Solutions to written works 2, 3

and Final Take-Home Exam

Ali Sinan Sertöz – MATH 591 Algebraic Geometry – Spring 2016

sertoz@[Link]

The following exercises are from Hartshorne’s Algebraic Geometry

I.1.6
Any nonempty open subset of an irreducible topological space is dense and
irreducible. If Y is a subset of a topological space X, which is irreducible
in its induced topology, then the closure Ȳ is also irreducible.

Let X be an irreducible topological space and Y a nonempty open subset.

If Y = X then Y is irreducible and dense as claimed. If however Y X,
then X \ Y is a nonempty closed proper subset of X and we can write
X = Y ∪ (X \ Y ). Since X is irreducible we must have X = Y , i.e. Y is
dense in X. Assume Y = Y1 ∪ Y2 where Yi are closed subsets of Y . There
are then closed subsets X1 , X2 of X such that Yi = Y ∩ Xi . If Yi Y , then
Xi X and X = X1 ∪ X2 renders X reducible which is not the case. Hence
either Y1 or Y2 must be equal to Y and Y is irreducible. (This proves the
first part of the problem which is Example 1.1.3.)

Next assume that Y is not irreducible when Y is irreducible in its induced

topology. Then Y = Y1 ∪ Y2 where Yi are proper closed subsets of Y , and
we have Y = (Y ∩ Y1 ) ∪ (Y ∩ Y2 ). Since Y is irreducible one of these subsets
is not proper, say Y = Y ∩ Y1 . Then Y ⊂ Y1 and Y ⊂ Y1 since Y1 is closed
but this contradicts the choice of Y1 as a proper subset of Y . Hence Y is
irreducible. (This second part of the problem is Example 1.1.4.)

I.1.8
Let Y be an affine variety of dimension r in An . Let H be a hypersurface
in An , and assume that Y 6⊆ H. Then every irreducible component of Y ∩H
has dimension r − 1. (See (7.1) for a generalization.)

Let Y be the zero set of the prime ideal p ⊂ k[x1 , . . . , xn ]. Let f ∈ k[x1 , . . . , xn ]
be an irreducible polynomial such that H = Z(f ). Set B = k[x1 , . . . , xn ]/p.
Let f¯ denote the residue of f in B. Then f¯ = f¯1 · · · f¯s where each fi is
a polynomial in k[x1 , . . . , xn ] and f̄i is irreducible in B. We have the de-

1
composition (f ) + p = [(f1 ) + p] · · · [(fs ) + p] and each irreducible compo-
nent Yi of Y ∩ H is of the form Z((fi ) + p). The coordinate ring of Yi is
k[x1 , . . . , xn ]/[(f ) + p] and is isomorphic to B/(f̄i ). None of the f̄i is a unit
or a zero divisor in B, and being irreducible generates a prime ideal. Hence
height(f̄i ) = 1 in B, by Theorem I.1.11A. Now invoking Theorem I.1.8A we
have height(f̄i ) + dim B/(f̄i ) = dim B. It follows that dim B/(f̄i ) = r − 1
which is the dimension of the irreducible component Yi of Y ∩ H, i = 1 . . . , s.

I.1.9
Let a ⊆ A = k[x1 , . . . , xn ] be an ideal which can be generated by r elements.
Then every irreducible component of Z(a) has dimension ≥ n − r.

We will prove this by induction. When r = 1, a is generated by a single

polynomial f = g1 · · · gs where each gi is irreducible in A. An irreducible
component of Z(a) is of the form Z(gi ) and is of dimension n − 1 by Propo-
sition I.1.13.

Now assume that the statement is true for r − 1. Let a be generated by

f1 , . . . , fr−1 , and f . An irreducible component X of Z(a) is of the form
Y ∩ Z(g) where Y is an irreducible component of Z(f1 , . . . , fr−1 ) and g is
an irreducible factor of f . Then by the previous exercise (I.1.8), dim X =
dim Y −1. Moreover dim Y ≥ n−(r−1) by the induction hypothesis. Putting
these together, we find that dim X ≥ n − r as claimed. This completes the
proof by induction.

I.1.11
Let Y ⊆ A3 be a curve given parametrically by x = t3 , y = t4 , z = t5 . Show
that I(Y ) is a prime ideal of height 2 in k[x, y, z] which cannot be generated
by 2 elements. We say Y is not a local complete intersection –cf. (Ex.
2.17).

Since t 7→ (t3 , t4 , t5 ) is a homeomorphism of A1 with Y , we have that dim Y =

1. Let B = k[x, y, z] and p = I(Y ). Then dim Y = dim B/p = 1. From the
identity (see Theorem 1.8A) height p + dim B/p = dim B, we get height p =
2, since clearly dim B = 3. By trial and error we find that the polynomials
of smallest degree in p are all linear combinations of xz − y 2 , yz − x3 and
z 2 − x2 y. Since these three polynomials are linearly independent, no two of
them can generate p. Hence I(Y ) cannot be generated by two elements even
though its height is two.

2
I.1.12
Give an example of an irreducible polynomial f ∈ R[x, y], whose zero set
Z(f ) in A2R is not irreducible (cf. 1.4.2).

The simplest case is when f = x2 + y 2 + 1 whose zero set is empty in A2R and
empty set is not considered irreducible.

A more convincing example can be constructed as follows. Take two real

numbers a and b, not both zero. Then consider the polynomial f = (x2 −
a2 )2 + (y 2 − b2 )2 = [(x2 − a2 ) + i(y 2 − b2 )][(x2 − a2 ) − i(y 2 − b2 )]. It is
straightforward to show that these factors are irreducible in C[x, y]. Since the
factorization in C[x, y] is unique, f cannot be factored into real components
and is thus f irreducible in R[x, y]. The zero set of f now has two components
if ab = 0 and has four components otherwise.

I.2.12
The d-Uple Embedding. For given n, d > 0, let M0 , . . . , MN be all the
mono-
mials of degree d in the n + 1 variables x0 , . . . , xn , where N = n+d n − 1. We
define a mapping ρd : Pn → PN by sending the point P = (a0 , . . . , an ) to
the point ρd (P ) = (M0 (a), . . . , MN (a)) obtained by substituting the ai in the
monomials Mj . This is called the d-uple embedding of Pn in PN . For exam-
ple, if n = 1, d = 2, then N = 2, and the image Y of the 2-uple embedding
of P1 in P2 is a conic.
(a) Let θ : k[y0 , . . . , yN ] → k[x0 , . . . , xn ] be the homomorphism defined by
sending yi to Mi , and let a be the kernel of θ. Then a is a homogeneous
prime ideal, and so Z(a) is a projective variety in PN .
(b) Show that the image of ρd is exactly Z(a). (One inclusion is easy. The
other will require some calculation.)
(c) Now show that ρd is a homeomorphism of Pn onto the projective variety
Z(a).
(d) Show that the twisted cubic curve in P3 (Ex. 2.9) is equal to the 3-uple
embedding of P1 in P3 , for suitable choice of coordinates.

(a)
Being the kernel of a homomorphism, a is prime. We need to show that
it is homogeneous. Let f = f0 + · · · + fm ∈ a, where each fi is homo-
geneous of degree i in the variables y0 , . . . , yN , and m = deg f . We have
f (M0 , . . . , MN ) = f0 (M0 , . . . , MN ) + · · · + fm (M0 , . . . , MN ) = 0 identically.
But each fi (M0 , . . . , MN ) is a homogeneous polynomial of degree id in the
variables x0 , . . . , xn . Since they all have different degrees, they cannot cancel
each other and hence they must be all identically zero. This gives fi ∈ a, for

3
i = 0, . . . , m, showing that a is a homogeneous ideal.

(b)
For this problem we fix an ordering of the monomials of degree d in the
variables x0 , . . . , xn . The map ρd is defined with this ordering. In particular
if
ρd (x) = ρd ([x0 : · · · : xn ]) = [· · · : xji11 · · · xjirr : · · · ] = y,
j1 jr
then the homogeneous coordinate ofy corresponding to the place of xi1 · · · xir
j · · · jr
will be denoted by y 1 . Here of course the i1 , . . . , ir , j1 , . . . , jr
i1 · · · ir
are integers with 0 ≤ i1 < · · · < ir ≤ n, and j1 , . . . , jr ≥ 0 with j1 + · · · + jr =
d. We will refer to such a collection of indices as admissible.

We want to show that ρd (Pn ) = Z(a).

If y = ρd (x) for some x ∈ Pn , then the entries of y are monomials of degree

d in the variables x0 , . . . , xn , hence any polynomials f ∈ a which vanishes
on these monomials will vanish at y. This shows that y ∈ Z(a), and the
inclusion ρd (Pn ) ⊂ Z(a) is easily established.
 
j1 · · · jr
Now assume that we have a point a = [· · · : a : · · · ] ∈ Z(a).
i1 · · · ir
We will show that there exists a point c = [c0 : · · · : cn ] ∈ Pn such that
 
j1 · · · jr
a = cji11 · · · cjirr ,
i1 · · · ir
for all admissible indices.

First observe that for any collection of admissible indices, we have

d  j j
d r
 
j1 · · · jr d 1
y − y ··· y ∈ a.
i1 · · · ir i1 ir
Since a ∈ Z(a), we must have
d  j j
d r
 
j1 · · · jr d 1
a = a ··· a ,
i1 · · · ir i1 ir
so not all of    
d d
a ,..., a
0 n

4
can be zero. After reordering the monomials if necessary and multiplying by
a suitable caontant we can assume that
 
d
a = 1.
0
We now pick a point c = [c0 : · · · : cn ] ∈ Pn as follows.
 
d−1 1
c0 = 1, ci = a , for i = 1, . . . , n.
0 i
We will show that
a = ρd (c),
which will show that Z(a) ⊂ ρd (Pn ) and will finish the proof.

We will show that  

j ··· jr
a 1 = cji11 · · · cjirr ,
i1 · · · ir
keeping in mind that c0 = 1.

First observe that since

  j−1  j
d−j j d d−1 1
y y − y ∈ a,
0 i 0 0 i
we must have    j
d−j j d−1 1
a = a = cji .
0 i 0 i
Moreover, in general we have
r−j1 −1  j j
d−1 1 r
  
j1 j2 · · · jr d d−1 1 2
y y − y ··· y ∈ a,
0 i2 · · · ir 0 0 i2 0 ir
and
r−1  j j
d−1 1 r
  
j ··· jr d d−1 1 1
y 1 y − y ··· y ∈ a.
i1 · · · ir 0 0 i1 0 ir
It then follows that
r−j1 −1 j j
d−1 1 r
   
j1 j2 · · · jr d d−1 1 2
a a = a ··· a
0 i2 · · · ir 0 0 i2 0 ir
= cji22 · · · cjirr ,

5
and
r−1 j j
d−1 1 r
   
j ··· jr d d−1 1 1
a 1 a = a ··· a
i1 · · · ir 0 0 i1 0 ir
= cji11 · · · cjirr .

This completes the proof.

(c)
We first show that ρd is one-to-one. Let c = [c0 : · · · : cn ] and e = [e0 : · · · : en ]
are points of Pn such that ρd (c) = ρd (e). At least one of the homogeneous
coordinates ei is different than zero. For convenience of notation assume
e0 6= 0. Since the images of c and e agree under ρd , we must have cd0 = ed0
and cd−1 d−1 d d
0 ci = e0 ei for i = 1, . . . , n. From c0 = e0 , we get c0 = ωe0 , where
d
ω = 1. From the other equations we get ci = ωei for i = 1, . . . , n. This
shows that c = e and hence ρd is one-to-one.

Next we show that ρd is continuous. In fact if Z(f1 , . . . , fm ) is a closed subset

of Z(a), where f1 , . . . , fm are homogeneous polynomials in y0 , . . . , yN , then
ρ−1
d (Z(f1 , . . . , fm )) = Z(f1 ◦ ρd , . . . , fm ◦ ρd ) is closed. Thus ρd is continuous.

We now have a continuous isomorphism from ρd (Pn ) onto Z(a). It is well

known that a continuous isomorphism is not necessarily a homeomorphism.
Therefore we have to check separately if ρd is a homeomorphism in this case.

What remains to be shown for ρd to be a homeomorphism is that it maps

closed sets to closed sets. Since every ideal in k[x0 , . . . , xn ] is finitely gen-
erated, every closed set is an intersection of finitely many hypersurfaces.
It therefore suffices to show that a hypersurface is mapped under ρd to a
closed set in Z(a). And for this we need to show that for any homogeneous
polynomial g ∈ k[x0 , . . . , xn ], there corresponds a homogeneous polynomial
G ∈ k[y0 , . . . , yN ] such that ρd (Z(g)) = Z(a) ∩ Z(G). We now describe a way
of obtaining G from g.

Let g ∈ k[x0 , . . . , xn ] be a homogeneous polynomial of degree m. Let

wt = xi00t xi11t · · · xinnt , t = 1, . . . , d.

be homogeneous monomials, not necessarily distinct, occurring in g with non-

zero coefficients, where i0t , . . . , int are non-negative integers with i0t + · · · +

6
int = m for t = 1, . . . , d. A typical monomial occurring in the polynomial g d
is of the form

w = w1 · · · wd = xi01 +···+i0d · · · xin1 +···+ind .

For s = 0, . . . , n, let
is1 + · · · + isd = us d + vs ,
where 0 ≤ us and 0 ≤ vs < d. Then we can write

w = (xd0 )u0 · · · (xdn )un (xv00 · · · xvnn ).

Since the degree of the monomial w is dm, we see that v0 + · · · + vn = `d,

where ` ≥ 0 is a non-negative integer. Since each vs < d, we have that
0 ≤ ` ≤ n. Therefore there exist integers 0 ≤ j0 < · · · < j` ≤ n such that we
can write
vje = vj0 e + vj00e , e = 0, . . . , `
in such a way that

v0 + · · · + vn = (v0 + · · · + vj0 −1 + vj0 0 ) + (vj000 + vj0 +1 + · · · + vj1 −1 + vj0 1 )+

· · · + (vj00` + vj` +1 + · · · + vn )

in such a way that each parenthesis adds up to d. Then

v0 vj00
xv00 · · · xvnn = (xv00 · · · xj0j0 ) · · · (xj` ` · · · xvnn ).

This proves that the monomial w of degree dm can be written as a product

of monomials of degree d in the variables x0 , . . . , xn . In fact let M0 , . . . , MN
be a list of monomials of degree d in x0 , . . . , xn , and let φ be the map

φ : k[M0 , . . . , MN ] → k[y0 , . . . , yN ],

sending each Mi to yi , then φ(w) is a monomial of degree m in the variables

y0 , . . . , yN . Define the polynomial G as

G(y0 , . . . , yN ) = φ(g d (x0 , . . . , xn ) ∈ k[y0 , . . . , yN ],

where the above process of writing g d as a polynomial in Mi is understood,

before φ is applied. Then G is uniquely defined and is homogeneous of degree
m.

7
It is now clear from the description of the maps that for any point a = [a0 :
· · · : an ] ∈ P n ,
G(ρd (a)) = φ(g d (a)).
It follows from this that a ∈ Z(g) if and only if ρd (a) ∈ Z(G) ∩ Z(a). This
shows that the closed set Z(g) is mapped onto the closed set Z(G) ∩ Z(a),
which completes the proof that ρd is a closed map.

Thus ρd is a homeomorphism from Pn onto Z(a)

(d)
Let [s : t] and [x0 : x1 : x2 : x3 ] be the homogeneous coordinates in P1 and P3
respectively. Let ρ3 ([s : t]) = [s3 : s2 t : s2 t2 : st3 : t3 ], and φ : U ⊂ P1 → A1
be given as usual by φ([s : t]) = t/s. Then Y = ρ3 ◦ φ−1 (t) = [1 : t : t2 : t3 ]
is an embedding of A1 into P2 such that Ȳ is the twisted cubic. Note that
Y also coincides with ρ3 (U ). The only point in P1 not in U is [0 : 1] which
maps under ρ3 to [0 : 0 : 0 : 1]. Since ρ3 is an isomorphism onto its image,
the closure of ρ3 (U ) contains only this extra point. Hence Ȳ is ρ3 (P1 ).

I.2.14
The Segre Embedding. Let ψ : Pr × Ps → PN be the map defined by
sending the ordered pair (a0 , . . . , ar ) × (b0 , . . . , bs ) to (. . . , ai bj , . . . ) in lexi-
cographic order, where N = rs + r + s. Note that ψ is well-defined and
injective. It is called the Segre embedding. Show that the image of ψ is
a subvariety of PN . [Hint: Let the homogeneous coordinates of PN be
{zij |i = 0, . . . , r, j = 0, . . . , s}, and let a be the kernel of the homomorphism
k[{zij }] → k[x0 , . . . , xr , y0 , . . . , ys ] which sends zij to xi yj . Then show that
Im ψ = Z(a).]

That ψ is well defined is immediate since if ai 6= 0 and bj 6= 0, then

zij = xi yj 6= 0. Moreover if ψ(a0 , . . . , ar ) × (b0 , . . . , bs ) = (. . . , zij , . . . ),
then ψ(λa0 , . . . , λar ) × (µb0 , . . . , µbs ) = (. . . , λµai bj , . . . ). Hence ψ gives the
same point in PN independent of which representative is used for the point
a × b ∈ Pr × Ps .

To show injectivity, assume ψ(a × b) = ψ(a0 × b0 ). Then for some i, j, we

have ai 6= 0 and bj 6= 0. Since zij = zij0 , we have ai bj = a0i b0j , hence a0i 6= 0 and
b0j 6= 0. Without loss of generality assume then that ai = bj = a0i = b0j = 1.
0
We have zuj = au and zuj = a0u for u = 0, . . . , r. Hence, a = a0 . Similarly,
0
exploiting the fact that ziv = ziv for v = 0, . . . , s gives b = b0 . Thus ψ is
injective.

8
Now we show that Im ψ = Z(a).

Let p = (. . . , pij , . . . ) ∈ Im ψ. Since pij = xi yj for some x = (x0 , . . . , xr ) ∈ Pr

and some y = (y0 , . . . , ys ) ∈ Ps , for every f ∈ a, we have f (p) = 0. Hence
p ∈ Z(a).

Conversely, assume that p = (. . . , pij , . . . ) ∈ Z(a). Without loss of generality

assume that p00 = 1. Since the polynomial z00 zuv − zu0 z0v is in a, we have
puv = pu0 p0v for all u = 0, . . . , r and v = 0, . . . , s. Let a = (1, p10 , . . . , pr0 ) ∈
Pr and b = (1, p01 , . . . , p0s ) ∈ Ps . Then p = ψ(a × b), hence p ∈ Im ψ.

Thus we get Im ψ = Z(a).

Remark: We now prove that a is generated by all polynomials of the form

zi1 j1 zi2 j2 − zi1 j2 zi2 j1 , where i1 , i2 run through 0 to r, and j1 , j2 run through 0
to s. Denote this ideal by J. We will show that J = a. Clearly J ⊆ a.

We want to show that a ⊆ J. For this, pick any homogeneous polynomial f

in a. Let M = zi1 j1 · · · zin jn be a homogeneous form of degree n appearing in
f with some non-zero constant coefficient.

Since f vanishes on every point of Im Ψ, there must be another homogeneous

form M 0 appearing in f with the same coefficient multiplied by −1 such that
if M 0 = zu1 v1 · · · zun vn , then {i1 , . . . , in } = {u1 , . . . , un } and {j1 , . . . , jn } =
{v1 , . . . , vn }.

We know claim that M − M 0 is in J. We do this by induction on the degree

n.

If n = 2, the statement is trivial.

Assume n > 2. Both M and M 0 evaluated on Im ψ will be equal to

xi1 · · · xin yj1 · · · yjn .

In particular xun and yvn are in this list, corresponding to zun vn term of M 0 .
So M has at least two terms of the form zun j , zivn for some i and j. Without
loss of generality assume that they are zin−1 jn−1 and zin jn . i.e. assume that
un = in−1 and vn = jn .

9
Then we can write
zu v
n n
0
M −M = (zi1 j1 · · · zin−2 jn−2 )[zin−1 jn−1 zin jn − zin jn−1 zin−1 jn ]
z }| {


− zi1 j1 · · · zin−2 jn−2 zin jn−1 − zu1 v1 · · · zun−1 vn−1 zun vn .

Now check that the term inside the parenthesis on the second line is ho-
mogeneous of degree n − 1 and satisfies the conditions about xi and yj as
M − M 0 satisfies. By induction hypothesis, it is inside J. It is now clear that
M − M 0 ∈ J. This shows that a ⊆ J and concludes the proof that they are
equal.

I.2.16
(a) The intersection of two varieties need not be a variety. For example,
let Q1 and Q2 be the quadric surfaces in P3 given by the equations
x2 − yw = 0 and xy − zw = 0, respectively. Show that Q1 ∩ Q2 is the
union of a twisted cubic and a line.
(b) Even if the intersection of two varieties is a variety, the ideal of the
intersection may not be the sum of the ideals. For example, let C be
the conic in P2 given by the equation x2 − yz = 0. Let L be the line
given by y = 0. Show that C ∩ L consists of one point P , but that
I(C) + I(L) 6= I(P ).

(a)
Let [w : x : y : z] be the homogeneous coordinates in P3 .

If x = 0 then either y = 0 or w = 0. In the subcase y = 0, we get the

points [1 : 0 : 0 : 0] and [0 : 0 : 0 : 1]. In the subcase w = 0, we get the line
[0 : 0 : y : z] where [y : z] ∈ P1 .

If x 6= 0, then wyz 6= 0. Without loss of generality we assume that w = 1

and get the solution space [1 : x : x2 : x3 ] where x ∈ k, which is the twisted
cubic in P3 together with the point [0 : 0 : 0 : 1].

(b)
The intersection of C and L is easily seen to be the point P = [0 : 0 : 1] if
[x : y : z] denotes the homogeneous coordinates of P2 . It is also clear that
I(P ) = (x, y). But I(C) + I(L) = (x2 − yz, y) = (x2 , y) (x, y) = I(P ).

10
 I.2.17
Complete Intersections. A variety Y of dimension r in Pn is (strict) com-
plete intersection if I(Y ) can be generated by n − r elements. Y is a
set-theoretic complete intersection if Y can be written as the intersection
of n − r hypersurfaces.
(a) Let Y be a variety in Pn , let Y = Z(a); and suppose that a can be
generated by q elements. Then show that dim Y ≥ n − q.
(b) Show that a strict complete intersection is a set-theoretic complete
intersection.
∗(c) The converse of (b) is false. For example let Y be the twisted cubic
curve in P3 (EX. 2.9). Show that I(Y ) cannot be generated by two
elements. On the other hand, find hypersurfaces H1 and H2 of degrees
2, 3 respectively, such that Y = H1 ∩ H2 .
∗∗(d) It is an unsolved problem whether every closed irreducible curve in P3
is a set-theoretic intersection of two surfaces. See Hartshorne [1] and
Hartshorne [5, III. §5] for commentary.

(a)
We will prove this by induction on q. When q = 1, this is the content of Ex.
I.2.8.

Assume true for q − 1. And let fq be an irreducible polynomial not in

(f1 , . . . , fq−1 ). Let X = Z(f1 , . . . , fq−1 ) and Y = Z(f1 , . . . , fq ), and let
S(X) = k[x0 , . . . , xn ]/(f1 , . . . , fq−1 ) and S(Y ) = k[x0 , . . . , xn ]/(f1 , . . . , fq )
be the corresponding projective coordinate rings.

From Ex. I.2.6 we know that dim S(X) = dim X +1 and dim S(Y ) = dim Y +
1. By induction hypothesis, we have dim S(X) = dim X +1 ≥ [n−(q−1)]+1.
We will show that dim S(Y ) ≥ [n − q] + 1.

Let p be the prime ideal in S(X) generated by fq . Note that we have

S(Y ) = S(X)/p.

By Theorem I.8A.b applied to the right hand side of this equation, we have

dim S(Y ) = dim S(X) − height p.

By Theorem I.1.11A, height p = 1. Putting in this with the induction hy-

pothesis that dim S(X) ≥ [n − (q − 1)] + 1, we get

dim S(Y ) ≥ n − q + 1,

11
as required.

(b)
Let Y be a strict complete intersection in Pn of dimension r. Then I(Y ) =
(f1 , . . . , fn−r ) where each fi is a homogeneous polynomial in the variables
x0 , . . . , xn . Let Hi = Z(fi ) be the hypersurface defined by fi , for i =
1, . . . , n − r. We then have

Y ⊂ H1 ∩ · · · ∩ Hn−r .

Conversely, let p be a point of the intersection of the hypersurfaces. Then

fi (p) = 0 for all i = 1, . . . , n − r. This says that p ∈ Y . Hence Y is the set
theoretic intersection of the hypersurfaces H1 , . . . , Hn−r .

(c)
Let Y ∈ P3 be the twisted cubic parametrized by

[s : t] 7→ [s3 : s2 t : st2 : t3 ] ∈ P3 ,

where [s : t] ∈ P1 . Using [x0 : x1 : x2 : x3 ] ∈ P3 as the homogeneous

coordinates, set

f1 = x0 x2 − x21 , f2 = x1 x3 − x22 , f3 = x0 x3 − x1 x2 .

It is easy to show that

I(Y ) = (f1 , f2 , f3 ).
Moreover let
φ = x3 f 3 − x2 f 2 .
It is also straightforward to see that

Y = Z(f1 ) ∩ Z(φ).

But
I(Y ) ! I(f1 , φ)
since f2 is not an element of I(f1 , φ). (Check however that f22 = x2 φ − x23 f1
and so belongs to I(f1 , φ).)

(d)
This problem is still unsolved at the time of this writing, 2016.

12
I.3.8
Let Hi and Hj be hyperplanes in Pn defined by xi = 0 and xj = 0, with
i 6= j. Show that any regular function on Pn − (Hi ∩ Hj ) is constant. (This
gives an alternate proof of (3.4a) in the case Y = Pn .)

Let φ be a regular function on Pn − (Hi ∩ Hj ). Restricting φ to Pn − Hi

we get a regular function there. Bur regular functions on Pn − Hi are of
f
the form m where f is a homogeneous polynomial in x0 , . . . , xn of degree
xi
m ≥ 0, and xi 6 |f . Similarly restricting φ to Pn − Hj we get a regular
g
function of the form r where g is homogeneous of degree r ≥ 0 and xj 6 |g.
xj
f g
On the intersection (Pn − Hi ) ∩ (Pn − Hj ) we must have m = r . This gives
xi xj
r m
f xj = gxi . But since i 6= j, xj does not divide the right hand side. This
forces r = 0. Similarly m = 0. So f = g are homogeneous of degree 0, i.e.
they are constants. On the other hand, φ being a constant on an open set is
constant throughout.

I.3.15
Products of Affine Varieties. Let X ⊆ An and Y ⊆ Am be affine varieties.
(a) Show that X × Y ⊆ An+m with its induced topology is irreducible.
[Hint: Suppose that X × Y is a union of two closed subsets Z1 ∪ Z2 . Let
Xi = {x ∈ X x × Y ⊆ Zi }, i = 1, 2. Show that X = X1 ∪ X2 and X1 , X2 are
closed. Then X = X1 or X2 so X × Y = Z1 or Z2 . ] The affine variety
X × Y is called the product of X and Y . Note that its topology is in
general not equal to the product topology (Ex. 1.4).
(a) Show that A(X × Y ) ∼= A(X) ⊗k A(Y ).
(c) Show that X × Y is a product in the category of varieties, i.e., show
(i) the projections X × Y → X and X × Y → Y are morphisms, and (ii)
given a variety Z, and morphisms Z → X, Z → Y , there is a unique
morphism Z → X × Y making a commutative diagram

(c) Show that dim X × Y = dim X + dim Y .

(a)
We follow the hint. Assume first that there is an x0 ∈ X such that x0 is

13
neither in X1 nor in X2 . Then define the sets Yi = {y ∈ Y x0 × y ∈ Zi },
i = 1, 2. Now Y = Y1 ∪ Y2 , and we show that Y1 , Y2 are closed subsets of
Y . Fix i = 1, 2. Let us use the notation (x, y) = (x1 , . . . , xn , y1 , . . . , ym ) for
coordinates in An+m . Let Ji be the ideal in k[y] generated by the polynomials
of the form fi (x0 , y) for all fi (x, y) belonging to the ideal of Zi in k[x, y]. Then
the zero set of Ji is precisely Yi , so Yi is closed. But Y is irreducible, so Y = Y1
or Y2 , which implies that x0 is in X1 or in X2 , which is in contradiction to
the way x0 was chosen. So no such x0 can be chosen and X = X1 ∪ X2 .

Next we show that X1 is closed. Let J1 be the ideal in k[x] generated by

polynomials f (x, y0 ) where f (x, y) is a polynomial vanishing on Z1 and yo is
a point on Y . Then X1 is the zero set of J1 and is therefore closed. Similarly
X2 is closed. But X being irreducible either X = X1 or X = X2 . But this
implies that either X × Y = Z1 or X × Y = Z2 , showing that X × Y is
irreducible.

(b) P
Define
P a map φ : A(X) ⊗k A(Y ) → A(X × Y ) as φ( fi (x) ⊗ gi (y)) =
fi (x)gi (y). This is a ring homomorphism. It is onto since φ(xi ⊗yj ) = xi yj
and they generate the ring P A(X × Y ). Now let r be the smallest integer
r
such that there exist F = i=1 fi (x) ⊗ gi (y) with φ(F ) = 0. From the
minimality of r, we see that the gi are not in the ideal of Y , so there is
a point y0 ∈ Y such that not all P gi (y0 ) are zero. Assume without loss of
r
generality that
Pr−1 r 0g (y ) 6
= 0. Then i=1 gi (y0 )fi (x) = 0 on X, and we have
fr (x) = i=1 [gi (y0 )/gr (y0 )]fi (x). Then we get
r−1
X
F = fi ⊗ {gi (y) + [gi (y0 )/gr (y0 )]gr (y)} ,
i=1

violating the minimality of r. This contradiction shows that φ is injective

and hence is an isomorphism.

(c)
Let πX : X × Y → X be the projection on the first component. Let φ be
∗
a regular function on X. Then (πX (φ))(x, y) = (φ ◦ πX )(x, y) = φ(x) is a
regular function on X × Y , so πX is a morphism of varieties. Similarly πY :
X × Y → Y is a morphism. Let Z be a variety with morphisms pX : Z → X
and pY : Z → Y . Define φZ → X × Y as φ(z) = (pX (z), pY (z) ∈ X × Y . We
then have πX ◦ φ)(z) = πX (pX (z), pY (z)) = pX (z) and similarly πY ◦ φ = pY ,

14
 making the given diagram commutative.

This also follows from the universal property of tensor products if we consider
the corresponding maps on the coordinate rings.

(d)
This follows from the fact that the dimension of a variety is the Krull dimen-
sion of its coordinate ring as follows.

dim X × Y = dim A(X × Y ) = dim A(X) ⊗k A(Y )

= dim A(X) + dim A(Y ) = dim X + dim Y.

I.3.19
Automorphisms of An . Let φ : An → An be a morphism of An to An given
by n polynomials f1 , . . . , fn of n variables x1 , . . . , xn . Let J = det |∂fi /∂xj | be
the Jacobian polynomial of φ.
(a) If φ is an isomorphism (in which case we call φ an automorphismof
An ) show that J is a nonzero constant polynomial.
∗∗(b) The converse of (a) is an unsolved problem, even for n = 2. See, for
example Vitushkin [1].

(a)
If φ is invertible, then the Jacobian matrix is also invertible at every point
in An . Therefore the determinant J of the Jacobian matrix must be nonzero
at every point of An . But J is a polynomial over an algebraically closed field
so will have at least one root unless it is a nonzero constant. Hence J is a
nonzero constant.

(b)
This is still an unsolved problem at the time of writing, April 2016.

I.3.20
Let Y be a variety of dimension ≥ 2, and P ∈ Y be a normal point. Let f
be a regular function on Y − P .
(a) Show that f extends to a regular function on Y .
(b) Show this would be false for dim Y = 1.
See (III, Ex. 3.5) for a generalization.

First we make an observation. Let A be a commutative domain with dimen-

sion at least 2. Then we can find two distinct prime ideals p1 and p2 with
0 ⊂ p1 ⊂ p2 . WE can consider p1 as a prime ideal in Ap2 . An element of

15
(Ap2 )p1 is of the form

a/b
, where a, b, c, d ∈ A, b, d 6∈ p2 , c 6∈ p1 .
c/d
But this can be written as
a/b ad
= , where now we have bc 6∈ p1 .
c/d bc
This shows that every element of (Ap2 )p1 can be considered as an element of
Ap1 . Conversely, every element of Ap1 is of the form
a
, where a, b ∈ A, b 6∈ p1 .
b
Since 1 6∈ p2 , we can rewrite a/b as (a/1)/(b/1) and then it can be considered
as an element of (Ap2 )p1 . Hence the two rings can be considered the same,

(Ap2 )p1 = Ap1 ,

which we will use below.

Now let U be any open neighborhood of P in Y , and let O(U ) be the ring of
regular functions on U . Let OP be the local ring of regular functions at P ,
with maximal ideal m. Let p be any prime ideal in OP with 0 ⊂ p ⊂ m. Such
prime ideals exist since dimension of Y is ≥ 2. There exist a prime ideal p0
and a maximal ideal m0 in O(U ) such that p = p0 OP and m = m0 OP . Since
OP = O(U )m0 , the above observation gives us the identity

(OP )p = (O(U )m0 )p = O(U )p0 .

Since P is a normal point, the ring OP is a noetherian normal domain. Then

we have \
OP = (OP )p .
p is prime
height p=1

This is Theorem 11.5 on page 81 of Matsumura’s Commutative Ring Theory.

Combining this with the previous identity we have
\
OP = O(U )p0 .
p0 is prime and p0 ⊂m0
height p0 =1

16
Now for any such p0 choose a point Q on Z(p0 ) other than P . Since f is regular
at Q, there exists polynomials g and h such that h(Q) 6= 0 and f = g/h in
some neighborhood of Q. But h 6∈ p0 since otherwise it would vanish at Q
contrary to our choice of h. This however says that f = g/h ∈ O(U )p0 . Since
f belongs to every such ring on the right hand side of the above equality, it
must belong to OP , which now means that it is regular at P .

If the dimension of Y is 1, then no prime ideal p as above exists so the

argument breaks. In fact the conclusion also fails in dimension 1 as the
example of f = 1/x on A1 − 0 shows.

I.7.1 – Final Take-Home Exam

(a) Find the degree of the d-uple embedding of Pn in PN (Ex. 2.12).
[Answer: dn ]
(b) Find the degree of the Segre embedding of Pr × Ps in PN (Ex. 2.14).
r+s


[Answer: r ]

(a)
Let φd : Pn → PN be the d-uple embedding. Here N = n+d


n
− 1. Let Pn,d (z)
n
be the Hilbert polynomial of φd (P ). If z is an integer, then a monomial of
degree z in the variables y0 , . . . , yN of PN pulls back via φd to a monomial of
degree zd in the variables x0 , . . . , xn of Pn . But the number of such monomials
is known, so
n−1
dn
 
n + zd 1 Y
Pn,d (z) = = (zd + n − `) = z n + lower degree terms in z.
n n! `=0 n!

Hence the degree of the d-uple embedding of Pn is dn .

Note that Pn,d (0) = 0, and as we will see in the next exercise, the arithmetic
genus of φd (Pn ) is zero.

As a special case consider the d-uple embedding of P1 into Pd+1 . Its Hilbert
polynomial is dz + 1. In particular the Hilbert polynomial of the twisted
cubic in P3 is 3z + 1.

(b)
Let ψr,s : Pr × Ps → PN be the Segre embedding. Here N = rs + r + s. Let
Σr,s = ψr,s (Pr × Ps ). Let Qr,s (z) be the Hilbert polynomial of Σr,s . If z is an
integer, then a monomial of degree z in the variables y0 , . . . , yN of PN pulls

17
back via ψr,s to a monomials of degree z on Pr and a monomial of degree z
on Ps . But the number of such monomials is known, so
  
z+r z+s
Qr,s (z) =
r s
(z + 1)(z + 2) · · · (z + r) (z + 1)(z + 2) · · · (z + s)
= ·
r! s!
1 r+s
= z + ··· + 1
r!s!
(r+s)!
= r!s!
z r+s + · · · + 1
(r + s)!
r+s


= r
z r+s + · · · + 1.
(r + s)!
r+s


This shows that the degree of Σr,s is r
.

I.7.2 – Final Take-Home Exam

Let Y be a variety of dimension r in Pn , with Hilbert polynomial PY . We
define the arithmetic genus of Y to be pa (Y ) = (−1)r (PY (0) − 1). This is
an important invariant which (as we will see later in (III, Ex. 5.3)) is
independent of the projective embedding of Y .
(a) Show that pa (Pn ) = 0.
(b) If Y is a plane curve of degree d, show that pa (Y ) = 12 (d − 1)(d − 2).
n
(c) More
generally, if H is a hypersurface of degree d in P , then pa (H) =
d−1
n .
(d) If Y is a complete intersection (Ex. 2.17) of surfaces of degrees a, b in
P3 , then pa (Y ) = 12 ab(a + b − 4) + 1.
(e) Let Y r ⊆ Pn , Z s ⊆ Pm be projective varieties, and embedd Y × Z ⊆
Pn × Pm → PN by the Segre embedding. Show that

pa (X × Z) = pa (Y )pa (Z) + (−1)s pa (Y ) + (−1)r pa (Z).

(a)  
n `+n 1
The Hilbert polynomial of P is PPn (`) = = [` + 1) · · · (` + n)
n n!
which counts the number of distinct monomials of degree ` in the variables
x0 , . . . , xn . Since PPn (0) = 1, the arithmetic genus is 0.

18
(b)
This is a special case of (c) which we solve below.

(c)
In the proof of Proposition 7.6, page 52, we saw that
   
z+n z−d+n
PH (z) = − .
n n

For the arithmetic genus we need to calculate PH (0).

   
n n−d
PH (0) = −
n n
1
= 1 − (−d + 1)(−d + 2) · · · (−d + n)
n!
(−1)n
= 1− (d − 1)(d − 2) · · · (d − n)
n!
(d − 1) · · · (d − n)(d − n − 1)!
= 1 − (−1)n
n!(d − n − 1)!
 
d−1
= 1 − (1−)n .
n

(−1)n
 
d−1
Hence the arithmetic genus is = (d − 1)(d − 2) · · · (d − n).
n n!

Going back to the special case of the (b) part above, putting in n = 2 gives
the required formula for a plane curve of degree d.

(d)
Let Ya and Yb are hypersurfaces in P3 of degrees a and b respectively with
Ya = Z(f ) and Yb = Z(g) where f and g are homogeneous polynomials of
degrees a and b. Assume that Y = Ya ∩ Yb is a complete intersection, i.e. the
two hypersurfaces Ya and Yb have no common components.

From the proof of Proposition 7.6, on page 52, we know the Hilbert polyno-
mials of Ya and Yb ; for example
   
z+n z−a+n
PYa (z) = − .
n n

19
We now consider the short exact sequence of graded S-modules
g
0 −→ (S/f )) (−b) −→ S/f −→ S/(f, g) −→ 0,

where S is the graded polynomial ring k[x0 , . . . , xn ], the first map is multi-
plication by g, and the second map is the natural quotient map. Noting that
S/f and S/(f, g) are the projective coordinate rings of Ya and Y respectively,
we have from the additivity of the Hilbert function

PY (z) = PYa (z) − PYa (z − b)

       
z+3 z−a+3 z−b+3 z−b−a+3
= − − − .
3 3 3 3

For the arithmetic genus we need to put z = 0.

1
PY (0) = 3! − (−1)3 [(a − 1)(a − 2)(a − 3) − (b − 1)(b − 2)(b − 3)
3!
+(a + b − 1)(a + b − 2)(a + b − 3)])
1 1
= 2ab − a2 b − ab2 .
2 2
Hence the arithmetic genus is
1
pa (Y ) = ab(a + b − 4) + 1.
2

(e)
Let the homogeneous coordinates of Pn and Pm be given by [x0 : · · · : xn ]
and [y0 : · · · : ym ]. The Segre embedding maps Pn × Pm into PN , where
N = nm + n + m, in the following fashion
ψ
[x0 : · · · : xn ] × [y0 : · · · : ym ] 7→ [· · · : xi yj : · · · ]

where we choose the lexicographical ordering for the placement of the xi yj .

Let the homogeneous coordinates of PN be [· · · : zij : · · · ] where zij occupies
the place of xi yi of the Segre embedding.

For any positive integer r, zijr = xri yjr . It is then clear that any monomial of
degree ` in the variables zij is the unique product of a monomial of degree `
in the variables xi with a monomial of degree ` in the variables yj .

20
The converse works modula a, where a is the ideal of the image of Pn × Pm
in PN under the Segre embedding, see the solution of Ex. I.2.14. This is how
it works. Let p = xa00 · · · xann be a monomial of degree ` = a0 + · · · + an , and
similarly let q = y0b0 · · · ym
bm
be another monomial of degree ` = b0 + · · · + bm .

Now we describe an algorithm to construct a monomial M of degree ` in the

variables zij starting with the above monomials p and q.

Input p and q as above and initialize M = 1.

Let i ≥ 0 be the smallest index such that ai > 0.

Let j ≥ 0 be the smallest index such that bj > 0.
Set cij = min{ai , bj }.
c
Replace M by M × zijij .
c
Replace p by p/xi ij .
cij
Replace q by q/yj .
Repeat until p = 1 or equivalently until q = 1.

We thus obtain a monomial of degree ` in the variables zij . Let M 0 be another

monomial of degree ` in the variables zij such that substituting xi yj for zij in
M 0 gives pq. Then M −M 0 vanishes on the image of the Segre embedding and
thus is in a. Therefore they represent the same monomial in the coordinate
ring of ψ(Pn × Pm ) which is k[. . . , zij , . . . ]/a.

We conclude that the number of distinct monomials of degree ` in the projec-

tive coordinate ring of ψ(Pn ×Pm ) is the product of the number of monomials
of degree ` on Pn and Pm . This gives the identity for the Hilbert functions
Pψ(Pn ×Pm ) (z) = PPn (z) PPm (z).
Finally returning to our case, we see that the above arguments work for Y ,
Z and ψ(Y × Z). Hence we have
Pψ(Y ×Z) (z) = PY (z) PZ (z).
The above algorithm shows that Pψ(Y ×Z) (z) = PY ×Z . Hence finally we have
PY ×Z (z) = PY (z) PZ (z).
Now we are ready to prove the claim of the exercise. Note that
pa (Y × Z) = (−1)r+s (PY ×Z (0) − 1)
= (−1)r+s (PY (0)PZ (0) − 1).

21
On the other hand we have
pa (Y )pa (Z) + (−1)s pa (Y ) + (−1)r pa (Z)
= [(−1)2(PY (0) − 1)] [(−1)s (PZ (0) − 1)] + (−1)[(−1)s (PY (0) − 1)]
+ (−1)r [(−1)s (PZ (0) − 1)]
= (−1)r+s (PY (0)PZ (0) − 1)
= pa (Y × Z),

as claimed.

22