Homework 5

ALGEBRAIC GEOMETRY
HOMEWORK 5
(1) Compute the points of intersection of the circle x2 + y 2 + x − y = 0 with the

lemniscate (x2 +y 2 )2 = x2 −y 2 using resultants (can you compute the affine
points of intersections directly?), and compute the multiplicities. Explain
which coordinate system you choose, how the equations look like, what the
resultant is, and what its factors are.
1. Let’s see how far we get with naive methods. Plugging x2 +y 2 = y −x
into the second equation gives us (y − x)2 = x2 − y 2 , hence 0 = 2y 2 − 2xy =
2y(y − x).
If y = 0, we get x2 + x = 0 from the first equation, hence the only
possibilities are (0, 0) and (−1, 0). Both of these points lie on the lemniscate
and are thus points of intersection.
If y = x, we get 2x2 = 0 from the first equation, hence x = 0, and then
the second equation tells us that y 4 = y 2 , which implies y 2 (y 2 − 1) = 0.
This gives the possible points (0, 0), (0, 1) and (0, −1), of which only the
first satisfies y = x and is on the lemniscate.
Thus the only affine points of intersection are (0, 0) and (−1, 0). The
points at infinity on the circle are [1 : ±i : 0]. Since both of them are also on
the lemniscate, we have exactly four points of intersection in the projective
plane: [0 : 0 : 1], [−1 : 0 : 1], [1 : i : 0], and [1 : −i : 0]. Using this method,
it is not possible to assign multiplicities, however.
2. The most naive way of using resultants is eliminating y from the
two equations. In fact, with f (x, y) = x2 + y 2 + x − y and g(x, y) =
(x2 + y 2 )2 − x2 + y 2 we get, using pari, Rf,g = 8x3 (x + 1). This is a
polynomial of degree 4, whose roots x = 0 and x = 1 lead to the two affine
points of intersection found above. The fact that the resultant has degree
4 although Bezout predicts 2 · 4 = 8 points of intersection already tells us
that something is wrong. Of course we cannot expect to find all points of
intersection if we work with the affine equation.
Let us therefore go projective. The two equations become
F (X, Y, Z) = X 2 + Y 2 + XZ − Y Z = 0,
G(X, Y, Z) = (X 2 + Y 2 ) − X 2 Z 2 + Y 2 Z 2 = 0,
and eliminating the variable Z using resultants we get, using the pari
command polresultant(F,G,Z) (after having entered F and G, of course):
Rf,g (X, Y ) = 2Y (Y − X)(X 2 + Y 2 )2 .

Again, something is wrong because the degree of the resultant is 6, not 8
as predicted by Bezout. In any case, solving this equation readily leads to
1
2 HOMEWORK 5
the same points as above. Assigning multiplicites will not work because we
only get 6 points of intersection even then, not the 8 points we need.
3. In order to find out what went wrong let us recall the first sentence
in Chapter 10:
Assume that we are given two curves CF and CG , where F and
G are homogeneous polynomials in K[X, Y, Z] of degrees m and
n, rspectively. Assume that the point [0 : 0 : 1] is on neither of
these curves. Then we can write
F (X, Y, Z) = Z m + a1 Z m−1 + . . . + am−1 Z + am ,
G(X, Y, Z) = Z n + b1 Z n−1 + . . . + bn−1 Z + bn ,
where the ai and bj are homogeneous polynomials of degree i and
j in K[X, Y ].
The condition we have to check before applying resultants is whether the
point [0 : 0 : 1] (that is, (0, 0) in the affine plane) is on one of the curves.
In fact we find that it lies on both curves. In particular, when we write the
homogenized polynomials as polynomials in Z over K[X, Y ], then we do
not get equations of the right degree: the degrees of F and G in solution 2.
were 1 and 2, respectively, instead of 2 and 4.
Substituting X 7−→ X − Z and Y 7−→ Y − Z gives us the equations
F (X, Y, Z) = (X − Z)2 + (Y − Z)2 + (X − Z)Z − (Y − Z)Z = 0,
G(X, Y, Z) = ((X − Z)2 + (Y − Z)2 )2 − (X − Z)2 Z 2 + (Y − Z)2 Z 2 = 0,
and now [0 : 0 : 1] is not a point on one of these curves. If we rewrite
F and G as polynomials in Z, we see that F and G have degree 2 and 4,
respectively, which is what we expected.
Eliminating Z using pari then gives
RF,G (X, Y ) = 8X(X − Y )3 (X 2 + Y 2 )2 .
(pari cannot factor the polynomial for you. But you expect the factors X,
X − Y and X 2 + Y 2 from the false starts above). For finding roots we have
to factor everything into linear factors, and we find
RF,G (X, Y ) = 8X(X − Y )3 (X − Y i)2 (X + Y i)2 .
Now X + Y i = 0 leads to [1 : i : 0] (there is no difference to the
coordinates found above because Z = 0, hence X = X − Z), which has
multiplicity 2; the same goes for the root [1 : −i : 0] corresponding to the
factor X − Y i.
Next X = 0 leads to Z 2 + (Y − Z)2 − Z 2 − (Y − Z)Z = 0, i.e., 0 =
2Z 2 − 3Y Z + Y 2 = (Y − 2Z)(Y − Z). If Z = 0, then Y = 0, which does not
give any point. Thus Z = 1 and therefore Y = 1 or Y = 2, and we get the
two points [0 : 1 : 1] and [0 : 2 : 1]. Plugging them into the second equation
we see that only the first on is on the lemniscate. Thus [0 : 1 : 1] is the
point of intersection corresponding to the factor X, and it has multiplicity
1.
Finally, setting Y = X in the equation of the circle gives 2(X − Z)2 = 0,
hence X = Z, and this leads to the unique point [1 : 1 : 1] with multiplicity
3.
ALGEBRAIC GEOMETRY 3
Returning to the original coordinate system, we find the following points

of intersection with multiplicity:
P mP (F, G)
[1 : i : 0] 2
[1 : −i : 0] 2
[0 : 0 : 1] 3
[−1 : 0 : 1] 1
In particular, we find 8 points of intersection, counting multiplicities, as
predicted by Bezout.
(2) Here you will learn how to solve cubics using resultants. Assume you want
to solve the cubic equation x3 + ax2 + bx + c = 0.
(a) Show that the substitution y = x + a/3 transforms the cubic into
y 3 + By + C = 0.
typing in
x = y - a/3; x^3 + a*x^2 + b*x + c
produces the output

y^3 + (-1/3*a^2 + b)*y + (2/27*a^3 - 1/3*b*a + c)
This is a cubic without quadratic term as desired.

(b) Now assume that we have to solve x3 + bx + c = 0. Using linear re-
lations y = rx ∗ b will not make the linear term disappear (without
bringing back the quadratic term). Thus we try to put y = x2 + cx + d
and then eliminate x using resultants: typing in
p = polresultant(x^3+b*x+c,y-x^2-d*x-e,x)
polcoeff(p,2,y)
we find that the coefficient of y 2 is 2b − 3e. Thus we have to put

e = 2b/3 in order to keep the quadratic term away.
p = polresultant(x^3+b*x+c,y-x^2-d*x-2*b/3,x)
polcoeff(p,1,y)
shows that the coefficient of y is −b2 /3 + bd2 + 3cd. This is a quadratic

equation in the unknown d. Since we know how to solve this, we can
find some d that makes the linear term disappear, and then we are left
with solving a pure cubic.
Note, by the way, that the discriminant of the quadratic resolvent
bd2 + 3cd − b2 /3 is (3c)2 + 4b3 /3 = 31 (4b3 + 27c2 ), an expression that
is all over the place in classical solutions of cubics by radicals.
(3) Now solve the quartic x4 + ax3 + bx2 + cx + d = 0.
We first transform away the a with x = y − a/4, and assume that the
quartic has the form x4 + bx2 + cx + d.
p = polresultant(x^4 + b*x^2 + c*x +d , y-x^2-e*x-f,x)
polcoeff(p,3,y)
4 HOMEWORK 5
gives 2b = 4f , hence f = b/2 will keep the cubic term away.

p = polresultant(x^4 + b*x^2 + c*x +d,y-x^2-e*x-b/2,x)
polcoeff(p,2,y)
shows that y 2 has coefficient b2 /2 + be2 + 3ec + 2d. This is a quadratic

equation in e (note that b, c, d are given constants), which can be solved;
for either choice of e this gives us a quartic of the form x4 + cx + d.
Now we go on:
p = polresultant(x^4 + c*x + d,y - x^3 - e*x^2 - f*x - g,x)
polcoeff(p,3,y)
shows that g = 3c/4 makes the cubic term disappear; trying to get rid
of the linear term, however, fails because the resulting equation −3c2 /8 +
3cef + 2de2 + 4df = 0, although being linear in f , also depends on e.
Thus we cannot solve the quartic this way. Mea culpa. Actually what
we have to do is use a Tschirnhaus transformation that gets rid of the linear
term: in fact, if we can transform the quartic into a biquadratic equation
x4 + ax2 + b = 0, then we are home because we can substitute x2 = X.
And this really works:
p = polresultant(x^4 + b*x^2 + c*x +d , y-x^2-e*x-b/2,x)
polcoeff(p,1,y)
shows that the coefficient of the linear term is −c2 + (e3 − 2 ∗ b ∗ e) ∗ c +

(4 ∗ e2 ∗ d − b2 ∗ e2 ). Since b, c, d are known, this is a cubic equation in
the unknown e. Solving this equation provides us with a value of e that
transforms the quartic into a biquadratic equation, and then we are home.

Homework 5

Uploaded by

Copyright:

Available Formats

Homework 5

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Homework 5

Uploaded by

Copyright:

Available Formats

ALGEBRAIC GEOMETRY

(1) Compute the points of intersection of the circle x2 + y 2 + x − y = 0 with the

Rf,g (X, Y ) = 2Y (Y − X)(X 2 + Y 2 )2 .

Returning to the original coordinate system, we find the following points

produces the output

This is a cubic without quadratic term as desired.

we find that the coefficient of y 2 is 2b − 3e. Thus we have to put

shows that the coefficient of y is −b2 /3 + bd2 + 3cd. This is a quadratic

gives 2b = 4f , hence f = b/2 will keep the cubic term away.

shows that y 2 has coefficient b2 /2 + be2 + 3ec + 2d. This is a quadratic

shows that the coefficient of the linear term is −c2 + (e3 − 2 ∗ b ∗ e) ∗ c +

You might also like