MA40188 ALGEBRAIC CURVES

2015/16 SEMESTER 1

ZIYU ZHANG

Abstract. This document contains the material for a 10-week course on algebraic
curves taught at the University of Bath in the first semester of the academic year
2015/16. The audience consists mostly of 3rd/4th year undergraduate students with
major in mathematics. The approach taken in this course is purely algebraic, therefore
assumes no prior knowledge of complex analysis. However, a solid background on ring
and ideal theory, as covered in Algebra 2B, is necessary. Topics discussed in this course
include affine and projective algebraic sets, along with lots of examples, including pro-
jective curves and surfaces of degree up to 3. This document contains all lecture notes,
exercise sheets and their complete solutions. There are 20 lectures and 10 exercise sheets
in total. Each subsection is covered in precisely one lecture of 50 minutes.

Contents

1. Affine Algebraic Sets 4

1.1. Affine spaces and affine algebraic sets 4

1.2. Noetherian rings and Hilbert basis theorem 7

Exercise Sheet 1 10

Solutions to Exercise Sheet 1 11

2. Nullstellensatz 13

2.1. Nullstellensatz and V − I correspondence 13

2.2. Prime ideals and maximal ideals 16

Exercise Sheet 2 19

Solutions to Exercise Sheet 2 20

3. Coordinate Rings 24

3.1. Coordinate rings and polynomial maps 24

Date: January 17, 2016.

1
3.2. Homomophisms of coordinate rings 27

Exercise Sheet 3 30

Solutions to Exercise Sheet 3 31

4. Projective Algebraic Sets 35

4.1. Projective spaces 35

4.2. Projective algebraic sets and projective Nullstellensatz 38

Exercise Sheet 4 41

Solutions to Exercise Sheet 4 42

5. Rational Maps 46

5.1. V − I correspondence and rational maps 46

5.2. Dominant rational maps and birational maps 49

Exercise Sheet 5 52

Solutions to Exercise Sheet 5 53

6. Function Fields 58

6.1. Bridge between affine and projective algebraic sets 58

6.2. Rational functions and function fields 61

Exercise Sheet 6 64

Solutions to Exercise Sheet 6 65

7. Non-singularity 68

7.1. Non-singularity of irreducible hypersurfaces 68

7.2. Non-singularity of varieties 71

Exercise Sheet 7 74

Solutions to Exercise Sheet 7 75

8. Algebraic Curves 80

8.1. Lines and conics 80

8.2. Cubics 83

Exercise Sheet 8 86
2
Solutions to Exercise Sheet 8 87

9. Elliptic Curves 91

9.1. The group law on non-singular cubics 91

9.2. Linear systems and associativity 94

Exercise Sheet 9 97

Solutions to Exercise Sheet 9 98

10. Algebraic Surfaces 102

10.1. Planes and quadric surfaces 102

10.2. Non-singular cubic surfaces 105

Exercise Sheet 10 108

Solutions to Exercise Sheet 10 109

Appendix A. Brief Review of Algebra 2B 112

Acknowledgements 115

References 115

3
 1. Affine Algebraic Sets

We introduce affine spaces and define an affine algebraic set as the common zeroes of a set
of polynomials. We study some basic properties of algebraic sets, and use the Hilbert basis
theorem to show that every algebraic set is the intersection of finitely many hypersurfaces.

1.1. Affine spaces and affine algebraic sets. In the entire course, a ring always means
a commutative ring with a multiplicative identity 1, and a field always means an alge-
braically closed field of characteristic 0, unless otherwise specified. Here a field k is
algebraically closed if every non-constant polynomial f (x) ∈ k[x] has a root in k. For
example, C is an algebraically closed field of characteristic 0, but R is not algebraically
closed. Although many theorems can be generalised to other fields, their statements are
often simpler with these extra assumptions on the underlying field.

Definition 1.1. Let k be a field, n ∈ Z+ . An n-dimensional affine space over k is the set
{(a1 , · · · , an ) | a1 , · · · , an ∈ k}.
denoted by Ank (or simply An if the field is understood in the context).

This notion is actually quite familiar. It is simply the set kn of n-tuples of elements in
k. However, we do not use the notation kn in algebraic geometry because we are not just
interested in its structure as a vector space. Indeed, the geometric objects that we will
study are some subsets of affine spaces. More precisely,

Definition 1.2. A subset X ⊆ Ank is called an affine algebraic set (or simply algebraic
set) if there is a set S of polynomials in k[x1 , · · · , xn ], such that
X = {(a1 , · · · , an ) ∈ Ank | f (a1 , · · · , an ) = 0 for all f ∈ S}.
In such a case we say X is the algebraic set defined by S and write X = V(S).

In this definition S could have finitely many or infinitely many elements. If S con-
tains only finitely many polynomials, say, S = {f1 , f2 , · · · , fr }, we usually write X =
V(f1 , f2 , · · · , fr ) instead of X = V({f1 , f2 , · · · , fr }) for simplicity. In particular we have

Definition 1.3. An algebraic set X ⊆ Ank is called a hypersurface if X = V(f ) for some
non-constant polynomial f ∈ k[x1 , · · · , xn ].

Example 1.4. Consider subsets of A1 . The set X1 = {5} is an algebraic set because
X1 = V(x−5). One can also say X1 = V((x−5)2 ), or even X1 = V(x(x−5), (x−1)(x−5)).
We see that different choices of S in Definition 1.2 could possibly define the same algebraic
set X. The set X2 = {5, 7} is an algebraic set because X2 = V((x − 5)(x − 7)). Many
other subsets of A1 are also algebraic sets. You will find all of them in an exercise.
4
Example 1.5. Consider subsets of A2 . Examples of algebraic sets are V(y − x2 ) which
is a parabola, and V(xy) which is the union of two coordinate axes. They are both
hypersurfaces in A2 . The algebraic set V(x − 5, y − 7) contains only one point. It is not a
hypersurface because we cannot define it by one non-constant polynomial (but we do not
prove this fact).

Example 1.6. Let k = Q (it is not algebraically closed but I just want to mention this
piece of history) and n = 2. For every m > 3, the set X = V(xm + y m − 1) ∈ A2Q is a
historically important algebraic set. Obviously X contains points (1, 0) and (0, 1) for all
m, and (−1, 0) and (0, −1) for even m. The fact that these are the only points in X is
one of the deepest results in mathematics. An equivalent formulation of this result is the
so-called Fermat’s Last Theorem, which was conjectured in 1637, and proved in 1995.

Here are some simple and useful properties of algebraic sets.

Proposition 1.7. We consider subsets in An .

(1) Let S1 and S2 be two sets of polynomials in k[x1 , · · · , xn ]. If S1 ⊇ S2 , then

V(S1 ) ⊆ V(S2 ). In other words, the correspondence V is inclusion-reversing.
(2) ∅ and An are both algebraic sets.
(3) The intersection of any collection of algebraic sets in An is an algebraic set.
(4) The union of finitely many algebraic sets in An is an algebraic set.

Proof. We leave the proof as an exercise. 

We introduce some algebraic language that we need to use later.

Definition 1.8. Let R be a ring (a commutative ring with 1).

(1) For any subset S ⊆ R, the ideal

I = {r1 f1 + · · · + rk fk | k ∈ Z+ ; r1 , · · · , rk ∈ R; f1 , · · · , fk ∈ S}

is called the ideal generated by S. We say S is a set of generators of I.

(2) An ideal I is said to be finitely generated if it is generated by a finite set S =
{f1 , · · · , fm } ⊆ R. We write I = (f1 , · · · , fm ).
(3) An ideal I is principal if it is generated by one element f ∈ R. We write I = (f ).

Notice that the notation in Definition 1.8 is slightly different from, indeed, simpler than
what we used in Algebra 2B (which was I = Rf1 + · · · + Rfm if I is finitely generated, or
I = Rf if I is principal). The notation here is more often used in algebraic geometry.
5
Example 1.9. Let I ⊆ Z be the ideal of all even integers. Then one can say I = (2), or
I = (−2), or I = (2, 4) (4 is obviously redundant), or I = (4, 6) (do you see why?). We
can even take S to be everything in I, then the ideal generated by S is still I. Upshot:
there are usually many choices for the generators of a given ideal.

Lemma 1.10. For any subset S ⊆ k[x1 , · · · , xn ], let I ⊆ k[x1 , · · · , xn ] be the ideal gen-
erated by S. Then V(S) = V(I).

Proof. We need to show mutual inclusions between V(S) and V(I). The inclusion in one
direction V(S) ⊇ V(I) follows from the fact that S ⊆ I and Proposition 1.7 (1).

We prove V(S) ⊆ V(I). For every point p = (a1 , · · · , an ) ∈ V(S), we need to show that
p ∈ V(I). Since I is generated by S, every element g ∈ I can be written in the form
g = r1 f1 + · · · + rk fk for some k ∈ Z+ , r1 , · · · , rk ∈ k[x1 , · · · , xn ] and f1 , · · · , fk ∈ S. By
assumption f1 (p) = · · · = fk (p) = 0, which implies g(p) = r1 (p)f1 (p)+· · ·+rk (p)fk (p) = 0.
Therefore p ∈ V(I). It follows that V(S) ⊆ V(I). 

This lemma shows that every algebraic set X ⊆ An can be defined by an ideal I ⊆
k[x1 , · · · , xn ]. Notice that different ideals could still define the same algebraic set.

Example 1.11. Consider X = {0} ⊆ A1 . Consider two principal ideals I1 = (x) and
I2 = (x2 ) in k[x]. Then X = V(I1 ) = V(I2 ).

Among the many ideals that define the same algebraic set, we will see next week which
one is “the best”. Stay tuned!

6
1.2. Noetherian rings and Hilbert basis theorem. We start with some algebra. But
eventually we will see its geometric applications.

Recall that a ring R is a principal ideal domain (or PID) if every ideal of R is generated
by one element. PIDs have many good properties. But unfortunately many interesting
rings in algebraic geometry, for example, k[x1 , · · · , xn ] when n > 2, are not PIDs. It will
be helpful to generalise the notion of PID to include examples like these.

Definition 1.12. A ring R is Noetherian if every ideal of R is finitely generated.

It is immediately clear from the definition that every PID is Noetherian. We want to see
more examples. A powerful tool to produce such examples is the following

Theorem 1.13 (Hilbert Basis Theorem). If a ring R is Noetherian, then R[x] is also
Noetherian.

Proof. Non-examinable. Interested reader can find the proof in [Section 3.3, Reid, Un-
dergraduate Algebraic Geometry] or [Section 1.4, Fulton, Algebraic Curves]. 

Corollary 1.14. For any field k and n ∈ Z+ , the ring k[x1 , · · · , xn ] is Noetherian.

Proof. We prove by induction on n. When n = 1, we know k[x1 ] is a PID, hence is

Noetherian. Assume Rn = k[x1 , · · · , xn ] is a Noetherian ring. We need to show that
Rn+1 = k[x1 , · · · , xn , xn+1 ] is also Noetherian. Notice that by collecting terms with respect
to the variable xn+1 , every polynomial in Rn+1 can be written as a polynomial in xn+1
with coefficients in Rn . In other words, we have Rn+1 = Rn [xn+1 ]. By Hilbert Basis
Theorem 1.13 and the induction assumption, we conclude that Rn+1 is Noetherian. 

There is yet another powerful tool very useful for producing examples of Noetherian rings.
Before stating it we need to give an equivalent description of a Noetherian ring.

Proposition 1.15. A ring R is Noetherian if and only if the following ascending chain
condition (or ACC) holds: for every ascending chain of ideals in R
I1 ⊆ I2 ⊆ I3 ⊆ · · · ,
there exists a positive integer N such that In = IN for all n > N .

Proof. (This proof is non-examinable and not covered in lectures.)

We first prove that the Noetherian condition implies ACC. Take any ascending chain of
ideals in R, say, I1 ⊆ I2 ⊆ I3 ⊆ · · · . Set I = ∪∞ n=1 In . We claim that I is an ideal in
R. Indeed, for any r ∈ R and a, b ∈ I, assume a ∈ Ii and b ∈ Ij . Then a, b ∈ Imax{i,j} .
It follows that a + b ∈ Imax{i,j} , hence a + b ∈ I. Moreover, ra ∈ Ii hence ra ∈ I. This
concludes that I is an ideal.
7
Since R is Noetherian, I is finitely generated, say, I = (f1 , · · · , fm ). Then each fi is an
element in Ini for some ni . Take N = max{n1 , · · · , nm }. We claim that IN = I. On one
hand fi ∈ Ini ⊆ IN for every i, hence r1 f1 + · · · + rm fm ∈ IN for any r1 , · · · , rm ∈ R,
which implies I ⊆ IN . On the other hand we have IN ⊆ I by the construction of I. It
follows that IN = I. For every n > N , we have IN ⊆ In ⊆ I = IN , hence In = IN .

We then prove that ACC implies the Noetherian condition. We use contradiction. Assume
R has an ideal J which is not finitely generated. We pick an element g1 ∈ J and define
I1 = (g1 ). Since J is not finitely generated we have I1 ( J, hence we can pick an element
g2 ∈ J\I1 and define I2 = (g1 , g2 ). Similarly we can pick g3 ∈ J\I2 and define I3 =
(g1 , g2 , g3 ). Repeat this process indefinitely, we get a chain of ideals I1 ( I2 ( I3 ( · · ·
where each Ii = (g1 , · · · , gi ). Every inclusion in the chain is strict, hence the chain never
stabilises, which is a contradiction to ACC. 

Now we are ready to state our second tool for producing examples of Noetherian rings.

Proposition 1.16. Let R be a Noetherian ring and I is an ideal in R. Then the quotient
ring R/I is also Noetherian.

Proof. We leave the proof as an exercise. 

Corollary 1.17. For any ideal I in k[x1 , · · · , xn ], k[x1 , · · · , xn ]/I is a Noetherian ring.

Proof. This is a consequence of Corollary 1.14 and Proposition 1.16. 

Why are we so interested in Noetherian rings? Can we understand more geometry from
the fact that k[x1 , · · · , xn ] is Noetherian? The following is the answer.

Theorem 1.18. Let X ⊆ An be an algebraic set, such that ∅ 6= X 6= An . Then X is the

intersection of finitely many hypersurfaces.

Proof. By Lemma 1.10, we can write X = V(I) for some ideal I in k[x1 , · · · , xn ]. By
Corollary 1.14, I is finitely generated, say, I = (f1 , · · · , fm ). By Lemma 1.10 again we
can write X = V(I) = V(f1 , · · · , fm ). Without loss of generality, we can assume every fi
is non-constant. Indeed, if a certain fi is zero, then we can simply remove it from the set
of generators; if a certain fi is a non-zero constant, then X = ∅ which is excluded by the
assumption. Notice that
X = V(f1 , · · · , fm )
= {p ∈ An | f1 (p) = · · · = fm (p) = 0}
= {p ∈ An | f1 (p) = 0} ∩ · · · ∩ {p ∈ An | fm (p) = 0}
= V(f1 ) ∩ · · · ∩ V(fm ).
8
Since each V(fi ) is a hypersurface in An , we conclude that X is the intersection of finitely
many hypersurfaces. 

Equivalently, we can say that every algebraic set in An can be defined by finitely many
polynomials (this even includes the algebraic sets ∅ and An , as they are defined by {1} and
{0} respectively). Notice that a geometric result like Theorem 1.18 cannot be obtained
without the algebraic theory of Noetherian rings. In fact, thoroughout this course, we will
always strive to build up a bridge, or a dictionary, between geometry and algebra. How
to translate a geometric question into algebra, and how to interpret an algebraic result in
the geometric language, will always be our main themes in this course.

9
 Exercise Sheet 1

This sheet will be discussed in the exercise class on 9 October. You are welcome to submit
your solutions at the end of the exercise class or anytime earlier.
Exercise 1.1. Examples of algebraic sets. For each of the following X ⊆ A2 , find a set
of polynomials S ⊆ k[x, y] such that X = V(S). You don’t need to justify your answer.

(1) X = {(0, 0), (0, 1), (1, 0), (1, 1)}.

(2) X = {(0, 0), (1, 1)}.
(3) X is the union of the x-axis and a single point (0, 1).
(4) For fun: describe the algebraic set V(xy, yz, zx) ⊆ A3 geometrically.
Exercise 1.2. Prove Proposition 1.7. Consider algebraic sets in An .

(1) Suppose S1 ⊇ S2 . Prove that V(S1 ) ⊆ V(S2 ).

(2) Prove that ∅ and An are algebraic sets in An .
(3) Prove that ∩α (V(Sα )) = V(∪α Sα ).
(4) Suppose S = {f g | f ∈ S1 , g ∈ S2 }. Prove that V(S1 ) ∪ V(S2 ) = V(S). Use induc-
tion to conclude that the union of finitely many algebraic sets is still algebraic.
Exercise 1.3. Examples of algebraic sets. Prove that algebraic sets in A1 are just the
finite subsets in A1 (including ∅) together with A1 itself. You can follow these steps:

(1) Verify that they are indeed algebraic sets.

(2) Prove that if an algebraic set in A1 is not A1 itself, then it contains at most
finitely many points. (Hint: you can use the following lemma in algebra: a non-
zero polynomial f (x) ∈ k[x] of degree d has at most d roots.)
(3) As an application of this exercise, give an example of infinitely many algebraic
sets, whose union is not an algebraic set.
Exercise 1.4. Prove Proposition 1.16. Prove that if R is a Noetherian ring, then R/I is
also Noetherian for any ideal I in R. You can follow these steps:

(1) We write the quotient ring homomorphism q : R → R/I (sending each r ∈ R to

the coset r + I). For any ideal J in R/I, prove that q −1 (J) is an ideal in R.
(2) For two ideals J1 ⊆ J2 in R/I, prove that q −1 (J1 ) ⊆ q −1 (J2 ).
(3) Suppose J1 ⊆ J2 ⊆ J3 ⊆ · · · is an ascending chain of ideals in R/I. Use (1), (2)
and the fact that R is Noetherian to show that this chain stabilises.
(4) Use Proposition 1.15 to conclude that R/I is Noetherian.
10
 Solutions to Exercise Sheet 1

Solution 1.1. Examples of algebraic sets. There are many possible answers.

(1) One possible answer is X = V(x(x − 1), y(y − 1)).

(2) One possible answer is X = V(x(y − 1), y(x − 1)).
(3) One possible answer is X = V(xy, y(y − 1)).
(4) This algebraic set is the union of the three coordinate axes. In other words, it is
the set of points (x, y, z) ∈ A3 with at least two zero coordinates.

Solution 1.2. Prove Proposition 1.7.

(1) Given any p ∈ V(S1 ), we have f (p) = 0 for every f ∈ S1 . Since every g ∈ S2 is
also an element in S1 , we have g(p) = 0. Hence p ∈ V(S2 ).
(2) We have that ∅ = V(1) and An = V(0).
(3) We first prove ∩α (V(Sα )) ⊆ V(∪α Sα ). Given any point p ∈ ∩α (V(Sα )), we have
p ∈ V(Sα ) for every α. Then for every f ∈ ∪α Sα , there exists some α0 such that
f ∈ Sα0 , therefore f (p) = 0 since p ∈ V(Sα0 ). This shows that p ∈ V(∪α Sα ).
We then prove ∩α (V(Sα )) ⊇ V(∪α Sα ). Given any point q ∈ V(∪α Sα ), we have
g(p) = 0 for every g ∈ ∪α Sα . In particular, for every α, we have p ∈ V(Sα ).
Therefore p ∈ ∩α (V(Sα )).


(4) We first prove V(S1 ) ∪ V(S2 ) ⊆ V(S). Given any p ∈ V(S1 ), we have f (p) = 0
for every f ∈ S1 . Therefore for every f g ∈ S with f ∈ S1 and g ∈ S2 , (f g)(p) =
f (p)g(p) = 0. Hence p ∈ V(S). This proves V(S1 ) ⊆ V(S). Similarly we have


V(S2 ) ⊆ V(S). Therefore V(S1 ) ∪ V(S2 ) ⊆ V(S).


We then prove V(S1 ) ∪ V(S2 ) ⊇ V(S). For every p ∈ V(S), we need to show
that p ∈ V(S1 ) ∪ V(S2 ). If not, then p ∈ / V(S1 ) and p ∈/ V(S2 ). This means there
exists some f0 ∈ S1 and g0 ∈ S2 , such that f0 (p) 6= 0 and g0 (p) 6= 0. It follows that
(f0 g0 )(p) = f0 (p)g0 (p) 6= 0. Since f0 g0 ∈ S, this implies p ∈
/ V(S). Contradiction.


This proves V(S1 ) ∪ V(S2 ) ⊇ V(S).
We then use induction to prove that V(S1 ) ∪ V(S2 ) ∪ · · · ∪ V(Sn ) is an algebraic
set for every positive integer n. When n = 1, V(S1 ) is by definition an algebraic
set. Assume the statement holds for n = k, then V(S1 ) ∪ V(S2 ) ∪ · · · ∪ V(Sk ) is
an algebraic set, say, V(S 0 ). When n = k + 1, we can write

V(S1 ) ∪ V(S2 ) ∪ · · · ∪ V(Sk ) ∪ V(Sk+1 )

= V(S1 ) ∪ V(S2 ) ∪ · · · ∪ V(Sk ) ∪ V(Sk+1 )
= V(S 0 ) ∪ V(Sk+1 )

which is still an algebraic set by the statement we just proved.

11
Solution 1.3. Examples of algebraic sets.

(1) We know that A1 and ∅ are algebraic sets by Proposition 1.7 (2). For any non-
empty finite subset of A1 , say, X = {c1 , c2 , · · · , ck }, we have X = V((x − c1 )(x −
c2 ) · · · (x − ck )), hence is an algebraic set.
(2) Say X = V(S) is an algebraic set in A1 . If S does not contain any non-zero
polynomial, then X = A1 . Otherwise, there is some f (x) ∈ S which is a non-zero
polynomial. Every point in X must be a root of f (x), hence X is a subset of the
all roots of f (x). Since f (x) has only finitely many roots, X has at most finitely
many elements.
(3) There are many possible counterexamples and here is one of them: for every
positive integer n, let Xn = {n} be a single-point set. Then Xn is an algebraic
set. But their union ∪n Xn is the set of all positive integers, which is an infinite
set, hence is not an algebraic set by part (2).

Solution 1.4. Prove Proposition 1.16.

(1) We check that q −1 (J) = {r ∈ R | r + I ∈ J} is an ideal in R. For any a1 , a2 ∈

q −1 (J), we have a1 + I, a2 + I ∈ J hence (a1 + a2 ) + I = (a1 + I) + (a2 + I) ∈ J,
which implies a1 + a2 ∈ q −1 (J). On the other hand, for any r ∈ R and a ∈ q −1 (J),
we have a + I ∈ J hence ra + I = (r + I)(a + I) ∈ J hence ra ∈ q −1 (J). Therefore
q −1 (J) is an ideal in R.
(2) For every a ∈ q −1 (J1 ), we have a + I ∈ J1 . Since J1 ⊆ J2 , we have a + I ∈ J2 .
Hence a ∈ q −1 (J2 ). This verifies that q −1 (J1 ) ⊆ q −1 (J2 ).
(3) Suppose J1 ⊆ J2 ⊆ J3 ⊆ · · · is an ascending chain of ideals in R/I. Then by parts
(1) and (2) we have q −1 (J1 ) ⊆ q −1 (J2 ) ⊆ q −1 (J3 ) ⊆ · · · is an ascending chain of
ideals in R. Since R is a Noetherian ring, this chain stablises by Proposition 1.15.
That means, there exists some positive integer N , such that q −1 (Ji ) = q −1 (JN )
for every i > N . In other words, q −1 (Ji ) and q −1 (JN ) contain precisely the same
cosets of I in R. Therefore Ji = JN for every i > N .
(4) We showed in part (3) that every ascending chain of ideals in R/I stabilises.
Therefore R/I is a Noetherian ring by Proposition 1.15.

12
 2. Nullstellensatz

We will introduct radical ideals, and use Nullstellensatz to establish the V − I correspon-
dence between radical ideals and algebraic sets. We will also see the geometric meaning
of prime ideals and maximal ideals.

2.1. Nullstellensatz and V − I correspondence. Recall the V map in Definition 1.2.

By Lemma 1.10, it defines a surjective map
V : {ideals in k[x1 , · · · , xn ]} −→ {algebraic sets in An }. (2.1)
However the map is not injective as different ideals could possibly define the same al-
gebraic set. Among all ideals that define the same algebraic set, we want to choose a
“good” one, so that we can establish a one-to-one correspondence between “good” ideals
in k[x1 , · · · , xn ] and algebraic sets in An . We start with some algebra.

Definition 2.1. Let I be an ideal in a ring R. The radical of I is

√
I = {f ∈ R | f n ∈ I for some n ∈ Z+ }.
√
An ideal I is said to be a radical ideal if I = I.
√
Lemma 2.2. Let I be an ideal in a ring R. Then I is an ideal in R containing I.

Proof. We leave it as an exercise. 

This definition does not look very intuitive at a first glance. But it will be clear why we
define it this way after we relate it to some geometry. We give a quick example.

Example 2.3. Consider the ideals I1 = (x) and I2 = (x2 ) in k[x]. It is not difficult to
√ √
find out that I1 = I2 = (x). Therefore I1 is a radical ideal in k[x] while I2 is not. We
leave the details in an exercise.

Definition 2.4. For any subset X ⊆ An ,

I(X) := {f ∈ k[x1 , · · · , xn ] | f (p) = 0 for all p ∈ X}
is called the ideal of X.

In other words, I(X) consists of all polynomials that vanish on X. Notice that this
definition makes sense for any subset X ⊆ An which is not necessarily algebraic.

Example 2.5. For the subset X = {0} ⊆ A1 , I(X) is the set of all f (x) ∈ k[x] such that
f (0) = 0. Therefore I(X) = (x) ⊆ k[x].

Lemma 2.6. The map I has the following properties:

(1) Let X1 and X2 be two subsets of An . If X1 ⊇ X2 , then I(X1 ) ⊆ I(X2 ).

13
 (2) For any subset X ⊆ An , I(X) is a radical ideal in k[x1 , · · · , xn ].

Proof. (1) For any f ∈ I(X1 ), we have that f (p) = 0 for every p ∈ X1 . In particular, since
X1 ⊇ X2 , f (p) = 0 for every p ∈ X2 . Hence f ∈ I(X2 ). It follows that I(X1 ) ⊆ I(X2 ).

(2) We first show I(X) is an ideal. For any f, g ∈ I(X) and r ∈ k[x1 , · · · , xn ], we have
p ∈ X. Therefore
(f + g)(p) = f (p) + g(p) = 0 and (rf )(p) = r(p)f (p) = 0 for all p
f + g, rf ∈ I(X), hence I(X) is an ideal. Then we need to show that I(X) = I(X). We
have that
p
f ∈ I(X) ⇐⇒ ∃ m ∈ Z+ such that f m ∈ I(X)
⇐⇒ ∃ m ∈ Z+ such that f (p)m = 0 for any p ∈ X
⇐⇒ f (p) = 0 for any p ∈ X
⇐⇒ f ∈ I(X).
p
It follows that I(X) = I(X), hence the ideal I(X) is radical. 

We return to the question at the beginning of the section. The V-map (2.1) hits all alge-
braic sets in An , but each algebraic set can be hit by many different ideals. However, the
I-map in Definition 2.4 assigns to each algebraic set in An a radical ideal in k[x1 , · · · , xn ].
Therefore if we only consider the radical ideals, there is hope that the two maps
V
/
{radical ideals I ⊆ k[x1 , · · · , xn ]} o {algebraic sets X ⊆ An } (2.2)
I

are inverse to each other, hence establish a one-to-one correspondence between radical
ideals in k[x1 , · · · , xn ] and algebraic sets in An . This holds as long as k is algebraically
closed. The proof relies on the so-called Nullstellensatz, which is a difficult theorem.
Definition 2.7. An ideal I in a ring R is proper if I 6= R.
Theorem 2.8 (Hilbert’s Nullstellensatz). For any algebraically closed field k,

(1) Let I be any proper ideal in k[x1 , · · · , xn ]. Then V(I) 6= ∅.

√
(2) Let I be any ideal in k[x1 , · · · , xn ]. Then I(V(I)) = I.

Proof. Non-examinable. Interested reader can find the proof in [Section 3.10, Reid, Un-
dergraduate Algebraic Geometry] or [Section 1.7, Fulton, Algebraic Curves]. 
Proposition 2.9. For any algebraically closed field k,

(1) Assume I is a radical ideal in k[x1 , · · · , xn ] and X is an algebraic set in An . Then

X = V(I) if and only if I = I(X).
(2) Assume I1 are I2 radical ideals in k[x1 , · · · , xn ], X1 = V(I1 ) and X2 = V(I2 ).
Then I1 ⊆ I2 (resp. I1 ( I2 ) if and only if X1 ⊇ X2 (resp. X1 ) X2 ).
14
 √
Proof. (1) We prove “=⇒”. By Nullstellensatz 2.8 we have I(X) = I(V(I)) = I =I
since I is a radical ideal.

We prove “⇐=”. The algebraic set X can be written as X = V(J) for some √ ideal
J ⊆ k[x
√ 1 , · · · , xn ]. By Nullstellensatz 2.8, V(I) =
√ V(I(X)) = V(I(V(J))) = V( J).
Since J ⊇ J by Lemma 2.2, we have V(I) = V( J) ⊆ V(J) = X by Proposition 1.7
(1). It remains to show that X ⊆ V(I). For every point p ∈ X, by the definition of V,
we need to show that f (p) = 0 for every f ∈ I. This is clear since I = I(X).

(2) The equivalence “I1 ⊆ I2 ⇐⇒ X1 ⊇ X2 ” follows from Proposition 1.7 (1) and Lemma
2.6 (1). By (1), we see that if one of the inclusions is an equality, then so is the other.
Therefore if one of them is a strict inclusion, then so is the other. 

In other words, Proposition 2.9 shows that V and I induce mutually inverse bijections
between radical ideals in k[x1 , · · · , xn ] and algebraic sets in An . Moreover, the bijection
is inclusion-reversing. Next time we will see how this correspondence relates algebra and
geometry.

15
2.2. Prime ideals and maximal ideals. We have established a one-to-one corrspon-
dence (2.2) between radical ideals in k[x1 , · · · , xn ] and algebraic sets in An . A major
benefit: we can read off some geometric properties of algebraic sets from algebraic prop-
erties of the corresponding radical ideals. We will see two such examples in this lecture.

Definition 2.10. Let I be an ideal in a ring R.

(1) The ideal I is prime if it is proper, and f g ∈ I implies f ∈ I or g ∈ I.

(2) The ideal I is maximal if it is proper, and for any ideal J satisfying I ⊆ J ⊆ R,
we have either J = I or J = R.

Example 2.11. We look at some ideals in k[x].

(1) Consider I1 = (x2 − x). I1 is not prime because x(x − 1) ∈ I1 , while x ∈

/ I1 and
2
x−1∈ / I1 . I1 is not maximal because (x − x) ( (x) ( k[x].
(2) Consider I2 = (x). We claim (x) is prime. Assume f g ∈ (x), then f g = xh for some
h ∈ k[x]. By unique factorisation, since x is irreducible, it must be a factor of f
or g. Hence f ∈ (x) or g ∈ (x). We claim (x) is maximal. Assume (x) ⊆ I ⊆ k[x].
If I 6= (x), then there exists f ∈ I\(x). Write f = a0 + a1 x + · · · + an xn , then
a0 6= 0, since otherwise f ∈ (x). We observe f − a0 = a1 x + · · · + an xn ∈ (x) ⊆ I.
It follows that a0 ∈ I, hence I = k[x] since a0 is a unit in k[x].
(3) Consider I3 = (0). I3 is prime because f g = 0 implies that either f = 0 or g = 0
as k[x] is an integral domain. I3 is not maximal because (0) ( (x) ( k[x, y].

Proposition 2.12. Let I be an ideal in the ring R.

(1) I is a prime ideal if and only if R/I is an integral domain. I is a maximal ideal
if and only if R/I is a field.
(2) Every maximal ideal is prime. Every prime ideal is radical.

Proof. (1) is non-examinable. (2) is an exercise. 

Under the corrspondence (2.2), we will find out what prime and maximal ideals correspond
to. Now we switch to geometry.

Definition 2.13. An algebraic set X ⊆ An is irreducible if there does not exist a de-
composition of X as a union of two stricly smaller algebraic sets. An irreducible (affine)
algebraic set is also called an affine variety. An algebraic set X ⊆ An is reducible if it is
not irreducible.

Example 2.14. We look at some algebraic sets in A2 .

16
 (1) The algebraic set V(xy) ⊆ A2 is the union of two coordinate axes. In other words,
V(xy) = V(x) ∪ V(y). Since each coordinate axis is an algebraic set stricly smaller
than V(xy), we conclude that V(xy) is reducible.
(2) The algebraic set V(x, y) ⊆ A2 consists of just one point, hence there is no way
to decompose it as the union of two strictly smaller algebraic sets. It follows that
V(x, y) is irreducible. Similarly, a point is always irreducible.

Next we show that prime ideals correspond to irreducible algebraic sets.

Proposition 2.15. Let I be a radical ideal in k[x1 , · · · , xn ] and X = V(I) the algebraic
set in An defined by I. Then I is prime if and only if X is irreducible.

Proof. In fact we prove the contrapositive: X is reducible ⇐⇒ I is not prime.

We first prove “=⇒”. Suppose X = X1 ∪ X2 with algebraic sets X1 , X2 ( X. Then X1 (

X implies that I(X1 ) ) I(X) by Proposition 2.9 (2). Hence there exists f1 ∈ I(X1 )\I(X).
Similarly X2 ( X implies that there exists f2 ∈ I(X2 )\I(X). The product f1 f2 vanishes
at all points of X, hence f1 f2 ∈ I(X). Therefore I = I(X) is not prime.

We then prove “⇐=”. Since I is not prime, there exist f1 , f2 ∈ / I such that f1 f2 ∈ I.
Consider the set S1 = I ∪ {f1 }. Then X1 = V(S1 ) is an algebraic set. Since S1 ⊇ I,
we have X1 ⊆ X by Proposition 1.7. Moreover, since f1 ∈ / I, there is some point p ∈ X
such that f1 (p) 6= 0, therefore p ∈
/ X1 . It follows that X1 ( X. Similarly we can consider
S2 = I ∪ {f2 }, then X2 = V(S2 ) ( X.

It remains to show that X1 ∪X2 = X. Since X1 and X2 are subsets of X, we have X1 ∪X2 ⊆
X. Conversely, for any p ∈ X, f (p) = 0 for every f ∈ I. Moreover f1 (p)f2 (p) = 0, which
implies f1 (p) = 0 or f2 (p) = 0. Therefore p ∈ V(S1 ) = X1 or p ∈ V(S2 ) = X2 . This
implies X ⊆ X1 ∪ X2 . 

Finally we show that maximal ideals correspond to points.

Proposition 2.16. Let I be a radical ideal in k[x1 , · · · , xn ] and X = V(I) the algebraic
set in An defined by I. Then I is maximal if and only if X is a point.

Proof. (This proof is non-examinable and not covered in lectures.)

In fact we prove the contrapositive: X is not a point ⇐⇒ I is not maximal.

We first prove “=⇒”. If X is not a point, then either X = ∅ or X contains more than
one point. If X = ∅, then by Proposition 2.9 (1), I = I(X) = k[x1 , · · · , xn ] is not a
proper ideal hence not maximal. If X contains more than one point, then we can pick a
subset Y of X containing only one point. Hence we have ∅ ( Y ( X. By Proposition
2.9 (2), we have k[x1 , · · · , xn ] = I(∅) ) I(Y ) ) I(X). Hence I = I(X) is not maximal.
17
We then prove “⇐=”. If I is not maximal, then either I is not a proper ideal, or there
exists an ideal J such that I ( J ( k[x1 , · · · , xn ]. If I is not proper then I = k[x1 , · · · , xn ],
hence X = V(I) = ∅ which is not a point.√ If I ( J ( k[x1 , · · · , xn ] for some ideal J,
then we claim that√we actually have I ( J ( k[x1 , · · · , xn ]. Indeed, by Lemma 2.2,
√ have I ( J ⊆ J. Moreover, by Nullstellensatz 2.8 (1), we have V(J) 6= ∅, hence
we
√ ( k[x1 , · · · , xn ]. Armed with
J = I(V(J)) √ this claim we apply Proposition 2.9 (2) to get
V(I) ) V( J) ) ∅. It follows that V( J) contains at least one point, hence X = V(I)
contains more than one point. 

In summary, the V − I correspondences induce bijections in each row of the diagram:

V
/
{radical ideals in k[x1 , · · · , xn ]} o {algebraic sets in An }
O I O

? ?
V
/
{prime ideals in k[x1 , · · · , xn ]} o {irreducible algebraic sets in An }
O I O

? ?
V
/
{maximal ideals in k[x1 , · · · , xn ]} o {points in An }
I

18
 Exercise Sheet 2

This sheet will be discussed in the exercise class on 16 October. You are welcome to submit
your solutions at the end of the exercise class or anytime earlier.

Exercise 2.1. Some proofs in lectures. We prove Lemma 2.2 and Proposition 2.12 (2).

(1) Let I be an ideal in a ring R. If am ∈ I and bn ∈ I for some a, b ∈ R and

m, n ∈ Z+ , show that (a + b)m+n ∈ I. (Hint: use the binomial expansion.)
√ √
(2) Let I be an ideal in a ring R. Prove that I is an ideal and I ⊆ I.
(3) Show that every maximal ideal is prime, and every prime ideal is radical.

Exercise 2.2. Examples of radical and prime ideals. Suppose a non-zero polynomial
f ∈ k[x1 , · · · , xn ] is factored as f = uf1k1 · · · ftkt for some 0 6= u ∈ k, k1 , · · · , kt ∈ Z+ , and
irreducible polynomials f1 , · · · , ft which are pairwisely coprime.

(1) Show that (f ) is a prime ideal if and only if f is an irreducible polynomial.

p
(2) Let f = f1 · · · ft . Show that (f ) = (f ). (Remark: this justifies Example 2.3.)
(3) Conclude that (f ) is a radical ideal if and only if f has no repeated factors.

Exercise 2.3. Examples of maximal ideals. Find all maximal ideals in k[x1 , · · · , xn ]. You
can follow these steps:

(1) For any fixed point p = (a1 , · · · , an ) ∈ An , consider the ring homomorphism ϕp :
k[x1 , · · · , xn ] → k; f (x1 , · · · , xn ) 7→ f (a1 , · · · , an ). Show that mp := ker(ϕp ) =
(x1 − a1 , · · · , xn − an ). Use Proposition 2.12 to show that mp is a maximal ideal.
(2) What is V(mp )? Use Proposition 2.16 to show that every maximal ideal in
k[x1 , · · · , xn ] is of the form mp for some p ∈ An . (Remark: historically, this
was proved before Nullstellensatz was established.)

Exercise 2.4. A famous example: the twisted cubic. Prove that the subset in A3 given
by X = {(t, t2 , t3 ) ∈ A3 | t ∈ k} is an affine variety. You can follow these steps:

(1) Show that X is the algebraic set V(I) for the ideal I = (y − x2 , z − x3 ) ⊆ k[x, y, z].
(2) Show that k[x, y, z]/I ∼
= k[t].
(3) Use Proposition 2.12 to conclude that I is a prime ideal, hence a radical ideal.
Use Proposition 2.9 to conclude that I = I(X). Use Proposition 2.15 to conclude
that X is an affine variety. (Remark: X is called the affine twisted cubic.)

(Remark: Exercise 3.4 will be a continuation of this one.)

19
 Solutions to Exercise Sheet 2

Solution 2.1. Some proofs in lectures.

(1) Using the binomial expansion, we have that (a + b)m+n = m+n m+n m+n−i i
P

i=0 i
a b.
m+n m+n−i i

m
For every term i a b , if i 6 n, then this term has a factor a , hence this
term is in I; if i > n, then this term has a factor bn , hence this term is also in I.
Since every such term is in I, it follows that their sum (a + b)m+n ∈ I.
√
(2) Let a, b ∈ I and r ∈ R. By Definition 2.1 there exist some m, n ∈ Z+ such √ that
m n m+n
a , b ∈ I. By part (1) we know that (a√+ b) ∈ I, hence√a + b ∈ I. We
m m m
also have (ra) √= r a ∈ I, hence ra ∈ I. It follows that I is an ideal. To
show that I√⊆ I, we just need to realise that for every a ∈ I, am ∈ I for m = 1.
Hence a ∈ I.
(3) Assume I is a maximal ideal in R, then R/I is a field by Proposition 2.12 (1).
Since every field is an integral domain, R/I is an integral domain. By Proposition
2.12 (1) again we conclude that I is a prime
√ ideal in I.
Assume J is a prime ideal. For any a ∈ J, there exists some n ∈ Z+ , such that
n
a ∈ J. We claim that a ∈ J. This can be shown by induction on n. When n = 1,
a ∈ J is automatic. Assume an ∈ J implies a ∈ J. If we have an+1 = a · an ∈ J,
n
√ either a ∈ J or a ∈ J. In either
then √ case we have a ∈ J. √ This shows that
J ⊆ J. By part (2) we also have J ⊆ J. It follows that J = J, hence J is a
radical ideal.

Solution 2.2. Examples of radical and prime ideals.

(1) Assume (f ) is a prime ideal. Since (f ) 6= k[x1 , · · · , xn ], f is not a constant

polynomial. If f is not an irreducible polynomial, then assume f = f1 f2 for some
non-constant polynomials f1 and f2 . Since f1 f2 = f ∈ (f ), it follows that either
f1 ∈ (f ) or f2 ∈ (f ). If f1 ∈ (f ), then f1 = f · g1 for some non-zero polynomial
g1 . Then f = f1 f2 = f g1 f2 which implies g1 f2 = 1. Hence f2 must be a constant,
which is a contradiction. If f2 ∈ (f ), the same argument implies f1 is a constant,
which is also a contradiction. This proves that f is irreducible.
Now assume f is an irreducible polynomial. We need to show (f ) is a prime
ideal. By definition an irreducible polynomial is not a constant, hence 1 ∈ / (f )
which means (f ) 6= k[x1 , · · · , xn ]. Let f1 f2 ∈ (f ) for polynomials f1 and f2 . Then
we can write f1 f2 = f g for some polynomial g. If g = 0, then either f1 = 0 ∈ (f )
or f2 = 0 ∈ (f ). If g 6= 0, then f is an irreducible factor in the factorisation of
f1 f2 , hence f is an irreducible factor of either f1 or f2 . Therefore we still have
f1 ∈ (f ) or f2 ∈ (f ). This proves that (f ) is a prime ideal.
20
 p
(2) We first show that (f ) ⊆ (f ). For any g ∈ (f ), there exists some poly-
nomial h such that g = f h = f1 · · · ft h. Let m = max{k p 1 , · · · , kt }. Then
m m m m m−k1 m−kt m
g = f1 · · · ft h = f · f1 · · ·p
ft h ∈ (f ), hence g ∈ p
(f ).
We prove the other inclusion (f ) ⊆ (f ). For any g ∈ (f ), there exists
some m ∈ Z+ such that g m ∈ (f ), that is, g m = f h = f1k1 · · · ftkt h for some
polynomial h. For every irreducible polynomial fi , since fi divides the right-hand
side, it must divide the left-hand side as well, i.e., fi divides g m . Therefore fi
divides g for every i. It follows that each fi appears in the factorisation of g,
hence g = f1 · · · fk g 0 = f g 0 ∈ (f ).
p
(3) (f ) is a radical ideal ⇐⇒ (f ) = (f ) ⇐⇒ (f ) = (f ) ⇐⇒ f and f differ by
a unit in k[x1 , · · · , xn ] (which is a non-zero constant). This holds if and only if
k1 = · · · = kt = 1; i.e. f has no repeated factors.

Solution 2.3. Examples of maximal ideals.

(1) We claim that every polynomial f (x1 , · · · , xn ) ∈ k[x1 , · · · , xn ] can be written in

the form
f = (x1 − a1 )g1 + · · · + (xn − an )gn + c
for some polynomials g1 , · · · , gn ∈ k[x1 , · · · , xn ] and a constant c ∈ k. There are
two ways to explain it (you can choose the one you like). The first approach: we
think of f as a polynomial in x1 and consider the Euclidean division of f by x1 −a1 .
We get f = (x1 −a1 )g1 +r1 where r1 has degree 0 in x1 , namely, r1 ∈ k[x2 , · · · , xn ].
Then we think of r1 as a polynomial in x2 , and consider the Euclidean division of
r1 by x2 − a2 , we get r1 = (x2 − a2 )g2 + r2 for some r2 ∈ k[x3 , · · · , xn ]. Repeat
this process to get
f = (x1 − a1 )g1 + r1
= (x1 − a1 )g1 + (x2 − a2 )g2 + r2
= ···
= (x1 − a1 )g1 + · · · + (xn − an )gn + rn
where rn is a constant. This justifies the claim. The second approach: we substi-
tute [(xi − ai ) + ai ] into each occurence of xi in f and expand the square brackets
leaving the round brackets untouched. In the expansion every non-constant term
has a factor of the form (xi − ai ). Then we can collect terms and write
f = (x1 − a1 )g1 + · · · + (xn − an )gn + c
where c is a constant. This justifies the claim.
Now we look at the image of f under ϕp . We have ϕp (f ) = f (a1 , · · · , an ) = c.
Therefore f ∈ ker ϕp ⇐⇒ c = 0 ⇐⇒ f = (x1 − a1 )g1 + · · · + (xn − an )gn ⇐⇒ f ∈
(x1 − a1 , · · · , xn − an ). This proves that mp = ker ϕp = (x1 − a1 , · · · , xn − an ).
21
 Moreover, ϕp is surjective, because every c ∈ k is the image of the constant
polynomial f = c. By the fundamental isomorphism theorem, we have
k = im ϕp = k[x1 , · · · , xn ]/ ker ϕp = k[x1 , · · · , xn ]/mp .
Since k is a field, we know that mp is a maximal ideal by Proposition 2.12 (1).
(2) V(mp ) = {p} is a single point set. By Proposition 2.16, there is a one-to-one
correspondence between maximal ideals in k[x1 , · · · , xn ] and points in An . Since
the ideals of the form mp have exhausted all points in An , they must be all maximal
ideals in k[x1 , · · · , xn ].

Solution 2.4. A famous example: the twisted cubic.

(1) We first show X ⊆ V(I). For every point (t, t2 , t3 ) ∈ X, we have y−x2 = t2 −t2 = 0
and z − x3 = t3 − t3 = 0. We then show V(I) ⊆ X. For every (x, y, z) ∈ V(I),
we have y − x2 = 0 and z − x3 = 0, hence y = x2 and z = x3 . It follows that
(x, y, z) = (x, x2 , x3 ) ∈ X.
(2) Consider the ring homomorphism
ϕ : k[x, y, z] −→ k[t]; f (x, y, z) 7−→ f (t, t2 , t3 ).
By the fundamental isomorphism theorem, we have
im ϕ ∼
= k[x, y, z]/ ker ϕ.
We need to find out im ϕ and ker ϕ.
We claim that ϕ is surjective, because for every p(t) ∈ k[t], it is the image of
p(x) ∈ k[x, y, z]. Therefore im ϕ = k[t].
To find out ker ϕ, we first claim that every f (x, y, z) ∈ k[x, y, z] can be written
in the form
f = (y − x2 )g1 + (z − x3 )g2 + h
where g1 , g2 ∈ k[x, y, z] and h ∈ k[x]. To see this, there are still two methods. The
first method: think of f as a polynomial in y, and consider the Euclidean division
of f by y − x2 . There is a quotient g1 ∈ k[x, y, z] and a remainder r1 ∈ k[x, z].
Then think of r1 as a polynomial in z, and consider the Euclidean division of r1
by z − x3 . There is a quotient g2 ∈ k[x, y, z] (in fact, in k[x, z]) and a remainder
h ∈ k[x]. In formulas,
f = (y − x2 )g1 + r1 = (y − x2 )g1 + (z − x3 )g2 + h.
The second method: we substitute [(y − x2 ) + x2 ] into each occurence of y in f
and substitute [(z − x3 ) + x3 ] into each occurence of z in f . We then expand the
square brackets leaving the round brackets untouched. In the expansion we collect
terms with a factor (y − x2 ) or (z − x3 ), and write
f = (y − x2 )g1 + (z − x3 )g2 + h
22
 where h ∈ k[x] does not involve y or z.
Armed with this claim, we find that the image of f under ϕ is given by
ϕ(f ) = (t2 − t2 )ϕ(g1 ) + (t3 − t3 )ϕ(g2 ) + h(t) = h(t).
Therefore ϕ(f ) = 0 ⇐⇒ h = 0 ⇐⇒ f = (y − x2 )g1 + (z − x3 )g2 ⇐⇒ f ∈
(y − x2 , z − x3 ). This means ker ϕ = (y − x2 , z − x3 ) = I.
Therefore the fundamental isomorphism theorem yields that k[t] ∼
= k[x, y, z]/I.
(3) Since k[t] is an integral domain, by Proposition 2.12, we conclude that I is a prime
ideal, hence a radical ideal. By part (1) and Proposition 2.9, X = V(I) implies
that I = I(X). Since I is a prime ideal, Proposition 2.15 shows that X is an
irreducible algebraic set, hence an affine variety.

23
 3. Coordinate Rings

We define polynomial functions and coordinate rings for algebraic sets. We will also study
polynomial maps between algebraic sets. Finally we will see how coordinate rings help us
understand polynomial maps.

3.1. Coordinate rings and polynomial maps. We look at functions on affine algebraic
sets. Roughly speaking, a function on an algebraic set X assigns to each point a value in k.
In algebraic geometry we are mostly interested in those functions defined by polynomials.
Definition 3.1. Let X ⊆ An be an algebraic set. A function ϕ : X → k is a polynomial
function if there exists f ∈ k[x1 , · · · , xn ] such that ϕ(p) = f (p) for every p ∈ X.
Remark 3.2. Two polynomials f, g ∈ k[x1 , · · · , xn ] define the same function on X if and
only if for every point p ∈ X, f (p) = g(p), or equivalently, f (p) − g(p) = 0. This holds if
and only if f − g ∈ I(X) by the definition of I. In other words, f and g define the same
polynomial function on X if and only if they are in the same coset of I(X) in k[x1 , · · · , xn ].
Therefore a polynomial function can be viewed as a coset of I(X), which is an element in
the quotient ring k[x1 , · · · , xn ]/I(X). This leads to the following definition.
Definition 3.3. Let X ⊆ An be an algebraic set. The quotient ring
k[X] := k[x1 , · · · , xn ]/I(X)
is called the coordinate ring of X.
Example 3.4. For any algebraic set X ⊆ An , the i-th coordinate defines a polynomial
function xi : X → k, which is called the i-th coordinate function. Since every polynomial
function is a polynomial in the coordinate functions, we can view the coordinate functions
as the generators of k[X]. This is where the name “coordinate ring” comes from.
Example 3.5. For the algebraic set X1 = V(x) ⊆ A2 , I(X1 ) = (x) since (x) is a prime
ideal hence is radical. Therefore the coordinate ring of X1 is k[X1 ] = k[x, y]/(x). We
show that it is isomorphic k[t]. Consider the ring homomorphism
ϕ : k[x, y] → k[t]; x 7→ 0, y 7→ t.
It is surjective because each p(t) ∈ k[t] is the image of p(y) ∈ k[x, y]. For any f (x, y) ∈
k[x, y], by collecting all terms involving x, we can write it as f (x, y) = xg(x, y) + h(y).
Its image ϕ(f (x, y)) = h(t). Hence f ∈ ker(ϕ) is equivalent to h(y) = 0, which is further
equivalent to f (x, y) ∈ (x). This shows ker(ϕ) = (x). By the fundamental isomorphism
theorem, we get k[X1 ] = k[x, y]/(x) ∼= k[t].
Example 3.6. For the algebraic sets X2 = V(y) and X3 = V(y − x2 ) in A2 , we can
similarly find out that k[X2 ] = k[x, y]/(y) ∼
= k[t] and k[X3 ] = k[x, y]/(y − x2 ) ∼
= k[t]. It is
not a coincidence that X1 , X2 and X3 have isomorphic coordinate rings. We will explain
this later.
24
Now we study maps between algebraic sets.

Definition 3.7. Let X ⊆ An and Y ⊆ Am be algebraic sets. A map ϕ : X → Y

is a polynomial map if there exist polynomial functions f1 , · · · , fm ∈ k[X], such that
ϕ(p) = (f1 (p), · · · , fm (p)) ∈ Y for every point p ∈ X.

Notice that a polynomial function on X is the same as a polynomial map from X to A1 .

Example 3.8. Let X ⊆ An be any algebraic set. The identity map idX : X → X;
(x1 , · · · , xn ) 7→ (x1 , · · · , xn ) is a polynomial map.

Example 3.9. Let W ⊆ A2 be any algebraic set and X = A1 . Then ϕ0 : W → X;

(x, y) 7→ xy is a polynomial map.

Example 3.10. Let X = A1 . Let Y1 = V(y−x2 ), Y2 = V(y 2 −x3 −x2 ) and Y3 = V(y 2 −x3 )
be algebraic sets in A2 . Then ϕ1 : X → Y1 ; t 7→ (t, t2 ) is a polynomial map from X to Y1 ,
since the point (t, t2 ) satisfies the defining equation of Y1 . Similarly, we can check that
ϕ2 : X → Y2 ; t 7→ (t2 − 1, t3 − t) and ϕ3 : X → Y3 ; t 7→ (t2 , t3 ) are both polynomial maps.

Remark 3.11. Let X ⊆ An , Y ⊆ Am , Z ⊆ Al be algebraic sets. Consider polynomial

maps
ϕ = (f1 (x1 , · · · , xn ), · · · , fm (x1 , · · · , xn )) : X → Y,
ψ = (g1 (y1 , · · · , ym ), · · · , gl (y1 , · · · , ym )) : Y → Z.
We can compose them to get a new polynomial map
ψ ◦ ϕ = (g1 (f1 , · · · , fm ), · · · , gl (f1 , · · · , fm )) : X → Z.

Example 3.12. To compose ϕ0 : W → X in Example 3.9 and ϕ1 : X → Y1 in Example

3.10, for any point p = (x, y) ∈ W , we have (ϕ1 ◦ ϕ0 )(x, y) = ϕ1 (xy) = (xy, x2 y 2 ). Hence
we get the polynomial map ϕ1 ◦ ϕ0 : W → Y1 ; (x, y) 7→ (xy, x2 y 2 ).

We can now describe when two algebraic sets “look the same”.

Definition 3.13. A polynomial map ϕ : X → Y between algebraic sets is an isomorphism

if there exists a polynomial map ψ : Y → X such that ψ ◦ ϕ = idX and ϕ ◦ ψ = idY . Two
algebraic sets X and Y are isomorphic if there exists an isomorphism between them.

Example 3.14. We show that ϕ1 : X → Y1 in Example 3.10 is an isomorphism. Let

ψ1 : Y1 → X; (x, y) 7→ x. Then the composition ψ1 ◦ ϕ1 : X → X is given by t 7→
(t, t2 ) 7→ t. Hence ψ1 ◦ ϕ1 = idX . The other composition ϕ1 ◦ ψ1 : Y1 → Y1 is given
by (x, y) 7→ x 7→ (x, x2 ). For every point (x, y) ∈ Y1 , we have y − x2 = 0, hence
(x, y) = (x, x2 ). This shows ϕ1 ◦ ψ1 = idY1 . We conclude that ϕ1 : X → Y1 (and
ψ1 : Y1 → X) is an isomorphism; in other words, X and Y1 are isomorphic.
25
Remark 3.15. If a polynomial map ϕ : X → Y is an isomorphism, then it induces a
bijection between the points in X and Y . However, it is important to note that the
converse is not true. We will see a counter example next time.

We will see next time that the coordinate ring captures a lot of geometry of the algebraic
set. In particular, whether two algebraic sets are isomorphic can be easily seen from their
coordinate rings.

26
3.2. Homomophisms of coordinate rings. We introduce a terminology which will be
very convenient in our discussion.
Definition 3.16. A finitely generated k-algebra is a ring that is isomorphic to a quotient
of a polynomial ring k[x1 , · · · , xn ]/I. A k-algebra homomorphism ϕ : k[y1 , · · · , ym ]/J →
k[x1 , · · · , xn ]/I is a ring homomorphism such that ϕ(c + J) = c + I for every constant
polynomial c ∈ k.

Recall that a polynomial function can be viewed as a polynomial map to A1 .

Definition 3.17. Let X ⊆ An and Y ⊆ Am be algebraic sets. Let ϕ : X → Y be a
polynomial map and g ∈ k[Y ] a polynomial function. The pullback of g along ϕ is the
polynomial function g ◦ ϕ ∈ k[X], denoted ϕ∗ (g).

The pullback map along ϕ sends a polynomial function on Y to a polynomial function on

X. We show that it preserves the ring structure and constants.
Lemma 3.18. For any polynomial map ϕ : X → Y , the pullback map
ϕ∗ : k[Y ] → k[X]; g 7→ g ◦ ϕ
is a k-algebra homomorphism.

Proof. We need to verify ϕ∗ preserves addition, multiplication and constants. For any
g1 , g2 ∈ k[Y ], we need to show (g1 + g2 ) ◦ ϕ = g1 ◦ ϕ + g2 ◦ ϕ. Indeed, for any point p ∈ X,
((g1 + g2 ) ◦ ϕ)(p) = (g1 + g2 )(ϕ(p)) = g1 (ϕ(p)) + g2 (ϕ(p)) = (g1 ◦ ϕ)(p) + (g2 ◦ ϕ)(p). Hence
ϕ∗ preserves addition. Replacing additions by multiplications shows that ϕ∗ preserves
multiplication. Now assume g is a constant function on Y , say, there exists some c ∈ k
such that g(q) = c for every q ∈ Y . Then (g ◦ ϕ)(p) = g(ϕ(p)) = c for every p ∈ X.
Therefore ϕ∗ (g) is the constant function on X which takes the same value as g. 
Example 3.19. The polynomial map ϕ : A1 → Y = V(y−x2 )(⊆ A2 ); t 7→ (t, t2 ) induces a
k-algebra homomorphism ϕ∗ : k[Y ] → k[A1 ], or more precisely, ϕ : k[x, y]/(y − x2 ) → k[t].
For any polynomial function f (x, y) on Y , ϕ∗ (f ) = f (t, t2 ) ∈ k[t]. In particular, for the
coordinate functions x and y on Y , we have ϕ∗ (x) = t and ϕ∗ (y) = t2 . For more examples,
the pullback of the polynomial function x + y is t + t2 ; the pullback of x2 y is t4 , and the
pullback of 3x3 + 5y + 1 is 3t3 + 5t2 + 1.

We have seen that every polynomial map ϕ : X → Y induces a k-algebra homomorphism

ϕ∗ : k[Y ] → k[X]. Next we show this is a one-to-one correspondence. This is the key
property of the “pullback” construction.
Theorem 3.20. Let X ⊆ An and Y ⊆ Am be algebraic sets. For every k-algebra homo-
morphism Φ : k[Y ] → k[X], there exists a unique polynomial map ϕ : X → Y , such that
Φ = ϕ∗ .
27
Proof. (This proof is non-examinable and not covered in lectures.)

We show the existence. For every coordinate function yi ∈ k[Y ], by assumption fi =

Φ(yi ) ∈ k[X] is a polynomial function on X. Since Φ is a k-algebra homomorphis, for any
polynomial function g(y1 , · · · , ym ) ∈ k[Y ], the image Φ(g) = g(f1 , · · · , fm ) ∈ k[X].

We consider the polynomial map ϕ = (f1 , · · · , fm ) : X → Am . To show it is a polynomial

map to Y , it must be checked that (f1 (p), · · · , fm (p)) ∈ Y for every p ∈ X; that is, it must
be checked that h(f1 (p), · · · , fm (p)) = 0 for every polynomial h ∈ I(Y ). Since h represents
the zero function in k[Y ], Φ(h) is also the zero function in k[X], hence Φ(h)(p) = 0 for
every p ∈ X. It follows that h(f1 (p), · · · , fm (p)) = Φ(h)(p) = 0, as desired.

To show Φ = ϕ∗ , it remains to show that Φ(g) = ϕ∗ (g) for every g ∈ k[Y ]. Indeed, for
any p ∈ X, Φ(g)(p) = g(f1 (p), · · · , fm (p)) = g(ϕ(p)) = (g ◦ ϕ)(p) = ϕ∗ (g)(p). Hence
Φ(g) = ϕ∗ (g), as required. This finishes the existence.
0
For uniqueness, assume there is another polynomial map ϕ0 = (f10 , · · · , fm ) : X → Y such
that Φ = (f ) . Then for each i, fi = (ϕ ) (yi ) = Φ(yi ) = ϕ (yi ) = fi . Hence ϕ0 = ϕ.
0 ∗ 0 0 ∗ ∗

This finishes the uniqueness. 

Remark 3.21. This theorem gives a one-to-one correspondence

( ) ( )
polynomial maps k-algebra homomorphisms
←→ .
ϕ : X −→ Y ϕ∗ : k[Y ] −→ k[X]

An application of this result is the following criterion for isomorphisms.

Proposition 3.22. A polynomial map ϕ : X → Y is an isomorphism if and only if

ϕ∗ : k[Y ] → k[X] is a ring isomorphism.

Proof. (This proof is non-examinable and not covered in lectures.)

Assume ϕ : X → Y is an isomorphism, then there exists ψ : Y → X such that ψ ◦ϕ = idX

and ϕ ◦ ψ = idY . By applying the pullback construction on both sides, we have ϕ∗ ◦ ψ ∗ =
(ψ ◦ ϕ)∗ = (idX )∗ = idk[X] . Similarly we have ψ ∗ ◦ ϕ∗ = idk[Y ] . Therefore ϕ∗ and ψ ∗ are
mutually inverse ring homomorphisms. Hence ϕ∗ : k[Y ] → k[X] is an isomorphism.

Assume ϕ∗ : k[Y ] → k[X] is a ring isomorphism, then there exists Ψ : k[X] → k[Y ] such
that ϕ∗ ◦ Ψ = idk[X] and Ψ ◦ ϕ∗ = idk[Y ] . By the existence in Theorem 3.20 we can write
Ψ = ψ ∗ for some polynomial map ψ : Y → X. Therefore we have (ψ ◦ ϕ)∗ = ϕ∗ ◦ ψ ∗ =
ϕ∗ ◦ Ψ = idk[X] = (idX )∗ . By the uniqueness in Theorem 3.20, we get ψ ◦ ϕ = idX .
Similarly we can get ϕ ◦ ψ = idY . Hence ϕ : X → Y is an isomorphism. 

This is a very powerful result as it allows us to show a certain polynomial map is an

isomorphism without writing down another one going backwards. It can also be used to
28
show a certain polynomial map is not an isomorphism, especially in some tricky situation
where the map is actually bijective on points, as shown in the following example:

Example 3.23. We consider the polynomial map ϕ : A1 → X = V(y 2 − x3 ) ⊆ A2 ; t 7→

(t2 , t3 ); see Example 3.10. One can show that it is bijective on points in A1 and X.
However, one can also show that ϕ∗ : k[X] → k[A1 ] is not an isomorphism of rings, hence
ϕ is not an isomorphism of algebraic sets. We leave the details as an exercise.

29
 Exercise Sheet 3

This sheet will be discussed in the exercise class on 23 October. You are welcome to submit
your solutions at the end of the exercise class or anytime earlier.
Exercise 3.1. Example: the graph of a polynomial function.

(1) Show that the projection map π : An → Ar , n > r, defined by π(a1 , · · · , an ) =

(a1 , · · · , ar ) is a polynomial map.
(2) Let X ⊆ An be an algebraic set and f ∈ k[X]. Define the subset G(f ) ⊆ An+1 by
G(f ) = {(a1 , · · · , an , an+1 ) ∈ An+1 | (a1 , · · · , an ) ∈ X and an+1 = f (a1 , · · · , an )}.
Show that G(f ) is an algebraic set. (Remark: G(f ) is called the graph of f .)
(3) Show that the map ϕ : X → G(f ); (a1 , · · · , an ) 7→ (a1 , · · · , an , f (a1 , · · · , an )) is a
polynomial map.
(4) Show that ϕ is an isomorphism of algebraic sets by writing down the inverse
polynomial map ψ : G(f ) → X, and checking both compositions are identities.
(5) Briefly explain why Example 3.14 is a special case of this.
Exercise 3.2. Example: a nodal cubic. Consider X = V(y 2 − x3 − x2 ) ⊆ A2 .

(1) Show that ϕ : A1 → X; t 7→ (t2 − 1, t3 − t) is a polynomial map.

(2) Show that ϕ is surjective but not injective on points, hence not an isomorphism.
(3) Show that y 2 − x3 − x2 is an irreducible polynomial. Use Exercise 2.2 to conclude
that I = (y 2 − x3 − x2 ) is a prime ideal, hence radical. Use Propositions 2.15 and
2.9 to conclude that X is an affine variety and I(X) = I.
Exercise 3.3. Example: a cuspidal cubic. Consider X = V(y 2 − x3 ) ⊆ A2 .

(1) Show that ϕ : A1 → X; t 7→ (t2 , t3 ) is a polynomial map.

(2) Show that ϕ is injective and surjective on points.
(3) Show that y 2 − x3 is an irreducible polynomial. Conclude that I = (y 2 − x3 ) is a
prime ideal, hence radical. Conclude that X is an affine variety and I(X) = I.
(4) Show that the k-algebra homomorphism ϕ∗ : k[X] → k[A1 ] is not surjective.
Conclude by Proposition 3.22 that ϕ is not an isomorphism.
Exercise 3.4. Example: the twisted cubic, revisited. This is a continuation of Exercise
2.4. We consider the polynomial map ϕ : A1 → X; t 7→ (t, t2 , t3 ).

(1) Show that ϕ is an isomorphism by writing down the inverse ψ : X → A1 and

computing the two compositions.
(2) Show that ϕ is an isomorphism by proving that ϕ∗ is an isomorphism.
30
 Solutions to Exercise Sheet 3

Solution 3.1. Example: the graph of a polynomial function.

(1) Every component of π is given by a polynomial, and the image of any point in An
is clearly in Ar , so π is a polynomial map.
(2) Since X is an algebraic set, we can write X = V(S) where S is a set of polynomials
in x1 , · · · , xn . Each polynomial in S can also be thought as a polynomial in
x1 , · · · , xn , xn+1 . Assume the polynoial function f is represented by a polynomial
F ∈ k[x1 , · · · , xn ]. Then consider the set of polynomials T = S ∪ {xn+1 − F } ⊆
k[x1 , · · · , xn , xn+1 ]. We claim G(f ) = V(T ).
To prove the claim, we need to show mutual inclusions. Given any point p =
(a1 , · · · , an , an+1 ) ∈ G(f ), we have (a1 , · · · , an ) ∈ X and an+1 = f (a1 , · · · , an ).
The former implies that p is a solution to all polynomials in S, and the latter
implies that p is a solution to the polynomial xn+1 − F . It follows that p ∈ V(T ).
Given any point q = (a1 , · · · , an , an+1 ) ∈ V(T ), since xn+1 does not occur
in any polynomial in S, we know that (a1 , · · · , an ) ∈ V(S). Moreover an+1 −
F (a1 , · · · , an ) = 0 implies that an+1 = F (a1 , · · · , an ) = f (a1 , · · · , an ). Hence
q ∈ G(f ). This finishes the proof of the claim G(f ) = V(T ), which implies G(f )
is an algebraic set.
(3) The first n components of ϕ are obviously polynomials in a1 , · · · , an . Since f is
a polynomial map, it can also be represented by a polynomial F ∈ k[x1 , · · · , xn ].
It remains to check the image of ϕ is always in G(f ), which is clear from the
definition of G(f ).
(4) We define ψ : G(f ) → X as the projection map to the first n components. Namely,
ψ(x1 , · · · , xn+1 ) = (x1 , · · · , xn ). It is clearly a polynomial map. We compute both
compositions. Given any p = (a1 , · · · , an ) ∈ X, we have
(ψ ◦ ϕ)(p) = ψ(a1 , · · · , an , f (a1 , · · · , an )) = (a1 , · · · , an ) = p.
Given any q = (a1 , · · · , an , an+1 ) ∈ G(f ), we have
(ϕ ◦ ψ)(q) = ϕ(a1 , · · · , an ) = (a1 , · · · , an , f (a1 , · · · , an )) = (a1 , · · · , an , an+1 ) = q.
Therefore ϕ (hence also ψ) is an isomorphism.
(5) Let X = A1 , and f (x) = x2 ∈ k[x], then part (4) recovers Example 3.14.

Solution 3.2. Example: a nodal cubic.

(1) Both components in ϕ are polynomials in t. Since

y 2 − x3 − x2 = (t3 − t)2 − (t2 − 1)3 − (t2 − 1)2
= t6 − 2t4 + t2 − t6 + 3t4 − 3t2 + 1 − t4 + 2t2 − 1 = 0,
31
 we conclude ϕ(t) ∈ X for every t ∈ A1 . Hence ϕ is a polynomial map.
(2) To show ϕ is surjective but not injective on points, take any point q = (x, y) ∈ X.
There are two cases. If x = 0, then by the defining equation of X we also have
y = 0. It is easy to see that the point q = (0, 0) is the image of the point t = 1 or
t = −1. Hence ϕ is not injective on points. If x 6= 0, then consider t = xy . To find
its image, notice that
y2 y 2 − x2 x3
t2 − 1 = − 1 = = = x;
x2 x2 x2
y
t3 − t = t · (t2 − 1) = · x = y.
x
Therefore ϕ(t) = (x, y), which means the point q = (x, y) is in the image of ϕ.
The two cases together show that ϕ is surjective on points. Since we have proved
ϕ is not injective on points, it cannot be an isomorphism by Remark 3.15.
(3) Use contradiction. Assume y 2 − x3 − x2 = f (x, y)g(x, y) for non-constant polyno-
mials f, g ∈ k[x, y]. Since the left-hand side has degree 2 in y, the degrees of f
and g in y must be either 2 and 0, or 1 and 1. In the first case we can write
y 2 − x3 − x2 = (y 2 f2 (x) + yf1 (x) + f0 (x)) · g(x).
Comparing coefficients of y 2 we find f2 (x)g(x) = 1, hence g(x) must be a constant.
Contradiction. In the second case we can write
y 2 − x3 − x2 = (yf1 (x) + f0 (x)) · (yg1 (x) + g0 (x)).
Comparing coefficients of y 2 we find f1 (x)g1 (x) = 1. Without loss of generality we
can assume f1 (x) = g1 (x) = 1. Comparing coefficients of y we find f0 (x) + g0 (x) =
0. Comparing constant terms we find −x3 − x2 = f0 (x)g0 (x) = −f0 (x)2 , hence
f0 (x)2 = x3 + x2 , which is also a contradiction since x3 + x2 = x2 (x + 1) is not a
square. So we conclude that y 2 − x3 − x2 is irreducible. By Exercise 2.2 (1) we
know I = (y 2 − x3 − x2 ) is a prime ideal. By Proposition 2.12 (2) we know I is a
radical ideal. By Proposition 2.9 (1) we know I(X) = I. By Proposition 2.15 we
know X is an irreducible algebraic set, i.e. an affine variety.

Solution 3.3. Example: a cuspidal cubic.

(1) Both components in ϕ are polynomials in t. Since

y 2 − x3 = (t3 )2 − (t2 )3 = 0,
we conclude ϕ(t) ∈ X for every t ∈ A1 . Hence ϕ is a polynomial map.
(2) To show ϕ is injective and surjective on points, take any point q = (x, y) ∈ X.
There are two cases. If x = 0, then by the defining equation of X we also have
y = 0. Assume ϕ(t) = (0, 0), then there is a unique point t = 0 whose image is
(0, 0). If x 6= 0, assume ϕ(t) = (x, y), then we must have t = xy , so there is at most
32
 one point whose image is (x, y). To check that its image is indeed (x, y), notice
that
y2 x3
t2 = = x;
x2 x2
y
t3 = t · t2 = · x = y.
x
Therefore ϕ(t) = (x, y), which means there is a unique point t ∈ A1 whose image is
the point q = (x, y). The two cases together show that ϕ is injective and surjective
on points.
(3) Use contradiction. Assume y 2 − x3 = f (x, y)g(x, y) for non-constant polynomials
f, g ∈ k[x, y]. Since the left-hand side has degree 2 in y, the degrees of f and g in
y must be either 2 and 0, or 1 and 1. In the first case we can write

y 2 − x3 = (y 2 f2 (x) + yf1 (x) + f0 (x)) · g(x).

Comparing coefficients of y 2 we find f2 (x)g(x) = 1, hence g(x) must be a constant.

Contradiction. In the second case we can write

y 2 − x3 = (yf1 (x) + f0 (x)) · (yg1 (x) + g0 (x)).

Comparing coefficients of y 2 we find f1 (x)g1 (x) = 1. Without loss of generality

we can assume f1 (x) = g1 (x) = 1. Comparing coefficients of y we find f0 (x) +
g0 (x) = 0. Comparing constant terms we find −x3 = f0 (x)g0 (x) = −f0 (x)2 , hence
f0 (x)2 = x3 , which is also a contradiction since x3 is not a square. So we conclude
that y 2 − x3 is irreducible. By Exercise 2.2 (1) we know I = (y 2 − x3 ) is a prime
ideal. By Proposition 2.12 (2) we know I is a radical ideal. By Proposition 2.9
(1) we know I(X) = I. By Proposition 2.15 we know X is an irreducible algebraic
set, i.e. an affine variety.
(4) By part (3) we have k[X] = k[x, y]/(y 2 − x3 ). To write down the pullback
map explicitly, we notice that ϕ∗ (x) = t2 and ϕ∗ (y) = t3 . Therefore for any
polynomial map on X represented by a polynomial f (x, y) ∈ k[x, y], its image
ϕ∗ (f ) = f (t2 , t3 ); that means, we simply replace every occurence of x by t2 and y
by t3 . It is clear that ϕ∗ (f ) is a polynomial in t. We claim that it has no term of
degree 1 in t. Indeed, the image of the constant term of f is still the same constant,
and the image of any other monomial of f is a monomial in t of degree at least 2.
This claim implies that ϕ∗ is not surjective, because any polynomial in t with a
non-zero degree 1 term is not in the image of ϕ∗ . In particular, t ∈ k[t] = k[A1 ] is
not in the image of ϕ∗ . Hence ϕ∗ is not an isomorphism. By Proposition 3.22, ϕ
is not an isomorphism.

Solution 3.4. Example: the twisted cubic, revisited.

33
(1) We define the polynomial map ψ : X → A1 by ψ(x, y, z) = x. It is clearly a
polynomial map as its only component is a polynomial. For any t ∈ A1 , we have
(ψ ◦ ϕ)(t) = ψ(t, t2 , t3 ) = t. For any (x, y, z) ∈ X, we have (ϕ ◦ ψ)(x, y, z) =
ϕ(x) = (x, x2 , x3 ) = (x, y, z). This shows that ϕ is an isomorphism.
(2) We first write down the pullback map ϕ∗ explicitly. By Exercise 2.4 (3), we have
k[X] = k[x, y, z]/I(X) = k[x, y, z]/(y − x2 , z − x3 ). We also have k[A1 ] = k[t].
The pullback of the coordinate functions are given by ϕ∗ (x) = t, ϕ∗ (y) = t2 and
ϕ∗ (z) = t3 . Therefore ϕ∗ is given by
ϕ∗ : k[x, y, z]/(y − x2 , z − x3 ) −→ k[t]; f (x, y, z) 7−→ f (t, t2 , t3 ).
We actually have proved in Exercise 2.4 (2) that ϕ∗ is an isomorphism. Indeed, ϕ∗
is surjective because every p(t) ∈ k[t] is the image of p(x) ∈ k[x, y, z] (or rather,
the coset p(x) + I(X) in the quotient ring). Moreover, ϕ∗ is injective because if the
image of f (x, y, z) is the zero polynomial in k[t], it must be in I(X), which means
that the only element in the kernel of ϕ∗ is the coset 0 + I(X), which is the zero
element in the quotient ring. Therefore by Proposition 3.22, we conclude that ϕ
is an isomorphism.

34
 4. Projective Algebraic Sets

Instead of affine spaces, it is more natural to study algebraic geometry in projective

spaces. We first introduce projective spaces, then study projective algebraic sets. There
is a similar projective Nullstellensatz and V − I correspondence.

4.1. Projective spaces. We will study algebraic geometry in projective spaces. We

prefer projective spaces because results in projective spaces are usually nicer. One such
example is that: two curves in A2 may or may not intersect each other. When they in-
tersect, the number of intersection is not known until one solves the system of equations.
However, in projective spaces P2 , two curves always intersect, and the number of intersec-
tion points can be easily read off from their equations. In this lecture we will understand
the projective space Pn from the following three different points of views:

• Pn is the set of 1-dimensional subspaces in An+1 (definition);

• Pn is covered by n + 1 subsets which are all An (aka from projective to affine);
• Pn is obtained by adding to An a “boundary at infinity”, whose points correspond
to “asymptotic directions” in An (aka from affine to projective).
Definition 4.1. For every integer n > 0, the projective space Pnk (or Pn if k is understood)
of dimension n over a field k is the set of 1-dimensional vector subspaces in An+1k .
Remark 4.2. Each point a = (a0 , a1 , · · · , an ) 6= (0, 0, · · · , 0) in An+1 determines a 1-
dimensional subspace. Two such points a = (a0 , a1 , · · · , an ) and b = (b0 , b1 , · · · , bn ) define
the same subspace if and only if there is some λ 6= 0 such that bi = λai for each 0 6 i 6 n.
We say two such points are equivalent, and write a ∼ b. Then points in Pn can be
identified with such equivalence classes. More precisely,
Pn = An+1 \{(0, · · · , 0)} / ∼ .

Definition 4.3. If a point p ∈ Pn is determined by (a0 , a1 , · · · , an ) ∈ An+1 \{(0, · · · , 0)},

we say that a0 , a1 , · · · , an are homogeneous coordinates of p, denoted p = [a0 : a1 : · · · : an ].
Remark 4.4. The homoeneous coordinates of p ∈ Pn are only determined up to a non-zero
scalar multiplication, so the i-th coordinate ai is not a well-defined number. However, it
is a well-defined notion to say whether ai is zero or non-zero; and if ai 6= 0, the ratios
aj /ai are also well-defined (since they remain unchanged under equivalence).

We want to relate projective spaces to our familiar affine spaces, so that we can “visualise”
them easily. There are two typical ways to do this.
Construction 4.5 (From projective to affine). We will see how to find subsets in Pn
which are affine spaces. For each 0 6 i 6 n, consider the subset
Ui = {[a0 : a1 : · · · : an ] ∈ Pn | ai 6= 0}.
35
Each point p ∈ Ui can be written as
 
a0 ai−1 ai+1 an
p= : ··· : :1: : ··· : .
ai ai ai ai
Since we insist that the i-th coordinate is 1, the other n coordinates are uniquely deter-
mined, which can be used to identify Ui with An . Moreover, since every point in Pn has at
least one non-zero homogeneous coordinate, it lies in at least one of the Ui ’s. This implies
Pn = ∪ni=0 Ui . (4.1)
So Pn is covered by n + 1 subsets, each of which looks just like An .

Definition 4.6. Each subset Ui = {[a0 : a1 : · · · : an ] ∈ Pn | ai 6= 0} of Pn is called a

standard affine chart of Pn . For every point p = [a0 : a1 : · · · : an ] ∈ Ui , the n-tuple

a0
ai
, · · · , ai−1
ai
, ai+1
ai
, · · · , aani are called the non-homogeneous coordinates of p with respect
to Ui . The cover Pn = ∪ni=0 Ui is called a standard affine cover of Pn .

Example 4.7. P1 has two standard affine charts. The point [2 : 3] ∈ P1 has non-
homogeneous coordinate 32 with respect to U0 , and 32 with respect to U1 . P2 has three
standard affine charts. The point [2 : 3 : 0] ∈ P2 has non-homogeneous coordinates ( 32 , 0)
with respect to U0 , and ( 32 , 0) with respect to U1 . This point is not in U2 because the
corresponding coordinate is 0.

Construction 4.8 (From affine to projective). We will see how to get Pn by adding
“points at infinity” to the affine space An . We work with U0 but each Ui works in the
same way. The complement of U0 in Pn is
H0 = Pn \U0 = {[0 : a1 : · · · : an ] ∈ Pn },
which can be identified with Pn−1 as each point in H0 is given by n homogeneous coor-
dinates which are not simultaneously zero. Hence Pn can be decomposed into an affine
= An and a set of “points at infinity” H0 ∼
space U0 ∼ = Pn−1 :
Pn = U0 ∪ H0 ∼
= An ∪ Pn−1 . (4.2)

Now we explain why we can view points in H0 as “asymptotic directions” of lines in

U0 = An . This is best illustrated for n = 2, but works for any positive integer n.

Example 4.9. Consider two lines V(x2 − x1 + 1) and V(x2 − x1 − 1) in A2 ∼ = U0 . They are
parallel since they have the same slope. We can regard x1 and x2 as the non-homogeneous
coordinates with respect to U0 , and substitute xi by aa0i . Then the defining equations of
the two lines become
a2 a1
− ± 1 = 0.
a0 a0
We clear the denominators to get
a2 − a1 ± a0 = 0.
36
Notice that after clearing the denominator, we no longer require a0 to be non-zero. There-
fore we could possibly get extra solutions corresponding to points in H0 . To see which
points in H0 satisfy the equation, we set a0 = 0. Then the equation becomes
a2 − a1 = 0.
Up to a non-zero scalar multiplication we get one extra solution [a0 : a1 : a2 ] = [0 : 1 : 1].
So we can say both lines pass through (and intersect at) the point [0 : 1 : 1] at infinity.
Since parallel lines always acquire the same point at infinity, we get an idea that points
in H0 correspond to “asymptotic directions”.

This example shows us how to understand points at infinity. We use the line V(x2 −x1 +1)
to preview some notions that will come up later. After clearing the denominators, we get
a polynomial a2 − a1 + a0 in which every monomial has the same degree. We say such a
polynomial is homogeneous. Its solutions in P2 is called a projective algebraic set. Since
it is obtained by adding the appropriate “points at infinity” to the affine algebraic set
V(x2 − x1 + 1), we say this projective algebraic set is the projective closure of the affine
algebraic set V(x2 − x1 + 1). In fact, every affine algebraic set in An (not necessarily a
line) has a projective closure in Pn obtained by adding the appropiate “points at infinity”,
which can be computed using a similar calculation. We will see more examples later.

37
4.2. Projective algebraic sets and projective Nullstellensatz. We develop the the-
ory of projective algebraic sets. Most of the results and proofs are similar to those in the
affine case. We will be brief on the similar part, but careful on a few special features.

Definition 4.10. A non-zero polynomial f ∈ k[z0 , z1 , · · · , zn ] is homogeneous of degree d

if each term of f has the same total degree d.

As easy examples, z2 − z12 is not homogeneous while z0 z2 − z12 is homogeneous of degree

2. The importance of this notion is the following. If f is homogeneous of degree d, then

f (λa0 , λa1 , · · · , λan ) = λd f (a0 , a1 , · · · , an ). (4.3)

In particular this means f (λa0 , λa1 , · · · , λan ) = 0 if and only if f (a0 , a1 , · · · , an ) = 0 for
any λ 6= 0. Therefore for any point p = [a0 : a1 : · · · : an ] ∈ Pn , the condition f (p) = 0 is
independent of the choice of its homogeneous coordinates. Hence the zero locus of f

{[a0 : a1 : · · · : an ] ∈ Pn | f (a0 , a1 , · · · , an ) = 0}

is also well-defined.

Remark 4.11. Since the zero polynomial satisfies (4.3) for every non-negative integer d, as
a convention, the zero polynomial is considered to be a homogeneous polynomial of any
degree. By doing so, we can avoid many unnecessary exceptions. For instance, the sum
of two homogeneous polynomial of degree d is again a homogeneous polynomial of degree
d when we include the zero polynomial.

Definition 4.12. For any non-zero polynomial f ∈ k[z0 , z1 , · · · , zn ] of degree m, we say

f = f0 + f1 + · · · + fm is the homogeneous decomposition of f , if for each i, 0 6 i 6 m, fi
is homogeneous of degree i. Each fi is called a homogeneous component of f .

Definition 4.13. An ideal I ⊆ k[z0 , z1 , · · · , zn ] is homogeneous if for every non-zero

polynomial f ∈ I, each of its homogeneous components fi ∈ I.

In practice, this condition for an ideal being homogeneous is not very easy to check. The
following criterion is usually more convenient.

Proposition 4.14. An ideal I ⊆ k[z0 , z1 , · · · , zn ] is homogeneous if and only if it can be

generated by a finite set of homogeneous polynomials.

Proof. We leave the proof as an exercise. 

Example 4.15. The ideals (x) and (x, y 2 ) in k[x, y] are both homogeneous, while the
ideal (x + y 2 ) in k[x, y] is not homogeneous, because the degree 1 part of x + y 2 is x, which
is not in this ideal.
38
Notice that an ideal could have many different sets of generators. The statement only
requires one set of generators consists of only homogeneous polynomials. It is still possible
that some other generating set is not given by homogeneous polynomials. Next we can
define the correspondences V and I.

Definition 4.16. For any homogeneous ideal I ⊆ k[z0 , z1 , · · · , zn ], the set

V(I) = {p ∈ Pn | f (p) = 0 for every homogeneous polynomial f ∈ I}

is called the projective algebraic set defined by I.

Similar to the affine case, the following result is usually convenient in practice.

Lemma 4.17. Suppose a homogeneous ideal I ⊆ k[z0 , z1 , · · · , zn ] is generated by a finite

set of homogeneous polynomials S = {f1 , · · · , fm }. Let

V(S) = {p ∈ Pn | f1 (p) = · · · fm (p) = 0}.

Then V(S) = V(I).

Proof. Similar to the proof of Lemma 1.10. We leave it as an exercise. 

Corollary 4.18. Every projective algebraic set X ⊆ Pn can be written as V(S) for a
finite set S of homogeneneous polynomials in k[z0 , · · · , zn ].

Proof. It follows immediately from Propositions 4.14 and 4.17. 

Example 4.19. In P1 , the projective algebraic set V(3z0 − 2z1 ) is the single-point set
{[2 : 3]}. In P2 , the projective algebraic set V(z2 − z1 + z0 ) is one of the affine lines in
Example 4.9 together with the corresponding point at infinity.

Definition 4.20. A projective algebraic set X ⊆ Pn is called a hypersurface if X = V(f )

for some non-constant homogeneous polynomial f ∈ k[z0 , z1 , · · · , zn ].

Definition 4.21. For any subset X ⊆ Pn , the set

( )
f (p) = 0 for every choice of homogeneous
I(X) = f ∈ k[z0 , z1 , · · · , zn ]
coordinates of every point p ∈ X

is called the ideal of X.

Lemma 4.22. For any subset X ⊆ Pn , I(X) is a homogeneous radical ideal.

Proof. The proof of Lemma 2.6 (2) works literally here to show I(X) is a radical ideal. To
show it is homogeneous, let f ∈ I(X) and write f = f0 +f1 +· · ·+fm for the homogeneous
39
decomposition of f where m is the degree of f . For each p = [a0 : a1 : · · · : an ] ∈ X and
λ ∈ k\{0}, we can also write p = [λa0 : λa1 : · · · : λan ], hence we have
0 = f (p) = f (λa0 , λa1 , · · · , λan )
m
X
= fi (λa0 , λa1 , · · · , λan )
i=0
m
X m
X
i
= λ fi (a0 , a1 , · · · , an ) = λi fi (p).
i=0 i=0

This means that every λ ∈ k\{0} is a root of the polynomial m i

P
i=0 fi (p)x ∈ k[x]. This
must be a zero polynomial, because the number of roots of any non-zero polynomial is at
most equal to its degree m. It follows that fi (p) = 0 for every 0 6 i 6 m, so fi ∈ I(X). 

Remark 4.23. We have used the same notation V and I in both affine and projective cases.
In practice it is usually clear which is meant; but if there is any danger of confusion, we
will write Vp and Ip for the projective operations, Va and Ia for the affine ones.

Now we state the projective Nullstellensatz. It is similar to the affine version, but there
is one point where care is needed. Clearly the trivial ideal (1) = k[z0 , z1 , · · · , zn ] defines
the empty set in An+1 , hence the empty set in Pn , as it should be. However, the ideal
(z0 , z1 , · · · , zn ) defines a single-point set {(0, · · · , 0)} in An+1 , which also corresponds
to the empty set in Pn . This ideal (z0 , z1 , · · · , zn ) is an awkward exception to several
statements in the theory, and is traditionally known as the “irrelevant ideal”. Keeping
that in mind, we state the projective version of Nullstellensatz.

Theorem 4.24 (Projective Nullstellensatz). Let k be an algebraically closed field. For

any homogeneous ideal I ⊆ k[z0 , z1 , · · · , zn ],
√
(1) V(I) = ∅ if and only if I ⊇ (z0 , z1 , · · · , zn ).
√
(2) If V(I) 6= ∅, then I(V(I)) = I.

Proof. This is an easy consequence of the affine Nullstellensatz. Non-examinable. Inter-

ested reader can find the proof in [Section 5.3, Reid, Undergraduate Algebraic Geometry]
or [Section 4.2, Fulton, Algebraic Curves]. 

40
 Exercise Sheet 4

This sheet will be discussed in the exercise class on 30 October. You are welcome to submit
your solutions at the end of the exercise class or anytime earlier.
Exercise 4.1. Get familiar with projective spaces. Answer the following quick questions.

(1) What is P0 ? Why does P1 have only one more point than A1 ? When k = C, can
you picture P1C as a bubble (or a ball, something like that)? Which points in Pn
belong to only one of the Ui ’s in the standard affine cover of Pn ?
(2) Follow Example 4.9 to find the points at infinity for the affine algebraic set Va (x22 −
x21 − 1) ⊆ A2 . Do the same for Va (x22 − x21 ) and Va (x22 − x31 ) in A2 .
Exercise 4.2. Properties of homogeneous polynomials and ideals.

(1) Let f ∈ k[z0 , · · · , zn ] be a non-zero homogeneous polynomial. Assume f = gh for

some g, h ∈ k[z0 , · · · , zn ]. Show that g and h are also homogeneous polynomials.
(2) Show that an ideal I ⊆ k[z0 , z1 , · · · , zn ] is homogeneous if and only if it can be
generated by a finite set of homogeneous polynomials.
(3) Suppose a homogeneous ideal I ⊆ k[z0 , z1 , · · · , zn ] is generated by a finite set of
homogeneous polynomials S = {f1 , · · · , fm }. Show that Vp (I) = Vp (S).
Exercise 4.3. Projective spaces are better than affine spaces! A line in P2 is a projective
algebraic set Vp (f ) defined by a homogeneous linear polynomial f = a0 z0 + a1 z1 + a2 z2 ∈
k[z0 , z1 , z2 ] for some a0 , a1 , a2 ∈ k not simultaneously zero.

(1) Show that two distinct points in P2 determine a unique line.

(2) Show that two distinct lines in P2 intersect at a unique point.

(Hint: How to compute the dimension of the null space of a matrix? Rank-nullity!)
Exercise 4.4. Example of projective algebraic sets. Recall that we always assume k is
algebraically closed. Prove that projective algebraic sets in P1 are just the finite subsets
in P1 (including ∅) together with P1 itself. You can follow these steps:

(1) Verify that they are indeed projective algebraic sets.

(2) Show that every non-constant homogeneous polynomial f (z0 , z1 ) ∈ k[z0 , z1 ] can
be factored into a product of homogeneous polynomials of degree 1. (Hint: you
can use the following lemma in algebra: a non-constant polynomial g(x) ∈ k[x]
can be factored into a product of polynomials of degree 1.)
(3) Show that if a projective algebraic set in P1 is not P1 itself, then it contains at
most finitely many points.
41
 Solutions to Exercise Sheet 4

Solution 4.1. Get familiar with projective spaces.

(1) Since there is only one 1-dimensional linear subspace in A1 , P0 is a point. P1 =

U0 ∪ H0 where U0 ∼ = A1 is an affine space and H0 ∼ = P0 is a point. Therefore P1
has just one more point than A1 . When k = C, U0 ∼ = A1C = C1 is the complex
plane. To view P1 as a bubble, imagine we remove a point from the surface of a
bubble (or a globe), the remaining part can be stretched into the complex plane.
A point p ∈ Pn belongs to only one of the standard affine chart Ui if and only
if p has only one non-zero homogeneous coordinate. We can assume this non-zero
homogeneous coordinate to be 1, otherwise we can divide all components by it.
So the point p can be given by p = [0 : · · · : 0 : 1 : 0 : · · · : 0] with 1 at a certain
position and 0 at all other positions. There are n + 1 such points.
z1
(2) We regard x1 and x2 as non-homogeneous coordinates and substitute x1 = z0
and
z22 z12
x2 = zz02 . The equation x22 − x21 − 1 = 0 becomes − − 1 = 0. We clear the
z02 z02
denominators to allow z0 to be zero, then we get z22 − z12 − z02 = 0. To find the
points at infinity, set z0 = 0, then we have z22 − z12 = 0, hence z2 = ±z1 . As points
in P2 we get two solutions [z0 : z1 : z2 ] = [0 : 1 : 1] or [0 : 1 : −1], which are the
points at infinity for Va (x22 − x21 − 1). This example tells us that a hyperbola has
two “asymptotic directions”, which is easy to understand since a hyperbola has
two asymptotes.
For Va (x22 −x21 ), we still substitute x1 = zz10 and x2 = zz20 . The equation x22 −x21 = 0
z2 z2
becomes z22 − z12 = 0. We clear the denominators to allow z0 to be zero, then we get
0 0
z22 −z12 = 0. To find the points at infinity, set z0 = 0, then we still have z22 −z12 = 0,
hence z2 = ±z1 . As points in P2 we get two solutions [z0 : z1 : z2 ] = [0 : 1 : 1]
or [0 : 1 : −1], which are the points at infinity for Va (x22 − x21 ). The result is not
surprising, because the polynomial x22 − x21 defines precisely the two asymptotes of
the hyperbola in the previous case.
For Va (x22 −x31 ), we still substitute x1 = zz10 and x2 = zz20 . The equation x22 −x31 = 0
z2 z3
becomes z22 − z13 = 0. We clear the denominators to allow z0 to be zero, then we
0 0
get z0 z22 − z13 = 0. To find the points at infinity, set z0 = 0, then we get −z13 = 0,
hence z1 = 0. As points in P2 we get one solution [z0 : z1 : z2 ] = [0 : 0 : 1], which
is the point at infinity for Va (x22 − x31 ).

Solution 4.2. Properties of homogeneous polynomials and ideals.

42
(1) We write the homogeneous decompositions of g and h as
g = gM + gM −1 + · · · + gm+1 + gm ,
h = hN + hN −1 + · · · + hn+1 + hn ,
where M and m are the maximal and minimal degrees of non-zero monomials in g
respectively; similarly N and n are the maximal and minimal degrees of non-zero
monomials in h respectively. Then the degree of every monomial in the product
f = gh is between m + n and M + N . Moreover, the sum of all degree M + N
monomials in f is given by gM hN , which is non-zero since both gM and hN are
non-zero. Similarly, the sum of all degree m + n monomials in f is given by gm hn ,
which is non-zero since both gm and hn are non-zero. If f is homogeneous, we
must have M + N = m + n, which is only possible when M = m and N = n.
Therefore both g and h are homogeneous.
(2) Assume I is a homogeneous ideal. Since k[z0 , · · · , zn ] is a Noetherian ring, I
is finitely generated. So we can write I = (f1 , · · · , fm ) for some f1 , · · · , fm ∈
I which are not necessarily homogeneous polynomials. However, each fi has a
homogeneous decomposition, say, fi = fi,0 + fi,1 + · · · + fi,di where di is the degree
of fi . We claim that I is generated by all the fi,j ’s; that is,
I = (f1,0 , · · · , f1,d1 , f2,0 , · · · , f2,d2 , · · · · · · , fm,0 , · · · fm,dm ).
On one hand, since I is a homogeneous ideal, each fi,j ∈ I, which proves “⊇”.
On the other hand, we notice that every element h ∈ I can be written as h =
f1 g1 + · · · + fm gm for some g1 , · · · , gm ∈ k[z0 , · · · , zn ], which can be expanded as
h = f1,0 g1 + · · · + f1,d1 g1 + · · · · · · + fm,0 gm + · · · + fm,dm gm , which proves “⊆”. The
claim shows that I can be generated by finitely many homogeneous polynomials.
Conversely, assume I = (p1 , · · · , pl ) for finitely many homogeneous polynomials
p1 , · · · , pl ∈ k[z0 , · · · , zn ], with deg pi = ei . Given any polynomial q ∈ I, assume
the homogeneous decomposition of q is q = q0 + · · · + qk , where k is the degree of q.
We need to show that every qj ∈ I. Since q ∈ I, we can write q = p1 r1 + · · · + pl rl
for some r1 , · · · , rl ∈ k[z0 , · · · , zn ]. For each j with 0 6 j 6 k, by comparing the
degree j terms we get qj = p1 r1,j−e1 + · · · + pl rl,j−el , where each ri,j−ei is the sum of
all degree j − ei monomials in ri . Since I = (p1 , · · · , pl ), we conclude that qj ∈ I
for every j, which implies I is a homogeneous ideal.
(3) Given any point p ∈ V(I), we have g(p) = 0 for every homogeneous polynomial
g ∈ I. In particular, fi (p) = 0 for every i. Therefore p ∈ V(S). This proves
V(I) ⊆ V(S).
On the other hand, given any point q ∈ V(S), we have fi (q) = 0 for every i.
For any homogeneous polynomial g ∈ I, we can write g = f1 g1 + · · · + fm gm for
some g1 , · · · , gm ∈ k[z0 , · · · , zn ]. Then g(q) = f1 (q)g1 (q) + · · · + fm (q)gm (q) =
0. (Rigorously speaking, one should argue that each gi can be chosen to be a
43
 homogeneous polynomial of degree equal to deg g − deg fi , which can be achieved
by replacing each gi with its homogeneous part of degree equal to deg g − deg fi .)
This proves that V(S) ⊆ V(I).

Solution 4.3. Projective spaces are better than affine spaces!

(1) Let the two points be p = [p0 : p1 : p2 ] and q = [q0 : q1 : q2 ]. A line V(a0 z0 + a1 z1 +
a2 z2 ) passes through these two points if and only if the following system of linear
equations in a0 , a1 , a2 hold
p0 a0 + p1 a1 + p2 a2 = 0,
q0 a0 + q1 a1 + q2 a2 = 0.
Since p and q are distinct points in P2 , the two rows in the coefficient matrix
!
p0 p1 p2
q 0 q 1 q2
are linearly independent, hence the matrix has rank 2. By the theorem of rank-
nullity, the solution space to the system has dimension 1. Let v = (a0 , a1 , a2 ) be
a non-zero solution, then every solution can be written as λv for some λ ∈ k.
The solution v defines a line V(a0 z0 + a1 z1 + a2 z2 ) through the points p and q. It
remains to show the uniqueness. When λ = 0, we have λv = (0, 0, 0) which does
not define a line. For every λ ∈ k\{0}, the line V(λa0 z0 + λa1 z1 + λa2 z2 ) is the
same as V(a0 z0 + a1 z1 + a2 z2 ). Therefore the line through p and q is unique.
(2) Let the two lines by V(a0 z0 + a1 z1 + a2 z2 ) and V(b0 z0 + b1 z1 + b2 z2 ). A point
[z0 : z1 : z2 ] lies on both lines if and only if it is a solution of the following system
of linear equations in z0 , z1 , z2
a0 z0 + a1 z1 + a2 z2 = 0,
b0 z0 + b1 z1 + b2 z2 = 0.
Since the two lines are distinct, the two rows in the coefficient matrix
!
a0 a1 a2
b0 b1 b2
are linearly independent, hence the matrix has rank 2. By the theorem of rank-
nullity, the solution space to the system has dimension 1. Let w = (z0 , z1 , z2 ) be
a non-zero solution, then every solution can be written as λw for some λ ∈ k.
The solution w defines a point [z0 : z1 : z2 ] of intersection. It remains to show the
uniqueness. When λ = 0, we have λw = (0, 0, 0) which does not define a point
in P2 . For every λ ∈ k\{0}, the point [λz0 : λz1 : λz2 ] is the same as the point
[z0 : z1 : z2 ]. Therefore the two lines meet at a unique point in P2 .

Solution 4.4. Example of projective algebraic sets.

44
(1) The empty set ∅ = V(1) and the entire P1 = V(0). For any non-empty finite
subset of P1 , say {[u1 : v1 ], [u2 : v2 ], · · · , [uk : vk ]}, it can be written as V(f ) for a
homogeneous polynomial f = (v1 z0 −u1 z1 )(v2 z0 −u2 z1 ) · · · (vk z0 −uk z1 ) ∈ k[z0 , z1 ].
Therefore every set stated in the question is a projective algebraic set in P1 .
(2) Let f ∈ k[z0 , z1 ] be a homogeneous polynomial of degree d. Assume z0e be the
highest power of z0 dividing f for some e 6 d. Then we can write
f = c0 z0d + c1 z0d−1 z1 + · · · + cd−e z0e z1d−e
z1d−e
 
d z1
= z0 · c0 + c1 + · · · + cd−e d−e .
z0 z0
We consider the polynomial g(x) = c0 + c1 x + · · · + cd−e xd−e . If g is constant,
then f = c0 z0d is a product of d homogeneous polynomials of degree 1. If g is not
a constant, then it can be factored into a product of polynomials of degree 1 as
g(x) = (a1 + b1 x) · · · (ad−e + bd−e x). Then we have
   
d z1 z1
f = z0 · a1 + b1 · · · · ad−e + bd−e ·
z0 z0
e
= z0 · (a1 z0 + b1 z1 ) · · · (ad−e z0 + bd−e z1 )
which is also a product of d homogeneous polynomials of degree 1.
(3) Let X ⊆ P1 be a projective algebraic set. By Corollary 4.18, we assume X = V(S)
for a finite set S of homogeneous polynomials in k[z0 , z1 ]. If S does not have
any non-zero polynomial then X = P1 . Otherwise, assume f ∈ S is a non-zero
homogeneous polynomial of degree d. By part (2) we can write f = (a1 z0 +
b1 z1 ) · · · (ad z0 + bd z1 ) (each factor z0 can be written as 1 · z0 + 0 · z1 ). For every
p = [u : v] ∈ X, we have f (p) = 0, hence a certain factor of f vanishes at p; more
precisely, ai u + bi v = 0 for some i. Therefore p = [bi : −ai ]. There are at most d
points of this kind, hence X contains only finitely many points.

45
 5. Rational Maps

We have seen projective algebraic sets. Now we study V − I correspondence for projective
algebraic sets and maps between them.

5.1. V − I correspondence and rational maps. We have introduced the projective

Nullstellensatz. The following notion is parallel to the same one in the affine case.

Definition 5.1. A projective algebraic set X ⊆ Pn is irreducible if there does not exist
a decomposition of X as a union of two stricly smaller projective algebraic sets. An irre-
ducible projective algebraic set is also called an projective variety. A projective algebraic
set X ⊆ Pn is reducible if it is not irreducible.

Not very surprisingly, we also have the projective version of V − I correspondences. Each
row in the following diagram is a bijection:
 
 homogeneous radical ideals  ( )
  V
/ non-empty projective
I ⊆ k[z0 , z1 , · · · , zn ] o
 with I 6⊇ (z , z , · · · , z ) 
  I algebraic sets X ⊆ Pn
0 1 n O
O

 ?   ? 
 homogeneous prime ideals
   non-empty 
/
 V
 
I ⊆ k[z0 , z1 , · · · , zn ] o irreducible projective
 with I 6⊇ (z , z , · · · , z )
 
 I  algebraic sets X ⊆ Pn
 

0 1 n

We summarise the content in the diagram in words for later reference.

Proposition 5.2. Let X be a non-empty projective algebraic set in Pn and I a homoge-

neous radical ideal in k[z0 , · · · , zn ] such that (z0 , · · · , zn ) 6⊆ I. Then X = V(I) if and only
if I = I(X). In such a case, X is irreducible if and only if I is prime.

Proof. Non-examinable. Interested reader can find the proof in [Section 5.3, Reid, Un-
dergraduate Algebraic Geometry]. 

Remark 5.3. Comparing with the affine V − I correspondence, the bijection between
maximal ideals and points is no longer valid in the projective setting. In fact, the only
homogeneous maximal ideal in k[z0 , z1 , · · · , zn ] is the irrelevant ideal (z0 , z1 , · · · , zn ), which
gives the empty set in Pn as we discussed above.

In practice it is usually not easy to determine whether a projective algebraic set is irre-
ducible. It is clear that Pn is irreducible since I(Pn ) = (0) is a prime ideal. In case of
hypersurfaces, the following result usually helps.
46
Lemma 5.4. Let I = (f ) ⊆ k[z0 , z1 , · · · , zn ]. Then I is a prime ideal if and only if f is
an irreducible polynomial; I is a radical ideal if and only if f has no repeated irreducible
factors.

Proof. It was proved in Exercise 2.2. 

Now we turn to maps between projective algebraic sets.

Definition 5.5. For projective algebraic sets X ⊆ Pn and Y ⊆ Pm , a rational map

ϕ : X 99K Y is an equivalence class of expressions [f0 : · · · : fm ] satisfying

(1) f0 , · · · , fm ∈ k[z0 , · · · , zn ] are homogeneous of the same degree;

(2) [f0 (p) : · · · : fm (p)] 6= [0 : · · · : 0] for some point p ∈ X;
(3) For each point p ∈ X, if [f0 (p) : · · · : fm (p)] is defined, then it is a point in Y .

Two such expressions [f0 : · · · : fm ] and [g0 : · · · : gm ] are equivalent if [f0 (p) : · · · :
fm (p)] = [g0 (p) : · · · : gm (p)] for every p ∈ X at which both are defined.

Definition 5.6. Let ϕ : X 99K Y be a rational map between projective algebraic sets.
We say ϕ is regular at p ∈ X if [f0 (p) : · · · : fm (p)] is well-defined for some expression
[f0 : · · · : fm ] representing ϕ.

Definition 5.7. For projective algebraic sets X and Y , a morphism ϕ : X −→ Y is a

rational map which is regular at every point in X.

Remark 5.8. The condition (1) in Definition 5.5 guarantees that the image is independent
of the choice of the homogeneous coordinates of p. More precisely, suppose fi ’s are
homogeneous of degree d, and p = [a0 : · · · : an ]. For any λ 6= 0, we can also write
p = [λa0 : · · · : λan ]. Then we have by (4.3) that

[f0 (λa0 , · · · , λan ) : · · · : fm (λa0 , · · · , λan )]

= [λd f0 (a0 , · · · , an ) : · · · : λd fm (a0 , · · · , an )]
= [f0 (a0 , · · · , an ) : · · · : fm (a0 , · · · , an )].

The condition (2) in Definition 5.5 guarantees that the expression [f0 : · · · : fm ] is defined
on a non-empty subset of X.

Remark 5.9. We can view a rational function ϕ : X 99K Y as a piecewise and partially
defined function. Each expression [f0 : · · · : fm ] representing ϕ is defined on a subset of
X. Two such expressions that agree on the locus where both are defined can be glued
together to represent the same function ϕ. However, there could still be some points in
X where none of the expressions is defined.
47
Example 5.10. We check the following is a morphism
ϕ: P1 −→ P2 ; [u : v] 7−→ [u2 : uv : v 2 ].
All components of ϕ are homogeneous polynomials of degree 2. For each point p = [u :
v] ∈ P1 , either u 6= 0 or v 6= 0, hence either u2 6= 0 or v 2 6= 0. Therefore ϕ is regular on
the entire P1 . Since the target is P2 , ϕ(p) ∈ P2 is automatic for every p ∈ P1 .

Example 5.11. Consider the projective algebraic set C = V(z0 z2 − z12 ) ⊆ P2 . We check
the following is a morphism
ϕ: P1 −→ C; [u : v] 7−→ [u2 : uv : v 2 ].
We need to check everything that we checked in Example 5.10. In addition we need to
check ϕ(p) ∈ C for every p ∈ P1 . To see that we need to show [u2 : uv : v 2 ] satisfies the
defining equation of C, which is clear since (u2 )(v 2 ) − (uv)2 = 0.

Example 5.12. For the same C as in Example 5.11, we check the following is a morphism
(
[z0 : z1 ] if z0 6= 0;
ψ : C −→ P1 ; [z0 : z1 : z2 ] 7−→
[z1 : z2 ] if z2 6= 0.
As we can see ψ is defined by two expressions, whose components are all homogeneous
polynomials of degree 1. They are both defined on a non-empty subset of C; e.g. both
are defined at [1 : 1 : 1] ∈ C. It is clear that the image is always in P1 . For any
point [z0 : z1 : z2 ] ∈ C with z0 6= 0 and z2 6= 0, we have z12 = z0 z2 hence z1 6= 0. Set
λ = zz01 = zz21 6= 0, then [z0 : z1 ] = [λz0 : λz1 ] = [z1 : z2 ]. Therefore the two expressions
agree on the locus where they are both defined. To show ψ is regular everywhere on C,
we observe that for any point p = [z0 : z1 : z2 ] ∈ C, z0 and z2 cannot be both zero, since
otherwise z12 = z0 z2 = 0 and p is not a valid point. This concludes that ψ is a morphism.

Example 5.13 (Cremona transformation). We check the following is a rational map

ϕ: P2 99K P2 ; [x : y : z] 7−→ [yz : zx : xy].
All components of ϕ are homogeneous of degree 2. For every point p ∈ P2 with at least
two non-zero coordinates, ϕ(p) is a well-defined point in P2 . Hence ϕ is a rational map.

48
5.2. Dominant rational maps and birational maps. We have seen rational maps
between projective algebraic sets. We now consider the composition of two rational maps.
Suppose f : X 99K Y and g : Y 99K Z are rational maps. It is not always true that they
can be composed to get g ◦ f : X 99K Z, because the image of f could be disjoint from
the locus where g is defined. We will deal with this problem.

Definition 5.14. Let X ⊂ Pn and Y ⊂ Pm be projective varieties. A rational map

ϕ : X 99K Y is dominant if there does not exist a projective algebraic set W ( Y , such
that ϕ(p) ∈ W for every p ∈ X where ϕ is defined.

Example 5.15. We claim the morphism ϕ : P1 −→ P2 in Example 5.10 is not dominant.

To see this, we consider W = V(z0 z2 − z12 ) ⊂ P2 . We see that W ( P2 because [1 : 1 :
0] ∈ P2 \W . But for every p ∈ P1 , ϕ(p) ∈ W because (u2 )(v 2 ) − (uv)2 = 0.

The definition is handy for showing a rational map is not dominant. The following
criterion is usually more convenient for showing a rational map is dominant.

Lemma 5.16. Let ϕ : X 99K Y be a rational map between projective varieties. Suppose
there exists a projective algebraic set Z ( Y , such that every q ∈ Y \Z can be written as
q = ϕ(p) for some p ∈ X. Then ϕ is dominant.

Proof. Suppose on the contrary that there exists some projective algebraic set W ( Y
such that ϕ(p) ∈ W for every p ∈ X at which ϕ is defined. It is clear Y ⊇ W ∪ Z. For
every q ∈ Y , if q = ϕ(p) for some p ∈ X, then q ∈ W ; otherwise q ∈ Z. It follows that
Y ⊆ W ∪ Z. Therefore Y = W ∪ Z where both W and Z are projective algebraic sets
strictly smaller than Y . This contradicts the irreducibility of Y . 

Remark 5.17. In explicit examples there are usually many possible choices for W in
Definition 5.14 and Z in Lemma 5.16. You can choose the one that you find easy to use.

Example 5.18. We consider the morphism ϕ : P2 99K P2 in Example 5.13. We know

P2 is a projective variety. We claim ϕ is dominant. If not, then we can find a projective
algebraic set W ( P2 , such that ϕ(p) ∈ W for every p ∈ P2 at which ϕ is defined.

We observe that the projective algebraic set Z = V(xyz) consists of all points in P2 with
at least one zero coordinate, so Z ( P2 . For every point [a : b : c] ∈ P2 \Z, all coordinates
are non-zero. It is in the image of ϕ since
ϕ([bc : ca : ab]) = [a2 bc : ab2 c : abc2 ] = [a : b : c].
It follows from Lemma 5.16 that ϕ is dominant.

Now we answer the question asked at the beginning and give a sufficient condition for the
existence of compositions.
49
Lemma 5.19. Let ϕ : X 99K Y and ψ : Y 99K Z be rational maps between projective
varieties. If ϕ is dominant, then ψ ◦ ϕ : X 99K Z is a rational map.

Proof. Non-examinable. Interested reader can find more details in [Section 4.10, Reid,
Undergraduate Algebraic Geometry]. 

The following is another special class of rational maps.

Definition 5.20. Let ϕ : X 99K Y be a rational map between projective varieties. It is

said to be a birational map if there exists another rational map ψ : Y 99K X, such that
ψ ◦ ϕ is a well-defined rational map equivalent to the identity map on X, and ϕ ◦ ψ is a
well-defined rational map equivalent to the identity map on Y . We say a birational map
ϕ is an isomorphism if both ϕ and ψ can be chosen to be morphisms.

Remark 5.21. More precisely, the condition that ψ ◦ ϕ is equivalent to idX means that
(ψ ◦ ϕ)(p) = p for every point p ∈ X at which ψ ◦ ϕ is defined. A similar condition holds
for the other composition ϕ ◦ ψ.

Example 5.22. We claim that the rational map ϕ : P2 99K P2 discussed in Examples 5.13
and 5.18 is a birational map. Let ψ be the same rational map as ϕ, then the composition
ψ ◦ ϕ is given by the expression
(ψ ◦ ϕ)([x : y : z]) = ψ([yz : zx : xy]) = [x2 yz : xy 2 z : xyz 2 ].
For any point [x : y : z] with all coordinates nonzero, we have (ψ ◦ ϕ)([x : y : z]) = [x2 yz :
xy 2 z : xyz 2 ] = [x : y : z]. The same is true for ϕ ◦ ψ. Therefore the claim holds.

Example 5.23. We claim that the morphism ϕ : P1 −→ C in Example 5.11 is an

isomorphism, with an inverse ψ given by the morphism in Example 5.12. For any [u :
v] ∈ P1 , either u 6= 0 or v 6= 0. If u 6= 0, then u2 6= 0, hence
(ψ ◦ ϕ)([u : v]) = ψ([u2 : uv : v 2 ]) = [u2 : uv] = [u : v].
If v 6= 0, then v 2 6= 0. We can similarly have
(ψ ◦ ϕ)([u : v]) = ψ([u2 : uv : v 2 ]) = [uv : v 2 ] = [u : v].
For the other composition, take any point [z0 : z1 : z2 ] ∈ C. We showed in Example 5.12
that either z0 6= 0 or z2 6= 0. If z0 6= 0, then
(ϕ ◦ ψ)([z0 : z1 : z2 ]) = ϕ([z0 : z1 ]) = [z02 : z0 z1 : z12 ] = [z02 : z0 z1 : z0 z2 ] = [z0 : z1 : z2 ].
If z0 6= 0, we can similarly have
(ϕ ◦ ψ)([z0 : z1 : z2 ]) = ϕ([z1 : z2 ]) = [z12 : z1 z2 : z22 ] = [z0 z2 : z1 z2 : z22 ] = [z0 : z1 : z2 ].
Therefore both compositions are equivalent to identity maps hence ϕ is a rational map.
Since ϕ and ψ are both morphisms, ϕ is in fact an isomorphism.
50
Definition 5.24. Two projective varieties X and Y are said to be birational if there
exists a birational map ϕ : X 99K Y . A projective variety X is said to be rational if it is
birational to Pn for some non-negative integer n.

Definition 5.25. Two projective varieties X and Y are said to be isomorphic if there
exists an isomorphism ϕ : X −→ Y .

Remark 5.26. In fact, being birational is an equivalence relation among projective vari-
eties. This is an extremely important and profound notion in algebraic geometry. Deter-
mining which projective varieties are in the same birational equivalence class, and finding
a good representative in each class, are the fundamental questions in a major branch of
algebraic geometry, called birational geometry. As these questions are in general very
difficult, a complete answer is far from being achieved. We will see some examples later.

51
 Exercise Sheet 5

This sheet will be discussed in the exercise class on 6 November. You are welcome to
submit your solutions at the end of the exercise class or anytime earlier.
Exercise 5.1. Example: linear embedding and linear projection.

(1) Show that ϕ : P1 −→ P3 ; [z0 : z1 ] 7−→ [z0 : z1 : 0 : 0] is a morphism. Is it

dominant? (Remark: in general, for any n 6 m, there is a linear embedding
from Pn to Pm by identifying homogeneous coordinates in Pn with a subset of
homogeneous coordinates in Pm and setting the remaining coordinates 0.)
(2) Show that ψ : P3 99K P1 ; [z0 : z1 : z2 : z3 ] 7−→ [z2 : z3 ] is a rational map. Is it
dominant? (Remark: in general, for any m > n, there is a linear projection from
Pm to Pn by choosing a subset of the homogeneous coordinates in Pm .)
(3) Is the composition ψ ◦ ϕ a well-defined rational map? Explain your reason.
Exercise 5.2. Example: the cooling tower. Consider Y = V(y0 y3 − y1 y2 ) ⊆ P3 .

(1) Show that y0 y3 − y1 y2 is irreducible. Conclude that Y is a projective variety.

(2) Show that ϕ : P2 99K Y ; [x0 : x1 : x2 ] 7−→ [x20 : x0 x1 : x0 x2 : x1 x2 ] is a rational
map. Show that ϕ is dominant. (Hint: first show that each point q = [y0 : y1 :
y2 : y3 ] ∈ Y with y0 6= 0 is in the image of ϕ, then use Lemma 5.16.)
(3) Show that ψ : Y 99K P2 ; [y0 : y1 : y2 : y3 ] 7−→ [y0 : y1 : y2 ] is a rational map. Show
that ψ is dominant. (Hint: first show that each point p = [x0 : x1 : x2 ] ∈ P2 with
x0 6= 0 is in the image of ψ, then use Lemma 5.16.)
(4) Show that ϕ and ψ are birational maps. Conclude that Y is rational.
Exercise 5.3. Example: the projective twisted cubic. Consider the projective variety
Y = V(y0 y2 − y12 , y1 y3 − y22 , y0 y3 − y1 y2 ) ⊆ P3 .

(1) Show that ϕ : P1 −→ Y ; [u : v] 7−→ [u3 : u2 v : uv 2 : v 3 ] is a morphism.

(2) Show that ϕ is an isomophism by finding the inverse morphism ψ : Y −→ P1 and
computing their compositions. Conclude that Y is isomorphic to P1 .
Exercise 5.4. A famous example: blow-up at a point. Consider the projective variety
Y = V(y0 y2 − y12 , y0 y4 − y1 y3 , y1 y4 − y2 y3 ) ⊆ P4 .

(1) Show ϕ : P2 99K Y ; [x0 : x1 : x2 ] 7−→ [x20 : x0 x1 : x21 : x0 x2 : x1 x2 ] is a rational map.

(2) Show that ϕ is a birational map by finding the inverse rational map ψ : Y 99K P2
and computing their compositions. Conclude that Y is rational.
(3) Show that ψ can be chosen to be a morphism. Show that ψ is surjective on points.
Find all points q ∈ Y , such that ψ(q) = [0 : 0 : 1].
52
 Solutions to Exercise Sheet 5

Solution 5.1. Example: linear embedding and linear projection.

(1) All components are given by homogeneous polynomials of degree 1. For every point
[z0 : z1 ] ∈ P1 , we have either z0 6= 0 or z1 6= 0, hence ϕ([z0 : z1 ]) = [z0 : z1 : 0 : 0]
has at least one non-zero coordinate, hence is clearly a point in P3 . Therefore
ϕ is a morphism. It is not dominant, because for the projective algebraic set
W = V(z2 , z3 ) ⊆ P3 , we have ϕ([z0 : z1 ]) ∈ W for every point [z0 : z1 ] ∈ P1 .
(2) All components are given by homogeneous polynomials of degree 1. The map is
not defined at every point in P3 , but for every point [z0 : z1 : z2 : z3 ] ∈ P3 with
z2 6= 0 or z3 6= 0, its image ψ([z0 : z1 : z2 : z3 ]) = [z2 : z3 ] has at least one non-zero
coordinate, and is clearly a point in P1 . Therefore ψ is a rational map. To see it is
dominant, we first claim that ψ is surjective. In fact, for every point [z2 : z3 ] ∈ P1 ,
we have that [z2 : z3 ] = ψ([z0 : z1 : z2 : z3 ]) for any choice of z0 , z1 ∈ k. Since ψ
is surjective, we can apply Lemma 5.16 and choose Z = ∅ to conclude that ψ is
dominant.
(3) The composition is not well-defined because for every [z0 : z1 ] ∈ P1 , we have
(ψ ◦ ϕ)([z0 : z1 ]) = ψ([z0 : z1 : 0 : 0]) = [0 : 0] which is not a point in P1 . This
shows that ψ ◦ ϕ is nowhere well-defined, which violates the second condition in
the definition of a rational map.

Solution 5.2. Example: the cooling tower.

(1) Assume we can write y0 y3 − y1 y2 = f g for some f, g ∈ k[y0 , y1 , y2 , y3 ]. Since the

polynomial y0 y3 − y1 y2 has degree 1 in y0 , the degrees of f and g in y0 should be 0
and 1 respectively. Without loss of generality we assume f = y0 f1 + f0 and g = g0 ,
where f1 , f0 , g0 ∈ k[y1 , y2 , y3 ]. By comparing the coefficients of terms of degree 1
and 0 in y0 , we get f1 g0 = y3 and f0 g0 = −y1 y2 . Therefore g0 is a common factor
of y3 and −y1 y2 , which has to be a constant. This implies g is a constant, hence
y0 y3 − y1 y2 is irreducible. Since it is a homogeneous polynomial, V(y0 y3 − y1 y2 )
is a projective algebraic set. By Lemma 5.4, the principal ideal I = (y0 y3 − y1 y2 )
in k[y0 , y1 , y2 , y3 ] is a prime ideal. Hence by Lemma 4.17, V(y0 y3 − y1 y2 ) = V(I),
which is a projective variety by Proposition 5.2.
(2) It is clear that all components of ϕ are given by homogeneous polynomials of
degree 2. For any point p = [x0 : x1 : x2 ] ∈ P2 , if x0 is non-zero, or x1 and x2
are simultaneously non-zero, the image ϕ(p) has at least one non-zero component.
Hence ϕ is defined on a non-empty subset of P2 . To show its image is always in
Y , we find that y0 y3 − y1 y2 = (x20 )(x1 x2 ) − (x0 x1 )(x0 x2 ) = 0. Therefore ϕ is a
rational map.
53
 To show that ϕ is dominant, we observe that every point q = [y0 : y1 : y2 :
y3 ] ∈ Y with y0 6= 0 is the image of the point p = [y0 : y1 : y2 ]. Indeed,
ϕ(p) = [y02 : y0 y1 : y0 y2 : y1 y2 ] = [y02 : y0 y1 : y0 y2 : y0 y3 ] = [y0 : y1 : y2 : y3 ] = q.
Set Z = V(y0 y3 − y1 y2 , y0 ), then Z ⊆ Y , and is strictly smaller than Y (e.g.
[1 : 0 : 0 : 0] ∈ Y \Z). And every point q ∈ Y \Z is in the image of ϕ. By Lemma
5.16, ϕ is dominant.
(3) We first realise that every component of ψ is a homogeneous polynomial of degree
1. ψ is well-defined at every point q = [y0 : y1 : y2 : y3 ] ∈ Y such that y0 , y1 , y2 are
not simultaneously zero (e.g. [1 : 0 : 0 : 0] is such a point). Hence it is defined on
a non-empty subset of Y . The image ψ(q) is always a point in P2 if it is defined.
Therefore ψ is a rational map.
To show ψ is dominant, we first observe that each point p = [x0 : x1 : x2 ] ∈ P2
with x0 6= 0 is the image of the point q = [x0 : x1 : x2 : xx1 x0 2 ]. Indeed, q is a
well-defined point since x0 6= 0, and q ∈ Y since it satisfies the defining equation
of Y . The expression that defines ψ gives ψ(q) = p. If we set Z = V(x0 ), then
Z ( P2 . Since every point in P2 \Z is in the image of ψ, we conclude that ψ is
dominant by Lemma 5.16.
(4) We show that ϕ and ψ are mutually inverse rational maps. For every point p =
[x0 : x1 : x2 ] ∈ P2 at which ψ ◦ ϕ is defined, we have (ψ ◦ ϕ)(p) = ψ([x20 :
x0 x1 : x0 x2 : x1 x2 ]) = [x20 : x0 x1 : x0 x2 ] = [x0 : x1 : x2 ] = p. For every point
q = [y0 : y1 : y2 : y3 ] ∈ Y at which ϕ ◦ ψ is defined, we have (ϕ ◦ ψ)(q) = ϕ([y0 :
y1 : y2 ]) = [y02 : y0 y1 : y0 y2 : y1 y2 ] = [y02 : y0 y1 : y0 y2 : y0 y3 ] = [y0 : y1 : y2 : y3 ] = q.
Therefore ϕ and ψ are mutually inverse birational maps. It follows that Y is
birational to P2 , hence Y is rational.

Solution 5.3. Example: the projective twisted cubic.

(1) All components of ϕ are homogeneous of the same degree 3. For every point
[u : v] ∈ P1 , we have either u 6= 0 or v 6= 0, therefore either u3 6= 0 or v 3 6= 0,
hence ϕ([u : v]) = [u3 : u2 v : uv 2 : v 3 ] is always a well-defined point. To show that
ϕ([u : v]) ∈ Y , we need to check all defining polynomial of Y are satisfied. Indeed,
we have

y0 y2 − y12 = (u3 )(uv 2 ) − (u2 v)2 = 0;

y1 y3 − y22 = (u2 v)(v 3 ) − (uv 2 )2 = 0;
y0 y3 − y1 y2 = (u3 )(v 3 ) − (u2 v)(uv 2 ) = 0.

We conclude that ϕ is a morphism.

(2) We define ψ : Y −→ P1 in the following way: for every point [y0 : y1 : y2 : y3 ] ∈ Y ,
let ψ([y0 : y1 : y2 : y3 ]) = [y0 : y1 ] or [y2 : y3 ]. We first check that ψ is a morphism.
54
 Both expressions used to define ψ are given by homogeneous polynomials of
degree 1. For any point [y0 : y1 : y2 : y3 ], if either y0 or y1 is non-zero (e.g.
[1 : 0 : 0 : 0]), then the first expression applies; if either y2 or y3 is non-zero (e.g.
[0 : 0 : 0 : 1]), then the second expression applies. This shows that both expressions
are defined on non-empty subsets of Y . Moreover, for any point [y0 : y1 : y2 : y3 ],
at least one of its coordinates is non-zero, hence at least one of the expressions
can be used to compute its image under ψ, hence ψ is defined at every point in
Y . The image ψ(q) for any point q ∈ Y is clearly a point in P1 .
To show ψ is a morphism, it remains to show that, if the two expressions are
both defined at a certain point q = [y0 : y1 : y2 : y3 ] ∈ Y , then they give the same
image. For such a point q, we claim y0 6= 0; otherwise y12 = y0 y2 = 0, which implies
the first expression is invalid. Similarly, we claim y3 6= 0; otherwise y22 = y1 y3 = 0,
which implies the second expression is invalid. Therefore y1 y2 = y0 y3 6= 0, which
implies y1 6= 0 and y2 6= 0. So all coordinates of q are non-zero. For such a point q,
we have [y0 : y1 ] = [y0 y3 : y1 y3 ] = [y1 y2 : y1 y3 ] = [y2 : y3 ], hence both expressions
give the same image of q.
Finally we check that ϕ and ψ are mutually inverse to each other. Given any
point p = [u : v] ∈ P1 , we have
(
3 2 2 3 [u3 : u2 v] = [u : v];
(ψ ◦ ϕ)(p) = ψ([u : u v : uv : v ]) =
[uv 2 : v 3 ] = [u : v].
For any point q = [y0 : y1 : y2 : y3 ] ∈ Y , we notice that y0 y12 = y0 · y0 y2 = y02 y2 and
y13 = y1 · y0 y2 = y0 · y1 y2 = y0 · y0 y3 = y02 y3 . Therefore if we use the first expression
that defines ψ, we have
(ϕ ◦ ψ)(q) = ϕ([y0 : y1 ])
= [y03 : y02 y1 : y0 y12 : y13 ]
= [y03 : y02 y1 : y02 y2 : y02 y3 ]
= [y0 : y1 : y2 : y3 ].
Similarly, noticing that y22 y3 = y1 y3 · y3 = y1 y32 and y23 = y2 · y1 y3 = y1 y2 · y3 =
y0 y3 · y3 = y0 y32 , we can use the second expression that defines ψ to compute
(ϕ ◦ ψ)(q) = ϕ([y2 : y3 ])
= [y23 : y22 y3 : y2 y32 : y33 ]
= [y0 y32 : y1 y32 : y2 y32 : y33 ]
= [y0 : y1 : y2 : y3 ].
The above calculation shows that ϕ and ψ are mutually inverse to each other,
hence they are birational. Since they are both morphisms, they are isomorphisms.
We conclude that Y is isomorphic to P1 .
55
Solution 5.4. A famous example: blow-up at a point.

(1) All components of ϕ are homogeneous of degree 2. Given a point p = [x0 : x1 :

x2 ] ∈ P2 , if x0 6= 0 or x1 6= 0, then x20 6= 0 or x21 6= 0, hence at least one component
of ϕ(p) is non-zero, which implies ϕ(p) is defined. When ϕ(p) is defined, we need
to check it is a point in Y . This can be verified by

y0 y2 − y12 = (x20 )(x21 ) − (x0 x1 )2 = 0;

y0 y4 − y1 y3 = (x20 )(x1 x2 ) − (x0 x1 )(x0 x2 ) = 0;
y1 y4 − y2 y3 = (x0 x1 )(x1 x2 ) − (x21 )(x0 x2 ) = 0.

This proves ϕ is a rational map.

(2) We first write down the formula for ψ, then prove ψ is a morphism, finally show
that the two compositions of ϕ and ψ are identities.
The morphism ψ : Y −→ P2 is defined as follows: for every point q = [y0 :
y1 : y2 : y3 : y4 ], let ψ(q) = [y0 : y1 : y3 ] or [y1 : y2 : y4 ]. It is clear that both
expressions in the definition of ψ are given by homogeneous polynomials of degree
1. When y0 , y1 and y3 are not simultaneously zero (e.g. [1 : 0 : 0 : 0 : 0]), then
the first expression applies. When y1 , y2 and y4 are not simultaneously zero (e.g.
[0 : 0 : 0 : 0 : 1]), then the second expression applies. Hence both expressions
are defined on non-empty subsets of Y . For every point q ∈ Y , at least one of its
coordinates is non-zero, which means at least one of two expressions is well-defined
at q. And the image of q is clearly a point in P2 , no matter which expression we
use to compute the image.
We still need to show that the two expressions define the same image of q when
they both apply. There are a few cases to consider. Case 1: if y0 , y1 and y3 are all
non-zero, then set λ = yy10 = yy21 = yy43 . Indeed, the three fractions are equal because
of the defining equations of Y . Then [y0 : y1 : y3 ] = [λy0 : λy1 : λy3 ] = [y1 : y2 : y4 ].
Case 2: if y0 = 0, then y12 = y0 y2 = 0 implies y1 = 0. Since we assumed
the expression [y0 : y1 : y3 ] is well-defined at q, we must have y3 6= 0. Then
y2 y3 = y1 y4 = 0 implies y2 = 0. Since we assumed the expression [y1 : y2 : y4 ] is
well-defined at q, we must have y4 6= 0. Now [y0 : y1 : y3 ] = [0 : 0 : y3 ] = [0 : 0 :
y4 ] = [y1 : y2 : y4 ]. Case 3: if y0 6= 0 and y1 = 0, then y0 y2 = y12 = 0 implies
y2 = 0, and y0 y4 = y1 y3 = 0 implies y4 = 0, then the expression [y1 : y2 : y4 ] is
not defined at q. Hence this case cannot happen. Case 4: if y0 6= 0 and y1 6= 0
and y3 = 0, then y0 y4 = y1 y3 = 0 implies y4 = 0. Set λ = yy01 = yy12 . Then
[y0 : y1 : y3 ] = [y0 : y1 : 0] = [λy0 : λy1 : 0] = [y1 : y2 : 0] = [y1 : y2 : y4 ]. In
summary, we always have [y0 : y1 : y3 ] = [y1 : y2 : y4 ]. This finishes the proof of
the fact that ψ is a morphism.
56
 We compute the two compositions of ϕ and ψ. Given any point p = [x0 : x1 :
x2 ] ∈ P2 at which ψ ◦ ϕ is defined, we have
(ψ ◦ ϕ)(p) = ψ([x20 : x0 x1 : x21 : x0 x2 : x1 x2 ])
(
[x20 : x0 x1 : x0 x2 ] = [x0 : x1 : x2 ];
=
[x0 x1 : x21 : x1 x2 ] = [x0 : x1 : x2 ].
Now pick any point q = [y0 : y1 : y2 : y3 : y4 ] ∈ Y at which ϕ ◦ ψ is defined. If we
use the first expression to compute ψ(q), then we have
(ϕ ◦ ψ)(q) = ϕ([y0 : y1 : y3 ]) = [y02 : y0 y1 : y12 : y0 y3 : y1 y3 ]
= [y02 : y0 y1 : y0 y2 : y0 y3 : y0 y4 ] = [y0 : y1 : y2 : y3 : y4 ].
If we use the second expression to compute ψ(q), then we have
(ϕ ◦ ψ)(q) = ϕ([y1 : y2 : y4 ]) = [y12 : y1 y2 : y22 : y1 y4 : y2 y4 ]
= [y0 y2 : y1 y2 : y22 : y2 y3 : y2 y4 ] = [y0 : y1 : y2 : y3 : y4 ].
The above calculation shows that ϕ and ψ are mutually inverse rational maps.
Hence Y and P2 are birational to each other. It follows that Y is rational.
(3) We have proved that ψ is a morphism. We first find all points q ∈ Y such that
ψ(q) = [0 : 0 : 1]. Let q = [y0 : y1 : y2 : y3 : y4 ] ∈ Y . Then depending on
which expression we use to compute ψ(q), there are two possibilities. If [y0 : y1 :
y3 ] = [0 : 0 : 1], then y0 = y1 = 0 and y3 6= 0. From y2 y3 = y1 y4 = 0 we obtain
y2 = 0. Hence q = [0 : 0 : 0 : y3 : y4 ] for any y3 6= 0 and y4 ∈ k. Similarly, if
[y1 : y2 : y4 ] = [0 : 0 : 1], theny1 = y2 = 0 and y4 6= 0. From y0 y4 = y1 y3 = 0 we
obtain y0 = 0. Hence q = [0 : 0 : 0 : y3 : y4 ] for any y3 ∈ k and y4 6= 0. Combining
the two cases, all points q ∈ Y satisfying ψ(q) = [0 : 0 : 1] are given by points of
the form q = [0 : 0 : 0 : y3 : y4 ] where y3 and y4 not simultaneously zero.
Finally we need to show that ψ is surjective. We have seen that [0 : 0 : 1] is in
the image of ψ. For any point p = [x0 : x1 : x2 ] ∈ P2 such that p 6= [0 : 0 : 1],
we claim that p = ψ(q) for q = [x20 : x0 x1 : x21 : x0 x2 : x1 x2 ]. Indeed, when
p 6= [0 : 0 : 1], we have either x0 6= 0 or x1 6= 0. In such a case, we have checked
in part (1) that q = [x20 : x0 x1 : x21 : x0 x2 : x1 x2 ] is a well-defined point in Y . It
remains to show ψ(q) = p. If x0 6= 0, then we can use the first expression of ψ to
get ψ(q) = [x20 : x0 x1 : x0 x2 ] = [x0 : x1 : x2 ] = p. If x1 6= 0, then we can use the
second expression of ψ to get ψ(q) = [x0 x1 : x21 : x1 x2 ] = [x0 : x1 : x2 ] = p. In
summary, p is always in the image of ψ. Hence ψ is surjective.

57
 6. Function Fields

We will study rational functions on projective varieties, and pullback of rational functions
along dominant rational maps. Similar to the affine case, we will see that the field of
rational functions determines the birational class of a projective variety.

6.1. Bridge between affine and projective algebraic sets. We have seen affine and
projective algebraic sets as subsets of affine and projective spaces defined by polynomial
equations. They are related in a way that is similar to affine and projective spaces. Recall
that Pn is covered by standard affine charts Ui for i = 0, 1, · · · , n.
Proposition 6.1 (From projective to affine). Let X ⊆ Pn be a projective algebraic set,
and Ui a standard affine chart of Pn . Then Xi := X ∩ Ui is an affine algebraic set in Ui .

Proof. Without loss of generality, we prove the statement for i = 0. Assume X =

Vp (f1 , · · · , fm ) for some homogeneous polynomials f1 , · · · , fm ∈ k[z0 , · · · , zn ]. Then
p = [a0 : · · · : an ] ∈ X ∩ U0 ⇐⇒ fj (a0 , a1 · · · , an ) = 0 for each j
 
a1 an
⇐⇒ fj 1, , · · · , = 0 for each j
a0 a0
 
a1 an
⇐⇒ gj ,··· , = 0 for each j
a0 a0
where gj = fj (1, z1 , · · · , zn ). Hence Xi = Va (g1 , · · · , gm ) is an affine algebraic set. 
Remark 6.2. As in the proof, given a homogeneous polynomial (i.e. fj ), we can set one of
its variables to be 1 to obtain a (not necessarily homogeneous) polynomial (i.e. gj ). This
process is often called dehomogenisation.
Definition 6.3. Let X ⊆ Pn be a projective algebraic set, and Ui a standard affine chart
of Pn . The affine algebraic set Xi = X ∩ Ui is called a standard affine piece of X. The
decomposition X = ∪ni=0 Xi is called the standard affine cover of X.
Example 6.4. Consider the projective algebraic sets X = Vp (xy − z 2 ) ⊆ P2 . By setting
one of the variables to be 1, we obtain the three standard affine pieces of X, which are
X0 = Va (y − z 2 ) ⊆ A2 , X1 = Va (x − z 2 ) ⊆ A2 , and X2 = Va (xy − 1) ⊆ A2 .

We turn to another relation between affine and projective algebraic sets. Recall that Pn
can be understood as An together with “points at infinity”. We have also seen in Example
4.9 how to find points at infinity for a line in A2 . More generally we have
Definition 6.5 (From affine to projective). For any affine algebraic set X ⊆ An , let I =
Ia (X) and I be the ideal in k[z0 , · · · , zn ] generated by the set of homogeneous polynomials
   
deg f z1 zn
z0 f ,··· , f (x1 , · · · , xn ) ∈ I .
z0 z0
58
Then the projective algebraic set X = Vp (I) is called the projective closure of X. The
points in {[z0 : · · · : zn ] ∈ X | z0 = 0} are called points at infinity for X.

Remark 6.6. We have already used the above modification of a polynomial in Example
4.9; that is, first replacing all non-homogeneous coordinates by ratios of homogeneous
coordinates, then clearing the denominators. This process is often called homogenisation.
More precisely, for a polynomial f (x1 , · · · , xn ) ∈ k[x1 , · · · , xn ], assume deg f = d and let
f = f0 + f1 + · · · + fd−1 + fd
be its homogeneous decomposition, then the homogenisation of f is given by
 
d z1 zn
z0 · f ,··· , = z0d f0 + z0d−1 f1 + · · · + z0 fd−1 + fd .
z0 z0
Example 6.7. The projective closure of An is Pn . The points at infinity are all points in
H0 , namely, all points {[z0 : z1 : · · · : zn ] ∈ Pn | z0 = 0}.

This definition is not easy to use in general, as it requires to homogenise infinitely many
polynomials in Ia (X). The following criterion is more convenient for computations.

Proposition 6.8. Let X = Va (f ) ⊆ An be an affine hypersurface for some polynomial

f ∈ k[x1 , · · · , xn ] of degree d. Let
 
d z1 zn
f (z0 , z1 , · · · , zn ) = z0 f ,··· ,
z0 z0
be the homogenisation of f . Then X = Vp (f ).

Proof. Non-examinable. 

Remark 6.9. In general, when an affine algebraic set X is defined by more than one
polynomial, the projective closure of X is not defined by homogenisation of the generators
of Ia (X). We will see an example in Exercise 6.3.

Example 6.10. In Example 4.9, we have seen that the projective closure of Va (x2 − x1 +
1) ⊆ A2 is Vp (x2 − x1 + x0 ) ⊆ P2 , and that the projective closure of Va (x2 − x1 − 1) ⊆ A2
is Vp (x2 − x1 − x0 ) ⊆ P2 . The point at infinity for both affine algebraic sets is [0 : 1 : 1].

Example 6.11. We compute the projective closure and points at infinity for the heart
curve X = Va ((x2 + y 2 − 1)3 − x2 y 3 ). We use z for the extra variable. By Proposition
6.8, the projective closure is given by one homogeneous equation of degree 6; that is
X = Vp ((x2 + y 2 − z 2 )3 − x2 y 3 z).
To find the points at infinity, we set z = 0. Then we have (x2 + y 2 )3 = 0, hence y =
√ √
± −1x. It follows that there are two points at infinity given by [x : y : z] = [1 : −1 : 0]
√
and [1 : − −1 : 0].
59
Finally we briefly mention the relation of the two constructions. They are almost inverse to
each other, subject to some assumptions. For simplicity, we only state the correspondece
in the case of varieties. We have the following bijection. Recall that H0 = Pn \U0 .

( ) ( )
projective varieties X ⊆ Pn Y =X∩U0
/ affine varieties Y ⊆ U0 ∼
= An
o
such that X 6⊆ H0 X=Y such that Y 6= ∅

We summarise the content of the correspondence in the following result:

Proposition 6.12. There is a bijection between projective varieties in Pn which are not
contained in H0 = Pn \U0 and non-empty affine varieties in U0 , given by the mutually
inverse correspondences of taking the standard affine piece in U0 and taking the projective
closure.

Proof. Non-examinable. Interested reader can find the proof in [Section 5.5, Reid, Un-
dergraduate Algebraic Geometry] or [Section 4.3, Fulton, Algebraic Curves]. 

The importance of the two constructions relating affine and projective varieties is that
they allow us to study some properties in a relatively easier context, i.e., either affine
or projective, and deduce some similar properties in the other context. We will see two
examples in future lectures.

60
6.2. Rational functions and function fields. As we have seen, polynomials cannot
be used to define functions on projective algebraic sets. Therefore we have to find a more
flexible way to define functions on them, namely, rational functions. For simplicity, we
only consider varieties. We will first define rational functions on affine varieties, then on
projective varieties.

For any affine variety X ⊆ An , I(X) is a prime ideal in k[x1 , · · · , xn ] by Proposition 2.15.
It follows that k[X] = k[x1 , · · · , xn ]/I(X) is an integral domain by Proposition 2.12 (1).

Definition 6.13. Let X ⊆ An be an affine variety. Its function field k(X) is the field of
fractions of the integral domain k[X]. In other words,
 
ϕ
k(X) := | ϕ, ψ ∈ k[X] with ψ 6= 0 / ∼,
ψ
where ∼ is an equivalence relation defined by
ϕ1 ϕ2
∼ ⇐⇒ ϕ1 ψ2 − ψ1 ϕ2 = 0 ∈ k[X].
ψ1 ψ2
An element in k(X) is called a rational function on X.

Remark 6.14. Recall that ϕ and ψ can be given by polynomials, so we can also write
 
f
k(X) = | f, g ∈ k[x1 , · · · , xn ] with g ∈
/ I(X) / ∼,
g
where ∼ is an equivalence relation defined by
f1 f2
∼ ⇐⇒ f1 g2 − g1 f2 ∈ I(X).
g1 g2

As a quick example, x1 defines a rational function on the affine variety X = A1 . Every

polynomial function is clearly a rational function which is defined everywhere on X. But
in general, a rational function is only a partially defined function on X.

Example 6.15. The coordinate ring of the affine variety X = An is k[An ] = k[x1 , · · · , xn ].
By Definition 6.13, its function field is the field of fractions of k[x1 , · · · , xn ], usually written
as k(An ) = k(x1 , · · · , xn ). A rational function on X = An is given by a fraction of the
form fg with g 6= 0. Two such fractions are considered to define the same rational function
if and only if they can be reduced to the same form after cancelling common factors in
the numerator and the denomirator.

We want to find out how to make a similar definition on projective varieties. Recall from
equation (4.3) that a non-constant homogeneous polynomial cannot define a function
on a projective algebraic set, because its value at a point depends on the choice of the
homogeneous coordinates. However, for two homogeneous polynomials f, g ∈ k[z0 , · · · , zn ]
61
of the same degree d, their ratio fg is well-defined at any point p = [a0 : · · · : an ] provided
that g(p) 6= 0, because for any λ 6= 0, we have
f (λa0 , · · · , λan ) λd f (a0 , · · · , an ) f (a0 , · · · , an )
= d = ,
g(λa0 , · · · , λan ) λ g(a0 , · · · , an ) g(a0 , · · · , an )
f
which is independent of the choice of the homogeneous coordinates of p. Therefore g
can
be thought as a partially defined function on a projective variety. More precisely,
Definition 6.16. Let X ⊆ Pn be a projective variety. The function field of X is
 
f
k(X) := f, g ∈ k[z0 , · · · , zn ] are homogeneous of the same degree, g ∈
/ I(X) / ∼,
g
where ∼ is an equivalence relation defined by
f1 f2
∼ ⇐⇒ f1 g2 − f2 g1 ∈ I(X).
g1 g2
An element in k(X) is called a rational function on X.

It is in general not easy to explicitly compute the function field of a projective variety.
However, the following result allows one to reduce the question to the affine situation.
Lemma 6.17. Let X ⊆ An be an affine variety and X ⊆ Pn its projective closure. Then
k(X) ∼
= k(X).

Sketch of proof. (This proof is non-examinable and not covered in lectures.)

We sketch a proof. For every rational function on X

f (x1 , · · · , xn )
∈ k(X),
g(x1 , · · · , xn )
assume m = max{deg f, deg g}, then we can get a rational function on X
z0m f ( zz01 , · · · , zzn0 )
∈ k(X),
z0m g( zz10 , · · · , zzn0 )
since it is the ratio of two homogeneous polynomials of degree m. In this way we can
define a map k(X) → k(X). On the other hand, for every rational function on X
p(z0 , · · · , zn )
∈ k(X),
q(z0 , · · · , zn )
we have a rational function on X
p(1, x1 , · · · , xn )
∈ k(X).
q(1, x1 , · · · , xn )
In this way we can define a map k(X) → k(X). We need to verify that both maps
are well-defined (i.e., independent of the choice of the representative in each equivalence
class), and are homomorphisms. More work is required to check that they are inverse of
each other hence are isomorphisms. 
62
Example 6.18. By Example 6.15 we know k(An ) = k(x1 , · · · , xn ). Since Pn is the
projective closure of An by Example 6.7, we have k(Pn ) ∼
= k(x1 , · · · , xn ) by Lemma 6.17.

Recall that polynomial maps can pullback polynomial functions on affine algebraic sets.
Similarly, a dominant rational map can pullback rational functions on projective varieties.

Definition 6.19. Let ϕ : X 99K Y be a dominant rational map between projective

varieties. For every rational function g on Y , the pullback of g along ϕ is the rational
function g ◦ ϕ on X, denoted ϕ∗ (g).

Example 6.20. Consider the dominant rational map ϕ : P2 99K P2 studied in Example
x
5.18. Then the pullback of the rational function y+z ∈ k(P2 ) along ϕ is
 
∗ x yz
ϕ = ∈ k(P2 ).
y+z zx + xy

Recall that two affine algebraic sets are isomorphic if and only if they have isomorphic
coordinate rings. A similar result holds for projective varieties.

Proposition 6.21. A rational map ϕ : X 99K Y between projective varieties is a bira-

tional map if and only if ϕ is dominant and ϕ∗ : k(Y ) −→ k(X) is a field isomorphism.
Two projective varieties X and Y are birational if and only if k(X) ∼
= k(Y ).

Proof. Non-examinable. Interested reader can find the proof in [Section 5.8, Reid, Un-
dergraduate Algebraic Geometry] or [Section 6.6, Fulton, Algebraic Curves]. 

63
 Exercise Sheet 6

This sheet will be discussed in the exercise class on 13 November. You are welcome to
submit your solutions at the end of the exercise class or anytime earlier.
Exercise 6.1. Example: the cooling tower, revisited. Consider the projective algebraic
set Y = V(y0 y3 − y1 y2 ) ⊆ P3 . We know by Exercise 5.2 (1) that Y is a projective variety.

(1) Write down all standard affine pieces of Y .

(2) Explain why its function field k(Y ) ∼
= k(x1 , x2 ). (Hint: you can use the results in
Exercise 5.2 and any results mentioned in lectures.)
Exercise 6.2. Example: irreducible cubic curves.

(1) Show that the affine algebraic set X = Va (y 2 − (x − λ1 )(x − λ2 )(x − λ3 )) ⊆ A2 is

an affine variety for any λ1 , λ2 , λ3 ∈ k.
(2) Find the projective closure X ⊆ P2 of X and the points at infinity. Use Proposition
6.12 to conclude that X is a projective variety.
Exercise 6.3. A caution for the projective closure. We demonstrate Remark 6.9.

(1) Let X = Va (I) ⊆ A3 for the ideal I = (f1 , f2 ) in k[x, y, z] where f1 = y − x2

and f2 = z − x3 . Using w as the extra variable, find polynomials f1 and f2 in
k[w, x, y, z] which are the homogenisations of f1 and f2 respectively.
(2) We have seen in Exercise 2.4 (3) that I = Ia (X). Let I be the homogeneous
ideal in k[w, x, y, z] defined as in Definition 6.5. Show that y 2 − xz ∈ I but
y 2 − xz ∈
/ (f1 , f2 ). Conclude that I 6= (f1 , f2 ). Show that X 6= Vp (f1 , f2 ).

Remark: this example demonstrates that the projective closure of an affine algebraic set
X is not obtained simply by homogenising the generators of Ia (X) in general.
Exercise 6.4. Geometric interpretation of the projective closure. We consider An as the
standard affine chart U0 ⊆ Pn . Then an affine algebraic set X ⊆ An can be thought as a
subset of Pn . Prove that its projective closure X is the smallest projective algebraic set
in Pn containing X. You can follow these steps:

(1) Let W ⊆ Pn be any projective algebraic set that contains X. Let g(z0 , z1 , · · · , zn ) ∈
Ip (W ) be a homogeneous polynomial and f (z1 , · · · , zn ) = g(1, z1 , · · · , zn ) the de-
homogenisation of g. Show that f ∈ Ia (X).
(2) Let f be the homogenisation of f . Show that g = z0k · f for some non-negative
integer k. Conclude that g ∈ I where I is the homogenisation of the ideal Ia (X)
defined as in Definition 6.5. Conclude that X ⊆ W .
(3) Conclude that X is the smallest projective algebraic set in Pn containing X.
64
 Solutions to Exercise Sheet 6

Solution 6.1. Example: the cooling tower, revisited.

(1) We can get the standard affine pieces Yi = Y ∩ Ui by setting yi = 1. Therefore the
standard affine pieces of Y are given by Y0 = Va (y3 − y1 y2 ), Y1 = Va (y0 y3 − y2 ),
Y2 = Va (y0 y3 − y2 ) and Y3 = Va (y0 − y1 y2 ).
(2) We proved in Exercise 5.2 that Y is birational to P2 . By Proposition 6.21 and
Example 6.18, we have k(X) ∼= k(P2 ) ∼
= k(x1 , x2 ).

Solution 6.2. Example: irreducible cubic curves.

(1) We claim that y 2 − (x − λ1 )(x − λ2 )(x − λ3 ) is an irreducible polynomial. Use

contradiction. Assume y 2 − (x − λ1 )(x − λ2 )(x − λ3 ) = f (x, y)g(x, y) for non-
constant polynomials f, g ∈ k[x, y]. Since the left-hand side has degree 2 in y, the
degrees of f and g in y must be either 2 and 0, or 1 and 1. In the first case we
can write

y 2 − (x − λ1 )(x − λ2 )(x − λ3 ) = (y 2 f2 (x) + yf1 (x) + f0 (x)) · g(x).

Comparing coefficients of y 2 we find f2 (x)g(x) = 1, hence g(x) must be a constant.

Contradiction. In the second case we can write

y 2 − (x − λ1 )(x − λ2 )(x − λ3 ) = (yf1 (x) + f0 (x)) · (yg1 (x) + g0 (x)).

Comparing coefficients of y 2 we find f1 (x)g1 (x) = 1. Without loss of generality we

can assume f1 (x) = g1 (x) = 1. Comparing coefficients of y we find f0 (x) + g0 (x) =
0. Comparing constant terms we find −(x − λ1 )(x − λ2 )(x − λ3 ) = f0 (x)g0 (x) =
−f0 (x)2 , hence f0 (x)2 = (x−λ1 )(x−λ2 )(x−λ3 ), which is also a contradiction since
the right-hand side is not a square. So we conclude that y 2 −(x−λ1 )(x−λ2 )(x−λ3 )
is irreducible. By Lemma 5.4 we know I = (y 2 − (x − λ1 )(x − λ2 )(x − λ3 )) is a
prime ideal. By Proposition 2.15 we know X is an irreducible algebraic set, i.e.
an affine variety.
(2) Using z as the extra variable, the projective closure is given by X = Vp (y 2 z −
(x − λ1 z)(x − λ2 z)(x − λ3 z)). To find points at infinity, we set z = 0 to get
−x3 = 0. It follows that x = 0, hence the only point at infinity for X is given by
[x : y : z] = [0 : 1 : 0]. One direction of Proposition 6.12 shows that the projective
closure of a non-empty affine variety is a projective variety. Hence by part (1), we
conclude that X is a projective variety.

Solution 6.3. A caution for the projective closure.

(1) The homogenisation of f1 and f2 are given by f1 = wy − x2 and f2 = w2 z − x3 .

65
 (2) We first claim y 2 − xz ∈ I = (y − x2 , z − x3 ). This can be seen by realising
y 2 − xz = (y 2 − x4 ) + (x4 − xz) = (y − x2 )(y + x2 ) − x(z − x3 ) which is a sum of
a term with y − x2 as a factor and a term with z − x3 as a factor. Since y 2 − xz
is an element in I, by Definition 6.5, the homogenisation of y 2 − xz is an element
in I. However, since y 2 − xz is already homogeneous, its homogenisation is still
y 2 − xz. Therefore y 2 − xz ∈ I.
We prove that y 2 − xz ∈ / (f1 , f2 ). Use contradiction. Assume we can write
y −xz = f1 ·g1 +f2 ·g2 = (wy −x2 )·g1 +(w2 z −x3 )·g2 for some g1 , g2 ∈ k[w, x, y, z].
2

There are many different ways to find a contradiction. Here is one approach: when
w = x = 0 and y = z = 1, the left-hand side is 1 while the right-hand side is 0,
which is a contradiction.
Finally we prove that X 6= Vp (f1 , f2 ). There are also many different approaches
to this. Here is one of them: On one hand, we can verify directly that f1 = 0
and f2 = 0 at the point [w : x : y : z] = [0 : 0 : 1 : 1], hence [0 : 0 : 1 : 1] ∈
Vp (f1 , f2 ). On the other hand, since X = Vp (I), a point in X has to be a solution
to every homogeneous polynomial in I, in particular, it has to be a solution to
the polynomial y 2 − xz by what we just proved. We can check directly that the
point [w : x : y : z] = [0 : 0 : 1 : 1] is not a solution to this polynomial, hence
[0 : 0 : 1 : 1] ∈
/ X. This finishes the proof.
Indeed, one can see that the value of z is irrelavant. For any λ ∈ k, the point
[w : x : y : z] = [0 : 0 : 1 : λ] would do the trick.

Solution 6.4. Geometric interpretation of the projective closure.

(1) We need to show that f (p) = 0 for every point p ∈ X. Let p = (a1 , · · · , an ) ∈ X,
where a1 , · · · , an ∈ k are the non-homogeneous coordinates of p as a point in
An ∼
= U0 . Then as a point in Pn , the homogeneous coordinates of p can be given by
p = [1 : a1 : · · · : an ]. Since X ⊆ W , we have p ∈ W , therefore g(p) = 0. In other
words, g(1, a1 , · · · , an ) = 0. Therefore we have f (a1 , · · · , an ) = g(1, a1 , · · · , an ) =
0, which proves f (p) = 0. Since p is an arbitrary point in X, we conclude that
f ∈ Ia (X).
(2) We assume g is a homogeneous polynomial with deg g = d. Assume that z0k is the
highest power dividing g, then k is a non-negative integer, and each term in g has
a factor of z0k . We collect terms in g which have the degree with respect to z0 , so
we can write

g = z0k · fd−k + z0k+1 · fd−k−1 + · · · + z0d−1 · f1 + z0d · f0

where fi ∈ k[z1 , · · · , zn ] is homogeneous of degree i for i = 0, 1, · · · , d − k, and

fd−k 6= 0. Since f is the dehomogenisation of g with respect to z0 , we have

f = fd−k + fd−k−1 + · · · + f1 + f0
66
 which is precisely the homogeneous decomposition of f . We observe that deg f =
d − k. Since f is the homogenisation of f with respect to z0 , we have
f = fd−k + z0 · fd−k−1 + · · · + z0d−k−1 · f1 + z0d−k · f0 .
Comparing the formula for g and f , we find out that g = z0k · f .
Now we prove g ∈ I. Since f ∈ Ia (X) by part (1), we have f ∈ I by Definition
6.5. Since I is an ideal, we have g = z0k · f ∈ I.
Since g is an arbitrary homogeneous polynomial in Ip (W ), we conclude that
every homogeneous polynomial in the ideal Ip (W ) is a homogeneous polynomial
in the ideal I. It follows that Vp (Ip (W )) ⊇ Vp (I). We have Vp (Ip (W )) = W by
Proposition 5.2, and Vp (I) = X by Definition 6.5. Therefore W ⊇ X.
(3) We proved in parts (1) and (2) that every projective algebraic set W that contains
X must contain X. Since X itself is also a projective algebraic set that contains
X (it is X together with points at infinity), we conclude that X is the smallest
one having this property.

67
 7. Non-singularity

The non-singularity is an algebraic version of smoothness in analysis. We will find out

how to determine the non-singularity of a variety from its defining equations, and study
the related notions of tangent spaces and dimensions.

7.1. Non-singularity of irreducible hypersurfaces. In this lecture we consider the

case of irreducible hypersurfaces. We start with the affine case. Let f ∈ k[x1 , · · · , xn ] be
a non-constant irreducible polynomial. By Lemma 5.4, we know that V(f ) ⊆ An is an
affine irreducible hypersurface.

Definition 7.1. Let X = V(f ) ⊆ An be an affine irreducible hypersurface defined by a

non-constant irreducible polynomial f ∈ k[x1 , · · · , xn ]. For any point p ∈ X, we say X is
∂f
singular at p if ∂x i
(p) = 0 for every i, 1 6 i 6 n; otherwise we say X is non-singular at
p. If X is non-singular at every point p ∈ X, then we say X is non-singular ; otherwise
we say X is singular.

Remark 7.2. From Definition 7.1 we see that the singular points in X = V(f ) form an
∂f ∂f
affine algebraic set Xsing = V(f, ∂x 1
, · · · , ∂xn
) ⊆ X. To find all singular points, we just
need to solve the system of equations given by f and all its partial derivatives.

Example 7.3. Consider the affine variety X = V(f ) ⊆ A2 where f = x3 + y 3 − 3xy. To

find all singular points, we need to solve the system of equations given by f = x3 + y 3 −
3xy = 0 and the partial derivatives ∂f∂x
= 3x2 − 3y = 0 and ∂f∂y
= 3y 2 − 3x = 0. From the
two partial derivatives we get x = y 2 = x4 , therefore x(x3 − 1) = 0, which implies x = 0
or x3 = 1. When x = 0, we have y = 0. It is clear that (x, y) = (0, 0) is a solution to the
system of equations. When x3 = 1, we have x3 + y 3 − 3xy = x3 + x6 − 3x3 = −1 6= 0.
Contradition. Therefore the only point at which X is singular is (0, 0).

The following result shows that X = V(f ) cannot be singular everywhere. Recall that we
always assume the underlying field k is an algebraically closed field of charasteristic 0.

Theorem 7.4. Let X = V(f ) ⊆ An be an affine hypersurface defined by a non-constant

irreducible polynomial f ∈ k[x1 , · · · , xn ]. Then the set of non-singular points in X is
non-empty.

Proof. The set of singular points in X is given by

 
∂f ∂f
Xsing = V f, ,··· , ⊆ X.
∂x1 ∂xn
∂f
Suppose on the contrary that Xsing = X, then ∂xi
∈ I(X) for every i.
68
Since f is an irreducible polynomial, (f ) is a prime ideal by Lemma 5.4. It follows by
Proposition 2.9 that I(X) = (f ). Therefore for every i, we have
∂f
= f · gi
∂xi
∂f
for some gi ∈ k[x1 , · · · , xn ]. Assume f has degree di in xi . If di > 0, then ∂xi
has degree
di − 1 in xi , while f · gi has degree at least di in xi . Contradiction. Therefore di = 0. In
other words, xi does not occur in f . Since this holds for every i, f must be a constant
polynomial. Contradiction. This finishes the proof of existence of non-singular points in
X = V(f ). 

Definition 7.5. Let X = V(f ) ⊆ An be an affine irreducible hypersurface defined by a

non-constant irreducible polynomial f ∈ k[x1 , · · · , xn ]. For any point p = (a1 , · · · , an ) ∈
X, the tangent space of X at p is the affine variety
 
∂f ∂f
Tp X := V (p) · (x1 − a1 ) + · · · + (p) · (xn − an ) ⊆ An .
∂x1 ∂xn
Example 7.6. Following Example 7.3, we compute the tangent spaces of X at two points
p1 = ( 43 , 32 ) and p2 = (0, 0). Recall that ( ∂f , ∂f ) = (3x2 − 3y, 3y 2 − 3x). It is easy to
∂x ∂y
compute that ( ∂f (p ), ∂f (p1 )) = ( 10
∂x 1 ∂y 3
, − 38 ) and ( ∂f (p ), ∂f (p2 )) = (0, 0). Therefore
∂x 2 ∂y
    
10 4 8 2
Tp1 X = V x− − y− = V(5x − 4y − 4),
3 3 3 3
Tp2 X = V (0 · (x − 0) + 0 · (y − 0)) = A2
are the tangent spaces of X at p1 and p2 respectively.

Remark 7.7. In Definition 7.5, when p is singular point of X, the defining equation of
Tp X is a zero polynomial hence Tp X = An , which has dimension n as a vector space over
k; when X is non-singular at p, the tangent space Tp X is a shift of the vector subspace
∂f ∂f
V ∂x 1
(p) · x1 + · · · + ∂xn
(p) · xn , which has dimension n − 1. Therefore we can say, the
irreducible hypersurface X ⊆ An is non-singular at p if and only if dim Tp X = n − 1;
X is singular at p if and only if dim Tp X > n − 1. We will generalise this conclusion to
arbitrary affine varieties in next lecture.

Finally we briefly mention the case of projective irreducible hypersurfaces. Let f ∈

k[z0 , · · · , zn ] be a non-constant homogeneous irreducible polynomial. By Lemma 5.4, we
know that V(f ) ⊆ Pn is a projective irreducible hypersurface.

Definition 7.8. Let X = V(f ) ⊆ Pn be a projective irreducible hypersurface defined

by a non-constant homogeneous irreducible polynomial f ∈ k[z0 , · · · , zn ]. For any point
p ∈ X, we say X is singular at p if the affine hypersurface Xi = X ∩ Ui is singular at
p for any standard affine piece Xi containing p; otherwise we say X is non-singular at
p. The tangent space Tp X of X at p is the projective closure of Tp Xi for any standard
69
affine piece Xi containing p. If X is non-singular at every point p ∈ X, then we say X is
non-singular ; otherwise we say X is singular.

Remark 7.9. A point p ∈ X could be contained in several standard affine pieces of X. To

check whether X is singular at p, and compute the tangent space of X at p, it suffices to
choose one standard affine piece of X containing p. The result does not depend on the
choice of the standard affine piece.

Example 7.10. Consider the projective variety Y = Vp (f ) ⊆ P2 where f = x3 +y 3 −3xyz.

The standard affine piece Y ∩ U2 is the affine variety in Examples 7.3 and 7.6. The results
in the two examples imply that Y is non-singular at p1 = [ 43 : 32 : 1] = [4 : 2 : 3] and
singular at p2 = [0 : 0 : 1]. Moreover, the tangent spaces of Y at p1 and p2 are given by
Tp1 Y = Vp (5x − 4y − 4z) and Tp2 Y = P2 .

70
7.2. Non-singularity of varieties. We generalise our discussion from last time and
study non-singularity of varieties. Similarly, we first consider the case of affine varieties.
for any affine variety X, we know by Corollary 1.14 that I(X) is finitely generated.
Definition 7.11. Let X ⊆ An be a non-empty affine variety. Assume I(X) = (f1 , · · · , fm )
for some f1 , · · · , fm ∈ k[x1 , · · · , xn ]. For any point p = (a1 , · · · , an ) ∈ X, the tangent
space of X at p is the affine variety
m n
!
\ X ∂fi
Tp X := V (p) · (xj − aj ) ⊆ An .
i=1 j=1
∂x j

Remark 7.12. We can view the tangent space Tp X as a shift of the linear subspace
m n
!
\ X ∂fi
V (p) · xj ⊆ An
i=1 j=1
∂x j

which is the null space of the matrix

 
∂fi
Mp := (p) .
∂xj 16i6m,16j6n

By the rank-nullity theorem, the dimension of Tp X is given by

dim Tp X = n − rank Mp .
Definition 7.13. Let X ⊆ An be a non-empty affine variety. The dimension of X is
dim X = min{dim Tp X | p ∈ X}.
For any point p ∈ X, we say X is singular at p if dim Tp X > dim X; we say X is non-
singular at p if dim Tp X = dim X. If X is non-singular at every point p ∈ X, then we
say X is non-singular ; otherwise we say X is singular.
Remark 7.14. By Remark 7.7, we find that Definition 7.1 for hypersurfaces is consistent
with the more general Definition 7.11. We also point out: although our definition of
tangent spaces and dimension involve a choice of generators in I(X), they are in fact
independent of the choice. In other words, different choices of generators in I(X) always
give the same tangent spaces and dimension.
Example 7.15. As a simple example, let X = An , then I(X) = {0}. For any point p ∈ X,
it is clear that Mp is a zero matrix and Tp X = An . Therefore dim Tp X = n−rank Mp = n.
It follows that dim X = n, and X is non-singular.
Example 7.16. Remark 7.7 together with Theorem 7.4 shows that dim X = n − 1 for
any irreducible hypersurface X ⊆ An .
Example 7.17. As another simple example, let X = {p} ⊆ An be a single point set,
where p = (a1 , · · · , an ). By Exercise 2.3 we know I(X) = (x1 − a1 , · · · , xn − an ). Then
we have Mp = In is the identity matrix, and that Tp X = ∩ni=1 V(xi − ai ) = {p}. It follows
that dim X = 0 and X is non-singular.
71
Now we consider projective varieties. Similar to the hypersurface case, the non-singularity
and dimension of a projective variety can be reduced to its standard affine pieces.

Definition 7.18. Let X ⊆ Pn be a non-empty projective variety. The dimension of X

is defined to be dim Xi for any non-empty standard affine piece Xi = X ∩ Ui , denoted
dim X. For any point p ∈ X, we say X is singular at p if Xi is singular at p for any
standard affine piece Xi = X ∩ Ui containing p; otherwise we say X is non-singular at p.
If X is non-singular at every point p ∈ X, then we say X is non-singular; otherwise we
say X is singular.

Remark 7.19. The dimension of a projective variety can be computed on any of its non-
empty standard affine piece. Similarly whether X is singular at p can be computed on
any of its standard affine piece containing p. Different standard affine pieces always give
the same answer. However, in order to find all singular points in a projective variety X,
we need to work with more than one standard affine piece to avoid missing any point.

A very surprising property of the dimension is its intrinsic nature.

Theorem 7.20. Let X and Y be (affine or projective) varieties. If k(X) ∼

= k(Y ), then
dim X = dim Y .

Proof. Non-examinable. Interested reader can find the proof in [Sections 6.7 and 6.8,
Reid, Undergraduate Algebraic Geometry] or [Section 6.5, Fulton, Algebraic Curves]. 

Remark 7.21. Theorem 7.20 shows that the dimension of a variety X only depends on its
function field k(X). In particular, by Proposition 6.21, if two projective varieties X and
Y are birational, then dim X = dim Y .

Definition 7.22. An affine (resp. projective) algebraic curve C ⊆ An (resp. C ⊆ Pn ) is

a finite union of affine (resp. projective) varieties of dimension 1.

Finally we look at a comprehensive example.

Example 7.23. Consider the projective variety X = Vp (w +x+y +z, w2 +x2 +y 2 +z 2 ) ⊆

P3 . We will show that X is a non-singular curve. By Definition 7.18, we need to show
every standard affine piece of X is non-singular of dimension 1.

We look at the standard affine piece X0 = X ∩ U0 = {p = [w : x : y : z] ∈ X | w 6= 0}.

Then X0 = Va (1 + x + y + z, 1 + x2 + y 2 + z 2 ) ⊆ A3 . To use Definition 7.11, we need to
know that Ia (X0 ) = (1 + x + y + z, 1 + x2 + y 2 + z 2 ). This can be verified by showing the
ideal (1 + x + y + z, 1 + x2 + y 2 + z 2 ) is prime and applying Proposition 2.9 (1). We skip
the proof of this step and simply assume it is true.
72
For any point p ∈ X0 , we have
!
1 1 1
Mp = .
2x 2y 2z
Since there are two rows in Mp and the first row is non-zero, we know that 1 6 rank Mp 6 2
for every point p ∈ X0 . We claim that rank Mp = 2 for every p ∈ X0 . Otherwise, assume
rank Mp = 1 for some p ∈ X0 , then the two rows must be proportional hence x = y = z.
However p ∈ X0 implies that 1 + x + y + z = 0 and 1 + x2 + y 2 + z 2 = 0, which become
1 + 3x = 0 and 1 + 3x2 = 0. It is easy to see that they do not have common solutions.
Hence such a point p does not exist. It follows that dim Tp X0 = 3 − rank Mp = 1 for every
p ∈ X0 . That means X0 is non-singular, and dim X = dim X0 = 1.

Since the defining equations of X are completely symmetric with respect to all variables,
the same computation would show that all other standard affine pieces of X are non-
singular. Therefore X is a non-singular curve.

73
 Exercise Sheet 7

This sheet will be discussed in the exercise class on 20 November. You are welcome to
submit your solutions at the end of the exercise class or anytime earlier.
Exercise 7.1. Examples of affine varieties. Find all singular points on the affine variety
X, if there is any. In parts (1) – (3), you can assume the polynomial f is irreducible. In
part (4), we know the two given polynomials generate Ia (X) by Exercise 2.4.

(1) X = V(f ) ⊆ A2 for f = (x2 + y 2 )3 − 4x2 y 2 ∈ k[x, y].

(2) X = V(f ) ⊆ A3 for f = xy 2 − z 2 ∈ k[x, y, z].
(3) X = V(f ) ⊆ A3 for f = xy + x3 + y 3 ∈ k[x, y, z].
(4) X = V(f, g) ⊆ A3 for f = y − x2 ∈ k[x, y, z] and g = z − x3 ∈ k[x, y, z].
Exercise 7.2. Example of projective varieties. Show that the projective variety X =
V(f ) ⊆ P2 for f = xy − z 2 ∈ k[x, y, z] is non-singular. Although one can achieve this
by showing all three standard affine pieces are non-singular, it is not necessary to check
every individual piece. Follow these steps for an easier approach.

(1) Show that the standard affine piece X0 = X ∩ U0 is non-singular.

(2) Find out all points in X\X0 . For each point p ∈ X\X0 , use a standard affine
piece of X that contains p to show X is non-singular at p.
(3) Using this method to find all singular points on the projective variety V(f ) ⊆ P2
for f = x3 z + x2 yz + y 3 z + x4 + y 4 . You do not need to prove the irreducibility of
any polynomial in this problem – just assume they are.
Exercise 7.3. Example: plane cubics. Find all singular points on the projective variety
V(f ) ⊆ P2 where f = y 2 z − (x − λ1 z)(x − λ2 z)(x − λ3 z) for some λ1 , λ2 , λ3 ∈ k, if there
is any. You do not need to prove irreducibility of any polynomial in this problem.

(1) λ1 , λ2 and λ3 are distinct.

(2) λ1 = λ2 6= λ3 .
(3) λ1 = λ2 = λ3 .
Exercise 7.4. Example: projective twisted cubic. Consider the projective variety Y =
Vp (y0 y2 − y12 , y1 y3 − y22 , y0 y3 − y1 y2 ) ⊆ P3 . Follow the method in Example 7.23 to

(1) Determine whether Y is non-singular or singular.

(2) Compute the dimension of Y .

Remark: For any standard affine piece Yi of Y , you can assume without proof that the
dehomogenisation of the above three polynomials generate Ia (Yi ).
74
 Solutions to Exercise Sheet 7

Solution 7.1. Examples of affine varieties.

(1) The singular points are defined by f = 0 and the two partial derivatives ∂f ∂x
=
∂f ∂f 2 2 2 2 2 2 2 2
∂y
= 0. We have ∂x = 6x(x + y ) − 8xy = 2x · (3(x + y ) − 4y ) and
∂f
∂y
= 6y(x2 + y 2 )2 · 2y − 8x2 y = 2y · (3(x2 + y 2 )2 − 4x2 ). If x = 0 or y = 0,
then f = 0 forces x = y = 0. The point (0, 0) satisfies all equations hence is a
singular point. If neither x nor y is 0, then we have 3(x2 + y 2 )2 = 4x2 = 4y 2 , hence
3(x2 + x2 )2 = 4x2 which implies x2 = 31 = y 2 . But then f = ( 31 + 13 )3 − 4 · 31 · 13 6= 0.
Therefore the only singular point is (0, 0).
(2) The singular points are defined by f = xy 2 − z 2 = 0, and ∂f
∂x
= y 2 = 0, ∂f
∂y
= 2xy =
∂f
0, ∂z = −2z = 0. From the second and fourth equations we have y = z = 0.
No matter what value x takes, (x, y, z) = (x, 0, 0) always satisfies all the four
equations. Therefore the singular points of V(f ) are all points of the form (x, 0, 0).
(3) The singular points are given by f = xy + x3 + y 3 = 0, and ∂f ∂x
= y + 3x2 = 0,
∂f
∂y
= x + 3y 2 = 0, ∂f
∂z
= 0. From ∂f∂x
= ∂f∂y
= 0 we get x = −3y 2 = −27x4 , hence
1
x = 0 or x3 = − 27 . If x = 0, then f = 0 forces y = 0. It is clear that every
point of the form (x, y, z) = (0, 0, z) is a solution to all the required equations
1
hence is a singular point on V(f ). If x 6= 0, then x3 = − 27 . Then we have
3 3 2 3 2 3 3 3 6 1 1 1 1
f = xy+x +y = x(−3x )+x +(−3x ) = −3x +x −27x = 9 − 27 − 27 = 27 6= 0.
Contradiction. Therefore (x, y, z) = (0, 0, z) are the only singular points of V(f ).
(4) At every point p = (x, y, z) ∈ X, we consider the matrix Mp given by the partial
derivatives !
−2x 1 0
Mp = .
−3x2 0 1
It is clear that the two rows of Mp are linearly independent, therefore rank Mp = 2
for every p ∈ X. It follows that dim Tp X = 3 − rank Mp = 1 for every p ∈ X.
Therefore dim X = 1 and dim Tp X = dim X for every p ∈ X. By Definition 7.13,
X is non-singular at every point p ∈ X.

Solution 7.2. Example of projective varieties.

(1) The standard affine piece X0 = X ∩ U0 is given by setting x = 1 in f . Hence

X0 = V(f0 ) where f0 = y − z 2 . For any point (y, z) ∈ X0 , ∂f
∂y
0
= 1 which never
vanishes. Therefore X0 does not have any singular point, hence is non-singular.
(2) The set of points in X\X0 is given by {[x : y : z] ∈ X | x = 0}. When x = 0,
f = xy − z 2 = 0 implies z = 0. Hence the only point in X\X0 is p = [x : y :
z] = [0 : 1 : 0]. This point is in the standard affine piece X1 = X ∩ U1 because
its y-coordinate is non-zero. The standard affine piece X1 is obtained by setting
75
 y = 1 hence X1 = Va (f1 ) where f1 = x − z 2 . The point p = [0 : 1 : 0] has
non-homogeneous coordinates p = (0, 0) in the standard affine piece X1 . To check
whether X1 is singular at p = (0, 0), we need to compute the partial derivatives
of the defining equation f1 . Notice that ∂f
∂x
1
= 1 which does not vanish at p.
We conclude that p is a non-singular point of X1 , hence by Definition 7.8, p is a
non-singular point of X.
Parts (1) and (2) together show that X = V(xy − z 2 ) ⊆ P2 is non-singular.
(3) We first consider the standard affine piece X0 = X ∩ U0 . By setting x = 1, we get
X0 = V(f0 ) ⊆ A2 where f0 = z + yz + y 3 z + 1 + y 4 . To find singular points in X0 ,
we need to consider the equations

f0 = z + yz + y 3 z + 1 + y 4 = 0;
∂f0
= z + 3y 2 z + 4y 3 = 0;
∂y
∂f0
= 1 + y + y 3 = 0.
∂z

We now solve the system. From the first equation we observe that f0 = z(1 + y +
y 3 ) + (1 + y 4 ) = 0. Together with the third equation we find that 1 + y 4 = 0. I
claim that the two equations 1 + y + y 3 = 0 and 1 + y 4 = 0 do not have a common
solution for y. There are many ways to prove the claim. One possible way is to
use the Euclidean division. We divide y 4 + 1 by y 3 + y + 1 to get

y 4 + 1 = y(y 3 + y + 1) − (y 2 + y − 1),

which implies y 2 + y − 1 = 0. We further divide y 3 + y + 1 by y 2 + y − 1 to get

y 3 + y + 1 = (y − 1)(y 2 + y − 1) + 3y,

which implies 3y = 0 hence y = 0. Therefore if the two equations have a common

solution for y then we must have y = 0, which is not a solution. This proves the
claim, which implies that X0 is non-singular.
Finally we need to check whether the points in X\X0 are singular points. To
find all points in X\X0 , we set x = 0 in f = 0. Then we get y 3 z + y 4 = 0, which
implies y = 0 or y + z = 0. Therefore there are two points in X\X0 , given by
p1 = [0 : 0 : 1] and p2 = [0 : −1 : 1] respectively. To check whether they are
singular points, we need to find a standard affine piece which contain them. Since
the z-coordinates of p1 and p2 are non-zero, we can choose X2 = X ∩ U2 . The
standard affine piece X2 = V(f2 ) where f2 = x3 + x2 y + y 3 + x4 + y 4 . The non-
homogeneous coordinates of p1 and p2 are given by p1 = (0, 0) and p2 = (0, −1)
76
 respectively. The partial derivatives of f2 are
∂f2
= 3x2 + 2xy + 4x3 ;
∂x
∂f2
= x2 + 3y 2 + 4y 3 .
∂y

It is easy to see that at the point p1 = (0, 0), we have f2 (p1 ) = ∂f

∂x
2
(p1 ) = ∂f
∂y
2
(p1 ) =
0. Therefore p1 is a singular point on X2 . At the point p2 = (0, −1), we have
∂f2
∂y
(p2 ) = −1 6= 0. Therefore p2 is a non-singular point on X2 . By Definition 7.8,
the only singular point of X is p1 = [0 : 0 : 1].

Solution 7.3. Example: plane cubics. There are three cases to deal with in this question.
Most of the calculations are the same in all the three cases. First of all we look at a
standard affine piece of X = V(f ) ⊆ P2 . You can choose any standard affine piece of X
to start with. For example, we choose the standard affine pice X2 = X ∩ U2 , which is
given by setting z = 1 in f . Therefore we have

X2 = V(y 2 − (x − λ1 )(x − λ2 )(x − λ3 )) ⊆ A2 .

To find the singular points on X2 , we need to solve the system

y 2 − (x − λ1 )(x − λ2 )(x − λ3 ) = 0;
−(x − λ2 )(x − λ3 ) − (x − λ1 )(x − λ3 ) − (x − λ1 )(x − λ2 ) = 0;
2y = 0.

The third equation implies y = 0, then the first equation implies x = λ1 or λ2 or λ3 . Now
there is some difference in the three cases.

(1) If λ1 , λ2 and λ3 are distinct, then it is clear that none of them is a solution to the
second equation. Therefore X2 is non-singular in this case.
(2) If two of the three are equal, say, λ1 = λ2 6= λ3 , then it is clear that x = λ1 (or
λ2 ) is a solution to the second equation while x = λ3 is not a solution. Therefore
X2 has a singular point (λ1 , 0), which has homogeneous coordinates [λ1 : 0 : 1] as
a point in X.
(3) If all the three are equal, then x = λ1 (or λ2 or λ3 ) is a solution to the sec-
ond equation. Therefore X2 has a singular point (λ1 , 0), which has homogeneous
coordinates [λ1 : 0 : 1] as a point in X.

It remains to consider the points in X\X2 . To find these points we set z = 0 in the
equation f = 0. We get −x3 = 0 hence x = 0. Therefore the only point in X\X2 is
p = [x : y : z] = [0 : 1 : 0]. Since the y-coordinate of p is non-zero, it is a point in the
standard affine piece X1 = X ∩ U1 , given by the non-homogeneous coordinates p = (0, 0).
77
To write down the defining polynomial for X1 we set y = 1 and get X1 = V(f1 ) ⊆ A2
where
f1 = z − (x − λ1 z)(x − λ2 z)(x − λ3 z).
Its partial derivative with respect to z is given by
∂f1
= 1 + λ1 (x − λ2 z)(x − λ3 z) + λ2 (x − λ1 z)(x − λ3 z) + λ3 (x − λ1 z)(x − λ2 z).
∂z
It is clear that at the point p = (0, 0), we have ∂f∂z
1
(p) = 1 6= 0. Therefore p = (0, 0) is a
non-singular point of X1 , hence p = [0 : 1 : 0] is a non-singular point of X. This holds in
all the three cases. We have the following conclusion:

(1) If λ1 , λ2 and λ3 are distinct, X is non-singular.

(2) If two of the three are equal, say, λ1 = λ2 6= λ3 , then X has a unique singular
point [λ1 : 0 : 1].
(3) If all the three are equal, then X has a unique singular point [λ1 : 0 : 1].

Solution 7.4. Example: projective twisted cubic. We first consider the standard affine
piece Y0 = Y ∩ U0 . By settin z0 = 1 we get

Y0 = Va (y2 − y12 , y1 y3 − y22 , y3 − y1 y2 ).

To find the dimension of the tangent space at any point p = (y1 , y2 , y3 ), we consider the
matrix of partial derivatives:
 
−2y1 1 0
Mp =  y3 −2y2 y1  .
 
−y2 −y1 1

We need to find rank Mp . First we compute the determinant of Mp :

det Mp = 4y1 y2 − y1 y2 − y3 − 2y13 = 4y1 y2 − y1 y2 − y1 y2 − 2y1 y2 = 0.

Therefore rank Mp 6 2. Notice that the first and third rows of Mp are linearly independent
(or the second and third columns). Therefore rank Mp = 2, which implies dim Tp Y0 = 1
at every p ∈ Y0 . It follows that Y0 is non-singular and dim Y = dim Y0 = 1.

Now we consider the points in Y \Y0 . Let p = [y0 : y1 : y2 : y3 ] be such a point, then y0 = 0,
which implies y12 = y0 y2 = 0 and y22 = y1 y3 = 0. Therefore the only point p ∈ Y \Y0 is
given by p = [0 : 0 : 0 : 1]. To determine whether p is a singular point, we need to look at
the standard affine piece Y3 = Y ∩ U3 . We could perform a similar calculation as above
to show that Y3 is non-singular. More precisely, we have

Y3 = Va (y0 y2 − y12 , y1 − y22 , y0 − y1 y2 ).

78
For any point q = (y0 , y1 , y2 ) ∈ Y3 , the matrix
 
y2 −2y1 y0
Mq =  0 1 −2y2  .
 
1 −y2 −y1
We notice that
det Mq = −y1 y2 + 4y1 y2 − y0 − 2y23 = −y1 y2 + 4y1 y2 − y1 y2 − 2y1 y2 = 0.
Therefore rank Mq 6 2. Moreover the second and the third rows are linearly independent,
hence rank Mq = 2 for every q ∈ Y3 . It follows that Y3 is non-singular. To summarise, Y
is non-singular and has dimension 1.

79
 8. Algebraic Curves

We study plane curves of degree up to 3.

8.1. Lines and conics. From now on we focus on plane curves.

Definition 8.1. A plane curve is a hypersurface C = V(f ) ⊆ P2 for some non-constant

homogeneous polynomial f ∈ k[x, y, z] without repeated factors. The degree of C is
defined to be deg f . Plane curves of degrees 1, 2, 3 and 4 are called lines, conics, cubics
and quartics respectively.

Example 8.2. Let [x : y : z] be the homogeneous coordinates in P2 . Every line is defined

by a polynomial f (x, y, z) = ax + by + cz for some a, b, c ∈ k which are not simultaneously
zero. A line is always irreducible.

Example 8.3. Every conic is defined by a non-zero polynomial of the form g(x, y, z) =
ax2 + 2bxy + cy 2 + 2dxz + 2eyz + f z 2 . It is sometimes more convenient to write it in the
matrix form   
  a b d x
g(x, y, z) = x y z  b c e  y  .
  
d e f z
We consider the factorisation of g into irreducibles. By Exercise 4.2 (1), each irreducible
factor of g is also homogeneous. There are three cases:

(1) If g is an irreducible polynomial, then V(g) is an irreducible conic;

(2) If g = g1 g2 for coprime irreducible homogeneous polynomials g1 and g2 of degree
1, then V(g) = V(g1 ) ∪ V(g2 ) is the union of two distinct lines;
(3) If g = g02 for an irreducible homogeneous polynomial g0 of degree 1. Since g has
repeated factors, V(g) is not a conic. Instead, V(g) = V(g0 ) is a line. However,
sometimes it is convenient to say that g defines a “double line”, just to indicate
that the factor g0 is repeated.

Definition 8.4. Let [x : y : z] be the homogeneous coordinates of any point in P2 . For a

fixed 3 × 3 invertible matrix A, define a new set of coordinates [x0 : y 0 : z 0 ] by the equation
   
x0 x
 0
y  = A y  .
 
z0 z
This is called the linear change of homogeneous coordinates defined by A.

Remark 8.5. Why it makes sense: Multiplication of [x : y : z] by any scalar λ ∈ k\{0}

results in the multiplication of [x0 : y 0 : z 0 ] by the same scalar λ, and x0 , y 0 , z 0 cannot be
all 0 unless x, y, z are all zero since A is nonsingular. So [x0 : y 0 : z 0 ] are a new system of
80
homogeneous coordinates for points in the projective plane. Why we care: We can often
reduce the defining equation of a curve to a very simple form by choosing a new system
of coordinates.

Lemma 8.6. Every line in P2 can be written as V(x) after a suitable linear change of
homogeneous coordinates. A non-zero homogeneous polynomial g(x, y, z) = ax2 + 2bxy +
cy 2 + 2dxz + 2eyz + f z 2 defines an irreducible conic if and only if the matrix
 
a b d
G = b c e 
 
d e f
has rank 3; g defines a union of two lines if and only if G has rank 2; g defines a double
line if and only if G has rank 1. Every irreducible conic in P2 can be written as V(xz − y 2 )
after a suitable linear change of homogeneous coordinates.

Proof. Non-examinable. The proof follows from the Gram-Schmidt orthogonalisation in

linear algebra. 

Proposition 8.7. A line (or an irreducible conic) is isomorphic to P1 , hence is rational.

Proof. By Lemma 8.6, we can assume the line is V(x) and the conic is V(xz − y 2 ) without
loss of generality. The case of a line is easy; we leave the details to the reader. The case
of a conic was proved in Example 5.23. 

The following results are special cases of a famous theorem.

Theorem 8.8. Let L be a line and D a plane curve of degree d. If L is not a component
of D, then L ∩ D has at most d distinct points. When counting with multiplicities, L and
D meet in precisely d points.

Proof. Assume L = V(ax+by +cz) where a, b and c are not simultaneously zero. Without
loss of generality, we can assume c 6= 0. Then a point p ∈ L can be written as p = [x :
y : − ac x − cb y]. Assume D = V(f ) where f (x, y, z) is a non-zero homogeneous polynomial
of degree d. Then p ∈ D if and only if f x, y, − ac x − cb y = 0. The left-hand side is a

homogeneous polynomial of degree d in x and y. By Exercise 4.4 (2), it can be factored

into a product of d homogeneous factors of degree 1 as
 
a b
f x, y, − x − y = (r1 x + s1 y) · · · (rd x + sd y) = 0.
c c
Each factor ri x + si y determines a solution [x : y] = [−si : ri ] which gives point pi = [−si :
ri : ac si − cb ri ] ∈ L ∩ D. Some of these points may be the same, so L and D meet in at
most d points. When counting with the number of times each distinct point occurs as a
solution, we have precisely d points. 
81
Remark 8.9. If p ∈ L ∩ D occurs m times as a solution, then we say L and D meet at
p with multiplicity m. The current proof provides a systematic method to compute all
intersection points of a line and a curve with multiplicities.

Remark 8.10. We briefly explain what it means by saying L is not a component of D. For
example, if D is a conic, it could be the union of two lines. If L happens to be one of them,
then L and D meet in more than d points, indeed, infinitely many points. The theorem
indicates that if L and D meet in more than d points, then L must be a component of D.

Proposition 8.11. Let D be an irreducible non-singular plane curve of degree d > 2. For
any point p ∈ D, the tangent line Tp D and D meet at p with multiplicity at least 2.

Proof. Non-examinable. But we will see some examples in exercises. 

Theorem 8.12. Let C be a conic and D a plane curve of degree d. If C and D have
no common component, then C ∩ D has at most 2d distinct points. When counting with
multiplicities, C and D meet in precisely 2d points.

Proof. Similar to the proof of Theorem 8.8. We leave it as an exercise. 

The more general version of the theorem is the following

Theorem 8.13 (Bézout’s Theorem). Let D1 and D2 be plane curves of degree d1 and d2
respectively. Assume D1 and D2 have no common component, then D1 and D2 meet in
at most d1 d2 distinct points. When these points are counted with multiplicities, D1 and
D2 meet in precisely d1 d2 points.

Proof. Non-examinable. Interested reader can find the proof in [Section 5.3, Fulton, Al-
gebraic Curves]. 

Remark 8.14. This theorem shows that the number of intersection points of two plane
curves can be read off easily from their defining equations without solving them, which
is a big advantage for projective spaces. A special case of this theorem is Exercise 4.3
(2), when both plane curves have degree 1. In the other direction, this theorem can be
generalised in many different ways, thus has become the starting point of a major branch
of algebraic geometry, called intersection theory.

82
8.2. Cubics. Now we consider cubic curves. We first give a classification.
Example 8.15. Every cubic curve is defined by a non-zero homogeneous polynomial f ∈
k[x, y, z] of degree 3. By Exercise 4.2 (1), each irreducible factor of f is also homogeneous.
There are a few cases:

(1) If f is an irreducible polynomial, then V(f ) is an irreducible cubic;

(2) If f is the product of two irreducible factors of degree 1 and 2 respectively, then
the cubic V(f ) = L ∪ C is the union of a line L and a conic C (in this case we still
say V(f ) is singular, although we have not discussed the singularity of reducible
algebraic sets);
(3) If f is the product of three irreducible factors of degree 1, then V(f ) could be the
union of three distinct lines, or the union of a single line and a double line, or a
triple line. The union of three distinct lines is a cubic. The other two are not.

We have seen that there is only one line and one irreducible conic up to linear changes of
homogeneous coordinates. The situation is different for irreducible cubics.
Lemma 8.16. Up to a linear change of homogeneous coordinates, every irreducible cubic
curve C can be written in one of the following three forms

(1) C0 = Vp (y 2 z − x(x − z)(x − λz)) for some λ ∈ k\{0, 1};

(2) C1 = Vp (y 2 z − x2 (x − z));
(3) C2 = Vp (y 2 z − x3 ).

Proof. Non-examinable. 
Remark 8.17. The defining equations in Lemma 8.16 are called the normal forms of
irreducible cubics. By Exercise 6.2, we see that these formulas do give irreducible cubics.
Moreover, by Exercise 7.3, C0 is always non-singular; C1 is singular at the point [0 : 0 : 1],
where C1 intersects with itself; C2 is singular at the point [0 : 0 : 1], where C2 has a corner.
They are known respectively as an non-singular cubic, the nodal cubic and the cuspidal
cubic. Each of them can be understood as the projective closure of the corresponding
affine variety Va (y 2 − x(x − 1)(x − λ)) or Va (y 2 − x2 (x − 1)) or Va (y 2 − x3 ), with the only
point at infinity [0 : 1 : 0].
Proposition 8.18. A nodal cubic curve (or a cuspidal cubic curve) is rational.

Proof. To show a nodal cubic is rational, by Lemma 8.16, we can assume the nodal cubic
is C1 = V (y 2 z − x2 (x − z)) without loss of generality. Consider the rational maps
ϕ1 : P1 99K C1 ; [u : v] 7−→ [u(u2 + v 2 ) : v(u2 + v 2 ) : u3 ]
ψ1 : C1 99K P1 ; [x : y : z] 7−→ [x : y].
83
We will verify they are rational maps and they are inverse to each other. They are
both given by homogeneous polynomials of the same degree. Moreover, ϕ1 is defined, for
example, at the point [1 : 0]; ψ1 is defined, for example, at the point [0 : 1 : 0]. The image
of ψ1 is always in P1 . To verify the image of ϕ1 is in C, one just needs to compute
[v(u2 + v 2 )]2 [u3 ] − [u(u2 + v 2 )]2 [u(u2 + v 2 ) − u3 ] = v 2 (u2 + v 2 )2 u3 − u2 (u2 + v 2 )2 uv 2 = 0.
Finally we show they are inverse to each other. For any point [x : y : z] ∈ C1 , we have
(ϕ1 ◦ ψ1 )([x : y : z]) = ϕ1 ([x : y]) = [x(x2 + y 2 ) : y(x2 + y 2 ) : x3 ].
By the equation of C1 we know y 2 z − x2 (x − z) = 0, which implies x3 = (x2 + y 2 )z.
Therefore
[x(x2 + y 2 ) : y(x2 + y 2 ) : x3 ] = [x(x2 + y 2 ) : y(x2 + y 2 ) : z(x2 + y 2 )] = [x : y : z].
Moreover, for any point [u : v] ∈ P1 , we have
(ϕ1 ◦ ψ1 )([u : v]) = ϕ1 ([u(u2 + v 2 ) : v(u2 + v 2 ) : u3 ]) = [u(u2 + v 2 ) : v(u2 + v 2 )] = [u : v].
This shows that C1 is birational to P1 , hence C1 is rational.

To show a cuspidal cubic is rational, by Lemma 8.16, we can assume the cuspidal cubic
is C2 = V (y 2 z − x3 ) without loss of generality. Consider the rational maps
ϕ2 : P1 99K C2 ; [u : v] 7−→ [uv 2 : v 3 : u3 ];
ψ2 : C2 99K P1 ; [x : y : z] 7−→ [x : y].
A similar proof shows C2 is rational. We leave the details as an exercise. 
Proposition 8.19. A non-singular cubic curve is not rational.

Proof. Non-examinable. The idea is to show that the function field of a non-singular
cubic is not isomorphic to that of P1 . Interested reader can find the proof in [Section 2.2,
Reid, Undergraduate Algebraic Geometry]. This is a fun proof. The method in the proof
is called “infinite descent”. There are a few famous applications of this method in the
√
history of mathematics. It was used to prove that 2 is not a rational number, which
unfortunately caused the first crisis in the foundations of mathematics. This crisis led
to the discovery of irrational numbers, which was a big step forward in the development
of mathematics. Another famous application of the descent method was in the proof
of Fermat’s last theorem. Fermat conjectured that the equation xm + y m = z m has no
solutions in positive integers for any positive integer m > 3. The proof of the theorem
in m = 3 and m = 4 cases was given by the descent method shortly after that. But it
took mathematicians more than 300 years to completely solve the problem. The Andrew
Wiles Building in University of Oxford was named after the British mathematician who
finally proved this conjecture. 

Finally we look at some special points on a non-singular cubic.

84
Definition 8.20. Given a non-singular cubic curve C, a point p ∈ C is said to be an
inflection point of C if the tangent line Tp C meets C at p with multiplicity 3.

Remark 8.21. Recall from Proposition 8.11 that Tp C meets C at p with multiplicity at
least 2. By Theorem 8.8, if p is an inflection point, then p is the only intersection point
of Tp C and C; if p is not an inflection point, then Tp C and C meet at another point with
multiplicity 1.

Example 8.22. We show that the point p = [0 : 1 : 0] is an inflection point on the

non-singular cubic C = Vp (f ) where f = y 2 z − x3 + xz 2 . First of all we need to find out
the tangent line Tp C, which can be computed on the standard affine piece C1 = C ∩ U1 =
Va (f1 ) where f1 = z − x3 + xz 2 . The non-homogeneous coordinates of p in U1 is p = (0, 0).
Since ∂f∂x
1
= −3x2 +z 2 and ∂f∂z
1
= 1+2xz, the tangent line Tp C1 = Va (0(x−0)+1(z −0)) =
Va (z). Its projective closure is Tp C = Vp (z). To find the intersection points of C and
Tp C, we follow the method in the proof of Theorem 8.8. A point on Tp C is given by
[x : y : 0]. It lies in C if and only if f (x, y, 0) = 0, where f (x, y, 0) = −x3 which has
one solution [x : y] = [0 : 1] with multiplicity 3. Therefore Tp C and C meet at the point
[0 : 1 : 0] with multiplicity 3, which proves p = [0 : 1 : 0] is an inflection point on C.

85
 Exercise Sheet 8

This sheet will be discussed in the exercise class on 27 November. You are welcome to
submit your solutions at the end of the exercise class or anytime earlier.
Exercise 8.1. Examples of rational curves. Complete proofs of Propositions 8.7 and 8.18.

(1) Show that L = V(z) ⊆ P2 is isomorphic to P1 . Conclude that L is rational.

(2) Show that ϕ2 and ψ2 defined in the proof of Proposition 8.18 are rational maps.
Show that they are mutually inverse to each other. Conclude that the cuspidal
cubic curve C2 = V(y 2 z − x3 ) ⊆ P2 is rational.
Exercise 8.2. Example: Fermat cubic. Consider the cubic curve C = V(x3 +y 3 +z 3 ) ⊆ P2 .

(1) Show that C is non-singular.

(2) Show that the line L = V(z) meets C at 3 distinct points. Find all of them.
(3) For any p = [a : b : c] ∈ C, show that the tangent line Tp C = V(a2 x + b2 y + c2 z).
(4) Show that every point you find in part (2) is an inflection point.
Exercise 8.3. Bézout’s theorem for conics. Prove Theorem 8.12 in these steps.

(1) If the conic C = L1 ∪ L2 is the union of two lines, use Theorem 8.8 to conclude
that C ∩ D comprises at most 2d distinct points; or precisely 2d points when
multiplicities are counted. (Remark: if L1 ∩ D and L2 ∩ D have a common point
p, the multiplicity at p is defined to be the sum of the two multiplicities at p.)
(2) If the conic C is irreducible, without loss of generality, we can assume C = V(xz −
y 2 ) by Lemma 8.6. We have proved in Example 5.23 that every point in C can
be given by [p2 : pq : q 2 ] for some [p : q] ∈ P1 . Use the method in the proof of
Theorem 8.8 to finish the proof.
Exercise 8.4. An interesting application of Bézout’s theorem. Let p1 , · · · , p5 ∈ P2 be
distinct points, and assume that no 4 of them are on the same line. Prove that there
exists exactly one conic through all 5 points. You can follow these steps.

(1) Show that there exists at least one conic through all 5 points. (Hint: rank-nullity.)
(2) Suppose there are two distinct conics C1 and C2 through all 5 points. Use Bézout’s
theorem to conclude that they have a common component.
(3) If one of them is an irreducible conic, which has only one component, then the
other must be the same irreducible conic, otherwise they cannot have a common
component. Therefore both conics must be unions of two lines. Explain why we
can assume C1 = L0 ∪ L1 and C2 = L0 ∪ L2 for distinct lines L0 , L1 and L2 .
Explain why this leads to a contradiction.
86
 Solutions to Exercise Sheet 8

Solution 8.1. Examples of rational curves.

(1) We claim that ϕ : P1 −→ L; [x : y] 7−→ [x : y : 0] is a morphism. It is given

by homogeneous polynomials of the same degree, and is everywhere defined, since
x and y cannot be both zero. The image of any point under ϕ lies in L because
the last coordinate is zero. This justifies the claim. Similarly we claim that
ψ : L −→ P1 ; [x : y : z] 7−→ [x : y] is a morphism. It is given by homogeneous
polynomials of the same degree. Since z = 0, x and y cannot be both zero, hence
it is defined for every point in L. The image of any point in L under ϕ is clearly
in P1 . This justifies the claim. Finally we check ϕ and ψ are inverse to each other.
For any point [x : y] ∈ P1 , (ψ ◦ ϕ)([x : y]) = ψ([x : y : 0]) = [x : y]. For any point
[x : y : z] ∈ L, (ϕ ◦ ψ)([x : y : z]) = ϕ([x : y]) = [x : y : 0] = [x : y : z] since
z = 0. Therefore L is isomorphic to P1 . In particular, they are birational, hence
L is rational.
(2) Define rational maps ϕ2 : P1 99K C2 by ϕ2 ([u : v]) = [uv 2 : v 3 : u3 ] and ψ2 : C2 99K
P1 by ψ2 ([x : y : z]) = [x : y]. To show ϕ2 is a rational map, we observe: all
components are homogeneous of degree 3; ϕ2 is defined at every point [u : v] ∈ P1
since either u3 or v 3 is non-zero; the image [uv 2 : v 3 : u3 ] is a point in C2 since it
satisfies the defining equation of C2 . To show ψ2 is a rational map, we observe:
all components are homogeneous of degree 1; ψ2 is well-defined at every point on
C2 except [0 : 0 : 1]; image of ψ2 is clearly in P1 . It remains to show ϕ2 and ψ2
are mutually inverse to each other. For every [u : v] ∈ P1 where ψ2 ◦ ϕ2 is defined,
we have (ψ2 ◦ ϕ2 )([u : v]) = ψ2 ([uv 2 : v 3 : u3 ]) = [uv 2 : v 3 ] = [u : v]. For every
[x : y : z] ∈ C where ϕ2 ◦ ψ2 is defined, we have (ϕ2 ◦ ψ2 )([x : y : z]) = ϕ2 ([x :
y]) = [xy 2 : y 3 : x3 ] = [xy 2 : y 3 : y 2 z] = [x : y : z]. Therefore C2 is birational to P1 ,
hence is rational.
Solution 8.2. Example: Fermat cubic.

(1) We consider the standard affine piece C0 = C ∩ U0 = Va (f0 ) ⊆ A2 where f0 =

1 + y 3 + z 3 . Since ∂f
∂y
0
= 3y 2 and ∂f
∂z
0
= 3z 2 , the two derivatives vanish if and only
if y = z = 0. But then f0 = 1 6= 0. Therefore f0 = ∂f ∂y
0
= ∂f
∂z
0
= 0 have no common
solution, which means C0 is non-singular. Since the equation of C is symmetric
with respect to the variables, the same calculation shows that all other standard
affine pieces are also non-singular. Therefore C is non-singular.
(2) A point on the line L can be given by p = [x : y : 0]. If√ p ∈ C, then we have
2π −1
x3 + y 3 = 0, hence y = −x or −ωx or −ω 2 x where ω = e 3 is a primitive third
root of unity. So the three points in L ∩ C are p1 = [1 : −1 : 0], p2 = [1 : −ω : 0]
and p3 = [1 : −ω 2 : 0].
87
 (3) At least one of the three coordinates is non-zero. Without loss of generality, we
can assume a 6= 0. Then the point p = [a : b : c] ∈ C0 = C ∩ U0 = Va (f0 ) ⊆ A2 ,
in which its non-homogeneous coordinates are given by p = ( ab , ac ). The tangent
space of p in the standard affine piece C0 is given by
b2 c2
 
b c
Tp C0 = Va 3 · 2 · (y − ) + 3 · 2 · (z − ) .
a a a a
The tangent space Tp C is the projective closure of Tp C0 , which is given by the
homogenisation of the above polynomial
b2 c2
 
b c
Tp C = Vp 3 · 2 · (y − x) + 3 · 2 · (z − x) .
a a a a
a3
Since we assumed a 6= 0, we can multiply this polynomial by 3
without changing
its vanishing locus. Then we get
Tp C = Vp (b2 (ay − bx) + c2 (az − cx))
= Vp ((−b3 − c3 )x + ab2 y + ac2 z)
= Vp (a3 x + ab2 y + ac2 z)
= Vp (a2 x + b2 y + c2 z).
In the last step above is valid since we assumed a 6= 0.
Since a, b and c are symmetric, a similar calculation will give the same equation
for the tangent space Tp C when b 6= 0 or c 6= 0.
(4) At the point p1 = [1 : −1 : 0], the tangent space Tp1 C = Vp (x + y). For any point
q = [x : y : z] ∈ Tp1 C, we have x = −y. If q ∈ C, we then have (−y)3 + y 3 + z 3 = 0
hence z 3 = 0, which has one solution with multiplicity 3. This means Tp1 C meet
C at one point with multiplicity 3, hence p1 is an inflection point.
Similarly, at the point p2 = [1 : −ω : 0], the tangent space Tp2 C = Vp (x + ω 2 y).
For any point q = [x : y : z] ∈ Tp2 C, we have x = −ω 2 y. If q ∈ C, we then have
(−ω 2 y)3 + y 3 + z 3 = 0 hence z 3 = 0, which has one solution with multiplicity 3.
This means Tp2 C meet C at one point with multiplicity 3, hence p2 is an inflection
point.
Moreover, at the point p3 = [1 : −ω 2 : 0], the tangent space Tp3 C = Vp (x + ωy).
For any point q = [x : y : z] ∈ Tp3 C, we have x = −ωy. If q ∈ C, we then have
(−ωy)3 + y 3 + z 3 = 0 hence z 3 = 0, which has one solution with multiplicity 3.
This means Tp3 C meet C at one point with multiplicity 3, hence p3 is an inflection
point.

Solution 8.3. Bézout’s theorem for conics.

(1) If C = L1 ∪ L2 , then every common point of C and D must be either a common

point of L1 and D, or a common point of L2 and D. We know by Theorem
88
 8.8 that L1 ∩ D comprises at most d points, or precisely d points when counting
with multiplicities; L2 ∩ D comprises at most d points, or precisely d points when
counting with multiplicities. Therefore C ∩ D comprises at most 2d points, or
precisely 2d points when counting with multiplicities.
(2) We have proved in Example 5.23 that C is isomorphic to P1 . In particular, every
point in C can be given by [p2 : pq : q 2 ] for some [p : q] ∈ P1 . Let D = V(f ) for
some homogeneous polynomial f (x, y, z) of degree d. Then [p2 : pq : q 2 ] ∈ V(f ) if
and only if f (p2 , pq, q 2 ) = 0. The left-hand side is a homogeneous polynomial of
degree 2d in p and q. By Exercise 4.4 (2), it can be completely factored into 2d
homogeneous factors of degree 1 as

f (p2 , pq, q 2 ) = (a1 p + b1 q) · · · (a2d p + b2d q) = 0.

Each factor ai p + bi q determines a point [p : q] = [bi : −ai ] ∈ P1 , hence f = 0

has at most 2d solutions [p : q] = [bi : −ai ] ∈ P1 , which give at most 2d points
[p2 : pq : q 2 ] = [b2i : −ai bi : a2i ] ∈ (C ∩ D). When counting the number of times
each point occurs as a solution, we get precisely 2d points.

Solution 8.4. An interesting application of Bézout’s theorem.

(1) By Example 8.3, every conic C is given by a homogeneous polynomial g(x, y, z) = 0

of degree 2 with 6 coefficients a, b, c, d, e and f . For each i, since pi = [xi : yi : zi ] ∈
C, we can plug in x = xi , y = yi and z = zi to get an equation g(xi , yi , zi ) = 0,
which is a homogeneous linear equation in a, b, c, d, e and f . In this way the 5
points give a system of 5 linear equations. Since there are 5 equations and 6
indeterminants, by the theorem of rank-nullity, there is a solution for a, b, c, d, e
and f such that they are not simultaneously zero. This solution determines the
homogeneous polynomial g(x, y, z) of degree 2. We claim that g has no repeated
factors. If g has repeated factors, then g is the square of a linear polynomial
hence gives a double line which passes through all the 5 given points. This is a
contradiction since no 4 of the given points are allowed to be on the same line.
Hence we conclude that g defines a conic.
(2) Assume that there are two distinct conics C1 and C2 , both of which pass through
the 5 points. By Theorem 8.12, if they do not have any common component, then
they can meet in at most 4 common points. Hence they must have a common
component.
(3) If either C1 or C2 is an irreducible conic, which has only one component, then the
other must be the same conic. Under the assumption that C1 and C2 are distinct
conics, both of them must be the unions of two lines. Since they have a common
component, the other component in the two conics must be distinct. Hence we
can assume C1 = L0 ∪ L1 and C2 = L0 ∪ L2 , where L0 , L1 and L2 are distinct lines.
89
We know the 5 points p1 , · · · , p5 are on both conics. For each pi , there are two
possibilities: pi ∈ L0 , or pi ∈
/ L0 . If the second possibility happens, then pi ∈ L1
since pi ∈ C1 , and pi ∈ L2 since pi ∈ C2 . This implies pi is a common point of
L1 and L2 . Since L1 and L2 are distinct lines, by Theorem 8.8, they have only
1 common point. It follows that among the 5 points p1 , · · · , p5 , at most one of
them is not on L0 ; in other words, at least 4 of them are on the line L0 . This is a
contradiction because no 4 of them are allowed to be on the same line.

90
 9. Elliptic Curves

A very special feature of a non-singular cubic curve C is the existence of an abelian group
structure on the set of points in C. We will see how that works.

9.1. The group law on non-singular cubics. Given any non-singular cubic C and any
point O ∈ C, there exists an abelian group structure on the set of points in C, with O
being the identity element in the group law. That means, there is a binary operation “+”
defined on the set of points in C, which satisfies the conditions required in the definition of
an abelian group. The identity element O in the group law is also called the neutral point.
We will first describe the operation geometrically, then show some explicit computations,
finally explain why the construction defines an abelian group structure.

Construction 9.1 (The group law). Given a non-singular cubic curve C with a point
O ∈ C, there is an abelian group law on the set of points on C such that O is the identity
element. For any two points A, B ∈ C, their sum A + B is obtained in two steps

(1) The line AB meets the cubic C at a third point R;

(2) The line OR meets the cubic C at a third point R = A + B.

If A = B (resp. O = R), then the line AB (resp. OR) is defined to be the tangent line
TA C (resp. TO C). 

We can follow the above construction to make explicity computations. In each step, we
need to write down the equation of a certain line, and compute its intersection points
with the cubic. The reason for the existence of the third intersection point of a line and
a cubic and the method for computing it has been discussed in the proof of Theorem 8.8.
To find the line AB (or similarly OR), we need Definition 7.8 if A = B, or the follow
simple result if A 6= B.

Lemma 9.2. Given two distinct points A = [a0 : a1 : a2 ] and B = [b0 : b1 : b2 ] in P2 ,

there is a unique line L passing through the two points, defined by the polynomial
 
x a 0 b0
f (x, y, z) = det y a1 b1  .
 
z a 2 b2

Proof. We have seen in Exercise 4.3 (1) that there is a unique line L passing through A
and B. It remains to verify that the given polynoial defines such a line. Notice that the
given polynomial is non-zero and homogeneous of degree 1 hence defines a line. When
[x : y : z] = [a0 : a1 : a2 ] or [b0 : b1 : b2 ], two columns of the matrix are identical hence the
determinant is zero. This shows that A and B are points on this line. 
91
Example 9.3. Consider the cubic C = V(y 2 z − x3 + 4xz 2 − z 3 ) with the identity element
O = [0 : 1 : 0]. Take two points A = [2 : 1 : 1] and B = [−2 : 1 : 1] on C. By Lemma 9.2,
the line AB is defined by
 
x 2 −2
det y 1 1  = −4y + 4z.
 
z 1 1
By the method in the proof of Theorem 8.8, we can find the third intersection point R of
AB and C to be R = [0 : 1 : 1]. By Lemma 9.2, the line OR is defined by
 
x 0 0
det y 1 1 = x.
 
z 0 1

By the method in the proof of Theorem 8.8, we can find the third intersection point R of
OR and C to be R = [0 : −1 : 1]. Therefore A + B = [0 : −1 : 1].

Construction 9.1 works for any non-singular cubic with any point on it as the identity
element. In some special cases, the group law becomes particularly nice and simple. This
simplified group law is applicable only when the following two conditions are satisfied

(1) The non-singular cubic is given by C = Vp (y 2 z − x3 − ax2 z − bxz 2 − cz 3 ) for some

a, b, c ∈ k, which is the projective closure of the affine curve C2 = Va (y 2 − x3 −
ax2 − bx − c) with the only point O = [0 : 1 : 0] at infinity;
(2) The point at infinity O = [0 : 1 : 0] is the identity element.

It is important to observe that the graph of C2 is symmetric with respect to the x-axis.

Construction 9.4 (Simplified group law). Let C = Vp (y 2 z − x3 − ax2 z − bxz 2 − cz 3 ) be

a non-singular cubic for some a, b, c ∈ k. Let O = [0 : 1 : 0] be the identity element of the
group law and C2 = Va (y 2 − x3 − ax2 − bx − c) a standard affine piece of C. Given two
points A, B ∈ C, we have:

(1) If A = O, then A + B = B; if B = O, then A + B = A;

(2) If A, B ∈ C2 , assume the line AB meet the cubic C at a third point R. If A = B,
the line AB is defined to be the tangent line TA C.
(a) If A and B are symmetric with respect to the x-axis, then A + B = O;
(b) Otherwise, let R = (p, q) ∈ C2 , then R = (p, −q) = A + B. 

Remark 9.5. The simplified group law 9.4 also gives an easy way to compute the inverse
of any point A ∈ C. If A = O, then −A = O. Otherwise, let A = (x, y) ∈ C2 , then the
inverse −A = (x, −y) ∈ C2 which is the reflection of A across the x-axis.
92
Example 9.6. We look at Example 9.3 again. It is clear that both conditions required for
the simplified group law are met. The affine curve C2 = Va (y 2 − x3 + 4x − 1). Neither A
nor B is the identity element O = [0 : 1 : 0]. In non-homogeneous coordinates, A = (2, 1)
and B = (−2, 1). The line AB in the affine plane is given by L2 = Va (y − 1). Solving the
system given by equations y 2 − x3 + 4x − 1 = 0 and y − 1 = 0, we get the third point of
intersection R = (0, 1). Therefore A + B = R = (0, −1), or in homogeneous coordinates
[0 : −1 : 1]. This answer is consistent with that of Example 9.3.

Definition 9.7. A non-singular cubic curve with a chosen point on it (hence a group law
is determined) is called an elliptic curve.

The theory of elliptic curves is extremely rich and deep, and provides a good example
of the profound connections between abstract algebraic geometry, complex analysis, and
number theory. It constitutes an active area of current research, and plays a crucial
role in the recent proof of Fermat’s Last Theorem. Elliptic curves also have important
applications in various aspects of cryptography, such as encryption, digital signatures,
(pseudo-)random generators and so on. There are other higher dimensional projective
varieties, on which there exist abelian group laws. They are called abelian varieties,
which is also a major branch of algebraic geometry.

93
9.2. Linear systems and associativity. We are aiming to prove that Construction 9.1
does define an abelian group law. The difficulty here is the associativity. We clear up the
easy bits first.

Proposition 9.8. In Construction 9.1 of the group law on a non-singular cubic curve C:
the addition is commutative; O is the identity element; and every point has an inverse.

Proof. For two points A, B ∈ C, there is no difference between the line AB and the line
BA, hence A + B = B + A is obvious. This justifies the commutativity.

To find A + O, the first step gives the third intersection point R of the line AO and C; the
second step gives the third intersection point of the line OR and C, which is A. Hence
A + O = A is also obvious. This justifies that O is the identity element in the group law.

Given any A ∈ C, we claim its inverse can be given like this: assume the tangent line
TO C meets C at a third point O, and the line AO meets C at a third point B, then B
is the inverse of A. We need to verify A + B = O. To compute A + B, the first step
gives the third intersection point of the line AB and C, which is O; the second step gives
the third intersection point of the line OO and C, which is O by Proposition 8.11. This
justifies A + B = O, hence the inverse of A is well-defined. 

Remark 9.9. Here is a special case that is worth mentioning: if O is an inflection point,
then TO C meet C at O three times hence O = O. In such a case the inverse of A is simply
the third intersection point of the line AO and the curve C.

It remains to check the associativity in the group law. This requires some preparation,
which is very interesting in their own stand.

Notation 9.10. Given finitely many points P1 , · · · , Pk ∈ P2 . For every d > 0, we write
( )
f is homogeneous of degree d
Sd (P1 , · · · , Pk ) := f ∈ k[x, y, z] .
f (P1 ) = · · · = f (Pk ) = 0

It is easy to see that Sd (P1 , · · · , Pk ) is a vector space over k, as it is closed under addition
and scalar multiplication. This vector space is sometimes called a linear system, but we do
not need this terminology. In the following results we will need to look at S3 (P1 , · · · , P8 ).

Lemma 9.11. Let C1 and C2 be two cubic curves whose intersection consists of precisely
9 distinct points P1 , · · · , P9 . Then dimk S3 (P1 , · · · , P8 ) = 2.

Proof. Non-examinable. We do not prove it but we explain what the proof is really
about. It is easy to find out that a homogeneous polynomial f ∈ k[x, y, z] of degree 3
is determined by 10 coefficients. For each given point Pi , the requirement f (Pi ) = 0
imposes one linear condition on the coefficients of f . If all the 8 linear conditions on the
94
coeffcients are independent, then the remaining freedom in the coefficient is 2, which is
precisely what we need. Therefore the whole point is to show that these linear conditions
are guaranteed to be independent given the assumptions. The key ingredient in the proof
is Bézout’s Theorem 8.13. Interested reader can find the proof in [Proposition 2.6, Reid,
Undergraduate Algebraic Geometry]. 
Lemma 9.12. Let C1 = V(F1 ) and C2 = V(F2 ) be two cubic curves whose intersection
consists of precisely 9 distinct points P1 , · · · , P9 . Then any cubic curve D = V(G) through
P1 , · · · , P8 also passes through P9 .

Proof. By Lemma 9.11, we have dimk S3 (P1 , · · · , P8 ) = 2. It is clear that F1 , F2 ∈

S3 (P1 , · · · , P8 ). Moreover F1 and F2 are linearly independent, as otherwise they would
define the same cubic. Therefore F1 and F2 form a basis of S3 (P1 , · · · , P8 ). Since
G ∈ S3 (P1 , · · · , P8 ), we can write G = λ1 F1 + λ2 F2 for some λ1 , λ2 ∈ k. Now G(P9 ) =
λ1 F1 (P9 ) + λ2 F2 (P9 ) = 0, hence D passes through P9 . 

Now we are ready to prove the associativity. To avoid excessive technicality while still
keeping a grasp of the main idea in the proof, we will prove it under an extra mild
assumption, which will be stated in the proof. Some extra work will be required if this
assumption is not met, which we do not discuss.
Proposition 9.13. In Construction 9.1 of the group law on a non-singular cubic curve
C, the addition is associative.

Proof. Let A, B, E ∈ C. The construction of (A + B) + E = S uses 4 lines:

L1 : ABR; L2 : ROR; L3 : ERS; L4 : SOS.
The construction of A + (B + E) = T uses 4 lines:
M1 : BEQ; M2 : QOQ; M3 : AQT ; M4 : T OT .
We need to show S = T , for which it suffices to show S = T . We consider two cubics
D1 = L1 ∪ M2 ∪ L3 and D2 = M1 ∪ L2 ∪ M3 .
Then by construction we have
C ∩ D1 = {A, B, E, O, R, R, Q, Q, S};
C ∩ D2 = {A, B, E, O, R, R, Q, Q, T }.
Now we need a mild assumption that the 9 points in C ∩ D1 are distinct. Then the
two cubics C and D1 satisfy the conditions of Lemma 9.12. Since the cubic D2 passes
through 8 of the 9 points, it must pass through S as well, which means S ∈ C ∩ D2 .
Therefore S = T since S cannot be any of the other points by the mild assumption that
we imposed. This finishes the proof under this assumption. Extra work has to be done
when this assumption is not met. 
95
Remark 9.14. This is a very beautiful piece of argument in projective algebraic geometry.
Bézout’s theorem plays a key role in the course of the proof, mostly in the proof of Lemma
9.11. A similar argument can be used to prove many other results, including the famous
Pascal’s theorem (aka the mystic hexagon), which we will see in the exercise.

96
 Exercise Sheet 9

This sheet will be discussed in the exercise class on 4 December. You are welcome to
submit your solutions at the end of the exercise class or anytime earlier.

Exercise 9.1. Example: understanding the simplified group law.

(1) Show that [0 : 1 : 0] is an inflection point of C in the simplified group law 9.4.
(2) In the simplified group law 9.4, explain briefly how to find all points P ∈ C such
that P + P = O.
(3) Consider the curve and the group law in Example 9.6. Let A = [2 : 1 : 1] and
B = [−2 : −1 : 1]. Use the simplified group law to find out −A, −B and A + B.

Exercise 9.2. Example of group law. Consider the non-singular cubic curve C = V(y 2 z −
x3 − 4xz 2 ) ⊆ P2 . Let O = [0 : 1 : 0] be the identity element in the group law.

(1) Find all points where C meets the line L1 = V(z) and specify their multiplicities.
Do the same for the lines L2 = V(x) and L3 = V(y − 2x).
(2) Find the order of the subgroup generated by the point P = [2 : 4 : 1] ∈ C.
(3) Find all points Q ∈ C such that Q + Q = O.

Exercise 9.3. Example: Tate’s normal form. Consider the projective closure C of the
cubic curve C2 = V(y 2 + sxy − ty − x3 + tx2 ) ⊆ A2 for some fixed s, t ∈ k where t 6= 0.
Assume C is non-singular. Let the point at infinity O = [0 : 1 : 0] be the identity element
in the group law on C.

(1) For any point P = (a, b) ∈ C2 , show that −P = (a, −b − sa + t) in the group law.
(2) Suppose Q = (0, 0) ∈ C2 . Show that Q + Q = (t, t(1 − s)) in the group law.

Exercise 9.4. Pascal’s mystic hexagon. Let X ⊆ P2 be an irreducible conic. Let

ABCDEF be a hexagon whose vertices are inscribed in X. Assume the three pairs
of opposite sides meet in points P, Q, R respectively. (To be precise, the lines F A and
CD meet at P ; the lines AB and DE meet at Q; the lines BC and EF meet at R.) Show
that P, Q, R are colinear. (That means, the three points are on the same line in P2 .) You
can follow these steps (the idea is already used in the proof of Proposition 9.13):

(1) Sketch a picture to illustrate the given situation.

(2) The three lines F A, BC and DE form a cubic curve C1 ; the three lines AB, CD
and EF form a cubic curve C2 . Find C1 ∩ C2 .
(3) Consider a third cubic C3 given by the union of the conic X and the line P Q.
Then apply Lemma 9.12.

97
 Solutions to Exercise Sheet 9

Solution 9.1. Understanding the simplified group law.

(1) We show that the point p = [0 : 1 : 0] is an inflection point on the non-singular

cubic C = Vp (f ) where f = y 2 z − x3 − ax2 z − bxz 2 − cz 3 . First of all we need
to find out the tangent line Tp C, which can be computed on the standard affine
piece C1 = C ∩ U1 = Va (f1 ) where f1 = z − x3 − ax2 z − bxz 2 − cz 3 . The non-
homogeneous coordinates of p in U1 is p = (0, 0). Since ∂f ∂x
1
= −3x2 −2axz−bz 2 and
∂f1
∂z
= 1−ax2 −2bxz−3cz 2 , the tangent line Tp C1 = Va (0(x−0)+1(z−0)) = Va (z).
Its projective closure is Tp C = Vp (z). To find the intersection points of C and
Tp C, we follow the method in the proof of Theorem 8.8. A point on Tp C is given
by [x : y : 0]. It lies in C if and only if f (x, y, 0) = 0, where f (x, y, 0) = −x3 which
has one solution [x : y] = [0 : 1] with multiplicity 3. Therefore Tp C and C meet at
the point [0 : 1 : 0] with multiplicity 3, which proves p = [0 : 1 : 0] is an inflection
point on C.
(2) First of all, since O is the identity element in the group law, we always have
O + O = O, so O is one of such point. It remains to find all such points P ∈ C2 .
The condition P + P = O can be interpreted as P = −P . If the non-homogeneous
coordinates of P in C2 is given by P = (x, y), then by the simplified group law 9.4,
−P = (x, −y). The condition P = −P holds if and only if y = 0. Therefore all
points P ∈ C satisfying P +P = O are precisely the identity element O = [0 : 1 : 0]
and those points P = (x, y) ∈ C2 such that y = 0.
(3) In the standard affine piece C2 = V(y 2 − x3 + 4x − 1), the non-homogeneous
coordinates of the two points are A = (2, 1) and B = (−2, −1). The line AB is
given by x − 2y = 0. To find its third intersection points with C2 , we need to solve
the system

y 2 − x3 + 4x − 1 = 0,
x − 2y = 0.

We substitute x by 2y in the first equation to get y 2 − 8y 3 + 8y − 1 = 0, which can

be factored as (y 2 −1)(1−8y) = 0. The solutions are y = ±1 and y = 81 . Therefore
the third intersection point is ( 14 , 81 ), whose reflection across the x-axis is the sum
of A and B; that is A + B = ( 14 , − 18 ), or [ 41 : − 18 : 1] in homogeneous coordinates
(or [2 : −1 : 8] if you prefer). The inverse −A is the reflection of A across the
x-axis, so −A = (2, −1), or [2 : −1 : 1] in homogeneous coordinates. The inverse
−B is the reflection of B across the x-axis, so −B = (−2, 1), or [−2 : 1 : 1] in
homogeneous coordinates.

Solution 9.2. Example of group law.

98
 (1) For L1 ∩ C, set z = 0 in the equation defining C to obtain x3 = 0, which gives
solutions [x : y] = [0 : 1] with multiplicity 3. Hence [x : y : z] = [0 : 1 : 0]
is the only intersection point with multiplicity 3. For L2 ∩ C, set x = 0 in the
equation defining C to obtain y 2 z = 0, which gives solutions [y : z] = [0 : 1]
with multiplicity 2 and [1 : 0] with multiplicity 1. Hence the line L2 meets C at
[0 : 0 : 1] with multiplicity 2 and [0 : 1 : 0] with multiplicity 1. For L3 ∩ C, set
y = 2x to obtain x(4xz − x2 − 4z 2 ) = 0, which can be written as −x(x − 2z)2 = 0.
Its solutions are [x : z] = [0 : 1] with multiplicity 1, and [x : z] = [2 : 1] with
multiplicity 2. Therefore L3 meets C at [x : y : z] = [0 : 0 : 1] with multiplicity 1
and [2 : 4 : 1] with mulplicity 2.
(2) We can use the simplified group law 9.4. The standard affine piece C2 = Va (f2 ) ⊆
A2 where f2 = y 2 − x3 − 4x. We first compute P + P . The non-homogeneous
coordinates of P are (2, 4). To compute the tangent line TP C2 , we find ∂f 2
∂x
=
2 ∂f ∂f ∂f
−3x − 4 and ∂y = 2y. Therefore ∂x (P ) = −16 and ∂y = 8. It follows that
2 2 2

TP C2 = Va (−16(x−2)+8(y−4)) = Va (−2(x−2)+(y−4)) = Va (−2x+y) ⊆ A2 . To

find the third intersection point of TP C2 and C, we solve the system of equations

y 2 − x3 − 4x = 0,
−2x + y = 0.

We substitute y by 2x in the first equation to get 4x2 − x3 − 4x = 0, which

is −x(x − 2)2 = 0. Therefore the system has a solution (x, y) = (2, 4) with
multiplicity 2 and a solution (x, y) = (0, 0) with multiplicity 1. The solution (2, 4)
corresponds to the point P , hence the third intersection point is R = (0, 0). The
sum P + P is the reflection R of R across the x-axis, which is still (0, 0). Hence
P + P = R = (0, 0) = R.
Now we compute R + R. Since R = (0, 0), by the simplified group law 9.4 (2a),
we immediately have R + R = O. Therefore P + P + P + P = O. It follows that
the order of P must divide 4, which can only be 1 or 2 or 4. Since P 6= O, the
order of P is not 1. Since P + P = R 6= O, the order of P is not 2. Therefore the
order of P is 4, which means the subgroup generated by P has order 4.
(3) To find all points Q ∈ C such that Q + Q = O, we use Exercise 9.1 (2). First of all
O = [0 : 1 : 0] is such a point. It remains to find all points Q = (x, y) ∈ C2 such
that y = 0. In the equation f2 = y 2 − x3 − 4x = 0 we set y = 0. Then we have
√ √
−x3 − 4x = −x(x2 + 4) = 0. Hence x = 0 or 2 −1 or −2 −1. The corresponding
√ √
points are Q = (0, 0) or (2 −1, 0) or (−2 −1, 0). In summary, we found 4 points
√
Q ∈ C such that Q + Q = O, which are [0 : 1 : 0], [0 : 0 : 1], [2 −1 : 0 : 1] and
√
[−2 −1 : 0 : 1].

Solution 9.3. Example: Tate’s normal form.

99
Notice that the defining polynomial of the cubic does not meet the conditions required
for using the simplified group law. So we need to use the group law 9.1.

(1) To find the inverse, we use the method in the proof of Proposition 9.8. We need to
find the third intersection point O of TO C and C, then find the third intersection
point of OP and C, which is −P .
Since C is the projective closure of C2 , we can write down its defining equation
as C = Vp (y 2 z + sxyz − tyz 2 − x3 + tx2 z) ⊆ P2 . It is easy to see that O = [0 : 1 : 0]
is the only point at infinity. To find the tangent line TO C, we need to consider
the standard affine piece C1 = C ∩ U1 which contains the point O. We have
C1 = Va (f1 ) ⊆ A2 where f1 = z + sxz − tz 2 − x3 + tx2 z and O = (0, 0) ∈ C1 .
Since ∂f∂x
1
= sz − 3x2 + 2txz and ∂f ∂z
1
= 1 + sx − 2tz + tx2 , we have ∂f 1
∂x
(O) = 0
∂f1 2
and ∂z (O) = 1, hence TO C1 = Va (z) ⊆ A . Taking its projective closure, we get
TO C = Vp (z) ⊆ P2 . To find the intersection points of TO C and C, we consider an
arbitrary point [x : y : z] = [x : y : 0] ∈ TO C. If this point is also in C, then we set
z = 0 in the defining equation of C to get −x3 = 0. Therefore TO C and C meet
at the only point [x : y : z] = [0 : 1 : 0] with multiplicity 3, which means that the
third intersection point O of TO C and C is still O = O = [0 : 1 : 0].
To find −P , we need to write down the line OP . We first make an observation.
Since P = (a, b) ∈ C2 , its coordinates have to satisfy the defining polynomial of
C2 , namely
b2 + sab − tb − a3 + ta2 = 0,
or equivalently
−a3 + ta2 = −b(b + sa − t).
The homogeneous coordinates of P are given by P = [a : b : 1]. By Lemma 9.2
the line is given by
 
x 0 a
det y 1 b  = x − az = 0.
 
z 0 1

To find the third intersection point of OP and C, we consider an arbitrary point

[x : y : z] = [az : y : z] ∈ OP . Since this point is also in C, we get

y 2 z + sayz 2 − tyz 2 − a3 z 3 + ta2 z 3 = 0.

Using the observation above, we get

y 2 z + (sa − t)yz 2 − b(b + sa − t)z 3 = 0

which can be factored into

z(y − bz)(y + (b + sa − t)z) = 0.

100
 The three solutions are [y : z] = [1 : 0], [b : 1] and [−b − sa + t : 1]. Since
x = az, the three intersection points of OP and C are [x : y : z] = [0 : 1 : 0],
[a : b : 1] and [a : −b − sa + t : 1]. The first two points are O and P , hence
−P = [a : −b − sa + t : 1]. The non-homogeneous coordinates of −P with respect
to C2 is −P = (a, −b − sa + t).
(2) To compute Q + Q, we need to find the tangent line TQ C. We know that Q ∈
C2 = Va (f2 ) ⊆ A2 where f2 = y 2 + sxy − ty − x3 + tx2 . The partial derivatives are
given by ∂f 2
∂x
= sy − 3x2 + 2tx and ∂f ∂y
2
= 2y + sx − t. At the point Q = (0, 0), their
∂f2 ∂f2
values are ∂x (Q) = 0 and ∂y (Q) = −t. Since t 6= 0, we have TQ C2 = Va (−ty) =
Va (y) ⊆ A2 , hence TQ C = Vp (y) ⊆ P2 . To find the third intersection point R of
the line TQ C and C, we consider an arbitrary point [x : y : z] = [x : 0 : z] ∈ TQ C.
When this point is also on C, we can set y = 0 in the defining equation of C to
get −x3 + tx2 z = 0. It has solutions [x : z] = [0 : 1] with multiplicity 2 and [t : 1]
with multiplicity 1. Therefore the intersection points of TQ C and C are given by
[x : y : z] = [0 : 0 : 1] with multiplicity 2 and [t : 0 : 1] with multiplicity 1. Hence
third intersection point R of TQ C and C is R = [t : 0 : 1].
It remains to find the third intersection point of OR and C, which is the sum
Q + Q. Fortunately we have done the computation in part (1). Indeed, we have
seen that, given a point P = [a : b : 1] ∈ C, the line OP (= OP ) meets C at a
third point [a : −sa + t − b : 1]. Let a = t and b = 0, then OR meets C at a third
point [t : −st + t : 1], or in non-homogeneous coordinates (t, −st + t). Therefore
Q + Q = (t, −st + t) = (t, t(1 − s)).
Solution 9.4. Pascal’s mystic hexagon.

(1) A picture has been given in the exercise class. You can also find the same picture
in [Section 2.11, Reid, Undergraduate Algebraic Geometry].
(2) From the picture we can see that C1 and C2 meet at 9 distinct points, i.e.
C1 ∩ C2 = {A, B, C, D, E, F, P, Q, R}.
Indeed, the first six points are distinct by the assumption. None of the last three
points is on X (otherwise a certain line meets X in 3 points), so none of them
can coincide with any of the first six points. The last three points must also be
distinct (otherwise two certain lines meet each other in 2 points).
(3) By assumption, the cubic curve C3 passes through 8 of the above 9 points with
the point R being the only possible exception. By Lemma 9.12, R must be on C3
as well. Therefore R is either on the conic X or the line P Q. We claim that R is
not on X. Otherwise, the line BCR and the conic X meet at three distinct points
B, C and R, which violates Bézout’s theorem 8.8. Therefore R is on the line P Q,
which means that the points P, Q, R are colinear.
101
 10. Algebraic Surfaces

We look at a few aspects of hypersurfaces in P3 of low degrees.

10.1. Planes and quadric surfaces. From now on we focus on hypersurfaces in P3 .

Definition 10.1. A hypersurface S = V(f ) ⊆ P3 defined by some non-constant homo-

geneous polynomial f ∈ k[z0 , z1 , z2 , z3 ] without repeated factors is called a surface. The
degree of S is defined to be deg f . Surfaces of degree 1, 2, 3 and 4 are called planes,
quadrics, cubics and quartics respectively.

Example 10.2. Let [z0 : z1 : z2 : z3 ] be the homogeneous coordinates in P3 . Every

plane is defined by a polynomial f (z0 , z1 , z2 , z3 ) = a0 z0 + a1 z1 + a2 z2 + a3 z3 for some
a0 , a1 , a2 , a3 ∈ k which are not simultaneously zero. A plane is always irreducible.

Example 10.3. Every quadric surface is defined by a non-zero homogeneous polyno-

mial g ∈ k[z0 , z1 , z2 , z3 ] of degree 2. Similar to the case of conics, it is sometimes more
convenient to write it in the matrix form

g(z0 , z1 , z2 , z3 ) = (z0 , z1 , z2 , z3 ) · M · (z0 , z1 , z2 , z3 )T

where M is a 4 × 4 symmetric matrix. The classification of quadric surfaces is controlled

by the rank of M .

There is a notion of linear change of homogeneous coordinates in P3 , which is literally

almost the same as Definition 8.4, with all vectors having 4 components and A being a
4 × 4 invertible matrix.

Lemma 10.4. Every plane in P3 can be written as V(z0 ) after a suitable linear change
of homogeneous coordinates. A non-zero homogeneous polynomial of degree 2

g(z0 , z1 , z2 , z3 ) = (z0 , z1 , z2 , z3 ) · M · (z0 , z1 , z2 , z3 )T

defines a non-singular irreducible quadric surface if and only if M has rank 4; g defines a
singular irreducible quadric surface if and only if M has rank 3; g defines a union of two
planes if and only if M has rank 2; g defines a double plane if and only if M has rank 1.
Every non-singular quadric surface can be written as V(z0 z3 − z1 z2 ) after a suitable linear
change of homogeneous coordinates.

Proof. Non-examinable. Application of Gram-Schmidt orthogonalisation again. 

Remark 10.5. A union of two planes can be thought as a singular algebraic set. A double
plane is not a quadric surface. So a “non-singular quadric surface” always means a “non-
singular irreducible quadric surface”.
102
Now we turn to the rationality problem. Recall from Proposition 8.7 that a line or a non-
singular conic is always isomorphic to P1 hence is rational. Something similar happens to
surfaces.

Proposition 10.6. A plane is isomorphic to P2 , hence is rational. A non-singular quadric

surface is birational to P2 , hence is rational.

Proof. By Lemma 10.4, we can assume the plane is V(z0 ) and the non-singular quadric is
V(z0 z3 − z1 z2 ) without loss of generality. It is easy to show that V(z0 ) is isomorphic to
P2 ; we leave the details to the reader. We have proved in Exercise 5.2 that V(z0 z3 − z1 z2 )
is birational to P2 . 

This result suggests that a non-singular quadric surface is not isomorphic to P2 . Indeed, it
follows from the fact that two curves in P2 always intersect while two curves in a quadric
surface could be disjoint. The details are left as an exercise. We would like to know
what precisely a quadric surface looks like. For that purpose we need the theory of multi-
projective spaces. We will not discuss the theory systematically. Instead, we will only
focus on this particular example and mention a few ingredients of the theory along the
way. Some details in the proof are left to the reader.

Proposition 10.7. A non-singular quadric surface is isomorphic to P1 × P1 .

Proof. We assume the quadric surface is S = V(z0 z3 − z1 z2 ). We need to find morphisms

ϕ : P1 × P1 → S and ψ : S → P1 × P1 , such that both compositions are identities.

The product P1 × P1 is the simplest example of a bi-projective space. A point in it is given

by a pair of points (p, q) in P1 . If p = [x0 : x1 ] and q = [y0 : y1 ], then the bi-homogeneous
coordinates of (p, q) are given by ([x0 : x1 ], [y0 : y1 ]). Notice that for any λ, µ ∈ k\{0}, we
have ([λx0 : λx1 ], [µy0 : µy1 ]) = ([x0 : x1 ], [y0 : y1 ]). We construct two morphisms:
ϕ : P1 × P1 −→ S; ([x0 : x1 ], [y0 : y1 ]) −→ [x0 y0 : x1 y0 : x0 y1 : x1 y1 ];



 ([z0 : z1 ], [z0 : z2 ]) if z0 6= 0;


([z : z ], [z
0 1 1 : z3 ]) if z1 6= 0;
ψ : S −→ P1 × P1 ; [z0 : z1 : z2 : z3 ] 7−→


 ([z2 : z3 ], [z0 : z2 ]) if z2 6= 0;

([z2 : z3 ], [z1 : z3 ]) if z3 6= 0.


We need to check they are morphisms. We have not defined the notion of a morphism
in this setting, but it is very similar to a morphism between two projective varieties. All
components of ϕ are homogeneous of the same degree with respect to the coordinates x0
and x1 of p, and the coordinates y0 and y1 of q (aka bi-homogeneous). All components
of ψ are also homogeneous of the same degree. We observe that ϕ and ψ are both well-
defined at every point in their domains (we leave the details to the reader). Moreover,
103
the image of ϕ satisfies the defining equation of S. Hence ϕ is a morphism. To show ψ
is a morphism, we need to verify that the image of any point in S is independent of the
choice of any valid expression. More precisely, we need to verify [z0 : z1 ] = [z2 : z3 ] and
[z0 : z2 ] = [z1 : z3 ], both of which follow from the defining equation z0 z3 = z1 z2 of S.

We check the composition ψ ◦ ϕ is identity. Given any point ([x0 : x1 ], [y0 : y1 ]) ∈ P1 × P1 ,

using the first expression of ψ, we have
(ψ ◦ ϕ)([x0 : x1 ], [y0 : y1 ]) = ψ([x0 y0 : x1 y0 : x0 y1 : x1 y1 ])
= ([x0 y0 : x1 y0 ], [x0 y0 : x0 y1 ])
= ([x0 : x1 ], [y0 : y1 ]).
Similarly we can check that ψ ◦ ϕ is identity in all the other three cases.

We check the composition ϕ ◦ ψ is identity. Given any point [z0 : z1 : z2 : z3 ] ∈ S, using

the first expression of ψ, we have
(ϕ ◦ ψ)([z0 : z1 : z2 : z3 ]) = ϕ([z0 : z1 ], [z0 : z2 ])
= [z02 : z0 z1 : z0 z2 : z1 z2 ]
= [z02 : z0 z1 : z0 z2 : z0 z3 ]
= [z0 : z1 : z2 : z3 ].
Similarly we can check ϕ ◦ ψ is identity in all the other three cases.

To summarise, ϕ and ψ are mutually inverse isomorphisms. Therefore a quadric surface

is isomorphic to P1 × P1 . 

Quadric surfaces are very useful in civil engineering. According to the literature, the
Shukhov water tower (in Polibino, Russia, 1896, designed by Shukhov) is the first structure
of this shape ever built in the world. Similar design can also be found at a few places
inside and outside Sagrada Famı́lia (in Barcelona, Spain, designed by Gaudi). Nowaways
numerous cooling towers in power plants are built in this shape.

104
10.2. Non-singular cubic surfaces. We have seen that non-singular cubic curves have
very rich geometry. The situation is similar for cubic surfaces. The theory of cubic
surfaces has a long history. It is known since 1849 that a non-singular cubic surface
contains 27 lines. This discovery is one of the first results on surfaces of higher degree and
is considered by many as the start of modern algebraic geometry. Many mathematicians
contributed to the understanding of rich geometry of non-singular cubic surfaces. In this
lecture we will take a glimpse of the theory of non-singular cubic surfaces via examples.

Definition 10.8. A line in P3 is a projective variety V(f, g), where f, g ∈ k[z0 , z1 , z2 , z3 ]

are non-zero homogeneous polynomials of degree 1 which are not proportional to each
other.

Remark 10.9. The definition shows that a line in P3 is defined by the system of equations
(
a0 z0 + a1 z1 + a2 z2 + a3 z3 = 0
b0 z0 + b1 z1 + b2 z2 + b3 z3 = 0
such that the coefficient matrix
!
a0 a1 a2 a3
b0 b1 b2 b3
has rank 2. We know from linear algebra that its reduced row echelon form has two pivots,
therefore the two variables corresponding to the pivots can be written as linear functions
of the other variables. For example, if the pivots are in the first two columns, then
(
z0 = r2 z2 + r3 z3
z1 = s2 z2 + s3 z3
for some r2 , r3 , s2 , s3 ∈ k.

Proposition 10.10. The Fermat cubic surface S = V(z03 + z13 + z23 + z33 ) contains exactly
27 lines.

Proof. Assume a line L in P3 is given by z0 = r2 z2 + r3 z3 and z1 = s2 z2 + s3 z3 for some

r2 , r3 , s2 , s3 ∈ k (i.e. pivots in first two columns). Such a line lies in S if and only if
(r2 z2 + r3 z3 )3 + (s2 z2 + s3 z3 )3 + z23 + z33 = 0
holds for all z2 , z3 ∈ k, hence is an identity. By comparing the coefficients, we get
r23 + s32 = −1 (1)
r33 + s33 = −1 (2)
r22 r3 = −s22 s3 (3)
r2 r32 = −s2 s23 (4)
105
If r2 , r3 , s2 , s3 are all non-zero, then (3)2 /(4) gives r23 = −s32 , in contradiction to (1). Hence
for a line in the cubic at least one of these numbers must be zero. By (3) r2 and r3 cannot
be both non-zero.

If r2 = 0, then by (1) s32 = −1, hence by (3) s3 = 0, which by (2) implies r33 = −1.
 √This
j k 2π −1
gives 9 solutions r2 = s3 = 0, s2 = −ω , r3 = −ω for 0 6 j, k 6 2 and ω = exp 3
is a primitive third root of unity. We thus obtain 9 lines given by

z0 + ω k z3 = z1 + ω j z2 = 0, 0 6 j, k 6 2.

If r3 = 0, we can similarly find out that s2 = 0 and r23 = s33 = −1, hence we obtain
another 9 lines given by

z0 + ω k z2 = z1 + ω j z3 = 0, 0 6 j, k 6 2.

As the equation of S is symmetric with respect to all variables, we can allow permutations
of variables to find other lines in the cubic (i.e. pivots not necessarily in first two columns).
Some of the lines show up repeatedly after permutations of variables, but we get 9 new
lines given by
z0 + ω k z1 = z2 + ω j z3 = 0, 0 6 j, k 6 2.
In summary, we have equations of all 27 lines. 

Proposition 10.11. The cubic surface S = V(z02 z1 + z12 z2 + z22 z3 + z32 z0 ) is rational.

Proof. We write down two mutually inverse rational maps

ϕ : S 99K P2 ; [z0 : z1 : z2 : z3 ] 7−→ [z0 z3 : z1 z2 : z2 z3 ];

ψ : P2 99K S; [r : s : t] 7−→ [rt(rt + s2 ) : −s(r2 s + t3 ) : t2 (rt + s2 ) : −t(r2 s + t3 )].

To check they are rational maps, we observe that they are both given by homogeneous
polynomials of the same degree. It is easy to check that ϕ([1 : −1 : 1 : −1]) = [1 : 1 : 1]
and ψ([1 : 1 : 1]) = [1 : −1 : 1 : −1]), hence both ϕ and ψ are defined on non-empty
sets. We need to show the image of ψ satisfies the defining equation of S, which can be
computed directly.

It remains to show that both ψ ◦ ϕ and ϕ ◦ ψ are identity maps on the loci where they are
well-defined. This is also a simple calculation. We leave the details to the reader. This
shows that S and P2 are birational. By definition, S is rational. 

The phenomenons in the above examples hold for every non-singular cubic surface. We
summarise it in the following result.

Theorem 10.12. Every non-singular cubic surface contains exactly 27 lines. Every non-
singular cubic surface is rational.
106
Proof. Non-examinable. Interested reader can find the proof in [Chapter 7, Reid, Under-
graduate Algebraic Geometry]. 

Remark 10.13. If we fix the degree and vary the dimension, there is major difference
between non-singular cubic curves and surfaces: the former is not rational while the
latter is rational. In higher dimensions, whether a cubic hypersurface is rational is a very
difficult question. (There is an answer in dimension 3, but mostly unknown in dimension
4 or higher.)

Moreover, if we fix the dimension, then the number of lines in a non-singular surface
depends on its degree: planes and non-singular quadric surfaces contain infinitely many
lines (which we will see in an exercise); a non-singular cubic surface has 27 lines; most
non-singular surfaces of higher degrees have no lines at all.

Counting special curves in various kinds of spaces turns out to be a fascinating topic in
algebraic geometry, which is usually called enumerative geometry. These questions are
not only interesting to mathematicians, but also have been extensively studied in physics,
as they play an important role in string theory. The 27 lines in non-singular cubic surfaces
is a first example of this type.

107
 Exercise Sheet 10

This sheet will be discussed in the exercise class on 7 December. You do not need to submit
your solutions.
Exercise 10.1. Infinitely many lines on planes.

(1) Without loss of generality, we consider the plane P = V(z0 ) ⊆ P3 . For every
[a : b : c] ∈ P2 , show that V(z0 , az1 + bz2 + cz3 ) defines a line in P .
(2) Show that two such lines always meet at exactly one point.
Exercise 10.2. Infinitely many lines on non-singular quadric surfaces.

(1) Without loss of generality, we consider the quadric surface Q = V(z0 z3 − z1 z2 ) ⊆

P3 . Show that for every [a : b] ∈ P1 , V(az0 + bz1 , az2 + bz3 ) defines a line in Q.
(2) Show that two such lines are always disjoint.
(3) Show that every point in Q lies on exactly one of such lines.
(4) Can you write down another family of pairwisely disjoint lines in Q, such that
every point in Q lies on exactly one of them?

Remark: the family of lines constructed in part (1) (or part (4)) is called a ruling on Q.
We have seen in Exercise 5.2 that Q is birational to P2 . This exercise shows that Q is not
isomorphic to P2 . The reason is: two lines in the same ruling on Q do not meet, while
any two curves on P2 meet by Bézout’s theorem. Since Q is isomorphic to P1 × P1 , it
follows that P1 × P1 is not isomorphic to P2 .
Exercise 10.3. Rationality of a cubic surface. Finish the proof of Proposition 10.11.

(1) Show that any point in the image of ψ satisfies the defining equation of S.
(2) Show that ψ ◦ ϕ and ϕ ◦ ψ are both identity maps on the loci where they are
well-defined.

Remark: there is a general method to find out the explicit formula for a birational map
between any given non-singular cubic surface S and P2 . For that purpose we need to know
the explicit equations of two disjoint lines on S. We do not discuss the details. However,
the formula is usually very messy. The example in Proposition 10.11 is one of the very
rare good-looking ones.
Exercise 10.4. Thank you and have a wonderful Christmas vacation!

Thank you all for your participation in this course. Please complete the Unit Evaluation
for this course whenever convenient. If you have any questions during your revision,
please feel free to ask me. There will be extra office hours after the vacation. You are
also welcome to contact me by email at any time. Good luck with your exams!
108
 Solutions to Exercise Sheet 10

Solution 10.1. Infinitely many lines on planes.

(1) Since z0 and az1 + bz2 + cz3 are both homogeneous polynomials of degree 1 and
not proportional to each other, L = V(z0 , az1 + bz2 + cz3 ) defines a line in P2 . To
show that the line L is in P , we just need to observe that every point on L satisfies
the equation z0 = 0, hence is a point in P .
(2) Let L = V(z0 , az1 + bz2 + cz3 ) and L0 = V(z0 , a0 z1 + b0 z2 + c0 z3 ) be two such lines,
where [a : b : c] 6= [a0 : b0 : c0 ]. If a point p = [z0 : z1 : z2 : z3 ] is an intersection
point of L and L0 , then its coordinates satisfy the system of equations

z0 = 0;
az1 + bz2 + cz3 = 0;
a0 z1 + b0 z2 + c0 z3 = 0.

The first equation fixes the z0 coordinate. For the other coordinates, we look at
the second and the third equations. We look at the coefficient matrix
!
a b c
.
a0 b 0 c 0

Since [a : b : c] and [a0 : b0 : c0 ] represent different points in P2 , both rows

are non-zero and linearly independent. Hence the matrix has rank 2. It follows
that the null-space has dimension 1, which means that there is a unique solution
for [z1 : z2 : z3 ] (up to scaling). Therefore there is a unique intersection point
[z0 : z1 : z2 : z3 ] for the lines L and L0 .

Solution 10.2. Infinitely many lines on non-singular quadric surfaces.

(1) It is clear that for every point [a : b] ∈ P1 , the two polynomials az0 + bz1 and
az2 + bz3 are non-zero and homogeneous of degree 1. They are not propotional to
each other, so V(az0 + bz1 , az2 + bz3 ) defines a line L in P2 . We still need to show
that every point in L is a point in Q. Since [a : b] ∈ P1 , we have either a 6= 0 or
b 6= 0. If a 6= 0, then a point p = [z0 : z1 : z2 : z3 ] ∈ L satisfies z0 = − ab z1 and
z2 = − ab z3 . Then
   
b b
z0 z3 − z1 z2 = − · z1 · z3 − z1 · − · z3 = 0.
a a
Hence p ∈ Q. If b 6= 0, a similar calculation shows that every point p ∈ L also
satisfies the equation z0 z3 − z1 z2 = 0 hence is a point in Q. We conclude that L
is a line in Q.
109
 (2) Consider two lines L = V(az0 + bz1 , az2 + bz3 ) and L0 = V(a0 z0 + b0 z1 , a0 z2 + b0 z3 )
where [a : b] and [a0 : b0 ] are two different points in P1 . If the two lines have a
common point [z0 : z1 : z2 : z3 ], then the system of equations

az0 + bz1 = 0,
az2 + bz3 = 0,
a0 z0 + b0 z1 = 0,
a0 z2 + b0 z3 = 0

must have a non-zero solution. ! However, the coefficient matrix for the first and
a b
the third equations is . Since [a : b] and [a0 : b0 ] are two different points
a0 b 0
in P1 , the two rows are both non-zero and linearly independent. Hence the matrix
has rank 2, which means that the only solution to these two equations is z0 =
z1 = 0. For the same reason the only solution to the second and fourth equations
is z2 = z3 = 0. Since the system of four equations has only a zero solution, L and
L0 do not have any common point. In other words, they are disjoint.
(3) For any point p = [z0 : z1 : z2 : z3 ] ∈ Q, we first show that p lies on a certain line
L = V(az0 + bz1 , az2 + bz3 ). There are two cases. Case 1. If z0 and z1 are not
simultaneously zero, then we choose [a : b] = [z1 : −z0 ] for the line L. We claim
that p ∈ L. Indeed, for such a choice of [a : b] we have az0 + bz1 = z1 z0 − z0 z1 = 0
and az2 + bz3 = z1 z2 − z0 z3 = 0. The claim holds. Case 2. If z0 and z1 are both
zero, then z2 and z3 are not simultaneously zero. We can choose [a : b] = [z3 : −z2 ]
for the line L. A similar calculation shows that p ∈ L. In both cases, the point p
lies on a certain line L = V(az0 + bz1 , az2 + bz3 ) for a suitable choice of [a : b].
It remains to prove that p lies on only one of such lines. This is clear because
we have seen from part (2) that two such lines are always disjoint.
(4) For every [a : b] ∈ P1 , V(az0 + bz2 , az1 + bz3 ) also defines a line. These lines are
pairwisely disjoint, and every point in Q lies on exactly one of them. The proof
can be obtained simply by switching z1 and z2 in the proof for the above three
parts.

Solution 10.3. Rationality of a cubic surface.

110
(1) We need to verify that every point in the image of ψ satisfies the defining equation
of S. Indeed, we have
z02 z1 + z12 z2 + z22 z3 + z32 z0
= −r2 t2 (rt + s2 )2 · s(r2 s + t3 ) + s2 (r2 s + t3 )2 · t2 (rt + s2 )
− t4 (rt + s2 )2 · t(r2 s + t3 ) + t2 (r2 s + t3 )2 · rt(rt + s2 )
= −(r2 t2 s + t5 ) · (rt + s2 )2 · (r2 s + t3 ) + (s2 t2 + rt3 ) · (r2 s + t3 )2 · (rt + s2 )
= −t2 (r2 s + t3 ) · (rt + s2 )2 · (r2 s + t3 ) + t2 (s2 + rt) · (r2 s + t3 )2 · (rt + s2 )
= −t2 · (rt + s2 )2 · (r2 s + t3 )2 + t2 · (r2 s + t3 )2 · (rt + s2 )2
= 0.
Therefore the statement holds.
(2) Let [z0 : z1 : z2 : z3 ] be a point in S. Then these coordinates satisfy
z02 z1 + z12 z2 + z22 z3 + z32 z0 = 0.
Then we have
(ψ ◦ ϕ)([z0 : z1 : z2 : z3 ])
= ψ([z0 z3 : z1 z2 : z2 z3 ])
= [z0 z2 z32 (z0 z2 z32 + z12 z22 ) : −z1 z2 (z02 z1 z2 z32 + z23 z33 ) :
: z22 z32 (z0 z2 z32 + z12 z22 ) : −z2 z3 (z02 z1 z2 z32 + z23 z33 )]
= [z0 z22 z32 (z32 z0 + z12 z2 ) : −z1 z22 z32 (z02 z1 + z22 z3 ) :
: z23 z32 (z32 z0 + z12 z2 ) : −z22 z33 (z02 z1 + z22 z3 )]
= [z0 z22 z32 (z32 z0 + z12 z2 ) : z1 z22 z32 (z32 z0 + z12 z2 ) :
: z23 z32 (z32 z0 + z12 z2 ) : z22 z33 (z32 z0 + z12 z2 )]
= [z0 : z1 : z2 : z3 ]
wherever the composition ψ ◦ ϕ is well-defined. This shows that ψ ◦ ϕ is equivalent
to the identity map on S.
Now let [r : s : t] be a point in P2 . Then we have
(ϕ ◦ ψ)([r : s : t])
= ϕ([rt(rt + s2 ) : −s(r2 s + t3 ) : t2 (rt + s2 ) : −t(r2 s + t3 )])
= [−rt2 (rt + s2 )(r2 s + t3 ) : −st2 (r2 s + t3 )(rt + s2 ) : −t3 (rt + s2 )(r2 s + t3 )]
= [r : s : t]
wherever the composition ϕ ◦ ψ is well-defined. This shows that ϕ ◦ ψ is equivalent
to the identity map on P2 .

111
 Appendix A. Brief Review of Algebra 2B

This is an outline of the topics in Algebra 2B that were reviewed during the exercise class
in the first week of the semester.

Ring. A ring is a set of elements with two operations: addition and multiplication,
which have to satisfy various algebraic laws. Check your Algebra 2B notes to make sure
you know the full definition. We are only interested in commutative rings with 1. More
precisely, we mainly focus on polynomial rings k[x1 , · · · , xn ] and their quotient rings. (In
particular, k[x1 , · · · , xn ] can be realised as a quotient of itself by the zero ideal.)

Ideal. An ideal I is a non-empty subset of a ring R, satisfying two closedness conditions:

“a, b ∈ I =⇒ a − b ∈ I”, and “r ∈ R, a ∈ I =⇒ ra ∈ I”. When R is a commutative ring
with 1, the first condition a − b ∈ I can be replaced by the equivalent condition a + b ∈ I.

Quotient ring. For any ideal I in a ring R, there is a quotient ring R/I, whose elements
are cosets r + I for any r ∈ R. Two cosets r1 + I and r2 + I are the same if and only if
r1 − r2 ∈ I. If R is a commutative ring with 1, then so is R/I.

Ring homomorphism. A homomorphism ϕ : R −→ S between two rings is a map

which preserves addition and multiplication. Nice and easy.

Special rings. We have “rings ⊃ integral domains ⊃ UFDs ⊃ PIDs ⊃ fields”. Make
sure you know the definition of each. It is important to us that k[x1 , · · · , xn ] is a UFD;
namely, every polynomial can be factored into a product of irreducible polynomials, which
is unique up to the order of factors and units (non-zero constants). It is a PID only when
n = 1. (We now know that it is a Noetherian ring for every n.)

Polynomial. A polynomial f (x1 , · · · , xn ) ∈ k[x1 , · · · , xn ] is a finite sum of monomials.

If f is not zero, then the degree of f is the highest degree of its non-zero monomials. But
the degree of the zero polynomial is quite arguable. There are different ways to treat this
problem. We will adopt one opinion and define the degree of the zero polynomial to be
any non-negative integer. Details will be explained in week 4.

Irreducible polynomial. When k is algebraic closed, the only irreducible polynomials

in k[x] are the ones of degree 1. For polynomial rings in more than 1 variable, there is no
such a general rule, but irreducible polynomials can still be determined in some cases.

Example A.1. We claim that y 2 − x3 + x ∈ k[x, y] is an irreducible polynomial. We

assume on the contrary that it can be written as the product of two non-constant factors.
As a polynomial in y with coefficients in k[x], y 2 − x3 + x has degree 2 in y. Hence the
two factors have degrees either 2 and 0 in y respectively, or 1 and 1 respectively. More
precisely,
y 2 − x3 + x = f2 (x)y 2 + f1 (x)y + f0 (x) · g(x) or f1 (x)y + f0 (x) · g1 (x)y + g0 (x) .




112
In the first case, we have the identity f2 (x)g(x) = 1, hence g(x) is a non-zero con-
stant. Contradiction. In the second case, we similarly have the identity f1 (x)g1 (x) = 1.
Therefore both factors are non-zero constants. Without loss of generality we can assume
f1 (x) = g1 (x) = 1. Then we have
y 2 − x3 + x = (y + f0 (x))(y + g0 (x)).
Comparing the coefficients of y we have f0 (x) + g0 (x) = 0, hence g0 (x) = −f0 (x). Com-
paring the terms without y we have f0 (x)g0 (x) = −x3 + x, hence f0 (x)2 = x3 − x =
x(x + 1)(x − 1). The right hand side is not a square. Contradiction. This concludes that
y 2 − x3 + x is an irreducible polynomial.

Algebra. You might not like the definition of a k-algebra, since it is kind of long and hard
to remember. We need to work with a special type of algebras called finitely generated
k-algebras. You might think the definition is even more involved, but it is actually very
simple and explicit. A finitely generated algebra is a ring which is isomorphic to some
k[x1 , · · · , xn ]/I. A k-algebra homomorphism ϕ : k[x1 , · · · , xn ]/I −→ k[y1 , · · · , ym ]/J is
simply a ring homomorphism that sends a coset c + I to c + J for every constant c. They
are formally defined in week 3.

Fundamental isomorphism theorem. The fundamental isomorphism theorem for

rings is the following statement: for a ring homomorphism f : R −→ S, there is a
canonical isomorphism
im(f ) ∼
= R/ ker(f ).
This is a very important theorem for our purpose. Look at the following example.

Example A.2. We claim that k[x, y]/(y − x2 ) ∼

= k[t]. To see this, we construct a ring
homomorphism (in fact, a k-algebra homomorphism)
ϕ : k[x, y] −→ k[t]; x 7−→ t; y 7−→ t2 .
This means that every monomial axi y j is sent to ati (t2 )j = ati+2j , where a ∈ k is the
coefficient. By the fundamental isomorphism theorem, we have
im(ϕ) ∼
= k[x, y]/ ker(ϕ).
We need to identify im(ϕ) and ker(ϕ).

For any p(t) ∈ k[t], we have ϕ(p(x)) = p(t). This shows ϕ is surjective, hence im(ϕ) = k[t].

For any f (x, y) ∈ k[x, y], I claim it can be written as

f = (y − x2 ) · g + h,
for some g(x, y) ∈ k[x, y] and h(x) ∈ k[x]. For this, one only need to replace every single
occurrence of y in f (x, y) by [(y − x2 ) + x2 ], and then multiply out the square brackets
113
leaving the terms in round brackets untouched. Armed with this claim, we see that
ϕ(f ) = ϕ(y − x2 ) · ϕ(g) + ϕ(h) = (t2 − t2 ) · ϕ(g) + h(t) = h(t).
It follows that ϕ(f ) = 0 ⇐⇒ h = 0 ⇐⇒ f ∈ (y − x2 ). Hence ker(ϕ) = (y − x2 ). Therefore
the fundamental isomorphism theorem implies that k[t] ∼= k[x, y]/(y − x2 ).

Field. A field is a commutative ring with 1 such that every non-zero element has a multi-
plicative inverse. Check your Algebra 2B notes to make sure you know the characteristic
of a field and the field of fractions of an integral domain (which are used in week 6).

114
 Acknowledgements

I would like to thank all 53 students in the class for their enthusiasm and participation
in this course. I am also grateful to Alastair Craw and Gregory Sankaran for sharing
their experience and lecture notes used for this unit in previous years, as well as David
Calderbank and Alastair King for their helpful advices.

References
[1] Alastair Craw, Lecture notes on algebraic curves, 2013.
[2] William Fulton, Algebraic curves – An introduction to algebraic geometry, Advanced Book Classics,
Addison-Wesley Publishing Company, Redwood City, CA, 1989.
[3] Miles Reid, Undergraduate algebraic geometry, London Mathematical Society Student Text 12, Cam-
bridge University Press, Cambridge, 1988.
[4] Gregory Sankaran, Lecture notes on algebraic curves, 2007.

115