ALGEBRAIC CURVES
A 2003 Math Circle session by Renzo
INTRODUCTION, MOTIVATION AND GOALS
We are going to make a rst few exploratory steps in the fascinating but
mysterious world of algebraic geometry. Algebraic geometry is a branch of
mathematics that studies geometric problems with the help of algebra (like
the more geometrical oriented p eople would say) or algebraic problems with
the help of geometry (like the algebraic p eople would say). Whether one
opinion is more correct than the other is a minor p oint. What's really co ol
is that there is such a close connection b etween these apparently unrelated
elds of mathematics.
We are going to lo ok in particular in the world of curves, which have
the big advantage of b eing \drawable" and hence a little more intuitive and
less scary. Algebraic curves have b een studied for a very long time. The
ancient Greeks were already pretty handy with them and had gured out
a fair amount of things ab out curves that allowed them to solve famous
geometrical problems (such as the trisection of the angle or the doubling of
the cub e, but this is another story...). Curves are still an extremely interesting
ob ject of study nowadays. Even though a LOT of things are known ab out
them, there are many questions that are still op en. For example we are
still trying to understand the geometry of various mo duli spaces of curves
(which you can think of as \spaces of curves with given characteristics". For
example, all circles in the plane are identi ed uniquely by the co ordinates
of the center and by the radius. Hence we can think that there is a whole
three dimensional space of circles: you give me three numb ers, and I will
pro duce you a corresp onding circle; you give me a circle, I'll tell you the
three corresp onding numb ers).
Along the way we'll have a chance to intro duce a \new world", where
there are no such things as parallel lines. Again, as bizzarre as this may seem,
it's an extremely handy things to do. And it's no funky mo dern invention:
the theory of pro jective geometry was initially develop ed by architects and
painters during the Renaissance, and heavily used for prosp ectic studies and
all sort of applications.
Have I convinced you that we are embarking in a worthwile, even though
mayb e a little mind quenching adventure? Let's go then!!
DEFINITION: An Algebraic Curve consists of the set of p oints in
the plane that satisfy a p olynomial equation.
2
C = f(x; y ) 2 R jP (x; y ) = 0g
The Degree of the curve is just what we exp ect it to b e: the degree of the
p olynomial de ning the curve. EXAMPLES:
a line many lines a parabola 2
ax+by+c=0 (what is the equation?) y=x
Q1: Given two curves C and C with equations f = 0,f = 0, what is
1 2 1 2
the equation for the curve C [ C ?
1 2
DEFINITION: A curve C is said Irreducible if it do es not break up
into the union of \simpler" curves.
Q2: How do es this translate in terms of the equation for C ?
Q3: Is the equation for an algebraic curve unique?
PROBLEM...Our \dictionary" do esn't work so well...We want a go o d
corresp ondence b etween algebra (p olynomial equations) and geometry (curves).
A dictionary that uses the words banana, bro ccoli and balalaika to mean the
exact same thing is not all that go o d, is it?
FIX 1:We are using the WRONG numb ers. From now on our numb ers
will b e all the complex numb ers and not just the real ones! (Alas, we will
have to continue to draw curves in the same way b ecause we yet haven't
discovered 4-dimensional chalk!)
Notice, though, that now curves will b e \ ob jects of complex dimension
1", which means they really are surfaces. Now mayb e you rememb er that we
know surfaces pretty well. We could even work out that all complex algebraic
curves are orientable, and hence must lo ok like a sphere or a multiholed
doughnut (eventually pinched or punctured),
but that's another story....
Q4: What happ ens now to
2 2
x + y = 0
2 2
x + y = 1?
Q5: Is the x a complete x?
If you p onder a little bit you will nd out that the answer is \almost".
In fact we are still de ning the same curve if we multiply the equation by a
constant...but that's not such a big deal, and we'll see later on that we'll b e
able to x this to o by just slightly mo difying our \universe".
For now we have estabilished the following bijective corresp ondence:
al g ebr aic cur v es eq uations up to scal ing
$
NOW FOR SOME MORE FUN EXAMPLES:
Cubics: What are the graphs (of the real trace) of:
2 3
y = x x;
2 3
y = x ;
2 3 2
y = x + x .
Pretty Flowers: Match the graphs with their equations:
2 2 3 2 2
(x + y ) 4x y = 0;
2 2 2
(x + y ) + 4xy = 0;
2 2 2 2 3
(x + y ) 3x y y = 0;
More pretty owers...: Can you pro duce a sensible guess for the equation
of the following gure?
Now that we are starting to get a little more familiar with Algebraic
Curves, let's start asking ourselvs interesting questions, and let's start easy
to then get harder and harder:
1. How many p oints of intersection can 2 lines in the plane have?
2. How many p oints of intersection do a line and a conic have in the plane?
3. How many p oints of intersection do es a line have with a curve C if
degree d?
4. How many lines through one p oint?
5. how many lines through two p oints?
6. How many p oints do you need to pin down a unique conic?
7. Will any con guration of the right numb er of p oints pin down a unique
conic?
8. What's a quick way to determine the equation of the conic through
p
1 1
P = ( 1; 0), P = (0; 0), P = (0; 1), P = ( 1; 1), P = ( 2 ; )?
1 2 3 4 5
2 2
9. How many p oints pin down a unique curve of degree d?
2
THE \IDEAL" UNIVERSE : P
C
Take the usual plane and \add" to it a new (pro jective) line, that we will
2
call line at in nity. Points of this line represent directions of lines in C .
point at infinity representing this direction
Notice that now every pair of lines meet at exactly one p oint!!!!
2
FORMALLY : We add an extra co ordinate, so that p oints in P will b e
C
identi ed by three complex numb ers
(a : b : c):
BUT (there's a big but!) we declare two p oints to b e the same p oint if the
co ordinates are prop ortional. So, for example
(1 : 2 : 3) = (2 : 4 : 6) = (23 : 46 : 69) = (k : 2k : 3k ):
Now we can see that
2 2
P = f(a : b : c); c 6= 0g [ f(a : b : 0)g = C [ l :
1
C
What happ ens to algebraic geometry now? We could a priori just take
2
p olynomials in three variables (for example P (x; y ; z ) : x + y + 3z + 4 = 0).
But obviously that's completely nuts!!
Q6: Why?
In general I have no hop e to make sense of the notion of the \value" of
2
a p olynomial at a p oint of P . But there are some sp ecial p olynomials such
C
that at least the zero-set is well de ned.
Q7: Can you guess which are they?
DEFINITION: A Homogoneous p olynomial is a p olynomial in which
all the monomials have the same degree.
Q8: Gimme some examples.
Q9: Why are homogeneous p olynomials the \go o d ones" for our pur-
p oses?
DEFINITION: A Pro jective Curve is the zero set of a homogeneous
p olynomial. For example:
2 2 2
X + 2XY + Y + 3Z = 0:
Q9: Can we go back and forth b etween our previous notion of curves and
this current one?
The answer is homogenize and dehomogenize !!!!
For example:
2 2 2 2 2
x + y 1 = 0 $ X + Y Z = 0;
2 2
y x = 0 $ Y Z X = 0:
Q10: Are homogeneization and dehomogeneization always inverse op er-
ations to each other?
And now for some more problems and interesting stu to discover!
1. Find algebraically the intersection p oints in pro jective space of the
following pair of lines:
y = x;
y = x + 1.
2. Find algebraically the intersection p oints in pro jective space of the
following line and hyp erb ola:
0 = x;
y x = 1.
2
3. Make sense of the statement: \All conics in P are p otato es".
C
2
4. P gives us a real easy way to decide whether a (non pro jective) conic
C
is a parab ola, ellipse or hyp erb ola. What is it?
5. Decide which of the following conics is a parab ola, ellipse or hyp erb ola!
2 2
x + 2xy + 1 + y + 2y = 0;
2 2
x + 3xy + 2y + 2y + 2x + 6 = 0;
2 2
x + xy + y + 4y + 3x 12 = 0.
A TINY BIT OF INTERSECTION THEORY
We want to de ne a notion of intersection of curves that resp ects the
following ideas:
1. If two curves intersect transversally we count exactly the numb er of
p oints of intersections.
1 1 1
1 1 1
Total intersection = 6
2. If one curve go es through a multiple p oint of the other curve, but
do esn't share a tangent with any \branch", then the p oint of intersec-
tion counts for exactly the multeplicity of the multiple p oint.
2
1
3. If two curves intersect at a p oint and share the tangent line, then the
intersection multeplicity go es up!! ( We could make this notion precise,
but we won't, it would take just to o long!)
>1
Q11: We actually can nd a way to make p oint 3 precise if one of the curves
is just a line. Can you think of how?
2 y=x4 y=x y=x3
EXERCISE: Find jC \ C j in the following cases:
1 2
C 1 C 1
C 2
C 2
C 1 C 2
C 2 C1
Now here comes our big guy. The following theorem is quite straightfor-
ward to prove in the case one of the curves is a line (think ab out why...), but
requires some sophistication to b e proven in full glory in the following form:
BEZOUT'S THEOREM: If C and C are two pro jective curves of
1 2
degree d and d then
1 2
jC \ C j = d d :
1 2 1 2
With the help of Bezout we are nally ready to tackle a bunch of really
interesting questions and problems and to prove the conjectures we started
out from:
1. If a conic has a double p oint, then it splits into 2 lines.
2. A conic cannot have a triple p oint.
3. If a cubic has a triple p oint, then it's the union of three lines.
4. A cubic with two double p oints must b e the union of a line and a conic.
5. A cubic with three distinct double p oints must b e...?
6. An irreducible quartic cannot have 4 double p oints.
7. Pascal's Theorem: The opp osite sides of a hexagon inscrib ed in a
conic meet in collinear p oints.
8. Pappus' theorem: Given two lines l ; l , and chosen randomly three
1 2
p oints on each line (call them P ; P ; P ; Q ; Q ; Q ), then the p oints
1 2 3 1 2 3
P Q \ P Q ; R =
2 3 3 2 1
R = P Q \ P Q ;
2 1 3 3 1
P Q \ P Q R =
1 2 2 1 3
are collinear.