MATH32062 Notes

1 Affine algebraic varieties

1.1 Definition of affine algebraic varieties

We want to define an algebraic variety as the solution set of a collection
of polynomial equations, or equivalently, a set where every element of a
set of polynomials vanishes. Then every combination of those polynomials
vanishes on the set, so we can use the ideal generated by those polynomials
to define the algebraic variety. This motivates the definition below. Using
ideals instead of concrete polynomials also turns out to be more useful for
proving most of the theorems about algebraic varieties.
In the following, K denotes an arbitrary field, unless specified otherwise.
Definition. An affine algebraic variety is a set of the form

V(I) = {(a1 , a2 , . . . , an ) K n | f (a1 , a2 , . . . , an ) = 0, f I},

where I is an ideal I / K[x1 , x2 , . . . , xn ].

Examples:
1. If I = {0}, then V(I) = K n .
2. If I = K[x1 , x2 , . . . , xn ], then V(I) = .
3. Let f1 , f2 , . . . , fm be polynomials of degree 1 in x1 , x2 , . . . xn , and let
I = hf1 , f2 , . . . fm i be the ideal generated by them. Then V(I) is the set of
solutions of the system of linear equations f1 = f2 = . . . = fm = 0. Such
a variety is called an affine subspace of K n . This, in particular, includes
set containing a single point, if P = (b1 , b2 , . . . , bn ) K n , then the ideal
I = hx1 b1 , x2 b2 , . . . xn bn i defines the variety V(I) = {P }.
4. Let I = hy 2 x3 i / R[x, y], then V(I) is the cuspidal cubic curve defined
by the equation y 2 = x3 in R2 , shown below on the left.

1
 4 4

2 2

-0.5 0.5 1.0 1.5 2.0 2.5 -1 1 2

-2 -2

-4 -4

5. Let I = hy 2 x3 x2 i / R[x, y], then V(I) is the nodal cubic curve defined
by the equation y 2 = x3 + x2 in R2 , shown above on the right.
6. Let I = hx2 + y 2 1i / R[x, y], then V(I) is the circle of radius 1 centred
at the origin defined by the equation x2 + y 2 = 1 in R2 .
7. If I = hx, yi / R[x, y], then V(I) = {(0, 0)} R2 .
8. If I = hx2 , y 2 i / R[x, y], then V(I) = {(0, 0)} R2 .
9. If I = hx2 + y 2 + z 2 1, x2 x + y 2 i / R[x, y, z], then V(I) R3 is a figure
of eight curve, shown below.

2
1
0
-1
-2
2

-1

-2
-2
-1
0
1
2

2
10. The special linear group SL(n, R) consisting of all n n real matrices
with determinant 1 is also an affine algebraic variety. The set of all n n
2
real matrices is Rn , the determinant is a polynomial in the entries of the
matrix, therefore SL(n, R) = V(hdet(A) 1i).
Note. Examples 7 and 8 show that different ideals can define the same variety.
The exact nature of the correspondence between ideals and varieties will be
investigated later in Section 1.3.
Proposition 1.1 Let K be a field. Let f1 , f2 , . . . , fm K[x1 , x2 , . . . , xn ]
and let I = hf1 , f2 , . . . , fm i. Then
V(I) = {(a1 , a2 , . . . , an ) K n | fi (a1 , a2 , . . . , an ) = 0, i, 1 i m}.

Proof. fi I for each i, 1 i m, so if (a1 , a2 , . . . , an ) V(I), then

fi (a1 , a2 , . . . , an ) = 0 for every i, 1 i m.
Conversely, assume that fi (a1 , a2 , . . . , an ) = 0 for every i, 1 i m. Any
Pm
f I can be written as f = fi gi for suitable gi K[x1 , x2 , . . . , xn ],
i=1
1 i m, therefore
m
X
f (a1 , a2 , . . . , an ) = fi (a1 , a2 , . . . , an )gi (a1 , a2 , . . . , an ) = 0,
i=1

so (a1 , a2 , . . . , an ) V(I).
Definition. A ring R is Noetherian if and only if all of its ideals can be
generated by finitely many elements. (Equivalently, R is Noetherian if and
only if for any increasing sequence of ideals I1 I2 I3 . . . In In+1 . . .
there exists an N such that In = IN for all n N .)
Theorem 1.2 (Hilbert Basis Theorem) (Not examinable) If K is a field,
then the polynomial ring K[x1 , x2 , . . . , xn ] is Noetherian for any n 0.
Note: The Hilbert Basis Theorem justifies why we only considered ideals
generated by a finite set of polynomials in Proposition 1.1. The combination
of Proposition 1.1 and Theorem 1.2 shows that any affine algebraic variety
is the set of solutions of finitely many polynomial equations.
The proposition below describes some methods which can be used to con-
struct further affine algebraic varieties from known ones.
Proposition 1.3 (i) Let V1 = V(I1 ), V2 = V(I2 ), . . . , Vk = V(Ik ) be affine
algebraic varieties in K n . Then
V1 V2 . . . Vk = V(I1 I2 . . . Ik ) = V(I1 I2 . . . Ik )

3
is also an affine algebraic variety.
(ii) Let V = V(I ), A be affine algebraic varieties in K n . Then
\ X 
V = V I
A A

is also an affine algebraic variety.

(iii) Let V1 = V(I1 ) K m , V2 = V(I2 ) K n be affine algebraic varieties.
Then V1 V2 K m K n = K m+n is also an affine algebraic variety.
Proof. (i) Lets assume first that k = 2. Let P V1 V2 and let f I1 I2 .
If P V1 , then f (P ) = 0 because f I1 , if P V2 , then f (P ) = 0 because
f I2 . Therefore in either case we have f (P ) = 0, so P V(I1 I2 ),
V1 V2 V(I1 I2 ). We also have V(I1 I2 ) V(I1 I2 ), therefore
V1 V2 V(I1 I2 ) V(I1 I2 ) (1)
(I1 I2 I1 I2 , since all the elements of the form i1 i2 , i1 I1 , i2 I2
are contained in both I1 and I2 by the definition of an ideal, and if all the
generators of I1 I2 are elements of I1 I2 , then necessarily I1 I2 I1 I2 . This
means that the elements of V(I1 I2 ) have to satisfy all the polynomials in
I1 I2 , therefore V(I1 I2 ) V(I1 I2 ).)
Let now P K n \ (V1 V2 ). Then P / V1 , so there exists f1 I1 such that
f1 (P ) 6= 0. Similarly, P
/ V2 , so there exists f2 I2 such that f2 (P ) 6= 0.
Then f1 f2 I1 I2 and (f1 f2 )(P ) = f1 (P )f2 (P ) 6= 0, therefore P / V(I1 I2 ).
This implies V(I1 I2 ) V1 V2 . By combining this with (1), we obtain
V1 V2 = V(I1 I2 ) = V(I1 I2 ).
For k > 2, use induction on k.
T
(ii) Assume that P V . Then P V = V(I ) for every A, so
A P
f (P ) = 0 for every A and f I . Any f I can be written
P A
as f = f with f I for every A and f = 0 for all but finitely
A  
P P P
many . Hence f (P ) = f (P ) = 0 = 0, so P V I . This
A A A
implies X 
\
V V I . (2)
A A

Assume now that P K n \

T
V . Then there exists 0 A such that
A P
P
/ V0 , therefore there exists f I0 such that f (P ) 6= 0. Now f I ,
A

4
 P
therefore P
/ V( I ). This implies that
A
X  \
V I V .
A A

By combining this with (2), we obtain

X  \
V I = V .
A A

(iii) Let x1 , x2 , . . . , xm be co-ordinates on K m , and y1 , y2 , . . . , yn co-

ordinates on K n . Let J1 = hI1 i / K[x1 , x2 , . . . , xm , y1 , y2 , . . . , yn ]. (I1 is an
ideal in K[x1 , . . . , xm ], J1 is the ideal generated by the elements of I1 in the
bigger ring K[x1 , x2 , . . . , xm , y1 , y2 , . . . , yn ].) We claim that V(J1 ) = V1 K n .
Let P = (a1 , a2 , . . . , am , b1 , b2 , . . . , bn ) V(J1 ). f (P ) = 0 for every f I1 ,
since I1 J1 . As f only involves the variables x1 , x2 , . . . , xm ,

0 = f (P ) = f (a1 , a2 , . . . , am , b1 , b2 , . . . , bn ) = f (a1 , a2 , . . . , am ).

Hence (a1 , a2 , . . . , am ) V(I1 ) = V1 , so P V1 K n , therefore V(J1 )

V1 K n .
Let now P = (a1 , a2 , . . . , am , b1 , b2 , . . . , bn ) V1 K n . Let f J1 . f can be
r
P
written as f = fi gi , where fi I1 and gi K[x1 , x2 , . . . , xm , y1 , y2 , . . . , yn ].
i=1
Now fi (P ) = fi (a1 , a2 , . . . , am , b1 , b2 , . . . , bn ) = fi (a1 , a2 , . . . , am ) = 0. (The
second equality holds because fi only involves on the variables x1 , x2 , . . . ,
xm , and the last equality holds because P V1 K n .) Therefore f (P ) =
r
fi (P )gi (P ) = 0, so P V(J1 ). This means V1 K n V(J1 ), by combining
P
i=1
this with V(J1 ) V1 K n proved above, we obtain V(J1 ) = V1 K n .
Similarly, if we define J2 = hI2 i/K[x1 , x2 , . . . , xm , y1 , y2 , . . . , yn ], then V(J2 ) =
K m V2 .
Hence

V1 V2 = (V1 K n ) (K m V2 ) = V(J1 ) V(J2 ) = V(J1 + J2 )

by (ii).
In practical calculations, the ideals defining the varieties are given by sets of
generators.

5
To find a generating set for an ideal defining the union of the varieties, take
all the possible products of the generators, one factor from each ideal.
To find a generating set for an ideal defining the intersection of the varieties,
take the union of the generating sets of the ideals.
To find a generating set for an ideal defining the Cartesian of the varieties,
take the union of the generating sets of the ideals, but remember that this
ideal will be in a different ring.
Examples:
1. Find an ideal I / R[x, y] such that V(I) = {(1, 1), (2, 3)}.
Let I1 = hx 1, y 1i and I2 = hx 2, y + 3i. Then V(I1 ) = {(1, 1)} and
V(I2 ) = {(2, 3)}. By Proposition 2.3 (i),

I = I1 I2 = h(x 1)(x 2), (x 1)(y + 3), (y 1)(x 2), (y 1)(y + 3)i

is a suitable ideal.
2. Let V1 = V(hx21 + x22 1i) R2 and let V2 = V(hy1 y2 i) R2 . Then
V1 V2 = Vhx21 + x22 1, y1 y2 i R4 is a cylinder.
Definition. Let K be a field and let V K n be an affine algebraic variety.
W is a subvariety of V if and only if W V and W is also an affine algebraic
variety.
Remark. By Proposition 1.3 (i) and (ii), affine algebraic varieties behave
like closed sets in a metric space. This motivates the introduction of the
Zariski topology on an affine algebraic variety, in which the closed sets are the
subvarieties and the open sets are the complements of subvarieties. There
are, however, certain differences from metric spaces, for example, any two
non-empty Zariski open sets in Rn have a non-empty intersection.

6
1.2 Affine spaces
Affine spaces are the simplest algebraic varieties, they are the solutions sets
of system of linear equations, i.e., they are defined by ideals generated by
degree 1 polynomials.
Definition. Let K be a field, and let V be a vector space over K. U V is
an affine subspace of V iff U = or U = u0 + W = {u0 + w | w W }, where
u0 U and W is a linear subspace of V .
Examples: K n as a subspace of itself, , any set containing a single point are
all affine subspaces of K n . The line y = 2x + 5 is an affine subspace of R2 ,
in this case we can take W = span{(1, 2)} and u0 = (0, 5).
Proposition 1.4 Let U = u0 +W be a non-empty affine subspace in a vector
space V . Then U = u + W for any u U , in other words, u0 can be chosen
to be an arbitrary element of U , but W is uniquely determined by U .
Proof. Let u U , then u = u0 + (u u0 ), so u u0 W . We have

u + W = {u + w | w W } = {u0 + (u u0 ) + w | w W }.

(u u0 ) + w W as W is a vector space, therefore u + w u0 + W for every

w W , hence u + W u0 + W . Similarly,

u0 + W = {u0 + w | w W } = {u + (u0 u) + w | w W } u + W,

therefore u + W = u0 + W = U .
W can be obtained as U u = {v u | v U } for any u U .
Definition. The dimension of an affine space U = u0 + W is defined to be
dim W . (Sometimes it is convenient to define dim = 1.)
Definition. Let K be a field. K n , considered as an affine subspace of itself
is called an n-dimensional affine space over K, and is denoted by An (K) or
An if the field is understood.
Definition. Let V1 , V2 be vector spaces. An affine map from V1 to V2 is
defined to be a function : V1 V2 which can be written in the form
(x) = T (x) + b, where T : V1 V2 is a linear transformation and b V2 .
It is easily checked that the composition of affine maps is also an affine
map. The invertible affine maps V V form a group. Within this group
the translations form a normal subgroup and the quotient by this normal
subgroup is the group of invertible linear transformations on V .

7
Proposition 1.5 Let V1 , V2 be vector spaces and let : V1 V2 be an
affine map.
(i) For any affine subspace U1 V1 , the image (U1 ) is also an affine subspace
of V2 .
(ii) For any affine subspace U2 V2 , the preimage 1 (U2 ) is also an affine
subspace of V1 .
Proof. Let x1 , x2 , . . . , xm be the co-ordinates on V1 , y1 , y2 , . . . , yn the
co-ordinates on V2 . Let (x) = T (x) + b, where T : V1 V2 is a linear
transformation and b = (b1 , b2 , . . . , bn )T V2 is a vector. Let {tij }1in,1jm
be the matrix of T with respect to the co-ordinates on V1 and V2 .
(i) Let U1 V1 be an affine subspace. If U1 = , (U1 ) = , too. Otherwise,
we can write U1 = u1 + W1 , where u1 U1 and W1 is a linear subspace of
V1 . Then

(U1 ) = {(u1 + w) | w W1 } = {T (u1 + w) + b | w W1 }

= {(T (u1 ) + b) + T (w) | w W } = (T (u1 ) + b) + T (W )

As T (W ) is a linear subspace of W2 , it follows that (U1 ) is an affine sub-

space, as claimed.
(ii) Let now U2 be an affine subspace of V2 .
There exist degree 1 polynomials f1 , f2 , . . . , fk in y1 , y2 , . . . , yn such that
U2 = {(y1 , y2 , . . . , yn ) V2 | fi (y1 , y2 , . . . , yn ) = 0, i, 1 i k}.
Then
 m
X m
X m
X T
T
((x1 , x2 , . . . , xm ) ) = b1 + t1j xj , b2 + t2j xj , . . . , bn + tnj xj ,
j=1 j=1 j=1

therefore ((x1 , x2 , . . . , xm )T ) U2 if and only if

 m
X m
X m
X 
f i b1 + t1j xj , b2 + t2j xj , . . . , bn + tnj xj =0
j=1 j=1 j=1

for every i, 1 i k. These are linear equations in the xi , 1 i m,

therefore 1 (U2 ) is an affine subspace of V1 , as claimed.

Definition. Let X, Y be subsets of a vector space V . X and Y are affine

equivalent if and only if there exist mutually inverse affine maps , : V
V such that (X) = Y and (Y ) = X. It is easy to verify that affine
equivalence is an equivalence relation.

8
Examples:
1. All affine subspaces of the same dimension are affine equivalent. This
can be proved by showing that any k-dimensional affine subspace is affine
equivalent to the linear subspace spanned by the first k standard unit vectors.
2. Any two triangles in R2 are affine equivalent. An easy way to prove
it is to show that any triangle is affine equivalent to the triangle with ver-
tices (0, 0), (1, 0) and (0, 1). See http://www.maths.manchester.ac.uk/
~gm/teaching/MATH32062/triangles.pdf for an example of calculating an
explicit affine equivalence between two triangles.
3. Similarly, any two parallelograms in R2 are because all parallelograms
affine equivalent to the unit square.
4. Plane conics
A conic (short for a conic section) is the set of points satisfying an equation
of the form ax2 + bxy + cy 2 + dx + ey + f = 0 with at least one of a, b, c
not 0.
By a sequence of change of co-ordinates corresponding to affine transforma-
tions and by multiplying the equation by a non-zero scalar, every such equa-
tion can be reduced to one of finitely many standard forms, which shows that
there are finitely many affine equivalence classed of conics. The tranforma-
tion of the points is given inverse of the change of co-ordinates, for example,
if the new co-ordinates are x0 = 2x and y 0 = y + 5, then its effect on the
points is the map (x, y) 7 (x/2, y 5).
The procedure consists of four steps. The co-ordinates after the ith step will
be denote by xi and yi , the coefficients by ai , bi , . . . , fi .
Step 1. If a 6= 0, let x1 = x, y1 = y. If a = 0 but c 6= 0, let x1 = y, y1 = x.
If a = c = 0, then let x1 = (x + y)/2, y1 = (x y)/2. In this case b 6= 0 and
bxy = b(x21 y12 ). We have achieved that a1 , the coefficient of x21 in the new
equation is not 0. If necessary, we can multiply the equation by 1 to make
a1 positive.
Step 2. Complete the square with respect to x1 , let
 
b1 d1
x 2 = a1 x 1 + y1 +
2a1 2a1
and y2 = y1 . After this step, the equation will be of the form x22 + c2 y22 +
e2 y2 + f2 = 0.
Step p3. If c2 6= 0, complete the square with respect to y2 , let x3 = x2 and
y3 = |c2 |(y2 +e2 /(2c2 )), then the equation will have the form x23 y32 +f3 = 0.
If c2 = 0 but e2 6= 0, let x3 = x2 and y3 = e2 y2 + f2 , so the equation becomes

9
x23 + y3 = 0. If c2 = e2 = 0, then let x3 = x2 and y3 = y2 , so the equation
will have the form x23 + f3 = 0.
p p
Step 4. If f3 6= 0, then let x4 = x3 / |f3 |, y4 = y3 / |f3 | and after the
change of variable divide the whole equation by |f3 |. After this, the form of
the equation will be the same, but f3 will be 1. If f3 = 0, then let x4 = x3 ,
y4 = y3 .
At the end of this step, either the equation is x24 + y4 = 0, or it contains x24 ,
possibly y42 and possibly 1 and no other term. x24 y42 1 = 0 can be
tranformed into x24 y42 + 1 = 0 by swapping x4 and y4 , so these two are
affine equivalent.
Therefore any variety in R2 defined by a degree 2 equation is affine equivalent
to one of the following:
Case 1. x2 + y 2 1 = 0 (ellipses)
Case 2. x2 + y 2 = 0 (single point)
Case 3. x2 + y 2 + 1 = 0 ()
Case 4. x2 y 2 1 = 0, (hyperbolas)
Case 5. x2 y 2 = 0 (two intersecting lines)
Case 6. x2 + y = 0 (parabolas)
Case 7. x2 1 = 0 (two parallel lines)
Case 8. x2 = 0 (double line)
Case 9. x2 + 1 = 0 ()
It is also true, but requires a separate proof, that these cases, apart from the
two empty sets are not affine equivalent. See http://www.maths.manchester.
ac.uk/~gm/teaching/MATH32062/conic.pdf for examples of transforming
the equation of a conic to one of the above forms.
Over C we can use essentially the same method to reduce the equation of a
conic to one of the above forms. The difference is that we do not need to
multiply the equation to ensure a1 > 0 in Step 1 and there is no need to take
the modulus of c2 and f3 in Steps 2 and 3, resp. If a term occurs in the final
equation, it will always have coefficient +1, so we will end up with one of
Cases 3, 6, 8 or 9.
Over C, the change of co-ordinates x0 = x, y 0 = iy converts Case 1 into Case
4 and vice versa, so all ellipses and hyperbolas are affine equivalent over C.
By a similar argument, Cases 1, 3 and 4 are affine equivalent over C, similarly
Cases 2 and 5, and Cases 7 and 9.

10
1.3 The correspondence between varieties and ideals
Definition. Let X An . The ideal of X is defined as
I(X) = {f K[x1 , x2 , . . . , xn ] | f (a1 , a2 , . . . , an ) = 0, (a1 , a2 , . . . , an ) X}.
I(X) is an indeed ideal of K[x1 , x2 , . . . , xn ].
It is clear from the definition that X1 X2 implies I(X1 ) I(X2 ).
Proposition 1.6
(i) If J / K[x1 , x2 , . . . , xn ], then J I(V(J)).
(ii) X V(I(X)) for any X An (K). Equality holds if and only if X is an
affine algebraic variety.
Proof. (i) If f J and P V(J) then f (P ) = 0 by the definition of V,
therefore f I(V(J)) by the definition of I, so J I(V(J)) as claimed.

(ii) It is also clear from the definitions that X V(I(X)) for any set X An
and that if equality holds then X is an affine algebraic variety.
The only thing that remains to be proved that if X is an algebraic va-
riety, then X = V(I(X)). Assume that X = V(J) for some ideal J /
K[x1 , x2 , . . . , xn ]. By (i), J I(V(J)), therefore X = V(J) V(I(V(J))) =
V(I(X)). We have already seen that X V(I(X)), therefore X = V(I(X)),
as required.
Definition. A field K is algebraically closed if and only if every polynomial
of degree at least 1 in K[x] has a root in K.
Examples: C and Q (the field of algebraic numbers in C) are algebraically
closed. Q, R and the finite fields are not algebraically closed.
Remark. For any field K, there exists a minimal algebraically closed field
containing it, which is unique up to isomorphism, this field is called the
algebraic closure of K. For example, the algebraic closure of R is C and the
algebraic closure of Q is Q
Definition. Let Rbe a (commutative) ring. Let I /R be an ideal. The radical
of I, denoted by I or rad I, is

I = {x R | n such that xn I}.

I is called a radical ideal if and only if I = I.
p
Algebraic
facts: For any ideal I, I is also an ideal, I I and I = I,
i. e., I is a radical ideal.

11
Theorem 1.7 (Hilberts Nullstellensatz) Let K be an algebraically closed
field.
(i) Every maximal ideal of K[x1 , x2 , . . . , xn ] is of the form hx1 a1 , x2
a2 , . . . , xn an i, where ai K, 1 i n.
(ii) If J is an ideal of K[x1 , x2 , . . . , xn ], J 6= K[x1 , x2 , . . . , xn ] , then V(J) 6= .
(iii) I(V(J)) = J for any J / K[x1 , x2 , . . . , xn ].
Proof. (i) Not proved here.
Note that hx1 a1 , x2 a2 , . . . , xn an i is a maximal ideal of K[x1 , x2 , . . . , xn ]
for any field K. We can define a ring homomorphism K[x1 , x2 , . . . , xn ] K,
f 7 f (aa , a2 , . . . , an ), whose kernel is exactly hx1 a1 , x2 a2 , . . . , xn an i.
Since the image is a field, the kernel is a maximal ideal. The substance of
this part is that over an algebraically closed field the converse is also true.
(ii) As J 6= K[x1 , x2 , . . . , xn ], there exists a maximal ideal m containing J.
By (i), m = hx1 a1 , x2 a2 , . . . , xn an i for some ai K, 1 i n.
Therefore {(a1 , a2 , . . . , an )} = V(m) V(J), so V(J) 6= .

(iii) If f J, then f k J for some k. Hence f k (P ) = (f (P ))k = 0 for
every P V(J), therefore
f (P ) = 0 for every P V(J), i. e., f I(V(J)).
This means that I I(V(J)) without assuming that K is algebraically
closed.

We need to prove the other inclusion, I(V(J)) I. Let f I(V(J)). Lets
introduce a new variable y. Let f1 , f2 , . . . , fr be a set of generators for J,
and let J1 = hf1 , f2 , . . . , fr , f y 1i / K[x1 , x2 , . . . , xn , y].
We claim that V(J1 ) = . Assume that P = (a1 , a2 , . . . , an , b) V(J1 )
An+1 . As fi J1 for every i, 1 i r, we have fi (a1 , a2 , . . . , an ) = 0 for
every i, 1 i r, so (a1 , a2 , . . . , an ) V(J). Therefore f (a1 , a2 , . . . , an ) = 0,
too, but then (f y 1)(P ) = 1, so P / V(J1 ) after all. Therefore V(J1 ) = ,
as claimed.
By (ii), this implies that J1 = K[x1 , x2 , . . . , xn , y], in particular 1 J1 . This
means that we can write
r
X
1 = (f y 1)g0 + gi fi
i=1

for suitable g0 , g1 , . . . , gr K[x1 , x2 , . . . , xn , y]. Let y N be the highest power

of y occurring in any of g0 , g1 , . . . , gr . Lets multiply the above equality
by f N . Whenever there is a power y, say y k , occurring in gi for some i,
0 i r, we can write y k f N as (f y)k f N k , so for each i, 0 i r,
f N gi (x1 , x2 , . . . , xn , y) = Gi (x1 , x2 , . . . , xn , f y) for a suitable polynomial Gi .

12
Therefore
r
X
N
f = (f y 1)G0 (x1 , x2 , . . . , xn , f y) + fi Gi (x1 , x2 , . . . , xn , f y).
i=1

By substituting y = 1/f into the above equality, we obtain

r
X
N
f = fi Gi (x1 , x2 , . . . , xn , 1).
i=1

The right-hand side is an element of J, therefore f J as required.
The function V from ideals of K[x1 , x2 , . . . , xn ] to affine algebraic varieties
in An is not injective, and the function I from subsets of An to ideals of
K[x1 , x2 , . . . , xn ] is neither injective nor surjective. However, the combina-
tion of Proposition 1.6 and the Nullstellensatz implies that V and I create
a bijection between radical ideals of K[x1 , x2 , . . . , xn ] and affine algebraic
varieties in An .

13
1.4 Irreducibility
Definition. An affine algebraic variety V is reducible if and only if it can
be written as V = V1 V2 , where V1 , V2 are also affine algebraic varieties,
V1 6= V 6= V2 . If V is not reducible, it is called irreducible. (The irreducibility
of V is equivalent to saying that if V = V1 V2 , where V1 , V2 are also affine
algebraic varieties, then V1 = V or V2 = V .)
Examples:
1.The variety V(hx2 y 2 i) R2 is reducible, since it is the union of V(hxyi)
and V(hx + yi), which are the lines y = x and y = x.
2. Any variety consisting of a single point is obviously irreducible.
3. Warning: Graphs can be useful in finding the irreducible components,
but they can also be misleading. The graph below shows the the curve
y 2 + (x2 4)(x2 1) = 0. Topologically it consists of two disjoint con-
nected components, but it is irreducible, the two parts cannot be separated
algebraically.
2

-2 -1 1 2

-1

-2

Theorem 1.8 Every affine algebraic variety V can be decomposed into a

union V = V1 V2 . . . Vk such that every Vi , 1 i k, is an irreducible
affine algebraic variety and Vi 6 Vj for i 6= j. The decomposition is unique
up to the ordering of the components.
Definition. The Vi , 1 i k, in the above theorem are called the irreducible
components of V .
Proof. Lets call a variety good if it can be written as the union of finitely
many irreducible affine algebraic varieties and bad otherwise. We shall prove
first that all affine algebraic varieties are good.
Assume that V is a bad affine algebraic variety. V cannot be irreducible, so
V = U1 W1 for some proper subvarieties. Furthermore, at least one of U1
and W1 must also be bad, since if both of them were good, their union would
also be good. We may assume that U1 is bad. Then by a similar argument,

14
U1 = U2 W2 , where U2 and W2 are proper subvarieties of U1 , and at least
one of them must be bad, we may assume it is U2 .
Continuing like this, we obtain an infinite strictly decreasing sequence U1
U2 U3 . . . of bad varieties. Ui Ui+1 implies I(Ui ) I(Ui+1 ) by the
remark after the definition of I. By Proposition 1.6 (ii), V(I(Ui )) = Ui and
V(I(Ui+1 )) = Ui+1 , therefore I(Ui ) 6= I(Ui+1 ). Thus

I(U1 ) I(U2 ) I(U3 ) . . . I(Un ) I(Un+1 ) . . .

is a strictly increasing sequence of ideals, but this contradicts the Noetherian

property of the polynomial ring (Theorem 1.2), so the assumption that V is
bad was incorrect. Therefore any affine algebraic variety can be written as the
union of finitely many irreducible affine algebraic varieties, and by discarding
varieties which are contained in others, we can write V = V1 V2 . . . Vk
such that every Vi , 1 i k, is an irreducible affine algebraic variety and
Vi 6 Vj for i 6= j. This proves the existence of the decomposition.
Let V = V10 V20 . . . Vl0 be another decomposition of V into irreducible
varieties with Vi0 6 Vj0 for i 6= j. Then

V1 = V1 V = (V1 V10 ) (V1 V20 ) . . . (V1 Vl0 ).

As V1 is irreducible, there exists i such that V1 = V1 Vi0 , i. e., V1 Vi0 .

By applying this argument to Vi0 , we obtain that there exists j such that
Vi0 Vj , therefore V1 Vj . But different components do not contain each
other, therefore j = 1 and V1 = Vi0 . By applying this to each Vr (1 r k)
and Vj0 (1 r l), we obtain a bijection between {V1 , V2 , . . . , Vk } and
{V10 , V20 , . . . , Vl0 }, so the decompositions are the same apart from possibly the
order of the components.
Definition. Let R be a commutative ring. An ideal I / R is called a prime
ideal if and only if ab I implies a I or b I.
Proposition 1.9 An affine algebraic variety W is irreducible if and only if
I(W ) is prime.
Proof. We shall prove the contrapositive, namely that W is reducible if and
only if I(W ) is not prime.
Assume that W is reducible, W = W1 W2 , where W1 and W2 are affine
algebraic varieties and W 6= W1 , W 6= W2 . W1 W implies I(W1 ) I(W ),
but V(I(W )) = W and V(I(W1 )) = W1 by Proposition 1.6 (ii), therefore
I(W1 ) 6= I(W ). Let f1 I(W1 ) \ I(W ). Similarly, let f2 I(W2 ) \
I(W ). If P W1 , then (f1 f2 )(P ) = 0 because f1 (P ) = 0, if P W2 , then

15
(f1 f2 )(P ) = 0 because f2 (P ) = 0, therefore (f1 f2 )(P ) = 0 for every P W ,
so f1 f2 I(W ). Neither f1 , nor f2 is an element of I(W ), therefore I(W )
is not prime. This shows that if I(W ) is prime, then W is irreducible.
Assume now that I(W ) is not prime. Let f1 , f2 be polynomials such that f1 ,
f2 / I(W ), but f1 f2 I(W ). Let W1 = V(hI(W ), f1 i), W2 = V(hI(W ), f2 i).
W1 , W2 are affine algebraic varieties. Clearly W1 W , W2 W , but
W1 6= W since f1 / I(W ) and similarly W2 6= W since f2 / I(W ). For any
P W , 0 = (f1 f2 )(P ) = f1 (P )(f2 )(P ), so f1 (P ) = 0 or f2 (P ) = 0. In the
first case, P W1 , in the second P W2 , therefore W = W1 W2 and W is
reducible. This shows that if W is irreducible, then I(W ) is prime.
Remark. The criterion involves I(W ), so if W = V(J) for some ideal J, the
primality or otherwise of J is not sufficient in general to tell whether Wis
irreducible or not. If the field is algebraically closed, then I(V(J)) = J
by the Nullstellensatz. If J is prime, then it is automatically radical, so
I(V(J)) = J and V(J) is irreducible. If J is radical but not prime, then
V(J) is reducible.
The functions V and I create a bijection between prime ideals of K[x1 , x2 , . . . , xn ]
and irreducible affine algebraic varieties in An .
Properties of irreducible varieties and examples
1. Irreducibility is preserved under affine equivalence. If X and Y are affine
equivalent affine algebraic varieties, then X is irreducible if and only if Y
is. Let : X Y be an affine equivalence and assume that X can be
written as X = X1 X2 , where X1 , X2 are also affine algebraic varieties
and X1 6= X 6= X2 , then Y = (X1 ) (X2 ), (X1 ), (X2 ) are also affine
algebraic varieties and (X1 ) 6= Y 6= (X2 ).
2. Let K be an algebraically closed field. Let H An (K) be a hypersur-
face, that is, an algebraic variety defined by an ideal generated by a sin-
gle polynomial f K[x1 , x2 , . . . , xn ], H = V(hf i). f can be factorised as
f = f11 f22 frr , where the fi , 1 i r, are irreducible polynomials such
that fi is not a scalar multiple of fj if i 6= j, and i , 1 i r, are positive
integers. This factorisation is unique up to scalar factors.
p p
We shall now show that hf i = hf1 f2 fr i. Let g hf i. g can be
factorised as g = g11 g22 gss , where the gi , 1 i s, are irreducible
polynomial such that gi is not a scalar p multiple of gj if i 6= j, and i , 1
i s, are positive integers. g hf i means that there exists a positive
integer m such that g m = g1m1 g2m2 gsms hf i, therefore for each i,
1 i r, fi |g1m1 g2m2 gsms . As fi is irreducible, it must divide gj for
some j, 1 j s, so fi |g. The fi are pairwise coprime, therefore it follows

16
that f1 f2 . . . fr |g, g hf1 f2 fr i. Conversely,
p if g hf1 f2 fr i, then
g max{1 ,2 ,...,r } hf i. This shows that I(H) = hf i = hf1 f2 fr i.
By Proposition 1.3 (i), ri=1 V(hfi i) = Vhf1 f2 fr i = H. By the argument
S
used previously, hfi i/K[x1 , x2 , . . . , xn ] is a prime ideal, therefore it is radical,
so I(V(hfi i)) = hfi i by the Nullstellensatz, by using the fact that hfi i is prime
again, it follows that V(hfi i) is irreducible. V(hfi i) does not contain V(hfj i)
if i 6= j, because fj is not a multiple of fi . This shows that the irreducible
components are V(hfi i), 1 i r.
In particular, H is irreducible if and only if r = 1, i. e., f is the power of a
single irreducible polynomial.
3. Ellipses, hyperbolas, parabolas are irreducible over C and R.
Lets consider first the unit circle given by x2 + y 2 1 = 0 over C. We shall
prove that x2 + y 2 1 is irreducible. Assume to the contrary that it can be
factorised, then the factors are necessarily linear, so

x2 + y 2 1 = (ax + by + c)(dx + ey + f )

for suitable a, b, c, d, e, f C. The x2 term in the product is adx2 , so

ad = 1. By dividing the first factor by a and multiplying the second by a, we
may assume that a = d = 1. By expanding the product on the right-hand
side and comparing coefficients of xy and y 2 we obtain b + e = 0 and be = 1,
which has solutions b = i, e = i and b = i, e = i. By symmetry, we may
assume b = i, e = i. Then from the coefficients of x and y we get c + f = 0
and if ic = 0, which give c = f = 0, but from the constant term we get
cf = 1, which is a contradiction, so x2 + y 2 1 is indeed an irreducible
polynomial.
It follows that hx2 + y 2 1i / C[x, y] is a prime ideal. Prime ideals are radical,
so by the Nullstellensatz (Theorem 1.7), I(V(hx2 + y 2 1i)) = hx2 + y 2 1i.
By Proposition 1.9, V(hx2 +y 2 1i) is irreducible over C as hx2 +y 2 1i/C[x, y]
is a prime ideal.
It can also be proved that I(V(hx2 + y 2 1i)) = hx2 + y 2 1i over R,
too. The main idea is that if x2 + y 2 1 does not divide f R[x, y],
then f (x, y) = x2 + y 2 1 = 0 only has finitely many solutions, but the
unit circle x2 + y 2 1 = 0 has infinitely many points over R, therefore
f / I(V(hx2 + y 2 1i)). The rest of the proof is the same as over C.
It can be proved similarly that x2 y 2 1 and x2 +y are irreducible polynomi-
als. By using the fact that irreducibility is preserved under affine equivalence,
it follows that all ellipses, hyperbolas and parabolas are irreducible varieties
over C and R.

17
4. Affine subspaces over an infinite field K are irreducible. (Note that al-
gebraically closed fields are infinite.) Every affine subspace is equivalent
to one of the form x1 = x2 = . . . = xk = 0 for some k. It can be
proved that I(V(hx1 , x2 , . . . , xk i)) = hx1 , x2 , . . . , xk i if K is infinite and that
hx1 , x2 , . . . , xk i/K[x1 , x2 , . . . , xn ] is a prime ideal directly from the definition.
5. Let C = V(hx2 y 2 + y 4 x2 2y 2 + 1i) A2 (C). This polynomial factorises
as

x2 y 2 + y 4 x2 2y 2 + 1 = (y 2 1)2 + x2 (y 2 1) = (x2 + y 2 1)(y 2 1)

= (x2 + y 2 1)(y 1)(y + 1)

y 1 and y + 1 have degree 1, so they are irreducible. We have already

showed that x2 + y 2 1 is irreducible. Therefore the irreducible components
of C are the lines V(hy 1i), V(hy + 1i) and V(hx2 + y 2 1i).

1.5

1.0

0.5

-2 -1 1 2

-0.5

-1.0

-1.5

Warning. The correspondence between the irreducible factors of the gener-

ator and the irreducible components described above only applies to hyper-
surfaces. An ideal generated by irreducible polynomials need not be prime.
Factorisation is still a useful tool in decomposition of varieties into irreducible
factors, but all elements of the ideal have to be considered, not just the gen-
erators.
6. V(hy 2 p(x)i) C2 is irreducible if p(x) C[x] is a polynomial of odd
degree.
It is sufficient to prove that y 2 p(x) is an irreducible polynomial in C[x, y].
Lets assume to the contrary that it factorises as y 2 p(x) = f (x, y)g(x, y)
for some f, g C[x, y]. The left-hand side has degree 2 in y, so there are two
possibilities, either one f , g has degree 2 in y and the other degree 0 (i. e., it
is a polynomial in x only), or both factors have degree 1 in y.
In the first case we can assume that f has degree 2 in y, so f (x, y) = a(x)y 2 +
b(x)y + c(x) and g(x, y) = d(x) for some polynomials a(x), b(x), c(x) and

18
d(x) C[x]. By considering the coefficient of y 2 in f (x, y)g(x, y) we obtain
a(x)d(x) = 1, so they are constants. Therefore g(x, y) = d(x) is a constant.
In the second case f (x, y) = a(x)y + b(x) and g(x, y) = c(x)y + d(x) for
some polynomials a(x), b(x), c(x) and d(x) C[x]. By considering the
coefficient of y 2 in f (x, y)g(x, y) we obtain a(x)c(x) = 1, so a(x) and c(x) are
constants. By dividing f (x, y) by a(x) and multiplying g(x, y) by a(x), we
can assume that a(x) = c(x) = 1. Then the coefficient of y in f (x, y)g(x, y)
is b(x) + d(x), this has to 0, so d(x) = b(x). The degree 0 term in y in
f (x, y)g(x, y) is b(x)d(x) = (b(x))2 , this would have to be equal to p(x),
but this is impossible, since p(x) has odd degree.
Hence y 2 p(x) is an irreducible polynomial and V(hy 2 p(x)i) is an irre-
ducible variety.
This includes as special cases V(hx3 + x2 y 2 i) (nodal cubic curve,), and
V(hx3 y 2 i) (cuspidal cubic curve) and elliptic curves (to be studied later in
the course).

4 4

2 2

-1 1 2 -0.5 0.5 1.0 1.5 2.0 2.5

-2 -2

-4 -4

Nodal cubic Cuspidal cubic

7. Let I = hx2 + y 2 + z 2 1, 3x2 + y 2 z 2 1i / C[x, y, z] and let W =

V(I) A3 be the variety defined by I. The graph below shows this variety,
the two surfaces are the sphere defined the equation x2 + y 2 + z 2 1 = 0 and
the hyperboloid of one sheet defined by 3x2 + y 2 z 2 1 = 0, W is their
intersection, the black curve.

19
 2
1
0
-1
-2
2

-1

-2
-2
-1
0
1
2

The generators themselves are irreducible, but

2 (3x2 + y 2 z 2 1) (x2 + y 2 + z 2 1)
2
x z = I,
2
too, and x2 z 2 = (x + z)(x z). Therefore by the argument of Proposition
1.9, W can be decomposed into the union of W1 = V(hx2 + y 2 + z 2 1, 3x2 +
y 2 z 2 1, x + zi) and W2 = V(hx2 + y 2 + z 2 1, 3x2 + y 2 z 2 1, x zi).
We shall prove that these varieties are irreducible.
Lets consider W1 = V(hx2 +y 2 +z 2 1, 3x2 +y 2 z 2 1, x+zi). The difference
of x2 + y 2 + z 2 1 and 3x2 + y 2 z 2 1 is divisible by x + z so we can omit one
of them from the set of generators, therefore W1 = V(hx2 +y 2 +z 2 1, x+zi).
The graph below show the intersection of the plane x + z = 0 with the sphere
and the hyperboloid.

20
 2
1
0
-1
-2
2

-1

-2
-2
-1
0
1
2

The plane x + z = 0 intersects both in the same curve, shown in black, this
is W1 . The intersection of a sphere with a plane is a circle. By rotating
W1 through /4 about the y-axis, we can transform it to the unit circle
x2 + y 2 1 = 0 in the xy-plane. We have already proved that the unit circle
is irreducible, the rotation is a Euclidean transformation, which is a special
case of an affine equivalence, therefore W1 is irreducible.
The irreducibility of W2 = V(hx2 + y 2 + z 2 1, 3x2 + y 2 z 2 1, x zi) =
V(hx2 + y 2 + z 2 1, x zi) can be proved similarly.
It is also clear that
neither
of W1 and W2 contains
theother, for exam-
ple,because(1/ 2, 0, 1/ 2) W1 , but
(1/ 2, 0, 1/ 2) / W2 , while
(1/ 2, 0, 1/ 2) W2 , but (1/ 2, 0, 1/ 2)
/ W1 . Therefore they are the
irreducible components of W .

21
2 Morphisms, co-ordinate rings, rational maps,
and function fields

2.1 Morphisms and co-ordinate rings

The obvious functions to consider on an affine algebraic variety V An (K)
are the polynomials. Two polynomials f , g define the same function V K
if and only if f g I(V ), this motivates the following definition.
Definition. Let V An (K) be an affine algebraic variety. Its co-ordinate
ring, denoted by K[V ], is defined to be K[V ] = K[x1 , x2 , . . . , xn ]/I(V ).
Definition. Let V Am (K) and W An (K) be affine algebraic varieties. A
morphism : V W is a function of the form (P ) = (1 (P ), 2 (P ), . . . , n (P ))
with i K[V ] for each i, 1 i n.
Elements of K[V ] can be considered as morphisms V A1 .
The composition of two morphisms : V W and : W X is a
morphism : V X.

Definition. The morphism : V W is called an isomorphism if and only

if it has an inverse morphism, i. e., if there exists a morphism : W V
such that = idV and = idW . V and W are called isomorphic if
and only if there exists an isomorphism between them.
Isomorphic affine algebraic varieties are the same for most purposes. One
of the goals of algebraic geometry is to classify algebraic varieties up to
isomorphism and to decide whether two varieties are isomorphic. One of the
tools is to find properties of algebraic varieties which are preserved under
isomorphism.
Examples:
1. Any affine map is a morphism. Invertible affine maps are isomorphisms.
Affine equivalent affine algebraic varieties are isomorphic.
2. Let K be an arbitrary infinite field.
Let t be the co-ordinate on A1 and x, y the co-ordinates on A2 . Let V = A1
and W A2 be the parabola defined by the equation y x2 = 0. Then
I(V ) = {0}, I(W ) = hy x2 i (this is where we need K to be infinite), so
K[V ] = K[t] and K[W ] = K[x, y]/hy x2 i.
: V W , t 7 (t, t2 ) is a morphism. t and t2 are elements of K[V ] and
x = t, y = t2 satisfy y x2 = 0 for any t. (This shows that (t) W for

22
every t V = A1 , so is a morphism V W , not just V A2 .)
: W V , (x, y) 7 x is also a morphism. x K[W ] (technically it should
be x + I(W ), but this distinction is usually not made in practice) and there
are no equations to check. ( )(t) = (t, t2 ) = t and ()(x, y) = (x) =
(x, x2 ) = (x, y) for (x, y) W , therefore and are inverses of each other,
so V and W are isomorphic.
3. Let K be an infinite field of characteristic other than 2, for example K = R
or C. Let V = A1 and let W A2 be the nodal cubic curve defined by the
equation x3 + x2 y 2 = 0.
4

-1 1 2

-2

-4

: V W , t 7 (t2 1, t(t2 1)) is a morphism. t2 1 and t(t2 1) are

elements of K[V ] and x = t2 1, y = t(t2 1) satisfy x3 + x2 y 2 = 0 for
any t, so is a function V W .
is not an isomorphism, since it is not injective, (1) = (1) = (0, 0).
(We need the characteristic to be different from 2 to ensure that 1 6= 1.)
In examples 45, K can be an arbitrary infinite field.
4. Let H A2 be the hyperbola defined by xy1 = 0, and let : H A1 be
the morphism (x, y) 7 x. is not an isomorphism, since it is not surjective
because 0
/ im .
Warning: The image of a morphism is not necessarily an algebraic variety.
In this example im = A1 \{0}, and this set is not a variety, since any proper
subvariety of A1 is finite.
5. Let V = A1 and let W A2 be the cuspidal cubic curve defined by the
equation y 2 x3 = 0. : V W , t 7 (t2 , t3 ) is a morphism. t2 and t3 are
elements of K[V ] and x = t2 , y = t3 satisfy y 2 x3 = 0 for any t, so is a
function V W .

23
 4

-0.5 0.5 1.0 1.5 2.0 2.5

-2

-4

is bijective as a function, but it is not

( an isomorphim. It has an inverse
y/x, if x 6= 0
as a function : W V , (x, y) = , but this does not
0, if x = 0
appear to be a morphism from that way it is defined. Lets assume that
: W V is a morphism, which is the inverse of . The is represented
by a polynomial f K[x, y] such that f (t2 , t3 ) = t, but this is impossible.
We shall prove later (see the example after Corollary 2.2) that V and W
are not even isomorphic, not just that this particular morphism is not an
isomorphism.
Warning. Any isomorphism is bijective, since it has an inverse as a function,
but as the above example shows, a bijective morphism is not necessarily an
isomorphism.
6. Let x, y be the co-ordinates on A2 (C) and u, v, w the co-ordinates on
A3 (C). Let V = A2 and W A3 be the variety defined by the equation
uw v 2 = 0, it is the cone shown below.

24
 : V W , (x, y) 7 (x2 , xy, y 2 )) is a morphism. x2 , xy, y 2 are elements of
K[V ] and u = x2 , v = xy and w = y 2 satisfy uw v 2 = x2 y 2 (xy)2 = 0
for all x, y, therefore (V ) W , so is indeed a morphism V W , not
just V A3 . is not an isomorphism because it is not injective, (x, y) =
(x, y) for any (x, y) A2 , but (x, y) 6= (x, y) unless (x, y) = (0, 0).
(This argument works as long as the characteristic of the field is not 2. In
characteristic 2, (x, y) = (x, y) and is bijective if the field is algebraically
closed, but it is still not an isomorphism.)
Theorem 2.1 Let V Am (K) and W An (K) be affine algebraic varieties.
Let y1 , y2 , . . . , yn be co-ordinates on An . For a morphism : V W , define
: K[W ] K[V ], (f ) = f . : 7 is a bijection between
morphisms of affine algebraic varieties : V W and ring homomorphisms
: K[W ] K[V ] preserving K.
Proof. Let f K[W ]. f = F + I(W ) for some F K[y1 , y2 , . . . , yn ]. Now
(f ) = f = F = F (1 , 2 , . . . , n ), where 1 , 2 , . . . , n K[V ] are
the components of . As F is a polynomial, (f ) = F (1 , 2 , . . . , n )
K[V ], too. If G K[y1 , y2 , . . . , yn ] is another polynomial such that f =
G + I(W ), then F G I(W ), so F ((P )) = G((P )) for every P V
and therefore (f ) is a well-defined element of K[V ]. This shows that is
indeed a function K[W ] K[V ].
() = for any K and (f +g) = (f )+ (g), (f g) = (f ) (g)
for all f , g K[W ] by the definition of , therefore is a homomorphism
of rings which preserves K.
Given a homomorphism : K[W ] K[V ] with () = for every K,

25
let i = (yi + I(W )) K[V ] for i = 1, 2, . . . , n. Then = (1 , 2 , . . . , n )
is clearly a morphism V An . We need to prove that (V ) W so that
is a morphism V W .
Let P V an arbitrary point. In order to prove (P ) W , we need to show
g((P )) = 0 for any g I(W ).
Let q : K[y1 , y2 , . . . , yn ] K[W ] = K[y1 , y2 , . . . , yn ]/I(W ) be the quotient
map q(f ) = f + I(W ). Then i = (q(yi )), so

g((P )) = g((q(y1 ))(P ), (q(y2 ))(P ), . . . (q(yn ))(P )).

g is a polynomial, composed of its variables, scalars in K, addition and mul-

tiplication, while and q are ring homomorphisms preserving K, therefore

g((q(y1 )), (q(y2 )), . . . , (q(yn ))) = (q(g(y1 , y2 , . . . , yn ))) = (0) = 0

since g I(W ) = ker q. If we substitute the co-ordinates of P , we obtain

g((P )) = 0. As this holds for every g I(W ), we have (P ) W , so is
indeed a morphism V W .
If = (1 , 2 , . . . , n ) is a morphism of algebraic varieties V W , then
i = (yi + I(W )), so the ring homomorphism constructed from is
exactly . Conversely, if is a ring homomorphism : K[W ] K[V ] and
: V W is the morphism of algebraic varieties constructed from , then
= .
This shows constructions associating a ring homomorphism to a homomor-
phism of algebraic varieties and a homomorphism of algebraic varieties to
a ring homomorphism are inverses of each other, therefore : 7 is a
bijection.
Note. 7 is contravariant, i. e., if : V W and : W X are
morphisms, then ( ) = : K[X] K[V ].
The definition of the homomorphism may seem complicated, but it can be
written down very easily. In Example 6 before the theorem, V = A2 , W =
V(huw v 2 i) A3 and : V W is the morphism (x, y) = (x2 , xy, y 2 ).
: K[W ] K[V ] is defined by (u + I(W )) = x2 , (v + I(W )) = xy,
(w + I(W )) = y 2 and since it is ring homomorphism that preserves K,
(f ) can be calculated for any f K[W ].
Corollary 2.2 : V W is an isomorphism of affine algebraic varieties if
and only if : K[W ] K[V ] is an isomorphism of rings. V and W are
isomorphic if and only if there is an isomorphism between K[V ] and K[W ]
which preserves K.

26
Examples: In these examples K can be an arbitrary infinite field.
1. Let H = V(hxy 1i) A2 be a hyperbola, we shall prove that it is not
isomorphic to A1 . It can be proved by direct calculation that I(H) = hxy1i.
(Over an algebraically closed field this follows from the Nullstellensatz, as
xy 1 is an irreducible polynomial.) In K[H], (x + I(H))(y + I(H)) =
1 + I(H), i. e., there exist invertible elements which are not scalars, whereas
in K[A1 ] the only invertible elements are the scalars. This shows that K[H]
and K[A1 ] are not isomorphic, therefore H is not isomorphic to A1 either.
2. We shall show that the cuspidal cubic (Example 5 before Theorem 2.1) is
not isomorphic to A1 .
Let t be the co-ordinate on A1 and x, y the co-ordinates on A2 . Let V = A1
and let W A2 be the cuspidal cubic curve defined by y 2 x3 = 0 and
let : V W be the morphism t (t2 , t3 ). We already proved that is
bijective, but it is not an isomorphism. Now we shall prove that W is not
isomorphic to V by showing that K[W ] is not isomorphic to K[V ] = K[t].
It can be proved by direct calculation that I(W ) = hx3 y 2 i. (Over an
algebraically closed field this follows from the Nullstellensatz, as x3 y 2 is
an irreducible polynomial.)
In order to simplify notation, let f = f + I(W ) for f K[x, y]. (x) = t2
and (y) = t3 , so (xa y b ) = t2a+3b . This way we can obtain all non-
negative powers of t except t itself. As the xa y b (a, b Z0 ) generate K[W ]
as a vector space over K, the image of is the subring S of K[t] consisting
of polynomials with no degree 1 term.
This gives another proof of the fact is not an isomorphism, since is not
an isomorphism, because it is not surjective.
We shall prove that K[W ] is not isomorphic to K[V ]
= K[t] by showing that
K[W ] cannot be generated by a single element and K as a ring. If there
existed an element in z K[W ] which together with K generated K[W ] as
a ring, then (z) and K would generate S, the image of K[W ] under .
(In fact, is an isomorphism between K[W ] and S.)
We shall prove that S cannot be generated by a single element and K as a
ring. Lets assume to the contrary that there exists such an element u S.
We can write u as u = a0 + a2 t2 + a3 t3 + . . . + ak tk for some k 0 and for
suitable coefficients a0 , a2 , a3 , . . . , ak K.
u and u a0 generate the same ring, so we may assume a0 = 0, so u =
a2 t2 + a3 t3 + . . . + ak tk . The elements of the ring generated by u and K are of
the form b0 + b1 u + b2 u2 + . . . bn un , for some n 0 and b0 , b1 , b2 , . . . , bn K.
If S is generated by K and u, then all elements of S, in particular t2 and t3

27
haver to be able to written in this form with suitable coefficients b0 , b1 , b2 ,
. . . , bn .
The degree 2 term in b0 + b1 u + b2 u2 + . . . bn un is b1 a2 t2 the degree 3 term
is b1 a3 t3 . In order to be able to get t2 with a non-zero coefficient, we need
to have a2 6= 0. Similarly, in order to get t3 with a non-zero coefficient, we
need a3 6= 0. This, however, means that if b1 = 0 then both t2 and t3 have
coefficient 0 in b0 + b1 u + b2 u2 + . . . bn un , while if b1 6= 0 then both t2 and t3
have a non-zero coefficient, we are not able to obtain just t2 or just t3 . This
contradiction shows that S cannot be generated by u and K.
K[t] = K[V ] is clearly generated by t and K, so we can conclude that
K[V ] 6= K[W ] and therefore V and W are not isomorphic either.

The correspondence between varieties and ideal can be re-written in terms

of varieties and co-ordinate rings.
Definition. An element r in a ring is called nilpotent if and only if there exists
a positive integer n such that rn = 0.
Algebraic fact: An ideal I in a ring R is radical if and only if R/I has no
nilpotent elements other than 0.
Co-ordinate rings of affine algebraic varieties over K are rings which contain
K, have no nilpotent elements other than 0 and are generated as rings by K
and finitely many other elements (the images of the co-ordinate functions).
Conversely, if we have such a ring R, we can take choose a set of elements
x1 , x2 , . . . , xn / K which together with K generate R. The fact that
they and K generate R means that there is a surjective ring homomorphism
K[x1 , x2 , . . . , xn ] R, let J be its kernel, then R = K[x1 , x2 , . . . , xn ]/J
by the First Ring Isomorphism Theorem. I is radical since quotient ring
K[x1 , x2 , . . . , xn ]/J has no nilpotent elements other than 0. If K is an al-
gebraically closed field, then J = I(V(J)) by the Nullstellensatz, so R =
K[V(J)].
If K is algebraically closed, then V 7 K[V ] gives a bijection between iso-
morphism classes of affine algebraic varieties and isomorphism classes of rings
which which contain K, have no nilpotent elements other than 0 and which
can be generated by K and finitely many other elements.
An affine algebraic variety V is irreducible if and only if I(V ) is a prime
ideal, which is equivalent to K[V ] being an integral domain, therefore in the
above correspondence irreducible varieties correspond to integral domains
which contain K and which can be generated by K and finitely many other
elements.

28
If a property of affine algebraic varieties can be expressed purely in terms
of algebraic properties of co-ordinate rings, then that property is preserved
under isomorphism.
Being an integral domain an algebraic property preserved by isomorphism,
therefore irreducibility is preserved by isomorphism of affine algebraic vari-
eties.

29
2.2 Rational functions and maps, function fields
Definition. Let V be an irreducible affine algebraic variety. The function
field of V , denoted by K(V ), is the set of fractions f /g f, g K[V ], g 6= 0,
considering two such fractions f1 /g1 and f2 /g2 equal if and only if f1 g2 = f2 g1
in K[V ]. (K(V ) is indeed a field with the natural operations.)
The elements of K(V ) are called rational functions on V .
A rational function on V is defined at P V if and only if it can be
represented in the form f /g such that g(P ) 6= 0, and in that case the value
(P ) is defined to be (P ) = f (P )/g(P ), and this value is independent of
the choice of f and g.
Example: If V = An , then K(An ) = K(x1 , x2 , . . . , xn ), the field of rational
functions in n variables.
Lemma 2.3 Let V be an irreducible affine algebraic variety. Let = f /g
f, g K[V ], g 6= 0. If for some point P V , f (P ) 6= 0 and g(P ) = 0, then
is not defined at P .
Proof. Let = f1 /g1 f1 , g1 K[V ], g1 6= 0 be another representation of
. By the definition of equality of rational functions, f1 g = f g1 in K[V ],
so f1 (P )g(P ) = f (P )g1 (P ). g(P ) = 0, so the left-hand side is 0, but then
f (P ) 6= 0 implies that on the right-hand side g1 (P ) = 0, because K[V ] is an
integral domain.
Remarks. 1. In general, all possible representations of a rational function
need to be considered when deciding whether it is defined at a point. For
example, let x, y, z, w be the co-ordinates on A4 . Let V = V(hxyzwi) A4 ,
and let (x, y, z, w) = x/z. x/z not defined at P = (0, 1, 0, 0), but x/z =
w/y, since xy = zw K[V ]. w/y is defined at P , so (P ) = 0/1 = 0.
2. However, if K[V ] is a unique factorisation domain, then can be written
as = f /g with f , g coprime. In this case it is sufficient to consider only this
one representation of , because all others are of the form = (f h)/(gh) for
some h K[V ], therefore is defined at a point P if and only if g(P ) 6= 0.
In particular, this applies if V = An for some n.
Definition. Let V Am (K) and W An (K) be affine algebraic varieties
and let V be irreducible. A rational map : V 99K W is a function defined
on a non-empty subset of V given by (P ) = (1 (P ), 2 (P ), . . . , n (P )) with
i K(V ) for each i, 1 i n, and such that if (P ) is defined at P , i. e.,
if i (P ) is defined for each i, 1 i n, then (P ) W .
Rational functions and rational maps are generally not functions on V , only

30
partial functions defined on a non-empty subset of V . The set where a
rational function or a rational map is not defined is a subvariety, the locus
where the denominators of all the representations of the rational function
or of the components of the rational map vanish. The set where a rational
function or a rational map is defined is the complement of the previous set,
therefore it is a Zariski open subset.
Definition. A rational map : V 99K W is called dominant if and only if its
image is not contained in any proper subvariety of W .
Remark. Let V , W be an irreducible affine algebraic varieties, and let :
V 99K W and : W 99K X be rational maps. If the composite : V 99K
X is defined, it is also a rational map. If is dominant, is always
defined.
Definition. A rational map : V 99K W is called a birational equivalence if
and only if there exists a rational map : W 99K V such that = idV
and = idW . (In this case and are necessarily dominant.) V and W
are birationally equivalent if and only if there exists a birational equivalence
between them. A variety is rational if and only if it is birationally equivalent
to An for some n.
Theorem 2.4 Let V Am (K) and W An (K) be irreducible affine al-
gebraic varieties. There exists a bijection between dominant rational maps
: V 99K W and field homomorphisms : K(W ) K(V ) preserving K
given by the following constructions: for a rational map : V 99K W ,
define : K(W ) K(V ), (f ) = f , and for a homomorphism
: K(W ) K(V ), define a rational map by ((y1 ), (y2 ), . . . , (yn )), where
y1 , y2 , . . . , yn are the co-ordinates on An .
Sketch of Proof: (Not examinable) We introduce an alternative view of ra-
tional maps. Let g K[V ], g 6= 0. Let x1 , x2 , . . . , xm be the co-ordinates
on Am . Let us consider Am+1 = Am A1 with a new co-ordinate z on
A and let Vg = V (I(V ), zg 1) Am+1 . (Vg is the graph of z = 1/g
1

over V .) Let K[V ][1/g] be the subring of K(V ) generated by K[V ] and
1/g, it consists of the elements of K(V ) that can be written as f /g k for
some f K[V ] and some non-negative integer k. There is an isomorphism
between g : K[V ][1/g] K[Vg ] given by g (f /g k ) = f z k . g : V 99K Vg ,
(x1 , x2 , . . . xm ) (x1 , x2 , . . . xm , 1/g(x1 , x2 , . . . , xm )) is a rational map, while
its inverse g : Vg V , (x1 , x2 , . . . xm , z) (x1 , x2 , . . . xm ) is a morphism.
For K[V ][1/g], g () = g and for K[Vg ], 1 g () = g . g and
g give a bijection between the Zariski open subset {P V | g(P ) 6= 0} in
V and Vg .

31
Let now : V 99K W be a dominant rational map. We can assume that
= (f1 /g, f2 /g, . . . , fn /g) for some g K[V ], if necessary we can multiply
the numerator and denominator of each component by a suitable factor to
get into this form. Now g : Vg 99K W is a rational map which is defined
at every point of Vg , therefore it is a morphism (problem sheet 4, question
5). By Theorem 2.1, there exists a corresponding homomorphism of rings
( g ) : K[W ] K[Vg ]. g is dominant as and are, therefore
( g ) is injective (problem sheet 4, question 4 (a)). By composing it
with the isomorphism 1 1
g , we obtain an injective morphism g ( g ) :
K[W ] K[V ][1/g]. As it is injective, it can be extended to a morphism of
function fields : K(W ) K(V ), and it can be checked that it has the
property (yi ) = fi /g = yi for each i, 1 i n, therefore (f ) = f
for any f K(W ).
Conversely, let : K(W ) K(V ) be a morphism of fields preserving K. Let
g K[V ] be such that (yi ) = fi /g for for a suitable fi K[V ] for each i,
1 i n. Then (K[W ]) K[V ][1/g], so by Theorem 2.8, g : K[W ]
K[Vg ] determines a morphism : Vg W . Moreover, is dominant since
g is injective (problem sheet 4, question 3 (a)), because is injective, as
it is a homorphism of fields. Therefore = g is a rational map V 99K W
and = .
Corollary 2.5 : V 99K W is a birational equvalence if and only if :
K(W ) K(V ) is an isomorphism of fields. V and W are birationally
equivalent if and only if K(V ) and K(W ) are isomorphic as extensions of K.
A variety V is rational if and only if K(V )
= K(t1 , t2 , . . . , tk ) for some k.
Birational equivalence is an equivalence relation on affine algebraic varieties,
which is weaker than isomorphism. Birational equivalence classes of varieties
are in bijection with the isomorphism classes of finitely generated extensions
of K.
Birational equivalence of V and W means that there exist g K[V ] and
h K[W ] such that Vg = Wh , so in a sense, V and W have isomorphic
non-empty Zariski open subsets.
Examples: Worked examples can be found in the separate handout at http:
//www.maths.manchester.ac.uk/~gm/teaching/MATH32062/rational.pdf
and the accompanying Mathematica file containing animated images can be
downloaded from http://www.maths.manchester.ac.uk/~gm/teaching/MATH32062/
rational.nb.

32
3 Tangent spaces, dimension and singulari-
ties
Let H An (K) be a hypersurface defined by the equation f (x1 , x2 , . . . , xn ) =
0, where f K[x1 , x2 , . . . , xn ], and and let P H. One of the possible ways
of defining that the line {P + tv | t K} (v = (v1 , v2 , . . . , vn ) K n \ {0})
is tangent to H at P is to require that f (P + tv), now simply a function of
t, has a stationary point at t = 0.
The Taylor expansion of f (P + tv) about t = 0 is

d f (P + tv) 
f (P + tv) = f (P ) + t P
+(terms of order 2 in t)
dt
n
X f 
= f (P ) + t v + (terms of order 2 in t).
P j
j=1
x j

(Derivatives of polynomials can be defined purely formally over any field

without using the concept of limits and Taylor expansion works for polyno-
mials over any field.)
n
X f 
Thus f (P +tv) has a stationary point at t = 0 if and only if v = 0.
P j
j=1
x j
  n
f f f X f 
By using the gradient f = , , , , the condition vj =
x1 x2 xn j=1
xj P
0 can be re-written as (f )P .v = 0.
Definition. Let V An (K) be an affine algebraic variety and let P V . A
line ` given in parametric form as {P + tv | t K}, (v K n \ {0}), is said
to be tangent to V at P if and only if (f )P .v = 0 for every f I(V ).
Definition. The tangent space to V at P , denoted by TP V is the union of
the tangent lines to V at P and the point P .
Theorem 3.1 Let V An be an affine algebraic variety. Let f1 , f2 , . . . , fr
be a set of generators for I(V ). Let P V and let v = (v1 , v2 , . . . , vn )
K n \ {0}. The line ` = {P + tv | t K} is tangent to V at P if and only if

33
JP v = 0, where J is the Jacobian matrix,

f1 f1 f1
x1
x2 xn
f2 f2 f2


J = x 1 x2 xn
... ..
.
... ..
.

fr fr fr

x1 x2 xn

and JP is J evaluated at P .
In particular, TP V is an affine subspace of dimension n rankJP .
Proof. Assume that ` is tangent to V at P . Then (f )P .v = 0 for every
f I(V ), in particular, (fi )P .v = 0 for every i, 1 i r. (fi )P .v is
exactly the ith component of JP v, therefore Jp v = 0.
r
P
Assume now that JP v = 0. Let f I(V ). Then f = gi fi for some
i=1
gi K[x1 , x2 , . . . , xn ], 1 i r. Now
r
X r
X
f = (gi fi ) = ((gi )fi + gi fi )
i=1 i=1

and so
r
X r
X
(f )P = ((gi )P fi (P ) + gi (P )(fi )P ) = gi (P )(fi )P .
i=1 i=1

In the last step we used that fi (P ) = 0 for every i, 1 i r. Hence

Pr
(f )P .v = gi (P )(fi )P .v, but (fi )P .v = 0, because it is the ith com-
i=1
ponent of JP v = 0, therefore (f )P .v = 0.
The set W = {v K n | JP v = 0} is a linear subspace, therefore TP V =
P +W is an affine subspace. Its dimension is dim Tp V = dim W = nrankJP
by the Rank-Nullity Formula.
Example: Let V A2 (C) be the cuspidal cubic curve defined by the equation
y 2 x3 = 0. It can be checked that y 2 x3 is irreducible, so hy 2 x3 i is a
prime ideal, therefore it is radical and then the Nullstellensatz implies that
that I(V ) = hy 2 x3 i. Therefore J = (3x3 2y).

34
 y

x
-0.5 0.5 1.0 1.5 2.0 2.5

-1

-2

At P = (1, 1), JP = (3 2), therefore the line {P + tv | t C} is tangent

to V at (1, 1) if and only if 3v1 + 2v2 = 0. Hence T(1,1) V is the line of slope
3/2 through P with equation y = 3x/2 1/2. It agrees with the tangent line
obtained by, for example, implicit differentiation.
At Q = (0, 0), however, something strange happens. JQ = (0, 0), therefore
every line through (0, 0) is a tangent line and T(0,0) V = A2 !
Let v = (, ) 6= (0, 0) be the direction vector of a line through Q = (0, 0).
Then Q + tv = (t, t) and if we substitute x = t, y = t into y 2 x3 we
obtain t2 2 t3 3 = t2 ( 2 t3 ), which has a zero of multiplicity at least
2 at t = 0, so any line through (0, 0) is a tangent line by our definition. If
= 0, then t2 ( 2 t3 ) has a zero of multiplicity 3 at t = 0, so the line
y = 0 has a higher order tangency to the curve than a general line through
(0, 0).
PropositionDefinition 3.2 Let V be an irreducible affine algebraic vari-
ety. If V 6= , there exists a proper subvariety of V (possibly ) such that
dim TP V is constant outside this subvariety. (In other words, dim TP V is
constant on a non-empty Zariski open subset of V .) This common value
is called the dimension of V , denoted by dim V . The points at which
where dim TP V = dim V are called non-singular, while the points where
dim TP V > dim V are called singular and the set of singular points of V is
denoted by Sing V . (Sometimes it is convenient to define dim = 1 and
Sing = .)
Proof. Let Vm = {P V | dim TP V m}, where m 0 is an integer. By
Theorem 3.1, Vm = {P V | rank JP n m}. It is a theorem in linear
algebra that a matrix has rank k if and only if all of its (k + 1) (k + 1)

35
minors vanish. Therefore
Vm = {P V | all (n m + 1) (n m + 1) minors of JP vanish}.
This shows that Vm is the affine algebraic variety defined by the ideal gener-
ated by I(V ) and by the (n m + 1) (n m + 1) minors of J. We have
V = V0 V1 V2 . . . Vn Vn+1 = . Choose d maximal such that
Vd = V . Then dim TP V = d for P V \ Vd+1 and dim TP V > d for P Vd+1 ,
so dim V = d and Sing V = Vd+1 .
Example: If V is an affine subspace of An , TP V = V at every point P V ,
so the dimension of V as a variety agrees with its dimension as an affine
space and it has no singular points. In particular, An is non-singular and has
dimension n.
Worked examples can be found in the separate handout at http://www.
maths.manchester.ac.uk/~gm/teaching/MATH32062/singularities.pdf.
Herwig Hausers gallery at http://homepage.univie.ac.at/herwig.hauser/
bildergalerie/gallery.html has many nice examples of singular surfaces.
The combination of the preceding results gives the following procedure for
finding the dimension and singular points of an irreducible variety V
An (K).
1. Choose a set of generators f1 , f2 , . . . , fr for I(V ). (In all the problems
you will have to solve, the field will be algebraically closed and the defining
ideal of V will be a radical ideal, so it will be equal to I(V ).)
2. Calculate the Jacobian matrix J.
3. For k = 0, 1, 2, . . ., find the points where all the (k + 1) (k + 1) minors
of J vanish and also f1 = f2 = . . . = fr = 0. These are the points where
rank JP k and therefore dim TP V n k. Do this until you find the
smallest k such that dim TP V n k for every point P V .
4. Then dim V = nk and Sing V is the set of points where dim TP V > nk,
i. e., the points of V where all the k k minors of J vanish.
Definition. Let V be an arbitratry (not necessarily irreducible) affine alge-
braic variety. The dimension of V , dim V , is the maximum of the dimensions
of the irreducible components of V .
The local dimension of V at P V is defined to be the maximum of the
dimension of the irreducible components of V containing P .
P is called non-singular if and only if dim TP V is equal to the local dimension
of V at P .
P is called singular if and only if dim TP V is greater than the local dimension
of V at P . The set of singular points of V is denoted by Sing V , just as in
the irreducible case.

36
Fact: Sing V consists of the singular points of the individual irreducible com-
ponents and of intersection points of different components.
This means that in general, one needs to decompose the variety into its
irreducible components in order to find the its dimension and its singular
points, however, by the proposition below and by problem sheet 5, question 3,
hypersurfaces over an algebraically closed field are an exception.
Proposition 3.3 Let K be an algebraically closed field. Let H An (K)
be a hypersurface, that is, H = V(hf i) for some non-constant polynomial
f K[x1 , x2 , . . . , xn ], then dim H = n 1.
Proof. f can be factorised as f = f11 f22 frr , where the fi , 1 i r, are
irreducible polynomials such that fi is not a scalar multiple of fj if i 6= j, and
i , 1 i r, are positive integers. The irreducible components are V(hfi i),
1 i r, so if we prove the proposition for f irreducible, then it will imply
the general case, because each component of H has dimension n 1.
Lets assume that f is irreducible, then H is also irreducible and I(H) = hf i.
 
f f f
We have J = , ,..., . rank JP can only be 0 or 1 and if
x1 x2 xn
rank JP = 1 for some P H, then dim H = n 1.
f f
Assume to the contrary that rank JP = 0 for every P H. Then , ,
x1 x2
f
..., I(H) = hf i, that is, all the partial derivatives are multiples of f .
xn
As their degree is smaller than the degree of f , the only way this can happen
is if they are all 0. This is not possible in characteristic 0 since then f would
be a constant, which has been excluded.
In characteristic p it is possible for all the partial derivatives to be 0, it
happens if and only if the exponent of every variable in every term is a
d xn
multiple of p. ( = nxn1 = 0 if and only if p divides n.) Then
dx
 X p
X pi1 pi2 pin i i i
f= i1 ,i2 ,...,in x1 x2 xn = p
i1 ,i2 ,...,in x1 x2 xn
1 2 n
,
i1 ,i2 ,...,in i1 ,i2 ,...,in

so f is reducible, contradicting our assumption that it was irreducible.

Theorem 3.4 Let V An be an affine algebraic variety and let P V . Let
mP = {F K[V ] | F (P ) = 0} be the maximal ideal of K[V ] corresponding
to P . Then dim TP V = dim(mP /m2P ). (The quotient ring K[V ]/m2P is a
finite dimensional vector space, mP /m2P is a subspace in it.)

37
Proof. Lets introduce the notation f = f + I(V ) for f K[V ]. Let
P = (a1 , a2 , . . . , an ) and let F mP . F = f for a suitable f K[V ] and
F (P ) = f (P ) = 0. We noted in the proof of the Nullstellensatz (Theorem
1.7) that f (P ) = 0 if and only if f hx1 a1 , x2 a2 , . . . , xn an i, therefore
n
P
there exist polynomials gi K[x1 , x2 , . . . , xn ] such that f = (xi ai )gi in
i=1
n
P
K[x1 , x2 , . . . , xn ]. Hence f = (xi ai )gi in K[V ]. xi ai mp for every i,
i=1
1 i n, obviously, therefore mp = hx1 a1 , x2 a2 , . . . , xn am i / K[V ].
In the rest of the proof we shall assume P = (0, 0, . . . , 0) for simplicity, then
we have mp = hx1 , x2 , . . . , xn i / K[V ].
We can also assume that TP V is the subspace x1 = x2 = . . . = xk = 0 of
K n , where k = n dim TP V , this can be achieved by a linear change of
co-ordinates. This means that the reduced row echelon form of JP is

1 0 0 0 0

0 1 0 0 0
. .
. . . . . .. .. ..
. . . . .

0 0 1 0 0 ,

0 0 0 0 0

. . .. .. ..
.. .. . . .
0 0 0 0 0

with 1s on the diagonal in the first k rows and all the other entries are 0.
The rows of the reduced row echelon form are linear combinations of the rows
of JP . By applying the same linear combinations to the generators of I(V )
we obtain a set of generators f1 , f2 , . . . , fr for I(V ) such that the Jacobian
of this set of generators is the reduced row echelon form above, therefore
fi = xi + (terms of degree 2) for 1 i k, and fi only contains terms of
degree 2 for k + 1 i r.
Lets define a function D : mp K nk in the following way. Let F
mp . Then F = f for some f hx1 , x2 , . . . , xn i / K[x1 , x2 , . . . , xn ]. f can
Pn
be written as f = i xi + (terms of degree 2) and we set D(F ) =
i=1
(k+1 , k+2 , . . . , n ).
We shall prove that D is a surjective linear map with kernel m2p , then it will
follow that mP /m2P
= K nk , so dim(mP /m2P ) = dim K nk = n k.
Step 1. First we need to show that the vector (k+1 , k+2 , . . . , n ) does not
depend on the choice of f , so that D is indeed a function.

38
Let g K[x1 , x2 , . . . , xn ] be another polynomial such that F = g. Then
f g I(V ), so there exist hi K[x1 , x2 , . . . , xn ], 1 i r, such that
Pr
f g = hi fi . As fi = xi + (terms of degree 2) for 1 i k, and fi
i=1
r
P
only contains terms of degree 2 for k + 1 i r, the sum hi fi has no
i=1
linear terms in xk+1 , xk+2 , . . . , xn , so the coefficients k+1 , k+2 , . . . , n are
uniquely determined and therefore D(F ) is indeed a function.
Step 2. D is clearly linear, D(F + G) = D(F ) + D(G) for any F, G mP
and D(F ) = D(F ) for any F mP and any K.
Pn
Step 3. D is surjective, since D( i xi ) = (k+1 , k+2 , . . . , n ) for any
i=k+1
k+1 , k+2 , . . . , n K.
Step 4. ker D = {F mp | D(F ) = 0} by definition. We shall prove that
ker D = m2p .
Pn
Let F = f , f = i xi + (terms of degree 2) as in the definition of D.
i=1
Then F ker D if and only if k+1 = k+2 = . . . = n . This implies that
Pk
f i fi only contains terms of degree 2, i. e., it is an element of the
i=1
ideal h{xi xj | 1 i j n}i / K[x1 , x2 , . . . , xn ]. This is equivalent to the
existence of polynomials gij K[x1 , x2 , . . . , xn ], 1 i j n, such that
k
X X
f i f i = gij xi xj
i=1 1ijn

in K[x1 , x2 , . . . , xn ]. Then in K[V ], we have

X
f= gij xi xj ,
1ijn

i. e., F = f h{xi xj | 1 i j n}i = m2P / K[V ]. This shows that

ker D m2p .
Conversely, lets now assume that F m2p . Then
X
F = Gij xi xj
1ijn

for suitable Gij K[V ], 1 i j n. Now let gij K[x1 ,P

x2 , . . . , xn ],
1 i j n, be such that gij = Gij , then we can take f = gij xi xj
1ijn

39
and it satisfiesf = F . As f has no linear terms, we have D(F ) = (0, 0, . . . , 0),
therefore F ker D. This shows m2p ker D, by combining it with the
previous result we obtain ker D = m2p .
We have proved that D is a surjective linear map mp K nk with kernel
m2p , therefore mp /m2p
= K nk , in particular dim(mP /m2P ) = dim K nk =
n k = dim TP V .
Corollary 3.5 Let : V W be an isomorphism between affine algebraic
varieties. Then dim TP V = dim T(P ) W for every P V , therefore dim V =
dim W and (Sing V ) = Sing W . In particular, a singular variety cannot be
isomorphic to a non-singular variety.
Proof. By Theorem 2.1, induces a morphism : K[W ] K[V ] and by
Corollary 2.2, if is an isomorphism of varieties then is an isomorphism
of rings. It easy to check also gives a bijection between mP / K[V ] and
m(P ) / K[W ] and also between m2P / K[V ] and m2(P ) / K[W ]. Hence mp /m2p
and m(P ) /m2(P ) are isomorphic vector spaces, therefore by Theorem 3.4,
dim TP V = dim mp /m2p = dim m(P ) /m2(P ) = dim T(P ) W .
The other assertions follow from this, since dimension and singularity are
defined in terms of the dimensions of the tangent spaces at the points of the
variety.
Examples:
1. The affine line A1 (C) and the cuspidal cubic curve defined by the equation
y 2 x3 = 0 in A2 (C) are not isomorphic, since the former has no singular
point, while the latter has a singular point at (0, 0).
2. Let W1 = V(hxy, xz, yzi) A3 (C) and let W2 = V(hxy(x + y)i) A2 (C).
2

1
y
0
2
-1

-2 W1
2
1
W2
1

x
0 -2 -1 1 2

-1
-1

-2
-2

-1
-2
0

The irreducible components of W1 are the co-ordinate axes, V(hx, yi), V(hx, zi)

40
and V(hy, zi). The irreducible components of W2 are the lines V(hxi), V(hyi)
and V(hx + yi). Both varieties consist of three lines meeting at a point, the
origin, but we shall prove that W1 and W2 are not isomorphic.
Lets calculate T(0,0,0) W1 and T(0,0) W2 . hxy, xz, yzi) and hxy(x + y)i are
radical ideals, so they are equal to I(W1 ) and I(W2 ), respectively.

y x 0
The Jacobian of W1 is J(W1 ) = z 0 x. At (0, 0, 0), J(0,0,0) (W1 ) =
0 z y
0 0 0
0 0 0, so it has rank 0 and therefore dim T(0,0,0) W1 = 3. (T(0,0,0) W1 =
0 0 0
3
A can be proved even without calculating the Jacobian. Any f I(W1 ) is
identically 0 on the three co-ordinate axes, therefore these axes are tangent
lines. If we know that T(0,0,0) W1 is an affine subspace, then it follows that
T(0,0,0) W1 = A3 as no other affine subspace contains all three co-ordinate
axes.)
2 2


The Jacobian

of W 2 is J(W2 ) = 2xy + y x + 2xy . At (0, 0), J(0,0) (W2 ) =
0 0 , so it has rank 0 and therefore dim T(0,0) W2 = 2.
If there existed an isomorphism : W1 W2 , then we would have (0, 0, 0) =
(0, 0), as (0, 0, 0) and (0, 0) are uniquely characterised as the intersection
points of the components in W1 and W2 , respectively. However, isomor-
phisms preserve the dimension of the tangent space and dim T(0,0,0) W1 = 3 6=
dim T(0,0) W2 = 2, therefore there does not exist an isomorphism between W1
and W2 .

The rest of the material in this section is not examinable.

For any vector space X over K, the dual space X is defined to be the vector
space of all linear maps X K with the obvious operations. For any linear
map : X Y , there is a dual map : Y X . What Theorem 3.4 really
proves, without mentioning dual spaces explicitly, is that TP V
= (mP /m2P ) .
While any two finite dimensional vector spaces of the same dimension are
isomorphic, the isomorphism between TP V and (mP /m2P ) is natural, it can
be defined without reference to bases.
Given an arbitrary morphism : V W , by Theorem 2.1 we have a
morphism of rings : K[W ] K[V ], which also induces a linear map
m(P ) /m2(P ) mP /m2P . Taking duals, we obtain the differential map
dP : TP V = (mP /m2P ) T(P ) W
= (m(P ) /m2(P ) ) . dP is similar
to the differential map betwen tangent spaces in differential geometry.

41
Isomorphisms can be characterised by using the differential map, is an
isomorphism if and only if it is bijective and dP is an isomorphism for every
P V.
The dimension of a variety can be defined in other ways as well, which are
more suited for certain purposes, but are less useful for practical calculations.
1. dim V is equal to the maximum d for which there exists a chain of irre-
ducible subvarieties V0 V1 . . . Vd V . Moreover, if V is irreducible, any
such chain that cannot be extended by inserting further subvarieties has the
same length. From this definition it is clear that the dimension of a proper
subvariety of an irreducible variety is smaller than the dimension of the whole
variety, which is hard to prove by using tangent spaces.
2. If V is irreducible, then dim V is transcendence degree of K(V ) over K,
i. e., the maximal number of elements x1 , x2 , . . . , xd K(V ) which do not
satisfy any non-zero polynomial with coefficients in K. As a consequence of
an algebraic result called the Noether Normalisation Lemma, any variety is
birationally equivalent to a hypersurface, therefore dim V = d if and only if
V is birationally equivalent to a hypersurface in Ad+1 . This interpretation of
the dimension shows that the dimension of irreducible varieties is invariant
under birational equivalence, not just under isomorphism.
Fact: For any irreducible variety V in characteristic 0, there exist a non-
singular variety V and a birational morphism : V V such that is
an isomorphism outside Sing V . This is called the desingularisation of V .
The existence of a desingularisation is a very hard theorem, for which the
Japanese mathematician Heisuke Hironaka received the Fields Medal in 1970.
The problem is still open in prime characteristic in dimension 3.

42
4 Projective space and projective varieties
One of the motivations for projective geometry is to understand perspective.
It is well-known phenomenon that parallel lines such as railway tracks ap-
pear to meet at a point, called the vanishing point. It was not until the
Renaissance that painters were able to give a geometrically correct represen-
tation of objects. The painting below, The delivery of keys by Pietro Perug-
ino in the Sistine Chapel (source https://en.wikipedia.org/wiki/File:
Entrega_de_las_llaves_a_San_Pedro_(Perugino).jpg) is a good exam-
ple of the use of perspective, the parallel lines on the ground converge towards
a point near the centre of the painting.

Projective geometry gives a nicer theory than affine geometry in many ways,
for example, two lines in the projective plane always meet, or more generally
two curves defined by degree m and degree n equations, resp., have mn
intersection points counted with multiplicity.
Projective geometry also has important practical applications, images are
combined by applying projective transformations in photo stitching, which
is used in panoramic photos, Google Street View and virtual reality.

43
 4.1 Basic properties of projective space and projective
varieties
The definition of projective space is also motivated by the idea that if a
camera is at the origin of the co-ordinate system, all points of a ray through
the origin will be mapped to the same point of the image, so points of the
image correspond to lines through the origin.
Definition. The n-dimensional projective space over a field K, denoted by
Pn (K) or by Pn if the field is understood, is the set of equivalence classes
of K n+1 \ {(0, 0, . . . , 0)} under the equivalence relation (x0 , x1 , . . . , xn )
(x0 , x1 , . . . , xn ) for any K \ {0}. (The points of Pn correspond to
lines through the origin in K n+1 .)
The equivalence class of a point (X0 , X1 , . . . , Xn ) K n+1 \ {(0, 0, . . . , 0)} is
denoted by (X0 : X1 : . . . : Xn ). X0 , X1 , . . . Xn are called homogeneous
co-ordinates on Pn .

Two ways of looking at projective space

1. Let U0 = {(X0 : X1 : . . . : Xn ) Pn | X0 6= 0}. (X0 : X1 : . . . : Xn ) =
(1 : X1 /X0 : . . . : Xn /X0 ) in U0 and X1 /X0 , X2 /X0 , . . . , Xn /X0 can take
arbitrary values in K, so the points of U0 are in bijection with the points of
An . The set Pn \ U0 = {(X0 : X1 : . . . : Xn ) Pn | X0 = 0} is clearly a copy
of Pn1 . Therefore Pn = An Pn1 as a set.
If n = 1, then
X1

U0 = {(X0 : X1 ) P1 | X0 6= 0} = { 1 : | X1 K, X0 6= 0} = {(1 : x) | x K},
X0
so we can identify U0 with A1 via (X0 : X1 ) X1 /X0 A1 .

P1 \ U0 = {(X0 : X1 ) P1 | X0 = 0} = {(0 : X1 ) | X1 K \ {0}} = {(0 : 1)},

because (0 : X1 ) = (0 : 1) for any K \ {0}. This is the decomposition,

P1 = A1 P0 , as P0 consists of just a single point.
We can identify A1 with K and denote the single point of P0 by , this way
we can identify P1 with K {}.
If n = 2, then
X 1 X2

U0 = {(X0 : X1 : X2 ) P2 | X0 6= 0} = { 1 : : | X1 , X2 K, X0 6= 0}
X 0 X0
= {(1 : x : y) | x, y K}

44
 so we can identify U0 with A2 via (X0 : X1 : X2 ) (X1 /X0 , X2 /X0 ) A2 .
The complement of U0 is

P2 \U0 = {(X0 : X1 : X2 ) P2 | X0 = 0} = {(0 : X1 : X2 ) | (X1 , X2 ) K 2 \{(0, 0)}},

this is a copy of P1 .
To understand these points, consider the parallel lines x+y = 0 and x+y2 =
0 in A2 = U0 . As x = X1 /X0 and y = X2 /X0 , we can rewrite the equations
X1 X2 X 1 X2
in terms of homogeneous co-ordinates as + = 0 and + 2 = 0.
X 0 X0 X 0 X0
After multiplying by X0 we get X1 + X2 = 0 and X1 + X2 2X0 = 0. The
solutions of this system of linear equations are X0 = 0, X1 = X2 , which
correspond to the point (0 : 1 : 1) P1 . Therefore these parallel lines in
A2 = U0 intersect in P2 \ U0 . Any other line parallel to them also contains
(0 : 1 : 1).
y

H0:1:-1L 4

x+y-2=0

x
-4 -2 2 4

-2

x+y=0

-4

Similarly, all lines with direction vector (X1 , X2 ) intersect at (0 : X1 : X2 )

P2 \ U 0 .
This is true in general for arbitrary n, the points of Pn \U0 = Pn1 correspond
to directions of lines in An or equivalence classes of parallel lines. (0 : X1 :
. . . : Xn ) is the point where all the lines with direction vector (X1 , X2 , . . . , Xn )
in An meet. (The single point of P0 corresponds to the only line in A1 .) For

45
this reason, these points are often called points at infinity, but this does not
mean that they are intrinsically different, the disctinction depends on the
choice of co-ordinates.
The youtube video https://www.youtube.com/watch?v=q3turHmOWq4 (also
linked to from the Animations and videos section of the course website)
contains a good explanation of the projective plane.

2. The other approach is is to consider the sets

Ui = {(X0 : X1 : . . . : Xn ) Pn | Xi 6= 0},

0 i n. Each one is a copy of An and their union is Pn , since there is no

point (0 : 0 : . . . : 0). These n + 1 copies are glued together by identifying
their points via rational maps ij : Ui 99K Uj so that P Ui is the same
point of Pn as ij (P ) Uj .
For example, if n = 1, let t = X1 /X0 be the co-ordinate on U0 , and u =
X0 /X1 the co-ordinate on U1 . We have the rational maps 01 : U0 99K U1 ,
t 7 1/t, and 10 : U1 99K U0 , u 7 1/u. A point t U0 is the same point in
P1 as 01 (t) U1 , whenever 01 (t) is defined and similarly, a point u U1 is
the same point in P1 as 10 (u) U0 , whenever 10 (u) is defined.
If n = 2, then let x = X1 /X0 , y = X2 /X0 be the co-ordinates on U0 ,
u = X0 /X1 , v = X2 /X1 the co-ordinates on U1 and s = X0 /X2 , t = X1 /X2
the co-ordinates on U2 . Now we have u = 1/x and v = y/x, which give
the rational map 01 : U0 99K U1 , (x, y) 7 (1/x, y/x), so that (x, y) U0
and 01 (x, y) U1 are the same point in P2 whenever the latter is defined.
The point (2 : 3 : 4) P2 is the same as (1 : 3/2 : 2), (2/3 : 1 : 4/3) or
(1/2 : 3/4 : 1), therefore (3/2, 2) U0 , (2/3, 4/3) U1 and (1/2, 3/4) U2
correspond to the same point of P2 and they get identified in the gluing
process.
In general, if F K[X0 , X1 , . . . , Xn ] and P = (X0 : X1 : . . . : Xn ) Pn ,
F (P ) cannot be defined, since F (X0 , X1 , . . . , Xn ) will give different values
for different values of K \ {0}. However, if F is homogeneous of degree d,
then F (X0 , X1 , . . . , Xn ) = d F (X0 , X1 , . . . , Xn ), so we can tell whether
F (P ) = 0 or not, and this is enough to define projective algebraic varieties.
Definition. An ideal I / K[X0 , X1 , . . . , Xn ] is called homogeneous if and only
if it can be generated by homogeneous elements. (Warning: This does not
mean that all elements of the ideal are homogeneous polynomials.)

46
 Definition. Let I /K[X0 , X1 , . . . , Xn ] be a homogeneous ideal. The projective
algebraic variety defined by I is the set

V(I) = {(X0 : X1 : . . . : Xn ) Pn | F (X0 , X1 , . . . , Xn ) = 0, F I, F homogeneous}.

Two ways of looking at projective varieties

1. Lets consider the variety V = V(hX1 X2 X02 i) P2 . If X0 6= 0, we can
divide X1 X2 X02 = 0 by X02 to get (X1 /X0 )(X2 /X0 ) 1 =, which can be
written as xy 1 = 0 in terms of the affine co-ordinates x = X1 /X0 and
y = X2 /X0 on U0 = A2 , so V0 = V U0 is a hyperbola. The points V \ V0 are
the points of V with X0 = 0. By substituting X0 = 0 into X1 X2 X02 = 0
we obtain X1 X2 = 0, therefore the points at infinity are (0 : 1 : 0), which
corresponds to lines parallel to the x-axis, and (0 : 0 : 1), which corresponds
to lines parallel to the y-axis. The asymptotes of the hyperbola pass through
these points at infinity. V consists of the hyperbola with two points at infinity
corresponding to the asymptotes.

H0:0:1L
y
4

xy-1=0

H0:1:0L
x
-4 -2 2 4

-2

-4

Starting with the equation xy 1 = 0 we can recover X1 X2 X02 = 0 by

substituting x = X1 /X0 and y = X2 /X0 into xy 1 = 0 and multiplying it
by X02 .
Lets now consider the variety V = V(hX12 X0 X2 i) P2 . If X0 6= 0, we can
divide X12 X0 X2 = 0 by X02 to get (X1 /X0 )2 (X2 /X0 ) = 0, which can
be written as x2 y = 0 in terms of the affine co-ordinates x = X1 /X0 and
y = X2 /X0 on U0 = A2 , so V0 = V U0 is a parabola. The points V \ V0 are
the points of V with X0 = 0. By substituting X0 = 0 into X12 X0 X2 = 0

47
we obtain X12 = 0, therefore the only points at infinity is (0 : 0 : 1), which
corresponds to lines parallel to the y-axis. Unlike the hyperbola, the parabola
has no asympotes, but for large x and y, its tangent direction gets approaches
the direction of the y-axis.

H0:0:1L
y

y-x2=0

x
-3 -2 -1 1 2 3

Starting with the equation x2 y = 0 we can recover X12 X0 X2 = 0 by

substituting x = X1 /X0 and y = X2 /X0 into x2 y = 0 and multiplying it
by X02 .
Given a homogeneous polynomial F K[x1 , x2 , . . . , xn ],

F [X0 , X1 , . . . , Xn ]
= F [1, x1 , x2 , . . . , xn ]
X0deg F
where xi = Xi /X0 , 1 i n, is called the dehomogenisation of F with
respect to X0 . If we dehomogenise all homogeneous elements of an homo-
geneous ideal J / K[x1 , x2 , . . . , xn ], we obtain an ideal defining the affine
algebraic variety V0 = V U0 An , called the affine piece X0 6= 0 of
V = V(J) Pn .
If we start with a polynomial f (x1 , x2 , . . . , xn ),

X0deg f f (X1 /X0 , X2 /X0 , . . . , Xn /X0 ) K[X0 , X1 , . . . , Xn ]

is a homogeneous polynomial, the homogenisation of f . Homogenisation can

also be done at the level of ideals, the projective algebraic variety in Pn
defined by the ideal generated by the homogenisation of the elements of an
ideal J / K[x1 , x2 , . . . , xn ] is called the projective closure of V(J) An . The

48
affine piece X0 6= 0 of the projective closure is exactly V(J), while the points
at infinity correspond to asymptotic directions of V(J) as we have seen in
the example.
Homogenisation and dehomogenisation are one-sided inverses of each other.
Homogenisation followed by dehomogenisation always yields the same poly-
nomial, and similarly for varieties, taking the projective closure of an affine
variety and then the affine piece X0 6= 0 gives the original variety.
Dehomogenising and then homogenising a homogeneous polynomial will give
the polynomial divided by the highest power of X0 dividing it. (For example,
the dehomogenisation of X0 X1 is x1 , whose homogenisation is just X1 .)
For varieties this means that taking projective closure of the affine piece
X0 6= 0 of a projective algebraic variety will give the union of irreducible
components of the original variety not contained in the hyperplane X0 = 0.
(Irreducibility for projective varieties will be defined later, but it is completely
analogous to the affine case.)
If a projective variety V has no irreducible components contained in the
hyperplane X0 = 0, then V is the projective closure the affine piece V0 =
V U0 and the points of V correspond to the points of V0 and the asympotic
directions of V0 .

2. The other approach is to consider all the affine pieces Vi = V Ui ,

0 i n, of a projective variety V . These are affine varieties and they are
glued together by identifying the points via the rational maps ij : Ui 99K Uj
defined previously. For example, by homogenising the equation y 2 x3 x2 =
0 of the nodal cubic in A2 , we obtain X0 X22 X13 X0 X12 = 0. The projective
curve defined by this equation is the projective closure V , whose affine piece
V0 = V U0 is original affine nodal cubic curve. Substituting X0 = 0 into
the equation gives X13 = 0, therefore the only point with X0 = 0 is (0 : 0 : 1),
the asymptotic direction of the y-axis.
x and y can be expressed as x = X1 /X0 , y = X2 /X0 in terms of the homo-
geneous co-ordinates. The affine co-ordinates on the other affine pieces are
u = X0 /X1 , v = X2 /X1 on U1 and s = X0 /X2 , t = X1 /X2 on U2 . Deho-
mogenising X0 X22 X13 X0 X12 = 0 with respect to X1 gives uv 2 1 u = 0,
dehomogenising it with respect to X2 gives s t3 st2 = 0. Therefore the
projective curve has the affine pieces shown below.

49
 y 2 x3 x2 = 0 uv 2 1 u = 0 s t3 st2 = 0
y v t
10 10 10

5 5 5

x u s
-10 -5 5 10 -10 -5 5 10 -10 -5 5 10

-5 -5 -5

-10 -10 -10

They are glued together by the maps 01 : U0 99K U1 , (x, y) 7 (1/x, y/x)
and 02 : U0 99K U2 , (x, y) 7 (1/y, x/y).

Definition. Let F K[X0 , X1 , . . . , Xn ] be a polynomial. The degree i ho-

mogeneous part of F , denoted by F[i] , is the sum of all the terms of degree i
in F . If i < 0 or i > deg F , F[i] is defined to be 0.
Example: Let F = X0 +X0 X2 X12 +X0 X1 X2 K[X0 , X1 , X2 ], then F[0] = 0,
F[1] = X0 , F[2] = X0 X2 X12 and F[3] = X0 X1 X2 .
Lemma 4.1 The ideal I / K[X0 , X1 , . . . , Xn ] is homogeneous if and only if
for any F I, all the homogeneous parts of F are also elements of I.
Proof. Assume that I is homogeneous, let F1 , F2 , . . . , Fr be a set of
Pr
homogeneous generators for I. Let F I, then F = Fi Gi for some
i=1
r
P
Gi K[X0 , X1 , . . . , Xn ], 1 i r. The F[d] = Fi G[ddeg Fi ] I for every
i=0
d, 0 d deg F , so all the homogeneous parts of F are also elements of I.
Assume now that F I implies that F[0] , F[1] , . . . , F[deg F ] are also in I. Let
F1 , F2 , . . . , Fr be an arbitrary set of generators for I. We claim that the set
deg
PFi
{Fi[j] | 1 i r, 0 j deg Fi } generates I. On one hand Fi = Fi[j] is
j=0
contained in the ideal generated by this set for each i, 1 i r, so this ideal
contains I. On the other hand, each of the generators is in I, so the ideal
generates by them is a subset of I. Therefore the ideal generated by this set
is exactly I and since these generators are homogeneous, I is a homogeneous
ideal.
Proposition 4.2 (Cf. Proposition 1.3)
(i) Let V1 = V(I1 ), V2 = V(I2 ), . . . , Vk = V(Ik ) be projective algebraic

50
 varieties in Pn . Then

V1 V2 . . . Vk = V(I1 I2 . . . Ik ) = V(I1 I2 . . . Ik )

is also a projective algebraic variety.

(ii) Let V = V(I ), A be projective algebraic varieties in Pn . Then
\ X

V = V I
A A

is also a projective algebraic variety.

Proof. Just imitate the proof of Proposition 1.3 (i) and (ii).
Warning: There is no direct analogue of Proposition 1.3 (iii) for projective
varieties because Pm Pn is very different from Pm+n .
Definition. The homogeneous ideal of a set Z Pn is the ideal I(Z) /
K[X0 , X1 , . . . , Xn ] generated by the set

{F K[X0 , X1 , . . . , Xn ] | F homogeneous, F (X0 , . . . , Xn ) = 0 (X0 : . . . : Xn ) Z}.

Theorem 4.3 (Projective Nullstellensatz, cf. Theorem 1.7)

Let K be an algebraically closed field and let J / K[X0 , X1 , . . . , Xn ] be a
homogeneous ideal.
(i) V(J) = if and
only if J
= K[X 0 , X1 , . . . , X n ] or J = hX0 , X1 , . . . , Xn i.
(ii) I(V(J)) = J unless J = hX0 , X1 , . . . , Xn i.
Idea of proof. For any homogeneous ideal J / K[X0 , X1 , . . . , Xn ] we can
consider the projective algebraic variety V(J) defined by in Pn and also the
affine algebraic variety defined by J in An+1 , called affine cone on V(J).
This gives an almost bijective correspondence between projective algebraic
varieties in Pn and certain affine algebraic varieties in An+1 , the only failure
of bijectivity is that K[X0 , X1 , . . . , Xn ] and hX0 , X1 , . . . , Xn i both define the
empty set as a projective variety.
Definition. A projective algebraic variety V is reducible if and only if it can
be written as V = V1 V2 , where V1 , V2 are also projective algebraic varieties,
V1 6= V 6= V2 . If V is not reducible, it is called irreducible.
Proposition 4.4 (Cf. Theorem 1.8) Every projective algebraic variety V
can be decomposed into a union V = V1 V2 . . . Vk such that every Vi ,
1 i k, is an irreducible projective algebraic variety and Vi 6 Vj for i 6= j.
The decomposition is unique up to the ordering of the components. The Vi ,
1 i k, are called the irreducible components of V .

51
Proof. Completely analogous to the proof of Theorem 1.8.
Lemma 4.5 A homogeneous ideal I / K[X0 , X1 , . . . , Xn ] is prime if and only
if for any homogeneous polynomials F , G K[X0 , X1 , . . . , Xn ], F G I
implies F I or G I.
Proof. If I is prime then F G I implies F I or G I for all F, G
K[X0 , X1 , . . . , Xn ] by the definition of a prime ideal.
Lets assume I is a homogeneous ideal which is not prime. Then there exist
polynomials P / I, Q / I such that P Q I. Let d 0 be the minimal
integer such that P[d]
/ I. There exists such a d since if P[j] I for every
deg
PP
j 0, then P = P[j] I, too. Similarly, let e 0 be the minimal integer
j=0
d+e
P
such that Q[e]
/ I. Then (P Q)[d+e] = P[j] Q[d+ej] . If j < d, then P[j] I,
j=0
so P[j] Q[d+ej] I, too. If j > d, then then Q[d+ej] I as d + e j < e, so
P[j] Q[d+ej] I, too. Therefore all the terms in the sum except for P[d] Q[e]
are in I, (P Q)[d+e] I by Lemma 4.1 since P Q I and I is homogeneous.
Hence P[d] Q[e] I, too, but P[d] / I and Q[e]
/ I. Therefore F = P[d] and
G = Q[e] are homogeneous polynomials with the property that F, G / I but
F G I.
Proposition 4.6 (Cf. Proposition 1.9) A projective algebraic variety V is
irreducible if and only if I(V ) is prime.
Proof. Imitate the proof of Proposition 1.9 and use Lemma 4.5.

52
4.2 Tangent spaces, dimension and singularities
There are two possible approaches.
1. One can define tangent lines and imitate the development of the affine
theory. The line through the points P 6= Q consists of the points P + Q,
where ( : ) P1 , and this line is tangent to V at P if and only if for any
homogeneous polynomial F I(V ), F (P + Q) contains no degree 1 term
in .
2. Let V Pn be an irreducible projective algebraic variety, and let P V .
Let (X0 : X1 : X2 : . . . , Xn ) be homogeneous co-ordinates on Pn . Choose
i such that P is not contained in the hyperplane Xi = 0. Let Vi = {(X0 :
X1 : . . . : Xn ) V | Xi 6= 0} be the affine piece Xi 6= 0 of V . The tangent
space TP V is defined as the projective closure of TP Vi , the local dimension
of V at P is the local dimension of Vi at P , and the P is a singular point
of V if and only if it is a singular point of Vi . (It needs to be proved that
these definitions are independent of the choice of i, which can be done by an
improved version of Theorem 3.5.)
Tangent spaces, dimension and singular points of projective varieties can be
calculated by using the Jacobian matrix in the same way as in the affine case.
Example: Find the singular points, if any, of the curve C defined by the
equation Y 2 Z X 3 X 2 Z = 0 in P2 (C).
Let F = Y 2 Z X 3 X 2 Z, then FX = 3X 2 2XZ, FY = 2Y Z and FZ =
Y 2 X 2 . The rank of the Jacobian J is 0 where all 3 partial derivatives
vanish, and 1 elsewhere.
The points P C where rank Jp = 0 are the solutions of F = FX = FY =
FZ = 0. FY = 0 implies Y = 0 or Z = 0. If Y = 0, then FZ = 0 implies
X = 0, and we get the point (0 : 0 : 1), which is indeed on C. (In projective
space, (0 : 0 : 1) = (0 : 0 : Z) for any Z 6= 0.) If Z = 0, then FX = 0 implies
X = 0, and then FZ = 0 implies Y = 0, but (0 : 0 : 0) is not a point in P2 .
Therefore rank J = 0 and dim TP C = 2 at (0 : 0 : 1), and rank J = 1 and
dim TP C = 1 at all other points of C. Hence dim C = 1 and (0 : 0 : 1) is the
only singular point of C. If we take the affine piece Z 6= 0 of C, we get the
familiar nodal cubic curve y 2 x3 x2 = 0, which has a singular point at
(0, 0), corresponding the singular point of the projective curve.

53
4.3 Functions, rational maps and morphisms
Definition. The homogeneous co-ordinate ring of a projective variety V Pn
is K[V ] = K[X0 , X1 , . . . , Xn ]/I(V ). If V is irreducible, the function field
of V , K(V ), is the set of equivalence classes of fractions F/G, where F ,
G K[V ], G 6= 0, F , G are homogeneous of the same degree, and two
fractions F1 /G1 , F2 /G2 are equivalent if and only if F1 G2 = F2 G1 in K[V ].
Warning: Unlike in the affine case, the elements of K[V ] are not functions
on V .
As I is a homogeneous ideal, we can also define homogeneous elements in
K[V ] degree j as the images of the homogeneous polynomials of degree j in
K[X0 , X1 , . . . , Xn ] and then every element of K[V ] can be written uniquely
as the sum of homogeneous elements just like polynomials.
The elements of K(V ) are partially defined functions on V . If F , G are both
homogeneous of degree d, then

F (X0 , X1 , . . . , Xn ) = d F (X0 , X1 , . . . , Xn )

and
G(X0 , X1 , . . . , Xn ) = d G(X0 , X1 , . . . , Xn ),
so the value of F/G does not depend on the representantive of the point
(X0 : X1 : . . . : Xn ) Pn chosen. Similarly, this value does not change if
another fraction representing the same element of K(V ) is chosen as long as
the denominator does not vanish.
Definition. Let V Pm (K) and W Pn (K) be projective algebraic varieties
and let V be irreducible. A rational map : V 99K W is a partial function
defined by an equivalence class of (n + 1)-tuples of homogeneous elements
of K[V ] of the same degree, (0 : 1 : . . . : n ) (0 : 1 : . . . : n )
if and only if i j = j i for every i, j, 0 i, j n. : V 99K W is
defined at P V if and only if can be represented by (0 : 1 : . . . : n )
such that i (P ) 6= 0 for some i, 0 i n, in this case we require that
(P ) = (0 (P ) : 1 (P ) : . . . : n (P )) W .
A rational map : V W is a morphism if and only if it is defined at every
point of V . (This definition of a morphism is not really the right one if K is
not algebraically closed, but it will do for our purposes. If K is algebraically
closed, our definition is analogous to the characterisation of morphisms of
affine algebraic varieties as rational maps which are defined everywhere, see
problem sheet 4, question 5, but for morphisms of affine algebraic varieties
there is a simpler definition which works over any field.)

54
A rational map : V 99K W is dominant if and only if its image is not
contained in any proper subvariety of W .
A rational map : V 99K W is a birational equivalence if and only if it has
an inverse rational map : W 99K V such that = idV , = idW .
In this case V and W are called birationally equivalent. A projective variety
is called rational if and only if it is birationally equivalent to Pn for some n.
A morphism : V W is an isomorphism if and only if it has an inverse
morphism : W V such that = idV , = idW . In this case V
and W are called isomorphic.

Worked examples of rational maps and morphisms between projective vari-

eties can be found on a separate handout at http://www.maths.manchester.
ac.uk/~gm/teaching/MATH32062/projexamples.pdf.
Remarks. 1. Rational maps between projective varieties can be defined in
terms of (n + 1)-tuples of elements of K(V ), just like in the affine case, but
it is more common to write them in terms of elements of K[V ].
2. The same degree condition in the definition of the rational map means
that 0 , 1 , . . . , n in one representation of must be homogeneous of the
same degree, but this common degree can be different for different represen-
tations of .
3. If K[V ] has unique factorisation into irreducibles (e. g., V = Pm so
K[V ] = K[X0 , X1 , . . . , Xn ]), then every rational map can be represented in
a form (0 : 1 : . . . : n ) such that 0 , 1 , . . . , n have no common fac-
tors and this representation is unique up to a scalar factor. It is sufficient
to consider this only representation of because all others are of the form
(F 0 : F 1 : . . . : F n ) for some homogeneous element F K[V ].
4. Morphisms are functions, whereas rational maps are only partially defined
functions, defined outside a proper subvariety.
5. The image of a projective algebraic variety under a morphism is also a
projective algebraic variety.
Theorem 2.4 and Corollary 2.5 can be adapted to the context of projective
varieties.
Warning: There is no equivalent of Theorem 2.2 and Corollary 2.3 for mor-
phisms between projective varieties.
Theorem 4.7 (Cf. Theorem 2.4) Let V Pm (K) and W Pn (K) be
irreducible projective algebraic varieties. Let (Y0 : Y1 : Y2 : . . . : Yn ) ho-
mogeneous co-ordinates on Pn and assume that W does not lie in the hy-
perplane Y0 = 0. There exists a bijection between dominant rational maps

55
 : V 99K W and field homomorphisms : K(W ) K(V ) preserving K
given by the following constructions: for a rational map : V 99K W ,
define : K(W ) K(V ), (f ) = f , and for a homomorphism
: K(W ) K(V ) preserving K, define the corresponding rational map
by (1 : (Y1 /Y0 ), (Y2 /Y0 ), . . . , (Yn /Y0 )).
Sketch of the proof. Let (X0 : X1 : X2 : . . . : Xm ) be homogeneous co-
ordinates on Pm . We can assume that V does not lie in the hyperplane X0 =
0. Let V0 be the affine piece X0 6= 0 of V and W0 the affine piece Y0 6= 0 of W .
Homogenisation with respect to X0 gives an isomorphism K(V ) K(V0 ).
Similarly, homogenisation with respect to Y0 gives an isomorphism K(W )
K(W0 ). If = (0 : 1 : . . . : n ), then = (1 /0 , 2 /0 , . . . , n /0 ) is a
rational map : V0 99K W0 , : K(W0 ) K(V0 ) gives : K(W ) K(V )
via the isomorphisms K(V ) = K(V0 ) and K(W ) = K(W0 ).
Conversely, given : K(W ) K(V ), there is a corresponding homomor-
phism K(W0 ) K(V0 ), which gives a rational map V0 99K W0 by Theorem
2.4. This rational map can be extended to the rational map V 99K W de-
scribed in the statement of the theorem.
Corollary 4.8 (Cf. Corollary 2.5) : V 99K W is a birational equvalence
if and only if : K(W ) K(V ) is an isomorphism of fields. V and W
are birationally equivalent if and only if K(V ) and K(W ) are isomorphic as
extensions of K. A variety V is rational if and only if K(V )
= K(t1 , t2 , . . . , tk )
for some k.

56
4.4 Projective transformations

a00 a01 a0n
a10 a11 a1n
Definition. Let A = .. .. be an (n + 1) (n + 1) invertible

.. . .
. . . .
an0 an1 ann
matrix with entries from the field K. The projective transformation defined
n n
by A is the rational map :P 99K P , (X0 : X1 : . . . : Xn ) = (Y0 : Y1 :
Y0 X0
Y1 X1 P
. . . : Yn ), where .. = A .. . Yi can be written as Yi = aij Xj , so

. .
Yn Xn
each is given by homogeneous degree 1 polynomials, therefore is indeed
a rational map. (We shall prove that the invertibility of A implies that
projective transformations are, in fact, morphisms.)
For any 6= 0 the matrix A defines the the same projective transformation
as A. The projective transformations of Pn form a group, called the projective
linear group, denoted by P GL(n + 1, K). It is the quotient of GL(n + 1, K),
the group of all invertible (n + 1) (n + 1) matrices by the normal subgroup
consisting of scalar multiples of the identity matrix.
Definition. Two sets V, W Pn are projectively equivalent if and only if
there exists a projective transformation : Pn 99K Pn such that (V ) = W .
Projective equivalence is an equivalence relation.
Examples:
1. The projective closure of the hyperbola xy 1 = 0 has equation X1 X2
X02 = 0 , while the projective closure of the parabola y x2 = 0 has equation
X12 X0 X2 = 0 . The projective transformation
(X0 : X1 : X2 ) 7 (X1 :
0 1 0
X0 : X2 ), corresponding to the matrix 1 0 0, maps one to the other. In
0 0 1
fact, all ellipses, parabolas and hyperbolas in P(R) or in P(C) are projectively
equivalent.
2. Any two sets of 4 distinct points in P2 with no 3 of them collinear are
projectively equivalent, this follows from question 2 on problem sheet 6.
However, the interior of the quadrilateral formed by one set of 4 points may
not be mapped to the interior of the quadrilateral formed by the other set of
4 points.

57
Proposition 4.9 Let n be a positive integer, let A be an invertible (n + 1)
(n + 1) matrix and let : Pn 99K Pn be the projective transformation defined
by A. Then is an isomorphism Pn Pn (an automorphism of Pn ).

X0 0
X1 0
Proof. As A is invertible, A .. = .. implies X0 = X1 = . . . = Xn = 0,

. .
Xn 0
therefore is defined at every point of Pn , so it is a morphism.
Let be the projective transformation defined by A1 . A1 is also invertible,
so by the previous argument, is also morphism Pn Pn .
Let (X0 : X1 : . . . : Xn ) =(Y0
: Y1 : . .
. : Yn
) and let
(Y 0 : Y1 : . .
. : Y
n) =
Y0 X0 Z0 Y0
Y1 X1 Z1 Y1
(Z0 : Z1 : . . . : Zn ), then .. = A .. and .. = A1 .. =

. . . .
Yn Xn Zn Yn

X0
X1
.. , therefore = idPn . Similarly, = idPn , so and are

.
Xn
inverses of each other, therefore they are isomorphisms.
Remark. By this theorem, projective transformations are automorphisms of
Pn . The converse is also true over algebraically closed fields, the automor-
phisms of Pn are exactly the projective transformations, see Theorem 4.12
for the case n = 1.

Projective transformations of P1 are of particular interest. If we identify P1

with K {} by (X0 : X1 ) X0 /X1 K (X1 6= 0) and  (X 0 : 0) , then
a b
the projective transformation given by the matrix can be written
c d
az + b
as (z) = with ad bc 6= 0. (Arithmetic involving or division by 0
cz + d
is handled in the obvious sensible way.) These maps are also called fractional
linear maps or Mobius transformations.
Definition. Let z1 ,z2 , z3 , z4 K {}, with at least 3 of them distinct. The
cross ratio of z1 ,z2 , z3 , z4 is defined as

(z1 z3 )(z2 z4 )
(z1 , z2 ; z3 , z4 ) =
(z1 z4 )(z2 z3 )

58
(Again, arithmetic involving or division by 0 is handled in the obvious
sensible way.) (http://en.wikipedia.org/wiki/Cross_ratio contains a
good description of the cross ratio and its properties.)
Proposition 4.10 Let z1 , z2 , z3 , z4 K {} with at least 3 of them
distinct. Then (z1 , z2 ; z3 , z4 ) = ((z1 ), (z2 ); (z3 ), (z4 )) for any projective
transformation : P1 P1 .
az + b
Proof. (z) = for some a, b, c, d K, ad bc 6= 0. If c = 0,
cz + d
a b bc ad 1 a
(z) = z + , while if c 6= 0, (z) = + . In either
d d c cz + d c
case, can be composed of maps of the form z 7 z + ( K), z 7 z
( K \ {0}) and z 7 1/z. It is easy to see that the first two preserve the
cross ratio and
     
1 1 1 1 z3 z1 z4 z2
 
1 1 1 1 z1 z3 z2 z4 zz zz
, ; , =    =  1 3  2 4 
z1 z2 z3 z4 1 1 1 1 z4 z1 z3 z2

z1 z4 z2 z3 z1 z4 z2 z3
(z3 z1 )(z4 z2 ) (z1 z3 )(z2 z4 )
= = = (z1 , z2 ; z3 , z4 ).
(z4 z1 )(z3 z2 ) (z1 z4 )(z2 z3 )

All three types of maps preserve the cross ratio, therefore so does .
Proposition 4.11 Let z1 , z2 , z3 , w1 , w2 , w3 be elements K {} such that
z1 , z2 , z3 are pairwise distinct and w1 , w2 , w3 are also pairwise distinct. Then
there exists a projective transformation : P1 P1 such that (zi ) = wi
for i = 1, 2, 3. In particular, any two sets of 3 distinct points in P1 are
projectively equivalent.
Proof. Let 1 (z) = (z1 , z2 ; z3 , z), then 1 is a projective transformation and
1 (z1 ) = , 1 (z2 ) = 0, 1 (z3 ) = 1. Similarly, let 2 (z) = (w1 , w2 ; w3 , z),
then 2 is a projective transformation and 2 (w1 ) = , 2 (w2 ) = 0, 2 (w3 ) =
1. Therefore = 1 2 1 is a projective transformation and it satisfies
(zi ) = wi for i = 1, 2, 3.
Remarks: 1. The projective transformation in the above proposition is
unique, since (zi ) = wi for i = 1, 2, 3, then (z1 , z2 ; z3 , z) = (w1 , w2 ; w3 , (z))
for any z K {} and this determines uniquely. If we write (z) =
az + b
, then the coefficients a, b, c, d are only determined up to multiplication
cz + d
by a non-0 scalar, but the function is unique.
2. For a higher dimensional analogue, see problem sheet 6, question 4.

59
Example: Find a projective transformation : P1 (C) P1 (C) such that
(3) = 4, (1) = 0, (0) = 1.
(3)(1 z) 3z + 3
Solution 1. Let 1 (z) = (3, 1; 0, z) = = . Let
(3 z)(1) z+3
(3)(z) 3z
2 (z) = (4, 0; 1, z) = = . The projective transformation
(4 z)(1) z + 4
4z
we want is = 1 1
2 1 . 2 (z) = , so
z+3

4 3z+3 12z+12


z+3 z+3 12z + 12 2z + 2
(z) = 3z+3 = 6z+12 = = .
z+3
+3 z+3
6z + 12 z+2

Solution 2. This is the recommended method as it is faster. As preserves

the cross ratio, (z) is characterised by the property that (3, 1; 0, z) =
3z + 3 3(z)
(4, 0; 1, (z)), i. e., = , (3z + 3)((z) + 4) = 3(z)(z + 3).
z+3 (z) + 4
2z + 2
Solving this equation for (z) gives (z) = .
z+2
Theorem 4.12 Any isomorphism P1 (C) P1 (C) is a projective transfor-
mation, i. e., of the form (X0 : X1 ) = (aX0 + bX1 : cX0 + dX1 ), where a, b,
c, d C with ad bc 6= 0.
Proof. Assume that : P1 P1 is an isomorphism. By Proposition 4.11,
there exists a projective transformation such that ((0 : 1)) = (0 : 1),
((1 : 0)) = (1 : 0) and ((1 : 1)) = (1 : 1). Let = .
can be written as (X0 : X1 ) = (0 (X0 , X1 ) : 1 (X0 , X1 )) with 0 , 1
coprime homogeneous polynomials of the same degree d. is bijective, in
particular it is injective, so there is a unique (X0 : X1 ) P1 such that (X0 :
X1 ) = (0 : 1), only (0 : 1) has this property, therefore 0 (X0 , X1 ) = X0d
for some C \ {0}. Similarly, 1 (X0 , X1 ) = X1d for some C \ {0}.
Now (1 : 1) = (1 : 1) implies = , so we may assume = = 1
and then (X0 : X1 ) = (X0d : X1d ). If is any dth root of unity, then
( : 1) = (1 : 1) = (1 : 1), so the injectitivity of implies that d = 1,
= idP1 . Therefore = 1 is a projective transformation.
Remark. This proof also works over any algebraically closed field of char-
acteristic 0 and the theorem is, in fact, true without any restriction on the
field, but the general proof requires Galois theory.

60
5 Geometry in the plane
One of the nice features of the projective plane is that any two lines meet
at exactly one point. We shall see that over an algebraically closed field,
this can be generalised to arbitrary curves, they have exactly the expected
number of intersection points if we count the intersections with appropriate
multiplicity. This is analogous to counting the zeros of a polynomial with
multiplicity in the Fundamental Theorem of Algebra.
Definition. Let V be an irreducible affine variety and let P V . The local
ring of V at P is OV,P = { K(V ) | is defined at P }. This means that
OV,P consists of the rational functions which can be written as = f /g with
g(P ) 6= 0.
The function P : OV,P K, 7 (P ) is a ring homomorphism. It is
surjective, therefore its kernel mV,P = { OV,P | (P ) = 0} is a maximal
ideal. Any element of OV,P \ mV,P is invertible, so any proper ideal of OV,P
must be contained in mV,P , therefore mV,P is the unique maximal ideal of
OV,P . (In algebra, a local ring means a ring with a unique maximal ideal.)
Definition. Let f , g K[x, y] and let P A2 . The intersection multiplicity
of f , g at P is IP (f, g) = dim OA2 ,P /hf, gi.
If C and D are curves in A2 , then let I(C) = hf i and I(D) = hgi, and the
intersection multiplicity of C and D at P is IP (C, D) = IP (f, g).
For homogeneous polynomials in K[X, Y, Z] or for curves C and D in P2 ,
their intersection multiplicity at P P2 is defined to be the intersection
multiplicity of their dehomogenisations or of the corresponding affine curves
in an affine piece of P2 containing P .
If f K[x], then the multiplicity of f at a K is the dimension of OA1 ,a /hf i,
so the intersection multiplicity defined above is indeed a generalisation of the
multiplicity of a zero of a polynomial at a point.
Key properties of Ip (C, D):
1. IP (C, D) is finite if and only if C and D have no common component
containing P .
2. IP (C, D) = 0 if and only if P
/ C or P / D.
3. IP (C, D) = 1 if and only if P is a non-singular point of both C and D
and the tangent lines to C and D at P are different.
4. IP (C, D) > 1 if and only if P is a singular point of at least one of C and
D, or P is a non-singular point of both C and D but the tangent lines to C
and D at P are the same.

61
Rules for calculating IP (f, g):
1. IP (f, g) only depends on the ideal hf, gi, so f , g can be replaced by another
pair of polynomials that generate the same ideal.
2. IP (f1 f2 , g) = IP (f1 , g) + IP (f2 , g).
3. If f (P ) 6= 0 or g(P ) 6= 0 then IP (f, g) = 0.
4. If neither f , nor g has a repeated factor that vanishes at P , the curves
f = 0 and g = 0 are non-singular at P and have distinct tangent lines at P ,
then IP (f, g) = 1.
5. If g = ax + b y (a, b K) and P = (x0 , y0 ), then IP (f, g) is equal to
the multiplicity of x0 as a root of f (x, ax + b) = 0. (Of course, it can also be
applied with the roles of x and y reversed.)
Examples:
1. Calculate the intersection multiplicity I(0,0) (x3 + x2 y 2 , x + y).
If we substitute y = x into x3 +x2 y 2 we obtain x3 , so by rule 5, I(0,0) (x3 +
x2 y 2 , x + y) = 3.
2

-2 -1 1 2

-1

-2

2. Calculate the intersection multiplicity

I(0,0) (x3 + x2 y 2 , x2 2xy + y 2 x3 ).

By using rules 1 and 2

I(0,0) (x3 + x2 y 2 , x2 2xy + y 2 x3 )

= I(0,0) (x3 + x2 y 2 , (x2 2xy + y 2 x3 ) + (x3 + x2 y 2 ))
= I(0,0) (x3 + x2 y 2 , 2x(x y))
= I(0,0) (x3 + x2 y 2 , x) + I(0,0) (x3 + x2 y 2 , x y).

I(0,0) (x3 +x2 y 2 , x) = I(0,0) (y 2 , x) = 2I(0,0) (y, x) = 2 by rules 1, 2 and 4, and

similarly I(0,0) (x3 +x2 y 2 , xy) = I(0,0) (x3 +x2 y 2 (xy)(x+y), xy) =

62
 I(0,0) (x3 , x y) = 3I(0,0) (x, x y) = 3, hence I(0,0) (x3 + x2 y 2 , x2 2xy +
y 2 x3 ) = 5.
2

-2 -1 1 2

-1

-2

Theorem 5.1 (Bezout) Let K be an algebraically closed field, and let F ,

G K[X, Y, Z] be homogeneous
X polynomials with no common factor of
positive degree. Then IP (F, G) = deg F deg G.
P P2

Remark. It is relatively easy to prove by using the resultant that the number
of intersection points without considering multiplicity is at most deg F deg G
over any field, and quite often this is sufficient.
Definition. Let f = am xm + am1 xm1 + . . . + a1 x + a0 and g = bn xn +
bn1 xn1 + . . . + b1 x + b0 be polynomials in K[x]. The resultant of f and g
is defined to be the (m + n) (m + n) determinant
 
am am1 am2 . . . a1 a0 0 0 ... 0 0 0
 
0 am am1 . . . a2 a1 a0 0 ... 0 0 0
 
0 0 am . . . a 3 a2 a1 a0 . . . 0 0 0 
 . .. .. .. .. .. .. .. .. .. 
 .
 . . . . . . . . . .
0 0 0 . . . am am1 am2 am3 . . . a1 a0 0 
 
0 0 0 ... 0 am am1 am2 . . . a2 a1 a0 
 
Res(f, g) = 
 bn bn1 bn2 . . . b2 b1 b0 0 ... 0 0 0

0 bn bn1 . . . b3 b2 b1 b0 ... 0 0 0
 
 . .. .. .. .. .. .. .. .. .. 
 .. . . . . . . . . .

0 0 0 . . . bn2 bn3 bn4 bn5 . . . b0 0 0 

0 0 0 . . . bn1 bn2 bn3 bn4 . . . b1 b0 0 

0 0 0 ... b b b b ... b b b 
n n1 n2 n3 2 1 0

The resultant has the property that Res(f, g) = 0 if and only if am = bn = 0

or if f and g have a common factor of positive degree, which is equivalent

63
to having a common zero if K is algebraically closed. (For more information
on resultants, see http://en.wikipedia.org/wiki/Resultant and http:
//mathworld.wolfram.com/Resultant.html.)
We can dehomogenise F and G with respect to Z by introducing x = X/Z
and y = Y /Z, let the f, g K[x, y] be the polynomials obtained this way.
We can write f (x, y) = am (y)xm + am1 (y)xm1 + . . . + a1 (y)x + a0 and
g(x, y) = bn (y)xn + bn1 (y)xn1 + . . . + b1 (y)x + b0 , where the coefficients
ai , 0 i m, and bj , 0 j n, are polynomials in y, which we can
consider as elements of the field K(y). Using these ai , 0 i m, and bj ,
0 j n, in the above determinant we obtain the resultant of f and g with
respect to x, denoted by Resx (f, g) to indicate that we are considering f and
g as polynomials in x. Resx (f, g) is a polynomial in y. As F and G have no
common factors, neither do f and g, not even in K(y)[x], therefore Resx (f, g)
is not the 0 polynomial. By considering the total degrees of f and g, we have
deg ai deg f i for every i, 0 i m deg bj deg g j for every j,
0 j n. Therefore every term in the determinant defining Resx (f, g) has
degree at most deg f deg g, so deg Resx (f, g) deg f deg g deg F deg G.
If f (x0 , y0 ) = g(x0 , y0 ) = 0, then the 1-variable polynomials in x f (x, y0 ) =
am (y0 )xm + am1 (y0 )xm1 + . . . + a1 (y0 )x + a0 and g(x, y0 ) = bn (y0 )xn +
bn1 (y0 )xn1 + . . . + b1 (y0 )x + b0 have x0 as a common zero, therefore their
resultant is 0, but this resultant is just Resx (f, g)(y0 ). In other words, the y
co-ordinates of the intersection points of the curves f (x, y) = 0 and g(x, y) =
0 are among the zeros of Resx (f, g). As Resx (f, g) is a non-0 polynomial
of degree at most deg F deg G in y, there are at most deg F deg G possible
values for the y co-ordinate of the intersection points.
By swapping the roles of x and y, we can similarly obtain that there are
at most deg F deg G possible values for the x co-ordinate of the intersection
points of the curves f (x, y) = 0 and g(x, y) = 0.
Therefore the number of intersection points is finite. The intersection points
of the affine curves f (x, y) = 0 and g(x, y) = 0 correspond exactly to the
intersection points of the projective curves F (X, Y, Z) = 0 and G(X, Y, Z) =
0 with Z 6= 0. Z is not a common factor of F and G, therefore at least one of
the curves F (X, Y, Z) = 0 and G(X, Y, Z) = 0 only intersects the line Z = 0
in finitely many points. Therefore the total number of intersection points of
F (X, Y, Z) = 0 and G(X, Y, Z) = 0 is finite.
Now lets change co-ordinates in such a way that in the new homogeneous
co-ordinates (X 0 : Y 0 : Z 0 ) none of the intersection points lie on the line
Z 0 = 0 and after dehomogenisation, the x0 axis is not parallel to any of the
lines connecting two intersection points. Now we do not lose any intersection

64
points by the dehomogenisation and disctinct intersection points have dis-
tinct y 0 co-ordinates, so if we repeat the above argument, we obtain that the
number of intersection points is at most deg F deg G. This argument works
over any field K. In the last step we may need to extend the field to find an
appropriate change of co-ordinates, but a bigger field can only increase the
number of intersection points, so if their number is still at most deg F deg G
after the field extension, the number of intersection points over the original
field K is at most deg F deg G.

Example: By Bezouts Theorem any two curves in P2 meet in at least

one point. This can be used to prove that P2 and the quadric surface
Q = V(hXY ZW i) P3 are not isomorphic. They were shown to be bi-
rationally equivalent in Example 4 of http://www.maths.manchester.ac.
uk/~gm/teaching/MATH32062/projexamples.pdf. However, Q contains the
lines X = Z = 0 and Y = W = 0, which do not intersect, so if Q and P2
were isomorphic, the images of these non-intersecting lines would give curves
in P2 which do not intersect, which is impossible.
In fact, the set of lines {X + W = Z + Y = 0 | ( : ) P1 } is
contained in Q and no two distinct lines in this set intersect. Similarly, no
two elements of the set {X + Z = W + Y = 0 | ( : ) P1 } intersect,
but every element of the first set intersects every element of the second set
in exactly one point.
The first picture below shows the two families of lines on an affine piece
of XY ZW = 0, the second one the families of lines on the projectively
equivalent hyperboloid. No two red lines intersect, similarly no two blue lines
intersect, but every red line meets every blue line.

65
The two pictures below show real life applications of the lines on the hyper-
boloid, the bridge across Corporation Street between the Arndale Centre and
Marks&Spencer and statues in Manlleu, Catalonia.

The youtube video https://www.youtube.com/watch?v=jymZ060T0iI (also

linked to from the Animations and videos section of the course website)
includes other examples of hyperboloids in architecture.
Theorem 5.2 Let P1 , P2 , P3 , P4 and P5 be distinct points of P2 and assume
that no 4 of them are collinear. Then there exists a unique conic (curve
defined by a degree 2 equation) in P2 passing through P1 , P2 , P3 , P4 , P5 .
Proof. First we prove the existence of such a curve. A degree 2 homogeneous
polynomial F K[X, Y, Z] has 6 coefficients. The condition that the curve
defined by F passes through a given point gives a homogeneous linear equa-
tion for the coefficients. Five homogeneous linear equations in 6 unknowns
always have a non-trivial (not all 0) solution, therefore there exists a degree
2 homogeneous polynomial F such that F (Pi ) = 0 for every i, 1 i 5.
F cannot be the square of a polynomial of degree 1, as the points are not
collinear, therefore the curve C = V(hF i) defined by F is indeed a conic.
Now for the proof of uniqueness lets assume that the curve D = V(hGi)
defined by the homogeneous degree 2 polynomial G also passes through P1 ,
P2 , P3 , P4 , P5 . By Bezouts Theorem, this can only happen if F and G have a
common factor of positive degree, otherwise C and D could only have at most
4 points in common. If this common factor has degree 2, then F and G are
scalar multiples of each other, so C = D. The only other possibility is that
F and G have a common factor of degree 1, so F = LM and G = LN , where
L, M and N are homogeneous polynomials of degree 1. The line L = 0 can

66
only contain at most 3 of the points P1 , P2 , P3 , P4 , P5 by the hypothesis of
the theorem, the remaining points, of which there are at least 2, must satisfy
M = 0 and N = 0. Two distinct points in the projective plane determine a
unique line, therefore M = 0 and N = 0 are the same line, so M and N are
scalar multiples of each other and C = D.
Theorem 5.3 (Chasles, CayleyBacharach) Let Pi , 1 i 8, be distinct
points of P2 and assume that no 4 of them lie on a line and no 7 of them lie
on a conic. Then there exists a point Q P2 , Q 6= Pi for 1 i 8 such that
all cubic (degree 3) curves passing through all of the points Pi , 1 i 8,
also pass through Q, or there exists i, 1 i 8, such that all such cubics
have a common tangent line at Pi .
In the picture below the black dots are P1 , P2 , . . . , P8 , the curves are some
of the cubics passing through them and the red dot is Q.

Michel Chasles (17931880) was a French mathematician, he is among the

72 French scientists and engineers whose names are written on the first level
of the Eiffel Tower. Arthur Cayley (18211895) was the most prominent
English pure mathematician of the 19th century, he was the fist Sadleirian
Professor of Pure Mathematics at Cambridge. Isaak Bacharach (18541942)
was a German mathematician. As he was a Jew, the Nazis sent him to the

67
 Theresienstadt (today Terezn, Czech Republic) concentration camp, where
he died.
Proof. Degree 3 homogeneous polynomials in K[X, Y, Z] have 10 coefficients,
therefore the set of these polynomials together with 0 form a 10-dimensional
vector space. Let
W = {F K[X, Y, Z] | F homogeneous, deg F = 3 or F = 0, F (Pi ) = 0, i, 1 i 8}.
The condition that the curve defined by F passes through a given point gives
a homogeneous linear equation for the coefficients of F . As the equations
are linear and homogeneous, W is a vector subspace. Eight homogeneous
linear equations in 10 unknowns always have at least a 2-dimensional space
of solutions, so dim W 2. We shall prove that under the hypotheses of the
theorem, dim W = 2.
Lets assume for the moment that we already know that dim W = 2. Let
F1 , F2 be a basis for W . We shall prove that they have no common factor of
positive degree. Lets assume to contrary that they have a common factor of
positive degree, let H be the greatest common divisor of F1 and F2 , the we
can write F1 = HH1 and F2 = HH2 with H1 , H2 coprime. If H had degree 3,
F1 and F2 would be scalar multiples of each other and they would not form
a basis, so the two possibilities are deg H = 1 and deg H1 = deg H2 = 2, or
deg H = 2 and deg H1 = deg H2 = 1. In the first case the line H = 0 can
contain at most 3 of the points P1 , P2 , . . . , P8 , while the conics H1 = 0 and
H2 = 0 have at most 4 points in common by Bezouts Theorem, so the cubic
curves F1 = 0, F2 = 0 cannot both contain all of P1 , P2 , . . . , P8 . In the
second case the conic H = 0 can contain at most 6 of the points P1 , P2 , . . . ,
P8 , while the lines H1 = 0 and H2 = 0 have one point in common, so the
cubics F1 = 0, F2 = 0 cannot both contain all of P1 , P2 , . . . , P8 . This shows
that F1 and F2 are coprime as claimed.
By Bezouts Theorem, the cubic curves F1 = 0 and F2 = 0 have 9 points of
intersection counted with multiplicity. (Bezouts Theorem only holds over al-
gebraically closed fields, but we can take intersection points over the algebraic
closure of K and because 8 of the 9 points of intersection have co-ordinates
in K, the 9th point must have, too.) They intersect at P1 , P2 , . . . , P8 , so
the possibilities are that the cubics have intersection multiplicity 1 at each
of these points and they also intersect at a 9th point Q, or IPi (F1 , F2 ) = 2
for some i, 1 i 8 and IPj (F1 , F2 ) = 1 for j 6= i. The equation of any
cubic passing through P1 , P2 , . . . , P8 , can be written as F = 1 F1 + 2 F2
for some 1 , 2 K not both 0. In the first case clearly F (Q) = 0 for any
such F , so the curve F = 0 also passes through P9 . In the 2nd case, at
least one of the curves F1 = 0 and F2 = 0 is non-singular at Pi , otherwise

68
the intersection multiplicity would be greater than 2. We can assume that
F1 = 0 is non-singular at Pi , then IPi (F1 , F2 ) = 2 implies that either F2 = 0
is also non-singular and has the same tangent line at Pi or F2 = 0 is singular.
In either case, the tangent line to F1 = 0 at Pi will also be tangent to F = 0
for any F = 1 F1 + 2 F2 . Therefore it is sufficient to prove that dim W = 2.
Lets assume to the contrary that dim W 3. This implies that for any
two points P9 and P10 there exists a homogeneous cubic polynomial G
K[X, Y, Z], G 6= 0 such that G(Pi ) = 0 for all i, 1 i 10.
Lets consider the lines P1 Pi for 2 i 8. As no 4 of the points P1 , P2 , . . . ,
P8 are collinear, each such line contains 1 or 2 of the points P2 , P3 , . . . , P8 .
As there are 7 points, there must be a line P1 Pi with does not contain Pj for
2 j 8, j 6= i, we may assume that P1 P2 is such a line.
Let now P9 and P10 be points on the line P1 P2 , different from P1 and P2 .
By our assumption, there exists a degree 3 homogeneous polynomial G1
K[X, Y, Z] such that G1 (Pi ) = 0 for all i, 1 i 10. Let L = 0, where L is
a homogeneous degree 1 polynomial, be the equation of the line P1 P2 . The
cubic curve G1 = 0 and the line L = 0 have at least 4 points in common,
namely P1 , P2 , P9 and P10 , so by Bezouts Theorem, G1 and L must have a
common factor, but as deg L = 1, this means that L divides G1 , G1 = LM ,
where M is homogeneous of degree 2. We have M (Pi ) = 0 for 3 i 8,
since G1 (Pi ) = 0 but L(Pi ) 6= 0 for 3 i 8.

P10
P9
P2
P1 P3
P4
L=0

P5 P8

M =0 P7

P6

M may be irreducible or it may factorise as M = M1 M2 , where M1 and M2

69
are degree 1 homogeneous polynomials. In the latter case the assumption
that no 4 of the points P1 , P2 , . . . , P8 are collinear implies that lines M1 = 0
and M2 = 0 are different, each one of them must contain exactly 3 of the
points P3 , P4 , . . . , P8 , and the intersection point of M1 = 0 and M2 = 0 is
not among the points P3 , P4 , . . . , P8 . We may assume that P3 , P4 , P5 are
on the line M1 = 0 and P6 , P7 , P8 are on the line M2 = 0.

P12 P12
P2 P2
P1 P3 P1
P4
N =0 P11 N =0
P3
P4
P5 P8 M1 =0 P5
P11
M =0 P7

M2 =0 P8
P7
P6 P6

If M is irreducible, let P11 be an arbitrary point of the conic M = 0 other

than P3 , P4 , . . . , P8 . If M is reducible, let P11 be the intersection point of
the lines M1 = 0 and M2 = 0. In both cases let P12 be a point such that
G1 (P12 ) 6= 0.
Now there exists a degree 3 homogeneous polynomial G2 K[X, Y, Z] such
that G2 (Pi ) = 0 for 1 i 8, i = 11 and i = 12.
If M is irreducible, then the cubic G2 = 0 and the conic M = 0 have at
least 7 points in common, namely P3 , P4 , . . . , P8 and P11 , so by Bezouts
Theorem, G2 and M must have a common factor. As M is irreducible, this
means that M divides G2 , G2 = M N , where N is a homogeneous polynomial
of degree 1.
If M is reducible, then the cubic curve G2 = 0 and the line M1 = 0 have
at least 4 points in common, namely P3 , P4 , P5 and P11 , so by Bezouts
Theorem, G2 and M1 must have a common factor. As deg M1 = 1, this
means that M1 divides G2 . By a similar argument M2 also divides G2 . M1
and M2 are coprime, therefore M = M1 M2 also divides G2 , so we can again
write G2 as G2 = M N , where N is a homogeneous polynomial of degree 1.
Now G2 (Pi ) = 0 and M (Pi ) 6= 0 for i {1, 2, 12}, therefore N (Pi ) = 0 for
i {1, 2, 12}. This implies that N = 0 is the line P1 P2 , however, P12 was

70
chosen such that it does not lie on the line P1 P2 . This is a contradiction,
therefore the assumption dim W 3 must be false. This proves the theorem.

The rest of this chapter is not examinable, but it contains some beautiful
geometry.
Algebraic geometry can also be used to prove many of the classical theorems
of projective geometry.
Theorem 5.4 (Pascal) Let C be an irreducible conic in P2 and let A1 , A2 ,
A3 , B1 , B2 , B3 be 6 distinct points of C. Let P1 be the intersection point of
the lines A2 B3 and A3 B2 , P2 that of A1 B3 and A3 B1 , P3 that of A1 B2 and
A2 B1 . Then P1 , P2 and P3 are collinear.

A2
A1

P3
A3 P2
P1 B1
B2

B3

Proof. Let Q be a homogeneous degree 2 polynomial defining C (I(C) =

hQi). Let L1 = 0 be the equation of the line A2 B3 , L2 = 0 the equation of
the line A3 B1 and L3 = 0 the equation of the line A1 B2 . Let M1 = 0 be the
equation of the line A3 B2 , M2 = 0 the equation of the line A1 B3 and M3 = 0
the equation of the line A2 B1 . Let F = L1 L2 L3 and let G = M1 M2 M3 . Let
R be a point of C, different from A1 , A2 , A3 , B1 , B2 , B3 .
Let H = F (R)G G(R)F . Then H 6= 0, since F and G are not scalar multi-
ples of each other because they define different curves. H(Ai ) = H(Bi ) = 0
for i = 1, 2, 3, since F (Ai ) = F (Bi ) = G(Ai ) = G(Bi ) = 0 for i = 1, 2, 3, but
H(R) = 0, too. This means that the cubic H = 0 and the conic C defined

71
by Q = 0 have at least 7 points in common, A1 , A2 , A3 , B1 , B2 , B3 and P .
By Bezouts Theorem, this is only possible if they have a common factor, as
Q is irreducible, the common factor must be Q, so H = QN , where N has
degree 1.
H(Pi ) = 0 for i = 1, 2, 3, since F (Pi ) = G(Pi ) = 0 for i = 1, 2, 3. Pi / C
because A1 , A2 , A3 , B1 , B2 , B3 are distinct, therefore Q(Pi ) 6= 0 for i = 1, 2, 3,
which implies N (Pi ) = 0 for i = 1, 2, 3. Therefore P1 , P2 and P3 all lie on
the line N = 0.
The analogous statement for reducible conics consisting of two lines is known
as Pappuss theorem, illustrated on the diagram below.

B3

B2

B1 P1
P
P3 2

A1
A2

A3

It can be proved similarly to Pascals Theorem, just take the point R in the
proof to be the intersection point of the two lines.
The converse of Pascals theorem is also true, given 6 distinct points A1 , A2 ,
A3 , B1 , B2 , B3 in P2 , if the points P1 , P2 and P3 constructed as in Pascals
theorem are collinear, then there exists a possibly reducible conics which
contains A1 , A2 , A3 , B1 , B2 , B3 . This is called the BraikenridgeMaclaurin
theorem.

72
Wikipedia has useful articles on Pascals and Pappuss theorems: http://
en.wikipedia.org/wiki/Pascals_theorem and http://en.wikipedia.
org/wiki/Pappuss_hexagon_theorem
There is also an interesting duality between points and lines in the projective
plane. Associating the line AX +BY +CZ = 0 to the point (A : B : C) gives
a bijection between lines and points, with collinear points corresponding to
lines passing through a point and vice versa. Theorems and their proofs
can be dualised by replacing a line through given points by the intersection
point of the corresponding lines and vice versa. The duals of points on an
irreducible conics are tangent to another conic. The dual of Pascals Theorem
is called Brianchons Theorem, stated below.
Let A1 , A2 , A3 , B1 , B2 , B3 be distinct points of an irreducible conic C in P2 .
Construct the line connecting the intersection point of the tangent lines to
C at Ai and at Bj and the intersection point of the tangent lines to C at Ai
and at Bj for (i, j) = (1, 2), (1, 3), (2, 3). These three lines meet at a point P .

A1
B3
B2

A2 A3

B1

The dual of Pappuss theorem is the following, it does not appear to have a
special name.

73
 l1 l2 l3
m1

m2
B m3

Let A and B be distinct points in P2 , let l1 , l2 , l3 be three distinct lines

passing through A and let m1 , m2 , m3 be three distinct lines passing through
B. Then the line connecting l1 m2 and l2 m1 , the line connecting l1 m3
and l3 m1 and the line connecting l2 m3 and l3 m2 meet at a point P .

74
6 Elliptic curves
Elliptic curves are not ellipses. The name comes from the elliptic functions
arising from the integrals used to calculate the arc length of ellipses. Elliptic
curves can be parametrised by elliptic functions in a similar way as circles
can be parametrised by sine and cosine.
Elliptic curves have the very special property that their points also have
a natural commutative group structure. Elliptic curves have been studied
intensively in number theory, they played a crucial part in Wiless proof
of Fermats Last Theorem. Elliptic curves over finite fields have a great
practical importance, too, their groups of points can be used for a public key
cryptography algorithm similar to RSA, but as the group structure is more
complicated, smaller primes can be used to achieve the same level of security.
Definition. An elliptic curve is a projective variety isomorphic to a non-
singular curve of degree 3 in P2 together with a distinguished point O E.
Definition. (Chord and tangent process) Let E be a non-singular curve of
degree 3 in P2 and fix O E. For A, B E define A + B E as follows:
let Q be the third intersection point of the line AB with E, and then A + B
is the third intersection point of the line OQ with E.

Q=A+B
Q

If A = B or O = Q, then the line AB or the line OQ is taken to be the

tangent line at A or O, resp. If any line in this construction is tangent to E,
the third intersection point is defined using intersection multiplicities.

75
The picture below shows the calculation of 2A. We take the tangent line at
A, let Q be the third intersection point of this tangent line with E, and then
2A is the third intersection point of the line OQ with E.

Q=2A Q

To calculate the negative of a point in this group structure, first take the
tangent line at O and let M be its 3rd point of intersection with E. To find
A, just take the line through A and M , its 3rd point of intersection with
E is A.

O
-A
A M

76
Worked examples of calculations on elliptic curves can be found in the hand-
out http://www.maths.manchester.ac.uk/~gm/teaching/MATH32062/Ellipticexample.
pdf. Although elliptic curves are projective, calculations are usually carried
out in affine coordinates.
If the affine equation has the special form y 2 = x3 + ax2 + bx + c, the
so-called Weierstra form, then its projective closure has equation Y 2 Z =
X 3 +aX 2 Z +bxZ 2 +Z 3 . The only point of intersection with the line Z = 0 is
(0 : 1 : 0), often called simply the point at . By convention, if the equation
is in this form, O is taken to be (0 : 1 : 0) even if it is not mentioned explicitly.
If a, b, c R, then x3 + ax2 + bx + c = 0 can have 1 or 3 real roots, in the
first case the elliptic curve y 2 = x3 + ax2 + bx + c has one real component
like the curve below on the left, in the second case the elliptic curve has two
real components like the curve below on the right.

y y

4 4

2 2

x x
-2 -1 1 2 3 4 -2 -1 1 2 3

-2 -2

-4 -4

The lines passing through (0 : 1 : 0) are the lines parallel to the y-axis in affine
terms. If the point Q in the addition process has co-ordinates (u, v), the line
through O and Q is the line x = u, and since as the curve is symmetric about
the x-axis, its 3rd intersection point with E is (u, v) = A + B. The line
Z = 0 and E have intersection multiplicity 3 at (0 : 1 : 0), therefore the point
M in the construction of A is also O, so if A = (s, t), then A = (s, t).
The calculations are much simpler if the equation is in Weierstra form, one
of our goals will be to transform elliptic curves into this form.
The diagrams below illustrate the addition of two points, doubling a point
and taking the negative of a point on an elliptic curve is Weierstra form.

77
 y y y

4 4 4

Q Q=2A
2
B A 2
A 2 A

x x x
-2 -1 1 2 3 4 -2 -1 1 2 3 4 -2 -1 1 2 3 4

-A
-2 -2 -2
Q
Q=A+B
-4 -4 -4

Theorem 6.1 The addition defined above makes the points of E into an
abelian group with O as the identity element.
Proof. Lets assume that K is algebraically closed, otherwise we can replace
it by its algebraic closure.
It follows directly from the addition process that O +A = A for every A E,
A + B = B + A for every A, B E and that the point A constructed
above satisfies A + (A) = 0. The substantial part of the proof is to prove
associativity.
Let A, B, C E. The diagram below shows the calculation of (A + B) + C
and A + (B + C), except for the last step. S is the 3rd intersection point
of the line C Q with E, so (A + B) + C is the 3rd intersection point of the
line OS with E. S 0 is the 3rd intersection point of the line AR with E, so
A + (B + C) is the 3rd intersection point of the line OS 0 with E. Therefore
in order to prove the equality (A + B) + C = A + (B + C) it is sufficient to
show that S = S 0 . Lets assume that O, A, B, C, Q, Q, R, R are all distinct
and they are not equal to S or S 0 .

78
 C
O
Q=A+B

R Q
A B R=B+C

S=S'

Let L1 = 0 be the equation of the line ABQ, let L2 = 0 be the equation of

the line ORR and let L3 = 0 be the equation of the line C QS. Let F be the
cubic curve defined by L1 L2 L3 = 0, consisting of the union of the green lines
in the diagram.
Let M1 = 0 be the equation of the line BCR, let M2 = 0 be the equation of
the line OQQ and let M3 = 0 be the equation of the line ARS 0 . Let G be
the cubic curve defined by M1 M2 M3 = 0, consisting of the union of the red
lines in the diagram.
E F {O, A, B, C, Q, Q, R, R, S}. As E is irreducible, it has no common
component with F , therefore the intersection must consist of exactly these
9 points by Bezouts Theorem. Then by Theorem 5.3, any cubic curve that
passes through O, A, B, C, Q, Q, R, R also passes through S.
Similarly, E G = {O, A, B, C, Q, Q, R, R, S 0 }. Since G contains O, A, B,
C, Q, Q, R, and R, as we noted above it must also contain S. Since S is
different from the other 8 points, this implies S = S 0 as required.
Now lets consider the assumption that O, A, B, C, Q, Q, R, R are all
distinct and they are not equal to S or S 0 . This implies A 6= O, then given
A, the points O, A, B, Q and Q will be all distinct for all but finitely
many choices of B. Similarly, for given A and B, the points O, A, B,
C, Q, Q, R, R will be all distinct and not equal to S or S 0 for all but
finitely many choices of C. Therefore (A + B) + C = A + (B + C) holds
for (A, B, C) in a non-empty subset of E E E. Over R or C we can use
continuity to prove (A + B) + C = A + (B + C) for all A, B and C. Over

79
any field, we can make E E E into a variety (this is not as simple as
in the affine case because Pm Pn 6= Pm+n , E E E will be a projective
variety in P26 ), and then we have two morphisms 1 , 2 : E E E E,
1 (A, B, C) = (A + B) + C, 2 (A, B, C) = A + (B + C), which agree outside
a proper subvariety of E E E E. As E E E E is irreducible,
this implies 1 (A, B, C) = 2 (A, B, C) for all A, B, C.
We noted earlier that if the equation of the elliptic curve is of the form
Y 2 Z = X 3 + aX 2 Z + bxZ 2 + Z 3 , then the line Z = 0 only intersects the
curve at (0 : 1 : 0), so by Bezouts Theorem, the intersection multiplicity at
that point must be 3. This implies that the tangent line to the elliptic curve
at (0 : 1 : 0) is the line Z = 0 and furthermore, that (0 : 1 : 0) is a special
point, since in general a curve and the tangent line to it at a point only have
intersection multiplicity 2.
Definition A non-singular point P of a plane curve C (affine or projective)
is called an inflection point if and only if the tangent line to C at P has
intersection multiplicity at least 3 with C at P .
Warning: Inflection points are preserved under affine or projective equiva-
lence, but not necessarily under more general isomorphisms of varieties.
Proposition 6.2 Let E be a non-singular curve of degree 3 in P2 and assume
that O E is an inflection point.
(i) A point P 6= O has order 2 in the group structure on E if and only if the
tangent line at P passes through O.
(ii) A point P 6= O has order 3 in the group structure on E if and only if it
is an inflection point.
Proof. (i) 2P = O is equivalent to P = P . As O is an inflection point,
P is simply the 3rd intersection point of the line OP with E as shown on
the diagram below on the left. It is equal to P if and only if the line OP is
tangent to E at P .

80
 O O

P Q=2P
P

Q=-2P
-P

(ii) 3P = O is equivalent to P = 2P . Let Q be the 3rd intersection point of

E with the tangent line to E at P . Then 2P = Q is the 3rd intersection point
of the line OQ with E as shown on the diagram above on the right. By the
method of calculating the negative of a point, Q = Q = 2P . P = 2P if
and only if P = Q, i. e., if and only if P is an inflection point.
In particular, (i) implies that if the equation of the curve is in the form
y 2 = x3 + ax2 + bx + c, the points of order 2 are the points where the tangent
line is parallel to the y-axis, which are the points whose y co-ordinate 0 and
whose x co-ordinate is one of the roots of x3 + ax2 + bx + c.
Example: Find the points of order 2 on the elliptic curve with affine equation
y 2 = x3 2x 4 over C.
The points of order 2 are the points of the form (, 0), where is a root
x3 2x 4. x3 2x 4 = (x 2)(x2 + 2x + 2), so the roots are 2 and 1 i,
hence the points of order 2 are (2, 0), (1 + i, 0) and (1 i, 0).
It follows from part (ii) of the above proposition that the 3rd point of inter-
section of a line through two inflection points of an elliptic curve is also an
inflection point.
Definition. Let f (x, y) K[x, y]. The Hessian of f is the determinant
 
fxx fxy fx 
 
fxy fyy fy  ,
 
 fx fy 0 
where fx denotes the partial derivative of f with respect to x, etc.
Let F (X, Y, Z) K[X, Y, Z] be a homogeneous polynomial. The Hessian of
F is the determinant  
FXX FXY FXZ 

 FXY FY Y FY Z  .

 FXZ FY Z FZZ 

81
Proposition 6.3 Let C be a curve in A2 or in P2 , defined by an irreducible
polynomial f (x, y) K[x, y] or by an irreducible homogeneous polynomial
F (X, Y, Z) K[X, Y, Z]. Let H be the Hessian of f or F and let D be the
curve H = 0. In the affine case the inflection points of C are the elements
of C D which are non-singular points of C, in the projective case the
same holds under the additional assumption that if the field K has finite
characteristic, the characteristic does not divide deg F 1.
Proof. In the affine case the slope of the tangent line to the curve f (x, y) = 0
is fx /fy by implicit differentiation. Inflection points are critical points of
fx /fy on C, which occur where the gradients of fx /fy and f are parallel,
including the possibility that the former is (0, 0).
 
fxx fy fxy fx fxy fy fyy fx
(fx /fy ) = , ,
fx2 fx2
so the condition for parallelarity with f = (fx , fy ) is

fx (fxy fy fyy fx ) fy (fxx fy fxy fx ) = 0.

The expression on the left-hand side is exactly the Hessian of f . A point

of C is singular if and only if fx = fy = 0 at that point, if this holds, then
H = 0 at that point automatically, but if fx , fy are not both 0, i. e., the
point is non-singular, then the above equation implies that (fx /fy ) is a
scalar multiple of f , so the slope of the tangent line has a critical point
there and the point is an inflection point.
In the projective case let f K[x, y], f (x, y) = F (x, y, 1) be the deho-
mogenisation of F with respect to Z, then F (X, Y, Z) = Z d f (X/Z, Y /Z),
where d = deg F . By using the chain rule and the product rule, we can
express the partial derivatives of F in terms of those of f and after some
calculation it turns out that
   
FXX FXY FXZ  fxx fxy fx  !
  2 3(d2)
  d 2
 FXY FY Y FY Z  = (d1) Z fxy fyy fy + (fxx fyy (fxy ) )f .
 fx fy 0  d 1
   
 FXZ FY Z FZZ 

Therefore the intersection points of C and D which do not lie on the line
Z = 0 correspond to the intersection points of the affine curve f = 0 and
the curve defined by the Hessian of f , so if they are non-singular points of
C, they are inflection points. The definition of the Hessian for homogeneous
polynomials is symmetric in X, Y and Z, therefore we can also use this
argument with X and Y to deduce that all elements of C D are inflection
points of C if they are non-singular points of C.

82
The diagram below shows a cubic curve (blue, thicker), its Hessian (red,
thinner), the inflection points, and the tangent lines to the original cubic
curve at the inflection points.

There are some special phenomena for cubic curves, which do not happen for
curves of higher degree. The tangent lines are also tangent to the Hessian
curve, and it is also true that the inflection points of the Hessian curve are
the same as those of the original curve, although this cannot be seen in this
diagram.
As the projective Hessian has degree 3(d 2), the number of inflection points
is at most 3d(d 2) and for general curves over an algebraically closed field
this number is achieved. If d = 3, the number of inflection points is 9 (if
the characteristic of the field is not 3). If the coefficients of the elliptic curve
E are real, then at least one inflection point has to be real, since non-real
complex solutions come in conjugate pairs. In fact, an elliptic curve over R
always has 3 real inflection points.
Example: Find the real inflection points of the curve y 2 = x3 + 4x2 + 3x 1.
Let f (x, y) = x3 + 4x2 + 3x 1 y 2 . Its Hessian is
 2

 6x + 8 0 3x + 8x + 3 
 
H =  0 2 2y  = 2(3x2 + 8x + 3)2 4y 2 (6x + 8).

3x2 + 8x + 3 2y 0 

To find the solutions f = H = 0, we consider

H 4(6x + 8)f = 2(3x2 + 8x + 3)2 4(6x + 8)(x3 + 4x2 + 3x 1)
= 6x4 32x3 36x2 + 24x + 50,

83
which only involves x. x = 1 is a root, the corresponding values of y are
y = 7. These are the real inflection points, together with the point at
infinity, (0 : 1 : 0). 6x4 32x3 36x2 + 24x + 50 = 0 has another real root
x 3.473, but the corresponding values of y are imaginary.
The diagram below show this elliptic curve, its Hessian curve, the two inflec-
tion points and the tangent lines there.
y

x
-3 -2 -1 1 2

-2

-4

Since if O is an inflection point, then the inflection points are exactly the
points P satisfying 3P = 0, so over an algebraically closed field of character-
istic other than 3, there are 9 such points. From Proposition 6.2 (i) and the
remarks after the proof it follows that if the equation of the elliptic curve is
in Weierstra form and the field K is algebraically closed, then then there
are 4 points satisfying 2P = O, O itself and the 3 points of order 2.
These are special cases of a more general phenomenon. The number of points
P on an elliptic curve such that nP = O is n2 if the field K is algebraically
closed and its characteristic does not divide n, in particular this holds over
C. If the characteristic of K is a prime p, then the number of points such
that pk P = 0 may be pk or just 1, in the latter case the curve is called
supersingular (not related to the definition of singular points, the elliptic
curve is always a non-singular variety).

84
Examples:
1. Let K be an algebraically closed field of characteristic 2. The elliptic curve
defined by the affine equation y 2 + y = x3 + ax2 + bx + c and O = (0 : 1 : 0)
is supersingular for any a, b, c K. It has no point of order 2, because
any line passing through O has affine equation x = for some K.
y 2 + y = 3 + a2 + b + c has two distinct solutions for any K, they
are of the form , + 1 for some K, so the line x = is never tangent
to the curve y 2 + y = y 2 + y = x3 + ax2 + bx + c.
2. Let K be an algebraically closed field of characteristic 3. The elliptic
curve defined by the affine equation y 2 = x3 + bx + c and O = (0 : 1 : 0)
is supersingular for any b K \ {0}, c K. In this case the Hessian turns
out to be 2b2 , a non-0 constant, so there are no inflection points other than
O = (0 : 1 : 0) and there are no points of order 3.
Theorem 6.4 If K is algebraically closed and its characteristic is not 2 or
3, any non-singular cubic curve in P2 is projectively equivalent to one with
an equation of the form Y 2 Z = X 3 + pXZ 2 + qZ 3 (y 2 = x3 + px + q in affine
form).
Proof. Step 1. Choose an inflection point of E and then do a projective
transformation that maps that inflection point to (0 : 1 : 0) and the tangent
line at the inflection point to the line Z = 0. This step eliminates the Y 3 ,
XY 2 and X 2 Y terms from the equation.
After this step the equation can be converted to affine form and the remaining
steps can be carried out in affine form, if preferred.
Step 2. Write the equation with the terms containing Y on one side and the
other terms, only containing X and Z on the other side. Let and be the
coefficients of Y 2 Z and X 3 , resp. Multiply the whole equation by 2 /3 and
then use X/ and Y / as new variables, this will make the coefficients
of Y 2 Z and X 3 equal to 1. (In the affine form the equation will look like
y 2 + a1 xy + a3 y = x3 + a2 x2 + a4 x + a6 , this is often called a Weierstra form,
too.)
Step 3. Complete the square with respect to Y to eliminate XY Z and
Y Z 2 terms (xy and y in the affine form). (This is where we need that the
characteristic is not 2.)
Step 4. Complete the cube with respect to X to eliminate the X 2 Z term (x2
in the affine form). (This is where we need that the characteristic is not 3.)

A worked example of this procedure form can be found at http://www.

85
maths.manchester.ac.uk/~gm/teaching/MATH32062/Ellipticequation.pdf.
As we noted previously, a cubic curve defined by an equation with real coef-
ficients always has real inflection points, so the 1st step can be carried out
over R and we obtain an equation with p, q real. However, if the coefficients
are rational, there is no guarantee that there exists an inflection point with
rational coefficients, so p, q may not be rational.
Remark. p and q are not unique, we can multiply the equation y 2 = x3 +px+q
by 6 for some K \ {0} and then we can rewrite it as (3 y)2 = (2 x)3 +
(4 p)(2 x) + 6 q, so the parameters 4 p and 6 q determine an isomorphic
curve.
4p3
Definition. The j-invariant of the curve y 2 = x3 +px+q is j = 1728 .
4p3 + 27q 2
The rest of this chapter is not examinable.
Another common form of the equation of the elliptic curve is the Legendre
form y 2 = x(x 1)(x ), where 6= 0, 1. j can be expressed in terms of
as
(2 + 1)3
j = 256 2 . (*)
( 1)2
is not uniquely defined, depending on which two roots of the cubic are
chosen to be mapped to 0 and 1, it can take 6 values, , 1 , 1/, 1/(1 ),
/( 1) and ( 1)/, but they all give the same value of j.
It is not clear from either of these definitions that j is really an invariant of the
curve E, since there are choices in various steps of the process of transforming
the equation to the Weierstra or Legendre form. If E is a non-singular curve
in P2 , then from any point E one can draw four tangent lines to E. Let
be the cross ratio of the slopes of these lines, then the formula (*) gives the
j-invariant. This implies that the j invariant is invariant under projective
transformations. It requires some more sophisticated tools to show that if
two cubic curves in P2 are isomorphic, then they are projectively equivalent.
Theorem 6.5 j is indeed an invariant of the elliptic curve, i. e., all possible
Weierstra and Legendre forms of the same curve give the same value for j.
If the field K is algebraically closed, two elliptic curves over K are isomorphic
if and only if they have the same j-invariant.
Non-singular cubic curves with different j invariant are not just not isomor-
phic, but they are not birationally equivalent either. Non-singular cubic
curves are not rational, i. e., they are not birationally equivalent to P1 .
The study of points with rational co-ordinates is a very active area of number

86
theoretical research. Mordells Theorem states that the group of rational
points is finitely generated, i. e., as an abstract group it is isomorphic to
T Zr , where T is a finite group and r 0 is a non-negative integer, called
the rank of the elliptic curve. Mazurs Theorem states that the group T has
at most 12 elements. It is not known whether r can be arbitrarily large, but
there exists an example with r 28.

87