INTRODUCTION TO ALGEBRAIC GEOMETRY, CLASS 2
RAVI VAKIL
Contents
1. A bit more commutative algebra 1
2. The correspondence between algebraic sets and radical ideals 2
3. More Nullstellensatz 4
1. A bit more commutative algebra
There was a little more commutative algebra I should have mentioned last time.
We won’t use it much soon, but I wanted to get it out of the way.
First of all, some of you suggested that I remind you of some facts, so you don’t
have to recall them quickly in the midst of the lecture.
If R is a ring, then a non-empty subset I is an ideal if it is closed under addition,
and if you multiply an element of I by any element of R, you get another element of
I. An ideal is prime if xy ∈ I implies x ∈ I or y ∈ I. Equivalently, I is prime if R/I
is an integral domain, which means that xy = 0 (in R/I) means x = 0 or y = 0 (in
R/I). A maximal ideal is a proper ideal maximal with respect to inclusion. Then
I is maximal if and only if R/I is a field, so maximal ideals are prime.
Localization. A subset S is multipliciative if xy ∈ S for all x, y ∈ S. Examples
are: (a) R − p where p is a prime ideal; (b) {f, f 2 , f 3 , . . . , } where f ∈ R. Then
you can define a new ring S −1 R as follows: elements are of the form (a, b), which
you should think of as being a/b, where b ∈ S, and a ∈ R; two elements (a, b) and
(c, d) are the same if s(ad − bc) = 0 for some element s ∈ S. In the two examples,
we write this ring as Rp and R(f ) .
(Experts should think about how prime ideals and localization interact.)
Examples of (a) are: R = Z, p = (3), so elements of Rp are fractions with no
power of 3 in the denominator. R = k[x], p = (x − 1), so elements of Rp are
quotients of polynomials with no power of (x − 1) in the denominator. If R is an
integral domain, then (0) is a prime ideal, and the localization with respect to this
ideal gives you all fractions, e.g. R = Z you get the rational numbers Q.
Date: September 14, 1999.
1
Examples of (b) are: R = Z, f = 3, so elements of R(f ) are fractions with only
a power of 3 in the denominator. R = k[x], p = (x − 1), so elements of R(f ) are
quotients of polynomials with only powers of (x − 1) in the denominator.
Warning: Be a little cautious: weird stuff can happen here. For example, if S
contains 0, then the localization of S is the zero ring. (That will be an exericse.) So
in particular, if f is nilpotent (i.e. f n = 0 for some n), then R(f ) is the zero-ring.
2. The correspondence between algebraic sets and radical ideals
Let’s recap where we are.
An algebraic set in An (k) is something cut out by polynomials. (We’ll see later
that we can add “a finite number of” to that sentence.)
From an algebraic set X (or indeed any subset of An (k)) we can cook up an ideal
I(X), the ideal of functions vanishing on X. It is a radical ideal.
From any radical ideal I (or indeed any subset of k[x1 , . . . , xn ]) we can cook up
an algebraic set V (I), the “vanishing set” of the ideal.
The Nullstellensatz implies that this gives an equivalence:
algebraic sets ↔ radical ideals in k[x1 , . . . , xn ].
Let’s begin to understand this dictionary, by looking in the plane A2 .
0) Ideal corresponding to a point (a, b): (x − a, y − b). (Maximal.)
If you have an algebraic set, say in A3 , the xy-plane union the z-axis, then this
gives you a radical ideal I. We’ll work out soon what that radical ideal is. On the
geometric side, you have a bunch of points, so on the algebraic side, you have a
bunch of maximal ideals. On the algebraic side, how do you see incidence, i.e. how
can you tell which points are on the algebraic set? Answer: those maximal ideals
containing I.
a) Caution: this is actually trickier than it looks. It’s low-tech, but subtle. It’s
also very important to understand the guts of the radical ideal / algebraic set “du-
ality”, so be sure to think this through.
Let’s understand how to interpret unions of algebraic sets in terms of algebra.
The union of two algebraic sets corresponding to radical ideals I1 , I2 is the radical
of the ideal I1 intersect I2 . Example: the two points (1,0) and (-1,0), give the ideal:
elements that are of the form (y, x + 1) and (y, x − 1).
Do it painfully: If something lies in both ideals, then it has a part that is a
multiple of y, plus a polynomial in x, which is divisible by both (x + 1) and (x − 1).
2
√ Notice another
√ way of doing it: (y 2 , (x + 1)y, (x − 1)y, (x2 − 1)); in general that
I1 ∩ I2 = I1 I2 .
√ √
Exercise. Show that I1 ∩ I2 = I1 I2 .
Exercise. If I1 and I2 are radical, show that their intersection is also radical.
Example. Find the ideal corresponding to the union of the point (1,1) and the
line y = 2. Ans: ((y − 2)(x − 1), (y − 2)(y − 1)).
Example. Find the ideal corresponding to the union of the two points (1,0) and
(-1,0).
Example. Find the ideal in k[x1 , x2 , x3 ] corresponding to the union of the z-axis
and the xy-plane. (Have them do this.)
Exercise. (In dimension 3.) Find the ideal corresponding to the x1 -axis union
the point (1,1,1).
Corollary. A finite union of points is an algebraic set.
b) This should remind you of high-school algebra, i.e. solving a bunch of equa-
tions in a bunch of unknowns, albeit in a more high-powered setting.
The intersection of two algebraic sets corresponding to radical ideals I1 , I2 is the
radical of the ideal I1 + I2 .
Example. y = 0, y = x2 − 1. Ideal is (y, y − x2 + 1) = (y, x2 − 1). You can see
the two points: (y, x + 1) and (y, x − 1).
Example. To see that you need take to radical, consider y = 0, y = x2 . Aside:
You want to say that the line intersects the parabola at one point with multiplicity
2. You already see the one point. Can you interpret the 2? This was a scheme-
theoretic intersection.
Exercise. Find the intersection of the x-axis and the cubic y 2 = x3 + x2 , using
ideals: find the ideal corresponding to those two curves, and compute the intersec-
tion. (This isn’t actually on the problem set, but resembles some that are.
Summary. We’ve defined unions and intersections algebraically.
Something else to think about: Maximal ideals correspond to points. What do
prime ideals correspond to (that aren’t maximal)? Here’s where to start: think in
the plane A2 . Name some prime ideals, other than those corresponding to points,
and the zero-ideal (which might confuse you).
Aside. Affine schemes!
radical ideals ⇔ algebraic sets (algebraic varieties) (in An )
3
ideals ⇔ affine schemes! (closed subschemes of An )
For example, (x2 , y). (x2 , xy, y 2 ). We can now define scheme-theoretic unions
and intersections, by thinking about ideals. More on this later.
3. More Nullstellensatz
Here’s a reminder of the various versions of the Nullstellensatz.
Nullstellensatz Version 1. Suppose F1 , . . . , Fm ∈ k[x1 , . . . , xn ]. If the ideal
(F1 , . . . , Fm ) 6= (1) = k[x1 , . . . , xn ] then the system of equations F1 = · · · = Fm = 0
has a solution in k.
We’ll prove this soon.
Nullstellensatz Version 2. The maximal ideals of k[x1 , . . . , xn ] are precisely
those maximal ideals which come from points, i.e. those ideals of the form (x1 −
a1 , . . . , xn − an ) for some a1 , . . . , an ∈ k.
Nullstellensatz Version 3 (the “Weak Nullstellensatz”). If I is a proper
ideal in k[x1 , ..., xn ], then V (I) is nonempty.
(In retrospect, this seems to be the same as version 1!)
√
Nullstellensatz Version 4. I(V (I)) = I, which implies that: Radical ideals
are in 1-1 correspondence with algebraic sets.
Nullstellensatz Version 5. A radical ideal of k[x1 , . . . , xn ] is the intersection
of the maximal ideals containing it.
Nullstellensatz Version 6. If F1 , ..., Fr , G are in k[x1 , ..., xn ], and G vanishes
wherever F1 , ..., Fr vanish, then some power of G lies in the ideal (F1 , . . . , Fr ), i.e.
GN = A1 F1 +...+Ar Fr for some N > 0 and some Ai in k[x1 , ..., xn ]. Put differently:
if I is a finitely generated√ideal of k[x1 , . . . , xn ], and G is 0 on the vanishing set of
I (G ∈ V (I)), then G ∈ I. (We’ll later see that all ideals are finitely generated
when we talk about Noetherian rings.)
Proof of Nullstellensatz, version 1.
We need a lemma:
Lemma. If a system F1 = · · · = Fm = 0 with Fi ∈ k[x1 , . . . , xn ] has a solution
in some finitely generated extension field K of k, then it has a solution in k.
Proof of Lemma. By structure of finitely-generated extension fields (Shafarevich
p. 280), K = k(t1 , . . . , tr , θ) = k(t, θ) where t1 , . . . , tr are algebraically independent
over k and θ is a root of a polynomial
4
P (t, U ) = p0 (t)U d + · · · + pd (t) ∈ k(t)[U ],
where P (t, U ) is irreducible over k(t).
Suppose (η1 , . . . , ηn ) is a solution where ηi ∈ K, so ηi = Ci (t, θ). Then
Fj (C1 (t, U ), . . . , Cn (t, U )) = P (t, U )Qj (t, U )
for some polynomial Qj . Subtitute values xi = αi ∈ k for all i, such that
(α1 , . . . , αn ) is not a zero of the denominators of any coefficient of any of the polyno-
mials that have turned up so far, i.e. P , Qi , Ci ∈ k(t)[U ], nor a zero of the leading
coefficient of P . Choose U = τ to be one of the roots of P (α1 , . . . , αn , τ ) = 0, and
set Cj (α1 , . . . , αn , τ ) = λj for all j. Then Fj (λ1 , . . . , λn ) = 0 for all j, so we’ve
found a solution of the system in k.
Proof of Nullstellensatz Version 1. If the ideal (F1 , . . . , Fm ) is not the entire ring
k[x1 , . . . , xn ] then it is contained in some maximal ideal m, and K = k[x1 , . . . , xn ]/m
is a field — and what a nice field it is: writing ηi for the image of xi in K, then
K = k(η1 , . . . , ηn ), and (η1 , . . . , ηn ) is a solution of the system F1 = · · · = Fm = 0.
By the lemma we get a solution. This proves the Nullstellensatz.
It’s important to see a proof of the Nullstellensatz in order to get a feel for the
flavour of the proof, but it isn’t necessary to really know it at well at this point, as
it differs in flavour from the rest of the course.