Algebraic Geometry

Karl-Heinz Fieseler and Ludger Kaup

Uppsala 2012

1
Contents
1 Introduction: Plane Curves 3

2 Algebraic Sets in k n 9

3 Integral Ring Extensions 21

4 Affine Varieties 29

5 Prevarieties 42

6 Algebraic Varieties 50
6.1 Digression: Complex Analytic Spaces . . . . . . . . . . 54

7 Projective Varieties 56

8 Dimension 64

9 Rational functions and local rings 76

10 Normal Varieties 82

11 Regular Points 87

2
1 Introduction: Plane Curves
In order to give an idea what algebraic geometry is about we discuss in the
introductory section plane curves in a not too rigorous manner:

Definition 1.1. Let k be a field. A plane curve C over k is a subset C ⊂ k 2 ,

which can be written

C = V (f ) := {(x, y) ∈ k 2 ; f (x, y) = 0}

with a polynomial f ∈ k[X, Y ] \ k.

Example 1.2. 1. For f = X 2 + Y 2 − 1 and k = R we find V (f ) ⊂ R2 is

the unit circle.

2. For f = X 2 + Y 2 and k = R we find V (f ) = {0}.

Only the unit circle seems to be an ”honest plane curve”. In order to avoid
points to be plane curves, we shall only consider an algebraically closed field
k. Nevertheless it turns out to be very useful to look at pictures in R2 ,
namely given f ∈ R[X, Y ] ⊂ C[X, Y ] one has V (f ) ⊂ C2 and draws its ”real
trace” V (f ) ∩ R2 . Only one has to understand, which features are due to
the fact that R is not algebraically closed and which ones really reflect the
complex situation.

Remark 1.3. The polynomial f can always be assumed to be square free.

A natural aim when studying a plane curve C is to try to parametrize it,

i.e. to write
C = {ϕ(t); t ∈ U ⊂ k},
where ϕ = (g, h) : U −→ C ⊂ k 2 .
In the next theorem we classify plane curves C = V (f ) resulting from
a polynomial of degree ≤ 2. Here the sign ”∼ =” means, that there is a
”reasonable” bijective map between the left and the right hand side.

Theorem 1.4. Let k be algebraically closed. Then we have V (f ) ∼

= k for
deg f = 1. For deg(f ) = 2 we have the following possibilities

1. V (f ) ∼
= k,
2. V (f ) ∼
= k ∗ := k \ {0},

3
 3. V (f ) ∼
= (k × {0}) ∪ (k × {1}) ⊂ k 2 is the union of two parallel lines,

4. V (f ) ∼
= (k × {0}) ∪ ({0} × k) ⊂ k 2 is the coordinate cross.

Proof. For deg f = 1, the curve V (f ) ⊂ k 2 is a line. For a reducible polyno-

mial f = gh of degree 2, we are in one of the cases 1, 3, 4, where the first
case corresponds to h ∈ k ∗ g. Let us now assume that f is irreducible. We
may assume that 0 ∈ V (f ) and write

f = f2 + f1

with homogeneous polynomials of degree 2 resp. 1. Since k is algebraically

closed we can factorize
f2 = gh
with homogeneous polynomials g, h ∈ k[X, Y ] of degree 1. If g, h only dif-
fer by a nonzero constant factor, we may assume - after a linear change of
coordinates - that f2 = X 2 , otherwise f2 = XY . In the first case we have

f = X 2 + aX + bY

with b 6= 0. We may solve for Y , and the map

t 7→ (t, −b−1 (t2 + at)),

is an isomorphism k −→ V (f ). In the remaining case we have

f = XY + aX + bY

with ab 6= 0. Then we have

V (f ) ∩ ({−b} × k) = ∅

and  
ab
t 7→ t − b, −a +
t
provides an isomorphism k ∗ −→ V (f ).

Then we might conjecture

4
 1. Every plane curve C ⊂ k 2 is the finite union

C = C1 ∪ ... ∪ Cr

of ”irreducible” curves Ci , i.e. of curves which do not admit a further

decomposition.

2. For every irreducible curve C there is a bijective map

k ⊃ U −→ C, t 7→ (g(t), h(t)),

where U ⊂ k has finite complement and g, h : U −→ k are rational

functions (quotients of polynomials) with poles outside U .

The first statement is true: If f1 , ..., fr are the irreducible divisors of f ,

we have Ci = V (fi ).
The second one is practically always wrong. Let us discuss an example:
Take
f = Y 2 − X 2 (X + a).
For k = C and a ∈ R>0 the curve V (f ) ∩ R2 ⊂ C2 is a noose. Given t ∈ k we
consider the line Lt := k(1, t), our parametrization ϕ : k −→ V (f ) associates
to t ∈ k the unique intersection point 6= (0, 0) in Lt ∩ V (f ). Solving

0 = f (λ(1, t)) = λ2 t2 − λ2 (λ + a)

we find
ϕ(t) = (t2 − a, t(t2 − a)).
Note that for a 6= 0 and char(k) 6= 2 this is unfortunately not an injective
parametrization: The self intersection point (0, 0) has the inverse images ±b
where b2 = a. On the other hand, for a = 0 the above parametrization
becomes ϕ(t) = (t2 , t3 ), it is bijective, but not an isomorphism, since its
inverse can not be defined by the restriction to V (f ) ⊂ k 2 of a polynomial
function on k 2 . In that case

V (f ) = {(x, y); y 2 = x3 }

is called Neil’s parabola, in the origin it has a ”cusp singularity”.

5
Definition 1.5. Let C = V (f ) ⊂ k 2 with a square free polynomial f . Then
a point (x, y) ∈ V (f ) is called a singular point if ∇f (x, y) = (0, 0). A curve
is called smooth if it has no singular points.
Example 1.6. 1. If C = C1 ∪ ... ∪ Cr with irreducible curves Ci then the
intersection points, i.e. the points in Ci ∩ Cj , j 6= i, are singular points.
2. Any curve has at most finitely many singular points.
3. The noose has the origin as its only singular point. If a 6= 0, it is a ”self
intersection point”: This concept may be defined algebraically, but here
we only give the definition for k = C. A singular point (b, c) ∈ C =
V (f ) (f being irreducible) is called a self intersection point, if there is
a neighbourhood U ⊂ C2 of (b, c) together with holomorphic functions
g, h : U −→ C with g(b, c) = 0 = h(b, c), such that f |U = gh and
V (g) 6= ∅ =
6 V (h). √ take U = {(x,
√ In the case of the noose y); |x| < |a|},
√
g(x, y) = y + x x + a, h(x, y) = y − x x + a, where .. : C \ R<0 is a
suitable branch of the square root.
4. A singular point of an irreducble curve which is not a self intersection
point, is called a cusp singularity, e.g. the origin is a cusp singularity
of Neil’s parabola.
5. Any nonsingular point of a complex plane curve has a neighbourhood
homeomorphic to a disc. This is an easy consequence of the holomor-
phic implicit function theorem.
The ”generic” version of our noose is the curve V (f ) with
f = Y 2 − p(X),
where p ∈ k[X] is a polynomial of degree 3 (resp. 2` + 1) with only simple
roots. In that case V (f ) is a smooth curve, it is called a plane elliptic
curve (resp. hyperelliptic curve for ` ≥ 2). For such curves parametrizations
U −→ V (f ) do not exist at all:
Theorem 1.7. Let C = V (f ) be an elliptic resp. hyperelliptic plane curve
over a field k of characteristic char(k) 6= 2. Then there is no nonconstant
map
(g, h) : U −→ V (f ),
where U ⊂ k is the complement of a finite set and g, h ∈ k(T ) := Q(k[T ])
are rational functions with poles outside U .

6
Proof. We may assume that g 0 6= 0 or h0 6= 0. (For char(k) = p > 0 we have
g 0 = 0 iff g = ĝ(T p ) with some ĝ ∈ k(T ). Then every g ∈ k(T ) can be written
r
g = g0 (T p ) with g00 6= 0. Now replace g with g0 . In order to see all that,
use the partial fraction decomposition of g.) Furthermore that U ⊂ k is the
maximal domain of definition for (g, h), i.e. that at a point a ∈ k \ U one of
the functions g, h has a pole. The equality
h2 = p(g)
shows that then the other function has a pole as well and that, if m, n are
the pole orders of g, h at a, we have 2n = (2` + 1)m. Formal differentiation
of the above equality gives
2hh0 = p0 (g)g 0 .
Let us now consider the rational function
2h0 g0
ϕ= 0 = ∈ k(T ).
p (g) h
Since p has only simple zeros, p0 (g) and h have no common zeros in U , in
particular the function ϕ is defined on U . We show that it can be extended to
k, hence is a polynomial. Consider a point a 6∈ U with m, n as above. Then
the function ϕ = g 0 /h has a removable singularity at a, since the pole order
m + 1 of g 0 is ≤ the pole order n of h: Indeed m + 1 ≤ n ⇐⇒ 1 ≤ (` − 21 )m,
the latter inequality holds, since m > 0 is even.
Finally we want to show that ϕ = 0 - this implies h0 = 0 = g 0 . We com-
pute the degree of ϕ. First of all we can define the degree of a rational func-
tion - the difference of the degrees of numerator and denominator polynomial.
Now denote m, n the degree of g, h. If m > 0, we have deg p(g) = (2` + 1)m
and once again 2n = (2` + 1)m. Since deg g 0 = m − 1 < m + 1 ≤ n, we
see that ϕ has negative degree, hence f = 0 and g 0 = 0. If m < 0, we have
deg p(g) = 0 and then deg h = 0 and deg g 0 ≤ m − 1 < 0, so deg(ϕ) < 0.
If m = 0 = n, we have deg(g 0 ) < 0 and obtain again deg(ϕ) < 0. Finally
in the case m = 0 > n, we have deg p(g) < 0 and thus deg p0 (g) = 0 - the
polynomials p and p0 don’t have common zeros - and are done even then since
deg h0 ≤ n − 1 < 0.

Holomorphic Parametrizations: For an elliptic curve V (f ) over C there

exists nevertheless a holomorphic parametrization
(g, h) : U −→ V (f ),

7
where Λ := C \ U is an infinite discrete set, indeed a lattice: A lattice Λ ⊂ C
is a subset of the form
Λ = Zω1 + Zω2 ,
where ω1 , ω2 are linearly independent over R, or, equivalently, ω1 /ω2 6∈ R. A
meromorphic function f on C is called Λ-periodic if f (z +ω) = f (z) holds for
all ω ∈ Λ. There are only constant holomorphic Λ-periodic functions: Such
a function would be globally bounded by its supremum on the parallelogram
{rω1 + sω2 ; 0 ≤ r, s ≤ 1}, and by Liouville’s theorem it follows that it is a
constant. But there are interesting meromorphic functions: The Weierstraß-
℘-function
1 X 1 1

℘(z) := 2 + − ,
z ω∈Λ∗
(z − ω)2 ω 2
where Λ∗ := Λ\{0}, is an example of a Λ-periodic function. It is holomorphic
outside Λ and has poles of order 2 at the lattice points. For
X 1 X 1
a=3 , b = 5
ω∈Λ
ω4 ω∈Λ
ω6

we consider the function

(℘0 )2 − 4℘3 + 20a℘ + 28b.
It is holomorphic in U , the Laurent expansion at the origin shows, that it
has a removable singularity there, indeed a zero. But then, being Λ-periodic,
it has to be ≡ 0. Some more detailed investigations show that
(℘, ℘0 ) : U −→ V (f ), f = Y 2 − 4X 3 + 20aX + 28b
is a surjective and locally homeomorphic parametrization; its fibres are ex-
actly the cosets z+Λ, z ∈ U . Now, in order to get a more satisfactory picture,
one compactifies V (f ) by adding a point ∞ to
V (f ) := V (f ) ∪ {∞}.
and gets a parametrization
(℘, ℘0 ) : C −→ V (f ),
sending the lattice points to ∞. Indeed it induces a homeomorphism
C/Λ −→ V (f ),

8
of the torus with the one point compactification of V (f ). Indeed, tradition-
ally one means the compactification V (f ) when speaking about an elliptic
curve. And it turns out that for every elliptic curve there is a corresponding
lattice such that we have a homeomorphism of the above type.
There is a further interesting remark: The homeomorphism V (f ) ∼ = C/Λ
endows on V (f ) the structure of an (additive) abelian group with the point
∞ as unit element. The group multiplication, it turns out, can also be
understood in a completely algebraic way: Take any line L ⊂ C2 and a
point a ∈ V (f ) ∩ L. We can define an intersection multiplicity: It is the
multiplicity of the polynomial function f (ϕ(t)) at 0 if ϕ : C −→ L is a
linear parametrization with ϕ(0) = a. Now counting the intersection points
in L ∩ V (f ) with multiplicities, there are at most 3 intersection points - two
for ”vertical” lines L = {a} × C, and three otherwise - and their sum in
V (f ) ∼
= C/Λ equals zero.
Hyperelliptic Curves: Now let us consider the case of a hyperelliptic curve:
The fact that the polynomial p(X) has odd degree guarantees, that the point
∞ ∈ V (f ) has a neighbourhood homeomorphic to a disc – if the degree is
even, it is a self intersection point, has a neighbourhood homeomorphic to
two copies of the unit disc E ⊂ C meeting at their centers, i.e. (E × {0}) ∪
({0} × E) ⊂ C2 with ∞ corresponding to (0, 0).
In that situation we have to replace C with the open unit disc E: There
is a surjective and locally homeomorphic parametrization
(g, h) : E −→ V (f )
given by two meromorphic functions, where the fibre of the point at infinity
consists of the points, where one of the functions g, h or both have a pole.
Indeed, there is a subgroup G ⊂ Aut(E), acting freely on E, such that the
parametrization is G-invariant and induces a homeomorphism
E/G ∼ = V (f ).
But for ` > 1 the extended curve V (f ) is not a group. Indeed V (f ) is
homeomorphic to a sphere with ` handles.

2 Algebraic Sets in k n
Algebraic geometry deals with objects, called ”algebraic varieties”, which
locally look like ”algebraic sets”, i.e. sets which can be described by a system

9
of polynomial equations. For this one has first of all to fix a base field:

Notation: We denote k an algebraically closed field; vector spaces are always

assumed to be vector spaces over k. In contrast to that, we use the letter K
for a not necessarily algebraically closed field.

In this section we define algebraic sets and prove that they correspond
to certain ideals in polynomial rings, the central result being ”Hilberts Null-
stellensatz”, Th.2.17.
To begin with let us mention the following fact: Given any infinite field
K the polynomial ring
K[T ] := K[T1 , ..., Tn ]
can be regarded as a subring
n
K[T ] ⊂ K K
n
of the ring K K of all functions K n −→ K.
Proof. For n = 1 this follows from the fact, that a polynomial f ∈ K[T ] has
at most deg f zeros. For n > 1 write
r
X
f= fi (T1 , ..., Tn−1 )Tni ∈ (K[T1 , ..., Tn−1 ]) [Tn ]
i=0

and assume f (x) = 0 for all x = (x0 , xn ) ∈ K n = K n−1 × K. Hence for any
x0 ∈ K n−1 we have 0 = f (x0 , Tn ) ∈ K[Tn ] resp. fi (x0 ) = 0 for all x0 ∈ K n−1 .
Using the induction hypothesis we conclude fi = 0 ∈ K[T1 , ..., Tn−1 ] for all
i = 0, ..., r and thus f = 0 ∈ K[T1 , ..., Tn ] as well.

Definition 2.1. A subset X ⊂ k n is called an algebraic set, if there is a

subset F ⊂ k[T ] of the polynomial ring k[T ] with

X = N (F ) := N (k n ; F ) := {x ∈ k n ; f (x) = 0, ∀f ∈ F } .

We indicate this by writing

X ,→ k n ,
and for a finite set F = {f1 , ..., fr } we also simply write

N (f1 , ..., fr ) := N (k n ; f1 , ..., fr ) := N (k n ; F ).

10
Example 2.2. 1. Both k n = N (k n ; 0) and the empty set ∅ = N (k n ; 1)
are algebraic sets.

2. One point sets

{a} = N (k n ; T1 − a1 , ..., Tn − an )

are algebraic sets.

3. Arbitrary intersections
\ [
N (Fλ ) = N ( Fλ )
λ∈Λ λ∈Λ

and finite unions of algebraic sets are algebraic, since

N (F G) = N (F ) ∪ N (G)

with F G := {f g; f ∈ F, g ∈ G}. The inclusion ”⊃” is obvious, while

for a point x 6∈ N (F ) ∪ N (G) we find f ∈ F and g ∈ G with f (x) 6=
0 6= g(x), so f g ∈ F G does not vanish at x either.

4. All finite sets are algebraic; for n = 1 this are all the proper algebraic
sets.

5. Vector subspaces W ⊂ k n are algebraic subsets.

6. Let X := N (k 2 ; T22 −p(T1 )) with a polynomial p(T1 ) ∈ k[T1 ]. For

p(T1 ) = T1 (T1 − 1)(T1 − λ) with λ 6= 0, 1 the algebraic set X ,→ k 2
is called an elliptic curve. For k = C and λ ∈ R draw a picture of
X ∩ R2 ⊂ R2 ! In the ”degenerate case” λ = 0, 1 the real picture
of X is a noose. For p(T1 ) = T13 it is also called Neils parabola, in-
deed, the latter can be viewed as degeneration of the nooses defined by
p(T1 ) = T1 (T1 − λ)2 with λ ∈ k ∗ .

7. The algebraic set X := N (k 3 ; T1 T3 − T22 ) is also called a quadric cone.

8. The algebraic set Z := N (k 4 ; T1 T4 − T2 T3 ) is called the Segre cone; if

we identify k 4 with the vector space k 2×2 of 2 × 2 matrices with entries
in k, then Z = k 2×2 \ GL2 (k) consists of all singular 2 × 2 matrices.

11
Remark 2.3. 1. According to Ex.2.2.1,3 and 4, the algebraic sets satisfy
the axioms of the closed sets of a topology on k n . The corresponding
topology is called the Zariski topology. If we want to express that a
subset X ⊂ k n is Zariski-closed we also write simply X ,→ k n .

2. Two nonempty open sets U, V ⊂ k n have nonempty intersection. Or

equivalently: The n-space k n is not the union of two proper closed
subsets X, Y ,→ k n . Otherwise we may assume X = N (f ), Y = N (g)
with f, g ∈ k[T ] \ {0}. Then f g ∈ k[T ] \ {0} satisfies k n 6= N (f g) =
N (f ) ∪ N (g).

3. If k = C, there is yet another p

topology
Pn both on Cn and all its subsets,
induced by the norm ||z|| := i=1 zi z i . We shall call it the strong
topology on Cn resp. X ⊂ Cn . Note that the strong topology is strictly
finer than the Zariski topology: Polynomials are continuous functions
with respect to both the Zariski and the strong topology.

4. A proper algebraic set X ,→ Cn has no interior points with respect to

the strong topology: Take a polynomial f ∈ C[T ] \ {0} vanishing on
ν
X. If a ∈ X is an interior point, then ∂∂T νf (a) = 0 for all ν ∈ Nn , and
thus Taylor expansion gives
X 1 ∂ νf
f= ν
(a)(T − a)ν = 0,
ν∈Nn
ν! ∂T

a contradiction!

5. We remark without proof that an algebraic subset X ⊂ Cn is connected

with respect to the Zariski topology iff it is with respect to the strong
topology.

An algebraic subset can always be described by finitely many polynomial

equations:

Proposition 2.4. For every algebraic subset X = N (F ) ,→ k n there is a

finite subset F0 ⊂ F mit X = N (F0 ).

In order to see that, we first enlarge the set of polynomials defining a

given algebraic set:

12
Remark 2.5. The ideal
( r )
X X
aF := k[T ]f := gi fi ; gi ∈ k[T ], fi ∈ F, r ∈ N
f ∈F i=1

of k[T ] generated by F ⊂ k[T ] satisfies N (aF ) = N (F ).

Proof. The inclusion F ⊂ aF implies obviously N (F )P ⊃ N (aF ); on the other
r
hand every polynomial
Pr h ∈ a F is a finite sum h = i=1 gi fi with fi ∈ F ,
whence h(x) = i=1 gi (x)fi (x) = 0 for all x ∈ N (F ) and thus N (F ) ⊂
N (aF ).

Hence, in order to study algebraic subsets one could try to investigate

first ideals a ⊂ k[T ] (or rather in K[T ], where K is any field). For n = 1
we know that every ideal a ⊂ K[T ] is a principal ideal a = K[T ] · f with a
polynomial f ∈ K[T ].

Digression to commutative algebra: Noetherian Rings

Definition 2.6. A commutative ring R with unity 1 is called noetherian, if

one of the following three equivalent conditions is satisfied:
1. Every ideal a ⊂ R is finitely generated, i.e., there is a finite set F ⊂
R : |F | < ∞, with a = aF .

2. Every ascending chain a1 ⊂ · · · ⊂ aj ⊂ . . . of ideals in R terminates,

i.e. there is a natural number j0 ∈ N with aj = aj0 for j ≥ j0 .

3. Every nonempty set A = {ai , i ∈ I} of ideals ai ⊂ R contains a maximal

element a0 , i.e. such that a0 ⊂ a ∈ A =⇒ a0 = a.
Proof. Exercise!
Remark 2.7. We leave it to the reader to check that given an ideal a =
aF ⊂ R in a noetherian ring R, we can choose a ”finite basis” F0 ⊂ F of the
ideal a, i.e. such that
aF0 = aF .
Note that the word ”basis” is used here in an old fashioned way, meaning
nothing but a system of generators, and not requiring linear independence of
the generators, as one does in linear algebra.

13
Example 2.8. A PID (:=Principal Ideal Domain) is obviously noetherian.
Theorem 2.9. The polynomial ring R[T ] in one variable over a noetherian
ring R is again noetherian.
Corollary 2.10 (Hilberts Basissatz). The polynomial ring K[T1 , ..., Tn ] over
a field K is noetherian.
Proof. K[T1 , ..., Tn ] = (K[T1 , ..., Tn−1 ])[Tn ].
Proof of 2.9. For an ideal a ⊂ R[T ] we define an ascending chain of ide-
als b0 ⊂ · · · ⊂ bj ⊂ . . . in R by
( )
X
bj := r ∈ R; ∃ rT j + ri T i ∈ a .
i<j

For every ideal bj ,→ R we choose generators (sj1 , . . . , sj`j ) and corresponding

polynomials X
pjk = sjk T j + rjik T i ∈ a.
i<j

Since R is noetherian, the chain of ideals (bj ) terminates at some j0 ∈ N.

Our candidate for a ”basis” is now
F := {pjk ; 0 ≤ j ≤ j0 , 1 ≤ k ≤ `j } .
Obviously we have then aF ⊂ a. It remains to show the opposite inclusion:
For p ∈ a we use induction on d := deg(p). For d < 0 we have p = 0. For
”d−1 =⇒ d” let m :=P min(d, j0 ). The leading coefficient of p belongs to bm ,
hence has the form `k=1
m
rk smk with suitable rk ∈ R. Then the polynomial
`m
X
d−m
q := p − T rk pmk ∈ a
k=1

satisfies deg(q) < d. So we know q ∈ aF and hence p ∈ aF as well.

Prop. 2.4 now follows from Corollary 2.9 and the remarks 2.7 and 2.5.

Let us now come back to geometry: As a consequence of 2.5 every alge-

braic set is of the form N (a). In the following lemma we study how N (a)
depends on the ideal a ⊂ k[T ].

14
Lemma 2.11. Let a, b ⊂ k[T ] and aλ ⊂ k[T ], λ ∈ Λ, be ideals. Then

1. N (a) = k n ⇐⇒ a = {0} ; a = k[T ] =⇒ N (a) = ∅ .

2. a ⊂ b =⇒ N (a) ⊃ N (b).
P T
3. N ( λ∈Λ aλ ) = λ N (aλ ).

4. N (ab) = N (a∩b) = N (a) ∪ N (b).

Proof. Exercise!

On the other hand any algebraic set X ,→ k n has an associated ideal, its
vanishing ideal I(X) ,→ k[T ], as well:

Definition 2.12. The vanishing ideal I(X) ⊂ k[T ] of an algebraic set X ,→

k n is
I(X) := {f ∈ k[T ]; f |X = 0} .

The next lemma investigates I(X) as a function of X:

Lemma 2.13. For X ,→ k n and an ideal a ⊂ k[T ] we have:

1. I(X) = {0} ⇐⇒ X = k n and I(X) = k[T ] ⇐⇒ X = ∅;

T P
2. I( λ Xλ ) ⊃ λ I(Xλ );

3. I(X∪Y ) = I(X) ∩ I(Y );

4. N (I(X)) = X;

5. I (N (a)) ⊃ a .

Example 2.14. In general a $ I (N (a)) and

X \
I(Xλ ) $ I( Xλ ).
λ∈I λ∈I

Namely:

1. For n = 1 and a := (T 2 ) ⊂ k[T ] we have N (a) = {0} and I(N (a)) =

(T ).

15
 2. For n = 2 and X1 := k×{0} = N (T2 ), X2 := N (T2 −T12 ) we obtain
X1 ∩X2 = {0} and I(X1 ∩X2 ) = (T1 , T2 ), while

I(X1 ) + I(X2 ) = (T2 , T2 −T12 ) = (T2 , T12 ) $ (T1 , T2 ) .

The intuitive reason for this is, that the line X1 and the parabola X2
do not meet ”transversally” at their common point 0: Both f = T2
∂f
and f = T2 −T12 satisfy ∂T 1
(0) = 0 and hence this holds as well for all
2
functions f ∈ (T2 , T2 −T1 ).

What is wrong in the first example? Obviously vanishing ideals I(X)

have the property
f m ∈ I(X) =⇒ f ∈ I(X) .
That leads us to the following:
Definition 2.15. An ideal a ,→ R is called a radical ideal, if for all f ∈ R
one has:
f m ∈ a for some m ∈ N =⇒ f ∈ a .
For every ideal a ,→ R the ideal
√
a := {f ∈ R ; ∃ m ∈ N with f m ∈ a}

is a radical ideal; it is called the radical of a.

p
Example 2.16. For n = 1 we have (T 2 ) = (T ) in k[T ].
Theorem 2.17 (Hilberts Nullstellensatz). For an ideal a ⊂ k[T1 , . . . , Tn ] we
have √
I (N (a)) = a .
In particular there is a bijection
N
{radical ideals in k[T ]} −→ {algebraic sets in k n }

with the map I : X 7→ I(X) in the opposite direction.

Example 2.18. Consider the ideal a := (T12 +T√22 ) ⊂ R[T1 , T2 ]. Since T12 +T22
is irreducible and R[T1 , T2 ] factorial, we find a = a. On the other hand
N (a) = {(0.0)} and thus
√
I(N (a)) = (T1 , T2 ) % a.

16
 For the proof of Hilberts Nullstellensatz Th.2.17 we need the following
description of maximal ideals m ⊂ k[T ].
Theorem 2.19 (Weak Nullstellensatz). For every maximal ideal m ⊂ k[T ]
there is a unique a ∈ k n with

m = ma := I({a}) = (T1 − a1 , ..., Tn − an ) .

In particular N (a) 6= ∅ for every proper ideal a $ k[T ].

Indeed, it is a consequence of the following purely algebraic result, to be
shown in the next section:
Theorem 2.20. Let K be any field and m ⊂ K[T ] a maximal ideal. Then
K ,→ L := K[T ]/m is a finite field extension.
Proof of Th.2.19. Since k is algebraically closed, we find even k ∼
= k[T ]/m,
hence have a natural ring homomorphism

ϕ : k[T ] −→ k[T ]/m ∼

= k,

which is nothing but evaluation at a = (a1 , ..., an ) ∈ k n with ai := ϕ(Ti ).

Corollary 2.21. If the polynomials f1 , ..., fr ∈ k[T ] = k[T1 , ..., Tn ] do not
have a common zero, then there are polynomials g1 , ..., gr with
r
X
gi fi = 1.
i=1

Proof. According to Th.2.19 we have (f1 , ..., fr ) = k[T ].

Proof of Th. 2.17. The equality

√
a = I (N (a))

for an ideal a ⊂ k[T ] is a consequence of the weak Nullstellensatz Th. 2.19:

The inclusion ”⊂” is immediate: From a ⊂ I (N (a)) we obtain
√ p
a ⊂ I (N (a)) = I (N (a)) ,

since vanishing ideals are radical ideals.

17
 For the inclusion ”⊃” we consider a function g ∈ I (N (a)) \ {0}. We
introduce an additional indeterminate S and consider in the ring k[T, S] :=
k[T1 , ..., Tn , S] the ideal
b := a[S] + (1−gS) ⊂ k[T, S] = (k[T ])[S]
P∞
with a[S] := i=0 a · S i . Its set of zeros is
N (k n ×k; b) = N (k n × k; a[S]) ∩ N (k n ×k; 1−gS)
with
N (k n × k; a[S]) = N (k n ; a) × k
and
N (k n ×k; 1−gS) = Γ1/g := (x, g(x)−1 ); x ∈ kgn ⊂ kgn ×k ⊂ k n × k,


the graph of the function g −1 : kgn := k n \ N (g) −→ k. Hence N (b) ,→ k n × k

is empty and thus b = k[T, S] according to Th.2.19. So there are elements
f0 , ..., fm ∈ a and h ∈ k[T, S] with
m
X
1 = fi S i + h·(1−gS).
i=0

Now we apply the algebra homomorphism

k[T, S] → k(T ) := Q(k[T ]), Tj 7→ Tj for j = 1, . . . , n, S 7→ 1/g ,
and obtain m
X
1 = fi g −i .
i=0
So, eventually
m
X
m
g = g m−i fi ∈ a
i=0
√
resp. g ∈ a.
Remark 2.22. If K is√any field and K its algebraic closure, Hilberts Null-
stellensatz I(N (a)) = a holds for ideals a ,→ K[T1 , ., , , Tn ] if the zero set
n
N (a) is taken in K instead of in K n . In that situation one has to consider
”K-algebraic sets” only, i.e. algebraic sets which are invariant under the
(componentwise) action of the automorphism group AutK (K).

18
 We end the first section with some easy topological considerations which
in particular apply to algebraic sets, endowed with the Zariski topology:
Definition 2.23. The Zariski topology on an algebraic set X ,→ k n is the
topology induced by the Zariski topology of k n .
Definition 2.24. 1. A topological space X is called noetherian, if every
decreasing chain X1 ⊃ X2 ⊃ . . . of closed subsets terminates.

2. A noetherian topological space X is called irreducible, if it is non-empty

and does not admit a decomposition X = X1 ∪X2 with proper closed
subsets Xi $ X.

3. A maximal irreducible subspace of a noetherian space is called an irre-

ducible component.
Example 2.25. 1. ”Affine n-space” k n is noetherian: Since k[T ] is noethe-
rian, the increasing sequence I(X1 ) ⊂ I(X2 ) ⊂ ... terminates and hence
the sequence of the Xi = N (I(Xi )) does so as well.

2. A noetherian space is quasicompact, i.e. every open cover admits a

finite subcover.

3. A subspace of a noetherian space is again noetherian. In particular

(a) Algebraic sets are noetherian.

(b) Any Zariski open set U ⊂ X of an algebraic set X is quasi-
compact.

4. An algebraic set X ,→ k n is irreducible iff I(X) ,→ k[T ] is a prime

ideal.

5. Irreducible spaces are connected, but not vice versa: X := N (k 2 ; T1 T2 ) =

(k × 0) ∪ (0 × k) is connected, but not irreducible.
Proof. We comment only on 4). If X = X1 ∪X2 is a nontrivial decomposition,
choose polynomials fi ∈ I(Xi ) \ I(X). Then f1 f2 ∈ I(X), so I(X) is not a
prime ideal. On the other hand if I(X) is not a prime ideal, take f1 , f2 6∈ I(X)
with f1 f2 ∈ I(X). Then X = X1 ∪ X2 with Xi := X ∩ N (fi ) $ X.

Proposition 2.26. Let X be a noetherian topological space.

19
 1. If X is irreducible, any non-empty open subset is dense.

2. If Z ⊂ X is an irreducible subspace, so is its closure Z. In particular

the irreducible components of X are closed.

3. X is the irredundant union of its finitely many irreducible components.

4. If X = ri=1 Ui with irreducible open sets Ui ⊂ X, such that Ui ∩ Uj 6= ∅

S
for all i, j, then X is irreducible.

Proof. If X = X1 ∪ X2 and the Xi ,→ X are the union of their irreducible

components, then so is X. Hence, if X is not the union of its irreducible
components, we can construct a strictly decreasing sequence of closed sub-
spaces Y0 % Y1 % Y2 % ..... enjoying the same property: Take Y0 := X.
Since Yn is not the union of its irreducible components, it is not irreducible,
so Yn = Z1 ∪ Z2 with proper closed subspaces Z1 , Z2 . W.l.o.g. then we
may assume that Z1 is not the union of its irreducible components and set
Yn+1 := Z1 . — Finally assume that X1 , ..., Xr are the (pairwise different)
irreducible components of X, and that already X = X1 ∪ ... ∪ Xr−1 . Then,
Xr = (Xr ∩ X1 ) ∪ ... ∪ (Xr ∩ Xr−1 ) being irreducible, we find Xr ⊂ Xi for
some i, 1 ≤ i < r. But then necessarily Xr = Xi , a contradiction.

Example 2.27. X := N (k 3 ; T3 T1 , T3 T2 ) = (k 2 × 0) ∪ (0 × 0 × k).

20
3 Integral Ring Extensions
Though the immediate aim of this section is to prove Th.2.20, the notions
we use are very important also later on. So we give a more general presen-
tation, starting with the following class of rings, which is basic for algebraic
geometry:

Definition 3.1. A K-algebra A is called affine, if it is a finitely generated

K-algebra:
A = K[t1 , ..., tn ]
with suitable elements t1 , ..., tn ∈ A, i.e., any element in A is a polynomial
(with coefficients in the field K) in the elements t1 , ..., tn , or equivalently, if
A is isomorphic to a factor algebra of a polynomial ring:

A∼
= K[T1 , ..., Tn ]/a

with some ideal a ,→ K[T1 , ..., Tn ].

The essential point now is to find for a given affine K-algebra A a subal-
gebra B ⊂ A (Th.3.10), such that

1. B is isomorphic to a polynomial algebra B ∼

= K[S1 , ..., Sd ].

2. A is a finite (:=finitely generated) B-module.

For the construction of the subalgebra B ⊂ A we have to introduce the

notion of an integral ring extension. We characterize first integral elements
in a ring extension R 0 ⊃ R (of commutative rings R, R 0 ) by three equivalent
conditions. For that we need the notion of a faithful module:

Definition 3.2. An R-module M is called faithful, if rM = 0 with r ∈ R

implies r = 0.

Example 3.3. 1. If R ,→ R 0 is a ring extension, then R 0 is a faithful

R-module.

2. The ideal (2) in R = Z/(4) is not a faithful R-module.

Theorem 3.4. Let R ,→ R 0 . For an element s ∈ R 0 the following three

conditions are equivalent:

21
 1. There is a monic polynomial p ∈ R [S] with p(s) = 0.

2. R [s] is a finitely generated R-module.

3. There is a faithful R [s]-module, which as an R-module(!) is finitely

generated.

Definition 3.5. Let R ,→ R 0 .

1. An element s ∈ R 0 is called integral over R, if one of the three equiva-

lent conditions in Th. 3.4 is satisfied.

2. The ring extension R 0 ⊃ R is called integral, if every s ∈ R 0 is integral

over R. Any equation p(s) = 0 with a monic polynomial p ∈ R [S] is
then called an integral equation for s.

Remark 3.6. In general there is no unique integral equation of minimal

degree for a given element s ∈ R 0 ⊃ R: Take a field K and consider the ring
∞
M
2 3
R := K[T , T ] = K ⊕ K · T i ⊂ R 0 := K[T ] ⊂ Q(R),
i=2

with K[T ] the polynomial ring in one variable. Then s := T ∈ R 0 has

p := S 2 − T 2 ∈ R[S] as an integral equation of minimal degree, but the
linear, non-monic polynomial q := T 2 · S − T 3 ∈ R[S] has s as its zero as
well, so p̃ := p + q is another integral equation of minimal degree.

Proof of 3.4. ”1) =⇒ 2)”: Take an element f (s) ∈ R [s], where f ∈ R [S].
The euclidian algorithm for monic polynomials in R [S] provides aP
decompo-
sition f = qp+r with deg(r) < m := deg(p). Hence f (s) = r(s) ∈ m−1 i
i=0 Rs .
Thus
m−1
X
R [s] = Rsi .
i=0

”2) =⇒ 3)”: The ring P R [s] is a faithful R [s]-module.

”3) =⇒ 1)”: If M = di=1 Rmi is the faithful R [s]-module, we may apply
Lemma 3.7. We obtain a monic polynomial p ∈ R [S] with p(s)M = 0. Since
M is faithful, we find p(s) = 0.

22
Lemma 3.7. Let M be an R [s]-module with a generator system m1 , . . . , md
over R. For every i = 1, . . . , d we choose elements rij ∈ R, j = 1, . . . , d,
with
Xd
smi = rij mj .
j=1

Then the determinant

p := det B ∈ R [S]
of the matrix
B := S · Ed − (rij ) ∈ (R [S])d×d
(with the d-dimensional unit matrix Ed ) is a monic polynomial of degree d
satisfying p(s)M = 0.
Proof. The Leibniz formula for determinants implies, that p ∈ R [S] is a
monic polynomial of degree d. Denoting B # the complementary matrix of
B, i.e.
B # = (−1)i+j (det Bji ) ∈ (R [S])d×d ,

where Bji ∈ (R [S])d−1×d−1 is the matrix obtained from B by deleting the

j-th row and the i-th column, we have

d×d
B # B = p·Ed ∈ R [S]·Ed ⊂ R [S] .
Now substitute s for S and multiply the equality from the right hand side
with the column vector m := (m1 , . . . , md )t ∈ M d . The choice of the entries
rij implies B(s)m = 0. Hence, in M d we obtain
0 = B # (s)B(s)m = p(s)m ,
respectively p(s)mi = 0 for all i, i.e., p(s)M = 0.
Note that for a field R = K and M := K d , mi := ei , i = 1, ..., d, the above
argument is a proof for the Cayley Hamilton Theorem: An endomorphism
of K d is annihilated by its characteristic polynomial. Take s : K d −→ K d
as the endomorphism given by the matrix (rij ), such that K[s] becomes a
commutative subring of End(K d ).

Remark 3.8. 1. If R 0 is integral over R and R 0 finitely generated as R-

algebra, then R 0 is even a finitely generated R-module. On the other
hand, if R 0 is a finitely generated R-module, then R 0 is also integral
over R.

23
 2. If R ,→ R 0 and R 0 ,→ R 00 are integral ring extensions, so is R ,→ R 00 .
3. For every ring extension R ,→ R 0 the integral closure
b := {s ∈ R 0 ; s is integral over R}
R
of R in R 0 is a subring of R 0 .
4. Let R ,→ R 0 be an integral extension of integral domains. Then R is a
field iff R 0 is.
Proof. 1) The second statement follows from condition 3) in 3.4 with M =
R 0 . For the first there is a chain of integral ring extensions
R ,→ R [s1 ] ,→ R [s1 , s2 ] ,→ . . . ,→ R [s1 , . . . , st ] = R 0 ,
which are finite modules over its predecessors. Since that property is ”tran-
sitive” the claim follows. P
2) For c ∈ R 00 let cn + n−1 j
j=0 sj c = 0 be an integral equation over R .
0

Then R [s0 , . . . , sn−1 ] is, according to 1) a finitely generated R-module and

R [s0 , . . . , sn−1 , c] a finitely generated R [s0 , . . . , sn−1 ] -module, so altogether
a finitely generated R-module. As extension of R [c] it is a faithful R [c]-
module, hence condition 3) in 3.4 is satisfied.
3) Assume s and s0 to be integral over R. We have to show, that s+s0 and ss0
are as well. For the integral extensions R ,→ R [s] ,→ R [s, s0 ] the R-algebra
R [s, s0 ] is according to 2) a finitely generated R-module as well as a faithful
module over R [s+s0 ] resp. over R [ss0 ]. Again 3.4 3) is satisfied.
4) Let K := R 0 be a field. The inverse s−1 ∈ K of an element s ∈ R \ {0}
satisfies an integral equation
n−1
X
s−n + rj s−j = 0
j=0

over R with some n ≥ 1. Multiplication with sn−1 yields

n−1
X
−1
s =− rj sn−1−j ∈ R.
j=0

If, on the other hand K := R is a field and s ∈ R 0 \ {0}, then, R 0 being

an integral domain, there is an integral equation p(s) = 0 with a polynomial
p ∈ R [S] not divisible by S. So p = Sq(S) + a with some a ∈ K ∗ := K \ {0}.
But then s(−a−1 q(s)) = 1.

24
Remark 3.9. An integral domain R is called integrally closed in its field of
fractions (or normal) if the ring extension Q(R) ⊃ R satisfies

Q(R) ⊃ R
b = R.

Indeed, if the integral domain R 0 ⊃ R is any extension of such a normal ring

the minimal polynomial of an integral element s ∈ R 0 ⊂ Q(R 0 ) over the field
of fractions Q(R) has coefficients in R and is the unique integral equation of
minimal degree. We mention furthermore, that UFDs are normal rings. The
proof is left as an exercise.

Here is the construction of the subalgebra B ⊂ A for a given affine K-

algebra A.

Theorem 3.10 (Noether normalization lemma). Let K be an infinite field,

A := K[t1 , . . . , tn ] an affine K-algebra and

ϕn : K[T1 , . . . , Tn ] −→ A, Ti 7→ ti

the corresponding surjective homomorphism of K-algebras. Then there ex-

ists a surjective linear map (but not necessarily a coordinate projection)
π : K n −→ K d , such that the pull back of functions

π ∗ : K[S1 , ..., Sd ] −→ K[T1 , ..., Tn ], f 7→ f ◦ π,

provides a finite injective homomorphism of K-algebras

π∗ ϕn
ϕd := ϕn ◦π ∗ : K[S1 , . . . , Sd ] −→ K[T1 , . . . , Tn ] −→ A ,

in particular the ring extension K[S1 , . . . , Sd ] ,→ A is integral.

Remark 3.11. We shall see later on that the number d ∈ N is uniquely

determined. If one admits pull backs with respect to ”nonlinear projections”,
the above theorem is even valid for finite fields K. But then polynomials
n
should be considered as K-valued functions on K , where K is the algebraic
closure of K.
Proof of Th.2.20. For A := K[T ]/m we choose B = π ∗ (K[S1 , ..., Sd ]) accord-
ing to Th.3.10. Since A is a field, so is B ∼
= K[S1 , ..., Sd ], hence d = 0 and A
is a finite field extension of K.

25
Proof of Th.3.10. We do induction on n, the number of generators of the K-
algebra A. For n = 0 we have A = K, so there is nothing to prove. Now let
A∼ = K[T ]/a with T = (T1 , ..., Tn ). If a = 0, we are done. Otherwise we take
some polynomial f ∈ a \ {0}. Indeed, we may assume that f is Tn -monic:
Lemma 3.12. Let K be an infinite field. For any f ∈ K[T ] \ {0} there is
a linear automorphism F : K n −→ K n such that g := F ∗ (f ) is an ”almost
Tn -monic” polynomial
X
g = cTnm + gi (T 0 )Tni ,
i<m

where c ∈ K ∗ and with T 0 := (T1 , ..., Tn−1 ).

Proof. Let f = m
P
i=0 pi be the decomposition of f as sum of homogeneous
polynomials pi of degree i and pm 6= 0. Since K is infinite, there is a
point x ∈ K n \ {0} with pm (x) 6= 0. We may even assume x = (x0 , 1) =
(x1 , . . . , xn−1 , 1) ∈ K n . Now with
F : K n −→ K n , t = (t0 , tn ) 7→ (t0 + tn x0 , tn )
we obtain
m
X
∗ 0 0
F (f ) = f (T + Tn x , Tn ) = pi (T1 + x1 Tn , . . . , Tn−1 + xn−1 Tn , Tn )
i=0
X
= cTnm + gi (T 0 )Tni =: g
i<m
0
with c := pm (x , 1) 6= 0.
Let us now continue with the proof of Th.3.10: If f has Tn -degree m, we
have
m−1
M
K[T ]/(f ) = K[T 0 ] · Tnν
ν=0
0
and choose π1 : K −→ K n n−1
,x = (x , xn ) 7→ x0 . Now consider
A0 := ϕn (K[T 0 ]) ⊂ A.
Then A, being a factor module of K[T ]/(f ), is a finite A0 -module. By induc-
tion hypothesis, there is a linear projection π2 : K n−1 −→ K d , such that
π∗ ϕn−1
2
K[S1 , ..., Sd ] −→ K[T 0 ] −→ A0
with ϕn−1 := ϕn |K[T 0 ] is finite and injective. Finally take π := π2 ◦ π1 .

26
Remark 3.13. We briefly indicate how the above proof has to be modified
if K is a finite field: In that case the pull back automorphism F ∗ : K[T ] −→
K[T ] in the proof of the auxiliary lemma 3.12 has to be replaced with σ :
K[T ] −→ K[T ], where

σ : Ti 7→ Ti + Tnmi for i ≤ n−1, and Tn 7→ Tn

with exponents mi := bn−i , where b is bigger than all the exponents of in-
determinates Ti in the monomials of the polynomial f . Then the maximal
Tn -power in X
σ(T α ) = Tnm + gi (T 0 )Tni
i<m
Pn
is m := m(α) = i=1 αi bn−i . Since different multi-exponents α give different
maximal Tn -exponents m(α), it follows that the maximal Tn -exponent in
σ(f ) is the maximal m(α) for a monomial T α appearing in f and thus σ(f )
is Tn -monic.

Finally, for use later on, we prove the following important result in the
theory of integral extensions:
Theorem 3.14. Let R ,→ R 0 be an integral ring extension. Then
1. given a prime ideal p in R, there is a prime ideal p0 ,→ R 0 lying above
p, i.e. p0 ∩R = p.

2. Furthermore, p ,→ R is a maximal ideal iff p0 ,→ R 0 is, and

3. if p01 ,→ p02 ,→ R 0 lie above p ,→ R, we have p01 = p02 .

Proof. 1.) First we assume that R is a local ring and p its (unique) maximal
ideal m. If then m0 ,→ R 0 is any maximal Ideal, the intersection m0 ∩R ,→ R
is a proper ideal, since it does not contain 1. In the commutative diagram

R ,→ R0
 ↓ 0 0
↓
0
.
R (R ∩ m ) ,→ R /m

the ring R 0 /m0 is a field, which is integral over R/(R ∩ m0 ). According to

Rem.3.8.4 R/(R ∩ m0 ) is a field as well, such that R ∩ m0 is a maximal ideal
in R and therefore m = R ∩ m0 , the ring R being local.

27
For general R and p ⊂ R we consider the multiplicative system S := R \ p.
Then m := S −1 p is the only maximal ideal in the local ring S −1 R, and the
ring extension S −1 R ,→ S −1 R 0 is integral. The first part provides a maximal
ideal m0 ,→ S −1 R 0 over m. Thus we obtain a commutative diagram
ι
p ,→ R ,→ R0 ←- p0
↓ ↓ ↓ .
S −1 ι
m ,→ S −1 R ,→ S −1 R 0 ←- m0

The inverse image p0 ⊂ R 0 of m0 with respect to the natural map R 0 −→

S −1 R 0 is again a prime ideal, lying actually over p: If jR : R −→ S −1 R and
jR 0 : R 0 −→ S −1 R 0 denote the vertical maps, then

p0 ∩ R = ι−1 jR−10 (m0 ) = jR−1 (S −1ι)−1 (m0 )

= jR−1 (m0 ∩ S −1R) = jR−1 (m) = p .

2.) This remark follows from Rem.3.8.4 applied to the integral ring extension
R/p ,→ R 0 /p0 .
3.) We may replace R with R/p and R 0 with R 0 /p01 resp. assume p = 0 = p01
and that R 0 (as well as R) is an integral domain. So we have to show
that for integral domains R ⊂ R 0 and a prime ideal p0 ,→ R 0 we have
p0 ∩ R = 0 =⇒ p0 = 0. Finally replacing R ⊂ R 0 with S −1 R ⊂ S −1 R 0 with
the multiplicative subset S = R \ {0}, Rem. 3.8.4 yields once again that in
the integral ring extension S −1 R ⊂ S −1 R 0 not only S −1 R, but also S −1 R 0 is
a field, so S −1 p0 = 0 resp. p0 = 0.

28
4 Affine Varieties

In this section we treat all the algebraic subsets X ,→ k n (with the ”embed-
ding dimension” n = n(X) depending on X) simultaneously as objects of a
category T A (no standard terminology!), thereby getting rid of any specific
inclusion into some affine space. From a given embedding X ,→ k n we only
keep the information about the functions obtained as restrictions of a poly-
nomial on k n . These regular functions provide the morphisms X −→ k, and
they are associated to the topological space X as an additional datum.

Definition 4.1. Let X ,→ k n be an algebraic set. A function f : X −→ k is

called regular if there is a polynomial g ∈ k[T ] with f = g|X .

A first observation is
Proposition 4.2. The regular functions on an algebraic set X ,→ k n form
a k-algebra O(X). Indeed there is an exact sequence
X %
0 −→ I(X) −→ k[T ] −→ O(X) −→ 0

with the inclusion I(X) ,→ k[T ] and the restriction homomorphism

%X : k[T ] −→ O(X), g 7→ g|X .

Proof. Exercise!

Example 4.3. Here we shall represent the ring O(X) for some examples
either as subalgebra or as extension of some polynomial ring.
1. The Neil parabola X := N (k 2 ; T22 −T13 ) admits a polynomial parametriza-
tion ϕ : k −→ X, s 7→ (s2 , s3 ), hence there is an induced pull back
map ϕ∗ : O(X) −→ k[S], an injection indeed, the map ϕ being onto
(or rather bijective!). Composing it with the restriction homomor-
phism %X : k[T1 , T2 ] −→ O(X) leads to the homomorphism k[T ] −→
k[S], T1 7→ S 2 , T2 7→ S 3 , whence
∞
O(X) ∼
= k[S 2 , S 3 ] ∼
M
=k·1⊕ k · S i.
i=2

29
 2. For char(k) 6= 2 the parametrization

k −→ X := N (k 2 ; T22 − T12 (T1 + 1)), s 7→ (s2 − 1, s(s2 − 1))

of the noose X induces an isomorphism O(X) ∼

= {f ∈ k[S]; f (1) =
f (−1)}.

3. For the reducible algebraic set X = (k × 0) ∪ (0 × k) take disjoint copies

Z1 , Z2 of the affine line k and look at

ϕ : Z1 ∪ Z2 −→ X, Z1 ∈ s1 7→ (s1 , 0), Z2 ∈ s2 7→ (0, s2 ).

It induces an isomorphism

O(X) ∼
= ϕ∗ (O(X)) = {(f1 , f2 ) ∈ k[S1 ] ⊕ k[S2 ]; f1 (0) = f2 (0)} .

Here k[S1 ] ⊕ k[S2 ] denotes the direct sum of the two polynomial rings
k[Si ] (in one variable Si !), the ring operations are componentwise.

4. If a parametrization is not available one can look at a projection: Con-

sider
pr1
ψ : X = N (k 2 , T22 − p(T1 )) ,→ k 2 −→ k.
Since ψ is surjective, the corresponding pull back of functions is injec-
tive, so we may identify ψ ∗ (k[T1 ]) with k[T1 ] and thus obtain
√
O(X) = k[T1 ] · 1 ⊕ k[T1 ] · T 2 ∼
= (k[T1 ])[ p]
2
with T 2 = p(T1 ).

With Hilberts Nullstellensatz Th. 2.17 we can characterize all the alge-
bras O(X) up to isomorphy:

Remark 4.4. A k-algebra A is isomorphic to the algebra O(X) of regular

functions of some algebraic subset X ,→ k n , i.e.

A∼
= O(X),

iff it is affine and reduced.

Here ”reduced” means:

30
Definition 4.5. A ring R is called reduced, if it does not admit non-zero
nilpotent elements, i.e. if √
0=0
holds in R.
√
Proof of 4.4. An affine algebra A ∼
= k[T ]/a is reduced iff a = a, the latter
being equivalent to a = I(X) for X = N (a).

Remark 4.6. An algebraic set X ,→ k n is irreducible iff O(X) is an integral

domain iff any two non-empty open subsets have points in common.

Let us now consider the following category T A of Topological spaces with

a distinguished Algebra of regular functions:

1. The objects are the pairs (X, A), where X is a noetherian topological
space and A ⊂ C(X) a subalgebra of the algebra C(X) of all k-valued
continuous functions on X (where k is endowed with the Zariski topol-
ogy, so a function f : X −→ k is continuous iff all the level sets
f −1 (c), c ∈ k, are closed.) satisfying in addition

A∗ = {f ∈ A; N (f ) = ∅},

i.e. a function f ∈ A is invertible iff it has no zeros. Functions f ∈ A

are also referred to as regular functions on X.

2. A morphism from (X, A) to (Y, B) is a continuous map

ϕ : X −→ Y ,

such that the pull back homomorphism

ϕ∗ : C(Y ) −→ C(X), f 7→ f ◦ϕ ,

maps B ⊂ C(Y ) into A ⊂ C(X), i.e.:

ϕ∗ (B) ⊂ A .

Then the pair (X, O(X)) belongs to this category for any algebraic set
X ,→ k n as a consequence of

31
Remark 4.7. 1. Any maximal ideal m ⊂ O(X) is of the form

m = ma := {f ∈ O(X); f (a) = 0}

with a unique point a ∈ X: Consider the restriction map % := %X :

k[T ] −→ O(X). Then %−1 (m) ,→ k[T ] is a maximal ideal as well, the
map % being onto, hence

%−1 (m) = I({a}) ,→ k[T ]

for some a ∈ k n , according to Th.2.19. Since I({a}) ⊃ I(X), we have

a ∈ X and thus m = %(I({a})) = ma .
2. If f1 , ..., fr ∈ O(X) are functions without a common zero, they are
not contained in any maximal ideal, henceP generate the unit ideal, i.e.
there are functions g1 , ..., gr ∈ O(X) with ri=1 gi fi = 1. In particular
regular functions without zeros are invertible.

For convenience of notation we shall from now on denote the objects

in T A simply by capital letters X, Y, ... and, in analogy to algebraic sets,
O(X), O(Y ), ... their associated algebras of regular functions.

Definition 4.8. An object Y ∈ T A is called an affine variety (over the field

k) if Y ∼= X ,→ k n with some Zariski closed set X in some affine n-space
k n . We denote AV ⊂ T A the full subcategory with the affine varieties as
objects.
Remark 4.9. 1. For an affine variety X the regular functions separate
points, i.e. given two different points x, y ∈ X there is a function
f ∈ O(X) with f (x) 6= f (y).
2. The algebra O(X) is a finite dimensional k-vector space iff |X| < ∞.
In that case we have a ring isomorphism O(X) ∼= k r , where r := |X|.
Example 4.10. Here are two examples of affine varieties X, which are given
independent of any explicit embedding into some k n :
1. Closed subsets of an affine variety are again affine varieties, also called
”closed subvarieties”: A closed subset X ,→ Y of an affine variety Y
inherits the structure of an affine variety: Take

O(X) := O(Y )|X ∼

= O(Y )/I(X)

32
 with the ideal

O(Y ) ←- I(X) := {f ∈ O(Y ); f |X = 0} .

2. An abstract version of k n : Every finite dimensional k-vector space V

is in a natural way a to k dim V isomorphic affine variety: The algebra
O(V ) is generated by the linear forms V −→ k, indeed

O(V ) ∼
= S(V ∗ )

is isomorphic to the symmetric algebra over the dual vector space V ∗ .

(Recall that S(W
L ) for ⊗qa vector space W is a factor algebra of the
tensor algebra ∞ q=0 W : Divide by the two sided ideal generated by
the elements w1 ⊗ w2 − w2 ⊗ w1 ∈ W ⊗2 = W ⊗ W .)
The topology on V is the coarsest topology such that all functions
f ∈ O(V ) become continuous.

Remark 4.11. Here are some morphisms in T A :

1. The inclusion j : X ,→ k n of an algebraic subset into k n is a morphism.

2. Let X ∈ T A. A map ϕ = (ϕ1 , ..., ϕn ) : X −→ k n is a morphism iff

ϕ1 , ..., ϕn ∈ O(X).

3. Let X ∈ T A and j : Y ,→ k n be the inclusion of an algebraic set. Then

a map ϕ : X −→ Y is a morphism iff j ◦ ϕ is.

4. A map ϕ : X −→ Y between embedded affine varieties X ,→ k m and

Y ,→ k n is a morphism iff it is the restriction ϕ = ϕ|
b X of a polynomial
m n
b : k −→ k (then necessarily satisfying ϕ(X)
map ϕ b ⊂ Y ).

5. Let X ∈ T A and Y ,→ k n be an affine variety. Then the morphisms

ϕ : X −→ Y correspond bijectively to the algebra homomorphisms
σ : O(Y ) −→ O(X). For Y = k n that follows from the fact that
algebra homomorphisms σ : k[T1 , ..., Tn ] −→ O(X) correspond to n-
tuples (ϕ1 , ..., ϕn ) ∈ O(X)n (with ϕi = σ(Ti )). In the general case
consider an embedding Y ,→ k n and note that k[T ] −→ O(X), Ti 7→ ϕi ,
factors through O(Y ) ∼ = k[T ]/I(Y ) iff the map ϕ = (ϕ1 , ..., ϕn ) : X −→
n
k satisfies ϕ(X) ⊂ Y .

33
 So it seems, all information about an affine variety X is contained in its
algebra O(X) of regular functions. Indeed:

Theorem 4.12. Denote FRA the category of affine reduced k-algebras. The
functor
O : AV −→ FRA
given by
X 7→ O(X), ϕ 7→ ϕ∗
defines an anti-equivalence of categories.

Proof. According to Rem.4.4 we know that every algebra A ∈ FRA is iso-

morphic to the regular function algebra O(X) of some affine variety X. It
remains to show that O : Mor(X, Y ) −→ Hom(O(Y ), O(X)) is bijective for
all affine varieties X and Y , but here we may refer to Rem.4.11.5.

Let us now describe a functor

Sp : FRA −→ AV

inverse to O : AV −→ FRA. For a reduced affine k-algebra A we set

Sp(A) := {m ,→ A, max. ideal} ,

the “(maximal) spectrum” of the ring A. We want to endow Sp(A) with a

topology and a distinguished algebra O(Sp(A)) of regular functions: Fix a
surjective homomorphism k[T1 , ..., Tn ] −→ A, denote a ,→ k[T ] its kernel and
X := N (a) ,→ k n its zero set. Then

X −→ Sp(A), x 7→ mx

is a bijection and hence can be used to define a topology and regular functions
on Sp(A). Indeed these data depend only on A: The closed sets are the sets

N (b) := {m ∈ Sp(A); b ⊂ m}

with an ideal b ,→ A. It remains to determine O(Sp(A)): For every m ∈

Sp(A) the map ı : k = k·1 ,→ A −→ A/m is an isomorphism. Hence every

34
f ∈ A provides a continuous function fb : Sp(A) → k, m 7→ ı−1 (f +m). If we
then declare these functions to be regular:

O(Sp(A)) := {fb : Sp(A) −→ k; f ∈ A},

we obtain, due to the fact that

\ \
m= mx = {0},
m∈Sp(A) x∈X

an isomorphism
∼
=
ΦA : A −→ O(Sp(A)), f 7→ fb.
A comment on notation: Though the elements in the set Sp(A) are ideals,
one usually prefers a geometric notation and denotes x ∈ Sp(A) its points,
such that, strictly speaking, x = mx .
Finally, given an algebra homomorphism σ : B → A between reduced
affine algebras B and A, the inverse image σ −1 (m) of a maximal ideal m ,→ A
is again maximal, since k ,→ B/σ −1 (m) ,→ A/m ∼ = k implies B/σ −1 (m) ∼= k.
Even better, the map

Sp(σ) : Sp(A) −→ Sp(B) , m 7→ σ −1 (m)

satisfies obviously
gb ◦ Sp(σ) = σ(g)
d
for any g ∈ B and thus

Sp(σ)∗ (O(Sp(B)) ⊂ O((Sp(A)),

so it really is a morphism in our category T A of topological spaces with

distinguished algebra of regular functions.
In order to see that O and Sp are inverse one to the other we have to
construct natural equivalences
∼
=
Φ : idF RA −→ O ◦ Sp.

as well as ∼
=
Ψ : idAV −→ Sp ◦ O.
The first one is defined for A ∈ FRA by the above

ΦA : A −→ O(Sp(A)), f 7→ fb,

35
while
ΨX : X −→ Sp(O(X)), x 7→ mx
is the natural equivalence over an affine variety X ∈ AV: We need to check
that given a morphism ϕ : X −→ Y , we have

ΨY ◦ ϕ = Sp(O(ϕ)) ◦ ΨX .

We evaluate the LHS and the RHS at a point x ∈ X and obtain

?
mϕ(x) = (ϕ∗ )−1 (mx ),

an equality readily verified.

Let us now briefly discuss some particular morphisms:

Definition 4.13. A morphism ϕ : X −→ Y is called

1. dominant, if Y = ϕ(X);
∼
=
2. a closed embedding, if it admits a factorization X −→ Z ,→ Y , where
Z ,→ Y is a closed subvariety of Y ;

3. finite, if O(X) is a finite(ly generated) O(Y )-module.

Proposition 4.14. Let ϕ : X −→ Y be a morphism of affine varieties.

1. The morphism ϕ is dominant, iff ϕ∗ : O(Y ) −→ O(X) is injective.

2. The morphism ϕ is a closed embedding, iff ϕ∗ : O(Y ) −→ O(X) is

surjective.

3. A finite dominant morphism ϕ : X := Sp(A) −→ Y := Sp(B) is

onto (:= surjective) and has finite fibers: For every y ∈ Y we have
|ϕ−1 (y)| ≤ r, if the finite B-module A can be generated by r elements.

4. A finite morphism is closed.

36
Proof. The first statement as well as the implication ”=⇒” of the second
statement is obvious. For the reversed one take Z := ϕ(X). Then X −→ Z is
an isomorphism, since the pull back homomorphism O(Z) −→ O(X) is. For
the third part take a point y ∈ Y . Since ϕ∗ is injective and O(X) ⊃ O(Y )
an integral extension, we may apply Th.3.14 and obtain a maximal ideal
m0 ,→ O(X) withP m0 ∩ O(Y ) = my . But m0 = mx for some x ∈ X and thus
y = ϕ(x). If A = ri=1 Bai , then A/my · A is generated by the residue classes
ai + my · A over B/my ∼ = k, the same holds for O(ϕ−1 (y)) ∼ = A/I(ϕ−1 (y)),
a factor ring of A/my · A. So O(ϕ−1 (y)) is a finite dimensional k-vector
space of a dimension ≤ r, hence the affine variety ϕ−1 (y) is finite with at
most r points. Finally let ϕ : X −→ Y be finite and Z ,→ X be a closed
subset. Then Z −→ ϕ(Z) is a finite dominant map, hence surjective and
thus ϕ(Z) = ϕ(Z).

We include here a geometric reformulation of the Noether normalization

theorem 3.10:

Corollary 4.15. For any irreducible variety X there is a finite surjective

morphism ϕ : X −→ k d . If X = N (k n ; f ) ,→ k n with a nonconstant polyno-
mial f we may choose d = n − 1.

Remark 4.16. Let us explain for k = C the geometry behind the notion of
a finite map:

1. First of all we note that a complex affine variety X is compact (w.r.t.

the strong topology) iff it is finite. The nontrivial implication follows
with the above geometric version of the Noether normalization lemma:
For compact X we have d = 0.

2. A morphism ϕ : X −→ Y of complex affine varieties is finite if and only

if it is (w.r.t. the strong topologies) proper, i.e. if for any compact set
K ⊂ Y the inverse image ϕ−1 (K) ⊂ X is compact as well. Obviously
that implies that ϕ has finite fibers, but that is definitely not sufficient:
The map
pr1
N (T1 T2 − 1) ,→ C2 −→ C
has finite fibers, but is not proper.

37
 3. The most suggestive characterization of properness is as follows: A
morphism ϕ : X −→ Y between complex affine varieties is proper if
and only if a point sequence (xn ) ⊂ X has points of accumulation
in X iff the sequence (ϕ(xn )) ⊂ Y has in Y . The implication ”=⇒”
makes use of the continuity of ϕ, while for ”⇐=” consider the compact
set K consisting of the points yn and the accumulation points of that
sequence. One could thus say, that ϕ : X −→ Y is proper iff the space
X has ”no holes over Y ”.

A closed subset Z ,→ X of an affine variety X is again an affine variety

with the regular function algebra

O(Z) := O(X)|Z ∼
= O(X)/I(Z).

But what can we do in order to give open subsets U ⊂ X also some structure,
say, to make them objects in the category T A? We have to look for a good
notion of regular functions on U . We could once again try with the following
algebra
O(U ) := O(X)|U .
But there is some embarrassing fact: If U is dense in X - this is always the
case for irreducible X and non-empty U - we get

O(X)|U ∼
= O(X).

In particular for the subalgebra O(X)|U ⊂ C(U ) it need not be true that
functions without zeros are invertible. So we have to enlarge that algebra.
For certain open subsets that is easily done:
Definition 4.17. Let X be an affine variety.
1. A principal open subset U ⊂ X is an open subset U of the form

U = Xf := X \ N (f )

with N (f ) := N (X; f ) for a suitable function f ∈ O(X).

2. For any f ∈ O(X) \ {0} we define

 
h
O(X)f := : Xf −→ k; h ∈ O(X), m ∈ N .
fm

38
Remark 4.18. 1. Any Zariski open set U ⊂ X is a finite union S of princi-
pal open sets: We can write X \ U = N (f1 , ..., fr ), hence U = ri=1 Xfi .

2. For an ideal a ,→ O(X) in the algebra of regular functions on an affine

variety X Hilberts Nullstellensatz holds as well:
√
I(N (a)) = a.

That follows immediately from Th.2.17 by embedding X ,→ k n into

some affine space.

3. O(X)f is a reduced affine k-algebra: If σ : k[T ] −→ O(X) is onto, so

is the extension k[T, S] −→ O(X)f , Ti 7→ σ(Ti ), S 7→ f −1 .

4. For principal open sets Xf ⊃ Xg we have

O(X)f |Xg ⊂ O(X)g .

The inclusion Xg ⊂ Xf tells us that g ∈ I(N (f )) and thus g ` ∈ (f ) for

some ` ∈ N by Hilberts Nullstellensatz. If, say, g ` = g̃f we have
h g̃ m h
=
fm g `m
on Xg .
Since in particular O(X)f = O(X)g for Xf = Xg we may define:
Definition 4.19. The algebra of regular functions O(U ) on a principal open
subset U = Xf of an affine variety X is defined as

O(U ) := O(X)f .

Remark 4.20. 1. A regular function on U = Xf without zeros is invert-

ible: We may assume that it is of the form h|U with h ∈ O(X). But
already the restriction of h to Xh ⊃ Xf is invertible!

2. The restriction O(X) −→ O(Xf ) is an injection if f ∈ O(X) is not a

zero divisor. In particular that holds true automatically if O(X) is an
integral domain.

3. If X = Sp(A) is irreducible, we obtain an inclusion O(Xf ) = Af ⊂

Q(A).

39
 So a principal open set U ⊂ X defines an object (U, O(U )) ∈ T A. Indeed,
it is again an affine variety:
Proposition 4.21. For a principal open subset Xf ⊂ X of an affine variety
X we have
Xf ∼= Sp(O(X)f ).

Before we prove Prop.4.21 we need cartesian products of affine varieties:

Remark 4.22. If X ,→ k m and Y ,→ k n are affine varieties, so is

X × Y ,→ k m × k n = k m+n .

Note that the homomorphism

O(X) ⊗ O(Y ) −→ O(X × Y )

induced by the bilinear map

O(X) × O(Y ) −→ O(X × Y ), (f, g) 7→ pr∗X (f ) · pr∗Y (g)

is an isomorphism (since linear independent functions fi ∈ O(X) and gj ∈

O(Y ) give linear independent functions pr∗X (fi ) · pr∗Y (gj )); hence the affine
variety X × Y depends only on the affine varieties X and Y and not on
the chosen embeddings. But note that the Zariski topology on X × Y is,
except in trivial cases, strictly finer that the product topology of the Zariski
topologies on the factors X and Y .
Proof of 4.21. Let X = Sp(A). It suffices to show that Xf is affine, indeed
isomorphic to the affine variety

Γ1/f := N (X × k; 1 − f S) ,→ X × k ∼
= Sp(A[S]),

the graph of the function 1/f : Xf −→ k considered as subset of X × k, the

map
1
Xf −→ Γ1/f , x 7→ (x, )
f (x)
being an isomorphism. Note that we even have

O(Xf ) ∼
= A[S]/(1 − f S).

40
 Finally, as a preparation of the next section, we prove that a locally
regular function on an affine variety is regular:

Proposition 4.23. Let X = ri=1 Ui be an open cover of an affine variety

S
by principal open subset Ui = Xfi . Then a function h : X −→ k is regular
iff all the restrictions h|Ui are regular for i = 1, ..., r.

Proof. Write h|Ui = hi /fimi with functions hi ∈ O(X). Then we have fimi h =
hi on Ui and fimi +1 h = fi hi even on all of X. Since the functions
Pr fimi +1 have
mi +1
no common zero, there are functions gi ∈ O(X) with i=1 gi fi = 1.
Then r r
X X
mi +1
h=( gi f i )h = gi fi hi ∈ O(X).
i=1 i=1

Remark 4.24. 1. Let ϕ : X −→ Y be a morphism of affine varieties.

The inverse image of a principal open subset V = Yg is obviously again
a principal open subset: U := ϕ−1 (V ) = Xϕ∗ (g) .

2. Let Y = sj=1 Vj be an open cover of an affine variety by principal open

S
subset Vj . Then ϕ : X −→ Y is a closed embedding resp. finite resp.
an isomorphism iff all the maps ϕ|Uj : Uj := ϕ−1 (Vj ) −→ Vj , j = 1, ..., s,
are.

Proof. Exercise!

41
5 Prevarieties
Affine varieties are special objects in the category T A of topological spaces
with a distinguished algebra of regular functions. In order to define algebraic
varieties we have to replace T A with the category of ringed spaces where one
has not only a distinguished subalgebra O(X) on the entire space X, but for
every open set U ⊂ X. That leads to the notion of a sheaf of functions:
Definition 5.1. Let X be a topological space.
1. A sheaf F of (continuous) k-valued functions on X consists of a choice
of a subalgebra F(U ) ⊂ C(U ) for every open subset U ⊂ X such that
(calling functions ∈ F(U ) regular functions)
(a) regular functions restrict to regular functions: F(U )|V ⊂ F(V ),
whenever V ⊂ U are open sets, and
S
(b) locally regular functions are regular: If f ∈ C(U ) for U = i∈I Ui
and f |Ui ∈ F(Ui ) for all i ∈ I, then already f ∈ F(U ).
2. A (k-)structure sheaf on X is a sheaf F of (continuous) k-valued func-
tions on X, such that
F(U )∗ = C(U )∗ ∩ F(U )
for all U ⊂ X, i.e. regular functions without zeros are invertible.
Example 5.2. 1. F(U ) := C(U ) is a sheaf of functions, denoted simply
C.
2. L(U ) := {f : U −→ k locally constant} is called the sheaf of locally
constant functions.
3. Let X be an affine variety. For principal open subsets U ⊂ X we have
already defined O(U ). For arbitrary open U ⊂ X we set then
O(U ) := {f : U −→ k; f |V ∈ O(V ) for all principal open subsets V ⊂ U }
S
and obtain a structure sheaf on X: Let V ⊂ U = i∈I Ui be a principal
open subset and f : U −→ S k with f |Ui ∈ O(Ui ). Since V is quasi-
compact, we may refine V = i∈I V ∩ Ui by a finite covering V = V1 ∪
... ∪ Vr with principal open subsets Vj ⊂ V . Then we have f |Vj ∈ O(Vj )
for j = 1, ..., r and Prop.4.23 tells us that f |V ∈ O(V ). Since that
holds for any principal open subset V ⊂ U , we get f ∈ O(U ).

42
 4. Take k = C and endow it with its strong topology. Then we define on
X = Cn (with the strong topology as well) two sheaves of functions H
and E, namely

H(U ) := {f : U −→ C holomorphic}

and
E(U ) := {f : U −→ C differentiable},
where a differentiable function means a C ∞ -function in the ”real sense”,
i.e. if we use C ∼
= R2 and Cn ∼ = R2n .
Note that for Zariski open U ⊂ Cn we have inclusions

O(U ) $ H(U ) $ E(U ).

Remark 5.3. Let U ⊂ X be an open subset of the affine variety X = Sp(A).

Then
O(U ) ⊂ O(X)h |U ,
whenever U ⊂ Xh , i.e. h ∈ O(X) does not vanish on U , but in general a
regular function f ∈ O(U ) need not admit a representation f = g/h on the
entire set U . If U = V1 ∪ ... ∪ Vr with special open sets Vi = Xhi , we can
write f = gi /hi on each Vi , such that for irreducible X, we may write
r
\
O(U ) = O(Vi ) ⊂ Q(A).
i=1

Only for a UFD A, that can be simplified to

r
\
O(U ) = Ahi = Ah ,
i=1

where h = gcd(h1 , .., hr ). In particular U ⊂ Xh .

Example 5.4. 1. With the above remark 5.3 we find for U := (k 2 )∗ :=

k \ {0} that O(U ) = O(k 2 )|U ∼
2
= k[T1 , T2 ]. Note that (U, O(U )) ∈ T A
is not an affine variety, since the maximal ideal m := (T1 , T2 ) is not of
the form ma for some a ∈ U .

43
 2. Let us consider the Segre cone X := N (k 4 ; T1 T4 − T2 T3 ) ,→ k 4 . Take
Xi := Xgi with gi := Ti |X . Consider now the open set

U := X \ 0 × 0 × k 2 = X1 ∪ X2 .

Since g1 g4 = g2 g3 we may define a regular function f ∈ O(U ) by

g3 g4
f |X1 := , f |X1 := ,
g1 g2

but there is no representation f = g/h with regular functions g, h ∈

O(X) with a denominator nowhere vanishing on U , or equivalently,
there is no proper principal open subset Xh ⊃ U . Indeed O(U )∗ = k ∗ .
For the proof consider the morphism ϕ : X −→ k 2 , (t1 , ..., t4 ) 7→ (t1 , t2 ).
It has fibers ϕ−1 (t1 , t2 ) = k(t1 , t2 ) ∼ = k for (t1 , t2 ) ∈ k 2 \ {0} and
ϕ−1 (0, 0) = 0 × 0 × k 2 ∼ = k 2 , furthermore a section σ : U −→ X, namely
σ(t1 , t2 ) := (t1 , t2 , 0, 0). We claim that ϕ∗ : O(k 2 \ {0})∗ −→ O(U )∗ is
an isomorphism. A regular function on k without zero is a constant,
hence f is constant along the fibers of ϕ and f = ϕ∗ (g) with g = σ ∗ (f ).
Finally we know from the first point, that O(k 2 \ {0}) ∼ = k[T1 , T2 ].
∼ 2 3
Indeed U = (k \ {0}) × k ⊂ k holds in T A.

3. There are even examples of affine varieties X admitting open subsets

U ⊂ X, such that O(U ) is not even an affine algebra!

Here is the analogue of the category T A:

Definition 5.5. A k-ringed space (X, O = OX ) is a topological space to-

gether with a k-structure sheaf O. A morphism between k-ringed spaces
(X, OX ) and (Y, OY ) is a continuous map ϕ : X −→ Y , such that

ϕ∗ (OY (V )) ⊂ OX (ϕ−1 (V ))

holds for all open subsets V ⊂ Y . We denote RS = RS k the category of

k-ringed spaces.

Proposition 5.6. The category AV of affine varieties is equivalent to a full

subcategory of RS k .

44
Proof. Given an affine variety X ∈ AV define its structure sheaf OX as in
the above example 5.2.3. Now assume that ϕ : X −→ Y is a morphism of
affine varieties. We have to show that ϕ is also a morphism in RS k . For a
principal open subset V = Yg we obviously have U := ϕ−1 (V ) = Xϕ∗ (g) and

ϕ∗ (O(V )) = ϕ∗ (O(Y )g ) ⊂ O(X)ϕ∗ (g) = O(U ).

Here we have, for convenience of notation, written O in order to denote OX

and OY , and ϕ∗ denotes the pull back OY (V ) −→ OX (ϕ−1 (V )) for any open
V ⊂Y.
In the general case write the open set V ⊂ Y as a finite union V =
V1 ∪ ... ∪ Vs of principal open subsets Vj ⊂ Y . Given f ∈ O(V ), we see that
ϕ∗ (f ) is regular on Uj := ϕ−1 (Vj ) and hence, using condition b) of Def.5.1.1.,
regular on U = ϕ−1 (V ) = U1 ∪ ... ∪ Us as well.
Finally Th.4.12 gives that any morphism X −→ Y in AV is induced by
exactly one morphism X −→ Y in RS k .

Remark 5.7. 1. Let X be a k-ringed space and Y be an affine variety.

Then ϕ 7→ ϕ∗ provides a bijection between the morphisms X −→ Y
and the algebra homomorphisms O(Y ) −→ O(X).
2. Every open subset U ⊂ X of a k-ringed space (X, OX ) is again a k-
ringed space with the structure sheaf OU satisfying OU (V ) = OX (V )
for open V ⊂ U , making the inclusion U −→ X a morphism.
3. An arbitrary subset Y ,→ X of a k-ringed space (X, OX ) is again a
k-ringed space as well: We want again that the inclusion Y ,→ X
becomes a morphism in RS k , but the definition of the structure sheaf
OY is more complicated: Call a function f : V −→ k on an open
subset V ⊂ Y locally extendible to X if any point y ∈ V admits an
open neighborhood U ⊂ X with U ∩ Y ⊂ V and a function g ∈ OX (U )
extending f |U ∩Y . Now set

OY (V ) := {f : V −→ k locally extendible to X} .

Warning: If U ⊂ X is open and V := U ∩ Y , the restriction homomor-

phism OX (U ) −→ OY (V ) need not be surjective!
Definition 5.8. An algebraic prevariety is a k-ringed space X admitting an
open cover X = U1 ∪ ... ∪ Ur by open subsets Ui which are affine varieties.

45
Example 5.9. 1. Open subsets U ⊂ X of an affine variety X are preva-
rieties, U being a finite union of principal open subsets, but not nec-
essarily affine (they are called instead ”quasi-affine”): The prevariety
U := k n \ {0} ⊂ X = k n is not affine for n > 1.

2. A closed subspace Y ,→ X of a prevariety X is again a prevariety. This

is an immediate consequence of:

Lemma 5.10. Let X = Sp(A) be an affine variety and Y ,→ X be a closed

subspace, endowed with the induced structure sheaf OY . Then Y ∼
= Sp(B)
with B = O(X)/I(Y ).

Proof. The restriction homomorphism A = OX (X) −→ OY (Y ) factors through

OX (X) −→ B := OX (X)/I(Y ) −→ OY (Y )

respectively j : Y −→ X as

Y −→ Z := Sp(B) −→ X.

As topological subspaces of X we have Y = Z, and we have to show that, for

all open subsets V ⊂ Y = Z, the inclusion OZ (V ) ⊂ OY (V ) is an equality.
Because of condition b) in Def.5.1.1 we may assume that V ⊂ Z is a special
open set; and writing it as V = Z ∩ U with a special open subset U ⊂ X, we
see that we may assume U = X and V = Z = Y . Take a function f ∈ OY (Y ).
By definition, we find principal open subsets U1 , ..., Ur ⊂ X covering Y = Z
and functions gi ∈ OX (Ui ) with gi |Vi = f |Vi , where Vi := Ui ∩ Z. Hence
f |Vi ∈ OZ (Vi ) for i = 1, ..., r resp. f ∈ OZ (Z).

Corollary 5.11. For a morphism ϕ : X −→ Y of prevarieties the following

statements are equivalent:

1. It is a closed embedding, i.e. admits a factorization

∼
=
X −→ Z ,→ Y,

where Z ,→ Y is the inclusion of a closed subspace.

2. For any affine open subspace V ⊂ Y the inverse image U := ϕ−1 (V ) ⊂

X is affine and ϕ∗ : O(V ) −→ O(U ) surjective.

46
 3. There is an open affine cover Y = si=1 Vi , such that Ui := ϕ−1 (Vi ) ⊂ X
S
is affine and ϕ∗ : O(Vi ) −→ O(Ui ) surjective for i = 1, ..., s.

Let us now come back to algebraic geometry and present some important
examples of prevarieties:

Quotient Constructions: A general idea how to produce new prevarieties

X from a given one Z is to look at quotient spaces X = Z/ ∼ of Z modulo an
equivalence relation ”∼” on Z, or, with other words, to take any surjective
map π : Z −→ X from Z to some set X. Then X is endowed with its
π-quotient topology:

U ⊂ X open : ⇐⇒ π −1 (U ) ⊂ Z open.

The construction of the structure sheaf OX uses the same idea:

OX (U ) := {f ∈ C(U ); π ∗ (f ) ∈ OZ (π −1 (U ))}.

Unfortunately, there is in general no cover of X by open affine subspaces, so

the resulting object is a priori just a k-ringed space. Nevertheless, sometimes
one has good luck:
Projective n-space: Projective n-space is a prevariety

Pn = Pn (k) = k n ∪ {∞L ; L ⊂ k n },

which is obtained from k n by adding to every line, i.e. one-dimensional

subspace, L ⊂ k n a separate ”point at infinity”, denoted ∞L . Indeed, we
take
Z := (k n+1 )∗ := k n+1 \ {0}
and
Pn = Pn (k) := L ,→ k n+1 one dimensional subspace


with the map

π : (k n+1 )∗ −→ Pn (k), (z0 , ..., zn ) 7→ [z0 , ..., zn ] := k(z0 , ..., zn ).

The numbers z0 , ..., zn are called the homogeneous coordinates of the point
[z0 , ..., zn ] ∈ Pn , they are unique up to a common nonzero multiple λ ∈ k ∗ . In

47
order to see that the k-ringed space Pn is a prevariety we consider the open
cover n
[
Pn = Ui
i=0
with the open subspaces
Ui := {[z]; zi 6= 0} ∼
= kn,
the latter isomorphism being
 
n z0 zi−1 zi+1 zn
Ui −→ k , [z] 7→ , ...., , , ..., .
zi zi zi zi
In particular we can identify
∼
=
k n −→ U0 , z = (z1 , ..., zn ) 7→ [1, z],
the point ∞L for L = kz being the point [0.z]. For n = 1 we thus get the
”projective line”
P1 = k ∪ {∞}
with ∞ := [0, 1]. For k = C the projective line P1 is (with respect to the
strong topology) homeomorphic to the 2-sphere S2 , the ”Riemann sphere” of
complex analysis. But note that Pn is not homeomorphic to S2n for n > 1.

We finish this section with the remark that the quotient map π : (k n+1 )∗ −→
Pn belongs to a particularly interesting class of morphisms:
Definition 5.12. A morphism π : E −→ X between prevarieties E S and X
is called a k -principal bundle if there is an open affine cover X = ri=1 Ui
∗

together with isomorphisms (”trivializations”)

τi : π −1 (Ui ) −→ Ui × k ∗
satisfying
1. prk∗ ◦ τi = π with the projection prk∗ : Ui × k ∗ −→ k ∗ , giving rise to
2. transitions
τj ◦ τi−1 : Uij × k ∗ −→ Uij × k ∗
looking as follows
(x, λ) 7→ (x, fij (x)λ)
with (necessarily nowhere vanishing) functions fij ∈ O(Uij ).

48
Example 5.13. The quotient map π : (k n+1S)∗ −→ Pn is a k ∗ -principal bun-
dle: Consider the standard open cover Pn = ni=0 Ui with the open subspaces
Ui := {[z]; zi 6= 0} ∼
= k n and define

τi : π −1 (Ui ) −→ Ui × k, (z0 , ..., zn ) 7→ ([z0 , ..., zn ], zi ).

Determine the corresponding transition functions fij !

49
6 Algebraic Varieties
In ”real” (i.e. non Zariski) topology reasonable spaces are Hausdorff, and
also in algebraic geometry one needs a corresponding condition: Reasonable
prevarieties, to be called later on simply (algebraic) varieties, should be ”sep-
arated”. But since even affine varieties, the local models for prevarieties, are
very far from being Hausdorff, we can not use literally the same definition
as in topology. First of all here is an example of the anomalies we want to
avoid:

Example 6.1. Starting with the disjoint union

Z := N k 2 ; T2 (T2 −1) = k×{0} ∪ k×{1} ∼

= Sp (k[T ] ⊕ k[S])

of two lines we consider the ringed space

X := Z/ ∼ ,

where
(x, i) ∼ (y, j) : ⇐⇒ x = y ∈ k ∗ or i = j.
Then
X = V0 ∪ V1 = V01 ∪ {o0 , o1 }
with the open subsets Vi := π(k × {i}) ∼ = k, so X is a prevariety, and
∼ ∗
V01 = V0 ∩ V1 = k . So X is an affine line with the two different origins
oi := π(0, i), i = 0, 1.

Now a topological space X is Hausdorff iff the diagonal

∆ := {(x.x); x ∈ X} ⊂ X × X

is a closed subspace of X × X, the product of the topological space X with

itself. So we could try to use that formulation: As we already have seen for
affine varieties X, Y , the Zariski topology on X × Y is in general strictly finer
than the product topology, and thus, the above condition is in the context
of prevarieties much weaker than the usual Hausdorff property. But first of
all we have to define the product X × Y for prevarieties X, Y .

Proposition 6.2. In the category of prevarieties (i.e. the full subcategory of

RS k with the prevarieties as objects) there exists products.

50
Proof. For affine X, Y we have already defined X × Y . In the general case we
have to endow the cartesian product X ×Y with theS structure of a prevariety.
In order to do so, we take open affine coverings X = ri=1 Ui and Y = sµ=1 Vµ
S
and ”patch together” the affine varieties Ui × Vµ : Consider the prevariety Z
defined as the disjoint union
[
Z := Ziµ ,
i,µ

of the affine varieties

Ziµ := Ui × Vµ .
Now endow X × Y with the π-quotient structure with respect to the natural
map π : Z −→ X × Y , which on Ziµ is just the inclusion into X × Y . If we
can show that π(Ziµ ) ⊂ X × Y is open and

π|Ziµ : Ziµ −→ π(Ziµ )

an isomorphism, we are done. To that end let

(Ziµ )jν := Uij × Vµν ⊂ Ziµ ,

considered as open (ringed) subspace of Ziµ , where Uij := Ui ∩ Uj and Vµν :=

Vµ ∩ Vν . We have to check that the identity map
id
(Ziµ )jν −→ (Zjν )iµ

is an isomorphism. That can be done locally: Take a point (x, y) ∈ Uij ×

Vµν . There are neighborhoods U ⊂ Uij of x and V ⊂ Vµν of y, which are
simultaneously principal open subsets of Ui and Uj resp. of Vµ and Vν : Take
for U the intersection of a principal open neighbourhood of x in Ui and a
principal open neighbourhood of x in Uj (using 6.3.1), and for V apply the
same recipe. Then the cartesian product U × V is an open subset of both
Ziµ and Zjν and the respective induced ringed space structures coincide with
that one belonging to U × V as the product of the affine varieties U and V ,
as follows from Lemma 6.3, part 2.
Lemma 6.3. 1. Let X be an affine variety, U = Xf and V = Ug be
principal open subsets of X resp. U . Then V is a principal open subset
of X.

51
 2. Let X, Y be affine varieties, f ∈ O(X), g ∈ O(Y ). Then

(X × Y )f ⊗g ∼
= Xf × Yg .

Proof. Do the first part yourself! For the second part, note that both the
RHS and the LHS are affine varieties with the same underlying set, hence it
suffices to check that the regular function algebras agree.

To finish the proof of Prop 6.2 we have to check that the prevariety X ×Y
is a product of X and Y in the category RS k . That follows from the fact,
that given morphisms ϕ : Z −→ X and ψ : Z −→ Y , the map (ϕ, ψ) : Z −→
X × Y is a morphism as well, since the restrictions (ϕ, ψ)|(ϕ,ψ)−1 (Ui ×Vµ ) :
(ϕ, ψ)−1 (Ui × Vµ ) −→ Ui × Vµ are.

So, eventually, we can define algebraic varieties:

Definition 6.4. A prevariety X is called separated or an algebraic variety

(over k) if the diagonal ∆ ⊂ X × X is closed in X × X.

Remark 6.5. 1. Affine n-space k n is separated, since

∆ = N (k n × k n ; S1 − T1 , ..., Sn − Tn ).

2. If ϕ : X −→ Y is an injective morphism from a prevariety X into an

algebraic variety Y , then X is an algebraic variety as well: We have
∆X = (ϕ × ϕ)−1 (∆Y ).

3. Open and closed subspaces of an algebraic variety are separated.

4. Affine and more generally quasi-affine varieties are separated.

5. S
In the general case of a prevariety X with an affine open cover X =
r
i=1 Ui we thus get that

r
[
∆ ,→ Ui × Ui
i=1

is a closed subspace of the open set

52
 Sr
i=1 Ui × Ui ⊂ X × X and that the diagonal map

δ : X −→ ∆, x 7→ (x, x),

is an isomorphism of prevarieties. Hence X is separated iff ∆ij :=

∆ ∩ (Ui × Uj ) is a closed subset of Ui × Uj for i 6= j. That leads to the
following reformulation of the ”diagonal criterion”:
Proposition 6.6. Assume X = ri=1 Ui is an open affine cover of the pre-
S
variety X. Then X is separated iff the intersections Uij := Ui ∩ Uj ⊂ X are
affine for all i, j and

O(Ui ) ⊗ O(Uj ) −→ O(Uij ), f ⊗ g 7→ f |Uij · g|Uij ,

is surjective for all 1 ≤ i, j ≤ r.

Proof. Let us first remark that the above ring homomorphism is nothing but
the pull back morphism

δ ∗ : O(Ui × Uj ) −→ O(Uij )

belonging to the diagonal map δ|Uij : Uij −→ Ui × Uj .

”=⇒”: If ∆ ⊂ X is closed, then Uij ∼ = ∆ ∩ (Ui × Uj ) ,→ Ui × Uj is affine and
the pullback of the inclusion surjective.
”⇐=”: Since the intersection Uij is affine, the given condition means that
the morphisms

δ|Uij : Uij = δ −1 (Ui × Uj ) −→ Ui × Uj

are closed embeddings. Since the Ui × Uj cover X × X, the morphism δ :

X −→ X × X is itself a closed embedding.

Proposition 6.7. Projective n-space Pn is an algebraic variety.

Proof. Denote Pn = ni=0 Ui the standard open cover of Pn . The intersection
S

Uij ∼
= k i × k ∗ × k j−i−1 × k ∗ × k n−j

is affine - we assume j > i. We treat the injections

π∗
O(U ) −→ O(π −1 (U ))

53
as inclusions and thus
O(U ) = O(π −1 (U ))0 := {f ∈ O(π −1 (U )); f (λt) = f (t), ∀λ ∈ k ∗ }.
In particular we have  
T0 Tn
O(Ui ) = k , ..., ,
Ti Ti
while  
Tµ Tν
O(Uij ) = (k[T ]Ti Tj )0 = k ; 0 ≤ µ, ν ≤ n .
Ti Tj
Hence the homomorphism
     
T0 Tn T0 Tn Tµ Tν
k , ..., ⊗k , ..., −→ k ; 0 ≤ µ, ν ≤ n ,
Ti Ti Tj Tj Ti Tj
Tµ Tν Tµ Tν
⊗ 7→
Ti Tj Ti Tj
is surjective.

6.1 Digression: Complex Analytic Spaces

This section is a (non-obligatory) supplement about the relationship between
complex algebraic and complex analytic geometry. No proofs are presented.

We consider on Cn (endowed with the strong topology) the structure

sheaf
H : U 7→ H(U ) := {f : U −→ C holomorphic}
of holomorphic functions. Following the recipe of Rem. 5.7 we then may
consider any subset of Cn as a ringed space. Actually we only need open
subsets W ⊂ Cn and analytic subsets Z ,→ W of an open set W ⊂ Cn , i.e.
(relatively) closed subsets, such that for any point a ∈ Z there is an open
neighbourhood U ⊂ W together with finitely many holomorphic functions
f1 , ..., fr ∈ H(U ) such that
Z ∩ U := N (U ; f1 , ..., fr ) := {z ∈ U ; f1 (z) = ... = fr (z) = 0}.
In particular, for a (relatively) open subset V ⊂ Z the algebra H(V ) consists
of all functions V −→ C, which locally can be extended to a holomorphic
function on some open subset of Cn .

54
Definition 6.8. 1. A complex analytic space (or complex analytic variety)
S
X is a C-ringed Hausdorff space admitting an open cover X = i∈I Ui
with open subspaces Ui ∼
= Zi ,→ Wi , where Zi is an analytic subset of
ni
the open subset Wi ⊂ C .
S
2. A complex n-manifold is a complex analytic space X = i∈I Ui with
open subspaces Ui ∼
= Wi , the Wi being open subspaces Wi ⊂ Cn .
3. A Riemann surface is a connected complex 1-manifold.

Remark 6.9. 1. An affine variety X ,→ Cn is an analytic subset of Cn ,

hence a complex analytic space, denoted Xh . So as a set Xh = X, while
the topology is ”strengthened” and regular functions are replaced with
holomorphic functions.
2. To a complex
Sr algebraic variety X we can, using an open affine cover
X = i=1 Ui , associate a complex analytic variety Xh with Xh = X
as sets. (Here separatedness of X gives the Hausdorff property for
Xh ). We thus obtain a functor X 7→ Xh from the category of complex
algebraic varieties to the category of complex analytic spaces.
3. A complex algebraic variety X is connected iff Xh is.
4. As a consequence of the (holomorphic) inverse function theorem an
algebraic subset
X := N (Cn ; f1 , ..., fr )
with polynomials f1 , ..., fr ∈ C[T ] provides a complex (n−r)-manifold
Xh if  
∂f
rank (a) = r
∂T
holds with the Jacobi matrix
 
∂f ∂fi
(a) := (a) ∈ Cr,n
∂T ∂Tj
for all points a ∈ X. In particular the isomorphisms ϕ : Ui −→ Wi ,
the ”local coordinates” or ”local charts” in classical terminology, can
be chosen always as restrictions of linear functions, but the Ui are in
general not Zariski open nor are the inverse local coordinates polyno-
mials.

55
 A complex manifold is a complex analytic space. Here are some remarks
how far the latter are from the former:

Definition 6.10. A point a ∈ X of a complex analytic space is called a

singular point (or a singularity) if it does not admit an open neighbourhood
isomorphic to an open subset in some Cn . A complex analytic space without
singular points is called nonsingular or smooth; its connected components
form complex manifolds.

Remark 6.11. The singular locus S(X) of a complex analytic space X is a

nowhere dense closed subset of X, locally given as the zero locus (:= set of
zeros) of finitely many holomorphic functions. Its complement X \ S(X) is
called the regular locus of X.

Example 6.12. For the Neil parabola X := N (C2 ; T22 − T13 ) and the noose
Y := N (C2 ; T22 − T12 (T1 + 1)) the only singular point of Xh resp. Yh is the
origin. The affine curves Z := N (C2 ; T22 − p(T1 )) induces a Riemann surface
Zh .

Remark 6.13. 1. The singular locus S(Xh ) of a complex algebraic vari-

ety X is even a Zariski closed subset and can be defined without the
passage to complex analytic spaces. Indeed, the definition works for
any field k, but it is not literally the analogue of the definition for
complex analytic spaces.

2. For a connected complex affine variety X the theorem of Liouville holds:

A bounded holomorphic function on Xh is constant. In particular a
bounded open set W ⊂ Cn is never isomorphic to the complex analytic
space Xh associated to a complex algebraic variety X.

7 Projective Varieties
Projective space Pn plays a rôle similar to that of affine space k n as an
ambient space for interesting algebraic varieties:

Definition 7.1. A projective variety is an algebraic variety isomorphic to a

closed subspace X ,→ Pn .

56
Remark 7.2. There are two ways to relate affine varieties to projective
varieties:
1. To every embedded affine variety

Y ,→ k n ∼
= U 0 ⊂ Pn
we can associate its projective closure

Y ,→ Pn = U0 ∪ P(0 × k n ).

2. The affine cone over a projective variety X ,→ Pn is the affine variety

C(X) := π −1 (X) = π −1 (X) ∪ {0} ⊂ k n+1 ,

where π : (k n+1 )∗ −→ Pn denotes the quotient morphism. A warning:

The affine cone C(X) does not only depend on X as algebraic variety,
but as well on the chosen embedding X ,→ Pn !
3. The above two cases are related as follows

C(Y ) = k ∗ · ({1} × Y ),

where k ∗ · A := {λx; λ ∈ k ∗ , x ∈ A} for a subset A ⊂ k n+1 .

The abstract definition of an affine cone is as follows:

Definition 7.3. An affine cone Z ⊂ k n (or for short an affine n-cone) is a
nonempty Zariski closed subset, such that λZ ⊂ Z for all λ ∈ k. It is called
nontrivial, if it contains points 6= 0.

Remark 7.4. 1. A nontrivial affine cone is a (finite or infinite) union of

lines through the origin.
2. There is a bijection between the non-empty closed subsets X ,→ Pn
and the nontrivial affine (n+1)-cones:

X 7→ C(X)

with the inverse

k n+1 ⊃ Z 7→ π(Z ∗ ) ⊂ Pn
with Z ∗ := Z \ {0}.

57
 The counterparts to affine n-cones on the level of ideals are the homoge-
neous ideals:

Definition 7.5. Denote k[T ]q := |α|=q k · T α the vector subspace of k[T ]

L

of all homogeneous polynomials of degree q ∈ N. An ideal a ,→ k[T ] is called

homogeneous if one of the two following equivalent conditions is satisfied:

1. The ideal a is generated by homogeneous polynomials.

2. The ideal a is the direct sum

∞
M
a= aq
q=0

of its homogeneous subspaces aq := a ∩ k[T ]q .

The second condition means that with a polynomial g = q g (q) ∈ a also

P

its homogeneous terms g (q) belong to a. Hence we may replace a system

(q)
f1 , ..., fr of generators with the fi , i = 1, ..., r, q ∈ N. On the other hand, if
a = (f1 , ..., fr ) with polynomials fi ∈ k[T ]qi , then any g = h1 f1 +...+hr fr ∈ a
satisfies
r
X
g (q) = (hi )(q−qi ) fi ∈ aq .
i=1

Proposition 7.6. 1. The zero set N (a) ,→ k n of a homogeneous ideal is

an affine n-cone.

2. An algebraic set Z ,→ k n is an affine cone iff its vanishing ideal I(Z)

is a homogeneous ideal.

Proof. The first part as well as the implication ”⇐=” of the second part
follow from the fact that a can be generated by homogeneous
P (q) polynomials.
Finally for an affine cone Z the assumption f = qf ∈ I(Z) implies
P q (q)
fλ = qλ f ∈ I(Z) for all λ ∈ k, where fλ (x) := f (λx). Evaluating
that equality at any point x ∈ Z gives f (q) (x) = 0 for all q ∈ N resp.
f (q) ∈ I(Z).

58
Corollary 7.7. Any projective variety X ,→ Pn can be described as the set
of zeros

X = N (Pn ; f1 , ..., fr ) := {[t]; f1 (t) = ... = fr (t) = 0},

of finitely many homogeneous polynomials f1 , ..., fr ∈ k[T0 , ..., Tn ].

Proof. Apply Prop. 7.6 to the affine cone C(X) ,→ k n+1 .

Note that homogeneous polynomials are functions only on k n+1 and not
on Pn , but since such a polynomial either vanishes identically on a punctured
line k ∗ · t (t 6= 0) or has no zeros there, the above description of a projective
variety makes sense nevertheless.

Remark 7.8. If Y = N (a) ,→ k n ∼ = U0 , then its projective closure can be

described as follows:
For a polynomial f ∈ k[T1 , ..., Tn ]\{0} its homogeneization fb ∈ k[T0 , ..., Tn ]
is the polynomial
deg(f )
fb := T0 f (T0−1 T1 , ..., T0−1 Tn ) ∈ k[T0 , ..., Tn ].

Then the ideal

X
a :=
b k[T0 , ..., Tn ] · fb ,→ k[T0 , ..., Tn ]
f ∈a

satisfies
a) = C(Y ).
N (b

Proof. Since Pn carries the quotient topology w.r.t. π : k n+1 \ {0} −→ Pn ,

we have
C(Y ) = k ∗ · ({1} × Y ).
Consider the projection
 
∗ n n t1 tn
k × k −→ k , (t0 , ..., tn ) 7→ , ..., .
t0 t0

The functions f ( tt10 , ..., ttn0 ), f ∈ a, then have

k ∗ · ({1} × Y )

59
as their common set of zeros in k ∗ × k n , and the same holds for the fˆ, f ∈ a.
Since fˆ ∈ k[T0 , ..., Tn ], it follows

k ∗ · ({1} × Y ) ⊂ N (b
a).

For the reverse inclusion we show

I(k ∗ · ({1} × Y )) ⊂ I(N (b

a))

Take any g ∈ k[T0 , ..., Tn ] vanishing on k ∗ · ({1} × Y ). Since both ideals are
homogeneous, we may assume, that g is homogeneous (of degree q) and not
divisible by T0 - a factor T0 does not contribute to the zeros on k ∗ × k n . Then
g1 := g(1, T1 , ..., Tn ) vanishes on Y , so (g1 )` ∈ a for some ` > 0. But then
g = gb1 resp. g ∈ I(N (b a)).

Example 7.9. If a = (f ) is a principal ideal, then b a = (fb), because of

c = fb · gb of the homogeneization. But it is not linear,
the multiplicativity gf
and thus a generator system f1 , ..., fr for a does not necessarily provide a
generator system fb1 , ..., fbr for b
a.

Given an algebraic variety Y , an ”extension” (no standard terminology!)

of Y is an algebraic variety X together with an isomorphism Y ∼ = U ⊂ X,
where U ⊂ X is a dense open subset. For an affine variety

Y ,→ k n ∼
= U0 ⊂ Pn .

such an extension is given by its projective closure:

X := Y ,→ Pn = U0 ∪ P(0 × k n ).

Indeed the projective closure yields a maximal extension, since, as we shall see
in a moment, projective varieties do not admit any further proper extensions:
They are ”complete”.

Definition 7.10. An algebraic variety X is called complete if it satisfies the

following property: For any algebraic variety Z the projection

60
prZ : X × Z −→ Z is a closed map, i.e. maps closed sets to closed sets.
Remark 7.11. 1. In order to check completeness of an algebraic variety
X, it obviously suffices to considers affine ”test varieties” Z.
2. A closed subvariety Y ,→ X of a complete variety is complete, the
induced map Y × Z −→ X × Z being a closed embedding as well.
3. The image ϕ(X) ⊂ Y of a morphism ϕ : X −→ Y with a complete
source X is closed in Y and complete: The graph Γϕ ⊂ X × Y of any
morphism ϕ : X −→ Y is the inverse image

Γϕ = (idX × ϕ)−1 (∆)

of the diagonal ∆ ,→ Y × Y with respect to the morphism idX × ϕ :

X × Y −→ Y × Y , hence closed, and ϕ(X) = prY (Γϕ ). Finally, if
C ,→ ϕ(X) × Z, consider the inverse image (ϕ × idZ )−1 (C) ,→ X × Z
and use the completeness of X.
4. A complete algebraic variety X has no nontrivial extensions: The image
∼
=
U of an extension homomorphism ϕ : X −→ U ⊂ Y is closed and dense,
hence U = Y .
5. A regular function on a connected complete variety X is constant,
i.e. O(X) = k: A regular function can be regarded as a morphism
f : X −→ P1 = k ∪ {∞} avoiding the value ∞. Thus the closed set
f (X) is finite.
6. An affine variety X is complete iff it is finite.
7. A complex algebraic variety X is complete iff Xh is compact.

Theorem 7.12. Projective varieties are complete.

Proof. According to the above Remark 7.11.2 it suffices to show that Pn is
complete. So let Z := Sp(C) be a w.l.o.g. affine variety and Y ,→ Pn × Z a
closed set. With B := prZ (Y ) we have the commutative diagram

Y ,→ Pn × Z
↓ ↓
?
B ,→ Z

61
and want to see, why we are allowed to remove the question mark. For z ∈ Z
we consider the sectional variety

Yz := {y ∈ Pn ; (y, z) ∈ Y } ,

such that
B = {z ∈ Z; Yz 6= ∅}.
Denote
C(Y ) ,→ k n+1 × Z
the closure of
(π × idZ )−1 (Y ) ,→ (k n+1 \ {0}) × Z,
where π × idZ : (k n+1 \ {0}) × Z −→ Pn × Z. Indeed
[
(π × idZ )−1 (Y ) = (C(Yz )∗ × {z})
z∈B

and !
[
C(Y ) = C(Yz ) × {z} ∪ ({0} × B).
z∈B

We want to show B = B, i.e. the second term is not really needed. The
below lemma assures that V = Z \ B is open:
Lemma 7.13. Let Z = Sp(C) be an affine variety and X ,→ k n+1 × Z be a
relative cone, i.e.

Xz := {t ∈ k n+1 ; (t, z) × X} ,→ k n+1

is a (possibly empty) affine cone. Then the set

V := {z ∈ Z; Xz ⊂ {0}}

is open.
Proof. We consider the vanishing ideal

a := I(X) ,→ C[T ] := C[T0 , ..., Tn ],

a graded ideal
a = (f1 , ..., fr )

62
generated by homogeneous polynomials fi ∈ C[T ] of degree di as well as its
”sectional ideals”
az := {f (z, T ); f ∈ a} = (f1 (z, T ), ..., fr (z, T )) ,→ k[T ], z ∈ Z.
Then
z ∈ V ⇐⇒ N (az ) ⊂ {0}
and furthermore
N (az ) ⊂ {0} ⇐⇒ az ⊃ m` for some ` ∈ N
with the ideal
m := (T0 , ..., Tn ) ,→ k[T0 , ..., Tn ].
The implication ”⇐=” is clear, while for the opposite direction we first note
that Hilberts Nullenstellensatz provides an exponent s ∈ N with Tis ∈ az
for i = 0, ..., n. Now take ` := (n + 1)s: The ideal m` is generated by the
monomials T α with |α| = `, and any such monomial divisible by some Tis .
As a consequence we obtain that
[∞
V = V`
`=1

is the ascending union of the sets

V` := z ∈ Z; az ⊃ m` .


Thus we are done if we can show that V` ⊂ Z is open for any ` ∈ N. In order
to see that we consider the linear maps
M r
Fz : E := k[T ]`−di −→ L := k[T ]`
i=1

(g1 , ..., gr ) 7→ g1 f1 (z, T ) + ... + gr fr (z, T ).

Since the homogeneous ideal m` is generated by k[T ]` , the inclusion az ⊃ m`
is equivalent to the surjectivity of the linear map Fz : E −→ L:
V` = {z ∈ Z; Fz : E −→ L is onto }.
Finally, the map Z −→ Hom(E, L), z 7→ Fz , is continuous (Hom(E, L) is a
k-vector space and the subset consisting of the surjective maps open, namely
using matrix representations, described by the nonvanishing of at least one
of the minors of Fz of size dim L). Thus V` ⊂ Z is open, as desired.

63
8 Dimension
A first basic invariant of an algebraic variety X is its dimension:

Definition 8.1. The dimension of an algebraic variety X at a point a ∈ X

is defined as

dima X := max{n ∈ N; ∃ X0 = {a} $ X1 $ ... $ Xn , Xi ,→ X irreducible},

while
dim X := max dima X.
a∈X

Remark 8.2. 1. Since a maximal strictly increasing chain of irreducible

subvarieties starts with a point, we have

dim X = max{n ∈ N; ∃ X0 $ X1 $ ... $ Xn , Xi ,→ X irreducible}.

2. If a ∈ U ⊂ X with an open subset U ⊂ X, then dima U = dima X.

This follows from the fact that Y 7→ Z := Y defines a bijection between
the irreducible subvarieties Y ,→ U containing a and the irreducible
subvarieties Z ,→ X containing a, the inverse map being given by
Z 7→ Z ∩ U .

3. dim X = 0 iff |X| < ∞.

4. If X1 , ..., Xr ,→ X are the irreducible components of X containing a,

then
dima X = max dima Xi .
i=1,...,r

5. If X1 , ..., Xr ,→ X are the irreducible components of X, then

dim X = max dim Xi .

i=1,...,r

We call X purely n-dimensional, if n = dim Xi for all irreducible com-

ponents Xi ,→ X.

6. A pure 1-dimensional algebraic variety is called a curve.

7. A pure 2-dimensional algebraic variety is called a surface.

64
 8. An irreducible variety is a curve iff the proper closed subsets are the
finite sets.
9. dim k = 1 and dima k n ≥ n for all a ∈ k n : Take Xi := a + (k i × 0) with
the origin 0 ∈ k n−i .

In order to determine the dimensions of varieties the following comparison

result is a standard tool:

Proposition 8.3. Let ϕ : X −→ Y be a finite surjective morphism of affine

varieties, b ∈ Y . Then we have

dimb Y = max
−1
dima X,
a∈ϕ (b)

and in particular
dim X = dim Y.
Proof. First note that for any dominant morphism ϕ : X −→ Y and closed
Z ,→ X we have
I(ϕ(Z))) = I(Z) ∩ O(Y ),
where we treat ϕ∗ : O(Y ) −→ O(X) as inclusion.
”dimb Y ≥ maxa∈ϕ−1 (b) dima X”: Consider a strictly increasing sequence of
irreducible subvarieties X0 = {a} $ X1 $ ... $ Xn on X. Then the (closed)
images Yi := ϕ(Xi ) are as well irreducible closed subvarieties of Y and Yi $
Yi+1 : Otherwise we would have

I(Xi ) ∩ O(Y ) = I(Yi ) = I(Yi+1 ) = I(Xi+1 ) ∩ O(Y ),

a contradiction to Th.3.14.3, since O(Y ) ⊂ O(X) is an integral extension.

”dimb Y ≤ maxa∈ϕ−1 (b) dima X”: Given any irreducible closed subvariety
Z ,→ Y , there is an irreducible component of ϕ−1 (Z) projecting onto Z, i.e.
the restriction of ϕ to that component has image Z. Now consider a strictly
increasing sequence of irreducible subvarieties Y0 = {b} $ Y1 $ ... $ Yn
on Y . By downward induction we construct a strictly increasing sequence
X0 $ X1 $ ... $ Xn on X lying above the Yi . Take Xn as an irre-
ducible component of ϕ−1 (Yn ) projecting onto Yn . If Xi is found, take
Xi−1 ,→ ϕ−1 (Yi ) ∩ Xi as an irreducible component projecting onto Yi−1 .
In particular X0 = {a} with some a ∈ ϕ−1 (b).
Corollary 8.4. We have dim k n = n = dima k n , where a ∈ k n , and

65
 dim X < ∞ for any affine variety X.
Proof. The second part of the statement follows from Prop. 8.3 together with
the Noether normalization theorem. The first part is obtained by induction,
the case n = 1 being obvious. Assume now X0 $ X1 $ ... $ Xm = k n is a
strictly increasing sequence of irreducible subvarieties in k n , n > 1. If we can
show dim Xm−1 < n, we have necessarily m − 1 ≤ n − 1 resp. m ≤ n: Since
Xm−1 6= k n , there is a polynomial f ∈ I(Xm−1 ) \ {0}. As a consequence of
the lemmata 3.12 and 8.3, we see that
dim Xm−1 ≤ dim N (k n ; f ) = n − 1.

Remark 8.5. 1. If U ⊂ X is a non-empty open subvariety of the S irre-

ducible variety X, then we have dim U = dim X. Write X = ri=1 Vi
with affine open sets Vi ⊂ X. Since dim X = maxi=1,...,r dim Vi and
dim U = maxi=1,...,r dim U ∩ Vi , we may assume that X is affine. For
X = k n the claim is obvious because of dima k n = n for all a ∈ k n . For
arbitrary affine X consider again a finite surjective map ϕ : X −→ k d .
Take V := k d \ϕ(X\U ) - a non-empty open subset, since dim ϕ(X\U ) =
dim(X \ U ) < dim X = n. But ϕ−1 (V ) −→ V being finite and sur-
jective, the inclusion U ⊃ ϕ−1 (V ) implies dim U ≥ dim ϕ−1 (V ) =
dim V = n.
2. Let E ⊃ K be a field extension. The transcendence degree trdegrK E ∈
N ∪ {∞} is defined as the maximal number n of over K algebraically
independent elements x1 , ..., xn ∈ E, i.e. such that there is an injective
K-algebra homomorphism K[T1 , ..., Tn ] ,→ E, and if such a maximal
number does not exist, then one sets trdegrK E = ∞. As a consequence
of the geometric Noether normalization theorem 4.15 we see that for an
irreducible affine variety X = Sp(A), we have dim X = trdegrk Q(A).

A minimal irreducible subvariety of a variety X is a point. But what

can we say about maximal proper irreducible subvarieties Z ,→ X of an
irreducible variety X, i.e. such that dim Z = dim X − 1? If X is affine,
there is a function f ∈ O(X) \ {0} vanishing on Z. Hence Z ,→ N (f ) is
an irreducible component of the ”hypersurface” N (f ) ,→ X - but in general
one can’t expect equality! On the other hand all the irreducible components
Z ,→ N (f ) of a given hypersurtface N (f ) have indeed codimension 1:

66
Theorem 8.6 (Krulls Hauptidealsatz). Let X be an n-dimensional irre-
ducible algebraic variety. Then the zero locus N (f ) ,→ X of a regular func-
tion f ∈ O(X) \ {0} is empty or purely (n−1)-dimensional, i.e. all its
irreducible components Z ,→ N (f ) satisfy dim Z = n − 1.

Proof. By Rem. 8.5.1 we may assume that X = Sp(A) is affine as well

as N (f ) ,→ X irreducible: Given an irreducible component Z ,→ N (f ) of
N (f ), replace X with an open affine subset U ⊂ X with U ∩ N (f ) = U ∩ Z.
For X = k n we may assume f ∈ k[T ] to be Tn -monic, hence the restricted
projection
N (f ) ,→ k n −→ k n−1
onto the first n − 1 coordinates is finite and surjective and we may refer to
Prop. 8.3.
In order to obtain the result in the general case we choose a finite sur-
jective morphism ϕ : X −→ k n and look for a polynomial g ∈ k[T ] with
ϕ(N (f )) = N (g): Then we may apply Proposition 8.3 to the morphism

ϕ|N (f ) : N (f ) −→ N (g).

We take g := N(f ) with the norm function

N : Q(A) −→ k(T ) := Q(k[T ])

of the field extension Q(A) ⊃ k(T ), see the below remark with B = K[T ] for
a short presentation. Since f |ϕ∗ (g) we have

N (f ) ⊂ ϕ−1 (N (g)) ,→ X

and thus ϕ(N (f )) ⊂ N (g). We consider the diagram

X ⊃ N (f ) ∪ ϕ−1 (V )
↓ ↓ ↓
k n ⊃ N (g) ∪ V

with a principal open subset V = khn ⊂ k n . Since ϕ(N (f )) ,→ k n is closed, it

suffices to show that the lower union is disjoint if the upper one is. But in
that case we have f ∈ O(ϕ−1 (V ))∗ and thus g = N(f ) ∈ O(V )∗ .

67
The norm function associated to a field extension: We consider a
finite field extension E ⊃ K fitting in a diagram

Q(A) = E ⊃ K = Q(B)
∪ ∪ ∪
A ⊃ B = Bb

with rings A, B and the integral closure B

b of B in Q(B). We define the norm
function N = N[E:K] : E −→ K of the extension E ⊃ K by

N(a) := det(µa : E −→ E),

where µa ∈ EndK (E) is the multiplication with a, i.e. µa (x) := ax. Then we
have

1. N(1) = 1,

2. N(ac) = N(a) · N(c),

3. N(A) ⊂ B,

4. N(A∗ ) ⊂ B ∗ , and

5. a|N(a) in A.

The first two statements are obvious, we comment on 3) and 5), while 4)
follows from the multiplicativity 2) and statement 3). We use:

Lemma 8.7. Denote pa ∈ K[S] the minimal polynomial of an element a ∈ E.

Then, for a ∈ A we have:
pa ∈ B[S]
and the characteristic polynomial χa ∈ K[S] of µa ∈ EndK (E) satisfies

χa = (pa )r ∈ B[S]

with r := [E : K[a]].

Since (−1)n N(a) = χa (0) (with n := [E : K]) is the constant term of the
characteristic polynomial χa and χa (a) = 0, that implies both 3) and 5).

68
Proof. The minimal polynomial pa ∈ K[T ] of an element a ∈ A ⊂ E lies in
B[S]: We may assume - by enlarging E if necessary - that pa splits into linear
factors over E. Then pa |f , if f ∈ B[S] is an integral equation for a ∈ A (i.e.,
a monic polynomial in B[S] with a as zero). As a consequence, all zeros of pa
are integral over R, hence its coefficients - being obtained from these zeros
by additions and multiplications - are both in K = Q(B) and integral over
B, but B is integrally closed in Q(B), so they are already in B.
Now let us consider the characteristic polynomial χa ∈ K[S] of µa ∈
EndK (E). The minimal polynomial pa of a is also the minimal polynomial
of µa , a divisor of the characteristic polynomial of µa . If E = K[a], both
polynomials are monic of the same degree, so they coincide. In the gen-
eral case, choosing a basis x1 , ...,
Lxr of E as K[a]-vector space provides a
µa -invariant decomposition E = di=1 K[a]xi , and the characteristic polyno-
mial of µa |K[a]xi is pa . So, altogether we obtain (pa )r as the characteristic
polynomial of µa .

Finally, in order to see that we may apply the above reasoning with
B = K[T ] we add

Proposition 8.8. A UFD B is integrally closed in its field of fractions.

Proof. Assume that a := p/q ∈ Q(B) with relatively prime p, q ∈ B has an

integral equation
Xn−1
n
a + ri ai = 0.
i=0

Multiplication with q n yields

n−1
X
n
p + ri pi q n−i = 0,
i=0

in particular q|pn resp. q|p. But that implies q ∈ B ∗ resp. a ∈ B.

Proposition 8.9. For any point a ∈ X of an irreducible affine variety X we

have dima X = dim X.

69
Proof. We may assume that X is affine, since for a ∈ U ⊂ X with an open
subspace U ⊂ X we have both dim U = dim X and dima U = dima X, and
use induction on dim X. The case dim X = 0 is trivial, since then X is
finite. Now assume dim X > 0. Take a regular function f ∈ ma \ {0} and let
Z ,→ N (f ) be an irreducible component of N (f ) ,→ X containing a. Then,
by induction hypothesis, dima Z = dim Z, and thus
dim X ≥ dima X > dima Z = dim Z = dim X − 1,
so necessarily dima X = dim X.
Corollary 8.10. 1. Let X := N (k n ; f1 , ..., fr ) ,→ k n be a subvariety de-
termined by r equations. Then either X is empty or every irreducible
component X0 ,→ X of X has dimension dim X0 ≥ n − r.
2. Let X := N (Pn ; f1 , ..., fr ) ,→ Pn be a projective variety determined by
r nonconstant homogeneous polynomials. If r ≤ n, then X is non-
empty and every irreducible component X0 ,→ X of X has dimension
dim X0 ≥ n − r.
Proof. 1.) The first part is proved by induction on the number r of equa-
tions, the case r = 0 being trivial. Now for r > 0 every irreducible com-
ponent X0 ,→ X is contained in an irreducible component Z0 of Z :=
N (k n ; f1 , ..., fr−1 ). By induction hypothesis we know that dim Z0 ≥ n−r +1,
while X0 is an irreducible component of N (Z0 ; fr ), so, according to 8.6, has
dimension at least n − r.
2.) With π : (k n+1 )∗ −→ Pn denoting the quotient morphism we have
X = π(N (k n+1 ; f1 , ..., fr ) \ {0}) 6= ∅,
since fi (0) = 0 for i = 1, ..., r and, according to the first part, all the irre-
ducible components of N (k n+1 ; f1 , ..., fr ) have dimension ≥ n + 1 − r ≥ 1.
Finally use that dim C(X) = dim X + 1, since
C(X)Ti ∼
= Xi × k ∗
holds for Xi := X ∩ Ui (with Ui := Pn \ N (Pn ; Ti )).

Let us now study the dimensions of fibres of a dominant morphism ϕ :

X −→ Y between algebraic varieties X and Y , where we use the convention
dim ∅ := −∞.

70
Simultaneously we investigate the sets ϕ(X) obtained as images of mor-
phisms. Here is an easy example showing some possible phenomena:

Example 8.11. The morphism ϕ : k 2 −→ k 2 , (s1 , s2 ) 7→ (s1 , s1 s2 ) maps a

line s2 = a parallel to the s1 -axis to a the line t2 = at1 through the origin.
In particular it satisfies

 {(t1 , t−1
 ∗
1 t2 )} , if (t1 , t2 ) ∈ k × k
−1
ϕ (t1 , t2 ) = {0} × k , if (t1 , t2 ) = (0, 0)
∅ , otherwise


and thus

 0 , if (t1 , t2 ) ∈ k ∗ × k
dim ϕ−1 (t1 , t2 ) = 1 , if (t1 , t2 ) = (0, 0) .
−∞ , otherwise


Furthermore for V := k ∗ × k, we get an isomorphism

∼
=
ϕ|ϕ−1 (V ) : ϕ−1 (V ) −→ V.

Note furthermore that ϕ(k 2 ) = V ∪ {(0, 0)} is neither open nor closed in
k 2 , not even locally closed and thus does not inherit in a natural way the
structure of an algebraic variety.

Proposition 8.12. Let ϕ : X −→ Y be a dominant morphism between the

irreducible varieties X and Y . Then

1. dim X ≥ dim Y ,

2. the irreducible components of a nonempty fibre ϕ−1 (y) have at least

dimension d := dim X − dim Y , and

3. there is a nonempty open subset W ⊂ Y , such that all fibers ϕ−1 (y) of
points y ∈ W are pure-d-dimensional.

Proof. 2.) We may assume that Y is affine and take a surjective finite map
π : Y −→ k r (so r = dim Y ). Look at ψ := π ◦ ϕ : X −→ k r . Since
w.l.o.g. ψ(y) = 0, we see with the first part of Cor.8.10, that all irreducible

71
components of ψ −1 (0) ,→ X have dimension at least dim X − r. On the other
hand, if π −1 (0) = {y1 = y, y2 , ..., ys }, then
s
[
ψ −1 (0) = ϕ−1 (yj ),
j=1

a disjoint union, so the fiber ϕ−1 (y) is a union of irreducible components of

ψ −1 (0), and we are done.
In order to prove the remaining points we need the following auxiliary
lemma:

Lemma 8.13. Let ϕ : X = Sp(A) −→ Y = Sp(B) be a dominant morphism

of irreducible affine varieties. Then, with d := dim X − dim Y , there is a
non-empty open subset W ⊂ Y and a factorization
ψ pr
ϕ|ϕ−1 (W ) = prW ◦ ψ : ϕ−1 (W ) −→ W × k d −→
W
W

with the projection prW : W × k d −→ W and a surjective finite morphism

ψ : ϕ−1 (W ) −→ W × k d . In particular the fibers ϕ−1 (y), y ∈ W , are pure
d-dimensional.
Proof. We have already seen that an irreducible component Z ,→ ϕ−1 (y) has
dimension at least d; on the other hand the morphism

Z ,→ ϕ−1 (y) −→ {y} × k d

is finite, whence dim Z ≤ d.

The pull back ϕ∗ : B −→ A being injective, we treat it as an injection
and create the following diagram:

B[a1 , ..., ar ] = A ⊂ (B \ {0})−1 A =: C = K[a1 , ..., ar ] ⊃ Ag

∪ ∪ ∪ ∪ ∪
B ⊂ Q(B) =: K ⊂ K[T ] ⊃ Bg [T ] .
∪
Bg

The algebra C is an affine K-algebra. So we may apply the Noether normal-

ization theorem 3.10 and obtain a finite injective homomorphism

K[T ] := K[T1 , ..., Td ] ,→ C,

72
where we even may assume Tj ∈ A for j = 1, ..., d. If the polynomials
qi ∈ (K[T ])[S], i = 1, ..., r, are integral equations over K[T ] for the (B-
algebra) generators ai ∈ A, we have qi ∈ (Bg [T ])[S] for a suitable
g ∈ B \ {0}: Take g as a common denominator P of the coefficients of the
polynomials fij ∈ K[T ], such that qi = S mi + j<mi fij S j . Then obviously,
Ag is a finite Bg [T ]-module, and we may take W = Yg , with the above
factorization corresponding to the algebra homomorphisms

Bg ,→ Bg [T ] ,→ Ag .

Since dim X = dim ϕ−1 (W ) = dim W × k d = dim W + d = dim Y + d, we

obtain d = dim X − dim Y .

We go on with the proof of 8.12:

1.) Take any affine open subset V ⊂ Y and U ⊂ ϕ−1 (V ) and apply the
lemma 8.13 to ϕ|U : U −→ V , finally remember that dim Y = dim V as well
as dim X = dim U .
3.) We may again assume that Y is affine and write X = ri=1 Ui with
S
affine open subsets Ui . Then we may find open subsets Wi ⊂ T Y such that
ϕ−1 (y) ∩ Ui is pure-d-dimensional for all y ∈ Wi . Now set W := ri=1 Wi .

It is natural to ask how dim ϕ(y) depends on y ∈ Y . We may for example

consider the sets
Tm (ϕ) := {y ∈ Y ; dim ϕ(y) ≥ m}
for m ∈ N ∪ {−∞}. Then one could hope that the Tm (ϕ) ,→ Y are closed
subsets, but T0 (ϕ) = ϕ(X) need not be closed. Instead we try it with the
sets
Sm (ϕ) := {x ∈ X; dimx ϕ−1 (ϕ(x)) ≥ m}.
In any case S0 (X) = X is closed, and

Tm (ϕ) = ϕ(Sm (ϕ)).

Since the definition of Sm (ϕ) is local in X, we succeed showing

Proposition 8.14. For any morphism ϕ : X −→ Y between algebraic vari-

eties X and Y the sets

Sm (ϕ) := {x ∈ X; dimx ϕ−1 (ϕ(x)) ≥ m}.

73
are closed:
Sm (ϕ) ,→ X.
Equivalently: The function X −→ N, x 7→ dimx ϕ−1 (ϕ(x)) is upper semicon-
tinuous.

Proof.SWe do induction on dim X, the case dim X = 0 being trivial. If

X = ri=1 Xi is the decomposition of X into irreducible components, then
r
[
Sm (ϕ) = Sm (ϕ|Xi ).
i=1

Hence we may assume that X is irreducible, and, obviously, that ϕ is domi-

nant and thus Y irreducible as well. According to Prop.8.12 we have Sm (ϕ) =
X for m ≤ d := dim X − dim Y . So let m > d. But 8.12 assures that there
is a nonempty open set W ⊂ Y , where all fibres ϕ−1 (y), y ∈ W , are pure-d-
dimensional. Consequently Sm (ϕ) ⊂ Z := X \ ϕ−1 (W ) with dim Z < dim X.
Since Sm (ϕ) = Sm (ϕ|Z ), we may use the induction hypothesis.
Finally we use Lemma 8.13 in order to characterize image sets ϕ(X) of
morphisms ϕ : X −→ Y between algebraic varieties X and Y . First of all we
call a subset Y ⊂ X of a topological space locally closed if it is of the form
Y = Z ∩ U with a closed subset Z ,→ X and an open subset U ⊂ X, i.e. Y
is closed in a suitable open neighborhood U ⊂ X or, the other way around,
Y is open in its closure Y .

Definition 8.15. A subset of an algebraic variety X is called constructible

if it is a finite union of locally closed subsets.

Lemma 8.16. The constructible sets of an algebraic variety X form an al-

gebra of sets C(X), i.e. C(X) ⊂ P(X) is closed with respect to finite unions
and intersections and with respect to taking the complement.

Proof. By definition C(X) is closed with respect to finite unions, and if we

can show A ∈ C(X) =⇒ X \ A ∈ C(X), the de Morgan rules give S that C(X)
is closed with respect to finite intersections as well. So let A = ri=1 Ai ∩Ui ∈
C(X) with closed Ai ,→ X and open Ui ⊂ X. Then
r
! r
[ \
X \A=X \ Ai ∩ Ui = (X \ Ai ∩ Ui )
i=1 i=1

74
 r
\
= (X \ Ai ) ∪ (X \ Ui )
i=1

!
[ \ \
= (X \ Ai ) ∩ (X \ Ui ) ∈ C(X).
I∈P({1,...,r}) i∈I i6∈I

Example 8.17. 1. C(X) ⊂ P(X) is the smallest algebra of sets contain-

ing all open subsets.

2. On any set X the sets which either are finite or cofinite (i.e. have
a finite complement) form an algebra of sets. From this and the
previous point it follows that for an irreducible curve constructible
sets are either finite or cofinite resp. either closed or open. If X
is a reducible
Sr curve, any constructible set is still locally closed: If
X = i=1 Xi is the decomposition into irreducible components, then
A ∈ C(X) ⇐⇒ Ai := A ∩ Xi ∈ C(Xi ) ∀i = 1, ..., r, so A is constructible
iff every Ai ⊂ Xi either is closed or open in Xi . Then it follows that
A is open in its closure A, which is the union of certain irreducible
components of X and finitely many isolated points.

3. The set A = (k ∗ × k) ∪ {0} ⊂ k 2 is constructible, but not locally closed,

since it is not open in its closure k 2 .

Proposition 8.18. The image ϕ(X) ⊂ Y of a morphism ϕ : X −→ Y

between algebraic varieties X and Y is a constructible set:

ϕ(X) ∈ C(Y ).

Proof. We use induction on dim X, the case dim X = 0, i.e. |X| < ∞,
being obvious. Furthermore we may assume X to be irreducible and ϕ to be
dominant and hence Y is irreducible as well. Then, according to Prop. 8.12
there is a non-empty open subset W ⊂ ϕ(X), so ϕ(X) = W ∪ ϕ(Z) with
Z := Y \ ϕ−1 (W ). Since then dim Z < dim X the induction hypothesis gives
that ϕ(Z) is constructible, hence so is W ∪ ϕ(Z).

75
9 Rational functions and local rings
On a complete algebraic variety X there are only locally constant regular
functions; so they do not provide any information about the geometry of X.
Instead we have to allow ”poles” and ”points of indeterminacy”: We consider
pairs (U, f ), where U ⊂ X is a dense open subset and f ∈ O(U ) a regular
function on U . Two such pairs are identified using the equivalence relation

(U, f ) ∼ (V, g) ⇐⇒ f |U ∩V = g|U ∩V .

In order to see that this really provides an equivalence relation, note that a
finite intersection of dense open subsets is again a dense open subset.
Definition 9.1. A rational function on an algebraic variety X is an equiv-
alence class of pairs (U, f ) with a dense open subset U ⊂ X and f ∈ O(U ).
We denote R(X) the set of all rational functions on X.
Remark 9.2. 1. Since the restriction O(U ) −→ O(W ) of functions is
injective for dense open subsets W ⊂ U , a rational function represented
by (U, f ) has a unique representative (Umax , fmax ) with a maximal open
subset Umax ⊂ X: Set
[
Umax := V , fmax |V := g.
(V,g)∼(U,f )

In order to simplify notation, we denote rational functions f and Df :=

Umax their domain of definition.

2. The rational functions form a k-algebra, the sum f +g resp. the product
f g having a domain of definition containing the (dense) intersection
Df ∩ Dg .

3. A morphism ϕ : X −→ Y , such that ϕ(X0 ) ,→ Y is an irreducible com-

ponent of Y for every irreducible component X0 ,→ X, induces a pull-
back homomorphism ϕ∗ : R(Y ) −→ R(X). In particular the inclusion
of an open subset U ⊂ X induces a a restriction map R(X) −→ R(U );
indeed it is an isomorphism R(X) ∼ = R(U ) if U ⊂ X is dense.
4. If X = Sp(A) is affine, then

R(X) ∼
= S −1 A

76
 with the localization of A with respect to the multiplicative set

S := {f ∈ A; f |X0 6∈ I(X0 ) ∀ irreducible components X0 ,→ X}

of all nonzero divisors in A. In particular

R(X) ∼
= Q(A),

if X is irreducible.

5. For an irreducible variety X the ring R(X) is a field

R(X) ∼
= Q(O(U ))

isomorphic to the field of fractions of the ring of regular functions on

any (non-empty) open affine subset U ⊂ X.

6. The morphism
f : Df −→ k
representing a nonzero rational function on an irreducible variety X
can be extended to a morphism

fb : U := Df ∪ D1/f −→ P1 = k ∪ {∞},

by defining
fb(x) := [(1/f )(x), 1] ∈ U1 ⊂ P1
for x ∈ D1/f . The points outside Df ∪ D1/f are called ”points of inde-
terminacy” of the rational function f . If X = Sp(A) is affine, a point
a ∈ X is a point of indeterminacy of f ∈ Q(A), iff g(a) = 0 = h(a) for
all representations of f as fraction f = g/h with g, h ∈ A.
For example take X := k 2 and f := T1 /T2 . Then Df = k × k ∗ and
D1/f = k ∗ × k, while the origin is the only point of indeterminacy.
Indeed the morphism fb : (k 2 )∗ −→ P1 turns out to be the quotient
morphism. In particular there is no extension to a morphism k 2 −→ P1 ,
since the restriction fb|k∗ t of fb to a punctured line k ∗ t can be extended
to the entire line kt only with 0 7→ [t], that value depending on the
chosen line.
In the sequel we shall use also the letter f in order to denote fb.

77
 7. If X = ri=1 Xi is the decomposition of X into irreducible components,
S
then the component covering
r
a
π:X
e := Xi −→ X
i=1

induces an isomorphism
r
R(X) ∼
M
= R(Xi ),
i=1

S e := π −1 (U ) ∼
since, with U := X \ i6=j Xi ∩ Xj and U = U , we have
∼ ∼ ∼
R(X) = R(U ) = R(U ) = R(X).
e e
In particular R(X) is a finite direct sum of fields.

Remembering the fact that affine varieties form a category anti-equivalent

to the category of affine reduced k-algebras, it is natural to ask whether the
category of k-algebras, which are finite sums of (as field) finitely generated
k-extensions, has a geometric counterpart as well. The point here is to find
the correct notion of a morphism. In analogy to rational functions there is
also the notion of a rational morphism:

Definition 9.3. A rational morphism between two algebraic varieties X, Y

is an equivalence class of pairs (U, ϕ) with a dense open subset U ⊂ X and
a morphism ϕ : U −→ Y , where the equivalence relation is

(U, ϕ) ∼ (V, ψ) ⇐⇒ ϕ|U ∩V = ψ|U ∩V .

As with rational functions every rational morphism ϕ : X −→ Y has a

a maximal domain of definition Dϕ , where it is an ”ordinary” morphism of
algebraic varieties. A rational morphism ϕ : X −→ Y between irreducible
varieties X, Y is called dominant if ϕ(U ) ⊂ Y is dense in Y for some (and
thus all) dense open subset(s) U ⊂ Dϕ .

Theorem 9.4. The functor R : X 7→ R(X) from the category of irre-

ducible algebraic varieties together with the dominant rational morphisms as
morphisms to the category of finitely generated field extensions of k is an
anti-equivalence of categories.

78
Proof. If K = k(a1 , ..., ar ), we have K = R(X) with X := Sp(k[a1 , ..., ar ]).
Assume now that σ : R(Y ) −→ R(X) is a k-algebra homomorphism. We
may replace X and Y with (nonempty) open affine subsets, i.e., we may
assume X = Sp(A), Y = Sp(B). If B = k[b1 , ..., bs ], choose a common
denominator g ∈ A for the elements σ(b1 ), ..., σ(bs ) ∈ Q(A). So σ(B) ⊂ Ag
and the morphism Xg −→ Y corresponding to the k-algebra homomorphism
σ|B : B −→ Ag defines the rational morphism X −→ Y we are looking for –
the easy proof of uniqueness is left to the reader.

Corollary 9.5. For irreducible algebraic varieties X, Y the following state-

ments are equivalent:

1. R(X) ∼
= R(Y )

2. There are nonempty open sets U ⊂ X and V ⊂ Y such that U ∼

= V as
algebraic varieties.

In that case we say also, that the varieties X and Y are birationally equiva-
lent.

Proof. ”⇐=”: R(X) ∼ = R(U ) ∼

= R(V ) ∼
= R(Y ).
”=⇒”: Let ϕ : U1 −→ Y and ψ : V1 −→ X be morphisms inducing the
above isomorphism in the respective directions. Take U := ϕ−1 (V1 ) ⊂ U1
and V := ϕ(U ) ⊂ V1 .

Definition 9.6. An n-dimensional irreducible algebraic variety X is called

rational, if it is birationally equivalent to k n , or equivalently, if it admits a
non-empty open subset U ⊂ X isomorphic to an open subset V ⊂ k n .

For the study of local properties of an algebraic variety resp. of rational

functions one needs even non-affine subrings of R(X), which are sufficiently
big in order to be understandable:

Definition 9.7. Let Y ,→ X be an irreducible subvariety and X0 ,→ X the

union of the irreducible components of X containing Y . The local ring OX,Y
of X at Y is defined as

OX,Y := {f ∈ R(X0 ); Y ∩ Df 6= 0}.

79
 Note that OX,Y is a subring, since any two nonempty open subsets of the
irreducible Y have nonempty intersection and thus

Df +g , Df g ⊃ Df ∩ Dg ←- (Df ∩ Y ) ∩ (Dg ∩ Y ) 6= ∅.

Furthermore:

Remark 9.8. 1. To justify its name, we remark that OX,Y is a local ring
with maximal ideal

mX,Y := {f ∈ R(X0 ); f |Y ∩Df = 0}.

2. If U ⊂ X is an open subset intersecting Y , we have OX,Y ∼

= OU,U ∩Y .

3. If X = Sp(A) with Y ⊂ X being contained in all irreducible compo-

nents of X, the ring OX,Y is isomorphic to the localization

OX,Y ∼
= S −1 A

of A with respect to the multiplicative subset S := A \ I(Y ).

4. The ring OX,Y is noetherian: If b ,→ OX,Y ∼ = S −1 A is an ideal, then

A being noetherian, we have a := b ∩ A = Ag1 + ... + Agr , and that
implies b = S −1 a = S −1 Ag1 + ... + S −1 Agr .

5. The ring OX,Y is never an affine algebra: The only local reduced affine
algebra is the base field k.

The extremal choices of Y are of particular importance: For minimal

Y 6= ∅, i.e. a one point set {a}, we obtain the local ring OX,a of X at the
point a:

Definition 9.9. The local ring OX,a of X at a ∈ X is defined as

OX,a := OX,{a} = {f ∈ R(X0 ); Df 3 a},

where X0 ,→ X is the union of all irreducible components of X containing a.

The ring OX,a is called the ”local ring of X at the point a ∈ X” or the ”stalk
of the structure sheaf OX at the point a ∈ X”.

80
Remark 9.10. 1. We can also describe OX,a as the direct limit
OX,a := lim OX (U ),
U 3a

taken with respect to the (by inclusion) partially ordered set of open
neighbourhoods of a. With other words the elements in OX.a are equiv-
alence classes of pairs (U, f ), where a ∈ U and f ∈ OX (U ) with
(U, f ) ∼ (V, g) iff f |W = g|W for some open neighbourhood W ⊂ U ∩ V
of the point a.
2. If k = C we can as well consider the stalk
Ha := lim H(U ),
U 3a

of the structure sheaf H of Xh , where now U ranges over all strongly

open neighbourhoods of a. We have
Ha ∼
= C{T1 , ..., Tn }
with the ring C{T1 , ..., Tn } of convergent power series on the RHS iff
a 6∈ S(Xh ). Furthermore we now can also define rigorously the notion
of a self intersection point: A point a ∈ X is a self intersection point
iff Ha is not an integral domain.
The local ring OX,a of an algebraic variety X at a point a ∈ X is an
important tool in the study of the local geometry of X near a; indeed there
is an analogue to Th. 9.4: We define the category of ”germs of algebraic
varieties ” to have as objects the pairs (X, a), where X is an algebraic variety
and a ∈ X a point. Furthermore a morphism from (X, a) to (Y, b) is an
equivalence class of morphisms ϕ : U −→ Y of algebraic varieties, where
U ⊂ X is an open neighborhood of a ∈ X and ϕ(a) = b, two morphisms
being equivalent if they agree on some open neighborhood of a. Obviously
such a morphism ϕ induces a pullback homomorphism ϕ∗a : OY,ϕ(a) −→ OX.a
depending only on the equivalence class of ϕ.
Theorem 9.11. The functor
(X, a) 7→ OX,a , [ϕ] 7→ ϕ∗a
is an anti-equivalence between the category of germs of algebraic varieties and
the full subcategory of the category of all k-algebras, whose objects are local
rings of algebraic varieties.

81
 The proof is analogous to that of Th. 9.4 and thus left as an exercise to
the reader. In particular we have as well the following corollary:
Corollary 9.12. For points a ∈ X, b ∈ Y in the algebraic varieties X, Y the
following statements are equivalent:
1. OX,a ∼= OY,b as k-algebras
2. There are open neighborhoods U of a ∈ X and V of b ∈ Y , such that
U∼= V with an isomorphism of algebraic varieties sending a to b.
Remark 9.13. 1. The local ring OX,a is an integral domain iff a is con-
tained in exactly one irreducible component of X. So all the local rings
of an algebraic variety X are integral domains iff X is the disjoint union
of its irreducible components.
2. For an irreducible variety we have OX,a ⊂ R(X) for all a ∈ X and
\
O(X) = OX,a .
a∈X

10 Normal Varieties
Definition 10.1. A point a ∈ X in an algebraic variety X is called a normal
point, iff the local ring OX,a of X at a is an integral domain, which is integrally
closed in its field of fractions Q(OX,a ) = R(X). An algebraic variety X is
called normal iff all its points are normal.
So a normal variety is the disjoint union of its irreducible components.
Proposition 10.2. 1. An irreducible affine variety is normal iff its alge-
bra of regular functions A is integrally closed in its field of fractions
Q(A).
2. The normal points of an algebraic variety X form a dense open subset.
In the proof we use the following basic facts:
Remark 10.3. 1. If A is normal, i.e. integrally closed in its field of frac-
tions, so is S −1 A for any multiplicative subset S ⊂ A: If f ∈ S −1 A
has integral equation p(f ) = 0 with a monic polynomial p ∈ S −1 A[T ],
then hf has integral equation hn p(h−1 T ), n = deg p. But for a suitable
choice of h ∈ S that polynomial has coefficients in A, and A being
normal, we get hf ∈ A resp. f = hf /h ∈ S −1 A.

82
 2. The integral closure A e ⊂ Q(A) of an affine k-algebra without zero
divisors in its field of fractions Q(A) is a finitely generated A-module,
thus in particular an affine algebra.

Proof. The first part follows from the fact, that a localization of a ring inte-
grally closed in its field of fractions enjoys the same property, as well as any
intersection – see Rem.9.13 – of such rings does.
For the second part consider a normal point a ∈ X. Since the irreducible
points of X, i.e. the points lying in only one of the irreducible components
of X, form a dense open subset, we may replace X with an irreducible affine
open neighborhood of a resp. assume that X = Sp(A) is irreducible and
affine. Then we have A e ⊂ OX,a for the integral closure A
e of A in its field of
fractions Q(A). But on the other hand it is a finitely generated A-module,
say Ae = Af1 + ... + Afr with f1 , ..., fr ∈ Q(A). Hence there is a common
denominator g ∈ A for f1 , ..., fr satisfying g(a) 6= 0 and thus Ag = A eg
is integrally closed in Q(Ag ) = Q(A), so the open neighborhood Xg of a
contains only normal points.

Remark 10.4. The affine variety X e := Sp(A)

e is called the normalization of
X, the morphism π : X e −→ X induced by the inclusion A ⊂ A e is finite and
surjective; it fits into a commutative diagram

Xe −→ X
∪ ∪ ,
∼
=
π −1 (U ) −→ U

where U ⊂ X is the open subset of all normal points of X.

Let us now pass to the case of a maximal proper irreducible subvariety

Y ,→ X of an irreducible normal variety X, i.e. Y has codimension 1 in X:
The ring OX,Y then is a ”discrete valuation ring”:

Definition 10.5. A discrete valuation ring R is a noetherian local integral

domain, whose maximal ideal m ,→ R is a principal ideal:

m = (π)

with an element π ∈ R \ {0}.

83
Remark 10.6. Every nonzero element x ∈ R in a discrete valuation ring R
can be written uniquely

(1) x = eπ n

with a unit e ∈ R∗ and n ∈ N. In particular R is a PID, the ideals being the

zero ideal and the ideals (π n ), n ∈ N. This is a consequence of the fact, that
an element in R is either a unit or divisible with π. We can write x = aν π ν
so long as possible; then the chain of ideals (aν ) is strictly increasing, whence
it follows that there is a final e := an , it is necessarily a unit. Furthermore
the field of fractions is a disjoint union
∞
[
Q(R) = Rπ = {0} ∪ R∗ π n .
n=−∞

Example 10.7. Let p ∈ R be a prime element of the UFD R. Then R(p) :=

{a/b ∈ Q(R); a, b ∈ R, gcd(p, b) = 1} is a discrete valuation ring.

Indeed there are more rings R, such that the localization Rp with respect
to a prime ideal p ⊂ R is a discrete valuation ring. Of course p has to be a
minimal nonzero prime ideal. Here we shall prove:

Theorem 10.8. For an irreducible normal variety X its local ring OX,Y at
a one codimensional irreducible subvariety Y ,→ X is a discrete valuation
ring.

Proof. Since R is a noetherian local ring, the lemma of Nakayama applies to

M := m ⊂ R and tells us that m 6= 0 implies m2 6= m. We take any element
f ∈ m \ m2 and claim
m = (f ).
First of all we prove:
mn ⊂ (f )
for n  0. Namely: We have f ∈ I(Df ∩ Y ) ⊂ O(Df ). Then Y ,→ N (f ),
and thus Y , being of codimension 1 in X, is an irreducible component of
N (f ) ,→ X. Hence we may choose a principal open subset U ⊂ Df with
U ∩ Y = U ∩ N (f ). Replacing X with the affine variety U , we see that we
may assume X = Sp(A) and f ∈ A with Y = N (f ).

84
 Now, if I(Y ) = (g1 , ..., gr ), then gi ∈ I(N (f )), so Hilberts Nullstellensatz
provides a number m ∈ N with gim ∈ (f ) for i = 1, ..., r. Then we have

I(Y )n ⊂ (f ) ,→ A

with n := mr and as well

mn ⊂ (f ) ,→ R.
We now show that the assumption m 6⊂ (f ) leads to a contradiction. Choose
n ∈ N minimal with mn ⊂ (f ). Obviously n ≥ 2 and there is an element
g ∈ mn−1 \ (f ) ⊂ m. We investigate how the element h := g/f ∈ Q(A) \ A
acts on m ,→ R:

1. We have
g·m mn
h·m= ⊂ ⊂ A,
f f
because of g ∈ mn−1 and mn ⊂ (f ),

2. but
h · m 6⊂ m.
Namely: Otherwise M := m is a faithful R[h]-module, which is finitely
generated as an R-module. But that means that h ∈ Q(R) is integral
over R, hence, R being integrally closed in its field of fractions Q(R),
we conclude that h ∈ R, a contradiction.

Hence the ideal h · m ⊂ R contains elements in R \ m = R∗ , and consequently

h · m = R resp. h−1 ∈ m and

f = f · h · h−1 = g · h−1 ∈ m · m = m2 ,

a contradiction, since f ∈ m \ m2 .

As a consequence of Th.10.8 we can define multiplicities: Given a normal

irreducible variety X we can for any codimension one irreducible subvariety
Y ,→ X define an order function

ordY : R(X) −→ Z ∪ {∞}

85
associating to a rational function h ∈ R(X) its ”order” (or ”multiplicity”)
along Y as follows: If OX,Y ⊃ m = (f ), then
∗

n , if h = ef n , with e ∈ OX,Y
ordY (h) := .
∞ , if h = 0
The order function satisfies
ordY (gh) = ordY (g) + ordY (h)
as well as
ordY (g + h) ≥ min {ordY (g), ordY (h)} .
Note that in particular for irreducible Y ,→ X of codimension 1 and a nonzero
f ∈ R(X) we have either Y ∩ Df 6= ∅ or Y ∩ D1/f 6= ∅ depending on whether
ordY (f ) ≥ 0 or ordY (f ) < 0. So the points of indeterminacy of a rational
function f form a closed set of codimension at least 2. For normal curves we
thus find:
Remark 10.9. 1. A rational function f ∈ R(X) on an irreducible normal
curve has no points of indeterminacy, hence defines a morphism f :
X −→ P1 = k ∪ {∞}.
2. Indeed, we may replace P1 with any complete variety Y and obtain
that any morphism ϕ : U −→ Y defined on a non-empty open subset
U ⊂ X of an irreducible normal curve into a complete variety Y can
be extended to a morphism ϕ̂ : X −→ Y .
3. As a consequence we obtain that the category of irreducible normal
complete curves is anti-equivalent to the category of finitely generated
field extensions of our base field k of transcendence degree 1.
4. Finally let us mention, that for k = C the functor X 7→ Xh is an
equivalence between the category of irreducible normal complete curves
and the category of compact Riemann surfaces.
Definition 10.10. For an irreducible affine variety X we denote D(X) the
free abelian group generated by the codimension one irreducible subvarieties
of X, i.e.
( r )
X
D(X) := ni Yi ; ni ∈ Z, Yi ,→ X irreducible, dim X = dim Yi + 1 .
i=1
Pr
The elements D = i=1 ni Yi in D(X) are called Weil divisors on X.

86
 Now given a rational function f ∈ R(X) \ {0} on an irreducible normal
variety X we can associate to it a divisor
X
Df := ordY (f ) · Y.
irred. 1-codim Y ,→X

Note that the above sum is finite, indeed ordY (f ) 6= 0, iff Y is an irreducible
component of N (f ) or of P (f ), where N (f ) = f −1 (0) is the set of the zeros
of f and P (f ) = f −1 (∞) the set of poles of f . Then

R(X)∗ −→ D(X), f 7→ Df ,

is a homomorphism of abelian groups: Df g = Df + Dg . The divisors of the

form Df are called principal divisors; its cokernel is the divisor class group
of X.

11 Regular Points
In order to study local properties of an affine algebraic variety X ,→ k n one
could - as in real analysis - think of “approximating“ X near a point a ∈ X
by an affine linear subspace a + Ta X, where Ta X ,→ k n is a vector subspace
of k n : For a polynomial f ∈ k[T ] and a ∈ k n denote
n
X ∂f
da f := (a) · Ti ∈ k]T ]1
i=1
∂T i

its ”differential“ at a ∈ k n , i.e. da f is the linear term in the Taylor expansion

of f (T + a) ∈ k[T ] at 0. Then we define provisionally(!) the tangent space
of X at the point a ∈ X as

Ta X := N (k n ; da f, f ∈ I(X)).

Indeed, since da (f g) = f (a)da g + g(a)da f we see that

Ta X = N (k n ; da f1 , ..., da fr ),

where the polynomials f1 , ..., fr generate the ideal I(X).

87
Example 11.1. For a hypersurface X = N (k n ; f ) with squarefree f ∈ k[T ]
(such that I(X) = (f )) we have Ta (X) = N (k n ; da f ). So with the ”singular
set ” S(X) := {a ∈ X; da f = 0} ,→ k n we have

dim X , if a ∈ X \ S(X)
dim Ta X = .
dim X + 1 , if a ∈ S(X)

For example for f = T22 − T13 and char(k) 6= 2, 3 we have da f = 2a2 T2 −

3a21 T1 = 0 iff a = (a1 , a2 ) = 0, so the Neil parabola X = N (k 2 ; T22 − T13 ) has
one singular point, the origin.

Though the geometric information encoded in the tangent space is essen-

tially related to Ta X as a subspace of k n , it is nevertheless useful to define
it in a functorial way as an abstract vector space, depending only on the
local ring OX.a : While the only isomorphy invariant of a vector space is its
dimension, the functoriality enables us to rediscover embedded versions of
the tangent space.
To begin with let us consider a point a ∈ X in an affine variety X =
Sp(A). We call a linear map δ : A −→ k a derivation at a if the Leibniz rule

δ(f g) = δ(f )g(a) + f (a)δ(g)

holds for all f, g ∈ A. Denote Dera (A) the vector space of all derivations of
A at a ∈ X.

Example 11.2. For t ∈ k n denote δta : k[T ] −→ k, f 7→ da f (t), the formal

derivative in the direction t ∈ k n . Then

Dera (k[T ]) = {δta ; t ∈ k n } ∼

= kn.

If % : k[T ] −→ A denotes the restriction map onto A = k[T ]/I(X), then

Dera (A) −→ Dera (k[T ]), δ 7→ δ ◦ %

is an injection with image

Dera (A) ◦ % = {δta ; t ∈ Ta X}.

Indeed, both, the left and the right hand side consist of the derivations of
k[T ] at a vanishing on I(X).

88
Lemma 11.3. There is a natural isomorphism

(ma /m2a )∗ −→ Dera (A), α 7→ α ◦ q,

where q : A = k ⊕ ma −→ ma /m2a is the composition of the projection onto

ma with the quotient map ma −→ ma /m2a .
Proof. Obviously the given map is injective, and one checks immediately that
α ◦ q really is a derivation at a ∈ X. On the other hand for any derivation
δ ∈ Dera (A) we have δ|m2a = 0 as a consequence of the Leibniz rule, so δ
induces a linear form δ : ma /m2a −→ k, f + m2a 7→ δ(f ), and since δ|k = 0
anyway, we obtain δ = δ ◦ q.
In the next step we replace ma ⊂ A with the maximal ideal mX,a ⊂ OX,a
of the local ring of X at a.
Remark 11.4. For a ∈ X = Sp(A) the natural map A −→ OX,a induces an
isomorphism
ma /m2a −→ mX,a /m2X,a .
To see this remember that for any multiplicative set the functor M 7→ S −1 M
is exact. Here we take S := A \ ma . Then mX,a = S −1 ma as well as
m2X,a = S −1 m2a , while S −1 (ma /m2a ) = ma /m2a , since f ∈ S acts on ma /m2a
by multiplication with f (a) ∈ k ∗ .
Definition 11.5. The Zariski tangent space of an algebraic variety X at a
point a ∈ X is defined as

Ta X := (mX,a /m2X,a )∗ .

Remark 11.6. 1. For any morphism ϕ : X −→ Y of algebraic varieties

and a ∈ X there is an induced homomorphism Ta ϕ : Ta X −→ Tϕ(a) Y .
Let b := ϕ(a). The pull back ϕ∗ : OY,b −→ OX,a maps mY,b to
mX,a as well as m2Y,b to m2X,a , so there is an induced homomorphism
mY,b /m2Y,b −→ mX,a /m2X,a ,. Its dual is our homomorphism Ta ϕ.

2. There is a covariant functor

T : (X, a) 7→ Ta X, ϕ 7→ Ta ϕ

from the category of germs of algebraic varieties to the category of

k-vector spaces.

89
 3. The tangent map Ta ι : Ta Z −→ Ta X induced by a closed embedding
ι : Z ,→ X of a closed subvariety Z containing a is injective: The ho-
momorphisms OX,a −→ OZ,a as well mX,a −→ mZ,a and mX,a /m2X,a −→
mZ,a /m2Z,a are surjective. Hence Ta ι, being the dual of the last map, is
injective.

Proposition 11.7. For the tangent space Ta X of an algebraic variety we

have
dim Ta X ≥ dima X.

Proof. If X1 , ..., Xr are the irreducible components of X passing through

a, we have dima X = max(dima X1 , ..., dima Xr ) and Ta Xi ,→ Ta X. So it
suffices to prove the claim in the case r = 1. Furthermore we may shrink X,
hence assume that X = Sp(A) is affine and that the vector space mX.a /n2X,a
has a basis f1 + m2X,a , ..., fn + m2X,a with regular functions fi ∈ O(X). Then,
according to the lemma of Nakayama we have even mX,a = (f1 , ..., fn ) resp.
after a further shrinking ma = (f1 , ..., fn ) ,→ A. But according to Cor.8.10
every irreducible component of N (X; f1 , ..., fn ) = {a} has at least dimension
dim X − n, whence n ≥ dim X.

Definition 11.8. A point of an algebraic variety is called a regular or smooth

point if dim Ta X = dima X, otherwise a singular point or a singularity. We
denote S(X) ⊂ X the set of all singular points of X, also called its singular
locus. The complement X \ S(X) then is the regular locus of X.

Example 11.9. 1. If X = N (k n ; f1 , ..., fr ) is of pure dimension

  d, then a
∂fi
point a ∈ X is a smooth point iff the Jacobi matrix ∂Tj (a)
1≤i≤r,1≤j≤n
has a non-vanishing (n−d) × (n−d)-minor.

2. The local ring OX,a of a curve is a discrete valuation ring iff a ∈ X is

a normal point iff a ∈ X is a smooth point: A discrete valuation ring
R is a UFD, hence integrally closed in its field of fractions, and the
quotient m/m2 is a one-dimensional vector space over R/m, since the
maximal ideal m ,→ R is a principal ideal. On the other hand Th. 10.8
ensures that the local ring at a normal point of a curve is a discrete
valuation ring. For a smooth point a ∈ X we have dimk mX,a /m2X,a = 1,
hence, according to Nakayamas lemma, mX,a is a principal ideal, while
OX,a is an integral domain according to Th.11.10 below and noetherian
anyway.

90
 For the next result we need a statement which we shall prove later on:

Theorem 11.10. The local ring OX,a of an algebraic variety at a smooth

point is an integral domain. Equivalently, a smooth point of a regular variety
is contained in exactly one irreducible component of X.

Theorem 11.11. The regular locus X \ S(X) of an algebraic variety is a

dense open set.

Proof. Let us first show that the singular locus is closed: S(X) ,→ X, resp.
that the regular locus X\S(X) is open. Since according to 11.10 the reducible
points of X, i.e. those which are contained in several irreducible components
of X, are singular we may assume that X is irreducible, indeed even affine.
But in that case the description of S(X) of Ex. 11.9.1 by minors gives the
result. It remains to show that any irreducible variety X admits smooth
points. We start with irreducible hypersurfaces:
Proposition 11.12. For an irreducible polynomial f ∈ k[T1 , ..., Tn ] there
are points a ∈ N (k n ; f ) with da f 6= 0. In particular the regular locus of
X = N (k n ; f ). is nonempty.
∂f
Proof. If da f = 0 for all a ∈ X, we have ∂T i
∈ I(X) for i = 1, ..., n resp.
∂f
f | ∂Ti . Hence all partial derivatives, having a total degree less than that of
f , have to vanish. And that means that our base field k has characteristic
p > 0 and f ∈ k[T p ] = k[T1p , ..., Tnp ], but k being algebraically closed we have
k[T p ] = (k[T ])p , in particular, f is not irreducible, a contradiction.
In order to apply that in the general case we need
Proposition 11.13. If L = k(a1 , ...., ar ) ⊂ k is a finitely generated field
extension, then there is a subset I ⊂ {1, ..., r}, such that the ai , i ∈ I, are
algebraically independent and L ⊃ E := k(ai ; i ∈ I) is a finite separable field
extension.
Proof. We use induction on r. We may assume that a1 , ..., a` is a maximal
set of over k algebraic independent elements. If ` = r there is nothing to
be shown. Otherwise denote f ∈ k[T1 , ...T` , Tr ] \ {0} a polynomial with
f (a1 , ..., a` , ar ) = 0. We may assume that f is irreducible. Hence, according
∂f
to Prop.11.12 some partial derivative ∂T i
, i = 1, ..., `, r, does not vanish,
indeed we may assume i = r. So ar is separable over k(a1 , ..., a` ) as well as
the field extension L ⊃ k(a1 , ..., ar−1 ). Now apply the induction hypothesis to

91
F := k(a1 , ..., ar−1 ) and use the fact that two successive separable extensions
give again a separable extension.

Finally we are done with the following

Lemma 11.14. Any irreducible algebraic variety X (of dimension n, say)
contains a non-empty open subset U isomorphic to an open subset V of a
hypersurface N (k n+1 ; f ) in affine n+1-space:

X⊃U ∼
= V ⊂ N (k n+1 ; f ).

Proof of Lemma 11.14. According to Prop.11.13 there is an intermediate

field E ∼
= k(T1 , ..., Tn ) of the field extension R(X) ⊃ k, such that R(X) ⊃ E
is a finite separable extension. The primitive element theorem now pro-
vides an element ϑ ∈ R(X) with R(X) = E[ϑ]; indeed we may even (after
possibly multiplying with some polynomial in k[T1 , ..., Tn ]) assume that ϑ
is integral over k[T1 , ..., Tn ]. Then we have f ∈ (k[T1 , ..., Tn ])[Tn+1 ] for the
minimal polynomial f of ϑ over k(T1 , ..., Tn ). Thus N (k n+1 ; f ) and X have
isomorphic rational function fields, and we may apply 9.5.

Remark 11.15. Let us comment on smooth points in complex algebraic

varieties. We may assume that X ,→ Cn is affine, let I(X) = (f1 , ..., fr ).
Consider a smooth point a ∈ X. W.l.o.g. da f1 , ...,  da f` form a basis of
Pr ∂fi
i=1 Cda fi ,→ C[T ]1 , and the Jacobi matrix ∂Tj
(a) is nonsingular.
1≤i,j≤`
Then
F : t 7→ (z1 , ..., zn ) := (f1 (t), ..., f` (t), t`+1 , ..., tn )
maps an open neighborhood U of a biholomorphically to a ball V centered
at F (a) and
F (X ∩ U ) ⊂ N (V ; z1 , ..., z` ) ,→ V.
But we may assume that dimb X = dima X for all b ∈ U , hence the functions
g`+1 := f`+1 ◦ F −1 , ..., gr := fr ◦ F −1 do not depend on z`+1 , ..., zn - otherwise
we would obtain dim Tb X < n − ` = dima X = dimb X for some points b ∈ U .
So gi (0, ..., 0, z`+1 , ..., zn ) = gi (F (a)) = 0 even for i ≥ ` + 1 and thus

F (X ∩ U ) = N (V ; z1 , ..., z` ) ,→ V,

92
i.e., the point a ∈ X admits with X ∩ U a strong neighborhood, which is, as
a complex analytic variety, isomorphic to an open ball in Cn−` .
In contrast to that it is quite far from being true – even for k = C – that a
smooth point a ∈ X admits a Zariski-neighborhood isomorphic (as algebraic
variety) to an Zariski-open subset of k n (where n = dima X). Indeed, this
would imply that (w.l.o.g. X being irreducible) R(X) ∼ = k(T1 , ..., Tn ) is a
purely transcendent extension of k.

In order to investigate the local structure of a variety near a point a ∈

X a little bit closer we replace the tangent space Ta X with the tangent
cone T Ca (X) ,→ Ta X. As before we start with an embedded description
for X ,→ k n at a = 0. While the tangent space T0 X is described by the
linear parts of the polynomials vanishing on X, every nonzero polynomial
f ∈ I(X) contributes to the description of the tangent cone T C0 (X) with its
(non-vanishing) Phomogeneous component f min of lowest degree: For such a
polynomial f = ri=q fi ∈ k[T ] = ∞
L
i=0 k[T ]i we define

f min := fq , if fq 6= 0.

Then
T C0 (X) := N (I0 (X)) ⊂ T0 X
with the homogeneous ideal
X
I0 (X) := k[T ] · f min .
f ∈I(X)∗

and I(X)∗ := I(X) \ {0}.

Example 11.16. Since (f g)min = f min · g min , the tangent cone of a hyper-
surface X = N (k n ; f ) with squarefree f is a hypersurface as well:

I0 (X) = (f min ) and T C0 (X) = N (k n ; f min ).

For example the noose X = N (k 2 ; T22 − T12 (T1 + 1) has tangent cone

T C0 (X) = N (k 2 ; T22 − T12 ) = k(1, 1) ∪ k(1, −1) ,→ k 2 = T0 X

a union of two lines (if char(k) 6= 2). In particular, the tangent cone can
detect the two branches of X at 0 in a purely algebraic way!

93
 On the other hand for the Neil parabola X = N (T22 − T13 ) we have
I0 (X) = (T22 )
and hence
T C0 (X) = k(1, 0).
So we see, that I0 (X) need not be a radical ideal!

Let us now present a description of T Ca (X) depending only on the local

ring OX,a of X at a. We start with an algebraic construction, the associated
graded algebra:
Definition 11.17. Let R be a ring and m ,→ R a maximal ideal. The
associated graded algebra grm (R) is the direct sum
∞
M
grm (R) := mq /mq+1 ,
q=0

the multiplication being the bilinear extension of

mq /mq+1 × ms /ms+1 −→ mq+s /mq+s+1 ,
(a + mq+1 , b + ms+1 ) 7→ ab + mq+s+1 .
If R is a local ring, then gr(R) := grm (R) with the unique maximal ideal
m ,→ R.
Remark 11.18. 1. Note that there is no natural ring homomorphism
R −→ gr(R). Nevertheless for f ∈ R we can define

f + mq+1 ∈ gr(R)q , if f ∈ mq \ mq+1 for some q ∈ N
gr(f ) := ,
0 , otherwise
but that does not yield a homomorphism. We note that gr(f g) =
) · gr(g) holds, if gr(R) is an integral domain and that gr(f ) = 0 iff
gr(f T
f∈ ∞ q
q=1 m .

2. The ideal m gives rise to a filtration (mq )q∈N of the ring R, but only if
we can find for every q a complementary additive subgroup Rq ⊂ mq ,
i.e. satisfying mq = Rq ⊕ mq+1 , suchL that Ri · Rj ⊂ Ri+j , then gr(R)
can be identified with the subring ∞ q=0 Rq of R. For example this is
true for R = k[T ] and m = m0 : Take simply Rq = k[T ]q , the vector
subspace of polynomials homogeneous of degree q.

94
 3. For the local ring S −1 R with S := R \ m we have

gr(S −1 R) ∼
= grm (R)

due to the fact that S −1 (mq /mq+1 ) ∼

= mq /mq+1 , the elements in S al-
ready acting as isomorphisms on the R/m-vector spaces mq /mq+1 .

Proposition 11.19. For X = Sp(A) ,→ k n we have

k[T ]/I0 (X) ∼

= gr(OX,0 ).

Proof. According to Rem.11.18.3 we have

gr(OX,0 ) ∼
= grm0 (A).

On the other hand the quotient homomorphism

% : k[T ] −→ A, f 7→ f |X

induces a surjective homomorphism

gr(%) : k[T ] ∼
= gr(k[T ]) −→ gr(A),

where the graded algebras are taken with respect to m0 ,→ A and m :=

%−1 (m0 ) ,→ k[T ]. We have to show that ker(gr(%)) = I0 (X). For a homoge-
neous polynomial f ∈ k[T ]q we have

gr(%)(f ) = 0 ⇐⇒ f |X ∈ mq+1
0 ⇐⇒ f ∈ mq+1 + I(X),

i.e. f = g min for some g ∈ I(X).

Definition 11.20. The reduction Red(R) of a ring R is defined as

√
Red(R) := R/ 0,
√
where 0 ,→ R denotes the nilradical of R.

1. The reduction Red(A) of a graded algebra A = ∞

L
Remark 11.21. q=0 Aq
√
is again a graded algebra, its nilradical 0 ,→ A being a graded ideal.

95
 2. A graded reduced affine algebra A = ∞
L
q=0 Aq is (as graded algebra)
isomorphic to the algebra of regular functions on an affine cone iff A,
as a k-algebra, is generated by the elements of degree 1.
Definition 11.22. The tangent cone of an algebraic variety X at a point
a ∈ X is defined as the affine cone

T Ca (X) := Sp (Red(gr(OX,a ))) .

Remark 11.23. There is a natural embedding

T Ca (X) ,→ Ta X,

since the linear map

mX,a /m2X,a = gr(OX,a )1 ,→ gr(OX,a ) −→ Red(gr(OX,a ))

extends uniquely to a graded epimorphism

S ∗ (mX,a /m2X,a ) −→ Red(gr(OX,a )) ∼

= O(T Ca (X))

and S ∗ (mX,a /m2X,a ) ∼

= O(Ta X) because of (Ta X)∗ ∼
= mX,a /m2X,a .
In contrast to the tangent space, the tangent cone has always the same
dimension as the original variety:
Theorem 11.24. For a pure d-dimensional variety X all the tangent cones
T Ca (X) at its points a ∈ X are pure d-dimensional as well.
Proof. The idea of the proof is as follows: First of all we may assume that
X = Sp(A) ,→ k n and a = 0. We then construct a morphism π : Z −→ k n ,
called ”the blow up of k n at the origin ”, such that
1. Z is irreducible and can be covered by n open subsets Wi ∼
= kn,
2. the morphism π : Z −→ k n induces an isomorphism

π|Z ∗ : Z ∗ −→ (k n )∗

with Z ∗ := π −1 ((k n )∗ ), and

3. the ”exceptional fiber ” π −1 (0) ,→ Z is a local hypersurface isomorphic

to projective (n−1)-space Pn−1 .

96
 b of X ,→ k n with respect to π is obtained from
The ”strict transform ” X
∗
X := X \ {0} as the closure
b := π −1 (X ∗ ) ,→ Z,
X
b ,→ π −1 (X) being proper in general. We shall prove that the
the inclusion X
isomorphism
π −1 (0) ∼
= Pn−1
induces an isomorphism
b ∩ π −1 (0) ∼
X = N (Pn−1 ; I0 (X)).

Hence, the projective variety N (Pn−1 ; I0 (X)), being isomorphic to a local

hypersurface with dense complement in the pure-d-dimensional variety X,b is
pure of codimension 1 in X.
b The tangent cone being the affine cone over it,
then is of pure dimension dim X.
Blow up: Let us give first a heuristic description of Z: It is obtained from
k n by a sort of surgery, one removes the origin and replaces it with a copy of
Pn−1 in such a way, that the closure of k ∗ · t in Z is obtained by adding the
limit point [t] ∈ Pn−1 .
We realize the variety Z as the closure

Z := Γϕ ,→ k n × Pn−1

of the graph
Γϕ ,→ (k n )∗ × Pn−1
of the quotient morphism

ϕ : (k n )∗ −→ Pn−1 , t 7→ [t]

in k n × Pn−1 . We have

Z = N (k n × Pn ; Si Tj − Sj Ti , 1 ≤ i < j ≤ n},

where the points in k n × Pn−1 are denoted (t, [s]) with t, s ∈ k n and Si Tj −
Sj Ti ∈ k[T1 , ..., Tn , S1 , ..., Sn ]. In order to understand the geometry of the
variety Z we consider the restricted projection
pr
p : Z ,→ k n × Pn−1 −→
2
Pn−1 .

97
Above the standard open subset Vi := Pn−1 \ N (Pn−1 ; Si ) there is a trivial-
ization
∼
=
τi : Vi × k −→ Wi := p−1 (Vi ),
([s], λ) 7→ (λs−1
i s, [s]).

Hence Z = ni=1 Wi with the open subsets Wi ∼ = Vi × k ∼

= k n−1 × k = k n .
S
Since each of the subsets Wi ⊂ Z is dense because of Wi ∩ Wj ∼= (Vi ∩ Vj ) × k,
we get the first part of our claim about Z, and the exceptional fiber is a local
hypersurface, since

π −1 (0) ∩ Wi = τi (Vi × 0) ,→ τi (Vi × k) ∼

= Wi .

Strict transform: We have an isomorphism

π ◦ τi : Vi × k ∗ −→ (k n )Ti

given by
([s], λ) 7→ λs−1
i s.

In order to determine X b ∩ Wi we look at Vi × k ∗ ⊂ Vi × k ∼

bi := X = Wi and
compute the closure of

(π ◦ τi )−1 (XTi ) ,→ Vi × k ∗ ⊂ Vi × k
−`
in Vi × k. A regular function in k[T ]Ti vanishing on
P XTi is of the formn Ti f
with a polynomial f ∈ I(X) ,→ k[T ]. If f = q fq ∈ I(X) ⊂ k with
homogeneous polynomials fq of degree q, we find
  X  
−` S S
f ◦ π ◦ τi = Λ f Λ = fq Λq−` .
Si q
S i

Now τi−1 (X
b ∩ Wi ) ,→ Vi × k is the zero locus of all regular functions on Vi × k,
whose restriction to Vi ×k ∗ vanishes on (π ◦τi )−1 (XTi ). These functions being
of the above form with only q ≥ `, either vanish on Vi ×0 or yield f min (Si−1 S).
Since that holds for i = 1, ..., n, we see that

(0 × Pn−1 ) ∩ X
b = 0 × N (Pn−1 ; I0 (X))

as desired.

98
 Let us now draw some conclusions:
Remark 11.25. If a ∈ Y ,→ X, then obviously T Ca (Y ) ,→ T Ca (X). Fur-
thermore, the closure of a finite union being the finite union of the separate
closures, we see that, if X1 , ..., Xr are the irreducible components of X passing
through a, then
[r
ϕ(T Ca (X)∗ ) = ϕ(T Ca (Xi )∗ )
i=1
n ∗
with the quotient morphism ϕ : (k ) −→ Pn−1 . It follows
r
[
T Ca (X) = T Ca (Xi ).
i=1

Note that the corresponding statement does not hold for the tangent spaces:
The union of the subspaces Ta (Xi ) ,→ Ta X will in general be no subspace
2
anymore - e.g.
Pr consider X = N (k ; T1 T2 ) at a = 0,2 but it is not even true
2
that Ta X = i=1 Ta (Xi ), as the example X = N (k ; T2 (T2 − T1 )) shows.

Theorem 11.26. A point a ∈ X in a variety X is a smooth point iff gr(OX,a )

is a polynomial ring.
Proof. Let g1 , ..., gn be a basis of mX,a /m2X,a . Then

σ : k[S1 , ..., Sn ] −→ gr(OX,a ), Si 7→ gi ,

is an epimorphism. With other words

 p 
T Ca (X) = Sp (Red(gr(OX,a ))) ∼
= N k n ; ker(σ) ,→ k n .

”⇐=”: If gr(OX,a ) is a polynomial ring, then T Ca (X) ∼

= k n and dimk gr(OX,a )1 =
dim T Ca (X), whence

dimk Ta X = dimk gr(OX,a )1 = dim T Ca (X) = dima X,

so X is smooth at a.
”=⇒”: If a ∈ X is a smooth point and dima X = n, the n-dimensional
subvariety T Ca (X) ,→ k n coincides with k n . In particular ker(σ) = {0} and
σ : k[S] −→ gr(OX,a ) is an isomorphism.

99
Remark 11.27. Note that for a smooth point a ∈ X with dima X = n
we have T Ca (X) ∼
= k n , but T Ca (X) ∼
= k n does not imply that a ∈ X is a
smooth point. For example for the Neil parabola X = N (k 2 ; T22 − T13 ) we
have T C0 (X) = k × 0 ∼
= k, while gr(OX,0 ) ∼
= k[T1 , T2 ]/(T22 ).
In order to prove Th, 11.10, we need a further result of commutative
algebra:

Theorem 11.28 (Krull Intersection Theorem). Let M be a finitely generated

module over the local noetherian ring R with maximal ideal m. Then
∞
\
mq M = {0}.
q=1

Proof of Th.11.10. Take f, g ∈ OX,a \ {0}. The Krull intersection theorem

gives gr(f ) 6= 0 6= gr(g), hence gr(OX,a ) being an integral domain, we obtain
0 6= gr(f ) · gr(g) = gr(f g), in particular f g 6= 0.
Proof of 11.28. Since R is noetherian, the module M is noetherian as well:
So the submodule ∞
\
D := mq M
q=1

is finitely generated. If we knew that mD = D, the lemma of Nakayama

would give the result, but we do not and have to use a more sophisticated
argument. The module M being noetherian, there is a maximal submodule
F ⊂ M with D ∩ F = 0. Furthermore, according to Nakayama’s lemma
D = 0 iff D/mD = 0. Hence we may replace M with M/mD resp. assume
that mD = 0. Now we are looking for some q ∈ N satisfying mq M ⊂ F .
Namely, if we succeed with that, then

D = D ∩ mq M ⊂ D ∩ F = 0.

We show that for every a ∈ m there is some exponent s ∈ N with as M ⊂ F .

Hence, if m = (a1 , ..., ar ) and si ∈ N for ai as above, then we have mq M ⊂ F
with q := r max(s1 , ..., sr ). So, given an element a ∈ m let us look for an
appropriate exponent s ∈ N. Indeed, if (as M + F ) ∩ D = 0, it already follows
from the maximality of F , that F = as M + F resp. as M ⊂ F . Now the

Mk := {u ∈ M ; ak u ∈ F }

100
form an increasing sequence of submodules Mk ⊂ M ; hence M being noethe-
rian, we have Ms = Ms+1 for some s ∈ N. For b = as m + f ∈ (as M + F ) ∩ D
we then obtain
ab = as+1 m + af ∈ aD ⊂ mD = 0,
i.e. ab = 0 and as+1 m = −af ∈ F resp. m ∈ Ms+1 = Ms and already
as m ∈ F . Hence b ∈ F and thus b ∈ F ∩ D = 0, i.e. b = 0 as desired.

Eventually we want to discuss a little bit more in detail the structure of

a local ring at a regular point.
Definition 11.29. Let R be a noetherian local ring. The m-adic completion
R
b of R is the ring
Y∞
n
R := lim R/m ⊂
b R/mn
←−
n=1

consisting of the sequences (an +m )n≥1 satisfying an+1 − an ∈ mn for all

n

n ∈ N>0 . With other words, a sequence in R b is obtained recursively as

follows: Start with any element in R/m, and given the n-th element take as
its immediate successor any inverse image of it with respect to the natural
map R/mn+1 −→ R/mn .
Example 11.30. 1. For R = OX.a with X = Sp(A) we have
bX,a ∼
O = lim A/(ma )n ,
←−

to be seen as follows: Since the elements in S := A \ ma become units

in the local ring A/(ma )n , we have

OX,a /(mX,a )n ∼
= S −1 (A/(ma )n ) ∼
= A/(ma )n .

2. For a ∈ k n we have
bkn ,a ∼
O = k[[T1 , ..., Tn ]],
where k[[T1 , ..., Tn ]] denotes the ring of formal power series in n inde-
terminates with coefficients in k. Assuming a = 0 apply the previous
point to X = k n .
Proposition 11.31. A point a ∈ X is a simple point iff the completion O
bX,a
is isomorphic to a formal power series ring.

101
Proof. Let a ∈ X be a smooth point. We may assume that X = Sp(A).
Take functions f1 , ..., fn ∈ ma providing a basis of ma /(ma )2 and consider the
map
k[T1 , ..., Tn ] −→ A, Ti 7→ fi .
It induces surjections k[T ] −→ A/(ma )n for every n ∈ N. On the other hand,
gr(OX,a ) being a polynomial ring in the indeterminates gr(fi ), we conclude
that k[T ]/(m0 )n and A/(ma )n have the same dimension as a k-vector space,
so the induced map
k[T1 , ..., Tn ]/(m0 )n −→ A/(ma )n
is an isomorphism for all n ∈ N. Now apply Ex.11.30.2. On the other hand
gr(OX,a ) ∼
= gr(O
bX,a )

gives with Th.11.26 the reverse implication.

Theorem 11.32. The local ring OX,a of an algebraic variety X at a smooth
point a is factorial.
A possibility to prove Th.11.32, is to use the fact that the formal power
bX,a ∼
series ring O = k[[T1 , ..., Td ]] is known to be factorial, but we prefer here
a geometric proof of 11.32. Of course we may assume that X = Sp(A) is an
affine irreducible variety. To any subvariety Z ,→ X passing through a we
associate the ideal
I(Z)a := {f ∈ OX,a ; f ∈ I(Df ∩ Y ) ⊂ O(Df )}.

Theorem 11.33. Let a ∈ X be a smooth point of the irreducible affine vari-

ety X = Sp(A). Then for every irreducible subvariety Z ,→ X of codimension
1 passing through a, the ideal I(Z)a ⊂ OX,a is a principal ideal.
Proof of Th.11.32. Since I(Z)a = (f ) for Z ,→ X irreducible of codimension
1 is a prime ideal, the generator f is a prime element. It remains to show
that every non-unit is a product of such generators. To see that we consider
the set of proper principal ideals (g) ⊂ OX,a with a generator g not admitting
a factorization. Take a maximal such ideal (g). Assuming g ∈ A we have
a ∈ N (g) ,→ X. Let Z ,→ N (g) be an irreducible component passing through
a. Then Z has codimension 1 according to Krull’s principal ideal theorem,
Th.8.6, and thus I(Z)a = (f ). Hence g = hf with (h) % (g). So h admits a
factorization and thus g as well: Contradiction!

102
Proof of Th.11.33. First let us reduce the problem to the case where

1. X ,→ k d+1 is a hypersurface and a = 0, and

2. the restriction ψ : X −→ k d of the projection k d+1 −→ k d onto the first

d coordinates is finite and Z ∩ ψ −1 (0) = {0}.

The following lemma describes how to reduce successively the embedding

dimension of X:
Lemma 11.34. Let X ,→ k n be a d-dimensional irreducible variety, 0 ∈ X
a regular point, and Z ,→ X a 1-codimensional irreducible subvariety passing
through 0. Then there is a projection πn−1 : k n −→ k n−1 such that for
ψ = ψn−1 := πn−1 |X the following holds:

1. ψ is finite,

2. the tangent map T0 ψ : T0 X −→ T0 k n−1 is injective,

3. for n ≥ d + 2 we have ψ −1 (0) = {0}, and ψ induces an isomorphism of

germs (X, 0) ∼
= (ψ(X), 0), finally
4. for n = d + 1 we have still ψ −1 (0) ∩ Z = {0}.

Proof. A projection (:= surjective linear map) π : k n −→ k n−1 is up to a

coordinate change in k n−1 determined by its kernel L := ker(π), a line. In
order to find the good choice of L we consider k n ∼
= U0 ⊂ Pn as a subset of
projective n-space. Denote X the projective closure of X and X∞ := X
b b \X
its part at infinity, furthermore

q : Pn \ {[1, 0, ..., 0]} −→ N (Pn ; T0 ), [t] 7→ [0, t1 , ..., tn ]

the projection onto N (Pn ; T0 ) ∼

= Pn−1 . For n ≥ d + 2 consider the set

E := q(X \ {0}) ∪ q(T0 X \ {0}) ∪ X∞

with the identification T0 X ,→ T0 k n ∼

= k n . It is a proper subset of N (Pn ; T0 ) ∼
=
Pn−1 for dimensional reasons: dim X∞ = d − 1, since X∞ ,→ X is a local b
hypersurface, and q(T0 X \ {0}) ∼ = Pd−1 , while dim q(X \ {0}) ≤ dim X = d.
For n = d + 1 let

E := q(Z \ {0}) ∪ q(T0 X \ {0}) ∪ X∞ .

103
Now for L = (q −1 (a) ∩ U0 ) ∪ {0} with a 6∈ E the conditions 2), 4) and
the first one of 3) are obviously satisfied. It remains to show the finiteness.
After a linear change of coordinates we may assume that L = ken resp. that
πn−1 is the projection onto the first n − 1 coordinates. Then we have to
find an integral equation for Tn |X over k[T1 , ..., Tn−1 ]. Since [0, ..., 0, 1] 6∈ X,b
there is a polynomial f ∈ I(X) ⊂ k[T1 , ..., Tn ], such that its homogeneization
fb = T0r f (T0−1 T ) satisfies fb(0, .., 0, 1) = 1 and thus fb ∈ (k[T0 , ..., Tn−1 ])[Tn ]
as well as f = fb(1, T1 , ..., Tn ) are Tn -monic - this gives the desired integral
equation. Finally let us show the second part of 3): Set Y := ψ(X) ,→ k n−1 .
Since ψ is finite and ψ −1 (0) = {0}, the sets ψ −1 (V ) form a neighbourhood
basis of 0 ∈ X. As a consequence, if O(X) = O(Y )f1 + ... + O(Y )fr , we have
as well OX,0 = OY,0 f1 + ... + OY,0 fr . Since T0 ψ : T0 X −→ T0 Y is injective,
ψ ∗ : mY,0 /m2Y,0 −→ mX,0 /m2X,0 is surjective. Now we apply Nakayama’s
lemma to the finite OX,0 -module mX,0 and obtain mX,0 = OX,0 · ψ ∗ (mY,0 ).
Hence
OX,0 /OX,0 ψ ∗ (mY,0 ) = OX,0 /mX,0 ∼= k.
So 1 is a generator of the finite OY,0 -module OX,0 since it is mod mY,0 , i.e.
ψ ∗ : OY,0 −→ OX,0 is surjective, while injectivity follows from the fact that
ψ : X −→ Y is dominant and ψ −1 (0) = {0}.
The special situation as described in the beginning is treated with the
tools of the proof of Krull’s principal ideal theorem. Let X = Sp(A) and
treat the pull back homomorphism ψ ∗ : k[T ] ,→ A as an inclusion. We
write I(ψ(Z)) = k[T ] · h with some polynomial h ∈ k[T ] and show that
I(Z)0 = OX,0 · h. If ψ −1 (0) = {x1 = 0, x2 , ..., xr }, we are done, if we succeed
in showing
1. Any function g ∈ I(Z) with g(xi ) 6= 0 for i = 2, ..., r lies in (h) ,→ OX,0 .
2. The ideal I(Z)0 is generated by functions g ∈ I(Z) as above.
1.) For the characteristic polynomial χ(T, S) ∈ (k[T1 , ..., Td ])[S] of the mul-
tiplication µg ∈ Endk(T ) (Q(A)) we have

χ(T, S) = p(T, S)S + (−1)` N(g) ∈ (k[T ])[S]

with the extension degree ` := [Q(A) : k(T )]. Due to the Hamilton-Cayley
theorem we have χ(µg ) = 0 resp. χ(g) = 0, hence

N(g) = (−1)`−1 p(T, g)g

104
yields N(g) ∈ I(ψ(Z)) = k[T ]·h; so it suffices to show that gb := p(T, g) ∈ A is
invertible in OX,0 ⊃ A or, equivalently, that gb(0) = p(0, g(0)) = p(0, 0) 6= 0.
To see that we consider the induced multiplication map

µg : A/(T1 , ..., Td ) −→ A/(T1 , ..., Td )

over the fiber ψ −1 (0) = N (X; T1 , ..., Td ) with g := g+(T1 , ..., Td ) ∈ A/(T1 , ..., Td ).
Note that the algebra A/(T1 , ..., Td ) need not be reduced, but in any case we
have
Red(A/(T1 , ..., Td )) ∼= O(ψ −1 (0)) ∼
= kr .
If now f (T, S) ∈ (k[T ])[S] denotes the minimal polynomial of Td+1 |X over
k(T1 , ..., Td ) (since k[T1 , ..., Td ] is integrally closed in its field of fractions
k(T1 , ..., Td ), it has coefficients in k[T1 , ..., Td ]), we have

A∼
= k[T1 , ..., Td+1 ]/(f (T, Td+1 )).

Hence
A/(T1 , ..., Td ) ∼
= k[Td+1 ]/(f (0, Td+1 )),
and r
Y
f (0, Td+1 ) = Td+1 (Td+1 − ai )mi ,
i=2

where xi = (0, ai ) ∈ k × k, i = 2, ..., r, are the points 6= 0 in the fiber ψ −1 (0).

d

The polynomial f (0, Td+1 ) has a simple zero at the origin, since otherwise
∂f
∂Td+1
(0) = 0 gives 0 × k ⊂ T0 X, which contradicts the injectivity of the
tangent map
T0 ψ : T0 X = ker d0 f −→ T0 k d = k d .
With the Chinese remainder theorem we decompose
r
A/(T1 , ..., Td ) ∼
M
=k⊕ k[Td+1 ]/((Td+1 − ai )mi )
i=2

and, correspondingly, g = (g 1 , ..., g r ). Since g(xi ) 6= 0, i = 2, ..., r, the element

g i ∈ k[Td+1 ]/((Td+1 − ai )mi ) is a unit, and thus the characteristic polynomial
of the isomorphism µgi has no zero for i = 2, ..., r, while that of µg1 = 0 is
S ∈ k[S]. Hence
χ(0, S) = Sq(S) = p(0, S)S
with p(0, 0) = q(0) 6= 0.

105
2.) Assume I(Z) = (h1 , ..., hs ). Choose c ∈ k, such that c2 6= hj (xi ) for
i = 1, ..., r, j = 1, ..., s and f ∈ I(Z) with f (xi ) = c for i = 2, ..., r - remember
that the restriction A = O(X) −→ O(Z ∪ {x2 , ..., xr }) is surjective. Take
now gj := hj + f 2 : According to Nakayama’s lemma they generate I(Z)0 ,
since their residue classes g j mod mX,0 I(Z)0 coincide with hj , hence generate
I(Z)0 /mX,0 I(Z)0 .

Remark 11.35. Let R := OX,a be the local ring of a regular point a ∈ X

of the algebraic variety X. According to the above argument we can embed
R as an intermediate ring

R0 := k[T ]m0 ⊂ R ⊂ k[[T ]].

More precisely, R looks as follows: Take a power series f ∈ k[[T ]] integral

over R0 . Then
R0 [f ] ⊃ R0
is a semilocal ring, i.e. it contains only finitely many maximal ideals, and

R = Q(R0 [f ]) ∩ k[[T ]] ,→ Q(k[[T ]]),

is the localization of R0 [f ] with respect to the maximal ideal m := R0 [f ] ∩

(T1 , ..., Tn ).

Trevlig Sommar!

106