1 ∧ a1 ∧ · · · ∧ ak ∧ b2 ∧ · · · ∧ b` .

Repeating this for each of the ` bj yields the result.

In particular, taking a = b and k odd, we have a ∧ a = −a ∧ a so that a ∧ a = 0.
Pn
8. Extend v0 , . . . , vk to a basis v0 , . . . , vn of V and then write wi = j=0 Aij vj . We plug this
into the given identity to get
X X
0= vi ∧ w i = Aij vi ∧ vj
i i≤k,0≤j≤n
X X
= Aij vi ∧ vj + Aij vi ∧ vj
0≤i,j≤k i≤k,j>k
X X
= (Aij − Aji )vi ∧ vj + Aij vi ∧ vj .
0≤i<j≤k i≤k,j>k

Now the products vi ∧ vj , for i < j ≤ k and i ≤ k, j > k are linearly independent so we
conclude that Aij = Aji for 0 ≤ i, j ≤ k and Aij = 0 when i ≤ k, j > k.
M231: Exercise sheet 10

Exterior algebra
1. Suppose v0 , v1 , v2 , v3 are linearly independent elements of V . Which of the following
elements of ∧2 V are decomposable?
(a) v0 ∧ v1 − v1 ∧ v2 + v2 ∧ v3 .
(b) v0 ∧ v1 − v1 ∧ v2 + v2 ∧ v0 .
(c) v0 ∧ v1 + v1 ∧ v2 + v2 ∧ v3 + v3 ∧ v0 .
Write the decomposable elements as a wedge product of two vectors in V .
2. Let dim V = n + 1. Show that any element of ∧n V is decomposable, that is, is of the
form v1 ∧ · · · ∧ vn , for v1 , . . . , vn ∈ V .
Hint: Look at the start of the proof of Theorem 3.4 in the on-line notes for the case
n = 2. This should provide some inspiration.
Around the Klein correspondence
3. Show that there is a bijection between the set of lines in a projective plane P(V ) and
P(∧2 V ).
How is this related to duality?
4. Let dim V = 4 and suppose that a ∈ ∧2 V has the property that a ∧ b = 0, for all
b ∈ ∧2 V . Show that a = 0.
Deduce that the Klein quadric is non-singular.
5. (a) Show that distinct α-planes in the Klein quadric meet in a single point.
(b) Show that distinct β-planes in the Klein quadric meet in a single point.
(c) Show that an α-plane and a β-plane intersect in a line or not at all.
(d) Show that a line in the Klein quadric lies in a unique α-plane and a unique
β-plane.

May 4, 2013

Home page: http://go.bath.ac.uk/ma30231

M231: Exercise sheet 10—Solutions

1. For each 2-vector a, we compute a ∧ a to see if it vanishes.

(a) Here (v0 ∧ v1 − v1 ∧ v2 + v2 ∧ v3 ) ∧ (v0 ∧ v1 − v1 ∧ v2 + v2 ∧ v3 ) = 2v0 ∧ v1 ∧ v2 ∧ v3 6= 0 so
this 2-vector is not decomposable.
(b) v0 ∧ v1 − v1 ∧ v2 + v2 ∧ v0 wedges with itself to zero (each summand wedges with every
other summand to zero) and so v0 ∧ v1 − v1 ∧ v2 + v2 ∧ v0 is decomposable. In fact, a
little thought gives

v0 ∧ v1 − v1 ∧ v2 + v2 ∧ v0 = (v0 + v2 ) ∧ (v1 − v2 ).

(c) We have

(v0 ∧ v1 + v1 ∧ v2 + v2 ∧ v3 + v3 ∧ v0 ) ∧ (v0 ∧ v1 + v1 ∧ v2 + v2 ∧ v3 + v3 ∧ v0 ) =
2v0 ∧ v1 ∧ v2 ∧ v3 + 2v1 ∧ v2 ∧ v3 ∧ v0 = 0.

So v0 ∧ v1 + v1 ∧ v2 + v2 ∧ v3 + v3 ∧ v0 is decomposable and in fact

v0 ∧ v1 + v1 ∧ v2 + v2 ∧ v3 + v3 ∧ v0 = (v0 − v2 ) ∧ (v1 − v3 ).

2. Let a ∈ ∧n V and consider the linear map V → ∧n+1 V given by v 7→ a∧v. Since dim ∧n+1 V =
1, the rank-nullity theorem says that this map has kernel of dimension at least n. Thus we
have linearly independent vectors v1 , . . . , vn with a ∧ vi = 0, 1 ≤ i ≤ n, which we extend to a
basis v0 , . . . , vn of V .
A basis of ∧n V is given by products of n of the n + 1 vj , listed in increasing order. Denote
by ωi the basis element comprising all the vj except the i-th. Thus ω0 = v1 ∧ · · · ∧ vn ,
ω1 = v0 ∧ v2 ∧ · · · ∧ vn and so on. PWe observe that ωi ∧ vj = 0, for i 6= j, while ωi ∧ vi =
±v0 ∧ · · · ∧ vn 6= 0. Now write a = j λj ωj and use a ∧ vi = λi v0 ∧ · · · ∧ vn = 0, for i ≥ 1, to
deduce that λi = 0, for i ≥ 1. Thus a = λ0 v1 ∧ · · · ∧ vn is decomposable.
3. The arguments in the lectures show that the map P(U ) 7→ ∧2 U from the set of lines in P(V ) to
P(∧2 V ) is a bijection onto the decomposable elements [u ∧ v] of P(∧2 V ). On the other hand,
we also know (cf. question 2) that, since dim V = 3, all elements of ∧2 V are decomposable.
Thus we have a bijection from the set of lines to P(∧2 V ).
Duality establishes a bijection from the set of lines in the plane P(V ) to P(V ∗ ) and we should
be able to see an identification of P(V ∗ ) and P(∧2 V ). In fact, there is a natural projective
transformation. To see this, fix a basis v0 , v1 , v2 of V and define a map T : ∧2 V → V ∗ by

a ∧ v = (T (a)v)v0 ∧ v1 ∧ v2 .

This is well-defined since dim ∧3 V = 1 so that a ∧ v is a scalar multiple of v0 ∧ v1 ∧ v2 .

Changing the basis changes T by a scalar multiple. One readily checks that T (a) is linear and
so an element of V ∗ and that T is a linear map. Since both ∧2 V and V ∗ are 3-dimensional,
T will be an isomorphism as soon as it injects. However, a ∈ ker T means that a ∧ v = 0,
for all v ∈ V . Now we know that a is decomposable, a = v0 ∧ v1 , say, and if a 6= 0,
v0 , v1 are linearly independent and so can be extended to a basis v0 , v1 , v2 of V . But then
a ∧ v2 = v0 ∧ v1 ∧ v2 6= 0. Thus ker T = {0} and T is an isomorphism inducing a projection
transformation τ : P(∧2 V ) ∼ = P(V ∗ ).
For completeness, we check that τ intertwines our identifications of the set of lines with these
two projective spaces. Thus, given a line P(U ) with u1 , u2 a basis of U and U = ker f , for
some f ∈ V ∗ , we show that τ ([u1 ∧ u2 ]) = [f ] or, equivalently U = ker τ ([u1 ∧ u2 ]). This last
means that u ∈ U if and only if u ∧ u1 ∧ u2 = 0 which is precisely the content of Lemma 3.1
in the notes.
P
4. Choose a basis v0 , . . . , v3 of V and write a = i<j aij vi ∧ vj . If a 6= 0, we may reorder the
basis to ensure that a01 6= 0. Now all basis vectors vi ∧ vj of ∧2 V except v0 ∧ v1 have either
vi or vj equal to v2 or v3 . Thus 0 = a ∧ (v2 ∧ v3 ) = a01 v0 ∧ v1 ∧ v2 ∧ v3 6= 0, a contradiction.
 The Klein quadric is defined by the bilinear form B in ∧2 V given by a∧b = B(a, b)v0 ∧v1 ∧v2 ∧v3
so we have just proved that B(a, b) = 0, for all b ∈ ∧2 V , implies that a = 0. Thus B is
nondegenerate and the Klein quadric is nonsingular.
5. (a) Distinct α-planes correspond to the lines in P(V ) through distinct points X0 , X1 , say.
Thus a point in the intersection of the α-planes corresponds to a line in P(V ) which goes
through X0 and X1 . There is a unique such line so the α-planes intersect in a single
point.
(b) Distinct β-planes correspond to the lines in P(V ) lying in distinct planes π0 , π1 in P(V ).
Thus a point in the intersection of the β-planes corresponds to a line lying in both π0
and π1 . However, π0 ∩ π1 is a single line so the β-planes intersect in a unique point.
(c) An α-plane corresponds to the lines in P(V ) through a point X while a β-plane corre-
sponds to the lines lying in a plane Y ≤ P(V ). If X does not lie on Y then there are
no lines through X and in Y so the intersection of the α-plane and β-plane is empty.
Otherwise the intersection corresponds to the lines L with X ≤ L ≤ Y and so is a line as
we saw in lectures.
(d) As we saw in lectures, a line in the Klein quadric corresponds to the set of lines in P(V )
through a point X and lying in a plane Y . Thus the line in the quadric lies in the α-plane
corresponding to X and the β-plane corresponding to Y .
For uniqueness, part (a) tells us that there is only one α-plane containing a given line
(because two such planes intersect in a point only) and part (b) gives uniqueness of the
β-plane.