LECTURES ON BASIC ALGEBRAIC GEOMETRY

VISHWAMBHAR PATI

Abstract. In these lectures, we will introduce some basic notions, developing the necessary algebra as we go
along.

1. Motivation
1.1. The Diophantine Problem. Let us briefly look at a simple algebraic problem, one that has been around
for centuries. Let f (X, Y, Z) be a homogeneous polynomial of degree d , with integer coefficients. One would
like to know if there are any solutions to the equation f (X, Y, Z) = 0 with X, Y, Z ∈ Z. (For example, the
Fermat problem asks whether such exist for f (X, Y, Z) = X d + Y d − Z d for d > 2). Note that (0, 0, 0) is always
a solution, and if (a, b, c) ∈ Z × Z × Z is a solution, so is (ma, mb, mc) for all m ∈ Z, by the homogeneity
of f . Thus it makes sense to consider only solutions of the kind (a, b, c) 6= (0, 0, 0), and concern ourselves
only with equivalence classes, denoted [a : b : c] , of such solutions, where the equivalence relation on the set
S = Z × Z × Z \ (0, 0, 0) is defined by (a, b, c) ∼ (ma, mb, mc) for all integers m 6= 0. There is also the obvious

Remark 1.1.1. Equivalence classes of solutions [a : b : c] ∈ S/ ∼ to the homogeneous equation f (X, Y, Z) = 0

with c 6= 0 (resp. a 6= 0, resp. b 6= 0) are in 1-1 correspondence with rational solutions (p, q) ∈ Q × Q to the
de-homogenised polynomial equation f (X, Y, 1) = 0 (resp. f (1, Y, Z) = 0, resp. f (X, 1, Z) = 0) in two
variables.

It is clear that all the integral solutions to the problem when d = 1 can be enumerated. If f (X, Y, Z) =
lX +mY +nZ is a general degree one polynomial, all the rational solutions to, say the de-homogenised equation
f (X, Y, 1) = 0 are precisely the pairs (p, −lp−n
m ) with p ∈ Q. Similarly for the other two de-homogenised versions
of f (X, Y, Z).

For the case d = 2 (conics), one should first observe that there need not be any √ non-trivial integral solutions.
For example, assume (a, b, c) is a solution to f (X, Y, Z) = X 2 + Y 2 − 3Z 2 . Since 3 is irrational, we can assume
that a, b, c are all non-zero. We may also assume, by scaling away, that the greatest common divisor of a, b, c
is 1. Since a2 + b2 = 3c2 , we have a2 + b2 congruent to 0 modulo 3. Now if a, b are any integers, a2 and b2
are congruent to 0 or 1 modulo 3, so that a2 + b2 is congruent to 0 mod 3 iff both a and b are divisible by
3. This will mean that a2 + b2 is divisible by 9, and hence c2 is divisible by 3, so that c is also divisible by 3,
contradicting that the greatest common divisor of a, b, c is 1.
In the degree 2 case however, there is a method of generating all the rational solutions to the de-homogenised
equation say g(X, Y ) = f (X, Y, 1) = 0 given one rational solution, say (p, q) ∈ Q × Q. Let us assume that
q 6= 0.
This is a geometric method, given as follows. Join the point P = (p, q) to the point (t, 0) by means
of a straight line, which moves in time with the time parameter t (this is why we took q 6= 0, otherwise
the line would be idle as the x-axis for all t ! ) keeping the point P fixed. For general t, this line will
intersect the conic g(X, Y ) = 0 in two points, one of which we know to be P = (p, q). Let Q(t) = (Xt , Yt )
be the other point where it intersects the conic. We note that for Q(t) to lie on this line, we must have
(p − t)Yt = q(Xt − t). Plug this equation for Yt into the equation g(Xt , Yt ) = 0. We will then have a quadratic
equation: α(t)Xt2 + β(t)Xt + γ(t) = 0 where α(t), β(t) and γ(t) are rational functions of t (a rational function
is a quotient of a polynomial in t by another non-zero polynomial in t.) Also, since g(X, Y ) = 0 had integer
coefficients, these rational functions in t have all rational coefficients. By well-known facts on quadratics, we
have p + Xt = −β(t)
α(t) , so that Xt also becomes a rational function of t with rational coefficients. Plugging into
the above expression for Yt , we have that Yt is also a rational function of t with rational coefficients. We thus
1
2 VISHWAMBHAR PATI

have a rational parametrisation of our conic. Now all the points on g(X, Y ) = 0 with both coordinates rational
are precisely (Xt , Yt ) with t ∈ Q.
Exercise 1.1.2. Use the method of projection from the rational point (0, 1) on the conic Y 2 −X 2 = 1 to obtain
2t 1+t2
the rational parametrisation for it given by Xt = 1−t 2 , Yt = 1−t2 . Use this to show that the Pythagorean

triples: {[m2 − n2 : 2mn : m2 + n2 ]} where m, n ∈ Z exhaust all the integral solutions to X 2 + Y 2 = Z 2 (upto
permutations of X and Y .)
In passing, we note another application of the above parametrisation to the evaluation of definite integrals.
Suppose we want a closed form solution to the indefinite integral:
Z
1
(ax2 + bx + c)− 2 dx
R
which maybe rewritten as dx 2 2 2 2
y , where y = ax + bx + c. Let g(x, y) = y − ax − bx − c, and by the above
method, find a rational parametrisation x(t) = p(t), y(t) = q(t), where both these functions are rational
R 0 (t)dt
functions of t. Then dx = p0 (t)dt and our integral becomes : p q(t) , which can be rewritten as the integral
of a rational function in t, and integrated by the method of partial fractions. (Try it out for a = 1, b = 0, c = 1,
using the exercise above).

The story for d = 3 changes completely. For example, in p. 7 of Shafarevic’s book (Basic Algebraic
Geometry), it is shown that the Fermat curve X d + Y d = 1 cannot be rationally parametrised for d > 2.
Indeed, the entire theory of elliptic curves and elliptic integrals was born with the effort to evaluate the
integral:
Z
dx
(ax + bx + c)1/2
3

the so called elliptic integral which arises when one solves the equations of motion of a pendulum without the
small-amplitude approximation.

1.2. Invariant Theory. At the turn of the last century, Felix Klein sought to understand geometry by means
of group theory, i.e. study the group of transformations under which the geometry remains invariant. For
example, for Euclidean geometry, the corresponding group is the Motion Group, for hyperbolic geometry it
is the group P SL(2, R), and for the Minkowski space-time, the Poincare Group. In these cases, the original
space is expressible as the quotient of the original group by a closed subgroup, and the groups listed above
are all given by algebraic (polynomial) conditions on matrix coefficients in a suitable realisation inside a large-
dimensional Euclidean space. So what sorts of objects does one get on taking a quotient like this, and what
are a complete set of invariants on the quotient space to describe all possible functions on it ? To give a simple
example, if one lets the permutation group Sn act on Euclidean n-space Rn by permutations of the coordinates,
then the quotient space X is an algebraic object, and polynomial functions on Rn which descend to X are
precisely the symmetric functions on Rn ,and it is well known that all symmetric polynomials are polynomials
in the the elementary symmetric functions σi (X1 , X2 , ..Xn ) , i = 1, 2, ..., n.

The objects of interest here are algebraic sets, which we define below:

Definition 1.2.1. Let k be a field, and let S be some subset of the polynomial ring in n variables, i.e.
k[X1 , X2 , ..Xn ]. The subset of k n defined by:
V (S) := {(a1 , ....an ) ∈ k n : f (a1 , ....an ) = 0 ∀ f ∈ S}
is called an affine algebraic set . If hSi denotes the ideal generated by S in k[X1 , X2 , ..Xn ], clearly V (S) =
V (hSi). If S = {f (X1 , ..Xn )} is a singleton, V (S) is denoted V (f ), and is called a hypersurface in k n . More
generally, if S = {f1 , .., fm } is a finite set, we denote V (S) by V (f1 , .., fm ). It will turn out by the Hilbert
Basis Theorem in the next section that all algebraic sets are of the form V (f1 , .., fm ).
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 3

Definition 1.2.2. Given a subset Z ⊂ k n , we can define the ideal of functions vanishing on Z denoted I(Z)
inside the k-algebra k[X1 , X2 , ..Xn ] by:
I(Z) := {f ∈ k[X1 , X2 , ..Xn ] : f (a1 , ....an ) = 0 ∀ (a1 , ....an ) ∈ Z}

A natural question to ask is whether the correspondences I 7→ V (I) and Z 7→ I(Z) are bijective. We will
address this issue in §4, via the Hilbert Nullstellensatz.

We recall that if X is a compact Hausdorff space, the Banach algebra of continuous complex-valued functions
C(X) completely determines the space X upto homeomorphism. In fact, there is a 1-1 correspondence between
closed ideals in C(X) and closed subsets of X, and maximal ideals of C(X) and points of X. Indeed, the set
of maximal ideals of C(X) (called the maximal spectrum of X) can be given a topology which makes it
homeomorphic to X. That result is known as the Gelfand-Naimark Theorem, and can be looked up in most
advanced analysis texts (e.g. Simmons, or Rudin).

In the situation of affine algebraic sets, the analogous facts are consequences of the Hilbert Nullstellensatz,
as we shall see in §4. Towards defining the algebraic ring of functions on an affine algebraic set, we make the
following definition:

Definition 1.2.3. Let Z ⊂ k n be an affine algebraic set, and let I(Z) be the ideal in k[X1 , X2 , ..Xn ] as defined
above. We define the coordinate ring k[Z] of Z to be the quotient ring k[X1 , X2 , ..Xn ]/I(Z). This coordinate
ring is obviously a k-algebra.

Since the restrictions of two polynomials f, g ∈ k[X1 , .., Xn ] to Z agree pointwise on Z iff their difference
lies in I(Z), the coordinate ring maybe viewed as the ring of all functions on Z which are restrictions of
polynomials from the ambient space k n , two such polynomial functions being identified if they agree pointwise
on Z. Elements of this coordinate ring are called regular functions on Z, and the coordinate ring k[Z] is
sometimes called the ring of regular functions on Z.

Now, the central theme of invariant theory is the following. Suppose a group G acts algebraically on an
affine algebraic set Z which has the coordinate ring k[Z]. To say G acts algebraically, for say, a finite group
G, means that each element g ∈ G induces a k-algebra automorphism of k[Z]. In this k-algebra k[Z], there is
then the subalgebra k[Z]G of functions that are G-invariant.

As an example, the polynomial ring k[X1 , .., Xn ] is the coordinate ring of the affine n-space Z = k n . If we
consider the action of the permutation group G = Sn given by permutation of coordinates (viz. for a polynomial
f ∈ k[Z] = k[X1 , .., Xn ], and σ ∈ Sn , we define σ.f (X1 , .., Xn ) = f (Xσ−1 (1) , .., Xσ−1 (n) ).), then k[Z]G is the
subalgebra ofPsymmetric Pnpolynomials. It is known to be generated as a k-algebra by the elementary symmetric
n
polynomials i=1 Xi , i6=j=1 Xi Xj , ..., X1 X2 ..Xn . Now, can one realise this subalgebra as the coordinate ring
of some affine algebraic set, call it Z/G? In this particular case, it is not too difficult to see that by viewing
the affine space k n as the space of all roots of all polynomial equations of the form T n + a1 T n−1 + ... + an ,
the quotient space is precisely the space of all such polynomials, which is parametrised by the coefficients
(a1 , .., an ), so that Z/G in this case is again k n . The natural quotient map Z → Z/G is the map:
n
X n
X
(λ1 , .., λn ) 7→ (− λi , λi λj , ..., (−1)n λ1 λ2 ..λn )
i=1 i6=j=1

where the right side quite predictably consists of all the elementary symmetric functions in n variables.

For a general finite group, this issue is addressed in the section §5.3 in the sequel.

Even more generally, G could be an algebraic group, and one could again ask for a description of Z/G where
Z is an affine algebraic group on which G acts algebraically. This consitutes the still fertile and deep area of
algebraic geometry known as invariant theory.
4 VISHWAMBHAR PATI

2. Some Algebra

2.1. Division. We shall always be dealing with commutative rings with identity. Of course all ring homomor-
phisms will be required to preserve the identity element. We first need some facts on polynomial rings.
Proposition 2.1.1 (Weak Division Algorithm). Let A be a ring, and let A[X] be the polynomial ring in one
variable over A. Let f (X) and g(X) be two polynomials in A[X], of degrees m and n respectively. Let a be
the coefficient of the highest degree term in g, and let k = max(m − n + 1, 0). Then there exist polynomials
q(X) and r(X) in A[X] such that :
ak f (X) = q(X)g(X) + r(X)
where r(X) = 0 or deg r(X) < n = deg g(X).

Proof: If m − n + 1 < 0, we may take k = 0 and q(X) = 0, with r(X) = f (X). If m − n + 1 ≥ 0, we can
induct on this quantity m − n + 1. Clearly the induction starts because if m − n + 1 = 0, we make the same
choices as above, namely, k = q(X) = 0, and f (X) = r(X). Assume by induction that the result is proven for
all r such that r − n + 1 < m − n + 1, i.e. r < m. Consider the polynomial
f1 (X) = af (X) − bX m−n g(X)
where b is the leading coefficient of f (X). This polynomial f1 (X) has degree m1 strictly less than m, and so
applying induction:
al f1 (X) = q1 (X)g(X) + r1 (X)
where l = m1 − n + 1 < k = m − n + 1, and r1 = 0 or deg r1 < n. Substituting f1 (X) = af (X) − bX m−n g(X),
we have :
al+1 f (X) = (q1 (X) + bal X m−n )g(X) + r1 (X)
Now multiplying both sides by ak−l−1 , and by setting
q(X) = ak−l−1 (q1 (X) + bal X m−n ) ; r(X) = ak−l−1 r1 (X)
we have the result. 2

Corollary 2.1.2 (Division Algorithm). In the setting above, let A = k a field. Then there exist polynomials
q(X) and r(X) in A[X] such that :
f (X) = q(X)g(X) + r(X)
where r(X) = 0 or deg r(X) < n = deg g(X).

Proof: Just divide by ak in the proposition above. 2

Remark 2.1.3. An integral domain in which the division algorithm holds is called a Euclidean Domain,
and a familiar example is the ring of integers Z. By the last corollary, k[X], where k is a field, is also an
Euclidean domain. In a Euclidean domain, one can take greatest common divisors, and also prove unique
prime factorisation. Unfortunately, the polynomial ring in several variables over a field is not an Euclidean
domain. But it is something weaker, called a unique factorisation domain, which we shall discuss in the next
section. In such a ring, one may still factorise uniquely into primes, take greatest common divisors, etc.

2.2. Factorisation. In this subsection, we look at factorisation in commutative rings. Necessarily, the account
is very condensed, and the reader is urged to consult Jacobson’s Basic Algebra, Zariski-Samuel’s Commutative
Algebra or Lang’s Algebra for more details.

In this subsection A denotes an integral domain.

Definition 2.2.1 (Units, irreducibles, primes). An element a 6= 0 in A is said to be a unit if there exists b ∈ A
such that ab = 1, i.e. a is invertible. An element a 6= 0 in A is said to be irreducible if a = bc implies either
b or c is a unit. This shows that the only divisors of an irreducible element a are units or a multiplied with a
unit. An element a 6= 0 in A is said to be prime if for every b, c ∈ A such that a divides bc, we have a divides
b or a divides c. (We shall henceforth write a|b to denote that a divides b).
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 5

It is an easy exercise to check that prime elements

√ are always irreducible. However the converse is false.
One
√ can easily
√ check that in the ring A = Z[ 5], the element
√ 2 is irreducible, but divides the product
( 5 + 1)( 5 − 1) = 4 without dividing either of the factors 5 ± 1, and hence not a prime in A.

Definition 2.2.2 (Unique Factorisation Domain). An integral domain A is called a unique factorisation do-
main (UFD) or a factorial ring if the following two conditions are satisfied:

(UFD1): If a 6= 0 is an element of A, then there exist irreducible elements pi , and positive integers ri for
i = 1, 2, ..n such that
a = up1r1 p2r2 ..pnrn
where u is a unit.

(UFD2): Every irreducible element in A is prime.

From the conditions above, the reader can also easily check that in a UFD, the factorisation into irreducibles
is unique, upto multiplication by units and ordering of the irreducible factors. We define, for a non- zero element
a ∈ A, A a UFD, p a prime, the non-negative integer vp (a) by vp (a) = r if p|a and r is the largest power of p
dividing a, and vp (a) = 0 if p does not divide a. Clearly
vp (ab) = vp (a) + vp (b)
and
vp (a + b) ≥ min(vp (a), vp (b))
To be consistent with the two properties above for vp , one defines vp (0) = ∞.

Thus, in a UFD, one may define their greatest common divisor as follows:
Definition 2.2.3. (Greatest common divisor in a UFD) Let A be a UFD and a, b ∈ A be two non-zero
elements. The greatest common divisor (gcd) of a and b, denoted (a, b), is :
(a, b) = Πp prime prp
where rp = min(vp (a), vp (b)) and the product is finite since rp = 0 for all but finitely many primes (=irre-
ducibles). Note that greatest common divisors are unique upto multiplication by a unit.

Exercise 2.2.4. Show that an integral domain A is a UFD iff the following condition holds:

(UFD) If a 6= 0 is any element of A, then there exist irreducible elements pi , and positive integers ri for
i = 1, 2, ..n such that
a = up1r1 p2r2 ..pnrn
where u is a unit. Further, if there is another such factorisation
a = vq1s1 q2s2 ..qnsn
into irreducibles, then there exists a permutation σ such that qi = vi pσ(i) for some unit vi , and si = rσ(i) , for
each i = 1, 2, .., n.

Exercise 2.2.5. Show that an integral domain A is UFD iff the following condition holds:

(UFD’) If a 6= 0 is any element of A, then there exist prime elements pi , and positive integers ri for i = 1, 2, ..n
such that
a = up1r1 p2r2 ..pnrn
where u is a unit.
6 VISHWAMBHAR PATI

√ √
Remark 2.2.6. We saw above that Z[ 5] is not a UFD. It is true that Z[ d]√ is not a UFD for d ≡ 1
mod 4. This is because this ring fails to be integrally closed in its quotient field Q[ d] (see later section §4 for
definitions, the Exercise 4.2.19 there.) On the other
√ hand, for d ≡ 2, 3 mod 4, it is unknown which are UFD’s.
It is however known that for d < 0, the only Z[ d] which is a UFD is when d = −1 or d = −2. (see Theorem
7.7 of Chapter 11 in Artin’s Algebra).

An important subclass of UFD’s are the so-called PID’s, which are defined below.

Definition 2.2.7 (Principal Ideal Domains). An integral domain A is said to be a principal ideal domain
(PID) or principal ring if every ideal in A is a principal ideal, i.e. is generated by a single element.
It is easy to see that the ring of integers Z and the polynomial ring k[X] in one variable over a field k are
PID’s, because of the Division Algorithm. Just take an ideal I in either of these rings. Let r 6= 0 be an element
in this ideal of least modulus (in the case of Z) or least degree (in the case of k[X]). Then every other element
a 6= 0 of this ideal is either divisible by r, or leaves a remainder r1 of modulus (resp. degree) less than that of
r. In the latter case, since a and r are in I, so is r1 . But this contradicts the choice of r. Thus every element in
I is divisible by r. Incidentally, there do exist PID’s which are not Euclidean domains, but we needn’t bother
with them presently. We enumerate some facts on PID’s below.

Proposition 2.2.8. Let A be a PID. Then every ascending chain of ideals:

a0 ⊆ a1 ⊆ ...ak ⊆ ak+1 ⊆ .....
is eventually stationary.
Proof: Let a = ∪∞i=0 ai , which is easily seen to be an ideal. Since A is a PID, we have an element a ∈ A such
that a = hai. Then a ∈ an for some n, and thus we have a ⊆ an , which implies aj = an for all j ≥ n. 2

Proposition 2.2.9. If A is a PID, then for two non-zero elements a, b of A, their greatest common divisor
d = (a, b) exists, and is a generator of the ideal ha, bi. It is unique upto multiplication by a unit.
Proof: Consider the ideal I = ha, bi generated by a and b. Since there exists a d ∈ A such that I = hdi, we
have that d = λa + µb, where λ, µ ∈ A. This shows that any element dividing both a and b must divide d.
On the other hand since a, b both belong to I = hdi, d divides both a and b. Thus d is the required greatest
common divisor. It is trivial to see that d is well-defined upto multiplication by a unit, because two generators
of a principal ideal differ by a unit. 2

Proposition 2.2.10. If A is a PID, then the condition UFD2 holds for A.

Proof: Let x be an irreducible element of A, and let x | ab. Let us assume that x does not divide a. Let
d = (x, a), the greatest common divisor guaranteed by the previous proposition. Since d | x and x is irreducible,
we have d = ux or d = v, where u, v are units. In the former case x would divide a, contrary to assumption.
Thus d = v. This means that v = λx + µa, and mutliplying with v −1 , we have 1 = v −1 (λx + µa), from which
it follows that b = v −1 (λxb + µab). Now since x | ab, we have that x divides the right hand side, so x | b . 2

Corollary 2.2.11. Let A be a PID. An ideal a = hxi with x 6= 0 in A is maximal iff x is prime.
Proof: Let x be prime. Then if hxi ⊆ hyi, we have that y divides x. Since x is prime, we have y is a unit, or
a unit times x. In the former case hyi = A, and in the latter case hxi = hyi, showing that a = hxi is maximal.
Conversely, suppose a = hxi is maximal, and x | ab and that x does not divide a. Then the ideal generated by
x and a, namely hx, ai, strictly contains the maximal ideal a = hxi. Thus hx, ai = A, and we have 1 = λx + µa
which implies b = λxb + µab. Since x | ab we have x divides the right hand side, and hence x | b. 2

Proposition 2.2.12. If A is a PID, then A is a UFD.

 LECTURES ON BASIC ALGEBRAIC GEOMETRY 7

Proof: We have already seen above that A satisfies (UFD2). It suffices to show the factorisation of a general
non-zero element x ∈ A into a unit times finitely many irreducibles. If x is prime, we are done. If not, by the
previous proposition 2.2.11 the ideal hxi is not maximal, and is included in a maximal ideal hp1 i, so that p1 is
a prime by the previous proposition 2.2.11, and hxi ⊂ hp1 i implies p1 is a prime divisor of x. Write x = p1 x1 .
Now apply the same reasoning to x1 in place of x. We then have an ascending chain of ideals

hxi ⊂ hx1 i ⊂ ....hxn i ⊂ ...

which must become stationary at some point, by the proposition 2.2.8 above. This proves UFD1, and the
proposition. 2

There are UFD’s which are not PID’s. For example, we shall see below that Z[X] is a UFD. However
(exercise!), the ideal h2, Xi is not a principal ideal.

Definition 2.2.13. Let A be a UFD, and let f (X) = a0 + a1 X + ...an X n be a polynomial in A[X]. We define
the content of f, denoted by c(f ) to be the greatest common divisor of a0 , a1 , ..., an . Note that c(f ) is defined
upto multiplication by a unit. We say a polynomial f (X) is a primitive polynomial if c(f ) is a unit.

Clearly every polynomial f (X) maybe written uniquely as c(f )f1 (X), where f1 (X) is a primitive polynomial.
We have the following :

Proposition 2.2.14 (Gauss’ Lemma). Let A be a UFD, and f (X) and g(X) be two polynomials in A[X].
Then :

c(f g) = c(f )c(g)

Proof: Write f = c(f )f1 and g = c(g)g1 , where f1 and g1 are primitive. Then f g = c(f )c(g)f1 g1 . It suffices
Pn f1 g1 is a primitive polynomial.
to show that Pm Let p be any prime factor dividing all the coefficients of f1 g1 . Let
f1 (X) = i=0 ai X i and g1 (X) = i=0 bi X i . Since f1 and g1 are primitive, there must be positive integers j
and l such that p | ai for 0 ≤ i ≤ j − 1 but p does not divide aj , and similarly p | bk for 0 ≤ k ≤ l − 1 but p
does not divide bl . Now note that the coefficient of X j+l is of the form (aj bl + terms divisible by p), by the
above choice of j and l. Since p divides this coefficient, it must divide aj bl . Since p is prime, it must divide
either aj or bl . This contradicts the choice of j and l. Hence f1 g1 is primitive, and we are done. 2

Proposition 2.2.15. Let A be a UFD and b 6= 0 be an element of A, and f (X) be a polynomial in A[X]. Let
g be a non-zero primitive polynomial such that g | bf . Then g | f .
Proof: g | bf implies that gh = bf , which implies c(g)c(h) = c(h) = c(b)c(f ) = bc(f ) by Gauss’ lemma and
primitivity of g. But this means that b | c(h). Thus h = c(h)h1 = bH, where H is some polynomial. Now
substituting in gh = bf , we have gH = f , and the proposition follows. 2

Corollary 2.2.16. Let A be a UFD, and Q denote its quotient field. If f (X) is an irreducible polynomial in
A[X], then it is irreducible in Q[X].

Proof: Write f = gh, with g, h ∈ Q[X]. Clearly, multiplying by the denominators occurring in the coefficients
of g and h, we have bf = GH where G, H ∈ A[X], and b ∈ A. Write G = c(G)G1 , where G1 is a primitive
polynomial in A[X] (since A is a UFD). Thus the primitive polynomial G1 divides bf .

By the previous proposition 2.2.15, it follows that G1 divides f in A[X]. Since f is irreducible in A[X], G1
must be a unit or a unit times f in A[X]. But it is easy to see that units in A[X] are precisely the units in A,
so G1 = u or G1 = uf , where u ∈ A is a unit in A. Since g is a non-zero multiple of G in Q[X], and thus a
unit times G1 in Q[X], we have that g is a unit or a unit times f in Q[X], i.e., f is irreducible in Q[X]. The
proposition follows. 2.
8 VISHWAMBHAR PATI

Proposition 2.2.17. If A is a UFD, then A[X] is also a UFD.

Proof: (UFD1) is easy. Write f = c(f )f1 , where f1 is primitive. By factorising c(f ) into irreducibles in A
which are also clearly irreducible in A[X], we are reduced to considering the case of f primitive of degree
greater than 0. If it is irreducible, we are done, or else it is factorisable as gh, where neither factor has degree
0 because f is primitive. Thus both g and h have degree strictly less than that of f , and are still of degree
greater than 0. We are through by induction.

To show (UFD2), let p be an irreducible polynomial dividing a product f g, and let us assume p does not
divide f . Consider the ideal a generated by p and f . Let φ be the polynomial of least degree in a. By the
weak division algorithm, proposition 2.1.1, we have ak f = bφ + r, where a is the leading coefficient of φ and
the degree of r is strictly less than that of φ if r 6= 0. Since ak f − bφ ∈ a, the case r 6= 0 would contradict
that φ has least degree in a, so r = 0. Thus ak f = bc(φ)φ1 where φ1 is primitive. By the proposition 2.2.15
above, φ1 divides f . Similarly, φ1 divides p. Since p is irreducible, we have that φ1 is either a unit, or a
unit times p. The latter possibility would imply that p | f , since we have seen that φ1 | f , contradicting our
assumption that p does not divide f . Thus φ1 is a unit, and φ = c(φ) := c ∈ A. Also c ∈ a implies c = αf + βp,
where α, β ∈ A[X]. Then cg = αf g + βpg. By assumption p divides the right hand side, so p | cg. Since p is
irreducible, and therefore primitive, the proposition 2.2.15 implies that p | g. Hence A[X] is a UFD. 2

Corollary 2.2.18. If A is a UFD, A[X1 , ...Xn ] is a UFD.

Proof: A[X1 , ..., Xn ] = A[X1 , ..., Xn−1 ][Xn ] and induction. 2

Pn
Exercise 2.2.19 (Eisenstein’s Criterion for Irreducibility). Prove that if f (X) = i=0 ai X i is a polynomial
in Z[X], and p ∈ Z is a prime satisfying:

(i): p does not divide an .

(ii): p | ai for i = 0, 1, .., n − 1.

(iii): p2 does not divide a0

then f (X) is an irreducible polynomial in Q[X]. If f is primitive, then it is irreducible in Z[X]. As a
corollary to this criterion, prove that the polynomial 1 + X + X 2 + ... + X p−1 is irreducible in Q[X]. for p a
prime. What happens if p is not a prime?

Exercise 2.2.20. Show that the only irreducible polynomials in C[X] are of degree ≤ 1, and the only irre-
ducible polynomials in R[X] are of degree ≤ 1 or degree 2 polynomials with no real roots.

Exercise 2.2.21 (Power Series Rings). Let A be a ring, and A[[X]] denote the ring of formal power series in
the variable X. That is, an element f (X) ∈ A[[X]] is a formal expression
X∞
f (X) = ak X k
k=0
P∞
with ak ∈PA. Addition is defined in the obvious manner (coefficientwise), and multiplication of f = k=0 ak X k
∞
and g = k=0 bk X k by collecting coefficients, viz.
∞
X
gh = ck X k
k=0
P
where ck = j+l=k aj bl . We leave it to the reader to make the analogous definition for the power series ring
A[[X1 , .., Xn ]] in n-variables, and check that it is isomorphic to A[[X1 , .., Xn−1 ]][[Xn ]] in a canonical manner.
Also note that the polynomial ring A[X1 , .., Xn ] is a subring of A[[X1 , .., Xn ]] via the natural inclusion.

Show that:
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 9

(i): If A is an integral domain, so is A[[X]].

P∞
(ii): f (X) = k=0 ak X k is a unit in A[[X]] iff a0 is a unit in A. In particular, if A = k, a field, then the
units in k[[X]] are precisely the power series with non-zero constant term.

(iii): If m is a maximal ideal in A[[X]], then X ∈ m. (Hint: Show that 1 + aX is a unit for each a ∈ A[[X]]).

(iv): If m is a maximal ideal in A[[X]], then mc := m ∩ A is a maximal ideal in A, and m is generated by mc

and X. (Hint: Use (iii) above to show that mc consists of precisely the constant terms of all the elements
of m, and apply the canonical homomorphism A[[X]] → A sending each power series to its constant
term.)

(v): Conclude that if A is a local ring, i.e., it has a unique maximal ideal m, then A[[X]] is also a local ring
(with unique maximal ideal hm, Xi). Since a field k is a local ring (with unique maximal ideal {0}), it
follows that k[[X]] is a local ring, and by induction, the power series ring k[[X1 , .., Xn ]] is also a local
ring, with unique maximal ideal m = hX1 , .., Xn i.

(vi): A local ring A whose maximal ideal satisfies

∞
\
mk = {0}
k=1

is called a complete local ring. Show that if A is a complete local ring, then so is A[[X]]. Conclude that
k[[X1 , .., Xn ]] is a complete local ring. Show that if f, g ∈ k[[X1 , .., Xn ]] satisfy f ≡ g (mod mk ) for all
k ≥ 1, (where m = hX1 , .., Xn i), then f = g. Note that determining f (mod mk ) amounts to knowing all
the terms in f of degree ≤ k − 1.

Before we investigate factoriality in k[[X1 , .., Xn ]], k a field, we need a very important result which allows
us to reduce many questions on formal power series to questions on polynomials. For notational convenience
we shall denote by f (0, 0, .., Xn ) the power series in k[[Xn ]] which is the image of f ∈ k[[X1 , .., Xn ]] under the
natural surjection
k[[X1 , .., Xn ]] → k[[X1 , .., Xn ]]/hX1 , .., Xn−1 i = k[[Xn ]]

Theorem 2.2.22 (Weierstrass Preparation Theorem). Let f ∈ k[[X1 , ..., Xn ]] be a formal power series satis-
fying f (0, 0, .., Xn ) 6= 0 in k[[Xn ]]. Then there exists a unit u ∈ k[[X1 , .., Xn ]], an integer d ≥ 0, and elements
bi ∈ k[[X1 , .., Xn−1 ]] for i = 0, .., d − 1 such that:
f = u(Xnd + bd−1 Xnd−1 + ... + b0 )
satisfying bi (0, .., 0) = 0 in k for 0 ≤ i ≤ (d − 1). Further, this expression for f is unique.

Proof: For notational convenience, denote the ring k[[X1 , .., Xn−1 ]] by A, and its unique maximal ideal
hX1 , .., Xn−1 i by m. Thus f ∈ A[[Xn ]], and we may write:
∞
X
f= ak Xnk
k=0

where ak ∈ A are formal power series in the variables (X1 , .., Xn−1 ).

Let d ≥ 0 be the smallest integer m such that am (0, .., 0) 6= 0 in k. Clearly d exists since we have assumed
f (0, 0, .., Xn ) is not identically 0. In the series expression of f above, it is the first coefficient am which is not
in m, i.e. has a non-zero constant term.

Let us write u = c0 + c1 Xn + ..cj Xnj + ..... If we require u to be a unit, we must have c0 a unit in A, by (ii)
of 2.2.21. Thus c0 is a unit modulo m.
10 VISHWAMBHAR PATI

We have to solve for bi and cj so that:

a0 = b0 c0
a1 = b0 c1 + b1 c0
.
ad−1 = b0 cd−1 + b1 cd−2 + ... + bd−1 c0
ad = b0 cd + b1 cd−1 + ... + bd−1 c1 + c0
.
ad+1 = b0 cd+1 + b1 cd + ... + bd−1 c2 + c1
am+d = b0 cm+d + b1 cm+d−1 + ... + bd−1 cm+1 + cm for all m ≥ 0
By (vi) of the Exercise 2.2.21 above, to determine bi and cj for i, j ≥ 1, it is enough to determine them
modulo mr for every r. The left hand sides of the first d equations above are all ≡ 0 (mod m), and since c0 6= 0
(mod m), we have b0 ≡ 0 (mod m). Plugging in the second equation, we get b1 ≡ 0 (mod m), and then moving
down the rest, we have b0 , .., bd−1 are all ≡ 0 (mod m). Plugging this into the remaining equations we obtain
that cj ≡ ad+j (mod m) for all j ≥ 0. Thus all the bi and cj are determined (mod m).

Now suppose, we have inductively determined all the bi and cj (mod mr ). We wish to determine them (mod
m ). Since b0 ≡ 0 (mod m), and we now know c0 (mod mr ), it is easy to see from the first equation a0 = b0 c0
r+1

that b0 gets determined modulo mr+1 . Combining this with the facts that c1 is already known (mod mr ) and
b0 ≡ 0 mod m, it again follows that b0 c1 is determined (mod mr+1 ). Plugging c0 (mod mr ), and b1 (mod mr )
in the second equation determines b1 (mod mr+1 ). Proceeding down to the remaining equations upto the d-th
one determines all the bj (mod mr+1 ). By putting in these bj (mod mr+1 ) in the (d + 1)-th equation onwards
determines all the cj (mod mr+1 ). The induction is complete, and the existence of a factorisation as claimed
is proved. The uniqueness is obvious, because d is uniquely determined by f , and the unit u is uniquely
determined as the coefficient of Xnd in any two such factorisations. Equating the coefficients of Xnd−1 , ..., etc
determines the bj uniquely. 2

Remark 2.2.23. The Weierstrass Preparation Theorem above crops up in many contexts, notably the theory
of holomorphic functions of several complex variables. It reduces may questions on holomorphic functions of
several variables into questions about polynomials. See, for example, the book of Gunning and Rossi entitled
Analytic Functions of Several Complex Variables for the holomorphic version of 2.2.22.

Exercise 2.2.24. Let k be an infinite field, and let f 6= 0 be a formal power series in the ring k[[X1 , .., Xn ]].
Show that we can make a linear non-singular change Xi = τi (Y1 , .., Yn ) of variables to Y1 , .., Yn such that the
transformed power series
fe(Y1 , .., Yn ) := f (τ1 (Y1 , .., Yn ), .., τn (Y1 , .., Yn )) ∈ k[[Y1 , .., Yn ]]
satisfies fe(0, 0, .., Yn ) 6= 0 in k[[Yn ]].

Now we can prove that k[[X1 , .., Xn ]] is a UFD. We prove it below only for an infinite field, but it is true
for all fields.

Proposition 2.2.25. Let k be an infinite field. The power series ring k[[X1 , .., Xn ]] is a UFD.

Proof: The case of n = 1 is trivial, because every element can be written as a unit times X k , and from this
it easily follows that k[[X]] is a principal ideal domain (in fact, every ideal is a power of the unique maximal
ideal m = hXi). From the proposition 2.2.12, it follows that k[[X]] is a UFD.

Inductively assume that A := k[[X1 , .., Xn−1 ]] is a UFD. Let f ∈ A[[Xn ]] = k[[X1 , .., Xn ]] be a non-zero
element. In view of exercise 2.2.24, we may assume without loss of generality that f (0, 0, .., Xn ) 6= 0.

By the Weierstrass Preparation Theorem 2.2.22, we may write:

f = ug
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 11

where u ∈ A[[Xn ]] is a unit, and g is a polynomial, lying in A[Xn ]. Since A is a UFD by the induction
hypothesis, so is the polynomial ring A[Xn ], by the proposition 2.2.18. Hence we may write:
g = vg1s1 g2s2 ...grsr
where v is a unit in A[Xn ] (hence in A), and gi are irreducible elements in A[Xn ]. Thus f is factorised as:
f = uvg1s1 g2s2 ...grsr
We claim that the polynomials gi ∈ A[Xn ] continue to be irreducible in A[[Xn ]]. This is because f (0, .., Xn ) 6= 0
implies (since A[[Xn ]] is an integral domain) that gi (0, 0, .., Xn ) 6= 0 for i = 1, .., r. If gi = G.H in A[[Xn ]],
with G, H ∈ A[[Xn ]], it follows that G(0, .., Xn ) 6= 0 and H(0, .., Xn ) 6= 0.

Thus, by the Weierstrass Preparation Theorem 2.2.22 applied to G, H, we have that G = u1 P , H = u2 Q

for some units ui ∈ A[[Xn ]] and P, Q ∈ A[Xn ]. Thus:
gi = wP.Q
where w = u1 u2 is a unit. Since gi is a polynomial in A[Xn ], the uniqueness assertion in the Weierstrass
Preparation Theorem implies that one of P or Q is a unit in A[[Xn ]]. So one of G or H is a unit in A[[Xn ]],
proving our assertion that gi are all irreducible in A[[Xn ]].

We now need to show that the factorisation of f above is unique. The factorisation f = ug above, by 2.2.22,
is unique, and the factorisation of g ∈ A[Xn ] into irreducibles is unique (upto units in A), since A[Xn ] is a
UFD, so that the factorisation of f above is unique. By the exercise 2.2.4, the ring k[[X1 , .., Xn ]] is a UFD. 2

Remark 2.2.26. The reader may wonder why one didn’t adopt induction to show that k[[X1 , .., Xn ]] is a
UFD as we did for polynomial rings. The reason is that there exist UFD’s A such that A[[X]] is not a UFD
(see Lang, Algebra, p. ). It is however known that if A is a PID, then A[[X1 , .., Xn ]] is a UFD. (See Bourbaki,
Commutative Algebra). In particular Z[[X1 , .., Xn ]] is a UFD, as is k[T ][[X1 , .., Xn ]] for k a field. If A is a
discrete valuation ring (viz. an integrally closed Noetherian domain with a unique non-zero prime-ideal), then
again A[[X1 , .., Xn ]] is a UFD.

2.3. Noetherian Modules and Rings.

Definition 2.3.1. A module M over a ring A is said to be Noetherian if every A-submodule N of M is finitely
generated. A ring is said to be Noetherian if it is Noetherian considered as a module over itself, i.e., if every
ideal of A is finitely generated.

Proposition 2.3.2. For an A module M , the following conditions are equivalent:

(i): Every non-empty set {Mα }a∈Λ of submodules of M has a maximal element.
(ii): Every ascending chain M0 ⊆ M1 ⊆ M2 ..... ⊆ Mn ⊆ ...of submodules of M becomes eventually station-
ary.
(iii): M is Noetherian

Proof:
(i)⇒(ii)
Let M0 ⊆ M1 ⊆ M2 ..... ⊆ Mn ⊆ .... be an ascending chain of submodules of M . Then the family {Mi }∞
i=1
must have a maximal element, say N . Then N = Mj for some j, and by maximality, Mk ⊇ Mj = N for all
k ≥ j implies that Mk = Mj for all k ≥ j.

(ii)⇒(iii)
Assume M is not Noetherian, so some submodule N ⊆ M is not finitely generated. Then, we can choose a
countable family of distinct elements {xi }i∈N of N , such that the chain of submodules of N ,
hx1 i ⊆ hx1 , x2 i ⊆ ... ⊆ hx1 , x2 , ..., xi i ⊆ ...
is strictly ascending, and so does not become stationary, contradicting (ii).

(iii)⇒(i)
12 VISHWAMBHAR PATI

Let {Mα }a∈Λ be a family of submodules of M . Order it by inclusion, and note that every totally ordered
subfamily Γ ⊂ Λ will have the submodule N = ∪α∈Γ Mα , as an upper bound. We claim that N = Mγ , for
some γ ∈ Λ, i.e. N is a member of this family Γ. By hypothesis M is Noetherian, so N is finitely generated,
i.e. N = hx1 , x2 , ..xn i. Since Γ is totally ordered, there will exist an Mγ for some γ ∈ Γ, which contains all of
the {xi }ni=1 . Clearly N = Mγ . This proves the claim. Now Zorn’s Lemma proves that the family {Mα }a∈Λ
has maximal elements, and the proposition follows. 2

Proposition 2.3.3. If M is a Noetherian module, then every quotient and submodule of M is Noetherian. If
there is a short exact sequence :
i π
0 −→ L −→ M −→ N −→ 0
of A-modules with L and N Noetherian, then M is Noetherian.
Proof: The first two assertions are clear by the definition.

Recall that the sequence above is said to be exact if Im i = ker π, i is injective and π is surjective. Now,
for the second statement, let P be a submodule of M . Then π(P ) is finitely generated since it is a submodule
of the Noetherian module N . Let {π(xi )}ni=1 be a finite set of generators for π(P ), where xi ∈ P . Since L
is given to be Noetherian, so is its isomorphic image i(L). Let {yj }m j=1 be a finite set of generators for the
Pn Pn
submodule P ∩ i(L) of i(L). Then, for x ∈ P , π(x) Pm = i=1 ai π(xi ), so that the element x − i=1 ai x i is in
P ∩ kerπ = P ∩ i(L), and is therefore equal to j=1 bj yj for some bj . Thus P is generated by the m + n
elements {x1 , ...xn , y1 ...., ym }, and the proposition is proved. 2

Example 2.3.4. Every field is a Noetherian ring, as is every PID. So Z, k[X], Z[i] are all Noetherian rings.

Example 2.3.5. The ring C([0, 1]) of continuous real or complex valued functions is not Noetherian. The
ideals:
am = {f : f (x) = 0 ∀ x ∈ [0, 1/m]}
for m ∈ N is an ascending chain of ideals which is not stationary.

Proposition 2.3.6. Let A be a Noetherian ring, and M be a finitely generated A-module. Then M is a
Noetherian module.
Proof: If M is generated by a single element, say x, then it is a quotient of the Noetherian module A, i.e. it
is isomorphic to A/a where the ideal a is defined by :
a = Ann x := {a ∈ A : ax = 0}
called the annihilator of x. Thus M is a quotient of the module A, which is a Noetherian module since it is a
Noetherian ring. So it is Noetherian, by the first part of the proposition 2.3.3. Now proceed by induction on
the number of generators, and the second part of proposition 2.3.3 about exact sequences. 2

We now prove a very crucial result about polynomial rings:

Proposition 2.3.7 (Hilbert Basis Theorem). If A is a Noetherian ring, then the polynomial ring A[X] is also
a Noetherian ring.
Proof: Let a be an ideal in A[X], and let I be the ideal of A generated by all the leading coefficients of
elements of a. Since A is Noetherian, there is a finite set of generators, say {ai }ni=1 , for I. Let fi (X) be a
polynomial in a with leading coefficient ai for i = 1, 2..., n, and let ri be the degree of fi (X).

Let r = max{ri : 1 ≤ i ≤ n} and M be the A−submodule of A[X] consisting of all polynomials of degree
less than or equal to (r − 1) (i.e. M = A + AX + ...AX r−1 ). Since M is finitely generated, and A is Noetherian,
we have that M is Noetherian,by proposition 2.3.6. Therefore the A-submodule a ∩ M of M is also finitely
generated as an A-module, by proposition 2.3.3.
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 13

Let {tj }m
j=1 be a set of generators for this A-module a∩M . We claim that the set S = {f1 , f2 , .., fn , t1 , .., tm }
generates a. If f (X) is any polynomial in a of degree less than or equal to r − 1, then f (X) lies in a ∩ M
and hence is an A-linear combination of the elements tj . Therefore,Pnassume that deg(f ) ≥ r. We induct on
k = deg(f ). Write f (X) = aX k + ..... + a0 . We have a ∈ I, so a = i=1 λi ai . Consider the element :
n
X
g(X) = f (X) − λi X k−ri fi .
i=1

which makes sense since k ≥ r ≥ ri for i = 1, 2, ..n. Since f , P

and fi all belong to a, this element g(X) also
n
belongs to a. On the other hand its leading coefficient is (a − i=1 λi ai ) = 0. P
Thus deg(g) < deg(f ), and by
n
induction hypothesis, g(X) belongs to the ideal generated by S. Also, so does i=1 λi X k−ri fi , and hence so
does f (X), and we are done. 2

Corollary 2.3.8. If A is Noetherian, so is A[X1 , ....Xn ], by induction on the number of variables and the
proposition above. If k is a field, k[X1 , ....Xn ] is a Noetherian ring.

Definition 2.3.9. Let k be a field. We say a k-algebra A is of finite type if A is generated as a k-algebra
by finitely many elements {y1 , y2 , ...yn }. That is, every element of A is a finite k−linear combination (not
necessarily unique) of elements in the set
{y1r1 y2r2 ...ynrn : ri ∈ Z+ }

Clearly, a k-algebra of finite type is the quotient of a polynomial ring k[X1 , ...Xn ], and hence we have the

Corollary 2.3.10. Every k-algebra of finite type is Noetherian.

Proof:
Let A be a k-algebra of finite type, generated by {yi }ni=1 , and let π : k[X1 , ...Xn ] → A be the k-algebra
homomorphism defined by π(Xi ) = yi . If a is an ideal in A, verify that π −1 (a) is an ideal in the polynomial
ring k[X1 , ...Xn ], and hence generated as an ideal by finitely many polynomials {fi (X1 , .....Xn )}m
i=1 . It is clear
that a = ππ −1 (a) is generated as an A-ideal by π(fi ) = fi (y1 , ..., yn ). 2

Example 2.3.11. It is possible for a Noetherian ring to contain a non-Noetherian subring. By the above
corollary, the polynomial ring k[X, Y ] in two variables is Noetherian. However, consider A ⊂ k[X, Y ] to be the
k-subalgebra generated by the infinitely many elements X, XY, XY 2 , ..., XY i , .... Then if we define the chain
of ideals ai = hX, XY, .., XY i i, we see that ai+1 strictly contains ai for all i, and this ascending chain does
not become stationary. Thus A is not Noetherian. Also, in particular, by the above Corollary 2.3.10, A is not
a k-algebra of finite type, even though it is a subalgebra of k[X, Y ], which is a k-algebra of finite type.

We now look at the Noetherian-ness of power series rings (see the Exercise 2.2.21 for the definitions).
Proposition 2.3.12. Let A be a Noetherian ring. Then the ring of formal power series A[[X]] is a Noetherian
ring.

Proof: Let a ⊂ A[[X]] be an ideal. For j = 0, 1, ..., define

qj := {a ∈ A : ∃ f ∈ a with f = aX j + ....higher order terms}
(Compare with the proof of 2.3.7, where we considered the ideal of leading coefficients, i.e. coefficients of
the degree terms). It is easy to check that qj is an ideal, and since Xf ∈ a for f ∈ a, it follows that:
q0 ⊆ q1 ⊆ ... ⊆ qj ⊆ qj+1 ....
is an ascending chain of ideals in A. By (ii) of the proposition 2.3.2, this chain becomes eventually stationary,
which implies qj = q for all j ≥ r.
14 VISHWAMBHAR PATI

Since A is Noetherian, q is finitely generated, so let {a1 , .., am } be a set of generators for q. Let f1 , .., fm ∈
qr = q such that fj = aj X r + ..(higher order terms).

For any power series f = a0 + a1 X + ..aj X j + .., define the order of f by

ord(f ) = min{k : ak 6= 0}
We claim that each f ∈ a with ord(f ) ≥ r is in the ideal hf1 , .., fm i ⊂ a. For let
f = br X r + ....
Pm
be an element of a. Then br ∈ qr = q, and we can write br = j=1 λ0j aj , where λ0j ∈ A. Then we have:
m
X
f− λ0j fj = br+1 X r+1 + .....
j=1
Pm
which is in a, so that br+1 ∈ qr+1 = q. Again write br+1 = j=1 λ1j aj , and note that
m
X m
X
f− λ0j fj − λ1j Xfj = br+2 X r+2 ....
j=1 j=1

That is,
m
X
f− (λ0j + λ1j X)fj = br+2 X r+2 + ...
j=1
Proceeding inductively, we find that
m
X
f= (λ0j + λ1j X + ..λij X i + ....)fj
j=1

which proves our claim.

Now let aj1 , .., ajmj be a set of generators for qj for j = 0, 1, .., r − 1. Let fj1 , .., fjmj be elements of qj such
that:
fjl = ajl X j + ....(higher order terms) for 1 ≤ l ≤ mj , 0 ≤ j ≤ r − 1
Pm0
Now if f = b0 + .. is any element of a, b0 ∈ q0 , and we can write b0 = l=1 µ0l a0l , with µ0l ∈ A. Then the
element:
m0
X
f− µ0l f0l
l=1
has order ≥ 1, and can be written as:
m0
X
f− µ0l f0l = b1 X + ....
l=1
and since this element is in a, we have b1 ∈ q1 . Proceeding by the inductive method outlined above, we will
finally have  
mj
r−1 X
X
ord f − µjl fjl  ≥ r
j=0 l=1
with µjl ∈ A. In view of the claim in the previous paragraph, this implies that a is generated by the finite set:
S = {fjl : 0 ≤ j ≤ r − 1, 1 ≤ l ≤ mj } ∪ {f1 , .., fm }
Thus a is finitely generated, and A[[X]] is Noetherian.
2

Corollary 2.3.13. Let k be a field. Then the power series ring k[[X1 , .., Xn ]] is Noetherian.

Proof: Apply the proposition above, inducting on n. 2.

Here are some additional exercises on polynomials, power series, Noetherian rings and modules.
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 15

Exercise 2.3.14. Let k[X] be the polynomial ring in one variable over a field k. Prove that the principal ideal
hf (X)i is a maximal ideal iff f is irreducible (⇔ f is prime).

Exercise 2.3.15. For A an integral domain, consider the following conditions:

(i): Every ascending chain of principal ideals:

ha1 i ⊂ ha2 i.... ⊂ han i ⊂ ....
becomes stationary.

(ii): Every a 6= 0 in A can be factorised as a unit times finitely many irreducibles.

Show that the condition (i) implies the condition (ii). In particular, if A is a Noetherian integral domain, then
every element factorises as a product of a unit and irreducibles. Thus the “key” condition for A to be UFD is
the condition UFD2.

What about (ii) ⇒ (i)?

Exercise 2.3.16. For the power series ring k[[X]] in one variable, prove that the division algorithm holds.
(Hint: Consider the function δ(f ) = ord(f ), where ord(f ), the order of f is defined in the proof of Proposition
2.3.12.)

Exercise 2.3.17. Let A be a Noetherian ring and f : A → B be a ring homomorphism. We may make B
an A-module via f , i.e. define a.b := f (a)b. Suppose that B is finitely generated as an A-module, with this
A-module structure. Then show that B is a Noetherian ring.

Exercise 2.3.18. Let f : A → B be a ring homomorphism, with A Noetherian. Consider B as an A-algebra

via f , as in the previous exercise. Suppose B is an A- algebra of finite type (i.e. is a quotient of some polynomial
algebra A[X1 , .., Xn ]). Again show that B is a Noetherian ring.

Exercise 2.3.19. Let A be a Noetherian ring, and let f : A → A be a surjective ring homomorphism. Then
show that f is also injective. Similarly, show that if f : M → M is a surjective module homomorphism of
a Noetherian A-module M , then f is injective. (Hint: Consider the sequence of ideals (resp. submodules)
ker f ⊂ ker f 2 ⊂ ... ⊂ ker f n ⊂ ...)

Exercise 2.3.20 (Germs of analytic functions). Let k = C, and consider the subring of C[[X]] defined by:
∞
X
O0,an = {f = an X n : lim sup | an |1/n < ∞}
n→∞
n=0

This subring is called the ring of germs of holomorphic functions at 0, and represents all holomorphic functions
whose Taylor expansions are convergent in a neighbourhood of 0 (by using the Taylor expansion). Show that
O0,an is also a local ring, with unique maximal ideal m consisting of all functions vanishing at 0 (i.e. those
with zero constant term). Is it a Euclidean domain? A PID? A UFD? Upto units, what are the irreducibles of
O0,an ?

There is an obvious analogous ring (of germs of real analytic functions at 0 with k = R. Check that the
analogous properties hold there.
16 VISHWAMBHAR PATI

Exercise 2.3.21 (Germs of smooth functions). Define the germ of a smooth real (resp. complex) valued
smooth function at 0 to be the equivalence class [f, W ] where W is a neighbourhood of 0 and f is smooth
on W . The equivalence class is defined by [f, W ] = [g, V ] iff there exists a neighbourhood U ⊂ V ∩ W with
f|U = g|U . Prove that the collection of real (resp. complex) germs of smooth functions at 0 is a local ring, and
not Noetherian. What are the irreducible elements of this ring? Is it a UFD? Is it a subring of R[[X]] (resp.
C[[X]]) as in the previous exercise? What is the story for “smooth” above replaced by “continuous” ?

ExerciseP2.3.22. Show that Q[[X]] is not contained in the quotient field of Z[[X]]. (Hint: Consider the
∞
element i=1 X i /pi , where pi is the i-th prime). What is the quotient field of Z[[X]]?
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 17

3. Algebraic Sets
Let k be a field in whatever follows.

3.1. The Zariski Topology.

Proposition 3.1.1. Let X = V (a) be an affine algebraic set in k n . Then
X = V (f1 , ..fm ) = ∩m
i=1 V (fi )

where {fi (X1 , ...Xn )}m

i=1 is any set of generators for the ideal a. Thus every affine algebraic set is defined by
finitely many equations, and is a finite intersection of hypersurfaces.
Proof: Clear from the Hilbert Basis Theorem 2.3.7 of the last section. 2

Proposition 3.1.2 (The Zariski Topology). We have the following:

(i): If a1 and a2 are two ideals in A = k[X1 , ...Xn ], then :
V (a1 ) ∪ V (a2 ) = V (a1 ∩ a2 )
(ii): If {aα }α∈Λ is a family of ideals in A = k[X1 , ...Xn ], then
V (Σα∈Λ aα ) = ∩α∈Λ V (aα )
(iii): V (A) = V ({1}) = φ
(iv): V (0) = k n

Proof:
We prove only (i), leaving the rest to the reader. Since a1 and a2 both contain a1 ∩ a2 we clearly have
V (a1 ∩ a2 ) ⊃ V (a1 ) ∪ V (a2 ). On the other hand, if (a1 , ...an ) ∈ / V (a1 ) ∪ V (a2 ), then there are polynomials
fi (X1 , ...Xn ) ∈ ai , where i = 1, 2, such that fi (a1 , ...an ) 6= 0. Then f1 f2 is a polynomial in a1 ∩ a2 , which does
not vanish at (a1 , ...an ), and hence (a1 , ...an ) ∈
/ V (a1 ∩ a2 ). 2

Thus the family of subsets:

{V (a) ⊂ k n : a an ideal in k[X1 , ...Xn ]}
satisfies the axioms for closed sets of a topology on k n called the Zariski topology on k n . If k = F (:= R or
C), it is clear that a Zariski closed subset of Fn is closed in the classical (Euclidean metric) topology on Fn ,
since polynomials are continuous in this classical topology on and algebraic sets are common zero sets of these.
However, the Zariski topology is much coarser, as we shall see shortly.

Example 3.1.3. If n = 1, the Zariski closed subsets of k n = k are precisely the finite subsets of k, i.e. the
Zariski topology is the cofinite topology, since polynomials in one variable have only finitely many zeros !

Example 3.1.4. If n = 2, the Zariski closed subsets of k 2 are k 2 , the empty set, and all algebraic curves,
and their arbitrary intersections. The reader should convince herself that this is not the product topology of
Zariski topology on k with itself, which is strictly coarser.

Notation : 3.1.5 (Affine n-space over k). The set k n with this topological structure, coming from the Zariski
topology, will be denoted An (k), and called n-dimensional affine space over k. A1 (k) is called the affine line
and A2 (k) is called the affine plane.

Proposition 3.1.6. Let Y ⊂ An (k) be any subset, and let I(Y ) be the ideal of polynomials f ∈ k[X1 , ...Xn ]
vanishing on Y (see definition 1.2.2). Then the algebraic set V (I(Y )) is the Zariski closure Y . In particular,
for a closed subset Y ⊂ An (k), we have V (I(Y )) = Y .
18 VISHWAMBHAR PATI

Proof: It is clear that V (I(Y )) is a Zariski closed subset of An (k) containing Y , and hence Y ⊂ V (I(Y )). On
/ Y . Then there exists a Zariski closed set V (a) ⊃ Y , which excludes p,
the other hand suppose p = (a1 , ...an ) ∈
and hence there is a polynomial f ∈ a such that f (p) 6= 0. Now V (a) ⊇ Y implies that f (q) = 0 for all q ∈ Y .
Thus f ∈ I(Y ). Hence p ∈ / V (I(Y )). 2

We can induce the Zariski topology from An (k) to any Zariski closed subset, i.e. any algebraic set X ⊂ An (k).
If, for example, X is an algebraic set, and I = I(X) is the corresponding ideal, then by the last proposition 3.1.6,
we have X = V (I). Thus the collection :

{Z ∩ X : Z closed in An (k)}
is precisely the collection:
{V (a + I) : a an ideal in k[X1 , ...Xn ]}
of Zariski closed subsets of X. If we define, for an ideal b in k[X] = k[X1 , ...Xn ]/I (=the coordinate ring of
X as defined in 1.2.3) the set V (b) := V (π −1 (b)) where π is the quotient homomorphism from the polynomial
ring k[X1 , ...Xn ] to the coordinate ring k[X], then the above collection of closed sets in the induced Zariski
topology on X is just the collection:
{V (b) : b an ideal in k[X]}
This is clear in view of the definition 1.2.3 and the fact that the natural surjection π from k[X1 , ...Xn ] to k[X]
sets up a 1-1 correspondence between ideals in k[X] and ideals in k[X1 , ...Xn ] containing I , and that a + I is
the most general ideal containing I

Exercise 3.1.7. Show that any infinite subset of A1 (k) is Zariski dense in it. Let K = R or C. Show that the
graph of the exponential function, i.e. the set {(z, ez ) : z ∈ K} is Zariski dense in A2 (K).

Exercise 3.1.8. Show that if a ⊂ b are ideals in k[X1 , .., Xn ], then V (a) ⊃ V (b). Similarly, if X ⊂ Y are any
two subsets in An (k), then I(X) ⊃ I(Y ).

The proposition 3.1.6 above says that if we start with an algebraic (Zariski closed) subset Y ⊂ An (k),
compute its ideal I(Y ), and then compute the closed set V (I(Y )) defined by this ideal, then we recover Y .
Clearly it would be desirable to also start with an ideal a ⊂ k[X1 , .., Xn ] and compute the ideal I(V (a)), and
compare it with a. This is the statement of the Hilbert Nullstellensatz, which will be proved in the course of
the next two sections. Before that, we need to understand the notion of prime ideals, and the related notion
of the radical of an ideal, which is addressed in the next section.
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 19

4. More Algebra
4.1. Prime Ideals.
Definition 4.1.1. Let A be a ring with 1, commutative, as usual. We say an ideal p in A is a prime ideal if
p 6= A and, for each pair of elements x, y ∈ A such that xy ∈ p, either x ∈ p or y ∈ p.

Note that this is equivalent to saying that A/p is a non-zero integral domain. Note that {0} is a prime ideal
iff A is an integral domain. Since m is a maximal ideal iff A/m is a (non-zero!) field, it is clear that every
maximal ideal is a prime ideal. We emphasize that maximal and prime ideals are always, by definition, proper
ideals of A.

Exercise 4.1.2. Show that a principal ideal hxi in a ring is prime iff x is not a unit and a prime element of A.
If A is a UFD (e.g. a polynomial ring over a field k), then the hxi is prime iff x is a non-unit and irreducible.
Thus an ideal generated by a single irreducible polynomial of degree greater than zero in a polynomial ring
k[X1 , .., Xn ] is prime. Show that if A is a PID, every non-zero ideal is prime iff it is maximal. Show that the
ideal generated by X 2 + 1 in R[X] is maximal (hence prime), but the ideal generated by X 2 + 1 in C[X] is not
maximal (hence not prime). Find all the maximal ideals in R[X].

Qm
Exercise 4.1.3. If p ⊃ i=1 ai for some ideals ai ⊂ A, then p ⊃ ai for some i. In particular, if p is a prime
ideal such that p ⊃ ∩m
i=1 ai , then p ⊃ ai for some i.

Exercise 4.1.4. If ∪mi=1 pi ⊃ a for some prime ideals pi and some ideal a in a ring A, then pi ⊃ a for some
i. (Note that for m = 2, the primeness assumption on pi is not necessary, the result is true in general. For
m ≥ 3, use induction, and the primeness of the pi . )

√
Definition 4.1.5. Let A be as above, and let a be an ideal in A. We define the radical of a , denoted a by
√
a = {x ∈ A : xk ∈ a for some k ≥ 1}
√
which is easily checked to be an ideal. An ideal a is said to be a radical ideal if a = a. For example, a prime
ideal is clearly a radical ideal.

The nilradical of A is defined as the radical of the zero ideal, and denoted by nil(A).

Exercise 4.1.6. Let A be a UFD. Show that a proper principal ideal hxi in A is a radical ideal iff x factors
into a product of distinct primes.

Exercise 4.1.7. Show that for a subset S ⊂ An (k), k a field, the ideal I(S) in k[X1 , .., Xn ] is a radical ideal.

Proposition 4.1.8. Let A be any ring. The nilradical nil(A) of A is the intersection of all the prime ideals of
A.
p
Proof: Clearly the nilradical = {0}, is the ideal of all the nilpotent elements of A. If x ∈ nil(A), then
xk = 0 belongs to every prime ideal of A, and by the definition of a prime ideal, so does x. Conversely, assume
x∈/ nil(A), so that no positive power of x is zero. Let
Σ = {I : I is an ideal containing no power of x}
Clearly {0} ∈ Σ so Σ is a non-empty collection, which can be partially ordered by inclusion. Let {Iα }α∈Λ be
a totally ordered subset of Σ. It is clear that ∪α∈Λ Iα is an ideal which is an upper bound for this chain. So,
by Zorn, Σ has a maximal element, say I. We claim that I is a prime ideal. For let us assume a ∈ / I, and
b∈/ I. Then the ideal I + Aa is an ideal strictly containing I, and hence not in Σ. Thus xr ∈ I + Aa. Similarly
20 VISHWAMBHAR PATI

xs ∈ I + Ab. Thus xr+s ∈ I + Aab. Thus I + Aab ∈ / Σ, which implies I + Aab 6= I, and hence ab ∈
/ I. Hence I
is a prime ideal not containing x, and the proposition is proved. 2.

Corollary 4.1.9. If A is any ring and a an ideal in A, then

√
a = ∩{p prime ideal : p ⊃ a}

Proof: Apply the above proposition to A/a , noting that prime ideals in this ring lift to exactly to prime
ideals in A that contain a . 2

Definition 4.1.10. The collection of all prime ideals of a ring A is called the prime spectrum of A and denoted
Spec A. The collection of all maximal ideals of A is called the maximal spectrum of A, and denoted Spm A.

Exercise 4.1.11. Let f : A → B be a ring homomorphism. Show that there is an induced map of prime
spectra:
f ∗ : Spec B → Spec A
p 7 → f −1 p
Give an example of a ring homomorphism f : A → B and a maximal ideal m ⊂ B such that f −1 m is not a
maximal ideal of A.

Exercise 4.1.12.
√
n
(i): Let a be an ideal in a Noetherian ring. Show that a ⊂ a for some n ≥ 1.
P∞
(ii): Let f (X) = i=0 ai X i be a formal power series in the power series ring A[[X]], where A is a Noetherian
ring. Show that ai nilpotent for each i implies that f is nilpotent.

(iii): Give an example of a ring A whose nilradical is not nilpotent (Note: an ideal a is said to be nilpotent
if there exists an n such that an = {0}).

We will return to further facts on prime ideals later.

4.2. Integral Dependence. We need a new algebraic notion here. To motivate it geometrically, let us
consider the following example. Let C = V (XY − 1) ⊂ C2 be the hyperbola in C2 . Thus the coordinate
ring k[C] = k[X, Y ]/hXY − 1i. Consider also the line L1 = V (Y ), the x-axis. Let π : C → L1 be the map
π(x, y) = x, the projection onto the first coordinate. The following facts are clear:

(i): π −1 (a) = {(a, a−1 )} for a 6= 0, whereas π −1 (0) = φ, so π is not surjective.

(ii): The inverse image of each closed disc around the origin {x :| x |≤ ε} in L1 is unbounded, i.e. the map
π is not a proper map (with respect to the classical topologies on C and L1 ).

Now let us change the line, and use the line L2 = V (X − Y ), the diagonal in C2 , and project from C to L2 .
That is, define
τ :C L2 →
1
(x, y) 7→ (x + y)
2
It is convenient, now, to introduce new coordinates z = 21 (x + y) and w = 12 (x − y), so that x = z + w, and
y = z − w. Then C can be rewritten as V (Z 2 − W 2 − 1) and τ (z, w) = z. Now note that :
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 21

√
(i): τ −1 (a) = {(a, ± a2 − 1)} (in the new coordinate system) for all a ∈ L2 , i.e. the map τ is surjective.
The fibre τ −1 (a) has two points for a 6= 0, and one point for a = ±1.

(ii): With respect to the classical topologies on C and L2 , the inverse image τ −1 (K) of every compact subset
K of L2 is compact, i.e. τ is a proper map.

Let us try to see what is going on, algebraically. In the first instance, the map π : C → L1 induces the ring
homomorphism π ∗ : k[L1 ] → k[C] between the coordinate rings, taking a polynomial function f ∈ k[L1 ] to the
function π ◦ f ∈ k[C]. More precisely, if we write:

k[L1 ] = k[X, Y ]/hY i = k[X], k[C] = k[X, Y ]/hXY − 1i

and denote the functions X (mod hXY − 1i) and Y (mod hXY − 1i) in k[C] as x and y respectively, the
homomorphism π ∗ : k[L1 ] → k[C] is just the natural inclusion:
k[X] ,→ k[X, Y ]/hXY − 1i
taking X to x. Note that in the right hand coordinate ring, the element y satisfies the equation xy − 1 = 0,
whose coefficients are in k[X], but whose leading coefficient, viz. x, is not 1. We now make the following:
Definition 4.2.1. We call a polynomial monic if the coefficient of the leading (degree) term is 1.
Thus y does not satisfy a monic polynomial with coefficients in k[X].
In the second situation, the homomorphism τ ∗ : k[L2 ] → k[C] is just the inclusion of rings:
k[Z] ,→ k[Z, W ]/hZ 2 − W 2 − 1i
taking Z to z := Z (mod hZ 2 − W 2 − 1i). In this case, the function w, which is defined by w = W (mod hZ 2 −
W 2 − 1i) does satisfy the monic degree 2 polynomial equation z 2 − w2 − 1 = 0, with coefficients in the subring
k[Z] = k[L2 ].

This paves the way for the next

Definition 4.2.2. An inclusion of rings B ⊃ A is called a ring extension. Let B ⊃ A be a ring extension. An
element y ∈ B is said to be integral over A if y satisfies a monic polynomial with coefficients in A. A ring
extension B ⊃ A is said to be an integral extension if every element in B is integral over A.

Note that whenever we have a ring extension B ⊃ A, B is an A-module in a natural fashion.

Example 4.2.3. From the discussion above, the element y ∈ k[C] is not integral over k[X] = k[L1 ], but the
element w = 12 (x + y) in k[C] is integral over the subring k[Z] = k[L2 ].

√ √
Example 4.2.4. The element
√ 5 in Z[ 5]√is integral over the subring Z. In fact, the reader may √ easily prove
2
(by squaring y = m + n 5 and eliminating 5 between
√ y and y ) that every such element y in Z[ 5] is integral
over the subring Z. Thus the ring extension Z[ 5] ⊃ Z is an integral extension.

√
1+ 5
√
Example 4.2.5. The element y = 2 in Q[ 5] is integral over the subring Z, and satisfies the famous
monic polynomial y 2 − y − 1 = 0.

Exercise 4.2.6. Show that if A ⊂ B is an integral extension of rings, and b ⊂ B is an ideal, then the extension
A/(b ∩ A) ⊂ B/b
is also an integral extension.

Example 4.2.7. In the ring extension Q ⊃ Z, no element in Q \ Z is integral over Z. This is because of the
following proposition.
22 VISHWAMBHAR PATI

Proposition 4.2.8. Let A be a UFD, and K be the quotient field of A (which is denoted Q(A)). Then if
y ∈ K is integral over A, y belongs to A.
Proof:
Let y ∈ K be written as pq where p, q are relatively coprime (that is, gcd(p, q) = 1). If y is integral over the
subring A, there is a monic polynomial equation:
 n  n−1
p p
+ a1 + ..an = 0
q q
where ai ∈ A. Multiply the equation by q n−1 , and transpose all terms from the second one on to the right
hand side to get :
pn
= −a1 pn−1 − a2 qpn−2 − .... − an q n−1
q
which shows that q divides pn in A. A being a UFD, any irreducible (prime) factor of q would therefore divide
pn , and hence p, contradicting that gcd(p, q) = 1. Thus q is a unit in A, and has an inverse u ∈ A, and
y = pq = up is in A. 2
√
The above proposition gives a handy way of showing that various √ rings are not UFD’s. For example, Z[ 5]
√ √ √
is not a UFD, because its quotient field is Q[ 5], and the element 1+2 5 in Q[ 5], which is integral over Z[ 5]
√
(in fact even over Z) is not in Z[ 5]. We know this fact already, of course.

Corollary 4.2.9. Denote the quotient field of k[X1 , ...Xn ] by k(X1 , ...Xn ). By the proposition 2.2.17 and the
proposition above, it follows that if a rational function p(X 1 ,...Xn )
q(X1 ,...Xn ) ∈ k(X1 , ..., Xn ) satisfies a monic relation
whose coefficients are polynomials, then this rational function is itself a polynomial.

Definition 4.2.10. We say that a module M over a ring A is faithful if aM = {0} for a ∈ A implies that
a = 0. That is, the annihilator ideal of M defined by:
Ann M := {a ∈ A : aM = {0}}
is the zero ideal in A.

We now state several equivalent conditions for integrality of an element.

Proposition 4.2.11. Let B ⊃ A be a ring extension, and let y ∈ B. Then the following are equivalent:
(i): y is integral over A.

(ii): The smallest subring of B containing A and y, which is denoted A[y] (caution: this is not the polynomial
ring over A in one variable, y may satisfy non-trivial relations) is finitely generated as an A-module.

(iii): There exists a subring R ⊂ B such that A[y] ⊂ R and R is a finitely generated A-module.

(iv): There exists a faithful A[y]-module M which is finitely generated as an A-module.

Proof:
(i)⇒ (ii)
For, let y ∈ B be integral over A, and let
y n + a1 y n−1 + ......an = 0
with ai ∈ A, be the monic polynomial relation over A which y satisfies. Then y n = −(a1 y n−1 + ......an ), and
the higher powers y r for r ≥ n can, by induction, all be written as A-linear combinations of 1, y, y 2 , ...y n−1 .
Since the elements of A[y] are precisely finite A-linear combinations of powers of y, it follows that every element
of A[y] is a finite A-linear combination of 1, y, y 2 , ..y n−1 . Thus A[y] = A + Ay + ...Ay n−1 , and is a finitely
generated A-module.

(ii)⇒(iii) Take R = A[y].

 LECTURES ON BASIC ALGEBRAIC GEOMETRY 23

(iii)⇒(iv) Let M = R. M is then clearly a faithful A[y] module since A[y] is a subring of R, and any element
of A[y] annihilating all of R would have to annihilate 1, and thus be zero.

(iv)⇒(i)
We need a monic polynomial relation for y. Since M is an A[y]-module, and hence the map y. (multiplication
by y) maps M to M and is A-linear. M is given to be finitely generated over A, so let {x1 , x2 , ..xn } be a set
of generators of M as an A-module. Then clearly we have the relations:
y.xi = Σnj=1 aij xj ; 1 ≤ i, j ≤ n, aij ∈ A
Now one has Cramer’s Rule, which holds in any commutative ring :

det T.1 = T adj T

adj
for any matrix T = [Tij ], and T is the transpose of the matrix of cofactors and 1 is the identity matrix.
Apply this to T = y1 − A, with entries from the ring A[y], where A = [aij ]. Then we have the matrix identity:
det(y1 − A)1 = (y1 − A)adj (y1 − A)
However, the matrix y1 − A kills all the generators xi of M , and hence acts as the zero operator on M . Thus
det(y1 − A).1 acts as the zero operator on M . Since M is faithful, it follows that det(y1 − A) = 0. But this
is a monic relation for y with coefficients being polynomials in the entries aij of the matrix A, i.e., elements of
the ring A. Thus (iv) implies (i). 2

Corollary 4.2.12. Let B ⊃ A be a ring extension, and let y1 , y2 be elements of B which are integral over A.
Then the elements y1 + y2 and y1 y2 are integral over A. Taking y2 = a ∈ A, we have ay1 is integral over A for
all a ∈ A.
Proof:
By (ii) of the last proposition, A[y1 ] = Az1 + ....Azn and A[y2 ] = Aw1 + ....Awm . We claim that A[y1 , y2 ] =
Σi,j Awi zj . This is because any power y1r = Σj aj zj for some aj ∈ A, and similarly y2s = Σi bi wi for some
bi ∈ A. Thus the product y1r y2s = Σaj bi wi zj , so any A-linear combination of monomials y1r y2s , i.e. any element
of A[y1 , y2 ], is an A-linear combination of the mn elements wi zj . Thus our claim is established and A[y1 , y2 ] is
finitely generated as an A-module. Now the subrings A[y1 +y2 ] and A[y1 y2 ] of B are contained in R = A[y1 , y2 ],
which is finitely generated, so by (iii) of the last proposition, y1 + y2 and y1 y2 are integral over A. 2

Corollary 4.2.13. If B ⊃ A is a ring extension, and {yi }ni=1 is a collection of elements with yi integral over
A for each i, then the subring A[y1 , ..., yn ] of B is an integral extension of A.
Proof: By applying the argument of the last corollary inductively, A[y1 , ..., yn ] is a finitely generated A-
submodule of B. If y ∈ A[y1 , ..., yn ] , then A[y] ⊂ A[y1 , ..., yn ], and by (iii) of the last proposition, y is integral
over A. 2

Corollary 4.2.14. If B ⊃ C ⊃ A is a tower of ring extensions such that B is an integral extension of C, and
C is an integral extension of A, then B is an integral extension of A.
Proof: Let y ∈ B. Let y n + c1 y n−1 + ...cn = 0 be a monic polynomial relation for y over C, i.e., ci ∈ C.
Then this becomes a monic relation for y over A[c1 , ..cn ]. Thus A[c1 , ..cn , y] is a finitely generated A[c1 , ..cn ]-
module, i.e., there exist elements {zj }m m
j=1 in A[c1 , ..cn , y] such that A[c1 , ..cn , y] = Σj=1 A[c1 , ..cn ]zj . By the last
corollary, since ci are integral over A, A[c1 , ..cn ] is a finitely generated A-module, so that A[c1 , ..cn ] = Σrk=1 Awk .
Thus A[c1 , ..cn , y] = Σj,k Awk zj is a finitely generated A-module, and since A[y] ⊂ A[c1 , ..cn , y], by (iii) of the
last proposition, y is integral over A. 2

Definition 4.2.15. Let B ⊃ A be a ring extension. We define the integral closure of A in B to be the set :
A = {y ∈ B : y is integral over A}
By the corollary 4.2.12 above, it is a subring of B. By definition, A ⊃ A is an integral extension.
24 VISHWAMBHAR PATI

Proposition 4.2.16. Let B ⊃ A be a ring extension. Then A is the largest integral extension of A contained
in B. Further A is integrally closed in B. In other words, the integral closure (in B) A = A.
Proof: We have already remarked that A is an integral extension of A. If y ∈ B is integral over A, then
A[y] ⊃ A is an integral extension. Also A ⊃ A is an integral extension. By corollary 4.2.14 above, A[y] ⊃ A is
an integral extension, so y is integral over A. Thus y ∈ A, proving the proposition. 2

Definition 4.2.17 (Integral closure, normal domains). The integral closure of a domain A, (when no over-
ring B is specified) is understood to be the integral closure of A in the quotient field Q(A) of A. Similarly,
saying that a domain is integrally closed without specifying an overring means that it is integrally closed in its
quotient field.

An integrally closed domain is often called a normal domain, and the integral closure A of a domain A inside
its quotient field is called its normalisation.

So, for example, a UFD is a normal domain by the Proposition 4.2.8.

Definition 4.2.18. An element α ∈ C which satisfies a monic polynomial with integer coefficients is called an
algebraic integer. By the previous Proposition 4.2.16, algebraic integers are elements of the integral closure of
Z in the overring C, and hence the set of algebraic integers is a normal domain inside C.

Exercise 4.2.19 (Quadratic integers and number fields). An algebraic integer α is said to be a quadratic in-
teger if it satisfies a monic degree 2 polynomial in Z[X] which is irreducible in Z[X]. Prove that:
√
(i): If α is a quadratic integer, then it is of the form 21 (m + n d) where m, n ∈ Z, and d ∈ Z is a square-free
integer. (A square-free integer d is one which satisfies vp (d) = 0 or 1 for all primes p).

(ii): For d a square-free integer, the set

√ √
Q[ d] := Q[X]/hX 2 − di = {r + s d : r, s ∈ Q}
is a field, and is called a quadratic number field. If d < 0, we call it an imaginary quadratic number field.

(iii): For d as above, show that every element of

√ √ √
Z[ d] := {m + n d ∈ Q[ d] : m, n ∈ Z}
√
is integral over Z, by explicitly writing down a monic quadratic equation for the element (m + n d).

(iv): Since d is square free, we have d ≡ 1, 2, 3 mod 4. Show that if d ≡ 1 mod 4, then the integral closure
√ √ √
of Z in Q[ d] is the subring { 12 (m + n d) : m, n ∈ Z, m ≡ n mod 2} = Z[ 1+2 d ].
√
In particular, Z[ d] is not a normal domain for d ≡ 1 mod 4, and hence cannot be a UFD. (We
already stated this in the Remark 2.2.6, now we have a proof.)
√ √
(v): If d is square free, and d ≡ 2, 3 √mod 4, show that Z[ d] is integrally √
closed in Q[ d].
The integral closure of Z in Q[ d] is called the ring of integers in Q[ d].

Remark 4.2.20 (Normal domains which are not√UFD’s). The above exercise completely determines the ring
of integers Ad√in the quadratic number field Q[ d]. Note that Ad is precisely the normalisation (=integral
closure) of Z[ d]. It is natural to ask which Ad ’s are UFD’s. From the√
Theorem 7.7 of Chapter 11 in Artin’s
Algebra, one learns that for the imaginary quadratic number fields Q[ d] (d < 0), the only ones which are
UFD’s are for d = −1, −2, −3, −7, −11, −19, −43, −67, −163 (a deep theorem in number theory). Hence for
any d < 0 which is not in the above list, Ad is a normal domain, but not a UFD. For d > 0, the complete
answer to which Ad ’s are UFDs is unknown !
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 25

Now we need to prove a most important lemma called Noether Normalisation. Recall the Definition 2.3.9
of a k-algebra of finite type, where k is a field as usual. We make some definitions first.

Definition 4.2.21. Let A be a k-algebra. We say that a subset S ⊂ A is k-algebraically independent or

simply algebraically independent, if there does not exist any polynomial relation with coefficients in k among
any of the elements of S. That is, if there is a finite subset F = {s1 , ....., sn } ⊂ S, and a polynomial
p(X1 , ..., Xn ) ∈ k[X1 , ..., Xn ], such that p(s1 , ....., sn ) = 0, then p is the 0 polynomial. If S is a singleton
{s}, we say s is transcendental over k. A set which is not k-algebraically independent is called k-algebraically
dependent, and if the singleton set x is k-algebraically dependent, we say it is algebraic over k.

Example 4.2.22. In the ring R, which is a Q-algebra (not of finite type, since it has uncountable dimension
as a Q-vector space), the set {π} is Q-algebraically independent (viz. π is transcendental),
√ though this is a
non-trivial fact. Also,
√ it is known
√ that e is transcendental over Q. On the other hand, 2 is algebraic over Q.
Hence the sets {e, 2} or {π, 2} are k-algebraically dependent.

Remark 4.2.23. If A is a k-algebra, and the subset {yi }ni=1 is algebraically independent over k iff the k-
subalgebra k[y1 , y2 , .., .yn ] of A generated by the yi0 s is isomorphic to the polynomial ring k[X1 , ..., Xn ] under
the k-algebra homomorphism Xi 7→ yi . For, by definition, the kernel of this k-algebra homomorphism is an
isomorphism onto its image iff it has no kernel, which is equivalent to demanding that yi are k-algebraically
independent.

Proposition 4.2.24 (The Noether Normalisation Lemma). Let A be a k-algebra of finite type. Then there
exists a k-algebraically independent subset {y1 , ....., yn } in A such that the extension:
A ⊃ k[y1 , ..., yn ]
is an integral extension. (Note: n could be = 0, in which case k[y1 , ..., yn ] is just k.)
Proof: The proof is by induction on the number of algebra generators for A. Suppose A is singly generated
as a k-algebra, say A = k[y] for some y ∈ A, then we have:
A = k[X]/hf (X)i
where f (X) is the generator of the kernel ideal of the natural surjective k-algebra homomorphism k[X] → A
taking X to y (this kernel ideal is singly generated since k[X] is a PID). Thus f (y) = 0. If f (X) is the
identically zero polynomial, then y is transcendental, and the ring extension A ⊃ k[y] is integral, since it is an
equality, and the required extension. If f (X) 6= 0 in k[X], then by multiplying with the inverse of the leading
coefficient, we may take f to be monic, and thus k[y] ⊃ k becomes an integral extension, with f (y) = 0 being
the monic relation of integrality for y over k. By Corollary 4.2.13, it is an integral extension.

Now assume, by induction, that we have the result for all k-algebras generated by m − 1 elements, and
assume A = k[y1 , .....ym ]. Again, if y1 , ....., ym are k-algebraically independent we take, as above, the ring
extension A ⊃ k[y1 , .....ym ], which is an equality, and hence an integral extension. If y1 , ....., ym is not a k-
algebraically independent set, we must have some polynomial f (X1 , ....., Xm ) ∈ k[X1 , ....., Xm ] which gives a
non-trivial relation, i.e. f (y1 , .....ym ) = 0. We have to create some new elements {zi }m
i=2 , which are m − 1 in
number, such that this equation becomes a monic equation for y1 with coefficients in k[z2 , ..., zm ]. To this end
define :

zi = yi − y1ri ; for i = 2, 3..., m

where the positive integers ri for i = 2, 3, ..m are to be chosen as follows. Let us consider the subset :
Γ = {b = (b1 , ..., bm ) ∈ (Z+ )m : the monomial X1b1 X2b2 ..Xm
bm
occurs inf }
which is a finite set. For b, c ∈ Γ, with b 6= c, the subset V (b, c) = {x ∈ Rm : x.b = x.c} is a hyperplane in
Rm whose intersection, if non-empty, with the plane {1} × Rm−1 is an affine subspace of dimension m − 2. The
finite union of the above hyperplanes, i.e., V = ∪{V (b, c) : b 6= c ∈ Γ} thus intersects {1} × Rm−1 in a finite
union of (m − 2)-dimensional affine subspaces of {1} × Rm−1 , and if we choose a lattice point (exercise: prove
26 VISHWAMBHAR PATI

that such points exist) r = (1, r2 , .., rm ) ∈ {1} × (Z+ )m−1 outside this union, we have r.b 6= r.c for all b 6= c
in Γ. This is the required choice of ri for the equations above.

Thus the positive integers {r.b : b ∈ Γ} are all distinct, and there will be an a ∈ Γ such that r.a > r.c for
all c ∈ Γ such that c 6= a. Now substitute yi = zi + y1ri for i ≥ 2 in the polynomial relation f (y1 , .....ym ) = 0,
to get
f (y1 , z2 + y1r2 , ....., zm + y1rm ) := g(y1 , z2 , ....zm ) = 0
A typical monomial term λb y1b1 y2b2 ...ym
bm
in f changes to
λb y1b1 +r2 b2 +....rm bm + (terms involving zi )
where b = (b1 , ...., bm ) ∈ Γ. By the choice of r and a, the term containing the highest power of y1 in
g(y1 , z2 , ....zm ) is λa y1a1 +r2 a2 +...rm am , and on dividing by λa ∈ k, λa−1 g(y1 , z2 , ...zm ) = 0 is the required monic
polynomial relation showing that y1 is integral over k[z2 , ..., zm ]. Thus:
A = k[y1 , ..., ym ] = k[y1 , z2 , ..., zm ] ⊃ k[z2 , ..., zm ]
is an integral extension.

Now, by the induction hypothesis, this last ring being generated by m − 1 algebra generators, there is
a subring k[w1 , ..., wr ] with {wj } being k-algebraically independent, and with k[z2 , ..., zm ] ⊃ k[w1 , ..., wr ] an
integral extension. By the Proposition 4.2.14 above, A ⊃ k[w1 , ..., wr ] is an integral extension and we are done.
2

Remark 4.2.25. If we assume that the field k in the proposition above is infinite, a simpler proof can be
given. We first make the observation that if F ∈ k[X2 , .., Xm ] is any non-zero polynomial in m − 1 variables,
then there exists a point (a2 , .., am ) ∈ Am−1 (k) such that F (a2 , .., am ) 6= 0. This is obvious for the 1-variable
case, since non-zero polynomials in 1-variable have only finitely many roots. In general, write :
F (X2 , .., Xm ) = Σdj=1 fj (X2 , .., Xm−1 )Xm
j

and, by induction, find a point (a2 , ..., am−1 ) such that the coefficient of the degree term does not vanish at
this point, i.e. fd (a2 , .., am−1 ) 6= 0. Then choose am to be any element in k outside the set of roots of the single
variable polynomial F (a2 , ..., am−1 , Xm ). Now, to return to the proof of the Normalisation lemma, proceed as
before by induction, and get the polynomial f (y1 , ..., ym ) = 0, as before. But now, we make the linear change
of variables :
zi = yi − ai y1 for 2 ≤ i ≤ m
where ai are chosen as follows. Let F (X1 , ...., Xm ) be the highest degree homogeneous term (of degree d, say)
in the polynomial f (X1 , .., Xm ). Choose (a2 , .., am ) so that F (1, a2 , ....am ) = λ 6= 0, which is possible by the
observation above. Verify that on substituting yi = zi + ai y1 for 2 ≤ i ≤ m, the equation f (y1, ....ym ) = 0
transforms to the equation :
j
λy1d + Σd−1j=0 Hj (z2, ...zm )y1 = 0.
Multiplying by λ−1 ∈ k, we have the required fact that y1 is integral over k[z2 , .., zm ]. Then the rest of the
proof is as before.

Proposition 4.2.26. Let B ⊃ A be an integral extension with B an integral domain. Then B is a field iff A
is a field.
Proof: Let A be a field, and let y ∈ B. By hypothesis, there is a relation
y n + a1 y n−1 + ...an = 0
where ai ∈ A. Since B is a domain, we may assume without loss of generality that an 6= 0, and since A is a
field, we may multiply the equation above by −a−1
n , transpose the last term to the right hand side to get :

y(−a−1
n y
n−1
− a1 a−1
n y
n−2
− ..... − an−1 an−1 ) = 1
which shows that the inverse of y is the term in brackets.
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 27

Conversely, suppose B is a field, and a ∈ A \ {0}. Then there exists an inverse a−1 ∈ B, since B is a field.
We claim it is actually in A. For, by the fact that a−1 is integral over A, there is a monic polynomial relation:
(a−1 )n + a1 (a−1 )n−1 + .... + an = 0
Multiply the entire equation by an−1 and transpose all but the first term to the right hand side to get a−1 =
−(a1 + a2 a1 + ....an an−1 ). Since the right hand side is clearly in A, a−1 is in A. Hence A is a field. 2

Remark 4.2.27. We note that the integral extension k ⊂ B where k is a field and B = k[X]/hX 2 i, shows
that the result above is false if one drops the assumption that B is an integral domain.

Here are some additional exercises on prime ideals, integral dependence and integral closure.

Exercise 4.2.28 (I.S. Cohen). Let A be a ring in which all prime ideals are finitely generated. Show that A is
Noetherian. (Hint: Assume that there exist non-finitely generated ideals, so the set Σ of non-finitely generated
ideals is non-empty. Apply Zorn’s Lemma to this set to get a maximal element a ∈ Σ. Claim that this maximal
element must be a prime ideal. Now suppose x 6∈ a. Prove that there exists a finitely generated ideal b ⊂ a
such that b + Ax = a + Ax. For any two ideals a, c in a ring, define:
a/c := {z ∈ A : yc ⊂ a}
Show that a/c is always an ideal, and contains a. Also show that in the above situation a = b + x.(a/Ax). If
y 6∈ a, show that the ideal (a/Ax) strictly contains a, and is therefore finitely generated. Conclude that xy
cannot be in a, and hence a is a prime ideal.

Exercise 4.2.29. Show that the inclusion of rings k[X] ⊂ k[[X]] is not an integral ring extension.

The following few exercises are a rapid primer on field extensions, and easily follow by reformulating the
Propositions 4.2.11, 4.2.12, 4.2.13, 4.2.14, 4.2.16, together with some other purely elementary considerations.
For more details on these matters, the reader may consult Zariski-Samuel’s Commutative Algebra, Vol. 1, Ch.
II.

Exercise 4.2.30 (Algebraic Field Extensions). If k ⊂ K is an inclusion of fields, we call it a field extension.
If S is a subset of K, k(S) denotes the smallest subfield of K containing k and S.
Show that, for an element x ∈ K, the following are equivalent :
(i): There is a unique irreducible monic polynomial f ∈ k[X] such that f (x) = 0 (called the minimal monic
polynomial of x ).

(ii): x is algebraic over k (see definition 4.2.21).

(iii): x integral over k.

(iv): k[x], the k-subalgebra of K generated by k and x, is a finite dimensional k-vector space.

(v): k[x] is contained in a k-subalgebra L of K with dimk L < ∞.

(vi): k[x] is the smallest subfield of K containing k and x, i.e. k[x] = k(x).

Exercise 4.2.31 (More on Algebraic Field Extensions). The field extension k ⊂ K is said to be an algebraic
field extension if every element in K is algebraic over k, viz. if k ⊂ K is an integral extension. If every element
of K \ k is transcendental over k, we say k ⊂ K is a transcendental extension. We say that it is a finite field
extension if the vector space dimension dimk K < ∞. This number is denoted [K : k] and called the degree of
the extension k ⊂ K. Show that :
(i): If k ⊂ K is a field extension with x ∈ K algebraic over k, then dimk k[x] = n, where n is the degree of
the minimal monic polynomial f of x.
28 VISHWAMBHAR PATI

(ii): If k ⊂ K is a finite field extension, then it is an algebraic extension.

(iii): If k ⊂ L and L ⊂ K are finite field extensions, then so is k ⊂ K, and that the degree of this extension
is given by:
[K : k] = [K : L][L : k]

(iv): If xi for i = 1, 2, .., n are elements of K algebraic over k, then k[x1 , .., xn ] is a field, and the field
extension k ⊂ k[x1 , .., xn ] is a finite algebraic extension. If k ⊂ L and L ⊂ K are any two (not necessarily
finite) algebraic extensions, then k ⊂ K is also an algebraic extension.

(v): If k ⊂ K is any field extension, the integral closure of k in K, call it L, is an algebraic extension of
k. Show L is the largest algebraic extension of k contained in K, and that the extension L ⊂ K is a
transcendental extension. Thus every field extension of k can be decomposed into a two step extension,
the first algebraic and the second transcendental.

Exercise 4.2.32. Let A ⊂ B be an integral extension of domains. Show that Q(A) ⊂ Q(B) is an algebraic
extension of their respective quotient fields.

Exercise 4.2.33 (Algebraic Closure of a Field). Let k be a field.

(i): For f ∈ k[X] an irreducible polynomial of degree d > 0, show that the ring K = k[X]/hf i is an algebraic
field extension of k with [K : k] = d, and containing a root of f (namely x, the image of X in K). Show
that if L is any field extension of k containing a root α of f , then the subfield k(α) of L is isomorphic to
K. Thus, the process above is the minimal way of “adjoining a root” of a given irreducible polynomial
to k, and the above construction shows that every irreducible polynomial with coefficients in k certainly
has a root in some algebraic extension of k.

(ii): Order the family :

{L : k ⊂ L is an algebraic extension}
by inclusion, and show that a chain Lα in this chain has the upper bound ∪α Lα . Conclude, using Zorn’s
Lemma that there is a maximal element k in this family, which is therefore an algebraic extension of k.
It is called the algebraic closure of k.

(iii): Show that k is algebraically closed, i.e. if L is an algebraic extension of k, then L = k. Equivalently,
every polynomial f with coefficients in k has a root in k, and hence breaks up into a product of linear
factors.

For example: the algebraic closure of R is C, a fact which is known as the Fundamental Theorem of
Algebra. It can be proved using the Liouville Theorem complex analysis or some elementary algebraic
topology. The algebraic closure Q is called the field of algebraic numbers, and has countable vector space
dimension over Q, and hence a countable set, and thus much smaller than C. Elements of C \ Q are the
transcendental numbers, such as π or e (again, non-trivial facts to prove!).

(iv): Every proper field extension of an algebraically closed field is a transcendental extension.

Exercise 4.2.34 (Transcendence bases and transcendental extensions). Let k be a field, and k ⊂ K be a field
extension in whatever follows. We say that a subset S ⊂ K is a k-transcendence set if the elements are k-
algebraically independent (see definition 4.2.21). We say that a k-transcendence set S ⊂ K is a transcendence
base for K over k (or k-transcendence base) if the extension k(S) ⊂ K is algebraic. (Here k(S) denotes the
smallest subfield of K containing k and S). Prove the following:
(i): The set {X1 , .., Xn } is a k-transcendence base for the extension k ⊂ k(X1 , .., Xn ) (this last field is defined
as the quotient field of the polynomial ring k[X1 , .., Xn ], and is called the field of rational functions in
n-variables over k).
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 29

(ii): If T ⊂ K is a k-transcendence set, then there exists a transcendence base S for K over k containing
T . (Apply Zorn’s lemma to the family of all transcendence sets containing T ).

(iii): If there exists a k-transcendence base S for K with card(S) = n < ∞, then every k-transcendence
base for K has cardinality n, and this number is called the transcendence degree of K over k and denoted
tr degk K. (Hint: Let S = {xi }ni=1 , and T = {yα } be another transcendence base. Assume that T
has cardinality > n. Choose some y1 ∈ T , and noting that it is algebraic over k(S), get an irreducible
polynomial relation f (y1 , x1 , .., xn ) = 0, i.e. the minimal monic polynomial for y1 over k(x1 , .., xn ) = k(S).
Since y1 is transcendental over k, it follows that some xi , say x1 , actually occurs in f . Now write f in
powers of x1 as:
f (y1 , x1 , .., xn ) = f0 (y1 , x2 , .., xn ) + f1 (y1 , x2 , .., xn )x1 + ...fd (y1 , x2 , .., xn )xd1 = 0 (1)
Now no irreducible factor of the polynomial fd (Y1 , X2 , .., Xn ) can vanish at (y1 , x2 , .., xn ), because
f (Y1 , x1 , .., xn ) is the unique irreducible polynomial satisfied by y1 over k(x1 , .., xn ). Thus fd (y1 , x2 , .., xn ) 6=
0, and the relation (1) above shows that x1 is algebraic over k(y1 , x2 , .., xn ). Hence K is algebraic over
k(y1 , x2 , .., xn ). Now proceed with y2 ∈ T , until one has K algebraic over k(y1 , .., yn ), and thus arrive
at the contradiction that any yj with j 6= 1, 2, .., n is k- algebraically dependent on y1 , .., yn . Thus
card(T ) ≤ card(S). Similarly, card(S) ≤ card(T ). The proof is similar to the proof that all bases of a
finite dimensional vector space have the same cardinality.)

(iv): Let k ⊂ L ⊂ K be a tower of field extensions, with K algebraic over L. Show that tr degk L = tr degk K.

(v): If k ⊂ L ⊂ K is a tower of field extensions, all of finite transcendence degree, then show that:
tr degk L = tr degL K + tr degk L
(Contrast with multiplicativity of degrees of finite algebraic extensions.)

Exercise 4.2.35. Let k = R, and let K be the quotient field of the formal power series ring R[[X]] (see the
exercise 2.2.21 for the definition). Show that tr degR K = ∞. (Hint: Construct a set S of exponential functions
of the kind eαX with α ranging in a suitable infinite set such that S is a R-transcendence set.)
30 VISHWAMBHAR PATI

5. Hilbert’s Nullstellensatz
We recall the question raised earlier in the subsection on algebraic sets, i.e., if I is an ideal in k[X1 , ..., Xn ],
k a field, then how does I(V (I)) compare with I ? For one thing, we noted in Exercise 4.1.7 that the ideal
I(V (I)) is a radical ideal, whereas the ideal I need not
√ have been radical. So it is natural to expect I(V (I))
to be just the smallest radical ideal containing I, i.e. I. (Compare with the situation of C([0, 1]), where the
ideal of continuous functions vanishing on the common zero set of an ideal I in C([0, 1]) turns out to be the
closure of that ideal in the sup-norm topology).

The Hilbert Nullstellensatz says that this expectation is realised, when the field k is algebraically closed.
5.1. The Weak Nullstellensatz. We first prove the weak nullstellensatz.
Proposition 5.1.1. Let k be an algebraically closed field, and m be a maximal ideal in k[X1 , ..., Xn ]. Then
there exists an (a1 , ...., an ) ∈ An (k) such that:

m = hX1 − a1 , ...., Xn − an i
Thus maximal ideals in the polynomial ring in n variables over k are in 1-1 correspondence with points in
An (k). In the notation of 4.1.10, Spm(k[X1 , .., Xn ]) is An (k), as a set.
Proof: Let m be a maximal ideal in k[X1 , ..., Xn ], which is a k-algebra of finite type. Then k[X1 , ..., Xn ]/m
is also a k-algebra of finite type, and also a field. Call it K. By Noether’s Normalisation lemma, there exists
a subset S = {yi }ri=1 such that K ⊃ k[y1 , ..., yr ] is an integral extension. By the Proposition 4.2.26 of the
previous section, since K is a field, so is k[y1 , ..., yr ]. Since the latter is a polynomial ring in r variables, it can
only be a field if r = 0. Thus k[y1 , ..., yr ] = k, and K ⊃ k is an integral extension. Thus every element y ∈ K
satisfies a monic polynomial over k, and is hence algebraic over k. But k is algebraically closed, so y ∈ k. Thus
K = k. Let τ be the quotient map :
k[X1 , ..., Xn ] → K = k
and let us denote τ (Xi ) = ai ∈ k. By definition, τ (Xi − ai ) = 0, so that (Xi − ai ) ∈ m for i = 1, 2, .., n. Thus
hX1 − a1 , ....., Xn − an i ⊂ m. On the other hand, the ideal hX1 − a1 , ....., Xn − an i is clearly maximal. This
follows by using the first order “Taylor formula”, i.e. for any polynomial f (X1 , ...., Xn ) we have:
f (X1 , ...., Xn ) = f (a1 , ...., an ) + Σnj=1 hj (X1 , ..., Xn )(Xj − aj )
where hj are some polynomials. This is obtained by writing:

f (X1 , ...., Xn ) = f (a1 + (X1 − a1 ), ...., an + (Xn − an ))

and expanding each monomial in f binomially. So for every polynomial f ∈ k[X1 , .., Xn ] we have:

f (X1 , ...., Xn ) ≡ f (a1 , ...., an ) ∈ k (mod hX1 − a1 , ....., Xn − an i)

so that this ideal hX1 − a1 , ....., Xn − an i is maximal.

Thus m = hX1 − a1 , ....., Xn − an i, and we are done. 2

Proposition 5.1.2 (Weak Nullstellensatz for general k). If K is a k-algebra of finite type which is a field,
then K is a finite algebraic extension of k. Thus, if m is a maximal ideal in a k-algebra A of finite type, then
the quotient A/m is a finite algebraic field extension of k.
Proof: Because K is a k-algebra of finite type, it is an integral extension of some k[y1 , ..., yr ] by Noether
normalisation and, as in the proof above, k[y1 , ..., yr ] must be a field and r must be 0. Thus K is an integral
extension of k, and hence an algebraic extension. If we let {xi }ni=1 denote the algebra generators of K over k,
we have K = k[x1 , ..., xn ], where all the xi are integral, and so by (ii) of Proposition4.2.11, and Proposition
4.2.14, we have K is a finitely generated k-module, i.e., a k-vector space of finite dimension, so a finite algebraic
extension of k. The second part is clear by considering K = A/m. 2

Exercise 5.1.3. Show that the above Proposition 5.1.2 is equivalent to the following statement: Let A be
a k-algebra of finite type. Then every embedding φ : k → L with L algebraically closed extends to a ring
homomorphism ψ : A → L.
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 31

Exercise 5.1.4. This exercise yields another consequence of the weak nullstellensatz for a general field k which
is not necessarily algebraically closed.

(i): Let K be a field, and let K ⊂ L be an algebraic extension. Show that if A is a K-subalgebra of L, then
A is a field.

(ii): Let k be a field (not necessarily algebraically closed) and K its algebraic closure. Show that there is a
1-1 correspondence between k-algebra homomorphisms f : k[X1 , .., Xn ] → K, and points (a1 , .., an ) ∈ Ln
where L ⊂ K is some finite extension of k, depending on f .

(iii): Let k, K be as in (ii) above. If A = k[X1 , .., Xn ]/a is a k-algebra of finite type, then show that there
is a 1 − 1 correspondence between k-algebra homomorphisms f : A → K, and points (a1 , .., an ) ∈ VL (a)
where L is a finite algebraic extension of k (depending on f ). (Note: We define the set:
VL (a) := {(a1 , .., an ) ∈ Ln : h(a1 , .., an ) = 0 for all h ∈ a}
The points of VL (a) are called the L-rational points of V (a).)

Corollary 5.1.5. The rational function field k(X1 , .., Xn ) in n variables, for n ≥ 1, is not a k-algebra of finite
type.
Proof: If it were so, the Proposition 5.1.2 would imply that k(X1 , .., Xn ) is an algebraic extension of k,
contradicting the fact that it is a transcendental extension of k (of transcendence degree n). 2.

Corollary 5.1.6. Let A = k[X1 , ..., Xn ]/a be a k-algebra of finite type, where k is algebraically closed. Then
maximal ideals in A are in 1-1 correspondence with points in V (a).
Proof: Let n be a maximal ideal in A, so that A/n is a k-algebra of finite type, which is also a field. Again, by
the argument in the weak nullstellensatz above, this field has to be k. Let τ : k[X1 , ...., Xn ] → A be the quotient
map. Then m = τ −1 (n) satisfies the relation : k[X1 , ...., Xn ]/m ' A/n = k, so that m is a maximal ideal in
k[X1 , ...., Xn ]. By the Proposition 5.1.1 above, m = hX1 − a1 , ....., Xn − an i, for some (a1 , ..., an ) ∈ An (k). Also
m contains the ideal a .

Thus we have that V (hX1 − a1 , .., Xn − an i) ⊂ V (a). But we saw above that V (hX1 − a1 , .., Xn − an i) =
(a1 , .., an ). So it follows that (a1 , ......an ) ∈ V (a). Conversely, if (a1 , .., an ) ∈ V (a), f (a1 , ..., an ) = 0 for all
f ∈ a. Thus f ∈ hX1 − a1 , .., Xn − an i, so that a ⊂ m. 2

Next, we have a couple of further consequences of the various nullstellensatzes above.

Proposition 5.1.7. Let φ : A → B be a k-algebra homomorphism, with A and B k-algebras of finite type. If
m is a maximal ideal in B, then n = φ−1 (m) is a maximal ideal in A
Proof: We note that since m is maximal, it is prime, and hence so is n = φ−1 (m). Thus A/n is an integral
domain, and the induced homomorphism φ : A/n → B/m is an injective k-algebra homomorphism. In fact
we have k ⊂ A/n, and so by Proposition 5.1.2 above, K = B/m is an integral extension of k. Thus it is also
an integral extension of the integral domain A/n. By Proposition 4.2.26, A/n is a field, and so n = φ−1 (m) is
maximal in A. 2

Remark 5.1.8. We note here that the above result is false if we drop the finiteness assumption on the
algebras. For example, φ : Q[π] ⊂ R is an inclusion of Q-algebras, though R is not a Q-algebra of finite type.
The 0-ideal is maximal in R, but φ−1 (0) = (0) is not maximal in Q[π], since this algebra is not a field (it is
isomorphic to Q[X], the polynomial ring over Q in one variable.
32 VISHWAMBHAR PATI

Geometrically, one interprets the Proposition 5.1.6 to mean that the points in V (a) are precisely the maximal
ideals in its coordinate ring A = k[X1 , .., Xn ]/a. Similarly, one interprets the statement of Proposition 5.1.7
to mean that an algebra homomorphism of two coordinate rings gives a map (in the opposite direction) of the
two algebraic sets: the image of a point is obtained by taking the inverse image of the corresponding maximal
ideal.

So one has the following:

Proposition 5.1.9. Let k be algebraically closed, and X ⊂ An (k) and Y ⊂ Am (k) be algebraic sets, with
coordinate rings k[X] = k[X1 , .., Xn ]/I(X) and k[Y ] = k[X1 , .., Xm ]/I(Y ) respectively. Each k-algebra homo-
morphism φ : k[Y ] → k[X] uniquely defines a set map:

φ∗ : X → Y

Such maps φ∗ : X → Y will be called k-morphisms or simply morphisms from X to Y .

Proof: We have seen from Proposition 5.1.6 that the set X is in bijective correspondence with the maximal
spectrum Spm k[X], and likewise for Y . For x ∈ X, corresponding to the maximal ideal mx , define φ∗ (x) = y
where y ∈ Y is the point corresponding to the ideal ny = φ−1 (mx ) (a maximal ideal by 5.1.7 above). Note that
ny is the ideal of all regular functions on Y that vanish at y, and φ−1 (mx ) = ny merely says that the composite
function f ◦ φ∗ is a regular function on X vanishing at x iff f is a regular function on Y vanishing at y = φ∗ (x).
2

Example 5.1.10 (Morphisms of affine spaces). If X = An (k) and Y = Am (k), then a morphism φ∗ : An (k) →
Am (k) corresponds precisely to a k-algebra homomorphism of polynomial rings:

φ : k[Y1 , .., Ym ] → k[X1 , ..., Xn ]

Such a k-algebra homomorphism is clearly determined uniquely by the images φ(Yi ), which is a polynomial
fi (X1 , .., Xn ) ∈ k[X1 , .., Xn ]. If a := (a1 , .., an ) ∈ An (k) is a point, then the Taylor formula implies that:
n
X
fi (X1 , .., Xn ) − fi (a1 , .., an ) = (Xj − aj )hij (X1 , .., Xn ) ∈ ma i = 1, , , m
j=1

Define the point b := (f1 (a1 , .., an ), .., fm (a1 , .., an )). The above relation reads as:

φ(Yi ) − bi ∈ ma , i = 1, ..., m

Since φ is a k-algebra homomorphism, φ(bi ) = bi , so we have:

φ(Yi − bi ) ∈ ma , i = 1, ..., m

so that (Yi − bi ) ∈ φ−1 (ma ) for i = 1, .., m. This implies that mb = φ−1 (ma ), or equivalently, b = φ∗ (a). Thus
the set map φ∗ : An (k) → Am (k) is just the expected map (a1 , .., an ) 7→ (f1 (a1 , .., an ), ...., fm (a1 , .., an ))

Thus morphisms of affine spaces are just coordinate-wise polynomial maps. Clearly these are far fewer than
all set maps. For example the set map C → C given by a 7→ ea is not a morphism.
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 33

5.2. The Strong Nullstellensatz. This is a bit of a misnomer, since it is actually equivalent to the “weak
nullstellensatz” stated above. First we need the :

Proposition 5.2.1. If A is a k-algebra of finite type, then for any ideal a ⊂ A,

√
a = ∩{m : m is maximal ⊃ a}

p
Proof: By arguing with A/a, we may assume that a = 0. Thus we need to show that {0} (which is the
same as nil(A) by Proposition 4.1.8 in §4) is also the intersection of all the maximal ideals in A (which is also
called the Jacobson radical of A, denoted Jac(A)).

Since every maximal ideal is prime (and Proposition 4.1.8), it follows that nil(A) ⊂ Jac(A). Suppose a ∈ A
such that a ∈ / nil(A). Then there is a prime ideal p such that a ∈ / p. Let x 6= 0 denote the image of a in
B := A/p under the quotient homomorphism π : A → B. Note B is an integral domain since p is prime and
also a k-algebra of finite type. Since it is a domain, it has a nontrivial quotient field Q(B). Consider the
inclusion of k-algebras of finite type: B ⊂ B[ x1 ], where B[ x1 ] denotes the smallest subring of Q(B) containing
B and x1 . Let m be a maximal ideal in B[ x1 ]. Since x is invertible in B[ x1 ], x ∈
/ m. By the Proposition 5.1.7
−1
above, n = B ∩ m is maximal in B, and does not contain x. Now π (n) is a maximal ideal in A which does
not contain a. Thus a ∈ / Jac(A). (This trick of producing ideals in B not containing an element x by taking
inverse images of ideals in B[ x1 ] is called the Rabinowitch Trick). 2

Remark 5.2.2. In a general ring, the nilradical is strictly contained in the Jacobson radical. For example, in
the power series ring k[[X]], the nilradical is {0}, but the Jacobson radical is hXi. In particular k[[X]] is not
a k-algebra of finite type.

Proposition 5.2.3 (Hilbert’s Nullstellensatz). Let k be an algebraically closed field. Then the following are
equivalent:

(i): The weak nullstellensatz

√
(ii): If a is an ideal in k[X1 , .., Xn ], then I(V (a)) = a.

(iii): If f1 , ..., fm are polynomials in k[X1 , ..., Xn ] such that V (f1 , .., fm ) = φ, then there exist elements
{gi }m m
i=1 in k[X1 , ..., Xn ] such that Σi=1 gi fi = 1.

Proof: (i)=⇒(ii)
In view of the Proposition 5.2.1 above, it suffices to prove that
I(V (a)) = ∩{m : m is maximal ⊃ a}
for an ideal a in k[X1 , ..., Xn ] when k is algebraically closed. By (i) m is maximal in k[X1 , .., Xn ] iff m =
hX1 − a1 , ..., Xn − an i, the ideal of all polynomial functions which vanish at some point (a1 , .., an ) ∈ An (k),
and such a maximal ideal contains a iff every f ∈ a vanishes at (a1 , .., an ) (recall the Taylor formula in the
proof of the weak nullstellensatz above), i.e, if (a1 , ..., an ) ∈ V (a). Thus the intersection
∩{m : m is maximal ⊃ a}
is precisely
√ the ideal of all polynomials which vanish at each point of V (a), which is by definition I(V (a)).
Thus a = I(V (A))

(ii)⇒(iii)
Let {f1 , .., fm } bep
elements of k[X1 , .., Xn ] such that V (f1 , ..., fm ) = φ. Then clearly we have I(V (f1 , ...., fm )) =
Pm
k[X1 , ..., Xn ]. Thus hf1 , ..., fm i = A, and so 1 ∈ hf1 , ..., fm i, i.e., 1 = i=1 gi fi .

(iii)⇒(i)
Let m be a maximal ideal in k[X1 , .., Xn ]. Then let f1 , ..., fm be ideal generators of m, by Proposition 2.3.7.
If V (f1 , ..., fm ) = φ, we would have 1 ∈ m by (iii), a contradiction. So there exists a point (a1 , ..., an ) ∈
34 VISHWAMBHAR PATI

V (f1 , .., fm ). Thus fi (a1 , .., an ) = 0 for all i = 1, ..., m, implying that fi ∈ hX1 − a1 , ..., Xn − an i. Hence
m ⊂ hX1 − a1 , ..., Xn − an i. However, since m is maximal, we must have m = hX1 − a1 , ..., Xn − an i. This
proves the proposition. 2

5.3. Localisation. We now need the notion of localisation of a ring, which is a generalisation of the construc-
tion of the quotient field of an integral domain. The idea is to invert certain elements which are non-zero.
Definition 5.3.1. Let A be a commutative ring with 1, and let S be a multiplicative set in A. i.e. :

(i): 0 ∈
/ S, 1 ∈ S

(ii): a, b ∈ S ⇒ ab ∈ S.

Define the localisation S −1 A of A at S as follows: On the set A × S, define the equivalence relation:
(a, s) ∼ (a1 , s1 ) if there exists a t ∈ S such that t(as1 − a1 s) = 0 (verify that this is an equivalence relation).
Denote the equivalence class of (a, s) as as . Make the set of these equivalence classes a ring by declaring
a1 as1 +a1 s
a
s + s1 = ss1 , and as . as11 = aa
ss1 . Again, verify that these operations are well defined, and that there is a
1

natural homomorphism:
A → S −1 A
a
a 7→
1

Example 5.3.2. If f ∈ A is not a nilpotent element, then the set {f k }∞ k=0 is a multiplicative set. In this case
S −1 A is denoted by Af . If A is a domain, this is the subring A[ f1 ] generated by A and f1 inside the quotient
field Q(A) of A.

Example 5.3.3. If p is a prime ideal in any ring A, then the set S = A \ p is a multiplicative set, and we
define the ring Ap := S −1 A, which, by abuse of language is called the localisation of A at p , though it should
be called the localisation of A at A \ p.

Remark 5.3.4. The natural homomorphism A → S −1 A may not be injective in general. For example, if we
take A = Z6 and S = {1, 3} (verify that this is a multiplicative set), then under the natural homomorphism
Z6 → S −1 Z6 , the element 2 goes to 0, since 3.2 = 0, showing that 21 = 0 in S−1 Z6 . (Exercise: compute S −1 Z6 ).
Of course, the natural map A → S −1 A is injective if and only if no element of S is a zero-divisor. In particular,
if A is an integral domain.

Example 5.3.5. If A is a domain and S = A \ {0}, then S −1 A is just Q(A), the quotient field of A. More
generally, for an arbitrary ring A, if S = {x : x is not a zero-divisor}, then S −1 A is the total quotient ring (or
total ring of fractions) of A. In this ring, every possible element which can be inverted, is inverted. By the
above, the natural map of a general ring A to its total quotient ring is an inclusion.

Exercise 5.3.6 (Rabinowitch trick again). Let S be a multiplicative set in a ring A, and let f : A → S −1 A
denote the natural homomorphism. Show that if q is a proper ideal (resp. prime ideal) in S −1 A, then f −1 a
is a proper ideal (resp. prime ideal) in A which is disjoint from S. In particular, for every maximal ideal
m ⊂ S −1 A, the ideal f −1 (m) is a prime ideal in A avoiding S. (The particular case of S = {1, x, x2 , .., xn , ..}
is the Rabinowitch trick used earlier in the proof of Proposition 5.2.1).
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 35

Exercise 5.3.7. Let A be any ring, and let D be the set of zero-divisors (and note that 0 is included in D).
Show that for every x ∈ D, there is a prime ideal p ⊂ D which contains x. Conclude that the set D is a union
of prime ideals. Conclude that if D 6= {0} in a ring A, then there exist prime ideals in A consisting entirely of
zero-divisors.
Exercise 5.3.8 (Rings of continuous functions). Consider the ring A = C([0, 1]) of continuous complex-valued
functions on the closed interval [0, 1]. Note that A is a complete normed linear space (Banach space) over C
with the sup norm
kf k := sup | f (x) |
x∈[0,1]

Prove that:

(i): For each a ∈ [0, 1], the ideal:

ma := {f ∈ C([0, 1]) : f (a) = 0}
is a maximal ideal.

(ii): Show that if a ⊂ A is an ideal, then the set:

V (a) = {x ∈ [0, 1] : f (x) = 0 for all f ∈ a}
is a closed set. Show that a ⊂ b iff V (a) ⊃ V (b).

(iii): Show that if a is a maximal ideal in A, then there exists a point a ∈ [0, 1] with a = ma .

(iv): Show that for each closed subset C ⊂ [0, 1], the ideal:
I(C) := {f ∈ A : f|C ≡ 0}
is an ideal in A. Check that it is a norm-closed ideal. That is, if kfn − f k∞ → 0 for a sequence fn ∈ I(C),
then f ∈ I(C).

(v): For any subset S ⊂ [0, 1] show that V (I(S)) = S. Show that for any ideal a ⊂ A, I(V (a) = a, where
the closure on the right is the closure in the norm k−k∞ .

(vi): Show that f ∈ A is a zero-divisor iff V (f ) has a non-empty interior.

(vii): Show by using the last two exercises that there exist prime ideals in A which are not maximal.

Exercise 5.3.9. Show that in the ring Ap , there is a unique maximal ideal, namely the ideal :
nx o
pAp = : x ∈ p and s ∈ A \ p
s

Definition 5.3.10. A ring with a unique maximal ideal is called a local ring. For example, the formal power
series ring k[[X1 , ..., Xn ]] is a local ring, for any formal power series is invertible iff it has non-zero constant
term, i.e., iff it lies outside the (clearly maximal) ideal hX1 , ..., Xn i, which is therefore the unique maximal
ideal. For a local ring A, the localisation Am at the unique maximal ideal m is isomorphic to the original ring
via the natural homomorphism A → Am defined in 5.3.1 above.
We have one last fact, which is a universal property of localisations.
Proposition 5.3.11 (Universal property of localisations). Let A be a ring, and S ⊂ A be a multiplicative set.
Let θ : A → S −1 A denote the natural map introduced in Definition 5.3.1.
36 VISHWAMBHAR PATI

(i): If f : A → B is a ring homomorphism such that f (s) is invertible (i.e. a unit) for each s ∈ S, then there
is a unique homomorphism fe : S −1 A → B such that the diagram:
A
f&
θ↓ B
fe%
S −1 A
commutes.

(ii): If f : A → B is a homomorphism of rings, and p ⊂ A, q ⊂ B are prime ideals such that f −1 q ⊂ p, then
there is a unique homomorphism fe : Ap → Bq such that the diagram:
f
A −→ B
θ1 ↓ ↓ θ2
fe
Ap −→ Bq
commutes.

Proof: For (i), define fe( as ) = f (a)f (s)−1 . It is trivial to check that this definition is independent of represen-
tative chosen for as . With this definition of fe, we have (fe ◦ θ)(a) = fe( a1 ) = f (a), so the diagram commutes as
claimed.

For (ii) note that the hypothesis implies that f (A \ p) ⊂ B \ q, and θ2 maps this last set into invertible
elements of Bq . Apply (i) to the map θ2 ◦ f . 2

Exercise 5.3.12. Here are a few simple exercises about localisation:

(i): Compute S −1 A for A = k[X, Y ]/hXY i and S = {1, x, x2 , ..., }, where x is the image of X in A.

(ii): Show that the natural map θ : A → S −1 A is an isomorphism iff every element of S is a unit.

(iii): Let f be a non-nilpotent element in a ring A, and S = {1, f, f 2 , ..}. Using the universal property (i)
of Proposition 5.3.11, prove that the localisation S −1 A is isomorphic to the ring A[X]/hXf − 1i.

(iv): Compute the coordinate ring k[X] of the hyperbola X = V (X1 X2 − 1) ⊂ A2 (k).

One can extend the notion of localisation to A-modules as well. More precisely:

Definition 5.3.13. Let S be a multiplicative set in a ring A, and let M be an A-module. One can also localise
this module at S, by defining an equivalence relation on M × S just as before, viz. (m, s) ∼ (m1 , s1 ) iff there
exists a t ∈ S such that t(s1 m − sm1 ) = 0 in M . Again, the equivalence classes are denoted by m s where
m ∈ M and s ∈ S, and the set of all these equivalence classes is denoted by S −1 M , and called the localisation
of M at S. It is easily checked to be an abelian group, and an S −1 A module by defining the addition and
scalar multiplication in the obvious manner. Thus it makes sense to define S −1 a for an ideal a in A, and for
example, the ideal p Ap defined above is nothing but S −1 p with S = A \ p. If M is an A-module, and p a prime
ideal in A, the localisation of M at the multiplicative set S = A \ p is denoted Mp .

Exercise 5.3.14. If
0→L→M →N →0
is an exact sequence of A-modules, show that the sequence :
0 → S −1 L → S −1 M → S −1 N → 0
is an exact sequence of S −1 A-modules.
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 37

Proposition 5.3.15. If B ⊃ A is an integral ring extension, then for S a multiplicative set in A (which also
makes it a multiplicative subset of B), S −1 B ⊃ S −1 A is an integral ring extension.
Proof: Let sb be an element of S −1 B and let bn + a1 bn−1 + ...an = 0 be a monic polynomial relation for b over
A. Divide this equation by sn and get :
 n  n−1
b a1 b an
+ + ... n = 0.
s s s s
b
as the monic relation for s over S −1 A. 2

We next have a crucial proposition about “lifting” prime ideals in integral extensions.
Proposition 5.3.16 (The Going-up Theorem of Cohen-Seidenberg). Let A ⊂ B be an integral ring extension.
Then, for each prime ideal p ⊂ A, there is a prime ideal p1 ⊂ B such that p1 ∩ A = p. (Notation: We say that
the ideal p1 lies over p. That is, the natural map Spec B → Spec A defined by q 7→ q ∩ A is surjective, with
the prime ideal p1 ⊂ B lying in the fibre of the prime ideal p ⊂ A).
Proof: Let us first do an easy case. Suppose A is a local ring, i.e. A has a unique maximal ideal, and also
assume that p is this maximal ideal. Let p1 be any maximal ideal in ideal in B. We claim p1 ∩ A = p. This is
because B/p1 ⊃ A/A ∩ p1 is also an integral extension. Since B/p1 is a field, so is A/A ∩ p1 , by Proposition
4.2.26, and hence p1 ∩ A is maximal. Since p is the unique maximal ideal of A, this forces p1 ∩ A = p . Hence
we are done in this easy case.

To do the general case we have to consider the localisation Ap of A at p. Again, by an abuse of notation,
we denote the Ap -module S −1 B by Bp , where S = A \ p (p is not a prime ideal in B !). Since Ap is a local
ring, with the maximal ideal being S −1 p = pAp , by the easy case above, there exists an ideal n in Bp such that
n ∩ Ap = S −1 p.

Let us consider the commutative diagram:

i : A → Ap
↓ ↓
j : B → Bp
where the vertical maps are inclusions. Let p1 = j −1 (n). We claim that p1 ∩ A = p.

Since n ⊃ S −1 p we have :
p1 = j −1 (n) ⊃ j −1 (S −1 p) = j −1 (pAp ) ⊃ p
so that p1 ∩ A ⊃ p.
On the other hand, let b ∈ p1 ∩ A. Then, 1b ∈ n ∩ Ap = S −1 p. Thus 1b = ps for some p ∈ p and s ∈ A \ p .
Thus t(bs − p) = 0, for some t ∈ S = A \ p. Since 0 ∈ p and p is prime, bs − p ∈ p. Since p ∈ p, this means
bs ∈ p, and since s ∈
/ p, it follows that b ∈ p. Thus p1 ∩ A = p, and we are done. 2

Exercise 5.3.17. Do the Exercise 4.2.29 using the Going-up Theorem 5.3.16 above.

Lemma 5.3.18. Let A ⊂ B be an integral extension of rings. Let q1 ⊂ q2 be prime ideals of B such that
p1 := q1 ∩ A = p2 := q2 ∩ A. Then q1 = q2 .

Proof: We note that with the given hypotheses, the extension

A1 := A/p1 ⊂ B1 := B/q1
is an integral extension of domains, and the prime ideal q := q2 /q1 in B1 satisfies the relation q ∩ A1 = {0}.
We need to show that q = {0}. So it is enough to prove that:

A ⊂ B is an integral extension of domains, and q is a prime ideal of B with p = q ∩ A = {0}, then q = {0}
38 VISHWAMBHAR PATI

Since A is a domain, the set S = A \ {0} is a multiplicative set, and S −1 A is precisely the quotient field of
A. Also, by the Proposition 5.3.15, the extension of domains
Q(A) = S −1 A ⊂ S −1 B
is an integral extension of domains. But since Q(A) is a field, it follows by the Proposition 4.2.26 that S −1 B
is a field. Since S −1 q is an ideal in S −1 B, it must be either all of S −1 B or {0}. It cannot be all of S −1 B,
because then we would have that 1 = a/s for some a ∈ q and s 6= 0 in A, and since B is a domain, this would
imply that a = s 6= 0, i.e. 0 6= s ∈ q ∩ A = {0}, a contradiction. So S −1 q = {0}. Again, since B is an integral
domain, this means q = {0} and the lemma is proved. 2

Remark 5.3.19. In the above Lemma 5.3.18, one cannot drop the restriction that all the ideals be prime. For
example, with A = k and B = k[x]/hx2 i, we see that for q1 = hxi and q2 = {0}, we have q1 ∩ A = q2 ∩ A = {0}.
However, q1 6= q2 , the reason being that q2 = {0} is not a prime ideal in B, since B is not a domain.

Corollary 5.3.20 (Going-up Theorem 2). Let A ⊂ B be an integral extension of rings, and let
p1 ⊂ p2 ⊂ .... ⊂ pn
be an ascending chain of prime ideals in A . Then there exists an ascending chain of prime ideals in B
q1 ⊂ q2 ⊂ ... ⊂ qn
with qi ∩ A = pi for all i. Furthermore, qi = qi+1 iff pi = pi+1 .

Proof: Inductive application of the Theorem 5.3.16 to the integral extensions A/pi ⊂ B/qi yields the ideal
qi+1 . The last assertion follows immediately from Lemma 5.3.18. 2

Here are some additional exercises on this section.

Exercise 5.3.21. Let A be a domain and p a prime ideal in A. Show that A ⊂ Ap is an integral extension
iff A = Ap . (Hint: First show that p must be a maximal ideal in A, and then show that it must be the only
maximal ideal in A, i.e. A must be a local ring).

Exercise 5.3.22. Prove that the polynomial ring in one variable k[T ] is not isomorphic to k[X, Y ]/hY 2 − X 3 i
as a k-algebra. (Hint: One is integrally closed in its quotient field, and the other is not). Show that the
morphism of the affine line to the cubic cusp defined by:
A1 (k) → V (Y 2 − X 3 ) ⊂ A2 (k)
t 7→ (t2 , t3 )
is a bijective morphism, but not an isomorphism of algebraic sets. (That is, there is no morphism which is an
inverse of this morphism). Thus a morphism of algebraic sets which is a bijection as a set map need not be
an isomorphism of algebraic sets! This is in sharp contrast to most algebraic objects like groups, rings, fields,
vector spaces etc.

Exercise 5.3.23. If k is not algebraically closed, prove that every closed subset of An (k) can be expressed as
V (G) for some polynomial G ∈ k[X1 , .., Xn ]. That is, every algebraic subset of An (k) can be defined by a single
equation. (Hint: For n = 1, this fact is true for arbitrary k. So assume n ≥ 2. First construct a polynomial
Fr ∈ k[T1 , .., Tr ] (where r ≥ 2)) whose only zero is the origin. Since k is not algebraically closed, find a
m
monic polynomial + a1 T m−1 + ...am ∈ k[T ] which has no roots in k. Prove that the only zero of
P f (T ) = Ti m−i
F2 (X1 , X2 ) := 0≤i≤m ai T1 T2 in A2 (k) is (0, 0). Inductively define Fr (T1 , .., Tr ) = F2 (Fr−1 (T1 , .., Tr−1 ), Tr )
and show that its only zero in A (k) is (0, .., 0). Now let X = V (f1 , .., fr ) be any closed subset of An (k). Set
r

G(X1 , .., Xn ) = Fr (f1 (X1 , .., Xn ), ..., fr (X1 , .., Xn ))

2
where Fr is the polynomial constructed above. For example, in the case of k = R, we could set f (T Pr) = 12+ T ,
2 2 2 2
leading to F2 (T1 , T2 ) = T1 + T2 , and more generally Fr (T1 , .., Tr ) = T1 + .. + Tr , and so G = i=1 fi does
the job).
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 39

Exercise 5.3.24. Let G be a finite group of automorphisms of a ring B, and let

A = B G := {b ∈ B : σ(b) = b ∀ σ ∈ G}
Show that:

(i): The ring extension A ⊂ B is Q

an integral extension (Hint: For an element x ∈ B, note that x satisfies
the monic polynomial f (X) = σ∈G (X − σ(x)).)

(ii): Let p ⊂ A be a prime ideal, and let:

Sp = {q ∈ Spec B : q ∩ A = p}
be the set of all prime ideals in A “lying over”p. Show that G acts transitively
Q on Sp (and hence Sp has
finite cardinality. Hint: Let q1 , q2 ∈ Sp . For each x ∈ q1 , note that σ (σ(x)) ∈ q1 ∩ B G = p ⊂ q2 . By
the primeness of q2 , it follows that σ(x) ∈ q2 for some σ ∈ G. Thus q1 ⊂ ∪σ∈G σ(q2 . Now use Exercise
4.1.4 and lemma 5.3.18 to conclude that q1 = σq2 .)

Exercise 5.3.25. Determine which of the following k-algebras (all are domains) are integrally closed:

(i): k[X, Y ]/hXY − 1i.

(ii): k[X, Y ]/hY 2 − X 3 i.

(iii): k[X, Y ]hY 2 − X 2 (X + 1)i.

Exercise 5.3.26. Let A ⊂ B be an integral extension of rings. Show that if q ⊂ B is a prime ideal of B lying
over a maximal ideal m ⊂ A, then q is a maximal ideal of B. Thus the natural map Spec B → Spec A, which
is a surjection, continues to be a surjection Spm B → Spm A.

Exercise 5.3.27. Give an example of an integral ring extension A ⊂ B such that B is not a finitely generated
A-module.
40 VISHWAMBHAR PATI

6. Topology of Affine Algebraic Sets

In this section, we will translate algebra into geometry. We will henceforth always assume that the field k
is algebraically closed, unless explicitly stated otherwise.
6.1. An equivalence of categories. The following proposition is obvious from the Proposition 5.2.3.
Proposition 6.1.1. Let k be algebraically closed, as per our assumption above. The correspondences a 7→
V (a) and Z 7→ I(Z) set up a one-one correspondence between radical ideals of k[X1 , .., Xn ] and closed algebraic
subsets of An (k).
Proof: Follows immediately from Propositions 5.2.3 and 3.1.6 that for a radical ideal a, we have I(V (a)) = a
and for a closed algebraic subset Z of An (k), we have V (I(Z)) = Z, so the two correspondences above are
inverses of each other. 2
Definition 6.1.2. We call a ring A reduced if nil A = {0}, i.e., the only nilpotent element in A is 0.
√
If Z = V (a) is a closed set, by replacing a by its radical a, we may as well assume that a is a radical
ideal, and equal to I(Z). Thus in this case, there are no non-zero nilpotents in the coordinate ring k[Z] =
k[X1 , X2 , ..., Xn ]/a of Z, and the coordinate ring of an affine closed set is a reduced k-algebra of finite type.
It is obvious that a reduced k-algebra of finite type is the quotient of some polynomial ring k[X1 , .., Xn ] by
a radical ideal. (This is, of course, not a unique representation, and will change if we change the algebra
generators of the k-algebra.) Thus every reduced k-algebra of finite type is the coordinate ring of some affine
closed set. The elements of the coordinate ring k[Z] are called regular functions on Z. If Z ∈ An (k), then these
are precisely the functions on Z which are restrictions of polynomials on An (k).

Going back to rings of continuous complex valued functions, it is known that every abelian, unital C ∗ -
algebra (i.e. a Banach algebra with conjugation which satisfies a norm relation) arises as C(X) for some
compact hausdorff space X. In the Gelfand-Naimark theory for a compact hausdorff topological space X, one
can reconstruct X and its topology from the ring of continuous complex valued functions C(X). In algebraic
geometry, the role of continuous functions is taken by regular functions,i.e. elements of the coordinate ring.
Thus, one would analogously like to reconstruct Z with its Zariski topology from the reduced k-algebra of finite
type given by the coordinate ring k[Z]. To this end, we have the following proposition:

Proposition 6.1.3. If Z = V (a) is an algebraic subset of An (k), where a is a radical ideal in k[X1 , .., Xn ],
then Z is in 1-1 correspondence with the maximal spectrum
Spm k[Z] = {m : m is a maximal ideal in k[Z]}
of the coordinate ring k[Z] = k[X1 , .., Xn ]/a. Further, the family of Zariski-closed subsets of Z corresponds to
the family of subsets:
{Ve (b) : b an ideal in k[Z]}
of Spm k[Z], where we define:
Ve (b) := {m : m is a maximal ideal in k[Z] containing b}
Conversely, given a reduced k-algebra A of finite type, Spm A is a topological space, and homeomorphic to
some algebraic set Z in some An (k) with its Zariski topology, and A is recovered as its coordinate ring k[Z].
Proof: The first part of the proposition is clear from the Proposition 5.1.6 (weak nullstellensatz), and we remark
(not necessarily radical) ideal a in k[X1 , ..,√Xn ], a maximal (in fact even prime)
here in passing that for a general √
ideal m contains a iff it contains a, and hence the sets Spm k[X1 , .., Xn ]/ a, and Spm k[X1 , .., Xn ]/a and the
set:
{m ⊂ k[X1 , .., Xn ] : m is maximal ⊃ a}
are all in bijective correspondence. To check the statements about the topologies, let
p : k[X1 , .., Xn ] → k[X1 , .., Xn ]/a
be the quotient homomorphism. Note that, b is an ideal in k[Z] = k[X1 , .., Xn ]/a iff the inverse image
b1 = p−1 (b) is an ideal in k[X1 , .., Xn ] containing a, and m is a maximal ideal in k[Z] containing b iff its inverse
image p−1 (m) is a maximal ideal in k[X1 , .., Xn ] containing b1 = p−1 (b). Thus the mapping m 7→ p−1 (m) giving
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 41

the bijection between Spm(k[X1 , .., Xn ]/a) → V (a) takes the set Ve (b) to the closed subset V (p−1 (b)) ⊂ V (a).
For the converse statement, just write a representation of A as A = k[X1 , .., Xn ]/a, where a is a radical ideal
in k[X1 , .., Xn ]. By definition, the coordinate ring of Z = V (a) is the reduced k-algebra A. This proves the
proposition. 2

Let A, B be reduced k-algebras of finite type, and ψ : A → B be a k-algebra homomorphism. We define a

map:
ψ ∗ : Spm B → Spm A
m 7→ ψ −1 (m)
By the Proposition 5.1.7, ψ −1 (m) is a maximal ideal in A. It is easily verified that (ψ ∗ )−1 (Ve (b)) = Ve (hψ(bi),
where hψ(b)i is the ideal generated by ψ(b) in B, so that ψ ∗ is continuous with respect to the Zariski topologies
on Spm A and Spm B.

Let Z1 , Z2 be algebraic sets. We recall from the Proposition 5.1.9 that a map φ : Z1 → Z2 is a morphism if
φ = ψ ∗ for some k-algebra homomorphism ψ : k[Z2 ] → k[Z1 ].

Remark 6.1.4 (Morphisms versus Zariski continuous maps). Though morphisms between algebraic sets are
continuous with respect to their Zariski topologies by definition (and the Proposition 6.1.3 above), not all
Zariski continuous maps are morphisms. For example, if k = C, any bijection from C to itself is continuous
with respect to the Zariski topologies (=cofinite topologies), whereas the only morphisms from C to itself are
polynomial maps, by Example 5.1.10.

We need one last lemma before we state the main proposition of this subsection.

Lemma 6.1.5. We have the following:

(i): Let k be an arbitrary field (not necessarily algebraically closed). Let A be a reduced k-algebra of finite
type. For f ∈ A, f = 0 iff f = 0 (mod m) for all m ∈ Spm A.

(ii): Let k be an algebraically closed field. If φ, ψ : A → B are both k-algebra homomorphisms of reduced
k-algebras of finite type, then φ = ψ iff φ∗ = ψ ∗ : Spm B → Spm A.

Proof: For (i), the only if part is obvious. So assume f = 0 (mod m) for all m ∈ Spm A. This means
f ∈ ∩ {m : m ∈ Spm A}. But this last intersection is exactly nil A, by Proposition 5.2.1 (which is true for an
arbitrary field). Since A is reduced, nil A = 0, and hence f = 0.

The only if part of (ii) is obvious again. So assume that φ∗ = ψ ∗ : Spm B → Spm A. This means that
ψ m = φ−1 m for all m ∈ Spm B. By the fact that φ−1 m ∈ Spm A, and φ is a k-algebra homomorphism, the
−1

map:
φ : A/φ−1 m → B/m
is precisely the identity map k → k. Thus for each x ∈ A, we have
x (mod φ−1 m) = x = φ(x) = φ(x) (mod m) for all m ∈ Spm B
Similarly,
x (mod ψ −1 m) = x = ψ(x) = ψ(x) (mod m) for all m ∈ Spm B
Since ψ m = φ m for all m ∈ Spm B, it follows that x (mod φ m) = x (mod ψ −1 m) for all m ∈ Spm B.
−1 −1 −1

By the two equations above, this means that φ(x) = ψ(x) (mod m) for all m ∈ Spm B and all x ∈ A. By the
part (i) above, this implies φ(x) = ψ(x) for all x ∈ A, i.e. φ ≡ ψ. The lemma follows. 2
42 VISHWAMBHAR PATI

Remark 6.1.6. (i) of the above Lemma is false if A is not assumed to be reduced. For example, if A =
k[X]/hX 2 i, and we denote the image of X in A by x, then Spm A is {hxi}, and clearly x (mod hxi) = 0 even
though x 6= 0.

Similarly (ii) fails if we don’t assume A and B reduced (Exercise, using the example above). It also fails if we
don’t assume k to be algebraically closed. For example, setting A = B = Q[i], a Q-algebra of finite type which
happens to be a field, we have Spm A = Spm B = {0}. Clearly the identity map and the complex-conjugation
map are distinct Q-algebra homomorphisms, but both induce the identity map on Spm A.

Proposition 6.1.7. Let Algred (k) be the category of reduced k-algebras of finite type with morphisms being
k-algebra homomorphisms, and let Zar be the category of all closed algebraic sets, with morphisms as defined
in Proposition 5.1.9. Consider the functor:

F : Zar → Algred (k)

Z 7→ k[Z]
∗ ∗
For a morphism φ : Z → Y , we define F (φ ) = φ : k[Y ] → k[Z], which makes sense by (ii) of Lemma 6.1.5.

Then:

(i): For each reduced k-algebra of finite type A ∈ Algred (k), there exists a Z ∈ Zar such that A = k[Z].
(Note: This Z is not unique, but is determined upto isomorphism in Zar as will follow from (iii) below.)

(ii): If two reduced k-algebras of finite type A and B are thus realised as k[Y ] and k[X] respectively, then
each k-algebra homomorphism φ : A → B is realised as F (f ) for some uniquely determined morphism
f : X → Y in Zar.

(iii): Algebraic sets X and Y in Zar are isomorphic in Zar iff F (X) = k[X] and k[Y ] are isomorphic in
Algred (k).

(iv): For each Z ∈ Zar, Z is homeomorphic to Spm F (Z), with the topology defined as in Proposition 6.1.3.
Thus the categories Zar and Algred (k) may be viewed as “equivalent categories”.
Proof: For (i), note that A ∈ Algred (k) implies A = k[X1 , .., Xn ]/a for some radical ideal a, which implies
that A = k[Z] for Z := V (a) ⊂ An (k). Since this expression of A as a quotient of a polynomial algebra is not
unique, Z is not unique.

(ii) follows immediately from the definition of a morphism in Proposition 5.1.9, (by taking f := φ∗ ) : X → Y ,
and of course (ii) of the preceding Lemma 6.1.5 which says that this φ∗ : X → Y uniquely pins down φ.

The “if” part of (iii) is trivial noting that φ∗ ψ ∗ = (ψφ)∗ and id∗ = id. For the only if part note that if
φ : X → Y and ψ ∗ : Y → X are morphisms which are inverses of each other, then (ψφ)∗ and (φψ)∗ are
∗

identity maps of Y and X respectively. By (ii) of Lemma 6.1.5, this implies that ψφ and φψ are both identity
maps of k[Y ] and k[X] respectively. Thus k[Y ] and k[X] are isomorphic.

(iv) follows directly from the Proposition 6.1.3. The proposition follows. 2

Exercise 6.1.8. Let ψ ∗ : X → Y be a morphism of affine algebraic sets. Show that ψ : k[Y ] → k[X] is
injective, iff ψ ∗ has Zariski dense image in Y . Similarly show that if ψ is a surjective map, then ψ ∗ is injective.
What about the converse ?

The Exercise 5.3.22 shows that bijective morphisms of affine algebraic sets are not necessarily isomorphisms,
i.e. do not necessarily induce isomorphisms of coordinate rings.
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 43

6.2. Noether Normalisation Lemma Revisited. We can now reformulate the Noether Normalisation
Lemma in geometric terms.

Proposition 6.2.1. Let Z = V (a) be an affine algebraic subset of An (k), where k is an algebraically closed
field, a a radical ideal in k[X1 , ..., Xn ]. Then there exists a morphism π : Z → Ar (k) which is surjective and
has finite fibres. If k = C, then π is also a proper map, with respect to the classical topologies on Z and Cr .

Proof: Let xi denote the coordinate functions (= image of Xi ) in the coordinate ring k[Z] = k[X1 , ..., Xn ]/a.
By the Noether Normalisation lemma, there is a subalgebra k[y1 , .., yr ] of the coordinate ring k[Z] = k[X1 , ..., Xn ]/a
such that the inclusion :
k[X1 , ..., Xn ]/a ⊃ k[y1 , .., yr ]
is an integral extension, and the elements yi are k-algebraically independent in k[Z]. Thus k[y1 , .., yr ] =
k[Ar (k)].

Let Y1 , .., Yr , which are polynomial functions of X1 , .., Xn , be elements of k[X1 , .., Xn ] which map to y1 , .., yr
respectively under the quotient map p : k[X1 , .., Xn ] → k[X1 , ..., Xn ]/a. Define :
π : An (k) → Ar (k)
(a1 , .., an ) 7→ (Y1 (a1 , .., an ), ..., Yr (a1 , .., an ))
In other words, π = i∗ , where i : k[y1 , .., yr ] ,→ k[Z] is the inclusion. Clearly π is continuous with respect to the
Zariski topology since it is a morphism. By its definition above, it also continuous with respect to the classical
topology if k = C.

By the Exercise 6.1.8 above, π is surjective. We need to show the last two finiteness/properness statements.

For these assertions, note that {xj }nj=1 are elements of k[Z] which are integral over k[y1 , .., yr ], so there are
monic polynomial relations :
n n −1
xj j + aj1 (y1 , .., yr )xj j + ..ajnj (y1 , .., yr ) = 0
for j = 1, 2, ..n, where aji (y1 , .., yr ) ∈ k[y1 , .., yr ]. For a fixed (y1 , .., yr ) = (b1 , .., br ), this is a polynomial with
coefficients in k, so has only finitely many roots. Thus the fibre of π is finite. If k = C, then as (y1 , .., yr ) ranges
in a compact set in Cr , each coefficient aji (y1 , .., yr ), which is polynomial and so continuous with respect to the
classical topology, also ranges in some compact subset of C, so the roots of all the above monic polynomials
also range in some compact subsets of C. (All roots of a monic polynomial xn +a1 xn−1 +..an = 0 with complex
coefficients have modulus less or equal to max{1, Σni=1 | ai |}, a fact which crucially depends on the monicity of
the polynomial !). So the inverse image π −1 (K) ⊂ Z of a compact subset K ⊂ Cr is compact, and π is proper.
This proves the proposition. 2

6.3. An application: Invariant Theory for Finite Groups. For reasons to be explained in the next
section, an affine algebraic set X whose coordinate ring k[X] is an integral domain is called an irreducible
algebraic set. The quotient field of this integral domain is denoted by k(X), and is called the field of rational
functions on X.

Let G be a finite group acting regularly on an irreducible affine algebraic set X, i.e., for each g ∈ G, the
left translation Lg is a regular morphism from X to X, with regular inverse Lg−1 . Thus it induces a natural
k-algebra automorphism L∗g : k[X] → k[X], given by L∗g f = f ◦ Lg−1 for f ∈ k[X]. For notational convenience,
we will denote L∗g f by g.f . Assume that the characteristic of the field k is not divisible by |G|. We would like
to make the quotient set X/G an affine algebraic set by describing its coordinate ring. Finally, we want the
natural quotient map π : X → X/G to be a surjective morphism with finite fibres.

It is clear that if any regular function h on the to-be X/G is composed with π, which is to be a regular
morphism, then this composite will be forced to be a regular function on X which will be constant on each
orbit, i.e. a function which is invariant under the action h 7→ g.h of G on k[X]. Conversely, we would like
every such G-invariant regular function on X to descend to the quotient X/G, in keeping with the universal
property of any quotient object in any category.
44 VISHWAMBHAR PATI

Introduce the averaging operator:

A : k[X] → k[X]
1 X
f 7 → g.f
|G|
g∈G

which makes sense since char(k) does not divide |G|. It is clear that A maps k[X] onto the G-invariant
subalgebra defined by:

k[X]G := {f ∈ k[X] : g.f = f for all g ∈ G}

Of course, A is not a ring homomorphism, but is easily checked to be a k[X]G -module endomomorphism of
k[X]. Let {yi }ni=1 be a set of k-algebra generators for k[X] (for example, the coordinate functions, if X is
realised as isomorphic to an affine closed subset of An (k)). Let us denote d := |G|.

Consider the single variable polynomial of degree d:

Pi (T ) = Πg∈G (T − g.yi )
This is a monic polynomial in T whose coefficients are the elementary symmetric polynomials in the roots
{g.yi }g∈G . G acts on this set of roots by permutations, so all the coefficients {aij }dj=1 of this polynomial are
in the invariant subalgebra k[X]G . Also note Pi (yi ) = 0. Hence all algebra generators yi of k[X] are integral
over k[X]G , and it follows that the inclusion homomorphism j : k[X]G ,→ k[X] is an integral extension.

We now claim that k[X]G is a k-algebra of finite type. Look at the set:
S = {A(y1i1 ...ynin ) : ij ≤ d − 1 ∀ 1 ≤ j ≤ n} ∪ {aij : 1 ≤ j ≤ d; 1 ≤ i ≤ n}
We claim that S is a set of k-algebra generators of k[X]G . For, if f = f (y1 , .., yn ) is any element of k[X]G ,
by using the monic relations Pi (yi ) = 0, we can convert it into a polynomial in aij and yi such that only powers
of yi which occur in f are ≤ d − 1. That is f = ΣλJI aJ y I where λJI ∈ k, aJ is a monomial in aij , y I = y1i1 ..ynin
with ir ≤ d − 1 for all 1 ≤ r ≤ n. Since f ∈ k[X]G , f = Af = ΣλJI aJ A(y I ), (since A is k[X]G linear, and
aJ ∈ k[X]G . This shows that S generates k[X]G as a k-algebra.

Since k[X]G is a subalgebra of k[X], it is also an integral domain, and by the Proposition 6.1.3, it is the
coordinate ring of an irreducible affine closed set which we denote by X/G. Since the injection j : k[X]G ⊂ k[X]
is an integral extension, it induces a finite surjective map j ∗ : X → X/G, which is the required regular quotient
map, which we will denote by π. It is a finite map by definition.

Those who are familiar with finite Galois extensions will note that the inclusion of the corresponding function
fields, namely the field extension k(X) ⊃ k(X)G is a Galois extension of degree | G |. This is by the fundamental
theorem of Galois theory. Also, the case of G = Sn acting by permutations on k n , is the classical one considered
by Galois in his work on the solvability of equations. If one views k n as the space of all roots of monic
polynomials of degree n, with G acting by permutation of these roots, the space (k n )G is interpreted as the
space of all monic polynomials
Qn of degree n, the quotient map π : k n → (k n )G simply maps (x1 , ..., xn ) to the
monic polynomial i=1 (X − xi ). If one identifies the space of monic polynomials of degree n with k n , the
polynomial f (X) = X n + Σni=1 ai X n−i being identified with (a1 , .., an ), we clearly have
π(x1 , .., xn ) = (−σ1 (x1 , .., xn ), .., (−1)n σn (x1 , .., xn ))
where the elementary symmetric functions σi are a set of k-algebra generators for k[X]G . Finally, we note
that for a monic polynomial f of degree n, π −1 (f ) consists of all permutations of the n-roots of f , and if
f is a monic polynomial with distinct roots, the cardinality of this fibre π −1 (f ) is n !. If it does not have
n!
distinct roots, the fibre will have cardinality d1 !..d k!
where di is the multiplicity of the i − th (distinct) root.
The degenerate polynomials f (X) = (X − a)n will have singleton fibres. Since k is algebraically closed, every
fibre is non-empty. The subset of polynomials whose fibres have cardinality strictly less than n ! constitutes a
closed
Q subset of X G = k n , namely, the hypersurface
Qn defined by the vanishing of the discriminant (defined as
i6=j (x i − xj ) for the polynomial f (X) = i=1 (X − xi ). It can be expressed as a polynomial in the coefficients
σi of f . For example, in the case of n = 2, the discriminant is σ12 − 4σ2 .)
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 45

6.4. Noetherian Spaces, irreducibility.

Definition 6.4.1. A topological space Z is said to be Noetherian if every descending chain of closed sets:
...Ci+1 ⊂ Ci ... ⊂ C1 ⊂ C0 = Z
becomes stationary, i.e., there exists an N such that Cj = Cj+1 for all j ≥ N .

Proposition 6.4.2. An algebraic set with its Zariski topology is Noetherian.

Proof:

Let ...Cj+1 ⊂ Cj ⊂ ....Z be a descending chain of closed subsets of Z. The sequence of ideals
I(Cj ) := {f ∈ k[Z] : f|Cj ≡ 0}
is an ascending chain of ideals in the Noetherian ring k[Z] (by Proposition 2.3.7) and thus becomes stationary.
Since Cj are Zariski closed, Cj = Ve (I(Cj )), and the descending chain above becomes stationary. 2

Definition 6.4.3. We call a space quasicompact if every open covering of it has a finite subcovering.
Usually such spaces are called compact. But Bourbaki, and other mathematicians of the French school,
included hausdorffness in their notion of compactness. So, for purely historical reasons, algebraic geometry
conventions dictate that such spaces be called quasicompact, since most spaces in the Zariski topology are not
hausdorff.

Corollary 6.4.4. Every open subset of a Noetherian topological space Z is quasicompact.

Proof: Let U be an open subset of Z, and let us assume that {Uα }α∈Λ is an open covering of U with no finite
subcovering. Then it is possible to find a sequence of opens {Uαi }∞ k
i=1 such that the open subsets Vk = ∪i=1 Uαi
c
satisfy Vk ⊂ Vk+1 is a proper inclusion. Thus the complements Ck = Vk give a strictly descending chain of
closed sets which is never stationary, which contradicts that Z is Noetherian. 2

Corollary 6.4.5. In a Noetherian topological space, every intersection of closed sets is a finite intersection.
Proof: If not, one easily obtains a strictly descending chain of closed sets. 2

Definition 6.4.6. A topological space Z is said to be irreducible if it cannot be expressed as the union of two
proper closed subsets. That is, if Z = C1 ∪ C2 , with Ci closed, then Z =C1 or Z = C2 .

Remark 6.4.7. Clearly, an irreducible space is connected, but not conversely. The algebraic set defined by
the union of the coordinate axes V (XY ) ⊂ C2 is connected, but expressible as V (X) ∪ V (Y ) and thus not
irreducible.

Proposition 6.4.8. If p is a prime ideal in any commutative ring, and p ⊃ ab for some ideals a, b ⊂ A, Qnthen
p ⊃ a, or p ⊃ b. In particular if a prime ideal p ⊃ a ∩ b, then p ⊃ a or p ⊃ b. More generally, if p ⊃ i=1 ai
for some ideals ai in A, then p ⊃ aj for some j. In particular, if p ⊃ ∩ni=1 ai , then p ⊃ aj for some j.
Proof: This was the Exercise 4.1.3, but here’s the proof anyway. Suppose p 6⊃ a and p 6⊃ b. Then let a ∈ a \ p
and b ∈ b \ p. Then ab ∈ ab ⊂ p, which contradicts that p is prime. The secondQassertion follows since
n
a ∩ b ⊃ ab. The third assertion follows by induction. The last follows since ∩ni=1 ai ⊃ i=1 ai . 2

√
Proposition 6.4.9. An algebraic set V (a) ⊂ An (k) is irreducible iff a is a prime ideal in k[X1 , .., .Xn ].
Equivalently, for a reduced k-algebra of finite type, Spm(A) is irreducible iff A is an integral domain.
46 VISHWAMBHAR PATI

Proof: √
Let A denote k[X1 , .., Xn ]. Assume a is prime. Let V (a) =C1 ∪ C2 , where Ci are√closed√subsets, and hence
equal to V (ai ) for i = 1, 2. Thus nullstellensatz a = a1 ∩ a2 ⊃ a1 ∩ a2 .
√ V (a) =V√(a1 ∩ a2 ), so that by the √
By the previous proposition, a ⊃ a1 or a ⊃ a2 . Thus V (a) = V ( a) ⊂ V (a1 ) = C1 or V (a2 ) = C2 , which
implies V (a) = C1 or C2 .
√
Conversely, assume V (a) is irreducible, and let f g ∈ a. Then V (f g) = V (f ) ∪ V (g) ⊃ V (a ), so that :
V (a) = (V (f ) ∩ V (a)) ∪ (V (g) ∩ V (a)) = V (a + Af ) ∪ V (a + Ag)
√ √
By
√ irreducibility,
√ either V (a) = V (a + Af ) or V √(a) = V (a + Ag), which implies
√ a = a +√Af ⊃ a + Af or
a = a + Ag ⊃ a + Ag. In the first case f ∈ a, and in the second g ∈ a, showing that a is prime.

The second assertion follows immediately from the Proposition 6.1.3. 2

Proposition 6.4.10. Every algebraic set is a finite union of irreducible closed sets.
Proof: Fix an n, and let Σ be the collection of all algebraic subsets Z ⊂ An (k) which are not finite unions of
closed irreducibles. We claim that Σ = φ.

Assume Σ 6= φ. Order it by inclusion. We claim that every chain Λ = {Zα } in Σ has a lower bound. Let
Z = ∩α Zα . By the Corollary 6.4.5 above, Z = ∩α∈F Zα , where F is a finite set. Since Λ is a chain, Z = Zγ
where Zγ is the smallest element of the finite family {Zα }α∈F . Thus Z ∈ Σ.

By Zorn’s lemma, Σ has a minimal element Z, say. We claim that Z is irreducible. For if Z = C1 ∪ C2 ,
with Ci closed and both proper subsets of X, by the minimality of X in Σ, it follows that C1 and C2 are not
members of Σ, and hence both are finite unions of closed irreducible sets, which means that Z is also a finite
union of closed irreducibles, contradicting that Z ∈ Σ. Thus Z is irreducible, and since it is closed, this means
Z∈/ Σ, which is again a contradiction. Thus Σ is empty, and the assertion is proved. 2

Remark 6.4.11. In the above proposition, if we express an algebraic set Z = ∪ki=1 Zi , where Zi are irreducible
algebraic sets, and make the union irredundant (i.e. Zi 6⊂ Zj for i 6= j), then the sets Zj are uniquely determined
upto a permutation, and called the irreducible components of Z. The proof is an easy exercise.

√
Corollary 6.4.12. Let a be an ideal in k[X1 , ..., Xn ]. Then there exist prime ideals {pi }ki=1 such that a=
∩ki=1 pi and pi 6⊂ pj for i 6= j. Further, this collection of primes is uniquely determined.

Proof: Consider Z = V (a). By the Propositions

p 6.4.9 and 6.4.10, it follows that V (a) = ∪ki=1 V (pi ) =
√ √
V (∩ki=1 pi ), from which it follows that a = ∩ki=1 pi = ∩ki=1 pi = ∩ki=1 pi . The uniqueness assertion follows
from the Remark 6.4.11 above. 2

The corollary above is a purely algebraic statement about polynomial rings, even though geometric ingre-
dients went into proving it. In the special case that a = hf i, i.e., a principal ideal, the corollary above can
be deduced from the unique prime factorisation in k[X1 , .., Xn ] proved in Proposition 2.2.17. Thus the above
corollary should be viewed as a way of factorising radical ideals into prime factors. There is thing called “pri-
mary decomposition” in arbitrary Noetherian rings which is the most general avatar of prime factorisation,
and we shall encounter it later.

Remark 6.4.13. Usually, one is used to decomposing elements into products of primes, so it would seem
natural to seek expressions of arbitrary ideals as ideal products of prime ideals. We note that for a prime ideal
p
n
Y
p⊃ ai ⇔ p ⊃ ai for some i ⇔ p ⊃ ∩ni=1 ai
i=1
(by 6.4.8) which implies (by 4.1.9) that the radical of a product of ideals is the same as the radical of their
intersection, for a radical ideal (or equivalently, a closed algebraic set), the statement of the last corollary is
true with intersection of prime ideals replaced by the radical of their product.
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 47

Incidentally, there are radical ideals in the world which cannot be expressed as a product of prime ideals.
Consider the:

Example 6.4.14. Let a = hXZ, Y Z, XY i ⊂ k[X, Y, Z]. It is easily verified that:

a = hX, Y i ∩ hY, Zi ∩ hZ, Xi
and each of the three ideals on the right is prime. Thus a is a radical ideal. However the product of the three
prime ideals on the right is the ideal:
hX 2 Y, XY 2 , X 2 Z, XZ 2 , Y 2 Z, Y Z 2 i
which is strictly contained in a.

Exercise 6.4.15. Let X be an affine closed set with irreducible components {Xi }m i=1 . Then show that the
coordinate ring k[X] is a k-subalgebra of ⊕mi=1 k[Xi ] (which is a reduced k-algebra in a natural manner by
coordinate-wise operations). Determine the coordinate ring of k[Z] where Z = V (XY ) ⊂ A2 (k) as a subalgebra
of k[X] ⊕ k[Y ].

Exercise 6.4.16.
(i): In the polynomial ring k[X, Y, Z], compute the radical of the ideal a = hXY 2 , Y 2 Zi.

(ii): Write down the decomposition of V (a) (a being the ideal in (i) above) into its irreducible components.

Exercise 6.4.17. Let X = ∪m i=1 Xi be the irredundant decomposition of an affine algebraic set X, and let
x ∈ Xi \ (∪j6=i Xj ). Show that there exists a regular function f ∈ k[X] which satisfies f (x) 6= 0, and f|Xj ≡ 0
for all j 6= i.

Now we would like to define a convenient basis for the open subsets of an algebraic set Z.

Proposition 6.4.18. Let Z ∈ An (k) be an algebraic set, and k[Z] its coordinate ring. Then for f ∈ k[Z],
define the subset
D(f ) = {p ∈ Z : f (p) 6= 0}
Then D(f ) is an open subset of Z, called a principal open set or basic open set of Z. The collection of
principal open sets
{D(f ) : f ∈ k[Z]}
is a basis for the Zariski topology on Z. In fact, for any open set U ⊂ Z, one may write:
U = ∪m
i=1 D(fi )

where fi ∈ k[Z] are some regular functions. Finally, Z = ∪m

P i=1 D(fi ) iff there exist hi ∈ k[Z] such that
m
i=1 hi fi = 1.

Proof: By the Proposition 6.1.3, the set Ve (f ) ⊂ Z is a closed subset of Z, and D(f ) is its complement, so it
is open. If U is an open subset of Z, then U = Z \ Ve (b), for some ideal b ⊂ k[Z]. Since k[Z] is Noetherian, we
have b = hf1 , .., fm i, so that Ve (b) = ∩m e m
i=1 V (fi ), which implies that U = ∪i=1 D(fi ).

For the last m e e

√ statement, note that Z = ∪i=1 D(fi ) iff V (hf1 , .., fm i) = φ = V (h1i), and this happens iff
the radical
Pm b of the ideal b := hf1 , .., fm i is the whole ring k[Z]. This, in turn, happens iff 1 ∈ b, i.e. iff
1 = i=1 hi fi . The proposition follows. 2

We have already defined regular functions on an affine algebraic set X as the elements of the coordinate
ring k[X]. Our next aim is to define the local ring at a point x ∈ X, as well as the ring of regular functions on
an open set U ⊂ X. First we observe the following easy fact about localisations.
48 VISHWAMBHAR PATI

Proposition 6.4.19. Let A be an integral domain, and let K = Q(A) denote its quotient field. For a prime
ideal p, we denote the localisation S −1 A at the multiplicative set S = A \ p by Ap (see Definitions 5.3.1 and
5.3.3). Clearly Ap ⊂ K for each prime ideal p of A. Then we have:
\ \
A= Ap = Am
p∈ Spec A m∈ Spm A
where the intersections on the right are taken in K.

Proof: Since A is a domain, it sits inside all its localisations, so it is clear that A is contained in the two
intersections on the right. Also since Spm A ⊂ Spec A, it follows that
\ \
Ap ⊂ Am
p∈ Spec A m∈ Spm A

Thus it is enough to prove that inside K, we have:

\
Am ⊂ A
m∈ Spm A
So let f ∈ K be an element of the left hand side. Then, for each m ∈ Spm A, we have the following equality
in K:
am
f=
bm
with bm ∈ A \ m. Consider the ideal in A defined by:
a := hbm im∈ Spm A
If this ideal a is a proper ideal of A, there will be a maximal ideal n of A containing it. But then n ∈ Spm A,
and so bn ∈ a. It follows that bn ∈ n, contradicting that bn ∈ A \ n. It followsP that a must be A. In particular,
r
there exist elements bi ∈ A \ mi , for mi ∈ Spm A, i = 1, 2, .., r, such that 1 = i=1 hi bi , for some hi ∈ A. Thus
r
X r
X
f= hi bi f = hi ai
i=1 i=1
since f = abii for each i. But the element on the right hand side above is clearly in A, which proves our assertion
and the proposition follows. 2

6.5. Local rings and regular functions on open sets. We have already defined regular functions on X to
be elements of the coordinate ring k[X]. It will be useful to define other kinds of functions, which are regular
at a point, or on an open subset, etc.

Definition 6.5.1 (Local ring at a point). Let X be an affine closed set, with the reduced affine coordinate
ring k[X]. For a maximal ideal mx corresponding to the point x ∈ X, the localisation k[X]mx (in the sense of
Example 5.3.3) is called the local ring of X at x. This local ring will be denoted by OX,x . Its unique maximal
ideal is mx OX,x , sometimes denoted by abuse of notation as mx . An element in this local ring is a quotient fg ,
where g ∈ k[X] \ mx , i.e. g(x) 6= 0. The image of a regular function f ∈ k[X], viz. the element f1 ∈ OX,x is
denoted fx , and called the germ of f at x. This germ fx is not be confused with f (x) ∈ k. An element of OX,x
is said to be a function regular at x.

Definition 6.5.2 (Function field of an irreducible affine algebraic set). If X ⊂ An (k) is an irreducible closed
algebraic set with X = V (p) where p ⊂ k[X1 , .., Xn ] is a prime ideal, its coordinate ring k[X] = k[X1 , .., Xn ]/p
is an integral domain. The quotient field Q(k[X]) makes sense, and is called the field of rational functions on
X, or function field of X, and denoted k(X). An element of k(X) can be written as fg where g 6= 0 in k[X],
i.e. g does not vanish identically on X. In this situation of X irreducible, each local ring OX,x is a subring of
k(X).
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 49

Example 6.5.3. For X = An (k), the function field is k(X) = k(X1 , X2 , .., Xn ), the field of rational functions
in n-variables.

We remark here that although the coordinate ring k[X] of X is a quotient of the polynomial coordinate ring
k[X1 , .., Xn ] of An (k), the function field k(X) of X generally bears no relation to the function field k(X1 , .., Xn )
of An (k).

We now list some immediate facts about these local rings.

Proposition 6.5.4 (Some facts on OX,x and k(X)).

(i): There is an evaluation map x : OX,x → k, sending fg to fg(x)(x)
, which is a k-algebra homomorphism. Its
kernel is precisely the unique maximal ideal mx OX,x (= the ideal localisation S −1 mx at the multiplicative
set S = k[X] \ mx ). mx OX,x is therefore the ideal of functions in OX,x which vanish at x.

(ii): If f, g ∈ k[X], and f (x) = g(x) for all x ∈ X, then f = g. In particular, if the germs fx = gx in OX,x
for all x ∈ X, then f = g.

(iii): If f : X → Y is a morphism of affine algebraic sets, induced by the k-algebra homomorphism f ∗ :

k[Y ] → k[X] of their coordinate rings, then for each x ∈ X, there is an induced homomorphism of local
rings denoted:
fx∗ : OY,f (x) → OX,x
which satisfies x ◦ fx∗ = f (x) . Thus (fx∗ )−1 (mx ) = mf (x) , where mx and mf (x) are the unique maximal
ideals of OX,x and OY,f (x) respectively.

(iv): Let f : X → Y be a morphism as in (iii), and X and Y be irreducible. Further assume that the
homomorphism f ∗ : k[Y ] → k[X] is injective. Then we have an induced injection of function fields
f ∗ : k(Y ) → k(X).

(v): If X is irreducible, we have:

\ [
OX,x = k[X], OX,x = k(X)
x∈X x∈X

where the intersection and union are taken inside the function field k(X).

Proof: To see (i), observe that there is the evaluation homomorphism x : k[X] → k[X]/mx = k. Also
it maps every element of the multiplicative set k[X] \ mx to k \ {0}, i.e. invertible elements in k. Thus
by (i) of the Proposition 5.3.11, there is a homomorphism OX,x → k, which we also denote x , satisfying
x (fx ) = x (f ) = f (x). For a typical element fg ∈ OX,x , with g 6∈ mx (i.e. g(x) 6= 0) we have x ( fg ) = fg(x)
(x)
∈ k.
The second statement of (i) is clear.

To see (ii), note that f (x) = g(x) for all x ∈ X means that f − g ∈ ∩x∈X mx . The right hand side is the
nilradical of k[X] by the Proposition 5.2.1, which is {0} since k[X] is reduced. Thus f = g. Since fx (x) = f (x)
for all x ∈ X, the second assertion of (ii) is clear.

To see (iii), note that the inverse image (f ∗ )−1 mx ⊂ k[X] is mf (x) ⊂ k[Y ] (by the Proposition 5.1.7 in §5).
Now we apply (ii) of the Proposition 5.3.11 to get a homomorphism ff∗ : OY,f (x) → OX,x of the corresponding
localised rings. We will also denote this homomorphism by f ∗ , for notational simplicity.

Since f ∗ is injective, and OX,x and OY,f (x) are domains, the inverse image of the prime ideal {0} in OX,x
is the prime ideal {0} in OY,f (x) . Now (iv) is immediate from (ii) of the Proposition 5.3.11.

The first part of (v) follows from the Proposition 6.4.19. The second part follows by observing that if
f
g ∈ k(X), with g 6= 0 in k[X], then there exists a point x ∈ X such that g(x) 6= 0. This is because the
intersection of all the maximal ideals in k[X] is the nilradical of k[X] (by Proposition 5.2.1), which is {0}, since
k[X] is a domain. Thus the element fg lies in OX,x . This proves the proposition. 2
50 VISHWAMBHAR PATI

Remark 6.5.5. The part (v) of the last proposition says that every rational function on an irreducible affine
algebraic set X is regular at some point, and a rational function that is regular at each point of x is a regular
function on X.

Remark 6.5.6. The requirement that k be algebraically closed is crucial. The reader can easily construct an
example of a function f ∈ k[A1 (k)] = k[T ], k a finite field, with f (x) = 0 for all x ∈ A1 (k) but f 6= 0. (Exercise:
If k is an infinite but not algebraically closed, and f ∈ k[X1 , .., Xn ] satisfies f (x) = 0 for all x ∈ An (k), is it
true that f = 0?)

Exercise 6.5.7. Describe the local ring OZ,0 of the pair of intersecting lines Z := V (XY ) ⊂ A2 (k).

Exercise 6.5.8. Using the map φ : A1 (k) → V (Y 2 − X 3 ) given by t 7→ (t2 , t3 ), describe the local ring OZ,0 of
the plane cusp Z := V (Y 2 − X 3 ) at the origin as a subring of k(t).

Proposition 6.5.9 (Regular functions on open sets). Let X be an irreducible affine algebraic set, and f ∈
k[X] be a regular function on X. Let D(f ) ⊂ X denote the principal open set of points at which f does not
vanish (see the Proposition 6.4.18). Then we have:
\
OX,x = (k[X])f
x∈D(f )

inside the function field k(X). The intersection on the left side is precisely the set of functions regular at each
point of D(f ), and denoted by k[D(f )]. An element of k[D(f )] is said to be a regular function on D(f ).

Proof: First note that a maximal ideal n in k[X]f is of the form n = S −1 m, where S is the multiplicative set
S := {1, f, f 2 , .., f m , ..}, and m is a maximal ideal in k[X] disjoint from S (See Exercise 5.3.6). On the other
hand m a maximal ideal in k[X] implies m = mx for some x ∈ X by the weak Nullstellensatz (Corollary 5.1.6).
Further S ∩ m = φ iff f 6∈ mx iff f (x) 6= 0 iff x ∈ D(f ). Thus the maximal spectrum of k[X]f is precisely:
Spm k[X]f = {n : n = S −1 mx : x ∈ D(f )}
Finally, for x ∈ D(f ), since S = {1, f, f 2 , ..} ⊂ k[X] \ mx , it also follows easily that for x ∈ D(f ) and
n = S −1 mx ∈ Spm k[X]f :


(k[X]f )n = S −1 k[X] S −1 m = k[X]mx = OX,x
x

Now we apply the second equality of the Proposition 6.4.19 to the domain A = k[X]f (whose quotient field is
the same as that of k[X], viz. k(X)) to conclude that:
k[X]f = ∩n∈Spm k[X]f (k[X]f )n = ∩x∈D(f ) OX,x
which proves our proposition. (Note that the proof of the last statement of Proposition 6.4.18 is just a special
case of this proof by setting f = 1). 2

Corollary 6.5.10. Let U ⊂ X be open, with X irreducible as above. Then the subalgebra k[U ] ⊂ k(X) of
rational functions which are regular on U (i.e. at each point of U ) can be described as the intersection:
∩ni=1 k[X]fi
where U = D(f1 ) ∪ D(f2 )... ∪ D(fn ) is an expression for U as a union of finitely many basic open sets (see
Proposition 6.4.18). By the proposition above, the subalgebra k[U ] consists of all functions in k(X) which are
regular at each point x ∈ U , and in particular, the expression above for k[U ] holds for every expression of U
as a union of principal opens D(fi ).
Proof: Clear from the foregoing proposition. 2
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 51

Exercise 6.5.11. Let X be an affine algebraic set with an irredundant decomposition X = ∪m i=1 Xi into
irreducible affine closed sets. Let x ∈ Xi \ (∪j6=i Xj ). Then show that the natural (restriction) homomorphism:
OX,x → OXi ,x
is an isomorphism. That is, the local ring at a point x lying in a unique irreducible component Xi “sees” only
that irreducible component. (Hint: Surjectivity of this homomorphism is easy. For injectivity, one needs to
construct a regular function which does not vanish at x, and vanishes identically on all Xj for j 6= i, which
was the assertion of Exercise 6.4.17.)

Exercise 6.5.12. Consider the plane curve Z := V (Y 2 − X 2 (X − 1)) (called the cubic node).

(i): Show that Z is irreducible and hence k[Z] is a domain.

(ii): Denote the images of X and Y in the coordinate ring k[Z] by x and y respectively. Show that the
rational functions y 2 /x ∈ k(Z) and y 2 /x2 ∈ k(Z) are regular functions on Z.

(iii): Show that the rational function y/x is not a regular function. Show using this fact that k[Z] is not
integrally closed, and in particular, not a UFD.

Exercise 6.5.13. Let X be an irreducible affine algebraic set. Show that every non-empty open subset U ⊂ X
is Zariski-dense in X.

Exercise 6.5.14. Let X be an irreducible affine algebraic set, and assume that X has a covering by basic
open sets, viz., X = ∪ni=1 D(fi ) for some regular functions fi ∈ k[X]. Show that the intersection:
∩ni=1 k[X]fi = k[X]
where k[X]fi denotes the localisation of k[X] at the multiplicative set {1, fi , fi2 , ..., }, and the intersection above
is taken in the function field k(X) of X. Conclude that if a rational function f ∈ k(X) is regular on each
D(fi ), then it is a regular function on X.

6.6. Affine Algebraic Varieties. Let X be an irreducible affine algebraic set, with the coordinate ring k[X].
In the Corollary 6.5.10 of the last subsection, we have seen how to define the ring k[U ] of functions regular on
U as a subring of the function field k(X). These regular functions on open sets U have the following property:

Proposition 6.6.1 (The sheaf of regular functions). Let X ⊂ An (k) be an irreducible affine algebraic set. For
each Zariski open subset U ⊂ X, we define O(U ) = k[U ]. We set O(φ) = {0}. Then:

(i): For each open subset U ⊂ X, and each x ∈ U there is a natural k-algebra homomorphism:
ρU,x : O(U ) → OX,x
f 7 → fx
mapping a function to its germ at x.

(ii): For U ⊂ X an open set, and a regular function f ∈ O(U ), f = 0 iff ρU,x (f ) = 0 for each x ∈ U . That
is, a function regular on an open set U is uniquely determined by its germ at each point.

(iii): For each pair of Zariski open subsets V ⊂ U of X are natural (restriction) ring homomorphisms:
ρU,V : O(U ) → O(V )
f 7 → f|V
mapping f to its restriction. These satisfy the compatibility conditions:

(a): For all opens W ⊂ V ⊂ U , we have ρV,W ◦ ρU,V = ρU,W . The homomorphism ρU,U : O(U ) → O(U )
is the identity map.
52 VISHWAMBHAR PATI

(b): For any x ∈ V ⊂ U , U, V open subsets of X, ρV,x ◦ ρU,V = ρU,x .

(c): (Patching of functions) If G ∈ O(U ) and H ∈ O(V ) are functions satisfying ρU,U ∩V (G) =
ρV,U ∩V (H), then there exists a unique F ∈ O(U ∪ V ) such that ρU ∪V,U (F ) = G and ρU ∪V,V (F ) = H.

Proof: Since O(U ) = ∩x∈U OX,x , it follows that there is a natural inclusion:
O(U ) ,→ OX,x
which is precisely the homomorphism ρU,x . This gives (i).

(ii) is now trivial, since all the maps ρX,x are inclusions. Indeed, it is worth remarking that if f ∈ O(U ),
and ρU,x f = 0 for any single x ∈ U , then f = 0 in O(U ), since the map ρX,x is an inclusion. This seems a bit
puzzling, but can be understood as follows. Write f ∈ O(U ) as f = g/h ∈ k(X), where g, h ∈ k[X], and h 6= 0.
Then hf = g in O(U ), and all the functions f, g, h ∈ O(U ). Assuming ρU,x f = 0, apply ρU,x to this equation
to obtain that ρU,x (g) = 0, where g also denotes g|U . This means α.g = 0 in k[X] for some α ∈ k[X] \ mx .
That is, α(x) 6= 0. Hence X = V (α) ∪ V (g), with V (α) 6= X since x 6∈ V (α). The irreducibility of X implies
that X = V (g), so that g ≡ 0, and hence f = 0.

For (iii), let V ⊂ U be open subsets of X. By the Proposition 6.4.18, we have:

O(U ) = ∩x∈U OX,x ⊂ ∩x∈V OX,x = O(V )
and this natural inclusion map is the restriction map ρU,V . The assertion (a) now follows easily, since composites
of inclusions are inclusions, and the inclusion of O(U ) into itself is the identity map of O(U ).

Part (b) is also trivial, since all maps concerned are inclusions. To see the patching condition (c), we first
note that by expressing V as a finite union of basic opens, it boils down to proving it for V = D(g) 6= φ, U 6= φ
arbitrary. So suppose we have U = ∪m i=1 D(gi ), and V = D(g). Note that since X is irreducible, U 6= φ implies
U is dense in X (by Exercise 6.5.13), as is D(g), if it is non-empty. Consequently, U ∩ D(g) is non-empty.

Let G ∈ O(U ), so that G = fi /gir for each i. (We may choose the largest r among the exponents of all
denominators, and make a common exponent for the denominators.) Also let H = f /g r ∈ O(D(g)) = k[X]g .
The fact that ρU,U ∩D(g) (G) = ρD(g),U ∩D(g) (H) implies that applying ρU ∩D(g),x to both sides produces the same
germ. By the part (b), this implies that Hx = Gx for all x ∈ U ∩ D(g). This means that for each x ∈ U ∩ D(g)
the following equations hold in k[X]:
hx (fi g r − f gir ) = 0
for some hx ∈ k[X] \ mx , and each i = 1, 2, .., m. Since hx 6∈ mx , we have hx (x) 6= 0, and since k[X] is a domain
by the irreducibility of X, we have:
fi (x)g r (x) − f (x) gir (x) = 0 1 ≤ i ≤ m, x ∈ U ∩ D(g)
Since U ∩ D(g) is non-empty and open, it is Zariski dense in X (by Exercise 6.5.13), and so it follows that:
fi (x)g r (x) = f (x) gir (x) 1 ≤ i ≤ m for all x ∈ X
Again this implies that for each i, the regular function fi g r − f gir ∈ ∩x∈X mx = {0}. Thus, in k(X), we have
the identity:
f1 fi f
= .... = r = .... = r
g1r gi g
If we call this rational function F , we have that F = G on U and F = H on D(g), and hence F is regular
at each point of U ∪ D(g). It is therefore the required regular function in O(U ∪ D(g)). The uniqueness of F
follows by (ii) above. This proves the proposition. 2
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 53

Definition 6.6.2. Let X be a topological space. A sheaf of rings F on X is an association of a ring F(U ) to
each open set U ⊂ X. For every pair of opens V ⊂ U , there is the restriction map ρU V : F(U ) → F(V ) which
is a ring homomorphism. Furthermore, these restriction homomorphisms are required to satisfy (a) and (c) of
(iii) in Proposition 6.6.1. Likewise, one may define sheaves of abelian groups, k-algebras, k-vector spaces etc.
The stalk of the sheaf F at a point x ∈ X, denoted Fx is defined as:
Fx = lim F(U )
→

where the direct limit is taken over the directed set of neighbourhoods U(x) of x ordered by U < V if U ⊃ V ,
and the maps of the direct system being the restriction maps ρU V . (For more on these matters, the reader
may consult a book on Sheaf Theory, or Hartshorne’s Algebraic Geometry).

The sheaf on an irreducible affine algebraic set X defined above in Proposition 6.6.1 is called the structure
sheaf of X, and denoted OX , or simply O, if no confusion is likely.

Exercise 6.6.3 (Sheaf of continuous functions). The reader may similarly construct, for a topological space
X, the structure sheaf of continuous functions on X. For U ⊂ X open O(U ) is defined as the ring of continuous
functions on the set U , and the local ring OX,x is the ring of germs of continuous functions at x. It consists
of equivalence classes hf, U i, where U is an open neighbourhood of x, and the equivalence relation is given
by hf, U i = hg, V i iff there exists a W ⊂ U ∩ V such that f|W = g|W . It is precisely the localisation of
O(X) = C(X) at the maximal ideal mx = {f : f (x) = 0}, and is therefore a local ring. The maps ρU,V are
defined by restriction, and ρU,x is the map taking a function f continuous on U to the germ hf, U i. All the
assertions above are easily checked in this situation by usual facts about patching of continuous functions etc.
The reader may similarly construct the sheaf of smooth functions on a smooth (differentiable) manifold X.

Definition 6.6.4 (Affine algebraic varieties). A topological space X, together with a sheaf of rings FX , such
that each stalk FX,x is a local ring, is called a locally ringed space. A locally ringed space (X, FX ) is called an
affine algebraic variety over k if there is an irreducible affine algebraic set Y , and a homeomorphism h : X → Y ,
and an isomorphism h∗ of FX with the natural structure sheaf of k-algebras OY of Y defined in the Proposition
6.6.1 above. That is, for each U ⊂ X open, there should be an isomorphism of rings h∗U : FX (U ) ' OY (h(U )),
compatible with all the restrictions ρU,V on both sides.

The reason for doing all this is to free the definition of an affine algebraic set from the ambient affine space
it sits inside. The important object attached to an affine variety is the structure sheaf, which determines it
completely. The ambient affine space is not important.

Exercise 6.6.5. Show that a basic open set D(f ) of an irreducible affine algebraic set X is an affine algebraic
variety. (In fact, if X ⊂ An (k), then one can make D(f ) isomorphic to a closed subset Y ⊂ An+1 (k). That is,
there is a bijection between h : D(f ) → Y , which is a homeomorphism of Zariski topologies, and such that for
any open set W ⊂ D(f ) and its image h(W ) ⊂ Y , the rings of regular functions OX (W ) and OY (h(W )) are
isomorphic as k-algebras. (Hint: See part (iii) of the Exercise 5.3.12.)

In particular, by viewing the group GL(n, k) as the principal open set D(det) inside the affine space
2
M (n, k) = An (k), we have that GL(n, k) is an affine algebraic variety.

Remark 6.6.6 (Open subsets of affines need not be affine). There are open subsets of affine varieties which
are not affine. For example, the open set U := A2 (k) \ {(0, 0)} (or for that matter the complement of the origin
in An (k) for all n ≥ 2) are not isomorphic to any affine algebraic set. The reason is as follows. Let X = A2 (k).
We first claim that OX (U ) = k[X, Y ]. Let V := D(X) and W := D(Y ) be the principal open subsets of A2 (k)
defined by the coordinate functions. Clearly U = V ∪W . By our earlier generalities, OX (U ) ⊂ OX (V )∩OX (W ).
If r ∈ k(X) is in OX (U ), it follows, since OX (V ) = k[X, Y ]X and OX (W ) = k[X, Y ]Y , that r = X −m f (X, Y )
on D(X) and r = Y −n g(X, Y ) on D(Y ). Thus at each point of V ∩ W = D(XY ), we have:
Y n f (X, Y ) = X m g(X, Y )
54 VISHWAMBHAR PATI

Again, both sides are regular functions on all of X = A2 (k), and D(XY ) is non-empty, open and hence Zariski
dense in the irreducible X, so the above equality holds on all of A2 (k), i.e. is an equality in k[X, Y ]. Since
k[X, Y ] is a UFD, and X and Y are irreducible, we have that Y n divides g(X, Y ) and X m divides f (X, Y ).
Thus r = X −m f (X, Y ) = Y −n g(X, Y ) is a polynomial in X, Y , viz., an element in k[X, Y ]. The claim follows.

We already know that for an affine algebraic set X, points of X are in 1-1 correspondence with maximal
ideals by the Corollary 5.1.6 of the Weak Nullstellensatz. Now if the open set U above were an affine variety,
the maximal ideal n := hX, Y i in k[U ] = k[X, Y ] would correspond to (functions vanishing at) some point
(a, b) ∈ U . But for p = (a, b) ∈ U , the maximal ideal mp corresponding to p is hX − a, Y − bi, which is not n
for any p ∈ U .

6.7. Birational Equivalence.

Definition 6.7.1. Closed irreducible algebraic sets whose function fields are k-isomorphic (i.e. as field exten-
sions of k) are said to be birationally equivalent. Clearly algebraic varieties which are isomorphic are birationally
equivalent. But birational equivalence is much weaker than isomorphism, as the following example shows.

Example 6.7.2 (The cubic cusp). Consider Y := V (Y 2 − X 3 ) ⊂ A2 (k), and X = A1 (k). We have the natural
map:
φ:X → Y
t 7 → (t2 , t3 )
The corresponding homomorphism between coordinate rings is:
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 55

k[X, Y ]
φ∗ : k[Y ] = → k[X] = k[T ]
hY 2 − X 3 i
x 7→ T 2
y 7→ T 3
where x and y are the images of X and Y in the quotient ring on the left. The map between the function fields
k(Y ) → k(X) induced by φ∗ is an isomorphism, with inverse map t 7→ xy . However, φ∗ : k[Y ] → k[X] is not an
isomorphism, as was noted in Exercise 5.3.22 (the reader can check that y/x is integral over k[Y ], but not in
it, so k[Y ] is not a normal domain). So there can be no k-algebra isomorphism between these rings.

We shall see more examples in the next section on plane algebraic curves. However, we can assert the
following:

Proposition 6.7.3. Let X and Y be two irreducible affine algebraic sets which are birationally isomorphic.
Then there exist non-empty Zariski open (and hence Zariski dense) subsets U ⊂ X and V ⊂ Y such that the
rings of regular functions OX (U ) and OY (V ) are isomorphic. In fact, if we restrict the structure sheaves OX
(resp. OY ) to U (resp. V ) and call them OU (resp. OV ), then (U, OU ) is isomorphic to (V, OV ) as a locally
ringed space.

Proof: Let k[X] be generated as a k-algebra by some regular functions x1 , .., xm (e.g. coordinate functions, if
X ⊂ Am (k)), and k[Y ] by the regular functions y1 , .., yn . If θ : k(Y ) → k(X) is an isomorphism, we may write:
fi (x1 , ..., xm )
θ(yi ) = , 1≤i≤n
gi (x1 , .., xm )
and also
pj (y1 , ..., yn )
θ−1 (xj ) = , 1≤j≤m
qj (y1 , .., yn )
where fi , gi ∈ k[X] and pj , qj ∈ k[Y ] are some regular functions.

Let
U1 = (∩ni=1 D(fi gi )) = (∩ni=1 D(gi )) ∩ (∩ni=1 D(fi ))
Then θ maps yi , and hence k[Y ] into OX (U1 ). In particular, θ(pj ) and θ(qj ) ∈ OX (U1 ) for each j = 1, .., m.
Now
θ(pj ) = pj (θ(y1 ), .., θ(yn )) = pj (f1 /g1 , .., fn /gn )
Pj (x1 , .., xm )
=
gI
where I = (i1 , .., in ) is some multiindex, and g I denotes the monomial g1i1 ...gnin . Similarly,
Qj (x1 , .., xm )
θ(qj ) =
gI
(Note that we can choose a common denominator g I for all the θ(pj )’s and θ(qj )’s, by taking the largest of the
exponents for each gi occurring in all of these expressions, and multiplying and dividing by a suitable positive
power of gi ) Now let U be the open set U1 ∩ (∩j D(Pj Qj )). Then, we have the following:

(i): For each j = 1, .., m, the rational functions Pj and Qj are regular functions on U , not vanishing anywhere
on U . Hence they are invertible elements of OX (U ).

(ii): For each i = 1, .., n, the functions fi (x1 , .., xm ) and gi (x1 , .., xm ) are regular and everywhere non-
vanishing on U . Thus they are also invertible elements of OX (U ).
56 VISHWAMBHAR PATI

By repeating the above mutatis mutandis for θ−1 , and defining V1 and V analogously, we obtain that
Fi (y1 , .., yn ) Gi (y1 , .., yn )
θ−1 (fi ) = J
; θ−1 (gi ) =
q qJ
Furthermore,

(iii): For each i = 1, .., n, the functions Fi and Gi are regular functions on V , not vanishing anywhere on
V . Hence they are invertible elements of OY (V ).

(iv): For each j = 1, .., m, the functions pj (y1 , .., yn ) and qj (y1 , .., yn ) are regular and everywhere non-
vanishing on V . Thus they are also invertible elements of OY (V ).

By definition of U , gi−1 ∈ OX (U ), and thus θ maps k[Y ] into OX (U ). Now OY (V ) is the localisation of k[Y ]
at the multiplicative set generated by
SY = {pj , qj , Fi , Gi }1≤i≤n; 1≤j≤m .
Pj Qj
Now θ(pj ) = gI
and θ(qj ) = gI
,and we have Pj , Qj and gj invertible elements of OX (U ), by (i) and (ii) above.
Hence θ(pj ) and θ(qj ) are invertible in OX (U ). Also note that Fi = q J θ−1 (fi ) implies θ(Fi ) = θ(q J )fi , and
since θ(qj ) is invertible in OX (U ) for each j (by the last line), and fi are invertible in OX (U ) by (ii) above, it
follows that θ(Fi ) is invertible in OX (U ) for each i. Similarly, θ(Gi ) is invertible in OX (U ) for each i.

By the universal property of localisations (see Proposition 5.3.11, (i)), we have that θ defines a map OY (V ) →
OX (U ). Interchanging the roles of θ and θ−1 , U and V , X and Y , and using (iii) and (iv) above, we similarly
obtain that θ−1 maps OX (U ) into OY (V ). Since θ and θ−1 are inverses of each other all over the respective
function fields, the maps θ and θ−1 above give an isomomorphism of OX (U ) with OY (V ) and the proposition
is proved. 2

Definition 6.7.4. An affine algebraic variety X is said to be rational if it is birationally equivalent to affine
n-space An (k), for some n. That is, its function field k(X) is isomorphic to the field k(x1 , .., xn ) of rational
functions in n-variables. We saw in the Example 6.7.2 that the cusp is a rational curve. In the next section,
we will encounter some non-rational varieties.

We close with some additional exercises on this section.

Exercise 6.7.5. Let X be an irreducible algebraic set. Define the system of neighbourhoods of x by
U(x) = {U ⊂ X : U is open and contains x}
Show that:
OX,x = ∪U ∈U (x) O(U )
where the union is taken inside the function field k(X) of X. Hence show that if there is a system of ring
homomorphisms into a ring R:
{fU : O(U ) → R}U ∈U (x)
satisfying fV ◦ ρU V = fU for all U, V ∈ U(x) with V ⊂ U , then there exists a unique homomorphism
fx : OX,x → R satisfying fx ◦ ρU,x = ρU for all U ∈ U(x). (This universal property, by the way, is the definition
of the direct limit, and shows that
OX,x = lim O(U )
→
and OX.x is therefore the stalk of the structure sheaf OX .)

Exercise 6.7.6. Show that the cubic node X := V (Y 2 − X 2 (X − 1)) ⊂ A2 (k) is birationally equivalent to
A1 (k) (i.e. is a rational curve). Determine the largest Zariski open subsets U ⊂ X and V ⊂ A1 (k) which
satisfy OX (U ) ' OA1 (k) (V ).

Exercise 6.7.7. Show that an irreducible plane conic is a rational curve. (See §1.1. Note that an irreducible
plane conic is defined to be V (f ) ⊂ A2 (k) where f ∈ k[X, Y ] is an irreducible polynomial of degree 2.)
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 57

7. A brief detour into plane curves

7.1. Generalities. In this section, let k = C, just to be specific, though much of what we say holds good for
any algebraically closed field.
Definition 7.1.1. A plane algebraic curve is just V (f ) for f a polynomial in k[X, Y ]. We say it is irreducible
if X is an irreducible algebraic set ⇔ the ideal hf i is prime ⇔ the polynomial f (X, Y ) is irreducible in k[X, Y ].

Exercise 7.1.2. Let X be an irreducible plane algebraic curve, and let k(X) denote its function field. Show
that the transcendence degree tr degk k(X) = 1.

The cases of deg f = 1 (lines) and deg f = 2 (conics, also called quadrics) are easier, and first courses on
analytic geometry study these in detail. Among the cubic plane curves, those of particular interest are of the
form V (Y 2 − f (X)), where f is a cubic polynomial. Later on, we shall be able to classify all planar cubics.

Right now it is useful to look at such curves with f (X) ∈ k[X] a general polynomial, and then specialise to
some interesting cases of f a cubic polynomial.

Proposition 7.1.3. Let f (X) ∈ k[X] be a polynomial. Let Z denote the plane curve V (Y 2 − f (X)) in A2 (k).
Then:

(i): f is not a square in k[X], (i.e. there does not exist g(X) ∈ k[X] such that f (X) = g(X)2 ), if and only
if hY 2 − f (X)i is a prime ideal in k[X, Y ]. In this event, the curve Z is irreducible, and its coordinate
ring is k[Z] = hY k[X,Y ]
2 −f (X)i , an integral domain.

(ii): Assume f is not a square in k[X], and thus by (i), B := k[Z] = hY k[X,Y ]
2 −f (X)i is a domain. Then B is

integrally closed (in its quotient field) iff f is square free (i.e. all the roots of f occur with multiplicity
one).

Proof: We note that A = k[X] is a domain, and since the ideal a := hY 2 − f i in A[Y ] = k[X, Y ] is generated
by a degree 2 polynomial in Y , the natural map A → A[Y ] → A[Y ]/a is an injective map. That is, the natural
map:
k[X, Y ]
k[X] →
a
is an inclusion.

Let B := k[X,Ya
]
, and let x (resp. y) denote the images of X (resp. Y ) in B. Then y 2 = f (x) in B. If
f (X) = g(X) , then we have (y − g(x))(y + g(x)) = y 2 − g(x)2 = y 2 − f (x) = 0. However, y + g(x) and y − g(x)
2

are 6= 0, since Y − g and Y + g are degree 1 polynomials in A[Y ], and cannot be divisible by Y 2 − f , which is
of degree 2. Thus B has zero-divisors, and is not a domain. This proves the “if” part of (i)

For the “only if” part, assume that f (X) is not a square in A = k[X]. WeP claim that every element of B
s
can be uniquely written as α(x) + β(x)y. By writing a polynomial F (X, Y ) as i=1 ai (X)Y i , and noting that
2
y = f (x), one sees that F (x, y) = α(x) + β(x)y in B. To see uniqueness, note that α(x) + β(x)y = 0 would
imply that:
β 2 y 2 = β 2 (x)f (x) = α2 (x)
Since the map k[X] → B mapping X to x is injective, this implies that β 2 (X)f (X) = α2 (X) in A = k[X].
This means that α 2
β is in k(X), and integral over k[X] (satisfying the monic polynomial T − f ). Since A = k[X]
α(X)
is a UFD, it is integrally closed in k(X), by the Proposition 4.2.8. Thus β(X) = g(X) ∈ k[X], and we have
f (X) = g(X)2 , contradicting the hypothesis on f .

Now if
(α(x) + β(x)y)(γ(x) + δ(x)y) = (α(x)γ(x) + β(x)δ(x)f (x)) + (α(x)δ(x) + β(x)γ(x))y = 0
58 VISHWAMBHAR PATI

then we have (α(x)γ(x) + β(x)δ(x)f (x)) = 0 = (α(x)δ(x) + β(x)γ(x)). Since k[X] → B is an inclusion, this
shows that the above two relations hold with x replaced by X. Multiplying the first of these equations by
δ(X), and the second one by γ(X), we find that β(X)(δ(X)2 f − γ(X)2 ) = 0 in k[X]. If the second factor is
0, then we will have (as in the last paragraph) that f (X) a square in k[X], a contradiction. If the first factor
β(X) = 0, then we will have α(X)δ(X) = α(X)γ(X) = 0. Which means that either α(X) = 0, or both γ(X)
and δ(X) = 0, in which case we have either α(x) + β(x)y = 0 or γ(x) + δ(x)y = 0, and thus B is an integral
domain. This proves (i)

Assume that f is not a square, and hence B is an integral domain. To see (ii), first assume that f (X) has
a repeated root, and by translating the X-variable to X − α if necessary, we may assume that 0 is a repeated
root. That is, f (X) = X 2 g(X). Then since y 2 = f (x) = x2 g(x) in B, we have that the rational function
y 2
x ∈ Q(B) (the quotient field of B) satisfies the monic polynomial T − g(X) over k[X], and hence is integral
over A = k[X]. Hence it is integral over B. On the other hand, xy ∈ B would imply that there is some
polynomial F (X, Y ) such that Y − XF (X, Y ) is divisible by Y 2 − f (X), which is again impossible by looking
at the Y -degree. Thus xy 6∈ B, and B is not integrally closed. This proves the “if” part of (ii).

Now suppose f is square free. It is easy to see that the quotient field of B is the field Q(B) = k(x)[y], where
y satisfies the monic polynomial y 2 − f (x). For some α(x), β(x) ∈ k(x), let the element α(x) + β(x)y ∈ Q(B),
be an element integral over B. Since B is integral over A, it will follow that this element is also integral over
k[x]. Now there is an automorphism of Q(B) defined by x 7→ x, y 7→ −y, which leaves k(x), and hence
k[x] fixed. Thus a monic relation for α(x) + β(x)y over k[x] would also imply the same monic relation for
α(x) − β(x)y over k[x]. So both α(x) + β(x)y and α(x) − β(x)y are integral over k[X]. Since we are assuming
k = C (or more generally, if k has characteristic 6= 2), we have α(x) and β(x)y are integral over k[x]. Since k[x]
is integrally closed, α(x) ∈ k[x]. Furthermore, β(x)y integral over k[x] implies β 2 (x)y 2 = β 2 (x)f (x) is integral
over k[X]. Write β(x) = p(x)/q(x), with p(X) and q(X) relatively prime. Then we have p2 (x)f (x)/q 2 (x) is
integral over k[x]. Again, by Proposition 4.2.8, this implies that p2 (X)f (X)/q 2 (X) is in k[X]. Thus q 2 (X)
divides p2 (X)f (X). If q(X) is not a constant polynomial, a linear factor of q(X) does not divide p(X) since
p, q are coprime. Thus this factor, and its square will have to divide f , contradicting that f is square free.
This proves the “only if” part, and the proposition follows. 2.

7.2. Some Cubic Plane Curves.

Notation : 7.2.1. From now on, we shall use lower case letters x, y, z etc. for variables in polynomial rings,
and also the same lower case letters for their images in coordinate rings. Upper case X, Y , Z etc. will be used
for algebraic sets or varieties.

Example 7.2.2 (The Cubic Cusp). This curve has already been discussed in the Example 6.7.2 above. Here
we have f (x) = x3 , the most degenerate cubic polynomial with one root of multiplicity 3. By (i) of Proposition
7.1.3 above, since f (x) = x3 is not a square in k[x], the ideal hy 2 − x3 i is a prime ideal in k[x, y], and so the
plane curve X := V (y 2 − x3 ) is irreducible. Also since x3 is not square free, the coordinate ring k[X] is not
integrally closed (in its quotient field k(X)). The proof of (ii) of the proposition above shows that xy ∈ k(X) is
integral over k[X] (since it satisfies the monic equation t2 − x = 0), but xy 6∈ k[X]. Recall that by the discussion
in the Example 6.7.2, the function field k(X) is isomorphic to that of A1 (k), i.e. k(t), the field of rational
functions in one variable, and thus the cubic cusp is a rational curve.

The reader may check that the map φ : A1 (k) → X defined in Example 6.7.2 is a bijection. In fact, it is
therefore a homeomorphism from A1 (k) to X, because the Zariski topologies on each of these is the cofinite
topology. However, it is not an isomorphism of curves because, as we saw in Example 6.7.2, and there can be
no k-algebra isomorphism between X and A1 (k) because the coordinate ring k[X] is not a UFD, whereas the
coordinate ring of A1 (k) is a polynomial ring in one variable, and hence a UFD.
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 59

Example 7.2.3 (The Cubic Node). This is defined as X = V (y 2 − x2 (x − 1)) inside A2 (k). In this situation,
the polynomial f (x) = x2 (x − 1) is slightly less degenerate, it has one root of multiplicity 2. Again this curve
is a rational curve. The morphism φ, from A1 (k) to X defined by the homomorphism:
k[x, y]
φ : k[X] = → k[t]
hy 2 − x2 (x − 1)i
x 7→ 1 + t2 ; y 7→ t(1 + t2 )
Again, for the same reason as in the example of the cusp above, is not an isomorphism. Since k[X] is not
integrally closed in its quotient field k(X) by (ii) of the Proposition 7.1.3 (the polynomial f (x) = x2 (x − 1) is
not square-free), it cannot be isomorphic to the coordinate ring of A1 (k). Specifically, the rational function xy
becomes integral over k[X] in its quotient field, but does not lie in k[X].

The example of f with all three roots distinct will be studied in the next subsection.

7.3. Smooth points and singular points. We begin with an algebraic notion.
Definition 7.3.1. Recall from the Definition 4.2.17 that an integral domain which is integrally closed (=inte-
grally closed in its quotient field) is called a normal domain (e.g. UFD’s are normal domains, by Proposition
4.2.8). An affine algebraic variety whose coordinate ring is normal is called a normal variety.

Example 7.3.2. By (ii) of Proposition 7.1.3, the curve X = V (y 2 − f (x)) in A2 (k) is a normal curve iff f is
square free in k[X]. As we saw above, the cubic cusp and cubic node are not normal varieties. Affine spaces
An (k), for all n, are normal varieties, since their (polynomial) coordinate-rings are UFD’s and hence normal
domains.
We shall soon investigate the geometric manifestation of normality for curves. We begin with a definition
which may be familiar from several-variable calculus.

Definition 7.3.3. A point (a, b) on a plane curve X (i.e. X = V (f ), where f (x, y) ∈ k[x, y] is a non-constant
polynomial) is said to be a smooth point or regular point if the vector:
 
∂f ∂f
gradf (a, b) := (a, b), (a, b) 6= (0, 0)
∂x ∂y
A point (a, b) ∈ X is called a singular point if it is not a smooth point. A plane curve is called smooth if every
point on it is smooth, and a singular curve if it is not a smooth curve.

The definition of a smooth point above corresponds to the geometric notion that at a smooth point (a, b) there
is a well-defined tangent line, viz., the space of vectors (v1 , v2 ) ∈ k 2 which satisfy : ∂f ∂f
∂x (a, b)v1 + ∂y (a, b)v2 = 0
is a line. It is easily checked that in the last two examples of the cubic cusp and node, that the origin (0, 0) is
the only singular point. The next proposition is another one that relates this geometric concept of smoothness
to a purely algebraic one. We prove it at present only for plane curves, but it is true in much greater generality,
as we shall see later.

Proposition 7.3.4. If X is a smooth irreducible plane algebraic curve, then its coordinate ring k[X] is a
normal domain.

Before proving this proposition, let us prove a couple of lemmas, which are very useful in general.

Lemma 7.3.5. Let A be an integral domain. Then the following are equivalent:
(i): A is a normal domain.

(ii): Ap is a normal domain for every prime ideal p in A

60 VISHWAMBHAR PATI

(iii): Am is a normal domain for every maximal ideal m in A

Proof: We first remark that the quotient field of any localisation Ap is just the quotient field of A. Let us call
it K.
a
(i)⇒(ii) Let y = b be an element of K which is integral over Ap . Then there is a monic relation:
a1 an
y n + y n−1 + ..... = 0; ai ∈ A, si ∈ A \ p
s1 sn
Multiply this equation by (s1 s2 ...sn )n , to get :
(s1 s2 ...sn y)n + b1 (s1 s2 ...sn y)n−1 + ....bn = 0
where bi := ai si1 .....sii−1 ...sin ∈ A for i = 1, 2, ..n. This is a monic relation for s1 s2 ...sn y over A, which is
integrally closed, so s1 s2 ...sn y = a for some a ∈ A. Thus y = s1 s2a..sn ∈ Ap since s1 s2 ...sn ∈ A \ p. Thus Ap is
integrally closed.

(ii)⇒(iii) clearly, since every maximal ideal is prime.

(iii)⇒(i) Let y ∈ K be integral over A. Since A ⊂ Am for all maximal ideals m, we have y is integral over
Am for all maximal ideals m. By hypothesis (iii) all these localisations are integrally closed in K, so y ∈ Am
for all maximal ideals m, i.e.,
\
y∈ Am
m∈SpmA

By the Lemma 6.4.19, we have y ∈ A, and A is integrally closed. This proves the proposition. 2

Proposition 7.3.6 (Nakayama’s Lemma I). Let a ⊂ A be an ideal, and M a finitely generated A-module.
Suppose that aM = M . Then there exists an a ∈ a such that 1 + a ∈ Ann M .

Proof: Let {xi }ni=1 be a set of generators for M . By hypothesis, for each i = 1, .., n, there are elements aij ∈ a
such that
n
X
xi = aij xj
j=1

That is, the matrix [δij − aij ] acts as the zero operator on each xi , and hence all of M . That is,
(IM − Λ)x = 0
for all x ∈ M , where Λ = [aij ]. Multiplying by the adjoint of IM − Λ, we have by Cramer’s rule in any
commutative ring that det(I − Λ)x = 0 for all x ∈ M . That is, det(IM − Λ) ∈ Ann M . Since A has entries in
the ideal a, it follows that det(IM − Λ) = 1 + a for some a ∈ a. The lemma follows. 2.

This has some interesting consequences for local rings.

Corollary 7.3.7 (Nakayama’s Lemma II). Let A be a local ring, with unique maximal ideal m. Then:

(i): If M is a finitely generated A-module with mM = M , then M = {0}.

(ii): (Krull’s Theorem) If A is a Noetherian local ring, then we have:

∩∞ n
n=0 m = {0}

(iii): If M is a finitely generated module, and there are elements {xi }ni=1 whose images xi generate M/mM ,
then {xi }ni=1 generate M .
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 61

Proof: By the Proposition 7.3.6 above, there exists a a ∈ m with (1 + a) ∈ Ann M . But since A is a local
ring, 1 + a is a unit in A. Thus 1 ∈ Ann M , and hence M = {0}. This proves (i).

If A is Noetherian, the ideal M := ∩∞ n

n=0 m is a finitely generated A-module. Also, by definition, it follows
that mM = M . By (i) above it is the zero ideal, proving (ii).

f := M/N . Since all the xi generate M/mM , it

Let N be the submodule of M generated by {xi }ni=1 , and let M
follows that N + mM = M . Hence every element x ∈ M can be written as ay + n, where a ∈ m, n ∈ N, y ∈ M .
That is Mf = mM f. By (i) above, M
f = {0} and M = N . This proves (iii) and the corollary. 2

Remark 7.3.8. (i) of the Proposition 7.3.7 above is false if one drops the finite generation assumption on
M . For example, if we take A = k[[X]] the power series ring, and M = AX (the localisation S −1 A at the
multiplicative set S = {1, X, X 2 , ..., X n , ..}) which is not finitely generated, then clearly we have XM = M for
all r ≥. Hence mM = M for the unique maximal ideal m = hXi ⊂ A. However M 6= 0.

Now we can prove the Proposition 7.3.4.

Proof of the Proposition 7.3.4:

We shall prove that whenever m is a smooth point of X, the local ring Am for the coordinate ring A = k[X]
is a unique factorisation domain, and hence by Propositions 4.2.8 and (iii) of the Lemma 7.3.5 above, we will
be done. In fact, we will show below that Am is a PID.

By translating the origin if necessary (which is an automorphism of A2 (k), i.e. of k[x, y]), we may assume
that the maximal ideal m is the origin (0, 0). Let f (x, y) be the irreducible polynomial with X = V (f ), so that
A = k[X] = k[x, y]/hf i. We also denote the images of the coordinate functions x, y ∈ k[x, y] in the integral
domain A by x, y respectively. By hypothesis, the point (0, 0) lies on X, and is a smooth point on it, so that
f (0, 0) = 0 and at least one of the partial derivatives fx (0, 0) = ∂f ∂f
∂x (0, 0), fy (0, 0) = ∂y (0, 0) are non-zero. To
be specific, assume that fy (0, 0) 6= 0.

In the expression for f (x, y) in the polynomial ring k[x, y], we can separate out the terms not involving y,
and the terms involving y. The first part maybe written as xg(x) (since f has no constant term), and the
second one as −yh(x, y) where h(0, 0) 6= 0, by the fact that fy (0, 0) 6= 0. Thus in the polynomial ring k[x, y],
f (x, y) = xg(x) − yh(x, y).

In the domain A, since f (x, y) = 0, we have:

h(x, y)y = xg(x)
/ m and is therefore a unit. Multiplying by h−1 on both sides
In the local ring Am , since h(0, 0) 6= 0, h(x, y) ∈
of the equation above, we have:
y = xr(x, y)
where r(x, y) ∈ Am . Since x, y generate the maximal ideal corresponding to (0, 0) in A, it follows that mAm
is generated by x, y. Since y = xr(x, y), it follows that the unique maximal ideal mAm is a principal ideal,
generated by x. To simplify notation, we will write m instead of mAm for the unique maximal ideal in the local
ring Am

Claim: Every non-zero element a(x, y) ∈ Am can be written uniquely as :

a(x, y) = xk v(x, y)
where v(x, y) ∈ Am is a unit. The integer k ≥ 0 is called the order of zero of a.

Proof of the Claim: If a(x, y) is a unit, we have k = 0 and a(x, y) = v(x, y), and we are done. If not,
then the ideal ha(x, y)i ⊂ Am is a proper ideal, and since Am has a unique maximal ideal m, it follows that
ha(x, y)i ⊂ m. Since ∩∞ k
k=0 m = {0} by (ii) of the Corollary 7.3.7, and a(x, y) 6= 0 by hypothesis, there is a
k ≥ 1 such that
a(x, y) ∈ mk ; a(x, y) 6∈ mk+1
62 VISHWAMBHAR PATI

Since m = hxi, as we saw above, mk = hxk i. a(x, y) ∈ mk implies a(x, y) = xk v(x, y), for some v(x, y) ∈ Am .
Since a(x, y) 6∈ mk+1 , it follows that v(x, y) 6∈ m. Thus v(x, y) is a unit, since Am is a local ring. This proves
the claim. 2

To get back to the proof of the proposition, i.e.,that Am is a UFD, we note that every element, by the claim
above, is uniquely expressible as xk u where u is a unit. Thus if a is any ideal, and since Am is Noetherian
(because it is the localisation of a Noetherian ring A), we have a = ha1 , .., ar i. But each aj = xkj vj , where vj
is a unit, so a = hxk i, where k = minj kj . That is, a is a principal ideal, and Am is a PID, and hence a UFD
by Proposition 2.2.12. This completes the proof of the proposition. 2

There is also a converse to the above Proposition 7.3.4. We will need the Lemma 5.3.18 about integral
extensions.

Proposition 7.3.9. Let X ⊂ A2 (k) be an irreducible plane curve, and let x = (a, b) ∈ X be a point on X,
associated with the maximal ideal m ⊂ k[X]. Then if the localisation k[X]m = OX,x is a normal domain, x is
a smooth point on X.

Proof: Let X = V (f ), where f (x, y) ∈ k[x, y] is an irreducible polynomial. We denote the coordinate ring
k[X] by A. We may assume, by translating coordinates, that x = (a, b) = (0, 0) ∈ X, so that f (0, 0) = 0, and
so f has no constant term. To show that (0, 0) is a smooth point, by definition, means to show that f (x, y)
has a non-vanishing linear term. That is, in the polynomial ring k[x, y], we have

f (x, y) = αx + βy + (terms of order ≥ 2)

where (α, β) 6= (0, 0). Now (α, β) = (0, 0) if and only if hf (x, y)i ⊂ n2 , where n = hx, yi ⊂ k[x, y] denotes the
maximal ideal in the polynomial ring k[x, y] corresponding to (0, 0). Since A = k[x, y]/hf i, hf i ⊂ n2 implies
that the (k[x, y]/n)-module (k-vector space) n/n2 maps isomorphically into the A/m-module (k-vector space)
m/m2 . But this will imply that the k-vector space m/m2 is two dimensional. Since mAm , the maximal ideal
in the local ring Am = OX,x , is generated by x := x/1, y := y/1 as an Am module, it follows that x and y
generate mAm /m2 Am as an Am /mAm = k module, so we have
mAm
dimk ≤2
(mAm )2
On the other hand, we claim that m/m2 injects into (mAm /m2 Am ). For this, it is enough to prove that
m ∩ (m2 Am ) = m2 .

Denote a := m ∩ (m2 Am ). Then clearly m2 ⊂ a ⊂ m. Since a ⊃ m2 , we have aAm ⊃ m2 Am . Also since

a ⊂ m2 Am , we have aAm ⊂ m2 Am . So aAm = m2 Am . If we let S = A \ m, since localisation preserves exact
sequences, we have S −1 (a/m2 ) = S −1 a/S −1 m2 = aAm /m2 Am = 0. In particular S ∩ Ann(a/m2 ) 6= φ. But since
ma ⊂ m2 , we have m ⊂ Ann(a/m2 ). Since S = A \ m intersects this annihilator, it follows that the Ann(a/m2 )
is an ideal strictly containing m and is therefore all of A. That is, a = m2 . This proves our claim, and hence
mAm 2
we have that dimk (mA m)
2 = 2 if (0, 0) is a singular point of X. This k-vector space mAm /(mAm ) , is called

the Zariski cotangent space to X at x, and what we have shown above is that at a singular point x of a plane
irreducible curve, the Zariski cotangent space is 2-dimensional. Thus it is enough to show the

Claim 0: Am integrally closed implies that

mAm
dimk =1
(mAm )2
We will prove a sequence of further claims, and the above Claim 0 will follow in the end.

The field k being algebraically closed, is an infinite field, and by applying a linear change of coordinates, we
may assume that (see the Noether Normalisation Lemma 4.2.24)
k[x] ⊂ A
is an integral extension (Note that x, y are irreducible in k[x, y], so hf i is properly contained in n = hx, yi, so
A cannot be k! Also f (x, y) = 0 shows that y is algebraic over k(x), so that the transcendence degree of the
function field Q(A) over k must be equal to that of k(x) over k, i.e. 1.)
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 63

Claim 1: Every non-zero prime ideal in A is a maximal ideal.

Since the polynomial ring k[x] is a PID, every non-zero prime ideal in k[x] is maximal. Let q be a non-zero
prime ideal in A. Since A is a domain and k[x] ⊂ A is an integral extension, it follows by Lemma 5.3.18 above
that p = q ∩ k[x] is a non-zero prime ideal in k[x]. Thus p is a maximal ideal in k[x], and k[x]/p is a field. Now
A/q is a domain which is an integral extension of this field, and from 4.2.26 it follows that A/q is also a field.
Hence q is a maximal ideal in A.

Claim 2: The only non-zero prime ideal in the local ring Am is its unique maximal ideal mAm .

If q is a non-zero prime ideal in Am , then clearly q ⊂ mAm , because the latter is the only maximal ideal in
Am . Then p = q ∩ A is a prime ideal in A. Since there exists 0 6= a/s ∈ q, it follows that a 6= 0 is in p, so that p
is a non-zero prime ideal. By Claim 1 above, p is maximal in A. Also q ⊂ mAm implies that p ⊂ A ∩ mAm = m.
The maximality of p implies p = m. Thus q ⊃ p implies q ⊃ pAm = mAm , and hence the Claim 2 follows.

Claim 3: Let R be a Noetherian integral domain which is a local ring, and a normal domain. Assume every
non-zero prime ideal of R is maximal (viz R is a “1-dimensional” ring). Then the unique maximal ideal n of
R is a principal ideal.

Let a ∈ n be any non-zero element. The radical of the proper non-zero ideal hai is the intersection of all
prime ideals in R containing hai. But by the hypotheses on R, every non-zero prime ideal in R is maximal,
and hence equal to n. Thus we have: p
hai = n
Since R is Noetherian, n = hx1 , .., xm i, and by the above we must have xm i
i
∈ hai, for each generator xi , for
r
some mi ≥ 1. From this it follows that n ⊂ hai, for all r ≥ m(maxi mi ). By Nakayama it easily follows that
nk is strictly contained in nk+1 for each k, so there exists an r such that nr ⊂ hai but nr−1 6⊂ hai. If r = 1, we
are done, so assume r ≥ 2. Choose an element b ∈ nr−1 \ hai, and consider the element x = a/b ∈ K := Q(R).
Since b 6∈ hai, x−1 = b/a 6∈ R. Since R is integrally closed in K, we have x−1 is not integral over R. If
x−1 n ⊂ n, then n would be a faithful R[x−1 ] module which is finitely generated as an R-module, and by the
characterisation (iv) of Proposition 4.2.11, we would have x−1 integral over R, a contradiction to the last line.
Thus x−1 n 6⊂ n. Since b ∈ nr−1 , it follows that bn ⊂ nr ⊂ hai, so that x−1 n = (b/a)n ⊂ R. Also x−1 n is
obviously an ideal of R, and since it is not contained in n, must be all of R. So x−1 n = R, implying that
n = Rx = hxi, proving our Claim 3 that n is a principal ideal.

By Claim 2, our Claim 3 applies to R = Am , and hence its unique maximal ideal mAm is generated by a
single element. The image of this generator in mAm /(mAm )2 generates it over Am /mAm = k. That is, the
Zariski cotangent space is a 1-dimensional k-vector space, and Claim 0, and hence the proposition follow. 2

Corollary 7.3.10. An irreducible plane curve is smooth iff its coordinate ring is a normal domain.

Proof: Immediate from 7.3.4, 7.3.9 and 7.3.5 above. 2

Remark 7.3.11. In all dimensions it is true that the coordinate ring of a smooth affine variety is a normal
domain, viz., that smooth affine varieties are normal varieties. The converse is not true in dimensions ≥ 2.
However, it is true in all dimensions that the “singular set” of a normal affine variety X is of codimension ≥ 2
in X, which is why it is empty if dim X = 1, i.e. X is a curve.

Corollary 7.3.12. The coordinate ring of a singular irreducible curve is not a UFD.

Proof: If the coordinate ring were a UFD, it would be integrally closed in its quotient field, by the Proposition
4.2.8, and by 7.3.10 above, X would be smooth. 2

The next question which one might naturally ask if the coordinate ring k[X] is itself a UFD for a smooth
irreducible curve. This is not true. Unlike normality of a domain, which is true iff each localisation is a normal
64 VISHWAMBHAR PATI

domain, it is possible for a domain not to be a UFD even if all its localisations are UFD’s. This is illustrated
in the following example:

Example 7.3.13 (An elliptic curve). Define the plane curve X = V (y 2 − x(x2 − 1)) ⊂ A2 (k). In this example,
it is important for k not to be of characteristic 2. It is easily checked that this curve is smooth, and that its
coordinate ring k[X] is integrally closed because x(x2 − 1) is square free, by (ii) of the Proposition 7.1.3.

We assert that the coordinate ring k[X], is not a UFD. The proof is divided into a series of claims.

Let x, y also denote the images of x, y ∈ k[x, y] in k[X] = k[x, y]/hy 2 − x(x2 − 1)i, so that y 2 = x(x2 − 1) in
k[X]. Note that x 7→ x and y 7→ −y defines an involution (=an automorphism of order 2) of the polynomial
ring k[x, y] which preserves the polynomial f (x, y) := y 2 − x(x2 − 1). Thus, it descends to an involution:

σ : k[X] → k[X]
x 7→ x
y 7→ −y

of the coordinate ring k[X].

Claim 1: k[x] is precisely the subalgebra of k[X] fixed by σ.

It is clear that k[x] is fixed by σ. For the converse, note that since y 2 = x(x2 − 1), any element b(x, y) of
k[X] can be written uniquely as (see the proof of (i) in Proposition 7.1.3 above). b(x, y) = a1 (x) + ya2 (x). If
σ(b) = b, then:
1 1
b = (b + σ(b)) = (a1 (x) + ya2 (x) + a1 (x) − ya2 (x)) = a1 (x)
2 2
which proves the claim 1 (dividing by 2 requires characteristic of k is 6= 2 !)

Claim 2: The subalgebra k[x] of k[X] is a polynomial algebra.

The proof of this is contained in the proof of (i) in Proposition 7.1.3.

We define a map N : k[X] → k[x], called the norm by N (g) = gσ(g). Clearly N (gh) = N (g)N (h) for all
g, h ∈ k[C]. (We verify that gσ(g) is fixed by σ, so that by the fact proved above that the fixed set of σ is k[x],
we have N (g) ∈ k[x] for all g ∈ k[X]). Also N (1) = 1.

Claim 3: If N (g) ∈ k ∗ for some g(x, y) ∈ k[X], then g ∈ k ∗ .

Write, as above, g(x, y) = a1 (x) + ya2 (x) ∈ k[X]. Then we have:

N (g) = a21 − y 2 a22 = a21 − x(x2 − 1)a22
If this polynomial in x is a scalar λ 6= 0 in k, then we would have x(x2 − 1)a22 = −λ + a21 . If a2 (x) is not
the zero polynomial, in view of claim 2 above, the left hand side, and hence the right hand side would be a
polynomial of degree 2k + 3, which is absurd, since the right hand side is a polynomial of even degree. Thus
a2 = 0, and g(x, y) = a1 (x), and a1 (x)2 = λ, and again by claim 2 above, this forces a1 (x) to be a constant
polynomial in k ∗ .

Claim 4: The units in k[X] are precisely the elements of k ∗ .

Clearly all elements of k ∗ = k \ {0} are units. On the other hand, if gh = 1 for some g, h ∈ k[C], then
N (g)N (h) = 1. But N (g), N (h) ∈ k[x], which was a polynomial ring, so both N (g) and N (h) are degree 0
polynomials, i.e. scalars, and also non-zero scalars, since their product is 1. This means, by claim 3 above,
that g, h are both elements of k, and also non-zero, since their norms are non-zero.

Claim 5: The elements x, y ∈ k[X] are irreducible.

 LECTURES ON BASIC ALGEBRAIC GEOMETRY 65

First note that, by claims 2 and 4, x and y are not units. First we prove that x is irreducible. Suppose
x = gh, for g, h ∈ k[X]. We have to show that one of g, h is a unit, viz. an element of k ∗ . x = gh implies
N (x) = x2 = N (g)N (h). Since N (g), N (h) ∈ k[x], which is a polynomial ring (and therefore a UFD), by claim
2 above, it follows that one of N (g), N (h) is either a unit, or a unit times x. Say it is N (g). Then in the
first case, by claim 3 above, g is itself a unit, and we are done. In the second case we would have, on writing
g = a1 + ya2 , that N (g) = λx with λ 6= 0. This implies that λx − a21 = −x(x2 − 1)a22 in k[x]. Again, the left
hand side can either be a polynomial of degree 1 or degree 2k, whereas the right hand side is a polynomial of
degree 0 or degree 2m + 3. This is impossible by claim 2. Thus N (g) cannot be λx (λ 6= 0) for any g in k[X].
This proves that x is irreducible.

To see that y is irreducible, apply norms to both sides of y = gh to obtain that N (g)N (h) = −x(x2 − 1). If
one of N (g), N (h) were of degree 0, we would have that the corresponding g or h is a unit, by claim 4, and we
are done. Since −x(x2 − 1) is a polynomial of degree 3, we may assume that one of N (g), N (h) is of degree
one. Say it is N (g) = a21 − x(x2 − 1)a22 where, as usual, we have written g(x, y) = a1 (x) + ya2 (x). This means
N (g) = λx, λ(x − 1), or λ(x + 1), for some λ ∈ k. If N (g) = λx, we have a21 − x(x2 − 1)a22 , a polynomial of
degree 2k or 2k + 3 equal to λx, a polynomial of degree 1. Thus λ = 0, a2 = 0, and this implies a1 = 0, a
contradiction. If N (g) = λ(x − 1), or λ(x + 1), the same reasoning applies. This proves that y is irreducible,
and the claim.

Claim 6: k[X] is not a UFD.

Since we have y 2 = x(x2 − 1) in k[X], we have x divides y 2 . If k[X] were a UFD, x being irreducible in
k[X], would therefore be prime, and thus divide y. But y is also irreducible by claim 5, so y = ax where a is
a unit, and thus an element in k ∗ . Taking norms, we would have that −x(x2 − 1) = ax2 , for a 6= 0 in k, a
relation that is impossible in the polynomial ring k[x] (claim 2). This proves our claim. 2

Exercise 7.3.14. Show that k[X] is the integral closure of k[x] inside the quotient field k(X) of k[X].

Remark 7.3.15. Integrally closed domains of dimension 1 (i.e. in which every non-zero prime ideal is maximal)
are called Dedekind domains. The Propositions 7.3.4 and 7.3.9 above assert that an irreducible plane curve
X is smooth iff its coordinate ring k[X] is a Dedekind
 √  domain.√Dedekind domains also arise as the rings of
integers inside algebraic number fields, e.g. Z 1+2 5 inside Q( 5). There is a unique factorisation available
for Dedekind domains, viz. that every non-zero ideal factorises uniquely into a product of prime ideals. See
Cor.9.4 in Chapter 9 of Atiyah-Macdonald’s Introduction to Commutative Algebra for a proof. So even though
the coordinate ring of, say, the elliptic curve V (y 2 − x(x2 − 1)) is not a UFD, it is pretty close to being a UFD.

Remark 7.3.16. The study of elliptic curves (over various fields k) is a vast and rich area of number theory,
algebraic geometry, complex analysis, representation theory, differential equations and differential geometry.
Elliptic curves are a meeting ground for nearly all of mathematics! For example, A. Wiles’ proof of Fermat’s
Conjecture is a statement about elliptic curves. It turns out that the projective closure (we will define this
later) of an elliptic curve acquires the structure of an abelian group, with the group operation getting defined
in purely geometric terms. (Incidentally, the involution σ constructed above in Example 7.3 corresponds to
taking the inverse in this group). The connections between this group structure and the geometry of the elliptic
curve leads to far-reaching number-theoretic consequences. For more on this topic, see books on elliptic curves,
e.g. the ones by Silverman, or Husemoller.
66 VISHWAMBHAR PATI

8. Quasiprojective varieties
8.1. Motivation. So far, we have only dealt with closed subsets of affine space. For reasons that will emerge
in the sequel, it is important to consider “points at infinity” on the same footing as finite points. The reader
may have already encountered this in complex analysis, where the completion of C by adding a point at infinity
gives us the Riemann sphere, thus providing a unified way of looking at rational functions, zeros and poles of
meromorphic functions, counting them, etc.

To give another reason, suppose one wants to develop a theory of intersections of algebraic sets. Intuitively,
it seems clear that if one intersects an irreducible plane curve C = V (f ) ⊂ A2 (k) defined by a degree d
polynomial f (X, Y ) with, say, a line L = V (aX − Y + c), then substituting Y = aX + c into f (X, Y ) will
lead to a degree d polynomial in X, and this polynomial will have d roots (since k is taken to be algebraically
closed) counted with multiplicity. Each such solution for X will lead to a unique value of Y = aX + c, so we
should expect C to intersect L in d points (counted with multiplicity).

Of course, this expectation is belied even in the case of two lines in A2 (k). The two parallel lines V (Y )
and V (Y + 1) in the plane do not intersect at all. The reason is that they meet at “infinity”, and we don’t
see this point in the affine plane. So the trick is to add one new point at infinity for each direction in the
plane, and then it turns out that this “completion” of the affine plane is a very interesting object called the
projective plane. (Incidentally, if one completes the affine line, since there is only one direction in the affine
line, only one point at infinity is attached to k to get the projective line. If k = C, then we get the Riemann
sphere as the corresponding projective line.) Now, in this new completed space, the count of d for the number
of intersection points of a curve of degree d and a line turns out to be correct, because the intersection points
at infinity get accounted for ! In fact, one gets the beautiful theorem of Bezout which says that two distinct
irreducible algebraic curves of degrees m and n in the projective plane intersect in mn points (counted with
multiplicity).

A fundamental property of projective space (and closed subsets of projective space) that is at the back of
many of the beautiful simplications which result is a geometric property called “completeness”, which is akin to
the property of compactness in topology. It is precisely this completeness that captures the points at infinity.
We shall discuss this in a future section.

For yet another illustration, it is hard to understand all the automorphisms of affine space An (k) for n ≥
2, which is impossibly huge. Indeed, the famous Jacobian conjecture (yet unsolved) asserts that a regular
(polynomial) map of An (k) (for n ≥ 2) with everywhere non-vanishing jacobian is an isomorphism. In contrast,
the only automorphisms of projective space are linear, so it is a very “rigid” space.

The foregoing discussion hopefully motivates the notion of projective space.

8.2. Closed Subsets of Projective Space.
Definition 8.2.1. Let k be a field. We look at the space An+1 (k) \ {0}, and introduce an equivalence relation
on it as follows : (x0 , x1 , .., xn ) ∼ (y0 , y1 , .., yn ) if there exists a λ ∈ k ∗ such that yi = λxi for all i = 0, 1, .., n.
This is clearly equivalent to saying that two points in An+1 (k) \ {0} are equivalent if they lie on the same
1-dimensional subspace of An+1 (k) (same line through the origin or “direction”). The quotient space is called
n-dimensional projective space over k, and denoted Pn (k), or simply Pn if the field is understood. A point p
on Pn (k) is denoted by p = [x0 : x1 : .. : xn ] which is the equivalence class of (x0 , x1 , .., xn ). It is therefore
understood that the xi (called homogeneous coordinates) in the expression [x0 : x1 : .. : xn ] for p are (i) not all
zero and (ii) well-defined only upto a common non-zero scalar multiplying them all.

More generally, for V a k-vector space, we will write P(V ) for the projective space of V , i.e. the space of all
1- dimensional subspaces of V .

It is sometimes convenient to think of projective space Pn (k) schematically as an n-simplex. The i-th vertex
can be thought of as the point [0 : 0 : .. : 1 : ..0] (with 1 at the i-th spot), the (n − 1)-faces as the various
coordinate hyperplanes ({Xi = 0} ' Pn−1 (k)), and so on. Their formal combinatorial relations, e.g. the
intersections of hyperplanes, etc. are the same. This formal picture should not be pushed too far ! We will see
shortly how a Zariski topology gets defined on Pn (k).
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 67

Let π : An+1 (k) \ {0} → Pn (k) be the quotient map. We would like to define the natural “quotient topology”
on Pn (k).

Let us provisionally define a subset Z ⊂ Pn (k) to be a closed subset of Pn (k) iff π −1 (Z) is a closed subset
of An+1 (k) \ {0}. That is,
π −1 (Z) = (An+1 (k) \ {0}) ∩ V (a)
for some ideal a ⊂ k[X0 , ..., Xn ]. We note that π −1 (Z) ∪ {0} is a union of lines through the origin in An+1 (k),
so this will impose a condition on the ideal a, which we describe next.

For every f ∈ a such that f vanishes at (a0 , ..., an ) ∈ π −1 (Z) ∪ {0}, we P must have f (λa0 , ..., λan ) = 0 for
all λ ∈ k ∗ , since (λa0 , ..., λan ) also lies in π −1 (Z) ∪ {0}. If we write f = d fd , where fd is the homogeneous
degree d part of f , we must have :
X
λd fd (a0 , ...., an ) = 0 for all λ ∈ k ∗
d

If one considers this as a polynomial in λ, it must be the trivial polynomial (since k is algebraically closed, and
therefore infinite). That is, fd (a0 , ..., an ) = 0 for all d.

Thus the condition on a is that it contains the homogeneous components of all its elements.

Definition 8.2.2. An ideal a ⊂ k[X0 , ..., Xn ] is called a homogeneous ideal if it contains the homogeneous
components of all its elements.

Exercise 8.2.3. Show that an ideal a ⊂ k[X0 , .., Xn ] is homogeneous iff it is generated by some finite set of
homogeneous generators. Prove that the ideal hY 2 −X 3 i ⊂ k[X, Y ] is not a homogeneous ideal in k[X, Y ]. Show
that the ideal hY 2 − X 3 , 3Y 2 + X 3 i ⊂ k[X, Y ] is a homogeneous ideal, and write down a set of homogeneous
generators. (It is generally not so trivial to decide the homogeneity of an ideal by looking at some set of
generators).

Nevertheless we have the following:

Remark 8.2.4. If a ⊂ k[X0 , .., Xn ] is a homogeneous ideal generated by r elements, then it is generated by
≤ r homogeneous elements. (Note that for r = 1, i.e. a principal homogeneous ideal a = hf i, then f must be
homogeneous).

Proof: Let a = hf1 , .., fr i. Clearly the homogeneity of a implies that the set of all the homogeneous components
of the fi , namely the finite set:
S = {g : g is a homogeneous component of some fi }
is also a generating set for the ideal a. Let {gj }m
j=1 be any subset of S which is a minimal generating set for
a. We claim that m ≤ r.

For, assume to the contrary that m > r. Since both {fi }ri=1 and {gj }m
j=1 generate the ideal a, let us expand:

r
X m
X
gj = Aji fi fi = Bil gl
j=1 l=1
where [Aji ] is an (m × r) and [Bil ] an (r × m) matrix with entries in k[X0 , .., Xn ]. Thus we get thatPthe (m × m)
matrix : C = Im − AB (where Im is the identity matrix of size m) annihilates every gj . That is, l Cjl gl = 0,
for all 1 ≤ j ≤ m.

Let dj denote the degree of the homogeneous polynomial gj and assume, by relabelling
P if necessary, that
dj ≤ dk for j ≤ k. For a fixed j, the degree dk homogeneous component of the relation l Cjl gl = 0 reads as :
X (0)
X (d −dl )
Cjl gl + Cjl k gl = 0
l:dl =dk l:dl <dk
68 VISHWAMBHAR PATI

(d) (0)
where Cjl denotes the degree d homogeneous component of Cjl . If any of the Cjl ∈ k occurring in the first
term above were non-zero, we could divide by it, and thus express the corresponding gl in terms of {gj }j6=l ,
(0)
contradicting the choice of {gj }m
j=1 as a minimal set of ideal generators. Thus, Cjl = 0 for all j and all l such
(0)
that dl = dk . Since k is arbitrary, it follows that that Cjl = 0 for all j and l.

Thus Cjl (0, 0, .., 0) = 0, and we have the relation of matrices with entries in k:
Im = A(0, 0, .., 0)B(0, 0, .., 0)
where A(0, 0, .., 0) is the matrix [Aji (0, 0, .., 0)] and B(0, 0, .., 0) has a similar meaning. Since m > r, the rank
of B is ≤ r, and the rank of AB is less or equal to r, whereas Im has rank m > r, a contradiction. Thus m ≤ r.
2

As before, let V (a) denote the zero set of the homogeneous ideal a in An+1 (k). Now we can finally define
the Zariski topology on Pn (k).

Definition 8.2.5 (Projective Closed Sets). We define a (Zariski) topology on Pn (k) by declaring its closed
subsets, called projective closed sets, to be all sets of the form:
 
VPn (k) (a) := π V (a) ∩ (An+1 (k) \ {0})
where a is a homogeneous ideal in k[X0 , .., Xn ]. Note that VPn (k) (h0i) = PN (k), and VPn (k) (h1i) = φ.

Clearly we have :
VPn (k) (a) = {[a0 : ... : an ] : f (a0 , .., an ) = 0 ∀ f ∈ a}

As a matter of notational convenience, we will often drop the subscript Pn (k), and just write the projective
closed set in question as V (a) whenever it is clear from the context whether we are in An+1 (k) or Pn (k).

Exercise 8.2.6. Verify that the closed sets defined as above define a topology on Pn (k).

Note that the maximal ideal hX0 , .., Xn i ⊂ k[X0 , ..., Xn ] defines the origin in An+1 (k). Hence this homoge-
neous ideal defines the empty set in Pn (k). Indeed, for any homogeneous ideal a whose radical is the maximal
ideal hX0 , .., Xn i, we have VPn (k) (a) = φ. More precisely, we have the following:

Lemma 8.2.7. Let Is be the ideal in k[X0 , ..., Xn ] generated by all homogeneous polynomials of degreeó s.
Then for a proper homogeneous ideal a ⊂ k[X0 , ..., Xn ], VPn (k) (a) = φ iff Is ⊂ a for some s ≥ 1 (⇔ a=
hX0 , .., Xn i = I1 .)

√
Proof: If Is ⊂ a for some s ≥ 1, then Xis ∈ a for each√i = 0, 1, .., n. Thus the radical a contains the
maximal ideal hX0 , .., Xn i, and since a is a proper ideal, a 6= h1i. Thus V (a) = {0} in An+1 (k), so that
V (a) ∩ (An+1 (k) \ {0}) = φ. Hence its image under π, namely VPn (k) (a) = φ.

On the other hand, if φ = VPn (k) (a) = π(V (a) ∩ (An+1 (k) \ {0}), then V (a) = φ or 0. The former case is
√
ruled out since a is a proper ideal. In the latter case of V (a) = {0}, the radical a = hX0 , .., Xn i. Hence there
exists an m ≥ 0 such that Xim ∈ a for all i. Thus if we choose s = (n + 1)m, any monomial of degree ≥ s will
contain a factor of Xim for some i and hence lie in a, so that Is ⊂ a for this s. 2

Thus, we make the following definition:

Definition 8.2.8. We call a homogeneous ideal a redundant if its radical is the maximal ideal hX0 , ..., Xn i. In
particular, it it is a proper ideal. Equivalently, a is redundant if it is a proper ideal containing the ideal Is for
some s ≥ 1. For example, the ideal hY 2 − X 3 , 3Y 2 + X 3 i ⊂ k[X, Y ] of the Exercise 8.2.3 above is redundant,
for it clearly contains I4 .
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 69

Again, given a subset S ⊂ Pn (k), one can consider the ideal of all polynomials that vanish identically on
−1 −1
π (S), i.e. what we denoted I(π (S)). This is again a homogeneous ideal, by the reasoning given at the
beginning of this subsection.

Notation : 8.2.9. By abuse of language denote this homogeneous ideal as I(S).

We have the following proposition:

Proposition 8.2.10 (Homogeneous Nullstellensatz). Let k be an √ algebraically closed field. If a is any non-
redundant homogeneous ideal in k[X0 , .., Xn ], then I(V (a)) = a. If S ⊂ Pn (k), then V (I(S)) = S, the
Zariski closure of S in Pn (k). There is a 1-1 correspondence between closed subsets of Pn (k) and homogeneous
non-redundant radical ideals of k[X0 , ...., Xn ]. (We note that the homogeneous ideal h1i is non-redundant by
our definition, and defines the empty set.)

Proof:
If VPn (k) (a) is a closed set in Pn (k), with a a homogeneous non-redundant ideal in k[X0 , .., Xn ], the ideal
I(VPn (k) (a)) is the ideal of polynomials I(π −1 VPn (k) (a)), and assuming VPn (k) (a) 6= φ, any polynomial f van-
ishing on π −1 (VPn (k) (a)) has each of its homogeneous components vanishing on it as well. In particular, all the
homogeneous components of f are of degree ≥ 1. But every homogeneous polynomial in k[X0 , .., Xn ] of degree
≥ 1 vanishes at the origin in An+1 (k), so
I(VPn (k) (a)) = I(π −1 VPn (k) (a)) = I(π −1 VPn (k) (a) ∪ {0}) = I(V (a))
√
But by the affine nullstellensatz Proposition 6.1.1, it follows that the right hand ideal I(V (a)) = a. This
proves the first part of the proposition.

The second assertion is straightforward using the definition of the Zariski topology on Pn (k), and the
corresponding fact for An+1 (k).

For the last assertion, the only thing to be noted is that the radical of a homogeneous
√ ideal in k[X0 , ..., Xn ]
is also homogeneous. For, let f = Σd≤m fd be a polynomial of degree m lying in a, where a is homogeneous.
Then f r ∈ a for some r. The√ highest degree homogeneous
√ term in f r is clearly fm
r
, so by homogeneity of a, we
r
have fm ∈ a so that fm ∈ a. Thus f − fm ∈ a and is of lower degree. Induction completes the proof. 2

Exercise 8.2.11. Prove that a homogeneous ideal a ⊂ k[X0 , .., Xn ] is prime iff for every pair of homogeneous
polynomials F , G such that F G ∈ a, and F 6∈ a then G ∈ a.

Example 8.2.12 (Quadrics). Let A = [aij ] be any n × n symmetric matrix with entries in k, and QA be
the associated quadratic form, viz. QA (X) = Σi≤j aij Xi Xj , where X = (X1 , .., Xn ) ∈ An (k). The zero-
set V (hQA i) in Pn−1 (k) is called a quadric. If A is a singular matrix, it is called a degenerate quadric, and
otherwise a non-degenerate quadric. For example, a very basic such non-degenerate quadric is the Segre variety
V (X1 X4 − X2 X3 ) in P3 (k) and, as we shall see later, is isomorphic to P1 (k) × P1 (k).

Example 8.2.13 (Determinantal Projective Sets). The affine space gl(n, k) of all n × n matrices with entries
in k has the coordinate ring given by the polynomial ring in n2 variables, which we denote k[Xij ]ni,j=1 . If we let
am be the homogeneous ideal of k[Xij ]ni,j=1 generated by the determinants of all m × m minors of the matrix
[Xij ], the projective closed subset V (am ) ⊂ P(gl(n, k)) consists of all elements
[: Xij ] := [X11 : X12 : ... : Xnn ] ∈ P(gl(n, k))
which are of rank ≤ (m − 1). Homogenous ideals of the type am are called determinantal ideals. Similar
examples of closed projective subsets of P(hom(k n , k m )) defined by
{A ∈ hom(k n , k m ) : rk A ≤ p}
can be constructed, by taking the ideal generated by all (p × p) minors.
70 VISHWAMBHAR PATI

If k = R or C, one can also put the classical topology on Pn (k) by taking the quotient topology of the
classical topology on An+1 (k) \ {0}. As an exercise, the reader may wish to check that P1 (C) with the classical
topology is homeomorphic to the Riemann sphere S 2 via the homeomorphism : [z0 : z1 ] 7−→ zz01 (the point [1; 0]
in P1 (C) maps to the point ∞ in the Riemann sphere.) The inverse of this map is : z 7−→ [z : 1], with ∞ going
to [1 : 0]. Another exercise is to check that P1 (R) with its classical topology is homeomorphic to the circle S 1 .

The following is a straightforward remark, whose proof is analogous to the corresponding statements in the
affine case, and left as an exercise.
Remark 8.2.14. Define a projective closed subset of Pn (k) to be irreducible if it is not the union of two proper
projective closed subsets. Then

(i): A projective closed set X ⊂ Pn (k) is irreducible iff X = V (p) where p is a non-redundant homogeneous
prime ideal of k[X0 , ..., Xn ].

(ii): A projective closed set is a Noetherian space, and finally,

(iii): every projective closed set in Pn (k) is the irredundant union of finitely many irreducible projective
closed sets. As a corollary, we obtain that every non-redundant homogeneous radical ideal in k[X0 , .., Xn ]
is an irredundant intersection of finitely many homogeneous non-redundant prime ideals.

Exercise 8.2.15. Show that two non-empty projective closed sets V (a) and V (b) (a and b homogeneous
non-redundant ideals in k[X0 , .., Xn ]) are disjoint iff Is ⊂ a + b for some s ≥ 1, i.e. if a + b is redundant.

The projective space Pn (k) has an open covering by n + 1 open sets {Ui }ni=0 , each of which (with the induced
Zariski topology from Pn (k)) is homeomorphic to the affine space An (k) (with its Zariski topology). Indeed,
define :
Ui = {[a0 :, .., : an ] ∈ Pn (k) : ai 6= 0}
The mappings [a0 :, ...., : an ] 7→ ( a0 , .., abi , .., an ) from Ui to An (k) (where the hat denotes omission) and
ai ai ai
(x1 , ..., xn ) 7→ [x1 : .. : 1 : .. : xn ] (where 1 is inserted at the i-th place) are clearly inverses of each other. For
a homogeneous degree d polynomial F (X0 , .., Xn ) in k[X0 , .., Xn ], the zero set of F intersected with Ui is the
set : {[a0 : .... : an ] : F (a0 , .., an ) = 0, ai 6= 0}, which is the same as the set :
{[a0 : ..., : 1 :, .. : an ] : F (a0 , ..., 1, ..an ) = 0}
by factoring out the common non-zero scalar ai from all the homogeneous coordinates and adi from the ho-
mogeneous polynomial F . If we agree to call the polynomial F (X0 , ..., 1, ...Xn ) as i F D (X0 , ..., X ci , ., Xn ) the
i-th dehomogenisation of F , (in general a non-homogeneous polynomial in k[X0 , ..., X ci , ., Xn ] ) then the closed
subset VPn (k) (F ) ∩ Ui of Ui maps precisely to the zero set V (i F D ) of the non-homogeneous polynomial i F D in
An (k) under the bijection of Ui with An (k) defined above. It is trivial to check that
D D D D D
i (F G) = iFi G , i (F + G) =i F +i GD for homogeneous polynomials F, G.

Conversely, the zero set of a non-homogeneous polynomial f (X0 , ..., X ci , ., Xn ]

ci , ., Xn ) of degree d in k[X0 , .., X
d X0 Xi−1 Xi+1 Xn
goes to the zero set of the degree d homogeneous polynomial Xi f ( Xi , ..., Xi , Xi , .. Xi ). This last polynomial
is called the i-th homogenisation of f , denoted i f H in k[X0 , .., Xn ]. Since all closed sets are intersections of
closed sets defined by single polynomials, (in both Pn (k), hence Ui , and in An (k)), this shows that the maps
defined above are homeomorphisms.

In fact, we make the following:

Definition 8.2.16. Let X = V (a) be a closed subset of affine n-space An (k) defined by the ideal a. We
define the homogeneous ideal aH , called the 0-th homogenisation of a to be the ideal generated by the 0-th
homogenisations of all polynomials in a. We define the projective closure or projectivisation of X in Pn (k) to
be the projective closed set defined by the homogeneous ideal aH in k[X0 , ...., Xn ]. It is denoted X. b Similarly,
if X = V (b) is a projective closed set defined by a non-redundant homogeneous ideal b ⊂ k[X0 , ..., Xn ], then
the ideal bD of k[X1 , .., Xn ], called the 0-th dehomogenisation of b, is defined as the ideal generated by the 0-th
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 71

dehomogenisations of all the homogeneous elements of b, (which is just the ideal {f (1, X1 , .., Xn ) : f ∈ b}).
V (bD ) is called the 0-th affine piece of X. Similarly, one can define the i-th affine piece of X. It is just the
intersection of X with Ui viewed as a closed subset of An (k) after making the above identification of Ui with
An (k).

Notation : 8.2.17. For a polynomial f ∈ k[X1 , .., Xn ], f H will always denote the 0-th homogenisation 0 f H
of f , for notational convenience. Likewise, for f ∈ k[X0 , .., Xn ], f D will always mean 0 f D ∈ k[X1 , .., Xn ].

Exercise 8.2.18. Check that (f g)H = (f H )(g H ). What is (f + g)H ? If a homogeneous ideal a in k[X0 , .., Xn ]
is generated by the homogeneous polynomials {fi }m i=1 , then show that the 0-th dehomogenisation a
D
inside
D m
k[X1 , .., Xn ] is generated by {fi }i=1 . Verify that for an ideal a ∈ k[X1 , .., Xn ], the 0-th homogenisation is
given by:  
X m 
k
aH = X0 j fjH : fj ∈ a
 
j=1

Remark 8.2.19. If an ideal b in k[X1 , .., Xn ] is generated by {fi }m H

i=1 , the 0-th homogenisation b in k[X0 , .., Xn ]
H m
may not be generated by {fi }i=1 . For example, the ideal a = hX1 , X2 i in k[X1 , X2 ] is also generated by the
elements f1 = X12 + X2 and f2 = X1 . The 0-th homogenisation aH is the ideal hX1 , X2 i ⊂ k[X0 , X1 , X2 ].
On the other hand f1H = X12 + X2 X0 and f2H = X1 . It is easily checked that X2 ∈ aH cannot lie in the
ideal generated by f1H and f2H . Hence, in the expression for aH in the last Exercise 8.2.18, one cannot replace
“fj ∈ a” by “fj in a generating set for a”. This is not surprising, since homogenisation behaves badly with
respect to addition, as the last exercise reveals.

Here is a lemma about composing the processes of homogenisation and dehomogenisation.

Lemma 8.2.20.

bi , .., Xn ], then i (i f H )D = f . For an ideal a ∈ k[X1 , .., Xn ], we have (aH )D = a.

(i): If f ∈ k[X1 , ..X

(ii): If F (X0 , ..., Xn ) is a homogeneous polynomial of degree d, then (Xil )i (i F D )H = F for some l ≥ 0. In
particular, if F is not divisible by Xi , then i (i F D )H = F .

(iii): If p is a homogeneous prime ideal not containing X0 , then (pD )H = p (where the homogenisation and
dehomogenisation on the left hand side are with respect to X0 ).

Proof:
We just do it for i = 0. If f ∈ k[X1 , ..., Xn ] is of degree d, then:
 
X1 Xn
f H (X0 , ..., Xn ) = X0d f , ....,
X0 X0

Thus (f H )D (X1 , .., Xn ) = f H (1, X1 , ..., Xn ) = f (X1 , ..., Xn ), hence proving the first statement of (i). For
the second statement, note that by the Exercise 8.2.18,
 
X m 
k
aH = X0 j fjH : fj ∈ a
 
j=1

Now, by definition,
X k X X
(aH )D = h{( X0 j fjH )D : fj ∈ a}i = h{ (fjH )D : fj ∈ a}i = h{ fj : fj ∈ a}i = a
j j j
72 VISHWAMBHAR PATI

To see (ii), we note that:

   
X1 Xn X1 Xn
D H
(F ) (X0 , ...., Xn ) = X0d−l (F D ) , .., d−l
= X0 F 1, , ..,
X0 X0 X0 X0
if d is the homogeneous degree of F and d − l ≤ d is the degree of F . Multiplying both sides by X0l and using
D

the homogeneity of F yields (ii). Clearly, if F is indivisible by X0 , l has to be 0.

To see (iii), let F be a homogeneous polynomial in p, a homogeneous prime ideal. Then write F =
X0d G(X0 , ..., Xn ) where G is homogeneous and indivisible by X0 . Since p is prime and does not contain
X0 , it follows (by Exercise 8.2.11) that G ∈ p. Thus, by (ii) above, (GD )H = G belongs to (pD )H . Thus
F = X0d G ∈ (pD )H . This shows p ⊂ (pD )H . On the other hand, if F is a homogeneous element of the form
(GD )H for some homogeneous G ∈ p, then for some l ≥ 0, X0l F = (X0l )(GD )H = G, which is in p. Since
X0 is not in p, and p is prime, we have F ∈ p, proving that all the generators of (pD )H are in p, and hence
(pD )H ⊂ p. This proves (iii), and the lemma. 2

Exercise 8.2.21.
(i): Give examples to show that the hypotheses in (ii) and (iii) of the above Lemma 8.2.20 cannot be
dropped.

(ii): Let p ⊂ k[X0 , .., Xn ] be a homogeneous non-zero prime ideal with X0 6∈ p. Then pD is also prime.

Some things need to be checked to ensure that the closing up of affine varieties in projective space by
homogenisation of the corresponding ideals is a well-behaved operation. More precisely,

Proposition 8.2.22. The following facts are true of projective closures and affine pieces:

(i): The projective closure of an affine closed set is just its Zariski closure in Pn (k) considered as a subset
of U0 ⊂ Pn (k).

(ii): The homogenisation of a prime ideal is prime, so that the projective closures of irreducible affine closed
sets are also irreducible.

(iii): The i-th affine piece of an irreducible closed projective set is an irreducible affine closed set.

Proof:
By the homogeneous nullstellensatz Proposition 8.2.10, we have that the Zariski closure of any subset S ⊂
Pn (k) is given by VPn (k) (I(S)), where I(S) is the homogeneous ideal of polynomials in k[X0 , ..., Xn ] vanishing
identically on S. Thus for an affine closed set X = VAn (k) (a) defined by the radical ideal a ⊂ k[X1 , .., Xn ], the
Zariski closure of X ⊂ U0 ⊂ Pn (k) is exactly VPn (k) (b) where :
b = hF (X0 , ., Xn ) homogeneous : F (1, a1 , ., an ) = 0 ∀ (a1 , ., an ) ∈ VAn (k) (a)}i

This last condition on a homogeneous polynomial F is equivalent to saying that F D vanishes identically on
VAn (k) (a), which by the affine nullstellensatz, implies F D ∈ a, since a was assumed to be a radical ideal. This
implies (F D )H ∈ aH . By (ii) of the previous Lemma 8.2.20, b ⊂ aH . Thus VPn (k) (b) ⊃ VPn (k) (aH ). This means
X ⊃ X. b On the other hand, since X b is Zariski closed and contains X, we have X ⊂ X. b This proves (i).

To see (ii), let p be a prime ideal in k[X1 , .., Xn ]. By the Exercise 8.2.11, to establish prime-ness of pH , it is
enough to show that if F and G are homogeneous polynomials with F G ∈ pH , then either F ∈ pH or G ∈ pH .
But F G ∈ pH implies (F G)D ∈ (pH )D = p, by (i) of the Lemma 8.2.20 above. Since p is prime, one of F D , or
GD ∈ p. Say F D ∈ p. Then (F D )H ∈ pH . By (ii) of the Lemma 8.2.20 we have that F = X0l (F D )H ) ∈ p.

To see (iii), let p be a homogeneous prime ideal in k[X0 , ..., Xn ], and let X = V (p) be the corresponding
irreducible projective closed set. If X0 ∈ p, we have that V (p) ⊂ V (X0 ), so that X ∩ U0 = φ, which is clearly
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 73

irreducible. So assume X0 ∈ / p. Then by the part (ii) of the Exercise 8.2.21 the dehomogenised ideal pD
is prime. Since this last ideal is the ideal defining the 0-th affine piece of X, (iii) follows. This proves the
proposition. 2

Remark 8.2.23. Note that it is possible for X ∩ Ui to be irreducible for each i, and for X to be reducible.
For example,
X = V (X0 X1 ) = {[1 : 0], [0 : 1]}
1
is a reducible closed subset of P (k), though its intersections with the affine opens U0 and U1 are singletons,
and irreducible.

Remark 8.2.24. Except for the redundant maximal ideal hX0 , ..., Xn i, none of the other maximal ideals in
k[X0 , ..., Xn ] are homogeneous. The maximal ideals among the homogeneous ideals are the ideals whose affine
pieces are either empty, or a single point. The maximal ideal defining the point [a0 : a1 : ..1.. : an ] in Pn (k) is
the ideal :
h{Xj − aj Xi : j = 0, 1, ..n; j 6= i}i
It can be written somewhat more symmetrically as the ideal generated by the homogeneous elements Xj ak −
aj Xk with k 6= j both ranging from 0 to n.

8.3. Maps and Morphisms between Projective Sets.

Notation : 8.3.1. In the sequel, we shall often call a homogeneous polynomial of degree m an m-form. Note
that 0-forms are the elements of k. If P ∈ k[X0 , ..., Xn ] is an m-form and x = [a0 : a1 : .. : an ] is a point in
Pn (k), we write P (x) = 0 to mean that P (a0 , .., an ) = 0. This condition clearly depends only on x and not on
the representative [a0 : a1 : .. : an ] chosen for x. Similarly P (x) 6= 0 is well-defined, and means P (a0 , .., an ) 6= 0.

There is a natural grading of the k-algebra R = k[X0 , .., Xn ] defined by

Rd = {f ∈ R : f is a d-form}
Thus R = ⊕d≥0 Rd . Note that R0 is k, and the redundant maximal ideal hX0 , .., Xn i is ⊕d≥1 Rd . Also R1
generates R as a k-algebra, and for two non-negative integers d, e, we have that
Rd .Re ⊂ Rd+e

Clearly an ideal a in R is a homogeneous ideal iff a∩Rd ⊂ a for all d ≥ 0. We denote a∩Rd by ad , and so
a is a homogeneous ideal iff a = ⊕d≥0 ad . If a is a homogeneous ideal in R, we can form another graded ring,
namely A = R/a = ⊕d≥0 Rd /ad .

Definition 8.3.2. In the case when X = V (p) is an irreducible projective closed set, the graded ring:
R/p = ⊕d≥0 Rd /pd
is called the graded coordinate ring of X.

This ring is precisely the coordinate ring of the irreducible affine closed subset :

VAn+1 (k) (p) = π −1 (X) ∪ {0}

of An+1 (k), where π : An+1 (k) \ {0} → Pn (k) is the quotient map introduced in the beginning.

The unfortunate thing is that, without bringing in line bundles and sections, there is no good elementary
interpretation of the graded coordinate ring of an irreducible projective closed set, as there was for a closed
irreducible affine set (where the coordinate ring was just the ring of functions on the closed set which were
restrictions of polynomials on the ambient affine space). The reason is not hard to see, if f is an m-form in
X0 , ..., Xn , it will have a well defined value at the point [a0 : ... : an ] ∈ Pn (k) iff f (λa0 , ..., λan ) = f (a0 , ..., an ),
for all λ 6= 0, λ ∈ k. But this is only possible if the homogeneous degree of f is zero, (i.e. f is a constant),
or f (a0 , .., an ) = 0. Thus, unless f is constant, f does not have a well-defined non-zero value at any point of
74 VISHWAMBHAR PATI

Pn (k). One therefore has to relax the notion of what one means by a “function” on Pn (k), by allowing it to
be undefined at certain points. We recall the analogous situation from complex analysis, where meromorphic
functions, for example, were not required to be everywhere defined.

We carry this out next. Denote Rd /ad by Ad , so that A = ⊕d≥0 Ad . If a is a proper ideal a0 = {0} and so
A0 = k. Note that the element 0 is considered to be of degree ∞, since it lies in all the Ad !

Definition 8.3.3. Let p be a homogeneous prime ideal in the polynomial ring R = k[X0 , ..., Xn ]. Consider
the graded ring A = R/p graded as above. Let S be the multiplicative set defined as S = (∪d≥0 Ad ) \ {0}.
We consider the homogeneous localisation of A, namely S −1 A. Thus we are just inverting all homogeneous
non-zero elements in A. This can also be made into a graded ring with grading from Z by defining for d ∈ Z
 
P (X0 , ..., Xn )
(S −1 A)d = : Q 6= 0, P, Q homogeneous, deg P − deg Q = d
Q(X0 , ..., Xn )
Again, S −1 A = ⊕d∈Z (S −1 A)d , and since p was assumed to be prime, A is an integral domain and the natural
map A → S −1 A is an inclusion which preserves grading. The ring of rational functions on the projective closed
set X = V (p) is defined as (S −1 A)0 , and denoted k(X). It is clearly a k-algebra which is a field, for the inverse
of a non-zero element QP
is Q
P.

This field is the closest thing to the field of rational functions on an irreducible closed affine algebraic set
P
defined in Definition 6.5.2. Note that an element f ∈ k(X) with f = Q is a priori undefined on the zero locus
of the denominator, so it isn’t really a “function” on X. But it is certainly defined on the non-empty Zariski
open subset where Q doesn’t vanish, which is dense in X, since X is irreducible. Thus every rational function
f on an irreducible projective closed set X is defined on a non-empty open (hence dense) subset U of X. For
a point x = [a0 : a1 : .. : an ] ∈ U , we can define the value of f at x to be f (x) = PQ(a (a0 ,....,an )
0 ,...,an )
, for since P and Q
are homogeneous of the same degree, f (λa0 , .., λan ) = f (a0 , ..., an ) for λ 6= 0 in k, as long as Q(a0 , .., an ) 6= 0.
Thus it makes sense to define the value of f at the point x = [a0 : a1 : .. : an ] of to be f (a0 , a1 , .., an ) if
Q(a0 , ..., an ) 6= 0, for this value is independent of the representative (a0 , .., an ) chosen for x.

Note that in the above discussion, f may be defined at a point x where Q vanishes, because f may have
another representation with a denominator not vanishing at x (because of quotienting by p, there are non-trivial
relations in A). For instance :

Example 8.3.4. Let p = hX02 − X1 X2 i in k[X0 , X1 , X2 ], and we denote by xi the images of the coordinate
functions Xi in A = k[X0 , X1 , X2 ]/p, the rational function f = xx02 seems not be defined at the point
x = [0 : 1 : 0] ∈ X = V (p) ⊂ P2 (k). However, in A, we have the relation x20 = x1 x2 , which implies that
this rational function is the same as the rational function xx01 , which is well-defined (and = 0) at the point
x = [0 : 1 : 0].

Before we go any further, it helps to have one unified notion which will cover both closed affine algebraic
sets and projective closed sets, among other things.

Definition 8.3.5. A quasiprojective variety X is an open subset of a projective closed set. An irreducible
quasiprojective variety is an open subset of an irreducible projective closed set.

Clearly, a projective closed set is a quasiprojective variety. If X is an an affine closed set, it is the open
subset X ∩ U0 of its projective closure X, and thus a quasiprojective variety. In fact, X is a quasiprojective
variety iff it is the difference of two projective closed sets. Note that a non-empty irreducible quasiprojective
variety X has to be dense in its projective closure. Indeed, if U is a non-empty open subset of an irreducible
(projective or affine) closed set X, then U is Zariski dense in X, otherwise we would have a decomposition
X = U ∪ (X \ U ) into proper closed subsets.
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 75

Example 8.3.6. The complement An (k) \ X of any affine algebraic set X ⊂ An (k) is a quasiprojective variety,
because it may be viewed as the difference Pn (k) \ (X ∪ V (X0 )) (where X is the projectivisation of X ⊂ U0 )
of projective closed sets. Clearly, it is irreducible.

Remark 8.3.7. We recall that the family of (Zariski) closed subsets of An (k) and Pn (k) are closed under finite
unions and arbitrary intersections. Quasiprojectives in Pn (k) fail on both counts. For example, in P1 (k), the
subsets An = P1 (k) \ {[1 : 0], [0 : 1/n]} are all quasiprojectives. But
∩ni=1 An = P1 (k) \ {[1 : 0], [0 : 1], [0 : 1/2], .., [0 : 1/n], ...}
and the right hand set cannot be expressed as the difference of two closed subsets of P1 (k) (verify). Similarly,
A1 = P2 (k) \ V (X0 X1 ) = U0 \ V (X1 ) = A2 (k) \ {X1 -axis} and A2 = [1 : 0 : 0] are quasiprojectives inside P2 (k),
but their union is not (verify).

Exercise 8.3.8. Show that a quasiprojective variety is irreducible iff it is not the union of two proper closed
subsets which are both quasiprojective varieties. Prove that every quasiprojective variety is a Noetherian
topological space, and a finite union of irreducible ones. (Just argue with the projective closure, and show that
for a quasiprojective variety X, the intersections Yi ∩ X are all non-empty for Yi an irreducible component of
X, and are precisely the irreducible components of X).

For this reason, we shall work only with irreducible quasiprojective varieties, the generalisations to arbitrary
quasiprojective varieties (if any) usually proceeds component by component.

Definition 8.3.9. Let X 6= φ be an irreducible quasiprojective variety. Define the field of rational functions
k(X) on X to be k(X). If fi ∈ k(X) for i = 1, 2, ..m, we say that the tuple f = (f1 , ..., fm ) is a rational map
f : X → Am (k). Note that it is not a map in the usual sense, for it is only defined on a dense open subset of
X.

Definition 8.3.10. A rational function f ∈ k(X) for X an irreducible quasiprojective variety is said to be
P
regular at x = [a0 : a1 :, .., : an ] ∈ X if there exists a representative of f as f = Q where Q(a0 , .., an ) 6= 0,
and P and Q are homogeneous of the same degree. Such a representative will be called a good representative
P (x)
at x. If f is regular at x, we define the value of f at x to be f (x) = Q(x) . This value is well-defined, i.e. the
same for all good representatives of f at x. Note that f regular at x implies that f is regular in a Zariski open
P
neighborhood U of x in X (i.e. the complement of V (Q), where f = Q is a good representative at x). We say
a rational function f ∈ k(X) is regular on a subset Y ⊂ X if f is regular at each point of Y . Thus a rational
function on X which is regular at x is also regular on a neighborhood of x. We denote the k-algebra (check
that it is one) of all rational functions on X which are regular at x by OX,x . Finally, the k-algebra of rational
functions on X which are regular on some subset A ⊂ X is denoted k[A].

Proposition 8.3.11. We have the following elementary facts for an irreducible quasiprojective variety X ⊂
Pn (k) :
(i): There is an evaluation map: x : OX,x → k which is a k-algebra homomorphism.

(ii): The kernel ker x is a maximal ideal, and the unique maximal ideal in OX,x , thus making it a local
ring. If f ∈ k[X], then f has the well-defined value x (f ) = f (x) at each point x ∈ X.

(iii): Let X be an irreducible closed projective set in Pn (k), and assume X 6⊂ V (X0 ). Consider the 0-th
affine piece of X, i.e. the quasiprojective variety Y = X ∩ U0 , where U0 is the affine subset {X0 6= 0} in
Pn (k). The field of rational functions k(Y ) defined above in Definition 8.3.9 coincides with the definition
of k(Y ) in Definition 6.5.2. Further, k[Y ] defined above coincides with the coordinate ring of Y as defined
in Definition 1.2.3, and OY,x = k[Y ]mx , where mx is the maximal ideal corresponding to the point x ∈ Y .

(iv): OX,x = ∪{k[U ] : U is a neighborhood of x}.

76 VISHWAMBHAR PATI

Proof:
(i) is trivial, using a good representative at x.

To see (ii), note x takes k ⊂ OX,x identically to itself, and is surjective. Its kernel is therefore a maximal
P
ideal in OX,x . To see that it is the unique maximal ideal, note that if f ∈ / kerx , then f (x) 6= 0. If f = Q is
a good representative for f at x = [a0 : a1 : ..an ], then we have P (a0 , ..., an ) 6= 0, and Q
P is a rational function
lying in OX,x , which is an inverse for f , showing that f is invertible in OX,x . Thus ker x is the unique maximal
ideal in OX,x , which is therefore a local ring. The second assertion of (ii) follows from the definition of k[X].
This proves (ii).

We now prove (iii). Let X = V (p) where p is a homogeneous prime ideal. Let xi denote the images of Xi
in A = k[X0 , .., Xn ]/p and note that, since we have assumed X is not contained in the hyperplane V (X0 ), the
P (x0 ,..,xn )
function x0 is a non-zero element of A1 ⊂ A. According to the Definition 8.3.9, k(Y ) = k(X). If f = Q(x 0 ,..,xn )
is an element of k(X), where P ∈ Ad and Q ∈ Ad and Q 6= 0, then we define the element Φ(f ) ∈ Q(k[Y ]) (the
quotient field of the affine coordinate ring k[Y ] of Y ) to be p(x1 ,..,xn ) D D
q(x1 ,..,xn ) , where p =0 P , and q =0 Q are in the
coordinate ring k[Y ] = k[X1 , .., Xn ]/pD . We leave it to the reader to check that :

(1): The indeterminacy of the polynomials P (X0 , .., Xn ) and Q(X0 , .., Xn ) upto homogeneous elements
in pd = p ∩ Rd leads to the indeterminacy of p and q upto elements in pD in k[X1 , ....., Xn ], so that
p(x1 , ..., xn ) and q(x1 , .., xn ) are well defined in k[Y ].
P
(2): Φ(f ) does not depend on the representative Q chosen for f in k(X).

(3): Q 6= 0 in Ad implies q(x1 , .., xn ) 6= 0 in k[Y ]. (Just use the Lemma 8.2.20), and finally Φ is a k-algebra
homomorphism.

Define the inverse map ψ : Q(k[Y ]) → k(X) by writing an element f ∈ Q(k[Y ]) as p(x1 ,..,xn )
q(x1 ,..,xn ) where p is an
inhomogeneous polynomial of degree d and q is an inhomogeneous polynomial of degree e, and letting
P (x0 , .., xn )
ψ(f ) =
Q(x0 , ..., xn )
where:

P (X0 , ..., Xn ) = X0e−d (0 pH (X0 , ..., Xn )) ; Q(X0 , ..., Xn ) = q H (X0 , ..., Xn ) if e ≥ d

P (X0 , ..., Xn ) = pH (X0 , ..., Xn ) ; Q(X0 , ..., Xn ) = X0d−e (pH (X0 , ..., Xn )) if e ≤ d

Again, the reader is urged to use the Lemma 8.2.20 to convince herself that ψ(f ) is a well-defined element
of k(X) and the inverse to Φ. This proves the first assertion of (iii), viz. that the definitions of k(Y ) arising
from Definitions 8.3.9 and 6.5.2 agree.

For the remaining assertions of (iii), let x = [1 : a1 : .. : an ] ∈ Y . Q(x) 6= 0 implies Q(1, a1 , ..., an ) 6= 0, i.e.
that q(a1 , ..an ) = QD (a1 , .., an ) 6= 0, so a rational function f ∈ k(X) being regular at a point x ∈ X ∩ U0 is
equivalent to Φ(f ) being regular at the point x of the affine closed set Y = X ∩ U0 according to the Definition
6.5.1. Thus OY,x is just the subring of functions f representable as pq ∈ Q(k[Y ]) whose denominator q does
not vanish at x, viz. the subring k[Y ]mx .

Finally, by the Remark 6.5.5, all rational functions on the irreducible affine closed set Y = X ∩ U0 which are
regular at all points of it are precisely the elements of the affine coordinate ring k[Y ]. This proves the second
and third assertions of (iii).

For (iv), note that a rational function that is regular on a neighborhood U of x is regular at x, so that the right
P
hand side is clearly contained in OX,x . On the other hand, if f ∈ OX,x , then let Q be a good representative for
f at x, i.e. Q(x) 6= 0. Then this representative shows that f is regular all over the neighborhood U = X \ V (Q)
of x and hence f ∈ k[U ]. This proves (iv) and the proposition. 2
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 77

Exercise 8.3.12. Let X ⊂ Pn (k) be an irreducible quasiprojective variety. Show that for x ∈ X, the local
ring OX,x is the same as OU,x for any neighbourhood U of x.

Remark 8.3.13. We note that in the case of irreducible affine closed sets, we first defined the k-algebra of
regular functions, which are restrictions of regular functions on the ambient affine space An (k) (viz. polynomi-
als) and then the rational functions to be the quotient field of this ring. For closed projective sets, one starts
first with the rational functions, and then defines regular functions. The reason for this is that the only regular
functions on Pn (k) are constants, as is proved in the following proposition ! (In fact, later we shall see that
the only regular functions on an irreducible closed projective set are constants.)

Example 8.3.14. Consider P1 (k), the projective line. It is covered by:

U0 = {[1 : z] : z ∈ k}; U1 = {[w : 1] : w ∈ k}

At a point x ∈ U0 ∩U1 , the affine coordinates z and w of x with respect to U0 and U1 respectively are related by
P (X0 ,X1 )
w = z1 . The rational function f = Q(X 0 ,X1 )
with deg P = deg Q, becomes, according to the proposition above,
p(z)
the rational function q(z) on the affine set U0 = k, where p(z) = P (1, z) and q(z) = Q(1, z) are polynomials
r(w)
in one variable, not necessarily of the same degree. It coincides with the rational function s(w) obtained by
1 p(z)
substituting w = in z and clearing denominators from numerator and denominator. If k = C, the reader
q(z)
will recognize these as meromorphic functions on C having at worst a pole, not an essential singularity, at ∞),
i.e. rational functions.

Proposition 8.3.15. The ring of regular functions k[Pn (k)] on projective space is k, the constant functions.

Proof: Let f ∈ k(Pn (k)) be regular on Ui . Then, by the Proposition 8.3.11 above, we have a polynomial
bi , .., Xn ) such that f (z0 , ., 1, .., zn ) = pi (z0 , ., zbi , .., zn ) for all (z0 , ., zbi , .., zn ) ∈ Ui . This means that
pi (X0 , ., X
f (X0 , ...., Xn ) = Pi (X0 ,...,X
di
n)
, where Pi are homogeneous polynomials in k[X0 , ..., Xn ] of degree di . This means
Xi
that for each point [a0 , ..., an ] ∈ Ui ∩ Uj , we have the relation :
d
aidi Pj (a0 , ..., an ) = aj j Pi (a0 , ..., an ) 0 ≤ i, j ≤ n
d
Thus the regular functions Xidi Pj (X0 , ..., Xn ) and Xj j Pi (X0 , ..., Xn ) on the affine space An+1 (k) agree on the
non-empty open subset An+1 (k) \ (V (Xi ) ∪ V (Xj )). Since this non-empty open set is Zariski dense in An+1 (k),
the relation above holds at all points of An+1 (k). That is :
d
Xidi Pj (X0 , ..., Xn ) = Xj j Pi (X0 , ..., Xn )
in the ring k[X0 , .., Xn ]. Since Xi are irreducible, and hence prime polynomials in the unique factorisation
domain k[X0 , .., Xn ] (Proposition 2.2.17), this implies that Pi is divisible by Xidi for all i = 0, 1, ..n. Since
deg Pi = di , we have Pi = ci Xidi for ci ∈ k, and the relation above says that ci = cj for all i, j. Thus letting
c = ci we have f = Pdii = c ∈ k, and the proposition follows. 2
Xi

Definition 8.3.16. Let X ⊂ Pn (k) and Y ⊂ Pm (k) be quasiprojective varieties. We say that a set map
f : X → Am (k) is a regular map if each component fi for i = 1, 2, ..m is a regular function on X, fi ∈ k[X]
for i = 1, .., m, as defined in the Definition 8.3.10 above. (By (ii) of the Proposition 8.3.11, a regular map
f : X → Am (k) has a well defined value in Am (k) at each x ∈ X). We say that a map f : X → Y is a regular
map if for each x ∈ X, and each affine open Ui ⊂ Pm (k) containing f (x), there exists an open neighborhood
U of x such that f (U ) ⊂ Ui and f : U → Ui ' Am (k) is a regular map, and f (X) ⊂ Y . (Incidentally, an open
subset of a quasiprojective variety is also a quasiprojective variety.)
78 VISHWAMBHAR PATI

We need to verify that the definition above makes sense, i.e. we have consistency with respect to choice of
affine open Ui on the right. This follows because if f (x) ∈ Ui ∩ Uj , then both the i-th and j-th homogeneous
coordinates of f are non-zero. Let U be an open neighborhood of x such that f (U ) ⊂ Ui and f : U → Ui
a regular map. Let V be a open neighborhood of x such that f (V ) ⊂ Uj . Then f : W → Ui ∩ Uj where
W = U ∩ V . If we write f = [f0 : ... : 1 : ...fm ] on U (with 1 at the i-th slot), the regularity of f implies
Pk
that the m functions f0 , .., fi−1 , fi+1 , ..fm are regular on U . We shrink U so that the good representatives Q k

for fk are valid all over U , i.e. Qk do not vanish at any point in U for each k = 0, 1, ..bi, ..m. In the affine
piece Ui ∩ Uj , this same map can be written as [ ff0j : ... : f1j : ..1 : .. ffmj ] (with 1 in the j-th slot, f1j in the i-th
slot). Clearly fj (y) 6= 0 for all y ∈ W since f (W ) ⊂ Uj . Thus if we use the affine open Uj on the right, the
representation for f on W is [g0 :, ... : gj−1 : 1 : gj+1 : .. : gm ], where gi = f1j , and gk = ffkj for k 6= i. If we
Pk Q Pk Qj
write fk = Q k
, with Pj nonvanishing all over W , we have gi = Pjj and gk = Qk Pj for k 6= i. Clearly these are
regular all over W since all the Qk and Pj are non-vanishing all over W .

There is an easy criterion for a map to be a regular map, without invoking affine pieces on the right. Namely,
Lemma 8.3.17. Let f : X → Y be a map, with X, Y as in the definition above. Then f is regular iff for each
x ∈ X there exists a neighborhood U of x in X, and homogeneous polynomials (= d-forms) Pi for i = 0, 1, ..m,
all of the same degree d, such that:

(i): f has a representation :

f ([a0 : ... : an ]) = [P0 (a0 , ..., an ) : .... : Pm (a0 , ...., an )]
for all [a0 : ... : an ] ∈ U

(ii): at least one of the Pi is everywhere non-vanishing on U .

Proof:
For each x ∈ X, we can find a neighbourhood U of x such that f (U ) is contained in one affine piece, say
Am (k) = U0 ⊂ Pm (k) for simplicity. Write f = (f1 , .., fm ) where fi are regular on U . Write good representatives
Pi
fi = Q i
for fi at x, so that Qi (x) 6= 0 for all i, and observe that by shrinking U if necessary, one can guarantee
that Qi never vanishes on U for each i. Then the element [1 : f1 : ... : fm ] ∈ U0 is the same as the element
[Q1 Q2 ..Qm : P1 Q2 ..Qm : ... : Pm Q1 Q2 ..Qm−1 ] by clearing the denominators, and this representative on U
satisfies the requirement of the lemma, since the first entry Q1 Q2 ..Qm is everywhere nonvanishing on U . The
converse is left as an exercise. 2

Remark 8.3.18. We note that the local representation of f from the lemma above is not unique, and two
representations [P0 : ... : Pm ] and [Q0 : ... : Qm ] will satisfy Qi Pj = Qj Pi for all i 6= j.

In line with the above Lemma 8.3.17, one may also make the following definition:
Definition 8.3.19. If X ⊂ Pn (k) is a quasiprojective variety, then a rational map f : X → Pm (k) is given by
an (m + 1)-tuple of homogeneous polynomials Pi 6= 0 ∈ k[X0 , .., Xn ], all of the same degree, and thus defining
f by :
[X0 : X1 : ... : Xn ] 7→ [P0 (X0 , ..., Xn ) : ... : Pm (X0 , ..., Xn )]
This is actually a set-mapping on the Zariski open subset X \∩m i=0 V (Pi ) where at least one of the Pi is non-zero.
Such a rational map will be called regular at x = [a0 : ... : an ] ∈ X if, as before, Pi (x) 6= 0 for some i, so that
the point defined by :
f (x) = [P0 (a0 , .., an ) : ... : Pm (a0 , .., an )]
is well defined, and is the value of f at x. It is easy to check that if one uses the identification of the affine pieces
of Pm (k) with Am (k), then this coincides with the definition of rational maps to Am (k) given in Definition
8.3.9. We call f : X → Y , where Y ⊂ Pm (k) is another quasiprojective variety, a rational map if it is a rational
map to Pm (k), and if f (x) ∈ Y for every x ∈ X at which f is regular. If X and Y are irreducible, it is easy to
check that a rational map f : X → Y will give rise to a field homomorphism (which is therefore an inclusion)
f ∗ : k(Y ) → k(X), and conversely any such field inclusion will give a rational map f : X → Y . Two irreducible
quasiprojective varieties will be called birationally equivalent if f ∗ is an isomorphism of k(Y ) with k(X), or
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 79

equivalently if there exists a rational map g : Y → X, such f ◦ g(y) = y for all y ∈ Y where g is regular and f
is regular at g(y), and g ◦ f (x) = x for all x ∈ X where f is regular and g is regular at f (x).

Definition 8.3.20. A regular map f : X → Y between irreducible quasiprojective varieties is said to be an

isomorphism if the inverse f −1 : Y → X exists, and is a regular map. In this case we say X and Y are
isomorphic. A quasiprojective variety which is isomorphic to an irreducible affine closed set is said to be an
affine variety. A quasiprojective variety which is isomorphic to an irreducible closed projective set is said to
be a projective variety.
Example 8.3.21. The quasiprojective variety U := A2 (k) \ {(0, 0)} is neither an affine variety nor a projective
variety. We already saw in Remark 6.6.6 that it is not an affine variety. We also saw noted there that the ring
of regular functions k[U ] = k[x, y]. To see that it is not a projective variety, we need the result that on an
irreducible closed projective set, the only regular functions are constants, i.e. elements of k. This fact, which
generalises Proposition 8.3.15, will be proved later. Since k[U ] = k[x, y] 6= k, U cannot be isomorphic to a
closed projective set either. Similarly for An (k) \ {0}, and n ≥ 2.

Exercise 8.3.22. Show that for n ≥ 2 and any point x ∈ Pn (k), the irreducible quasiprojective variety
Pn (k) \ {x} is neither an affine nor a projective variety. (See the Remark 6.6.6 of §6)
We will see some more examples at the end of this subsection.
Exercise 8.3.23. Show that two irreducible quasiprojective varieties are birationally equivalent iff they contain
isomorphic open subsets. (Compare with the Proposition 6.7.3.) Thus, for example, the (projective closures
of) the cubic node and the cubic cusp are both birationally equivalent to P1 (k).
It is convenient to be able to reduce various arguments about rational or regular maps between quasipro-
jective varieties to the situation of affine closed sets. To this end we have the following propositions:

Proposition 8.3.24 (A basis for the Zariski topology on quasiprojective varieties).

(i): Let Y ⊂ Pn (k) be an irreducible projective closed set. For 0 ≤ i ≤ n, let Yi be the affine pieces of Y .
Then the family of open sets :
{D(f ) = Yi \ V (f ) : f ∈ k[Yi ], 0 ≤ i ≤ n}
constitute a basis for the Zariski topology on Y , and are again called basic open sets of Y . These are
again all irreducible affine varieties . Another basis is the collection:
{DY (F ) = Y \ V (F ) : F ∈ k[X0 , .., Xn ] homogeneous}
(ii): Let Y be an irreducible quasiprojective variety. Then Y has a basis consisting of open sets which are
irreducible affine varieties. We will refer to an affine basic open containing a point x ∈ Y as an affine
neighbourhood of x.

Proof: The asssertion (i) follows directly from the Proposition 6.4.18, because Y = ∪ni=0 Yi and the affine
pieces Yi are are open in Y . Since Y is irreducible, any non-empty open subset of Yi is open in Y , and hence
dense in Y , so has closure equal to Y . So all the basic opens of each Yi are irreducible quasiprojective varieties.
By the Exercise 6.6.5, these basic opens are isomorphic to irreducible affine closed sets.

To see (ii), note that Y irreducible and quasiprojective implies it is an open subset of a closed projective set
Z, and so any open subset of Y is an open subset of Z, and may be expressed as a union of D(fi )’s. Thus the
subfamily of {D(f )} where f ∈ k[Zi ] for some affine piece Zi of Z, and such that D(f ) ⊂ Y will constitute a
basis for Y . Each member of this collection is an affine variety. This proves the proposition. 2
80 VISHWAMBHAR PATI

Remark 8.3.25. It is natural to enquire whether DX (F ) = X \ V (F ) for X a projective closed subset in

Pn (k), and F 6= 0 a homogeneous polynomial, is an affine variety. If F is linear, i.e. homogeneous of degree 1,
then clearly DX (F ) is an affine piece of X, and thus an affine variety (just complete F to a set of coordinates
{Xi } on An+1 (k), so that X0 = F ). That D(F ) is an affine variety in general for F 6= 0 any homogeneous
polynomial will be seen later after introducing the Veronese embedding.

Lemma 8.3.26. Let X be a topological space, with X = ∪α∈Λ Uα ,where Uα are open subsets of X. Then a
subset C of X is closed in X iff C ∩ Uα is closed in Uα for each α ∈ Λ.
Proof:
Only if is clear. If C ∩ Uα is closed in Uα for each α, write C ∩ Uα = Zα ∩ Uα , where Za is closed in X. Let
Tα = X \ Uα , which is also closed in X for each α ∈ Λ. Now verify that C = ∩α∈Λ (Zα ∪ Tα ), which is clearly
closed in X. 2

Corollary 8.3.27. Let f : X → Y be a map of topological spaces, and let {Uα }α∈Λ and {Vβ }β∈Γ be open
coverings of X and Y respectively, such that for each α ∈ Λ, there exists a β(α) ∈ Γ such that f (Uα ) ⊂ Vβ(α) .
Then f is continuous iff f : Uα → Vβ(α) is continuous for each α ∈ Λ.

Proof:
Again, only if is clear. If C is closed in Y , C ∩ Vβ is closed in Vβ for each β ∈ Γ. Since we have that
f : Uα → Vβ(α) is continuous, it follows that f −1 (C) ∩ Uα is closed in Uα for every α. Thus f −1 (C) is closed
in X by the previous lemma. 2

Exercise 8.3.28 (Topology of Quasiprojectives).

(i): Verify that for X ∈ Pn (k) a quasiprojective variety, the collection of sets:
{DX (f ) : f is regular on X}
where
DX (f ) := {x ∈ X : f (x) 6= 0}
constitutes a basis for a topology on X which agrees with the subspace topology on X induced from
Pn (k). (Compare with the Proposition 6.4.18.)

(ii): Show that regular maps between quasiprojective varieties are continuous (with respect to the Zariski
topology).

(iii): Show that open (resp. closed) subsets of a quasiprojective variety are also quasiprojective varieties.

We now come to an important characterisation of regular maps between quasiprojective varieties.

Proposition 8.3.29. Let f : X → Y be a mapping of irreducible quasiprojective varieties, with X ⊂ Pn (k)
and Y ⊂ Pm (k). The following are equivalent:

(i): f is regular.

(ii): f is continuous, and for each x ∈ X, the map g 7−→ f ∗ g := g ◦ f is a k-algebra homomorphism of
OY,f (x) → OX,x .
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 81

Proof:

(i)⇒(ii)
By the Definition 8.3.16 of a regular map, for each x ∈ X, there is an open neighbourhood U of x in X
and an affine open set Ui = D(Xi ) ⊂ Pm (k) such that f : U → Ui is regular. By Propositions 8.3.24, and
8.3.27, it is enough to show that a regular map f : U → Ui is continuous for each affine basic open set U in X
and Ui = D(Xi ) ⊂ Pm (k). By (iii) of the Proposition 8.3.24, U and Ui are irreducible affine varieties, so we
are reduced to showing that a regular map from an irreducible affine closed set to Am (k) is continuous with
respect to the Zariski topologies. But this follows from the Proposition 5.1.9, Remark 6.5.5 and the discussion
following Proposition 6.1.3, because any such regular mapping is just a morphism as defined there. Thus a
regular map is continuous.

Similarly, any k-algebra homomorphism f ∗ : A → B between k-algebras of finite type has the property that
if mx is a maximal ideal in B, (f ∗ )−1 (mx ) is a maximal ideal in A, denoted by mf (x) , (see Propositions 5.1.7
and 5.1.9). It follows that f ∗ maps the multiplicative set A \ mf (x) to the multiplicative set B \ mx , and the
corresponding localisations Amf (x) to Bmx . So the property we desire is true of morphisms between irreducible
affine closed sets, and since we have reduced to this case, we are done, by Exercise 8.3.12.

(ii)=⇒(i) Again, by continuity, and (iii) of the Proposition 8.3.24 one can reduce the question to f : X → Y
a continuous map of irreducible affine closed sets. We are given that f ∗ : OY,f (x) → OX,x is a k-algebra
homomorphism for each x ∈ X. Since the coordinate ring k[Y ] is contained in all the local rings OY,f (x) ,
it follows that the image of k[Y ] lies in ∩x∈X OX,x . By (v) of the the Proposition 6.5.4, this intersection is
precisely k[X]. Thus f ∗ : k[Y ] → k[X] is a k-algebra homomorphism, and each coordinate function fi = f ∗ (xi )
(where xi is the i-th coordinate function on the affine closed set Y ⊂ Ak (m)) is an element of k[X], and
therefore a regular function on X, so that the mapping f = (f1 , .., fm ) is a regular map into Y . This proves
the proposition. 2

Corollary 8.3.30 (The category of quasiprojective varieties). The composite of regular maps between quasipro-
jective varieties is a regular map, and that the identity map of a quasiprojective variety is regular, as is clear
from (ii) of the foregoing proposition. Thus quasiprojective varieties together with regular maps as morphisms
form a category. In the foregoing, we have frequently restricted ourselves to irreducible quasiprojective va-
rieties, and one has to make the necessary straightforward modifications to those assertions to obtain the
corresponding facts for quasiprojective varieties.

We next go through some examples of quasiprojective varieties and rational maps between them to clarify
the foregoing concepts.

Example 8.3.31 (Automorphisms of P1 (k)). First note that all maps P1 (k) → P1 (k) of the form:
[z0 : z1 ] 7→ [az0 + bz1 : cz0 + dz1 ]
with ad−bc 6= 0, are automorphisms of P1 (k), i.e. regular maps with regular inverses. These are called projective
∗
linear transformations, and form a group  denoted P GL(2, k). In fact the group P GL(2, k) = GL(2, k)/k , by
a b
observing that scaling of the matrix leads to the same map. We claim that all automorphisms of
c d
P1 (k) are of this form. If φ : P1 (k) → P1 (k) is an automorphism, we may postcompose it with a projective
linear transformation to ensure that φ([1 : 0]) = [1 : 0]. So it is enough to show that an automorphism of P1 (k)
fixing the point [1 : 0] is a projective linear transformation.

Now U1 = P2 (k) \ {[1 : 0]} is the affine space A1 (k), and φ : U1 → U1 is thus an automorphism of A1 (k).
In particular, it is injective, and therefore defined by a degree 1 polynomial z 7→ az + b (by Example 5.1.10,
all morphisms of affine spaces An (k) are given by polynomials). Thus φ is the projective linear transformation
[z : 1] 7→ [az + b : 1], or what is the same thing, [z0 : z1 ] 7→ [az0 + bz1 : z1 ], a projective linear transformation.
Hence our assertion that Autk (P1 (k)) = P GL(2, k).
82 VISHWAMBHAR PATI

This same fact is true for all Pn (k), i.e. the group of regular automorphisms of Pn (k) is P GL(n + 1, k) =
GL(n + 1, k)/k ∗ . One needs a little more machinery to prove it, so it is postponed till we get to divisors,
intersections etc.
Example 8.3.32 (The Cremona Transformation of P2 (k)). Let us define a rational map of P2 (k) to itself by
the assignment :
[X0 : X1 : X2 ] 7→ [X1 X2 : X0 X2 : X0 X1 ]
Clearly, this map is regular wherever at least one of the monomials X1 X2 , X0 X2 , X0 X1 are non-zero, i.e.
at all points [X0 : X1 : X2 ] where at least two of the homogeneous coordinates Xi are non-zero. This set is
P2 (k) \ {P, Q, R} where P = [1 : 0 : 0], Q = [0 : 1 : 0], R = [0 : 0 : 1]. This transformation is called the
Cremona transformation. It is a birational transformation, whose rational inverse is again the same Cremona
transformation, viz. :
[X0 : X1 : X2 ] 7−→ [X1 X2 : X0 X2 : X0 X1 ]
which is again regular on P2 (k)\{P, Q, R}. This map “blows down” the coordinate hyperplanes V (X0 ), V (X1 ),
V (X2 ) to the points P , Q, R respectively. Similarly, it “blows up” the points P, Q, R to the coordinate hyper-
planes. The tranformation sets up an isomorphism between the basic open sets D(X0 X1 X2 ) (the complement
of the three coordinate hyperplanes) and itself.

Example 8.3.33 (Projection from a plane in Pn (k)). Let π : An+1 (k) \ {0} → Pn (k) be the natural quotient
map. For a k-subspace V ⊂ k n+1 , we denote the image π(V \ {0}) in Pn (k) as P(V ), and call it the projective
space on V . Let {Li }mi=0 be m + 1 linearly independent linear functionals on k
n+1
. Let E := ∩m i=0 ker Li , a
n+1
k-subspace of k of dimension n − m. Define the rational map:

p : Pn (k) → Pm (k)
[X0 : X1 : .. : Xn ] 7→ [L0 (X0 , .., Xn ) : .. : Lm (X0 , .., Xn )]
which is clearly regular on Pn (k) \ P(E), where P(E) ' Pn−m−1 (k). The reason this map is called projection
from E (or more precisely, P(E)) is that each point v = (a0 , .., an ) ∈ k n+1 can be expressed uniquely as a
(direct) sum :
v = T p(v) + w
where T : k m+1 → k n+1 is a k-linear map satisfying pT (x)) = x for all x ∈ k m+1 , and p = (L0 , L1 , .., Lm )
and w ∈ E. This is just a vector space splitting of the linear map p = (L0 , .., Lm ), so that T is an injective
mapping of k m+1 = Im p into k n+1 . T maps k m+1 isomorphically onto the subspace H := Im T ⊂ k n+1 , a
vector space complement of E in k n+1 . With this above decomposition, the map p just kills the component
w. In particular, if v ∈
/ E, p(v) 6= 0, and hence T p(v) 6= 0, and so T p(v) determines a line in im T = H, i.e.
an element of P(H). Equivalently, if v 6∈ E, the k-subspace spanned by v and E is a subspace of dimension
n − m + 1, and thus intersects the m + 1 dimensional subspace H in a 1-dimensional space, viz. the element
p(v) ∈ P(H).

A particular case, of course, is the projection [X0 : .. : Xn ] 7→ [X0 : ... : Xm ] of Pn (k) \ E to Pm (k), from the
(n − m − 1) dimensional projective space P(E), where E is the (n − m) dimensional subspace consisting of all
points [0 : 0 : ...0 : Xm+1 : .. : Xn ] in Pn (k).

Example 8.3.34 (The Veronese Embedding). Let Rm be the m-th graded component of R = k[X0 , .., Xn ],
that is, the k-vector space of all homogeneous polynomials of degree m. Since every homogeneous degree
m monomial in X0 , ..., Xn occurs exactly once in the coefficient of tm in the expression f (X0 , .., Xn , t) =
Πni=0 (1 − tXi )−1 , it follows that the number of these monomials is precisely the coefficient of tm in the product
f (1, 1..1, t) = (1 − t)−n−1 , which is easily checked to be the binomial coefficient (n+m
m ), which we shall denote
by N . Consider the projective space P(Rm ) ' PN −1 (k). It is convenient to write a point of this projective
space as x = [: YI ]I∈S , where S is defined by :
S = {I = (i0 , .., in ) : Σnk=0 ik = m}
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 83

We now define a mapping:

j : Pn (k) → P(Rm )
[X0 : ... : Xn ] 7→ [: X I ]I∈S
where X I := X0i0 ....Xnin . This map is clearly regular, since, for example if Xj 6= 0, the monomial X I where
I = (0, 0, .., m, 0, 0) with m at the j + 1-th place and zero elsewhere, will be just Xjm , which will be non-zero.

For two multi-indices I, J ∈ S, we define I + J = (i0 + j0 , .., in + jn ). Clearly, since X I X J = X K X L

whenever I + J = K + L, the image of j lies in the projective closed set V (a) where a is the homogeneous ideal
generated by all the quadratic expressions YI YJ − YK YL with I + J = K + L, in the N -variable polynomial
ring k[YI ]I∈S .

Claim: The image of j is precisely V (a).

It is convenient to introduce the following notation. Define the length l(I) of a multi-index I ∈ S to be the
number of non-zero entries in I.

We first claim that if a point [: YI ]I∈S ∈ V (a), then there exists a multi-index I = (0, 0, ..m, ..0), of length
1, such that YI 6= 0. For, if say Y(m,0,0,..0) = 0, then using the equation :
2
Y(m−r,r,..,0) = Y(m−2r,2r,0,0..0) Y(m,0,0..,0)
one sees that Y(m−r,r,0,...,0) = 0 for any r ≤ m 2 , since the right hand side is zero by assumption. Interchanging
the roles of r and m − r, one sees that all Y(i,m−i,0,..,0) = 0. By repeating this argument for other multi-indices
I = (0, ..., m, .., 0) of length 1, entry, we conclude that YI = 0 for all multi-indices I ∈ S with l(I) ≤ 2. Assume
inductively that we have proved that YI = 0 for all multi-indices with l(I) ≤ r. Consider for example a multi-
index I = (i0 , i1 , .., ir , ..0) with r + 1 non-zero entries. Let us assume, to be specific, that i0 ≤ i1 , otherwise we
interchange the roles of i0 and i1 . Again, we have the equation:
YI2 = Y(2i0 ,i1 −i0 ,..,ir ,..,0) Y(0,i0 +i1 ,i2 ,..,ir ,..,0)
where the extreme right hand term is zero by induction hypothesis, so YI = 0. One can repeat this argument
for any other multi-index with length r + 1. Thus we will have that all the homogeneous coordinates are zero,
a contradiction. This proves the claim.

Thus for a point [: YI ]I∈S to lie in V (a), we must have YI 6= 0 for at least one of the length 1 multi-indices
I = (0, 0, .., m, .., 0). Again, to be specific, say Y(m,0,..,0) 6= 0. Define X0 = Y(m,0,..,0) , and Xi = Y(m−1,0,..,1,..,0) ,
where the entry 1 occurs in the (i + 1)-th spot in the subscript on the right. We now claim that:
j([X0 : ... : Xn ]) = [: YI ]I∈S
Again, we first claim that :
 r
Y(m−r,r,0,..,0) X1
=
Y(m,0,...,0) X0
This is true for r = 1 by definition, and for other r we inductively apply:
Y(m−r,r,0,..,0) Y(m−1,1,0,..,0) = Y(m,0,..,0) Y(m−(r+1),r+1,0,..,0)
Similarly for the other multiindices I = (m−r, 0, .., r, 0, .., 0) with r at the (i+1)-th spot. Now if I = (i0 , i1 , .., in )
with i0 + i1 + .. + in = m is any multi-index, we have the relations:
 i  i  i    
X0 0 X1 1 Xn n Y(m−i1 ,i1 ,0,..,0) Y(m−in ,0,..,in )
... = ...
X0 X0 X0 Y(m,0,..,0) Y(m,0,..,0)
Now we we use the relation:
Y(m−i1 ,i1 ,..,0) Y(m−i2 ,0,i2 ,...,0) = Y(m,0,...,0) Y(m−i1 −i2 ,i1 ,i2 ,0,..,0)
to obtain that the right hand side is :
    
Y(m−i1 −i2 ,i1 ,i2 ,..,0) Y(m−i3 ,0,..i3 ,..0) Y(m−in ,0,..,in )
....
Y(m,0,..,0) Y(m,0,..,0) Y(m,0,..,0)
84 VISHWAMBHAR PATI

and by similar collapsings, and the fact that i0 = m − i1 − ... − in , this reduces to the final equation:
 I  
X Y(i0 ,i1 ,..,in )
=
X0m Y(m,0,..,0)
for all I ∈ S. This proves the assertion that the image of j is precisely V (a). In fact, we define the map:
ψ : V (a) → Pn (k)
[: bI ]I∈S ) 7 → [a0 : a1 : .. : an ]
b
where a0 = b(m,0,..,0) and ai := (m−1,0,..1,..0)
b(m,0,...0) if b(m,0,..,0) 6= 0 (resp. the suitable modification of this definition
if b(0,0,..m,...0) 6= 0). This is clearly well-defined because of the relations of a, and regular on the affine piece
D(b(m,0..,0) ) ∩ V (a) (resp. D(b(0,..,m,..0) ) ∩ V (a) which is a neighbourhood of the point x = [: bI ]I∈S . Since this
point x is arbitrary, we have that ψ is a regular inverse to j. Thus the projective space Pn (k) and V (a) are
isomorphic, and j and ψ are, in particular, homeomorphisms with respect to Zariski topologies.

We now list some applications of the Veronese embedding. The first is the reduction of questions on degree
m hypersurfaces in Pn (k) to questions on hyperplane sections, i.e. degree one hypersurfaces, on the Veronese
subvariety V (a) defined above.

Example 8.3.35 (The space of degree m hypersurfaces in Pn (k)). The closed projective set V (F ) ⊂ Pn (k),
wherePF (X0 , X1 , ...Xn ) is an m-form in k[X0 , ..., Xn ], is called a degree-m hypersurface in Pn (k). If we write
F = I∈S aI X I , it is clear that V (F ) is uniquely determined as a closed projective set by the point [: aI ]I∈S
in PN −1 (k), where N = (n+m m ). Thus there is a 1-1 correspondence between degree-m hypersurfaces and P
N −1
.
N
Furthermore, if we let G be the linear functional on k defined by G = ΣI∈S aI YI , then the intersection
V (G) ∩ j(Pn (k)) is precisely:
j({[X0 : X1 : .. : Xn ] : ΣI aI X I = 0}) = j(V (F ))
which shows that the hypersurface V (F ) in Pn (k) can be recovered as a hyperplane section of the Veronese
variety V (a) = j(Pn (k)).
As a consequence, we have the following:

Corollary 8.3.36. Basic open subsets of closed projective sets are affine. That is, if X is a closed projective
set in Pn (k), then DX (F ) = X \ V (F ) is affine. For, by using the Veronese map j, X is isoomorphic to the
closed subset j(X) ⊂ PN −1 (k), and thus DX (F ) is isomorphic to the set j(X) \ V (G), (in the notation of the
example above), which is a closed subset of the affine space PN −1 (k) \ V (G) ' AN −1 (k) (see Remark 8.3.25).
Thus it is isomorphic to an affine closed set, and so DX (F ) = X \ V (F ) is an affine variety, for X a closed
projective set. This answers the question in Remark 8.3.25.

Remark 8.3.37. If one takes a general quasiprojective variety X ⊂ Pn (k), and a homogeneous polynomial
F ∈ k[X0 , ..., Xn ], then it is not true that DX (F ) = X \ V (F ) is affine. For example, if one takes the
quasiprojective variety X = P2 (k)\{[1 : 0 : 0]}, and F = X0 , then DX (F ) = U0 ∩X, which is just A2 (k)\{(0, 0)},
and is not affine by the Example 8.3.21 above.

Exercise 8.3.38 (Structure sheaf of a quasiprojective). Let X ⊂ Pn (k) be an irreducible quasiprojective va-
riety. Define a sheaf OX by declaring:
OX (U ) := {f ∈ k(X) : f is regular on U } = ∩x∈U OX,x
Verify that this defines a sheaf on X, and makes X into a locally ringed space. In contrast with the case
of an irreducible affine closed set, if X is an irreducible projective closed set, we shall see later that OX (X),
the global sections of the sheaf OX or what is the same thing, the k-algebra of functions regular on all of X,
consists only of constants.
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 85

8.4. Products of Quasiprojective Varieties. For the affine spaces An (k), Am (k), their product An+m (k)
is another affine space, with its topology and structure as an affine variety, so nothing needs to be done in
this situation. Note that the coordinate ring of the product, namely k[X1 , .., Xn , Y1 , .., Ym ] is just the tensor
product (as k-algebras) k[X1 , .., Xn ] ⊗k k[Y1 , .., Ym ] of the coordinate rings of An (k) and Am (k) respectively.

For projective spaces, the product is no longer a projective space, but can be given the structure of a closed
projective set. More precisely,

Definition 8.4.1 (Segre Embedding). Denote a point on P(n+1)(m+1)−1 (k) by [: Wij ] where i = 0, 1, ..n and
j = 0, 1, ...m. It is convenient to think of this point as the line defined by the (n + 1) × (m + 1) matrix [Wij ].
Define a map:
j : Pn (k) × Pm (k) → P(n+1)(m+1)−1 (k)
([X0 : ... : Xn ], [Y0 : ... : Ym ]) 7 → [: Xi Yj ]
Since some Xk 6= 0 and some Yl 6= 0, the homogeneous coordinate Xk Yl on the right will be non-zero, so this
map is certainly regular. We need to find its image. Consider the homogeneous ideal defined by:
a = hWij Wkl − Wil Wkj : 0 ≤ i, k ≤ n; 0 ≤ j, l ≤ mi
It is clear that the image of j is contained in V (a). On the other hand, it is easy to define a map ψ :
V (a) →Pn (k) × Pm (k) which is an inverse for j. If, for example, p = [: wij ] is a point on V (a) with, say, the
entry wrs 6= 0, then we may define:
ψ([: wij ]) = ([w0s : .. : wrs : ...wns ], [wr0 : ... : wrs : ... : wrm ])
(picking out the row and column of the matrix [wij ] that contain wrs ). This makes sense regardless of choice
of the non-zero entry wrs because of the relations in the ideal a, which say that if [: wij ] ∈ V (a), all the 2 × 2
minors of the matrix [wij ] are zero, so it is of rank 1, and hence all rows are multiples of one row, and likewise
all columns.

It is easy to show that if X = V (a) ⊂ An (k) and Y = V (b) ⊂ Am (k) are affine closed sets defined by the
radical ideals a and b respectively, then X × Y is an affine closed subset of An+m (k). In fact, if c is the radical
of the ideal in k[X1 , .., Xn , Y1 , .., Ym ] generated by a and b (regarded as subsets of k[X1 , .., Xn , Y1 , .., Ym ] via
the obvious inclusions), then it is easily checked that X × Y = V (c). Note that with this structure of an affine
closed set, the projection p1 : X × Y → X is a surjective and corresponds to the ring homomorphism:
p∗1 : k[X] = k[X1 , .., Xn ]/a → k[X1 , .., Xn , Y1 , .., Ym ]/c = k[X × Y ]
and is therefore regular. Likewise, for the other projection p2 : X × Y → Y .

Exercise 8.4.2. Check that for affine closed sets X and Y , the coordinate ring k[X × Y ] is the k-algebra
tensor product k[X] ⊗k k[Y ].

We again emphasise that the (Zariski) topology on X × Y is not the product of the Zariski topologies on X
and Y . However, if U is open in X and V is open in Y , then the product U × V = p−1 −1
1 (U ) ∩ p2 (V ) is open in
X × Y , since the maps pi are regular, and hence continuous with respect to the respective Zariski topologies,
in view of Proposition 8.3.29. Thus the Zariski topology on X × Y is finer than the product topology.

Next we define the product of two projective closed sets. Let X = V (a) and Y = V (b) be closed projective
sets defined by the homogeneous non-redundant radical ideals a in k[X0 , .., Xn ] and b in k[Y0 , ..., Ym ]. A point:
(x, y) = ([a0 : .. : an ], [b0 : ... : bm ])
n m
of P (k) × P (k) lies in X × Y iff F (a0 , ..., an ) = 0 for each homogeneous F ∈ a and G(b0 , ..., bm ) = 0 for
each homogeneous G ∈ b. This is equivalent to the statement that (x, y) ∈ X × Y iff F (a0 bj , ..., an bj ) = 0 and
G(ai b0 , .., ai bm ) = 0 for each i, j and all F ∈ a, and G ∈ b. From this it follows that X × Y = V (c) where c is
the homogeneous ideal in k[Wij ]0≤i≤n,0≤j≤m generated by the homogeneous polynomials F (W0j , .., Wnj ) for
F ∈ a, G(Wi0 , ..., Wim ) for G ∈ b, and the polynomials Wij Wkl − Wil Wkj for 0 ≤ i, k ≤ n; 0 ≤ j, l ≤ m. This
makes the product of two projective closed sets, more generally two projective varieties, a projective variety.
Thus we have the following proposition:
86 VISHWAMBHAR PATI

Proposition 8.4.3. If X ⊂ Pn (k) and Y ⊂ Pm (k) are quasiprojective varieties, then their product X × Y ,
contained in Pnm+n+m (k), is also a quasiprojective variety.
Proof:
Clearly, X = X1 \ X2 where Xi are closed subsets of Pn (k), and Y = Y1 \ Y2 , where Yi are closed subsets of
P (k). Thus, Xi ×Yj are closed subsets of Pnm+n+m (k), and since X ×Y = (X1 ×Y1 )\[(X2 × Y1 ) ∪ (X2 × Y1 )],
m

it is a difference of two closed projective subsets of Pnm+n+m (k), and hence a quasiprojective variety. 2

Lemma 8.4.4. The closed subsets of various products are as follows:

(i): closed subsets of An (k) × Am (k) are common zero sets of polynomials in n + m variables.

(ii): closed subsets of An (k) × Pm (k) are common zero sets of polynomials in n + m + 1 variables X1 .., Xm ,
Y0 , ..., Ym which are homogeneous in the last m + 1 variables Yi .

(iii): closed subsets of Pn (k) × Pm (k) are common zeros of polynomials in n + m + 2 variables which
are bihomogeneous, i.e. separately homogeneous in the variables Xi and Yj (which is stronger than
homogeneity in all the variables together).

Proof:
(i) is obvious, since An (k) × Am (k) = An+m (k).

For (ii), just note that a subset C of An (k) × Pm (k) is closed iff its intersection with each of the affine pieces
A (k) × Ui where Ui = D(Yi ) ⊂ Pm (k) is closed. That is, C ∩ (An (k) × Ui ) is precisely the common zero set of
n

some polynomials F (X , .., X , Y0 , .., c

j 1 n Yi
Yi
, .. Ym ). But this implies that (x , .., x , y , ..., y ) ∈ C iff it is a zero
Yi Yi 1 n 0 m
of the i-th partial homogenisation in the last m + 1 variables, which we may write as Fej (X1 , .., Xm , Y0 , .., Ym )
for each j. And conversely, if C is the common zero set of such polynomials, its intersection with An (k) × Ui
is a closed subset thereof, for each i = 0, 1, .., m. This proves (ii). The proof of (iii) is similar, and left as an
exercise. 2

Proposition 8.4.5. Let f : X → Y be a regular map of quasiprojective varieties. Then the graph of f , defined
as :
Γf = {(x, y) ∈ X × Y : y = f (x)}
is a closed subset of X × Y .

Proof: We first claim that the diagonal in Y , defined as :

∆Y = {(y, y) ∈ Y × Y }
is closed in Y × Y . For Y = V (a) ⊂ Pn (k) a closed projective set defined by the homogeneous ideal a, this
is clear, since by (iii) of Lemma 8.4.4 above, Y is the common zero set of the bihomogenous polynomials:
F (X0 , ..., Xn ) for F ∈ a and the bihomogeneous polynomials Xi Yj − Yi Xj for i, j = 0, 1, .., n. For an arbitrary
quasiprojective variety Y = Y1 \ Y2 , we note that ∆Y = ∆Y1 ∩ (Y × Y ), which is a closed subset of the open
subset Y × Y of Y1 × Y1 .

Now if f : X → Y is a regular map, it is easy to check that the map:

f × idY : X × Y → Y × Y
is also a regular map, and hence continuous. Γf = (f × idY )−1 (∆Y ) is therefore closed in X × Y . 2
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 87

8.5. Main Theorem of Elimination Theory. We now come to an important proposition about regular
maps from projective varieties, and has to do with eliminating variables. Suppose one has some homogeneous
polynomials {Fi (X0 , .., Xn )}m n
i=0 , and one subjects the point [a0 : a1 : .. : an ] ∈ P (k) to lie in some closed
n
projective set X ∈ P (k). That is [a0 : .. : an ] are subjected to some “constraints”, i.e. required to be
zeros of some homogeneous polynomials {gk (X0 , .., Xn )}rk=1 . Then one wants to know if the image points
[F0 (a0 , .., an ) : ... : Fm (a0 , .., an )] are all the solutions to some polynomials {hj (Y0 , .., Ym )}sj=1 . This amounts
to solving the equations:

Yi − Fi (X0 , .., Xn ) = 0, (0 ≤ i ≤ m)
gk (X0 , .., Xn ) = 0, (1 ≤ k ≤ r)
in the form:
hj (Y0 , .., Ym ) = 0, (0 ≤ j ≤ s)
That is, we want to eliminate (X0 , .., Xn ) from the equations above, and express the general solution as relations
among the Yi ’s. The next proposition answers this in the affirmative.

Proposition 8.5.1 (Main Theorem of Elimination Theory). Let f : X → Y be a regular map, with X a
projective variety, and Y a quasiprojective variety. Then the image of f is a closed subset of Y .
Proof:
Let us say X ⊂ Pn (k) and Y ⊂ Pm (k). Since the inclusion i : Y → Pm (k) is a regular (hence continuous)
map, it is enough to prove it for Y = Pm (k). Also, note that the image of f , for f : X → Pm (k), is the image
of the graph Γf of f under the second projection:

p2 : X × Pm (k) → Pm (k).
So, in view of the Proposition 8.4.5 above, it is enough to show that for X a closed projective set, the above
projection p2 is a closed map. Thus it is enough to show that if C is closed in X × Pm (k), then p2 (C) ∩ Ui is
closed in Ui , (where Ui is the i-th affine piece of Pm (k)), by the Lemma 8.3.26. Since the pieces C ∩ (X × Ui )
are closed in X × Ui , and the images of these pieces are precisely p2 (C)∩ Ui , we may as well just verify that
the second projection:
p2 : X × Am (k) → Am (k)
is a closed map, for X a closed projective set. Since the inclusion i : X ⊂ Pn (k) is a closed embedding
(i.e., it is homoeomorphism onto its image, and its image is closed), it is easy to check that the inclusion
X × Am (k) ⊂ Pn (k) × Am (k) is a closed embedding. Thus it is enough to check that the second projection:
p2 : Pn (k) × Am (k) → Am (k)
is a closed map. Let C ⊂ Pn (k) × Am (k) be a closed set, defined as the common zero set of some polynomials
{Fi (X0 , .., Xn , Y1 , .., Ym )}ri=1 , which are homogeneous in the Xj variables, in accordance with (ii) of Lemma
8.4.4. For each y = (y1 , .., ym ) ∈ Am (k), let ay denote the homogeneous ideal in k[X0 , .., Xn ] generated by
the homogeneous polynomials Fi (X0 , ., Xn , y1 , ., ym ), 1 ≤ i ≤ r. For convenience, denote these polynomials by
Fi (X0 , ., Xn ; y). For each s ≥ 0, define the set:
Ds = {y = (y1 , .., ym ) ∈ Am (k) : ay 6⊃ Is }
Note that since Is ⊇ It for t ≥ s, we have Dt ⊂ Ds for t ≥ s. Now, y ∈ p2 (C) iff the zero set V (ay ) is
non-empty, which is true iff ay does not contain Is for any s ≥ 0, by the Lemma 8.2.7. Thus p2 (C) = ∩s Ds .
We claim that the sets Ds are closed, for s sufficiently large. This will imply that p2 (C) is closed. Choose s
to be greater than or equal to maximum of {dj }rj=1 where dj is the homogeneous X-degree dj of Fj . It is
enough to show that for such an s, the set Dsc of all y ∈ Am (k) such that ay ⊇ Is is open. Enumerate all the
monomials
{X I : I = (i0 , .., in ) with Σij = s}
as the index set Λ. Then, if ay ⊇ Is , we have relations:
X I = Σrj=1 aIj (X0 , .., Xn ; y)Fj (X0 , .., Xn ; y)
for some elements aIj (X0 , ..., Xn ; y) which are polynomials in the X-variables (though not necessarily in the
y-variables), and which we can assume to be homogeneous of degree s − dj in the X-variables, for each I ∈ Λ.
88 VISHWAMBHAR PATI

Expand Fj (X0 , .., Xn ; y) = Σ|J|=dj Fj,J (y)X J , where Fj,J (y) are polynomials in y, and expand aIj (X0 , .., Xn ; y) =
Σ|K|=s−dj aK K K
Ij (y)X , where aIj (y) are some k-valued functions of y. Plugging these into the above equation,
we get:
X I = ΣL∈Λ (Σj,|K|=s−dj ,L≥K aK Ij (y)Fj,L−K (y))X
L

where L ≥ K means that each entry Lk of L is greater than the entry Kk of K. This implies that:
Σj,|K|=s−dj aK
Ij (y)Fj,L−K (y) = δI,L for all I, L ∈ Λ

This means that the matrix A defined by AjK,L (y) := Fj,L−K (y) for L ≥ K, and = 0 otherwise, with rows
indexed by Γ = {jK : |K| = s − dj } and columns indexed by L ∈ Λ has a left-inverse, and so is of of full rank
=N , say. Conversely, if this matrix has a left-inverse for some y, then by reversing all the steps above we get
X I , for I ∈ Λ, as combinations of the Fj (X0 , .., Xn ; y), so that ay ⊇ Is . Thus ay ⊇ Is iff at least one principal
N × N minor of this matrix AjK,L (y) is non-singular. Let us denote the principal N × N minors of AjK,L (y)
by Mβ (y), with β in some finite set S. Thus Dsc = ∪β∈S {y : det(Mβ (y)) 6= 0}. However, det(Mβ (y)) is a
polynomial in the non-zero entries Fj,L−K (y) of AjK,L (y). Since Fj,L−K (y) are polynomials in y, det(Mβ (y))
is a polynomial in y = (y1 , .., ym ) for all β ∈ S. This proves that Dsc is open for all s ≥ maxrj=1 dj , and the
proposition. 2

Remark 8.5.2. Of course, the second projection p2 : An (k) × Am (k) → Am (k) is not a closed map. For
example, for the map p2 : A1 (k) × A1 (k) → A1 (k), the image of the closed hyperbola V (X1 X2 − 1), in A2 (k),
is the subset A1 (k) \ {0}, which is not closed in A1 (k). (Where does the above proof break down) ?

In particular, if we consider the problem of eliminating (X1 , X2 ) from the equations:

Y = X2
X1 X2 − 1 = 0
by writing one or more polynomial relations for Y , we find that it cannot be done. We just get the rather silly
result Y 6= 0. The homogeneity of all the polynomials being considered (i.e. the fact that we are the setting of
closed projective sets as opposed to affine closed sets) is crucial to the truth of the main theorem of elimination
theory.
We now have a generalisation of the Proposition 8.3.15. Namely,
Corollary 8.5.3. Let X be an irreducible projective variety, and Y an affine variety, and let f : X → Y be a
regular map. Then f is a constant.
Proof: It is enough to take X ⊂ Pn (k) an irreducible closed projective set, and Y ⊂ Am (k) an affine closed set.
Since the coordinate functions {yi }m i=1 are regular functions on Y , it is enough to prove that the components
of f , namely the regular functions fi = yi ◦ f , are constant. We will prove every regular map g : X → k is
constant. i.e. k[X] = k for an irreducible closed projective set (and hence irreducible projective variety) X.
Consider the composite j ◦ g : X → P1 (k) where j is the inclusion A1 (k) ⊂ P1 (k) as the affine piece
U0 = D(X0 ). Clearly, j ◦ g is a regular map from X to P1 (k), and its image is contained in U0 , so cannot be
all of P1 (k). By the previous proposition, it is therefore a proper closed subset of P1 (k). But the only proper
closed subsets of P1 (k) are finite sets. Since the point at infinity [0 : 1] is not in the image of g, we have
Im g = {a1 , ..., ar } for some ai ∈ A1 (k) = U1 . This means X = ∪ri=1 g −1 (ai ). Since g −1 (ai ) are closed subsets
of X, which is irreducible, we have r = 1, and Im g = {a1 }, so that g is a constant map. 2

The above corollary is an analogue of the Liouville theorem in complex analysis, which says that a holomor-
phic map on a connected compact complex manifold is a constant.
Corollary 8.5.4. Let X ⊂ Pn (k) be a quasiprojective variety. If X is a projective variety, then X is a closed
subset of Pn (k).
Proof: Since X is a projective variety, there is a closed projective set Y ⊂ Pm (k) with regular a map f : X → Y
having a regular inverse g : Y → X. The composite j ◦ g, where j : X ⊂ Pn (k) is the inclusion, is a regular
map from the closed projective set to Pn (k), so by the foregoing proposition its image, namely X, must be
closed. 2
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 89

Contrast the above corollary with the affine situation, where the affine variety A1 (k) \ {0} = D(X) is not a
closed subset of A1 (k).

Corollary 8.5.5. Let X be a projective variety, and Y any quasiprojective variety. Then the second projection
pY : X × Y → Y is a closed map.

Proof: In the beginning of the proof of Proposition 8.5.1, we showed that the second projection is closed for
the case Y = Pm (k), which reduced to proving it for Y = Am (k). For Y ⊂ Am (k) an affine closed set, a closed
subset C of X × Y , by the Proposition 8.4.4, will actually be a closed subset of X × Am (k), so applying the
earlier case we are done. This implies the result for Y any affine variety. Now, for Y a general quasiprojective
variety, use an affine open covering, and the Proposition 8.3.26. 2

Corollary 8.5.6. In the space P N −1 (k) of all degree m hypersurfaces in Pn (k) described in Example 8.3.35,
the reducible ones form a proper closed subset.


Proof: Let Nk denote the binomial coefficient n+k k . Consider the map :

φk : PNk −1 (k) × PNm−k −1 (k) → PNm −1 (k)

which maps the pair ([: aJ ]|J|=k , [: bK ]|K|=m−k ) to ([: cL ]|L|=m ) where cL = ΣJ+K=L aJ bK . This map is
clearly well defined and regular. The hypersurface V (F ) is reducible iff the homogeneous degree m polynomial
F = Σ|L|=m cL X L is a product of a degree k homogeneous polynomial H = Σ|J|=k aJ X J and a degree
m − k homogeneous polynomial G = Σ|K|=m−k bK X K . In other words, the point [: cL ]|L|=m representing the
hypersurface V (F ) in PNm −1 (k) is in the image of φk for some k = 1, 2, ..., m − 1. But, by the Proposition
m−1
8.5.1 above, these images are all closed, and so is their union ∪k=1 (Im φk ). Since there exist irreducible
hypersurfaces, the subset is proper. This proves the result. 2

We close this subsection with some additional exercises.

Exercise 8.5.7. We saw in Proposition 8.5.1 that the image of a regular map from closed projective set into a
projective space is a closed projective set. On the other hand, if we allow the domain to be just a quasiprojective,
the image need neither be an affine nor a projective variety, indeed need not even be a quasiprojective! For
example, consider the regular map:

f : A2 (k) → P2 (k)
(a, b) 7→ [a : ab : 1]

Compute the image of f and show that it is not a quasiprojective in P2 (k).

Exercise 8.5.8. Let X = V (X02 + X12 + X22 ) ⊂ P2 (k), and consider the rational map:

f :X → P1 (k)
[a0 : a1 : a2 ] 7→ [a20 + a21 : a1 a2 ]

Find the largest open subset U of X on which f is regular, and compute f (U ).

Exercise 8.5.9. Determine the parameter space of all conics (=degree 2 hypersurfaces) in P2 (k) that pass
through 2 distinct given points p1 , p2 ∈ P2 (k).
90 VISHWAMBHAR PATI

9. Plane Projective Curves and Bezout’s Theorem

9.1. Plane projective Curves.
Definition 9.1.1. We call a closed projective set C ⊂ P2 (k) a plane projective curve if C = V (F ) for F a non-
zero homogeneous polynomial in k[X0 , X1 , X2 ] which is square free (i.e. F is a product of distinct irreducible
factors, so that hF i is a radical homogeneous ideal of k[X0 , X1 , X2 ]). For example, the projectivisations of all
the plane affine curves discussed earlier will be plane projective curves. A plane projective curve is said to be
smooth if all its affine pieces are smooth.

Exercise 9.1.2. Verify that a plane projective curve V (F ) is smooth iff the only zero of
 
∂F ∂F ∂F
grad F = , ,
∂X0 ∂X1 ∂X2
in k 3 is the point (0, 0, 0). Show that a smooth plane projective curve is irreducible.

Proposition 9.1.3. Let C be a smooth irreducible plane projective curve. Then any rational map
f : C → Pn (k) is regular.
Proof: By definition, for [x0 : x1 : x2 ] ∈ C, we have the representation:
f ([x0 : x1 : x2 ]) = [P0 (x0 , x1 , x2 ) : .... : Pn (x0 , x1 , x2 )]
where Pi are all homogeneous polynomials of the same degree. Let us consider a point p ∈ C ∩ U0 , say
p = [1 : a : b], and show that f is regular at p. Let pi (X, Y ) be the 0-th dehomogenisations of Pi , i.e.
pi (X, Y ) = Pi (1, X, Y ). Since the affine piece C ∩ U0 is smooth, we have that the polynomial functions
pi (X, Y ) are elements of the affine coordinate ring A = k[C ∩ U0 ], and are certainly regular at p. By 7.3.4,
there is a function t ∈ Amp (in fact t = x or y, from the proof given there) such that pi (x, y) = tdi ui where
ui (p) 6= 0, ui ∈ Amp and x, y are the images of X, Y respectively in A. Writing ui = srii (x,y) (x,y) , where ri , si ∈ A,
we have :
f (x, y) := f ([1 : x : y]) = [td0 r0 s1 s2 ..sn : td1 s0 r1 s2 ..sn : ... : tdn s0 s1 ..sn−1 rn ]
where ri (a, b) 6= 0 for all i and si (a, b) 6= 0 for all i. If dj = min {di }, one can divide out by tdj throughout,
and obtain:
f ([X0 : X1 : X2 ]) = [td0 −dj Q0 : td1 −dj Q1 : ... : Qj : ... : tdn −dj Qn ]
where Qi (X0 , X1 , X2 ) := X0mi (s0 s1 ..ri ..sn )H (the superscript denoting 0-th homogenisation) and mi is defined
by:
mi + deg(s0 s1 ..ri ..sn ) = m
where m =maxi {deg(s0 s1 ..ri ..sn )}. Since:
Qj (p) = Qj ([1 : a : b]) = s0 (a, b)s1 (a, b)..rj (a, b)..sn (a, b) 6= 0
we have a representation for f at p which shows that it is regular at p. This proves the proposition. 2

Corollary 9.1.4. Let φ : C1 → C2 be a birational equivalence between two smooth irreducible plane projective
curves Ci ⊂ P2 (k). Then φ is an isomorphism.

Proof: By definition, we have a rational map φ : C1 → P2 (k), and by the Proposition 9.1.3 above, this map is
regular at all points of C1 . Similarly, φ−1 : C2 → P2 (k) is regular. Since φ is an isomorphism of an open (and
thus dense) subset U of C1 with an open dense subset V of C2 , and φ being regular on C1 makes it continuous,
it follows that φ(C1 ) = φ(U ) ⊂ φ(U ) = V = C2 . Similarly, φ−1 (C2 ) ⊂ C1 , and thus φ and φ−1 are regular
inverses of each other and C1 and C2 are isomorphic curves. 2

Corollary 9.1.5. The projectivised elliptic curve X = V (y 2 z − x(x2 − z 2 )) ⊂ P2 (k) is not a rational curve,
i.e., is not birationally equivalent to P1 (k). In particular it is not birationally equivalent to the projectivised
cubic cusp V (y 2 z − x3 ) or the cubic node V (y 2 z − x2 (x − z)).
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 91

Proof: By the Corollary 9.1.4 above, if X were birationally equivalent to P1 (k), it would be isomorphic to it,
since X is easily checked to be a smooth curve. In that case, the affine piece X0 = V (y 2 − x(x2 − 1)) = X \ {[0 :
1 : 0]} of X would be isomorphic to P1 (k) \ {p} = A1 (k) (where the point p := φ([0 : 1 : 0])). We saw in the
Example 7.3 that the coordinate ring of X0 is not a UFD, whereas that of A1 (k) is k[x], which is a UFD. Thus
X0 cannot be isomorphic to A1 (k), and X cannot be birationally equivalent with P1 (k).

Since the projectivised cubic cusp and projectivised cubic node are birationally equivalent to P1 (k), it follows
that X cannot be birationally equivalent to either of these curves. 2

9.2. The resultant of two polynomials.

Pn Pm
Lemma 9.2.1 (The Resultant). Let p(z) = i=0 ai z i and q(z) = i=0 bi z i be two polynomials of degrees n
and m respectively, in k[z] where k is algebraically closed as always. Then p and q have a common root if and
only if the (m + n) × (m + n) resultant defined as:

 
a0 a1 .. an 0 0 .. 0
 0 a0 a1 .. an 0 .. 0 
 
 0 0 a0 a1 .. an .. 0 
 
 .. .. .. .. .. .. .. .. 
 
 0 0 0 .. a0 a1 .. an 
R(p, q) := det 



 b0 b1 .. .. bm−1 bm .. 0 
 0 b0 b1 .. bm−2 bm−1 bm 0 
 
 0 0 b0 b1 .. bm .. 0 
 
 .. .. .. .. .. .. .. .. 
0 0 0 .. b0 b1 .. bm

(with the first m rows of a’s and the last n-rows of b’s) is 0.

Proof: Let λ be a common root of p and q. Then, factorising out (z − λ), we have two polynomials p1 (resp.
q1 ) of degrees (n − 1) (resp. (m − 1)) satisfying:

p(z) = (z − λ)p1 (z); q(z) = (z − λ)q1 (z)

so that p1 (z)q(z) = q1 (z)p(z), upon eliminating (z − λ). Conversely, if there exist p1 , q1 of degrees (n − 1) and
(m − 1) respectively, satisfying p1 (z)q(z) = q1 (z)p(z), then we may cancel out common linear factors of p1 and
q1 and assume they are coprime of degrees ≤ (n − 1) and ≤ m − 1 respectively. Thus we have have:

p(z) q(z)
=
p1 (z) q1 (z)
= a polynomial of degree ≥ 1 = (say) (az + b)r(z)

and since a 6= 0, we get p(z) = (az + b)r(z)p1 (z) and q(z) = (az + b)r(z)q1 (z) have the common root λ = −b/a.

Pn−1 Pm−1
Now let us write p1 (z) = i=0 ci z i and q1 (z) = i=0 di z i . We substitute this into the relation p1 (z)q(z) =
p(z)q1 (z), and equate coefficients to obtain:

c0 b0 = a0 d0 ⇒ a0 (−d0 ) + b0 c0 = 0
c0 b1 + c1 b0 = a0 d1 + a1 d0 ⇒ a0 (−d1 ) + a1 (−d0 ) + b0 c1 + b1 c0 = 0
.... .. ....
92 VISHWAMBHAR PATI

and so on. These (m + n) equations can be written in matrix form as:

 
a0 a1 .. an 0 0 .. 0
 0 a0 a1 .. an 0 .. 0 
 
 0 0 a0 a1 .. an .. 0 
 
 .. .. .. .. .. .. .. .. 
 
 0 0 0 .. a0 a1 .. an 
[−d0 , ..., −dm−1 , c0 , .., cn−1 ] 
 b0 b1
=0

 .. .. bm−1 bm .. 0 
 0 b0 b1 .. bm−2 bm−1 bm 0 
 
 0 0 b0 b1 .. bm .. 0 
 
 .. .. .. .. .. .. .. .. 
0 0 0 .. b0 b1 .. bm
Clearly, this system of linear equations has a non-zero solution for the vector (−d0 , ..., −dm−1 , c0 , .., cn−1 ) iff
the matrix on the right, viz. is singular. This happens iff its determinant R(p, q) = 0. The lemma follows. 2

Pn
Corollary 9.2.2. Let p(z) = i=0 ai z i be a polynomial of degree n, with an 6= 0. Then p(z) has a repeated
root iff a polynomial ∆(p), which is a polynomial in the ai vanishes. ∆(p) is called the discriminant of p.

P
Proof: It is easy to see that p(z) has a repeated root iff p(z) and its derivative p0 (z) = i iai z i−1 have a
common root. Thus p has a repeated root iff the resultant R(p, p0 ) = 0. This is clearly a polynomial in the
ai ’s. As an exercise, the reader may wish to compute the discriminants of quadratic and cubic polynomials. 2

Pn i
Pm i
Corollary 9.2.3. Suppose f = i=0 ai z and g = i=0 bi z are two polymomials in k[x, y, z] such that
ai = ai (x, y) is a homogeneous polynomial in x, y of degree (n − i), and bi = bi (x, y) is homogeneous of degree
(m − i). Then if R(f, g) 6≡ 0, it is a homogenous polynomial R(x, y) in x, y of degree nm.

Proof: The homogeneity hypotheses of ai and bi imply that ai (tx, ty) = tn−i ai (x, y) and bi (tx, ty) =
tm−i bi (x, y). The determinant expression for R(tx, ty), is therefore:
 
tn a0 (x, y) tn−1 a1 (x, y) .. an (x, y) 0 0 .. 0
 0 tn a0 (x, y) tn−1 a1 (x, y) .. an 0 .. 0 
 
 0 0 tn a0 (x, y) tn−1 a1 (x, y) .. an (x, y) .. 0 
 
 .. .. .. .. .. .. .. .. 
 
 0 0 0 .. tn a0 (x, y) tn−1 a1 (x, y) .. an (x, y) 
det 



 tm b0 (x, y) tm−1 b1 (x, y) .. .. tbm−1 bm .. 0 
 0 tm b0 (x, y) tm−1 b1 (x, y) .. t2 bm−2 (x, y) tbm−1 bm 0 
 
 0 0 tm b0 (x, y) tm−1 b1 (x, y) .. bm (x, y) .. 0 
 
 .. .. .. .. .. .. .. .. 
0 0 0 .. tm b0 (x, y) tm−1 b1 (x, y) .. bm (x, y)
If we multiply the first row by t−n , the second by t−n−1 ,..,the m-th now by t−n−m+1 , the (m + 1)-st row again
by t−m , the (m + 2)-nd by t−m−1 ..., the (m + n)-th row by t−m−n+1 , then we will have a common factor of
t−1 from the second column, of t−2 from the third, and a factor of t−m−n+1 from the last column. Thus we
have:
t−[(n+(n+1)+...+(n+m−1)+(m+(m+1)+...+(m+n−1)] R(tx, ty) = t−1−2+...−(m+n−1) R(x, y)
That is
1 1
t− 2 [m(2n+m−1)+n(2m+n−1)] R(tx, ty) = t− 2 [(m+n)(m+n−1)] R(x, y)
which implies that t−mn R(tx, ty) = R(x, y), i.e. that R(tx, ty) = tmn R(x, y). Thus R is homogenous of degree
mn in (x, y), and the corollary follows. 2

Now we can prove Bezout’s Theorem.

 LECTURES ON BASIC ALGEBRAIC GEOMETRY 93

Theorem 9.2.4 (Bezout’s Theorem). Let C and D be two projective plane curves, and assume that C and
D have no irreducible components in common. Then if C has degree n (i.e. C = V (F ) where F (X0 , X1 , X2 ) is
an n-form), D is of degree m, the set C ∩ D is a finite set of cardinality nm, provided each point of intersection
is counted with its multiplicity (to be explained in the proof).

Proof: First note that if C = V (f ) and D = V (g) are distinct irreducible P affine plane curves in A2 (k), then
n
f (x, y) and
Pmg(x, y) are not scalar multiples of each other. Write f (x, y) = i=0 ai (x)y i with an (x) 6= 0, and
i
g(x, y) = i=0 bi (x)y with bm (x) 6= 0. If (λ, µ) ∈ C ∩ D, it follows that the 1-variable polynomials f (λ, y) and
g(λ, y) have the common root µ. Thus, from the Lemma9.2.1 it follows that the resultant R(f (λ, −), g(λ, −)) =
0. If we consider the polynomial in x defined by G(x) := R(f (x, −), g(x, −)), this polynomial G is not identically
zero since C 6= D, and hence has only finitely many roots. For each of these finitely many roots λi , there are
only finitely many common roots of f (λi , −) and g(λi , −). Thus C ∩ D is a finite set.

Now if C and D are any two affine plane curves with no irreducible components in common, then the
cardinality of C ∩ D is bounded by the sum of the cardinalities of Ci ∩ Dj , where Ci and Dj are the irreducible
components of C and D respectively. And for each i, j, since Ci 6= Dj , we have that Ci ∩ Dj is a finite set, by
the last paragraph, so C ∩ D is also finite.

Finally, if C and D are two plane projective curves, with no common irreducible components, then by
applying the above affine case to the affine plane curves Ui ∩ C and Ui ∩ D, we find that C ∩ D is a finite set.
Define S := C ∩ D.

By a line in P2 (k), we mean a linear subspace of the kind V (F ), where F 6= 0 is a 1-form. Since both C
and D have finitely many irreducible components, we may choose a line L such that L is not an irreducible
component of either C or D.

Given any pair p, q of distinct points in P2 (k), there is a unique line containing both p and q. Consider the
lines Lpq joining p, q ∈ S for p 6= q, and note that these are finitely many lines. We can choose a point P such
that P does not lie on L, does not lie on C, does not lie on D, and also does not lie in Lpq for each pair of
points p, q ∈ S with p 6= q. We can define a coordinate system (X0 , X1 , X2 ) (by a linear change of coordinates
if necessary) so that P = [1 : 0 : 0], and L = V (X0 ).

Now let us write:

F (X0 , X1 , X2 ) = a0 (X1 , X2 ) + a1 (X1 , X2 )X0 + .... + an (X1 , X2 )X0n
G(X0 , X1 , X2 ) = b0 (X1 , X2 ) + b1 (X1 , X2 )X0 + .... + bm (X1 , X2 )X0m
where ai (X1 , X2 ) is homogeneous of degree (n−i) and bi (X1 , X2 ) is homogeneous of degree (m−i). Thus we are
in the setting of the Corollary 9.2.3. Note that if the scalar an = 0, then since ai (0, 0) = 0 for 0 ≤ i ≤ (n − 1),
we would have P = [1 : 0 : 0] in C, contrary to the choice of P . Hence an 6= 0. Similarly bm 6= 0, so F and
G are polynomials of degree n and m in the first variable X0 . Again, if a0 ≡ 0, we would have L = V (X0 )
becoming an irreducible component of C, which contradicts the choice of L. So a0 6≡ 0. Similarly b0 6≡ 0.

Now consider the resultant R(X1 , X2 ) = R(F (−, X1 , X2 ), G(−, X1 , X2 ). If this were identically zero, then
for every choice of (a, b) 6= (0, 0), we would have that there exists a common root λ(a, b) of F (−, a, b) and
G(−, a, b). This gives infinitely many points [λ(a, b) : a : b] ∈ C ∩ D, contradicting that C ∩ D is a finite set.

Thus, by the Corollary 9.2.3, R(X1 , X2 ) is a non-zero homogeneous polynomial of degree nm, and for each
[c : a : b] ∈ C ∩ D = S, we first observe that (a, b) 6= (0, 0), since [c : 0 : 0] = [1 : 0 : 0] = P 6∈ C ∩ D. Thus for
each [c : a : b] ∈ S, we have [a : b] ∈ P1 (k) is a point on V (R), by the Lemma9.2.1. Thus we have defined a
map τ : S → V (R) which takes [c : a : b] ∈ S to [a : b] ∈ V (R).

This map τ is surjective, for if the point [a : b] ∈ P1 (k) is a point on V (R), then F (−, a, b) and G(−, a, b)
have a common root c, and [c : a : b] ∈ S, again by Lemma 9.2.1. Then τ ([c : a : b]) = [a : b].

Finally, τ is injective. For if there are two distinct points p = [c1 : a : b] and q = [c2 : a : b] which lie in
S = C ∩ D, we have that p 6= q, and the line Lpq is the line V (aX2 − bX1 ). Since the point P = [1 : 0 : 0]
94 VISHWAMBHAR PATI

clearly satisfies the equation aX2 − bX1 = 0, we have that P ∈ Lpq , contrary to choice of P . Thus the map
τ : S → V (R) defined by [c : a : b] 7→ [a : b] is a bijection of sets.

Now since R(X1 , X2 ) is a non-zero homogeneous polynomial of degree nm, V (R) ⊂ P1 (k) has exactly nm
points, where each point is counted with its multiplicity. Associate to the point p = [c : a : b] ∈ S the
multiplicity of [a : b] as a root of R. Then, by the last para, the cardinality of S, counting each point with this
multiplicity, is exactly nm. Hence the proposition. 2

Remark 9.2.5. We note that the proof above associated to each point in S = C ∩ D the multiplicity of
a certain corresponding point in V (R) after a choice of good coordinate system was made. It is not clear
that these multiplicities are independent of the choice of coordinate system. There are both algebraic and
topological ways of defining these multiplicities which establish their independence of the choice of coordinate
system. For details, the reader may look up the proof of the general Bezout Theorem given in the chapter on
intersection theory in Shafarevich’s Basic Algebraic Geometry. We will return to this point soon.

Corollary 9.2.6. Every projective plane curve C intersects every other projective plane curve D. If C is a
line in the projective plane, then either C is an irreducible component of D, or intersects it in at most m points.

Proof: If C and D are two projective plane curves of degrees n and m respectively, then if they have a common
irreducible component, they cetainly intersect. If they don’t have a common irreducible component, by the
Bezout theorem above, the cardinality of their intersection counted with multiplicity, is nm, and hence non-
zero. If C is line, it is a projective plane curve V (F ) where F is a degree 1 polynomial, so of degree 1, and
hence C is irreducible. Hence it is either an irreducible component of D, or meets it in at most m points. 2

Example 9.2.7. If Li = V (ai X0 + bi X1 + ci X2 ) with (ai , bi , ci ) 6= (0, 0, 0) and i = 1, 2 are two distinct lines
in P2 (k), they intersect in the unique point
[b1 c2 − b2 c1 : a2 c1 − a1 c2 : a1 b2 − a2 b1 ]
Note that all three of the homogeneous coordinates above cannot be zero since we have assumed L1 and L2 to
be distinct lines.

9.3. Multiplicity of a point on a plane curve. We now define an important geometric notion for a point
on a plane curve. It is enough to look at plane affine curves to define this notion, and then carry it over to
projective plane curves by looking at affine pieces.

Let f (x, y) be a polynomial in k[x, y], of degree m. Assume that f (0, 0) = 0, and let P = (0, 0) be the origin
of A2 (k). Thus the curve V (f ) passes through P . Let us write f as:
f (x, y) = f1 (x, y) + f2 (x, y) + .... + fm (x, y)
where fi is a homogeneous polynomial of degree i. Note f0 = 0 since f (0, 0) = 0.

Now let L := L(a,b) = {(ta, tb) : t ∈ k} be any line passing through the origin in A2 (k), where (a, b) 6= (0, 0).
We substitute x = ta and y = tb in f (x, y) = 0 to get the equation:

f (ta, tb) = tf1 (a, b) + t2 f2 (a, b) + .... + tm fm (a, b)

We define:
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 95

Definition 9.3.1. The order of contact of the line L(a,b) with the curve C := V (f ) at the origin P is defined
to be the highest power of t dividing f (ta, tb). It is clearly equal to d, where d is the smallest j for which
fj (a, b) 6= 0. We will denote it as oP (L, C). If a line L does not pass through P , we set oP (L, C) = 0. Note
that by definition, 0 ≤ oP (L, C) ≤ m for any line L = V (ax + by + c) in A2 (k).

Finally, we define the multiplicity of P on C to be:

mP (C) = inf{oP (L, C) : L is a line passing through P }
Again, by definition 1 ≤ mP (C) ≤ m. If P is a point with mP (C) = 1, we say P is a simple point on C, if it
is 2, a double point and in the extreme case it is m, an m-ple point.

What we are doing in defining the multiplicity is viewing the line L through P as affine 1-space A1 (k), and
regarding the restriction of the polynomial f (x, y) to L as a regular function on L, and taking the order of the
zero of this regular function at P ∈ L.

Remark 9.3.2.
(i): If P = (c, d) is not the origin, but some other point on C, we can again define the order of contact
of a line through P with C, and the multiplicity of P on C, by a translation of coordinates. That is,
substitute x = c + ta, y = d + tb in the equation f (x, y) = (0, 0), and again look at the highest power of
t which factors out of the resulting polynomial of degree m in t, etc.

(ii): If C is a plane projective curve in P2 (k), and P is a point on C, we again define for a line L in P2 (k)
through P the order of contact oP (L, C) by taking an affine piece Ci := C ∩ Ui of C in which P lies, and
similarly Li := L∩Ui (where Ui = {[X0 : X1 : X2 ] ∈ P2 (k) : Xi 6= 0}), and setting oP (L, C) = oP (Li , Ci ).
Similarly, define mP (C) = mP (Ci ). Again it is easy to check that these definitions do not depend on
which affine piece containing P we choose.

Example 9.3.3. For example let us consider the three affine cubics we encountered in §7. For P = (0, 0) on
C = V (y 2 − x3 ), we note that every line L = L(a,b) = {(ta, tb) : t ∈ k} with b 6= 0 has oP (L, C) = 2. When
b = 0, the line is the x-axis V (y), and for this line oP (L, C) = 3. Thus mP (C) = 2. All other points P 6= (0, 0)
on C have mP (C) = 1. For the node V (y 2 − x2 (1 + x)), and P = (0, 0), the f (ta, tb) = t2 (b2 − a2 − a3 t). Thus
if b 6= ±a, we have oP (L(a,b) , C) = 2. On the other hand, for L = L(a,±a) , a 6= 0, we have oP (L, C) = 3. Again
mP (C) = 2, and for all P 6= (0, 0), we have mP (C) = 1. Finally for the elliptic curve V (y 2 − x(x2 − 1)), we
leave it to the reader to see that mP (C) = 1 for all P ∈ C. At each of these points P ∈ C, there is exactly
one line (the tangent line at P ) whose order of contact with C at P is ≥ 2, and all other lines through P have
order of contact 1 at P .

Proposition 9.3.4. Let P ∈ C, where C is an affine or projective plane curve. Then mP (C) = 1 (i.e. P is a
simple point on C) iff P is a smooth point on C. In this case there is a unique line L through P whose order
of contact at P is ≥ 2, and this line is the tangent to C at P . If P is a simple point and the order of contact
of the tangent line at P to C is ≥ 3, we call P a point a point of inflexion on C.

Proof: We will just do the affine case, since the projective case follows from the affine case by the definitions.
As before, we may assume P = (0, 0) ∈ C = V (f ), and write:
f = tf1 (a, b) + t2 f2 (a, b) + ... + tm fm (a, b)
If mP (C) = 1, there is a line L(a,b) = {(ta, tb) : t ∈ k} which has order of contact 1 with C at (0, 0). That is
f1 (a, b) 6= 0. But in the linear term f1 (x, y) = αx + βy of f , we have α = ∂f ∂f
∂x (0, 0) and β = ∂y (0, 0). Thus
∂f ∂f
f1 (a, b) = αa + βb = (0, 0) a + (0, 0) b 6= 0
∂x ∂y
implies that ( ∂f ∂f
∂x , ∂y )(0, 0) 6= (0, 0), and P = (0, 0) is a smooth point on C. Further, if for some line L = L(a,b)
∂f ∂f
we have oP (L, C) ≥ 2, it will follow that for this (a, b) we have f1 (a, b) = ∂x (0, 0) a + ∂y (0, 0) b = 0, which
96 VISHWAMBHAR PATI

implies that L(a,b) is the tangent line to C at (0, 0) from elementary calculus. It is unique because it is defined
by the equation ∂f ∂f
∂x (0, 0) x + ∂y (0, 0) y = 0.

Conversely, if mP (C) ≥ 2, then for every line L(a,b) through P = (0, 0), we have order of contact oP (L(a,b) , C)
greater than or equal to 2, so that f1 (a, b) = a ∂f ∂f
∂x (0, 0)+b ∂y (0, 0) = 0 for all (a, b) 6= (0, 0), which clearly implies
that (grad f )(0, 0) = ( ∂f ∂f
∂x , ∂y )(0, 0) = (0, 0) and hence P = (0, 0) is a singular point on C. This proves the
proposition. 2

Corollary 9.3.5. If C is smooth affine plane curve, then all of its irreducible components are pairwise disjoint.
If C is a smooth projective plane curve, then it is irreducible.

Proof: Let C = V (f ) be an affine plane curve, and let P ∈ C1 ∩ C2 , where C1 = V (g) and C2 = V (h) are
two intersecting irreducible components of C. Then we have f is divisible by gh (by the UFD property of
k[x, y]), and for any line L through P , we will have oP (L, C) ≥ oP (L, C1 ) + oP (L, C2 ). From this it follows that
mP (C) ≥ mP (C1 ) + mP (C2 ). Since P ∈ C1 and P ∈ C2 , we have mP (Ci ) ≥ 1 for i = 1, 2, so that mP (C) ≥ 2.
By the previous Proposition 9.3.4 P is a singular point on C.

If C is a projective plane curve with any pair of distinct irreducible components C1 and C2 , by the Corollary
9.2.6, they must intersect nontrivially. Take P in this intersection, an affine open Ui containing P , and apply
the above affine case to the i-th affine pieces of C1 and C2 to conclude that C must be singular at P . 2

Proposition 9.3.6. Let L be any line in P2 (k), and C = V (F ) be a projective plane curve, where deg F = m.
Then, if L is not an irreducible component
P Ps of C, we know by Bezout that C ∩ L = {P1 , .., Ps }. In this event,
s
i=1 oPi (L, C) = m. Also we have i=1 m Pi (C) ≤ m.

Proof: That the intersection C ∩ L is a finite set follows from Bezout’s theorem 9.2.4, since we are assuming
that L is not an irreducible component of C. By a linear change of coordinates assume Pm that this line is
L = V (X0 ). Thus X0 does not divide F . Then it is obvious that the resultant of F = i=0 ai (X1 , X2 )X0i and
G = X0 is the determinant of the (m + 1) × (m + 1) matrix:
 
a0 a1 .. am 0
 0 1 0 0.. 0 
 
 0 0 .. 0 0 
 
 .. .. .. .. .. 
0 0 0 .. 1
since b0 = 0 and b1 = 1 in the notation of Corollary 9.2.3. But this determinant is precisely a0 (X1 , X2 ), which
is the homogeneous degree m polynomial F (0, X1 , X2 ).
Ps
In the proof of the Bezout theorem, we had seen that i=0 ν(Pi ) = m, where ν(Pi ) is the multiplicity of
the root (ai , bi ) in the resultant above, for each Pi = [0 : ai : bi ].PBut since the resultant is F (0, X1 , X2 ), we
have that ν(Pi ) = oPi (L, C), where L = V (X0 ). This proves that i oPi (L, C) = m. The inequality of the last
statement follows from the fact that oPi (L, C) ≤ mPi (C) for each i by definition. The proposition follows. 2

Remark 9.3.7. As one would expect, the first (equality) part of the proposition above is false for affine plane
curves. For example, if we look at C = V (y 2 − x3 ), and consider its intersection with the line L = V (x) which
is the y − axis, we have L meeting C only at P = (0, 0), and also putting (x, y) = (0, t) a general point on L
into f (x, y) = y 2 − x3 gives t2 , so that the sum on the left side is oP (L, C) = 2. On the other hand m = 3.

If we projectivise this curve to C = V (Y 2 Z − X 3 ), we find that the projectivisation of the line x = 0 is the
line X = 0 in P2 (k). Another point Q = [0 : 1 : 0] in the hyperplane at infinity V (Z) occurs in C ∩ L. At this
point, one checks that the order of contact is 1, and now we have oP (L, C) + oQ (L, C) = 2 + 1 = 3. All part
of the magic of projective space!
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 97

Finally, we note that the second inequality of the Proposition 9.3.6 can be a strict one, even in projective
space. For if L is a line tangent to C at say P1 , and P1 is a smooth point on C, we will have oP1 (L, C) ≥ 2
which is strictly greater than mP1 (C) = 1, since P1 is a smooth point.

Corollary 9.3.8. Let C = V (F ) be a plane projective curve of degree m (viz. deg F = m). Then the number
of singular points on C is finite. If m = 1, C has no singular points. If m = 2, C can have at most one
singular point, and in that case C is a pair of intersecting straight lines and their point of intersection is the
only singularity. If m = 3, and C is irreducible, then again C can have at most one singular point.

Proof: Let C = ∪ri=1 Ci be the irredundant decomposition of C into its irreducible components. From the
Proposition 9.3.4, a point P ∈ C is singular iff mP (C) ≥ 2. If P is a point of Ci which is not on any other
Cj 6= Ci , then mP (C) = mP (Ci ), so that P is a singular point of Ci . On the other hand, if P ∈ Ci ∩ Cj for
i 6= j, we have seen in the Corollary 9.3.5 that P is a singular point. Thus we have that the set Σ(C) of C is
the union of the singular points Σ(Ci ) and the union of Ci ∩ Cj for i 6= j. The second set is finite by Bezout’s
Theorem. So we are reduced to showing that for an irreducible plane projective curve C, its singular set Σ(C)
∂F
is finite. So if C = V (F ), with F irreducible, we consider the polynomials Gi := ∂X i
for i = 0, 1, 2. All of them
cannot be identically 0, because that would mean F ≡ 0. So say G0 6≡ 0, and is homogeneous of degree m − 1.
Since F is irreducible, V (G0 ) and C = V (F ) have no common irreducible components. By Bezout C ∩ V (G0 )
is a finite set. Since the set of singular points of C is the set C ∩ V (G0 ) ∩ V (G1 ) ∩ V (G2 ) ⊂ C ∩ V (G0 ), we
have that this set is finite. This proves the first part of the proposition.

Clearly if m = 1, every point P of C has multiplicity 1 ≤ mP (C) ≤ m = 1, and thus by Proposition 9.3.4,
every point on it is smooth.

If m = 2, then there are two cases. In the first, when C is reducible, i.e. the degree 2 polynomial F
defining C = V (F ) is reducible, we have that F must be a product of two distinct degree 1 polynomials. Thus
C = L1 ∪ L2 is the union of two distinct lines, these lines must intersect exactly at one point, by Example
9.2.7, so that C is a pair of straight lines intersecting at a point.

If m = 2, and C is irreducible, we claim that C must be smooth. For let P ∈ C is any point of C. Let Q be
another point of C, with P 6= Q. Clearly there is a line L passing through P and Q, and since C is irreducible,
L is not an irreducible component of C. We know by Bezout that the set L ∩ C can have cardinality at most 2,
so that L ∩ C = {P, Q}. By the second part of the Proposition 9.3.6 that mP (C) + mQ (C) ≤ 2. However, since
both mP (C) and mQ (C) are ≥ 1 by definition, they are both exactly one. By Proposition 9.3.4, P is a simple
point. Since P is arbitrary, C is smooth.
P2 In fact, a little linear algebra and changes of coordinates shows that
we can express F (X0 , X1 , X2 ) = i=0 Xi2 (we are assuming k algebraically closed, and char k = 0 here).

In the case m = 3, and C irreducible, let there be two singular points P and Q in C, if possible. Then the
line L joining P and Q is not an irreducible component of C, since C is irreducible. By the second inequality
of 9.3.6, it follows that mP (C) + mQ (C) ≤ 3. If P and Q are both singular points, then by 9.3.4, we have both
mP (C) and mQ (C) ≥ 2, which is a contradiction. The proposition follows. 2

9.4. Points of Inflexion. From now on, for the rest of this section, we will assume that k is algebraically
closed and of characteristic zero. Our aim is to completely classify plane projective curves of degree ≤ 3. We
first need a criterion to determine when a point on a smooth curve is a point of inflexion.

Proposition 9.4.1. Let C = V (F ) ∈ P2 (k) be a plane projective curve, with F (X, Y, Z) ∈ k[X, Y, Z] a
homogeneous polynomial of degree m ≥ 1. Let P ∈ C be a smooth point of C. Then P is a point of inflexion
iff the determinant of the hessian matrix at P defined by:
 
FXX (P ) FXY (P ) FXZ (P )
HP (F ) := det  FY X (P ) FY Y (P ) FY Z (P ) 
FZX (P ) FZY (P ) FZZ (P )
∂2F
is zero. (Here FXX (P ) denotes ∂X 2 (P ) etc. ).
98 VISHWAMBHAR PATI

Proof: We recall from the Proposition 9.3.4 that P is a point of inflexion if it is a smooth (=simple ⇔
mP (C) = 1) point, and if the tangent line L at P has order of contact oP (L, C) ≥ 3.

It is an easy exercise to check that if we make a linear change of coordinates (X, Y, Z) 7→ A(X, Y, Z) =:
(X1 , Y1 , Z1 ) where A ∈ GL(3, k) is a nonsingular (3 × 3) matrix, then the old coordinate partials are related
to the new ones by
∂ ∂ ∂ ∂
= A11 + A21 + A31
∂X ∂X1 ∂Y1 ∂Z1
and so on, so that the old hessian determinant is related to the new one by HP (F ) = (det A)2 HP1 (F1 ), where
F1 (X1 , Y1 , Z1 ) = F (X, Y, Z). Since A is nonsingular, HP (F ) = 0 iff HP1 (F1 ) = 0, where P1 = A.P . Thus we
may linearly change coordinates and assume that the point P = [0 : 0 : 1], and the tangent line is L = V (Y ).
We may also ensure that Z is so chosen that V (Z) is not an irreducible component of C.

The affine piece of the curve C2 := C ∩ U2 is defined by the dehomogenisation f (x, y) of F (X, Y, Z) by
setting Z = 1. That is f (x, y) = F (X, Y, 1). The point P = [0 : 0 : 1] ∈ C ∩ U2 is a smooth point of C, so the
point Q = (0, 0) ∈ V (f ) is a smooth point of C2 = V (f ), by definition. The affine piece L2 = L ∩ U2 of the
line L is the line L ∩ U2 = V (y), the x-axis. Now, by the fact that L2 is tangent to C2 = V (f ) at (0, 0), we
have that:
f (x, y) = αy + f2 (x, y) + ... + fm (x, y)
where α ∈ k and α 6= 0 by the fact that (0, 0) is a smooth point, and fd is homogeneous of degree d in x, y.
Thus we have, on homogenisation, and the fact that V (Z) is not an irreducible component of C, that
F (X, Y, Z) = αZ m−1 Y + Z m−2 f2 (X, Y ) + ... + fm (X, Y )
and that fm (X, Y ) 6≡ 0. Now one can calculate:

FXX (0, 0, 1) = f2,XX (0, 0) + ....fm,XX (0, 0) = fxx (0, 0)

FXY (0, 0, 1) = FY X (0, 0, 1) = f2,XY (0, 0) + .. + fm,XY (0, 0) = fxy (0, 0)
FY Y (0, 0, 1) = f2,Y Y (0, 0) + ... + fm,Y Y (0, 0) = fyy (0, 0)
FXZ (0, 0, 1) = FZX (0, 0, 1) = (m − 2)f2,X (0, 0) + (m − 3)f3,X (0, 0) + .. + fm−1,X (0, 0) = 0
FY Z (0, 0, 1) = (m − 1)α + (m − 2)f2,Y (0, 0) + .. + fm−1,Y (0, 0) = (m − 1)α
FZZ (0, 0, 1) = α(m − 1)(m − 2).0 + (m − 2)(m − 3)f2 (0, 0) + ... + fm−2 (0, 0) = 0
(Note that for any homogeneous polynomial g of degree ≥ 1, we have g(0, 0) = 0, which is why fr (0, 0) =
fr,X (0, 0) = fr,Y (0, 0) = 0 for all r ≥ 2.) Now the hessian matrix at P = [0 : 0 : 1] is:

 
fxx (0, 0) fxy (0, 0) 0
 fyx (0, 0) fyy (0, 0) (m − 1)α 
0 (m − 1)α 0

whose determinant HP (F ) = −(m − 1)2 α2 fxx . Since α 6= 0, HP (F ) = 0 iff fxx (0, 0) = 0.

From the equation f (x, y) = αy + f2 (x, y) + ..fm (x, y), we note on substituting the parametric point (t, 0) on
the tangent L2 that f (t, 0) = t2 f2,xx (0, 0) + t3 g(t), so that L2 has order of contact oQ (L2 , C2 ) ≥ 3 at Q = (0, 0)
if and only if f2,xx (0, 0) = fxx (0, 0) = 0. Thus P is a point of inflexion of C iff HP (F ) = 0 and the proposition
is proved. 2

Corollary 9.4.2. The proof of the Proposition 9.4.1 above also yields the following criterion for (0, 0) to be a
smooth inflexion point of the affine curve C = V (f ) to have the x-axis V (y) as a tangent at P with order of
contact ≥ 3, i.e. that f (0, 0) = fx (0, 0) = fxx (0, 0) = 0 but fy (0, 0) 6= 0. This condition may be familiar from
single variable calculus.
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 99

Corollary 9.4.3. Let C ⊂ P2 (k) be a projective plane curve. If C is of degree 1, i.e. a line, every point on it
is a point of inflexion. If C is of degree 2 (i.e. a conic), then it has no point of inflexion iff it is irreducible. If
it is reducible (i.e. a union of two lines), then every smooth point is a point of inflexion. If C is of degree 3,
then there exists at least one point of inflexion.

Proof: If C = V (F ) is a degree 1 curve, i.e. a line, F is a degree 1 homogeneous polynomial, so HP (F ) ≡ 0.

If C is a conic, denote the determinant HP (F ) by A. If write (X, Y, Z) ∈ k 3 , as a column vector P , we have

F (X, Y, Z) = P t AP , in matrix notation. Also we get a bilinear form B(P, Q) = P t AQ on k 3 . By elementary
linear algebra, A is singular iff the map P 7→ B(P, −) is non- invertible, i.e. has a kernel. So if HP (F ) = 0, we
have A is singular, and hence there is a 0 6= P ∈ k 3 such that B(P, −) is the zero map. Since the derivative
(FX (P ), FY (P ), FZ (P )) = B(P, −) = P t A (as a row vector), we have that P is a singular point on C. In the
Proposition 9.3.8 above, we have seen that this implies that C is reducible, i.e. is a pair of intersecting straight
lines. In that case, every smooth point is a point on exactly one of these two lines, and is a point of inflexion
by the last para. Thus if C is irreducible, A is a non- singular matrix, and HP (F ) = det A is not zero and
there are no inflexion points.

For a cubic polynomial F (X, Y, Z), the polynomials FXX , FXY etc. are all degree 1 homogeneous polynomi-
als, so the hessian determinant HP (F ) is a cubic polynomial in P = (X, Y, Z). If this polynomial is identically
zero (as can happen if for example F (X, Y, Z) = X 3 + Y 3 ), then certainly every smooth point on C is a point
of inflexion. If it is not identically zero, the curve D := V (HP (F )) is a curve of degree 3 in P2 (k). By the
Bezout theorem, the intersection C ∩ V (HP (F )) ⊂ P2 (k) is non-empty, and we are done. (Note that in this
last case, there are at most nine points of inflexion). 2

Exercise 9.4.4. Let C ⊂ P2 (k) be a plane projective curve of degree m. Assuming that the set of inflexion
points in C is finite, show that this set has cardinality at most 3m(m − 2).

9.5. Classification of plane projective curves of degree ≤ 3. A degree 1 curve in P2 (k) is a line. All
lines in P2 (k) are isomorphic, via a linear coordinate change (i.e. an automorphism of P2 (k)) to V (X0 ).

For a degree two curve there are two possibilities. Either it is reducible, and equal to a pair of lines (not
necessarily distinct), or it is irreducible and smooth (by the Corollary 9.3.8). In this case, C = V (F ) where F
is a non-degenerate quadratic form. The reader may show (using the fact that k is of characteristic zero and
P2
algebraically closed) that a linear change of coordinates brings F to the normal form F (X0 , X1 , X2 ) = i=0 Xi2 .
Thus every smooth plane quadric is equivalent to V (X02 + X12 + X22 ).

The classification of cubics falls into three cases. We disregard the first case when it is reducible, because
it has irreducible components of the kind considered above, i.e. lines and quadrics. The second case is that
of singular irreducible cubics, and the third of smooth irreducible cubics. The second case is easier to handle,
and is the content of the next proposition.
Proposition 9.5.1 (Singular cubics). Let C be an irreducible singular cubic curve in P2 (k). Then, after a
linear coordinate change in P2 (k), we have that C = V (Y 2 Z − X 3 ) (cusp) or C = V (Y 2 Z − X 2 (X − Z)) (node).

Proof: From the Proposition 9.3.8, we see that if the irreducible cubic C is singular, it has exactly one singular
point. Let us linearly change coordinates so that this singular point is P = [0 : 0 : 1]. Since P is a singular
point of C, we have by the Proposition 9.3.4 that mP (C) ≥ 2. Now, write the U2 affine piece of C, viz
C2 = C ∩ U2 = V (f ) by setting Z = 1 in F (X, Y, Z) to get:
f (x, y) := F (X, Y, 1) = f2 (x, y) + f3 (x, y)
(P ∈ C2 implies f (0, 0) = 0, and P a singular point of C2 implies f1 (x, y) = fx (0, 0)x + fy (0, 0)y ≡ 0). Since
C is irreducible, and P ∈ C, it follows that (0, 0) ∈ C2 , and so C2 6= φ, and is also irreducible. In particular
f2 (x, y) 6≡ 0 (otherwise f = f3 (x, y) and V (f ) is the union of 3 lines meeting at (0, 0) and thus reducible).
Hence f2 (x, y) = (αx + βy)(γx + δy) with (α, β) 6= (0, 0) and (γ, δ) 6= (0, 0). Two cases arise:
100 VISHWAMBHAR PATI

Case 1: [α : β] = [γ : δ]. In this case we have f2 = λ(αx + βy)2 , λ 6= 0, and there is exactly one line
L = V (αx + βy) through the origin whose order of contact oP (L, C2 ) = 3. In this event let us make a linear
change of coordinates (x, y) (which corresponds to a linear change of the homogeneous coordinates X, Y, Z in
P2 (k) such that Z 7→ Z) so that L = V (y). Then:
f (x, y) = λy 2 + f3 (x, y)
We may rescale y so that we have f (x, y) = y 2 + f3 (x, y). If the coefficient of x3 in f3 (x, y) were zero, L = V (y)
would become an irreducible component of C2 , contradicting the irreducibility of C2 , so this coefficient is
non-zero. Rescaling x and leaving y as it is we may assume that this coefficient is 1. Thus
f (x, y) = y 2 + (x3 + ax2 y + bxy 2 + cy 3 )
which on homogenising, reads:
F (X, Y, Z) = Y 2 Z + (X 3 + aX 2 Y + bXY 2 + cY 3 )
By a linear change X 7→ X + λY , Y 7→ Y and Z 7→ Z, we can get rid of the X 2 Y term. Thus we have:
F (X, Y, Z) = Y 2 Z + (X 3 + bXY 2 + cY 3 )
where b and c are not the old b and c. This last equation maybe rewritten as:
F (X, Y, Z) = Y 2 (Z + bX + cY ) + X 3
Now set Z1 = −(Z + bX + cY ), to obtain F (X, Y, Z1 ) = X 3 − Y 2 Z1 . This is precisely the projective cusp.

Case 2: [α : β] 6= [γ : δ]. In this case we make a linear change of coordinates so that αx + βy 7→ x and
γx + δy 7→ y. In this case then:
f (x, y) = xy + f3 (x, y)
Again since f is irreducible, the coefficients of x3 and y 3 in f3 are both non-zero. We can rescale x and y so
that f2 becomes λxy, λ 6= 0 and f3 (x, y) = x3 + ax2 y + bxy 2 + y 3 . On homogenising we have:
F (X, Y, Z) = λXY Z + (X 3 + aX 2 Y + bXY 2 + Y 3 )
= λXY Z + (X + Y )3 + (a − 3)X 2 Y + (b − 3)XY 2
= XY (λZ + (a − 3)X + (b − 3)Y ) + (X + Y )3
= XY Z1 + (X + Y )3

 
Now note XY = 41 (X + Y )2 − (X − Y )2 , so changing to (X + Y ) 7→ X, X − Y 7→ Y , and 4Z1 7→ −Z, we
have F (X, Y, Z) = X 3 − (X 2 − Y 2 )Z = −(Y 2 Z − X 2 (X − Z)), and thus C = V (Y 2 Z − X 2 (X − Z)) which is
precisely the projective cubic node. This proves the proposition. 2

It remains to tackle the third case, that of the smooth cubics. This is done next.

Proposition 9.5.2. Let C = V (F ) ⊂ P2 (k) be a smooth cubic. Then after a linear change of coordinates, we
have C = V (ZY 2 − X(X − Z)(X − µZ)) for some µ 6= 0, 1.

Proof: Note that a smooth cubic is irreducible, by the Corollary 9.3.5. By the last part of Corollary 9.4.3,
there exists a point of inflexion P on C. By a linear change of coordinates let us assume that (i) P = [0 : 1 : 0]
and (ii) the tangent line L at P which has order of contact oP (L, C) ≥ 3 is the line V (Z). If we consider
the affine chart U1 = D(Y ) = {[X : Y : Z] : Y 6= 0}, we again have that P = (0, 0) is an inflexion point
for the non-empty irreducible affine piece C1 = C ∩ U1 = V (f ) which is defined by the dehomogenisation
f (x, z) = F (X, 1, Z). Since (0, 0) ∈ C1 , we have f (0, 0) = 0, and since the line z = 0 is tangent to C1 at (0, 0),
we have fx (0, 0) = 0. Since (0, 0) is a point of inflexion, the Corollary 9.4.2 we have fxx (0, 0) = 0. Thus we
have:
f (x, z) = γz + (2αzx + 2βz 2 ) + f3 (x, z)
 LECTURES ON BASIC ALGEBRAIC GEOMETRY 101

where γ 6= 0. Since C2 is irreducible, the coefficient of x3 in the degree 3 homogeneous polynomial f3 is non-
zero. By scaling z and x, we may assume that γ = 1 and the coefficient of x3 in f3 is also 1. We homogenise
with respect to Y and get:
F (X, Y, Z) = ZY 2 + Y (2αZX + 2βZ 2 ) + f3 (X, Z)
= Z(Y 2 + 2αXY + 2βZY ) + f3 (X, Z)
 
= Z (Y + αX + βZ)2 − (αX + βZ)2 + f3 (X, Z)
= Z(Y + αX + βZ)2 + g3 (X, Z)
where g3 (X, Z) := f3 (X, Z) − Z(αX + βZ)2 is another homogeneous polynomial in X, Z of degree 3. Note
that the coefficient of x3 in g3 is the same as its coefficient in f3 , which is 1. Now change coordinates linearly
by X 7→ X, (Y + αX + βZ) 7→ Y and Z 7→ Z. Then our F looks like
F (X, Y, Z) = ZY 2 + g3 (X, Z)
where g3 (X, Z) = X 3 +aX 2 Z +bXZ 2 +cZ 3 . Clearly, we may write g3 (X, Z) = (X −λ1 Z)(X −λ2 Z)(X −λ3 Z).
Changing linearly X − λ1 Z to X, and keeping Z and Y unchanged, we have
F (X, Y, Z) = ZY 2 − X(X − λZ)(X − µZ)
Now if both λ and µ are zero, we have C is a cusp, which is a contradiction since our C is smooth. If one of
λ or µ is zero, we have C is a node, which is also a contradiction for the same reason. Thus both λ and µ are
non-zero. Scaling Z, we may assume λ = 1. This will alter the term ZY 2 to λ−1 ZY 2 , but rescaling Y will
again restore it to ZY 2 . Thus we now have:
F (X, Y, Z) = ZY 2 − X(X − Z)(X − µZ)
where µ 6= 0 and µ 6= 1, for both of these situations would give us a node, which is singular. This proves the
proposition. 2

Remark 9.5.3. The issue addressed in the foregoing subsection is the purely algebraic problem:

Let F (X0 , X1 , X2 ) be a homogeneous polynomial of degree ≤ 3. Making only non-singular linear changes
of coordinates (X0 , X1 , X2 ), bring F to standard form, and classify the standard forms thus possible.

As pointed out at the outset of this subsection, the question yields to pure algebraic methods for deg F ≤ 2.
If one wanted to tackle the case of cubics, i.e. say F is irreducible of degree 3, by purely algebraic means,
the task would be completely hopeless. However, the Propositions 9.5.2 and 9.5.1 show that the full power
of geometry, viz. the use of Bezout, the notion of orders of contact, smoothness, multiplicity, singular and
inflexion points, can make the problem soluble, and indeed one just gets the three irreducible cubics we have
been discussing all along.

In the case of singular irreducible cubics, there is a unique normal form, viz. node or cusp. In the smooth
case of the elliptic curve V (Y 2 Z − X(X − Z)(X − µZ)) (in Proposition 9.5.2), the reader may wonder how
much freedom the parameter µ has. It turns out that two of these with parameters µ and λ are equivalent by
µ
a linear coordinate change in P2 (k) iff λ = µ, µ1 , 1−µ 1
, 1−µ , 1−µ
µ or 1 − µ. The reader is urged to empirically
demonstrate that at least these values lead to equivalent elliptic curves. The fact that these are the only ones
possible is a slightly deeper fact.