P1: ZCK Revised Pages Qu: 00, 00, 00, 00

Encyclopedia of Physical Science and Technology EN001H-20 May 26, 2001 14:20

Algebraic Geometry
Rick Miranda
Colorado State University

I. Basic Definitions of Affine Geometry

II. Projective Geometry
III. Curves and Surfaces
IV. Applications

GLOSSARY ALGEBRAIC GEOMETRY is, broadly speaking, the

study of the solutions of polynomial equations. Since poly-
Affine space The space of n-tuples with coordinates in a nomials are ubiquitous in mathematics, algebraic geome-
field K . try has played a central role in the development and study
Algebraic set The set of solutions to a collection of poly- of many mathematical ideas. Its origins can be traced to
nomial equations in affine or projective space. ancient Greek mathematics, when the study of lines and
Coordinate ring Ring of polynomial functions on an al- conics in the plane began. From these relatively simple
gebraic set. beginnings, modern algebraic geometry has grown into a
Curve An algebraic variety of dimension one. subject which draws on and informs nearly every other dis-
Genus The fundamental invariant of curves, equal to the cipline in mathematics, and modern practitioners develop
dimension of the space of regular 1-forms. and use some of the most sophisticated “mathematical ma-
Moduli space A topological space (usually an algebraic chinery” (including sheaves, cohomology, etc.) available.
variety itself) whose points represent algebraic vari- However, the subject is still rooted in fundamental ques-
eties of a specific type. tions about systems of polynomial equations; the complex-
Projective space The space of (n + 1)-tuples with coor- ity and apparent abstraction of some of the modern tools
dinates in a field K , not all zero, up to scalar factor. of practicing algebraic geometers should not be confused
The natural compactification of affine space, which in- with a lack of concern for basic and accessible problems.
cludes points at infinity.
Quadric An algebraic variety defined by a single
I. BASIC DEFINITIONS OF
quadratic equation.
AFFINE GEOMETRY
Rational function A ratio of polynomial functions de-
fined on an algebraic set in affine or projective
A. Affine Space
space.
Regular function A rational function whose denomina- To be more precise about solutions to polynomial
tor does not vanish at a subset of an algebraic set. equations, one first specifies a field k of numbers where the

465
P1: ZCK Revised Pages
Encyclopedia of Physical Science and Technology EN001H-20 April 20, 2001 17:14

466 Algebraic Geometry

coefficients of the polynomials lie, and where one looks the most common algebraic structure is the ring, which
for solutions. A field is a set with two binary operations is a set with addition and multiplication, but not neces-
(addition and multiplication) in which all of the usual rules sarily division. The ring of importance in affine algebraic
of arithmetic hold, including subtraction and division (by geometry is the ring K [x] = K [x1 , x2 , . . . , xn ] of polyno-
nonzero numbers). The main examples of interest are the mials in the n variables x = (x1 , . . . , xn ). Sometimes this
field Q of rational numbers, the field R of real numbers, is called the affine coordinate ring, since it is generated
and the field C of complex numbers. Another is the finite by the coordinate functions xi . It is from this ring that we
field Z/ p of the integers {0, 1, . . . , p − 1} under addition draw the subset S ⊂ K [x] of polynomials whose zeros we
and multiplication modulo a prime p. want to study.
Solutions to polynomials in n variables would naturally An ideal J in a ring R (like K [x]) is a subset of the
be n-tuples of elements of the field. If we denote the field ring R with the special properties that it is closed under
in question by K , the natural place to find solutions is addition and “outside multiplication”: if f and g are in J
affine n-space over K , denoted by K n , AnK , or simply An : then f + g is in J , and if g is in J then f g is in J for any
f in the ring. An example of an ideal is the collection of
An = {z = (z 1 , z 2 , . . . , z n ) | z i ∈ K }.
all polynomials that vanish at a particular point p ∈ An .
When n = 1 we have the affine line, if n = 2 we have the If S is a subset of the ring, the ideal generated
 by S is
affine plane, and so forth. the set of all finite “linear combinations” i h i f i where
the h i ’s are in the ring and the f i ’s are in the given subset
S. The reader may check that this set, denoted by S, is
B. Affine Algebraic Sets
closed under addition and outside multiplication, and so
Let f (x) = f (x1 , . . . , xn ) be a polynomial in n variables is always an ideal.
with coefficients in K . The zeros of f are denoted by Returning to the setting of algebraic sets, one can easily
Z ( f ): see that if S is any collection of polynomials in K [x],
and J = S is the ideal in K [x] generated by S, then
Z ( f ) = {z ∈ An | f (z) = 0}.
S ⊂ J and Z (S) = Z (J ): the set S and the ideal J have
For example, Z (y − x 2 ) is a parabola in the plane, and exactly the same set of common zeros. This allows alge-
Z (x 2 − y 2 − z 2 − 4) is a hyperboloid in 3-space. braic geometers to focus only on the zeros of ideals of
More complicated geometric objects in affine space polynomials: every algebraic set is of the form Z (J ) for
must be defined by more than one polynomial. Let S be an ideal J ⊂ K [x]. It is not the case that the ideal defining
a set of polynomials (not necessarily finite) all having co- the algebraic set is unique, however: it is possible that two
efficients in the field K . The common zeros of all the different ideals J1 and J2 have the same set of common
polynomials in the set S is denoted by Z (S): zeros, so that Z (J1 ) = Z (J2 ).
Z (S) = {z ∈ An | f (z) = 0 for all f ∈ S}.
An algebraic set in A , or simply an affine algebraic set, is a
n
D. Algebraic Sets Defining Ideals
subset of An of the form Z (S) for some set of polynomials
S in n variables. That is, an affine algebraic set is a set We saw in the last paragraph that ideals may be used to
which is exactly the set of common zeros of a collection define all algebraic sets, but that the defining ideal is not
of polynomials. Affine algebraic sets are the fundamental unique. In the construction of algebraic sets, we look at
objects of study in affine algebraic geometry. a collection of polynomials and take their common zeros,
The empty set is an affine algebraic set [it is Z (1)] and which is a set of points in An . Now we turn this around,
so is all of affine space [it is Z (0)]. The intersection of an and look instead at a subset X ⊂ AnK of points in affine
arbitrary collection of algebraic sets is an algebraic set. space, and consider all the polynomials that vanish on X .
The union of a finite number of algebraic sets is also an We denote this by I (X ):
algebraic set. Therefore the algebraic sets in affine space
form the closed sets in a topology; this topology is called I (X ) = { f ∈ K [x] | f (z) = 0 for all z ∈ X }.
the Zariski topology.
No matter what kind of subset X we start with, this is
always an ideal of polynomials. Again there is the same
C. Ideals Defining Algebraic Sets
possibility of non-uniqueness: it may be that two differ-
The “algebraic” part of algebraic geometry involves the ent subsets X and Y of An have the same ideal, so that
use of the tools of modern algebra to study algebraic sets. I (X ) = I (Y ). This would happen if every polynomial that
Algebra is the study of sets with operations on them, and vanished on X also vanished on Y and vice versa.
P1: ZCK Revised Pages
Encyclopedia of Physical Science and Technology EN001H-20 April 20, 2001 17:14

Algebraic Geometry 467

The ideal I (X ) for a subset X ⊂ An has a special rational numbers, the field R of real numbers, and all finite
property not shared by all ideals: it is closed under roots. fields are not algebraically closed.
That is, if f is a polynomial, and f m vanishes on all the
points of X , then it must be the case that f itself vanishes. Theorem 1 (Hilbert’s Nullstellensatz). If K is an
This means that if f m ∈ I (X ), then f ∈ I (X ). An ideal algebraically closed field, then for any ideal J ⊂ K [x],
with this special property is called a radical ideal. I (Z (J )) = rad(J ).
The power of the Nullstellensatz occurs because in
E. The Z–I Correspondence many applications one has access to a polynomial f which
is known to be zero when other polynomials g1 , . . . , gk
The two operations of taking an ideal of polynomials and
are. The conclusion is then that f is in the radical of the
using Z to get a subset of affine space, and taking a subset
ideal generated by the gi ’s. Therefore there is a power of
of affine space and using I to get an ideal of polynomials
f which is a linear combination of the gi ’s, and so there
form the theoretical foundation of affine algebraic geom-
is an explicit equation of the form
etry. We have the following basic facts: 
fm = h i gi .
r If J1 ⊂ J2 ⊂ K [x] are ideals, then Z (J1 ) ⊃ Z (J2 ). i
r If X 1 ⊂ X 2 ⊂ An , then I (X 1 ) ⊃ I (X 2 ). The Nullstellensatz permits a more detailed correspon-
r If J ⊂ K [x] is an ideal, then I (Z (J )) ⊃ J . dence of properties between the algebraic set X and its
r If X ⊂ An , then Z (I (X )) ⊃ X . ideal I (X ). Typical statements when K is algebraically
closed are as follows:
This last statement can be sharpened to give a crite-
rion for when a subset of An is algebraic, as follows. r There is a one-to-one correspondence between
If X is algebraic, equal to Z (J ) for some ideal J , then algebraic subsets of An and radical ideals in K [x].
I (X ) = I (Z (J )) ⊃ J , so that Z (I (X )) = Z (I (Z (J ))) ⊂ r X is empty if and only if I (X ) = K [x].
Z (J ) = X , which forces Z (I (X )) = X . Conversely if r X = An if and only if I (X ) = {0}.
Z (I (X )) = X , then X is obviously algebraic. Hence, r X consists of a single point if and only if I (X ) is a

X is an algebraic subset of An if and only if Z (I (X)) = X. maximal ideal.

r X is an irreducible algebraic set (that is, it is not a
This statement addresses the question of which points are union of two proper algebraic subsets) if and only if
additional zeros of all the polynomials which vanish on X , I (X ) is a prime ideal [that is, I (X ) has the property
stating that if X is algebraic there are no additional such that if f g ∈ I (X ) then either f ∈ I (X ) or g ∈ I (X )].
points.
All these statements relate geometric properties of the
F. The Nullstellensatz and the Basis Theorem algebraic set X to algebraic properties of the ideal I (X ).
This is the basic theme of algebraic geometry in general.
The Nullstellensatz turns the previous question around The other main theorem concerning these general ideas
and asks which additional polynomials always vanish at is the basis theorem, which addresses the question of the
the points which are common zeros of a given ideal of finiteness of the set of equations which may define an
polynomials. This is a much more subtle question, and the algebraic set:
answer is the fundamental theorem in elementary affine
geometry. It uses the concept of the radical of an ideal J , Theorem 2 (Hilbert’s basis theorem). Every ideal
rad(J ), which is the set of all polynomials for which some J in the ring of polynomials K [x] is finitely generated:
power is in J : there is a finite set S of polynomials such that J = S.
rad(J ) = { f ∈ K [x] | f m ∈ J for some m ≥ 1}. As a consequence, every algebraic set X ⊂ An may be
defined as the set of common zeros of a finite set of poly-
It is always the case that rad(J ) ⊃ J , and that rad(J ) is a nomials.
radical ideal. Clearly I (Z (J )) ⊃ rad(J ): if f m ∈ J , then f
will certainly vanish wherever everything in J (in partic-
G. Regular and Rational Functions
ular wherever f m ) vanishes.
An algebraically closed field is a field that contains all Any polynomial f ∈ K [x] can be considered as a
roots of all polynomials in one variable. The field C of K -valued function on affine space An . If X ⊂ An is an
complex numbers is algebraically closed; the field Q of algebraic set, such a polynomial may be restricted to X
P1: ZCK Revised Pages
Encyclopedia of Physical Science and Technology EN001H-20 April 20, 2001 17:14

468 Algebraic Geometry

to obtain a function on X . Functions on X which are re- single quadratic equation in two variables, and they define
strictions of polynomial functions are called polynomial a curve in the affine plane. The classification and study of
functions. The polynomial functions on X form a ring, conics date to antiquity. Most familiar is the classification
denoted by K [X ] and called the affine coordinate ring of of nonempty irreducible conics over the real numbers R:
X . For example, the affine coordinate ring of affine space we have either an ellipse, a parabola, or a hyperbola. Again
is the entire polynomial ring: K [An ] = K [x]. an appreciation of the points “at infinity” sheds light on
There is an onto ring homomorphism (restriction) from this classification: an ellipse has no real points at infinity,
K [x] to K [X ], whose kernel is the ideal I (X ). There- a parabola has one real point at infinity, and a hyperbola
fore K [X ] is isomorphic as a ring to the quotient ring has two real points at infinity.
K [x]/I (X ). Over any field, if an irreducible conic C [given by
Irreducible algebraic sets are often referred to as f (x, y) = 0] is nonempty, then it may be parametrized
algebraic varieties. If X is an algebraic variety, then I (X ) using a rational function of one variable. This is done by
is a prime ideal, so that the coordinate ring K [X ] is an in- choosing a point p = (x0 , y0 ) on C, writing the line L t
tegral domain (if f g = 0 then either f = 0 or g = 0). The through p with slope t [given by y − y0 = t(x − x0 )], and
field of fractions of K [X ], denoted by K (X ), represent intersecting L t with C. This intersection will consist of
rational functions on the variety X . Rational functions are the point p and one other point pt which depends on t.
more useful than polynomial functions, but they have the One may solve easily for the coordinates of pt as ratio-
drawback that any given rational function may not be de- nal functions of t, giving a rational parametrization for the
fined on all of X (where the denominator vanishes). If a conic C. From the point of view of maps, this gives a ratio-
rational function is defined at a point p, one says that the nal map φ : A1 → C which has an inverse: for any point
function is regular at p. (x, y) on C, the t-value is t = (y − y0 )/(x − x0 ). Invertible
The rational function field K (X ) of an algebraic variety rational maps are called birational, and most of the clas-
X is an extension field of the base field K , and as such has sification efforts of algebraic geometry are classifications
a transcendence degree over K : this is the largest num- up to birational maps.
ber of algebraically independent rational functions. This Applying this procedure to the unit circle C (defined
transcendence degree is the dimension of the variety X . by x 2 + y 2 = 1) and choosing the point p to be the point
Points have dimension zero, curves have dimension one, (−1, 0), one finds the parametrization
surfaces have dimension two, and so forth.  
Polynomial and rational functions are used to define 1 − t2 2t
pt = , .
maps between algebraic sets. In particular, maps between 1 + t2 1 + t2
two affine spaces are simply given by a vector of functions.
We see here the origins of algebraic number theory, in
Maps between affine algebraic sets are given by restriction
particular the formulas for the Pythagorean triples. If
of such a vector of functions. Depending on the type of
we look at the case when t = p/q is a rational number,
functions used, such maps are called polynomial or ratio-
we find (clearing the denominator q) that ((q 2 − p 2 )/
nal maps. Again, rational maps have the property that they
(q 2 + p 2 ), 2 pq/(q 2 + p 2 )) is a point on the unit circle.
may not be defined everywhere. A map is called regular if
Again clearing the denominator q 2 + p 2 , we see that this
it is given by rational functions but is defined everywhere.
means

H. Examples (q 2 − p 2 )2 + (2 pq)2 = (q 2 + p 2 )2 ,

The most common example of an affine algebraic variety which gives the standard parametrization of Pythagorean
is an affine subspace: this is an algebraic set given by linear triples. This style of argument in elementary number the-
equations. Such a set can always be defined by an m × n ory dates the ancient Greeks, in particular to Appollonius
matrix A, and an m-vector b, as the vanishing of the set of and Diophantus.
m equations given in matrix form by Ax = b. This gives the In higher dimensions, an algebraic variety given by a
geometric viewpoint on linear algebra. One consequence single quadratic equation is called a quadric. Although the
of the geometric point of view is a deeper understanding classification of affine quadrics is not difficult, it becomes
of parallelism for linear spaces: geometrically, one begins clearer that the use of projective techniques simplifies the
to believe that parallel lines might be made to intersect matter considerably.
“at infinity.” This is the germ of the development of pro- Algebraic varieties defined by either more equations
jective geometry. or equations of degree larger than two present a much
No doubt the first nonlinear algebraic sets to be studied greater challenge. Even the study of cubic curves in the
were the conics. These are the algebraic sets given by a plane [given by a single equation f (x, y) = 0 where f
P1: ZCK Revised Pages
Encyclopedia of Physical Science and Technology EN001H-20 April 20, 2001 17:14

Algebraic Geometry 469

has degree three] is still a subject of modern research, [x0 : x1 : . . . : xn ], with each xi ∈ K , not all equal to zero,
especially over number fields. subject to the equivalence relation that
[x0 : x1 : . . . : xn ] = [λx0 : λx1 : . . . : λxn ]
I. Affine Schemes
for any nonzero λ ∈ K . The xi ’s are called the homoge-
The theory exposed above was developed in the hundred neous coordinates of the point [x]; note that they are not
years ending in the middle of the twentieth century. In the individually well-defined, because of the scaling condi-
second half of the twentieth century a different foundation tion. However, it does make sense to say whether xi = 0
to algebraic geometry was developed, which more closely or not.
follows the algebra of the rings and ideals in question. If x0 = 0, we can scale by λ = 1/x0 and assume x0 = 1;
Recall that there is a correspondence between algebraic then all the other n coordinates are well-defined and cor-
subsets of affine space and radical ideals in the polyno- respond to a unique point in affine n-space An :
mial ring K [x]. If the ground field K is algebraically
closed, points correspond to maximal ideals, and irre- p = (x1 , . . . , xn ) ∈ An corresponds to
ducible algebraic sets to prime ideals. The ring of polyno- [1 : x1 : . . . : xn ] ∈ Pn .
mial functions on X is naturally the ring K [x]/I (X ). This
ring is a finitely generated K -algebra, with no nilpotent However, we have a host of new points where x0 = 0 in
elements (elements f = 0 such that f k = 0 for some k). Pn ; these are to be thought of as points “at infinity” of
From this ring one can recover X , as the set of maximal An . In this way the points at infinity are brought into view
ideals. and in fact become no different than any other point, in
The idea of affine schemes is to start with an arbitrary projective space.
ring R and to construct a geometric object X having R
as its natural ring of functions. Grothendieck’s theory, de-
veloped in the 1950s and 1960s, uses Spec(R), the set of C. Projective Algebraic Sets
prime ideals of R, as the set X . This gives the notion of Let K [x] = K [x0 , x1 , . . . , xn ] be the polynomial ring gen-
an affine scheme. erated by the homogeneous coordinates. We can no longer
view these polynomials as functions since (because of the
scaling issue) even the coordinates themselves do not have
II. PROJECTIVE GEOMETRY well-defined values. However, suppose that a polynomial
F(x) is homogeneous; that is, every term of F has the
A. Infinity same degree d. Then F(λx) = λd F(x) for any nonzero
λ, and hence whether F = 0 or not at a point [x] ∈ Pn is
Consider the affine line A1 , which is simply the ground
well-defined.
field K as a set. What should it mean to approach infinity
We therefore define a projective algebraic set to be the
in A1 ? Consider a ratio x/y ∈ K ; if y = 0 this is an element
set of common zeros of a collection S of homogeneous
of K , but as y approaches zero for fixed x, this element
polynomials:
will “approach infinity.” However, we cannot let y equal
zero in this ratio, since one cannot divide by zero in a Z (S) = {[x] ∈ Pn | F(x) = 0 for all F ∈ S}.
field. It makes sense then to separate the numerator and
denominator in this ratio and consider the ordered pair: Correspondingly, given a subset X ⊂ Pn , we define the
let us write [y : x] for this. Since we are thinking of this homogeneous ideal of X , I (X ), to be the ideal in K [x] gen-
ordered pair as representing a ratio, we should maintain erated by all homogeneous polynomials F which vanish
that [y : x] = [ay : ax] for any nonzero a ∈ K . The ordered at all points of X :
pair [1 : x] will represent the number x ∈ K ; the ordered
I (X ) = {homogeneous F ∈ K [x] | F| X = 0}.
pair [0 : 1] will represent a new point, at infinity. We re-
move the ordered pair [0 : 0] from the discussion, since The reader will see the possibility of developing exactly
this would represent an undefined ratio. the same type of theory, complete with a Z −I correspon-
dence, a basis theorem, and a projective version of the
Nullstellensatz, all properly interpreted with the extra at-
B. Projective Space
tention paid to the homogeneity conditions at every turn.
The construction above generalizes to n-dimensional This is in fact what happens, and projective geometry
affine space An , as follows. Let us define projective attains an algebraic foundation as solid as that of affine
space Pn to be the set of all ordered (n + 1)-tuples geometry.
P1: ZCK Revised Pages
Encyclopedia of Physical Science and Technology EN001H-20 April 20, 2001 17:14

470 Algebraic Geometry

D. Regular and Rational Functions In general, one expects that each time one adds a new
equation that must vanish, the dimension of the set of ze-
In the context of projective geometry, the polynomial ring
ros goes down by one. Therefore, in projective space Pn ,
K [x] is called the homogeneous coordinate ring. It is a
which has dimension n, one expects that the projective
graded ring, graded by degree: K [x] = ⊕d≥0 Vd where Vd
algebraic set defined by exactly n homogeneous polyno-
is the vector space of homogeneous polynomials in x of
mials F1 = F2 = · · · = Fn = 0 will have dimension zero;
degree exactly d. If X ⊂ Pn is a projective algebraic set,
this means that it should be a finite set of points.
the homogeneous ideal I (X ) is also graded: I (X ) = ⊕d Id ,
Bezout’s theorem deals with the number of such in-
and the graded quotient ring K [X ] = K [x]/I (X ) is called
tersections. It may happen that a point of intersection is
the homogeneous coordinate ring of X .
counted with multiplicity; this is the same phenomenon
As we have noted above, a polynomial (even a homo-
as when a polynomial in one variable has a double root: it
geneous one) F does not have well-defined values at
counts for two when the number of roots is being enumer-
points of projective space. However, a ratio of polyno-
ated. In more than one variable, there is a corresponding
mials r = F/G will have a well-defined value at a point
notion of multiplicity; this is always an integer at least one
p if both F and G are homogeneous of the same degree,
(for a common isolated root).
and G( p) = 0. Such a ratio is called a rational function of
degree zero, and these functions form the foundation of
Theorem 3 (Bezout’s theorem). Suppose the
the function theory on projective algebraic sets. A rational
ground field K is algebraically closed. Let Fi , i = 1, . . . ,
function whose denominator does not vanish at p is said
n, be homogeneous polynomials in the n + 1 homogeneous
to be regular at p.
variables of projective n-space Pn . Suppose that Fi has
degree di . Then,
E. Homogenization
Let us suppose now that we have an affine algebraic set, (a) The common zero locus X = Z ({F1 , . . . , Fn }) is
given by the vanishing of a polynomial f (x1 , . . . , xn ). We nonempty.
want to investigate how this algebraic set looks at infinity, (b) If X is finite, the cardinality of X is at most the
in projective space. Let d be the largest degree of terms product of the degrees d1 d2 · · · dn .
of f . To each term of f , we multiply in the appropriate (c) If each point of X is counted according to its
power of the new variable x0 to make the term have de- multiplicity, the sum of the multiplicities is exactly
gree exactly d. This produces a homogeneous polynomial the product of the degrees d1 d2 · · · dn .
F(x0 , x1 , . . . , xn ), which for x0 = 1 has the same zeros as
the original polynomial f . However, now for x0 = 0 we For example, three quadrics (each of degree two) in P3
have new zeros, at infinity, in projective space. always intersect. If the intersection is finite, it is at most
For a general affine algebraic set X defined by more than eight points; if the points are counted according to their
one equation, we do this process for every polynomial in multiplicity, the sum is exactly eight.
I (X ), producing a collection of homogeneous polynomi- As another example, a line and a cubic in the plane will
als S. Then Z (S) ⊂ Pn is the projective closure X̄ of X intersect three times, counted with multiplicity. This is the
and exactly adds the minimal set of points at infinity to X basis for the famous “group law” on a plane projective cu-
to produce a projective algebraic set. bic curve X : given two points p and q on X , the line joining
For example, let us consider the hyperbola in the affine p and q will intersect the curve X in a third point, and this
plane defined by f (x, y) = x 2 − y 2 − 1 = 0. We homog- operation forms the basis for a group structure on X .
enize this to F(z, x, y) = x 2 − y 2 − z 2 , a homogeneous
polynomial of degree two. For z = 0, we recover the
two points [0 : 1 : 1] and [0 : 1 : −1] at infinity (where III. CURVES AND SURFACES
x 2 − y 2 = 0), up to scalar factors.
A. Singularities
F. Bezout’s Theorem
Let X be an irreducible algebraic variety, of codimension
One of the consequences of adding in the points at infinity r , in An . If xi are local affine variables at a point p ∈ X ,
to form projective space is that algebraic sets which did not and the ideal of X is generated near p by polynomials
intersect in affine space now tend to intersect, at infinity, f 1 , . . . , f k , then the Jacobian matrix J = (∂ f i /∂ x j ) will
in projective space. The classic example is that of two have rank at most r at every point of X and will have
parallel lines in the affine plane: these intersect at one maximal rank r at all points of X away from a subalge-
point at infinity in the projective plane. braic set Sing(X ). At points of X − Sing(X ), the variety
P1: ZCK Revised Pages
Encyclopedia of Physical Science and Technology EN001H-20 April 20, 2001 17:14

Algebraic Geometry 471

is said to be nonsingular, or smooth; at points of Sing(X ), projective plane. Plücker’s formula addresses the question
the variety is said to be singular, and Sing(X ) is called the of the genus of this curve in relation to the degree of the
singular locus of X . Over the complex numbers, the non- polynomial F:
singular points are those points of X where X is a complex
manifold, locally analytically isomorphic to an open set in Theorem 4 (Plücker’s formula). Suppose X is a
Cd , where d is the dimension of X . A common situation smooth projective plane curve defined by an irreducible
is when Sing(X ) is empty; then the variety is said to be polynomial F(x, y, z) of degree d. Then the genus of X is
smooth. At smooth points of algebraic varieties, there are equal to (d − 1)(d − 2)/2.
local “analytic” coordinates y1 , . . . , yd equal in number
to the dimension of the variety. Plücker’s formula has been generalized to curves with
singularities, in particular with simple singularities like
nodes (locally like x y = 0 or y 2 = x 2 ) and cusps (locally
B. 1-Forms
like y 2 = x 3 ). Then the formula gives the genus of the
A 1-form on a smooth algebraic  variety is a collection of desingularization of the curve: if a plane projective curve
local expressions of the form i f i (y) dyi , where {yi } are of degree d has a finite number of singularities which are
local coordinates on the variety and the f i ’s are rational all either nodes or cusps, and there are ν nodes and κ cusps,
functions; this collection of expressions must be all equal then the genus of the desingularization is
under changes of coordinates, and at every point of the (d − 1)(d − 2)
variety at least one of the expressions must be valid. For a g= − ν − κ.
2
curve, where there is only one local coordinate y, a local
1-form expression has the simple form f (y) dy. The set
of 1-forms on a variety form a vector space. E. Elliptic and Hyperelliptic Curves
Using Plücker’s formula, we see that smooth plane pro-
C. The Genus of a Curve jective curves of degree one or two have genus zero. We
have in fact seen above that conics may be parametrized by
Let X be a smooth projective curve. Let ω be a 1-form on lines, and this is possible precisely because they have the
X . If at every point of X there is a local expression for ω same genus. Smooth projective plane cubic curves have
of the form f (y) dy where y is a local coordinate and f is genus one, and they cannot be parametrized by lines. Ev-
a regular function, then we say that ω is a regular 1-form. ery smooth projective plane curve over C has exactly nine
The set of regular 1-forms on a smooth projective curve inflection points, and if an inflection point is put at the
form a finite-dimensional vector space over K , and the projective point [0 : 1 : 0], and if the line of inflection is
number of linearly independent regular 1-forms, which is the line at infinity, the affine equation of the cubic curve
the dimension of this vector space, is the most important can be brought into Weierstrass form
invariant of the curve, called the genus.
If K is the complex numbers, the genus has a topological y 2 = x 3 + Ax + B
interpretation also. A smooth projective curve is a compact for complex numbers A and B, such that 4A3 + 27B 2 = 0
complex manifold of dimension one, which is therefore a (this is the smoothness condition). The form of this equa-
compact orientable real manifold of dimension two. As tion shows that a cubic curve is also representable as a
such, it is topologically homeomorphic to a sphere with double cover of the x-line: for every x-value there are two
g handles attached; this g is the topological genus, which y-values each giving points of the curve.
is equal to the genus. If g = 0, the curve is topologically a In general, there are curves of every genus representable
sphere; if g = 1, the curve is topologically a torus; if g ≥ 2, as double covers, using similar affine equations of the form
the curve is topologically a g-holed torus.
y 2 = f d (x),
The simplest smooth projective curve is the projective
line itself, P1 . It has genus zero. where f d is a polynomial with distinct roots of degree
d. Such a curve is called hyperelliptic and has genus
g = [(d − 1)/2]. In particular these constructions show the
D. Plane Curves and Plücker’s Formula
existence of curves of any genus g ≥ 0.
The most common projective curves studied over the Higher-degree coverings of the projective line, such as
centuries are the plane curves, which are defined by a sin- a curve given by an equation of the form y 3 = f (x), are
gle irreducible polynomial f (x, y) = 0 in affine 2-space, also important in the classification of curves. Coverings of
and then closed up with points at infinity to a projective degree three are called trigonal curves, tetragonal curves
curve defined by the homogenization F(x, y, z) = 0 in the are covers of degree four, and so forth.
P1: ZCK Revised Pages
Encyclopedia of Physical Science and Technology EN001H-20 April 20, 2001 17:14

472 Algebraic Geometry

F. Rational Functions, Forms, jective line P1 ; (b) all curves of genus one are isomor-
and the Riemann-Roch Theorem phic to smooth plane cubic curves; (c) all curves of genus
The most celebrated theorem in the theory of curves is two are hyperelliptic, given by an equation of the form
the theorem of Riemann-Roch, which gives precise infor- y 2 = f 6 (x), where f 6 is a sextic polynomial; (d) all curves
mation about the rational functions on a smooth projective of genus three are either hyperelliptic or smooth plane
curve X . The space of rational functions K (X ) on X forms quartic curves; (e) all curves of genus four are either hy-
a field of transcendence degree one over K , and as such perelliptic or the intersection of a quadric surface and a
is an infinite-dimensional vector space over K . In order cubic surface in P3 ; and (f) all curves of genus five are
to organize these functions, we concentrate our attention hyperelliptic, trigonal, or the intersection of three quadric
on the zeros (where the numerator vanishes) and the poles threefolds in P4 . Of special interest is the so-called canon-
(where the denominator vanishes). Specifically, given a ical embedding of a smooth curve, which shows that a
point p ∈ X and a positive integer m, we may look at all curve of genus g either is hyperelliptic or can be embed-
the rational functions on X with a zero at p, no zero or ded as a curve of degree 2g − 2 in Pg−1 .
pole at p, or a pole at p of order at most m. We denote this
space by L(mp). If m = − n is a negative integer, we de- G. The Moduli Space Mg
fine L(mp) = L(−np) to be the space of rational functions
Much of the research in algebraic geometry since 1960 has
with a zero at p of order at least n.
focused on the study of the moduli spaces for algebraic
A divisor on X is a function from the points of X to
varieties. In general, a moduli space is a topological space
the group of integers Z, such that all but finitely many
M whose points represent particular geometric objects;
values are zero. For any divisor D, we define the vector
the topology on M is such that points which are close in
space L(D) to be the space L(D) = ∩x∈X L(D(x) · x). In
M represent geometric objects which are also “close” to
plain English this is the space of rational functions with
each other. Of particular interest has been the moduli space
restricted poles (to the set of points with positive D-values)
Mg for smooth projective curves of genus g. For exam-
and prescribed zeros (as the set of points with negative D-
ple, since every curve of genus zero is isomorphic to the
values). Part of the Riemann-Roch theorem is that these
projective line, M0 consists of a single point. The moduli
spaces are finite-dimensional.
space M1 classifying curves of genus one is isomorphic
We can make the same construction with rational
to the affine line A1 : the famous j-invariant of elliptic
1-forms also. Let us recall that a rational 1-form has local
curves classifies curves of genus one by a single num-
expression f (y) dy and use the function part f to define
ber. [For a plane cubic curve of genus one in Weierstrass
the zeros and poles. Given a divisor E, we may then con-
form y 2 = x 3 + Ax + B, j = 4A3 /(4A3 + 27B 2 ).] For
sider the space 1 (E) of rational 1-forms with restricted
g ≥ 2, Mg is itself an algebraic variety of dimension
poles and prescribed zeros as E indicates. Again, this is a
3g − 3.
finite-dimensional space of forms.
Of particular interest has been the construction and
study of meaningful compactifications of various moduli
Theorem 5 (Riemann-Roch). Let X be a smooth
spaces. For Mg , the most natural compactification Mg
projective curve of genus g. Let D be a divisor on
 X , and was constructed by Deligne and Mumford and the ad-
denote by d the sum of the values of D: d = x D(x).
ditional points represent stable curves of genus g, which
Then,
are curves without continuous families of automorphisms,
dim(L(D)) = d + 1 − g + dim(1 (−D)). and having only nodes as singularities. Even today the con-
struction and elementary properties of moduli spaces for
The inequality dim(L(D)) ≥ d + 1 − g was proved by higher-dimensional algebraic varieties (e.g., surfaces) is
Riemann; Roch supplied the “correction term” related to a challenge. More recently attention has turned to mod-
1-forms. One of the main uses of the Riemann-Roch the- uli spaces for maps between algebraic varieties, and it is
orem is to guarantee the existence of rational functions an area of very active research today to compactify and
with prescribed zeros and poles, in order to intelligently understand such spaces of maps.
embed the curve in projective space: given rational func-
tions f 0 , . . . , f n on a curve X , the mapping sending x ∈ X
H. Surfaces and Higher Dimensions
to the point [ f 0 (x), f 1 (x), . . . , f n (x)] will always be well-
defined on all of X and under certain circumstances will The construction and understanding of the moduli spaces
embed X as a subvariety of Pn . Mg for smooth curves is tantamount to the successful clas-
Using this technique, we can show for example that sification of curves and their properties. The classification
(a) all curves of genus zero are isomorphic to the pro- of higher-dimensional varieties is not anywhere near as
P1: ZCK Revised Pages
Encyclopedia of Physical Science and Technology EN001H-20 April 20, 2001 17:14

Algebraic Geometry 473

complete. Even surfaces, for which a fairly satisfactory widespread use of general schemes, sheaves, and homo-
classification exists due to Enriques, Kodaira, and others, logical algebra, as well as slightly more general notions
presents many open problems. of algebraic spaces and stacks.
The Enriques classification of smooth surfaces essen- After this period of partial introspection, a vigorous
tially breaks up all surfaces into four categories. The first return to application areas in algebraic geometry began in
category consists of those surfaces with a family of genus the 1980s. We will close this article with brief discussions
zero curves on them. Since genus zero curves are all iso- of a sampling of these.
morphic to lines, such surfaces are known as ruled sur-
faces, and a detailed understanding of them is possible. A. Enumeration
The prototype is a product surface X × P1 for a curve X .
One of the fundamental questions of geometry, and of
The second category consists of surfaces with a nowhere-
many other subjects in mathematics, is the “how many”
vanishing regular 2-form, or finite quotients of such sur-
question: in algebraic geometry, this expresses itself in
faces. These are the so-called abelian surfaces, K 3 sur-
counting the number of geometric objects with a given
faces, Enriques surfaces, and hyperelliptic surfaces. The
property. In contrast to pure combinatorics, often the geo-
third category consists of surfaces with a family of genus
metric approach involves counting objects with multiplic-
one curves on them. Some techniques similar to those
ity (e.g., roots of a polynomial).
used in the study of ruled surfaces are possible, and since
Typical enumerative questions (and answers) ask for the
genus one curves are very well understood, again a rather
number of flexes on a smooth cubic curve (9); the number
detailed description of these surfaces, called elliptic sur-
of lines on a smooth cubic surface (27); the number of
faces, is available. The last category is the surfaces of
conics tangent to five given conics (3264); and the number
general type, and most surfaces are in this category. Mod-
of lines on a smooth quintic threefold (2875).
uli spaces have been constructed, and many elementary
Recent breakthrough developments in intersection the-
invariants are known, but there is still a lot of work to do
ory have enabled algebraic geometers to compute such
to understand general-type surfaces. An exciting current
numbers of enumerative interest which have stumped prior
application area has to do with the connection of algebraic
generations. New excitement has been brought by totally
surfaces (which over the complex numbers are real four-
unexpected relationships with string theory in theoretical
dimensional objects) and the study and classification of
physics, where the work of Witten and others has found
4-manifolds.
surprising connections between computations relating el-
For varieties of dimension three or more, there are
ementary particles and generating functions for enumera-
essentially no areas of complete classification. Basic
tive problems of the above sort in algebraic geometry.
techniques available for curves and surfaces begin to
break down for threefolds, in particular because several
B. Computation
fundamental constructions lead immediately to singular
varieties, which are much more difficult to handle. How- Computation with polynomials is a fundamentally algo-
ever, since the 1980s, starting with the groundbreaking rithmic process which permits in many cases the design of
work of Mori, steady progress has been made on funda- explicit algorithms for calculating a multitude of quantities
mental classification constructions. of interest to algebraic geometers. Usually these quantities
either are related to enumerative questions such as those
mentioned above or are the dimensions of vector spaces
IV. APPLICATIONS (typically of functions or forms) arising naturally either in
elementary settings or as cohomology spaces.
The origins of algebraic geometry, and the development With the advent of sufficient computing power as
of projective geometry in the Renaissance, were driven in represented by modern computers and computer algebra
large part by applications (of the theory of multivariable software packages, it has become more possible to actually
polynomials) to a variety of problems in geography, art, perform these types of computations by means of com-
number theory, and so forth. Later, in the 1800s, newer puter. This has led to an explosion of activity in designing
problems in differential equations, fluid flows, and the efficient and effective algorithms for the computations of
study of integrals were driving development in algebraic interest. These algorithms usually rely in some way on the
geometry. In the 1900s the invention of algebra as we know theory of Gröbner bases, which build on a multivariable
it today caused a rethinking in the foundations of alge- version of the division algorithm for polynomials.
braic geometry, and the energies of working researchers Software is now widely available to execute many cal-
were somewhat siphoned into the development of new culations, and it is typically user-customizable, which
structures and techniques: these have culminated in the has enabled researchers around the world to make
P1: ZCK Revised Pages
Encyclopedia of Physical Science and Technology EN001H-20 April 20, 2001 17:14

474 Algebraic Geometry

fundamental contributions. Software packages which are These three properties tend to act against one another, and
widely used include Macaulay, CoCoA, and Schubert. the theory of error-correcting codes is directed toward the
classification and analysis of possible codes and coding
schemes.
C. Mechanics
Algebraic geometry has found an application in this
Many problems in mechanical engineering and construc- area, by taking the field K to be a finite field, and taking
tion involve the precise understanding of the position of the code to be certain natural spaces of functions or forms
various machine parts when the machine is in motion. on an algebraic variety. Most successful have been at-
Robotic analysis is especially concerned with the position tempts to use algebraic curves; this was initiated by Goppa
and velocity of robot extremities given various parame- and has been successful in producing codes with desirable
ters for joint motion and extension. There are a few basic properties and in aiding in the classification and uniform
motions for machine joints, and these are all describable treatment of several families of previously known codes.
by simple polynomials [e.g., circular motion of radius r
occurs on the circle (x 2 + y 2 = r 2 )]. It is not difficult to see
E. Automatic Theorem Proving
therefore that in suitable coordinate systems, virtually all
such problems can be formulated by polynomial systems. It is often the case, especially in elementary geometry, that
However, mechanical devices with many joints can cor- geometric statements can be expressed by having some
respond to algebraic sets in affine spaces with many vari- polynomial vanish. For example, three points (x1 , y1 ),
ables and many equations, which makes the geometric (x2 , y2 ), and (x3 , y3 ) in the plane are collinear if and only
analysis rather complicated. It is therefore useful to be if the determinant of the matrix
able to apply more sophisticated techniques of the theory  
1 1 1
to reduce the complexity of the problem, and this is where  
the power of algebraic geometry can come into play. x1 x2 x3  ,
A specific example of the type of problem is the follow- y1 y2 y3
ing so-called n-bar configuration. Let us consider a cycli- which is the polynomial x2 y3 − x3 y2 − x1 y3 + x3 y1 +
cal arrangement of n stiff rods, joined at the ends by joints x1 y2 − x2 y1 , is zero.
that can actuate only in a planar way locally. For what ini- A typical theorem therefore might be viewed as a
tial positions of this arrangement does the configuration collection of hypothesis polynomials h 1 , . . . , h k , and a
actually move and flex? In how many different ways can conclusion polynomial g. The truth of the theorem would
it be made to flex? When n = 3, the configuration must lie be equivalent to saying that wherever the hypothesis poly-
in a plane, and no flexing or motion is possible; for n = 4, nomials all vanish, the conclusion polynomial is also zero.
the flexing configurations, although known, have not been This exactly says that g ∈ I (Z ({h 1 , . . . , h k })), using the
completely analyzed. language of the Z −I correspondence of affine algebraic
geometry.
D. Coding Theory If the field is algebraically closed, the Nullstellensatz
says that the conclusion polynomial g is in this ideal if
A code is a collection of vectors in K n for some n, where K and only if some power of g is a linear combination of the
is a field; each vector in the collection is a code word. The h i ’s; that is, there is some equation of the form
Hamming distance between two code words is the number 
of positions where the code words differ. If each code gr = fi h i ,
word is intended to represent some piece of irreducible i

information, then desirable properties of a code are as where f i are other polynomials. A “proof” of the theorem
follows: would be given by an explicit collection of polynomials
f i , which exhibited the equation above for some r , and this
(a) Size: There should be many code words so that a can be easily checked by a computer. Indeed, algorithms
large amount of information can be represented. exist which check for the existence of such an expression
(b) Distinctness: The Hamming distance between any for g and are therefore effective in determining the truth
two code words should be large, so that if a code of a given proposed theorem.
word is corrupted in some way, the original code
word can be recovered by changing the corrupted
F. Interpolation
code word in a small number of positions.
(c) Efficiency: The ambient dimension n, which is the A general problem in approximation theory is to con-
number of positions, should be as small as possible. struct easy-to-evaluate functions with specified behavior.
P1: ZCK Revised Pages
Encyclopedia of Physical Science and Technology EN001H-20 April 20, 2001 17:14

Algebraic Geometry 475

Typically this behavior involves the values of the func- has lately become standard fare for computer graphics
tion at prescribed loci, the values of the derivatives of the specialists and novices alike.
functions, or both. Polynomial interpolation is the method Although initially the application of the ideas of
whereby polynomial functions are used in this manner, and projective geometry were rather elementary, more sophis-
algebraic geometry has found recent applications and new ticated techniques are now being brought to bear, espe-
problems of interest in this field in recent decades. cially those involving problems in computer vision and
Lagrange interpolation involves finding polynomial pattern recognition. In particular it now seems feasible
functions with specified values ck at a specified set of to use subtle projective invariants to discriminate between
points pk ∈ An . In one variable, there is a relatively easy scene objects of either different types at the same perspec-
formula for writing down the desired polynomial, and this tive or the same type at different perspectives.
is taught in most first-year calculus courses. In higher di-
mensions, no such formula exists. However, the problem is
a linear problem, and so it is a straightforward application SEE ALSO THE FOLLOWING ARTICLES
of linear algebra techniques.
Hermite interpolation involves finding polynomial ABSTRACT ALGEBRA • BOOLEAN ALGEBRA • CALCULUS
functions with specified derivative values. This is a signif-
icantly more complicated generalization, and open ques- BIBLIOGRAPHY
tions exist even for polynomials in two variables.
Spline interpolation involves finding piecewise poly- Beauville, A. (1996). “Complex Algebraic Surfaces,” 2nd ed., Cambridge
nomial functions, which stitch together to have a certain Univ. Press, Cambridge, UK.
degree of global smoothness, but which also have speci- Cox, D., Little, J., and O’Shea, D. (1997a). “Ideals, Varieties, and
fied behavior at given points. Cubic splines are the most Algorithms: An Introduction to Computational Algebraic Geometry
and Commutative Algebra,” 2nd ed., Springer-Verlag, New York.
popular in one variable, and with the advent of computer Cox, D., Little, J., and O’Shea, D. (1997b). “Using Algebraic Geometry,”
graphics, two- and three-variable splines are becoming Graduate Texts in Mathematics, Vol. 185, Springer-Verlag, New York.
more widely known and used in elementary applications. Eisenbud, D. (1996). “Commutative Algebra with a View toward Alge-
In all these settings, techniques of algebraic geometry braic Geometry,” Graduate Texts in Mathematics, Vol. 150, Springer-
are brought to bear in order to determine the dimension Verlag, New York.
Fulton, W. (1997). “Intersection Theory,” 2nd ed., Springer-Verlag, New
of the space of suitable interpolating functions. Usually it York.
is a simple matter to find a lower bound for such dimen- Griffiths, P., and Harris, J. (1978). “Principles of Algebraic Geometry,”
sions, and the more difficult problem is to find sharp upper Wiley (Interscience), New York.
bounds. Harris, J. (1992). “Algebraic Geometry: A First Course,” Springer-Verlag,
New York.
Harris, J., and Morrison, I. (1998). “Moduli of Curves,” Springer-Verlag,
G. Graphics New York.
Hartshorne, R. (1979). “Algebraic Geometry,” Springer-Verlag, New
The origins of projective geometry lie in the early York.
Renaissance, when artists and architects began to be inter- Kirwan, F. (1992). “Complex Algebraic Curves,” Cambridge Univ. Press,
ested in accurately portraying objects in perspective. The Combridge, UK.
Miranda, R. (1995). “Algebraic Curves and Riemann Surfaces,” Am.
correct treatment of horizon lines and vanishing points in Math. Soc., Providence.
artwork of the period led directly to a new appreciation of Reid, M. (1988). “Undergraduate Algebraic Geometry,” Cambridge Univ.
points at infinity and eventually to a mathematically sound Press, Combridge, UK.
theory of projective geometry. Shafarevich, I. (1994). “Basic Algebraic Geometry,” 2 vols., Springer-
With the recent explosion of capabilities in computer Verlag, New York.
Sturmfels, B. (1996). “Grobner Bases and Convex Polytopes,” Am. Math.
speed, resolution of displays, and amount and interest of Soc., Providence.
visualizing data, an appreciation of projective geometry, Ueno, K. (1999). “Algebraic Geometry I: From Algebraic Varieties to
homogeneous coordinates, and projective transformations Schemes,” Am. Math. Soc., Providence.