Algebraic Geometry by Robin Hartshorne

Exercises solutions by Jinhyun Park

Chapter I Varieties
Section 1. Affine Varieties

1.1. (a) Note that A(Y) = k[x,y]/(y-x^2) i

p
Define phi : k[x] -> A(Y) = k[x,y]/(y-x^2) by the composition k[x] -> k[x,y]
-> k[x,y]/(y-x^2)
Claim: phi is injective; Let phi(f)=phi(g) for f,g in k[x]. Then, f(x)-g(x)
in (y-x^2) <=> f(x) - g(x)= h(x,y) (y-x^2) for some h in k[x,y]. But, if h
is not zero, deg_y h(x,y) (y-x^2) >=1, and deg_y LHS = 0. Hence h ==0, i.e.
f(x)=g(x)./
Claim: phi is surjective; Let h(x,y) + (y-x^2) be in A(Y). Then, if h(x,y) =
sum_{i=0 to n} f_i (x) y^i, f_i are in k[x]. Note that h(x,y)-h(x,x^2) =
sum_{i=0 to n} f_i(x) (y^i - x^2i) = sum_{i=0 to n} f_i(x) (y-x^2)(y^(i-1) +
y^(i-2) x^2 + ... + y (x^2)^(i-2) + (x^2)^(i-1)) is in (y-x^2). Hence,
h(x,y) + (y-x^2) = h(x,x^2) + (y-x^2). Let g(x)=h(x,x^2), then, phi(g) =
h(x,y) + (y-x^2)./

(b) Note that A(Z) = k[x,y]/(xy-1). Assume that phi: A(Y) -> A(Z) is an
isomorphism. Since phi is surjective, there are f,g in k[x] s.t. phi(f(x))=
x+(xy-1), phi(g(x))=y+(xy-1). => phi(f(x)g(x)) = xy + (xy-1) = 1+(xy-1) =
unity of A(Z). Since phi is an isomorphism, f(x)g(x)= unity of A(Y), i.e.
f,g are in k. Then for any h(x,y) + (xy-1) in A(Z), h(f,g) is in k, and
phi(h(f,g)) = h(x,y) +(xy-1) i.e. phi|k (k) = k[x,y]/(xy-1), but, it is a
contradiction.//

(c) Let f(x,y) in k[x,y] be an irreducible conic polynomial. Let's write

f(x,y) = ax^2 + 2bxy + cy^2 + dx + ey + g, a,b,c,d,e,g in k, and not all of
a,b,c are zero. For this f, define D(f) = b^2-ac. We prove the following
claims:
1) Whether D(f) =0 or D(f) !=0 is stable under the following operations on
f:
i) multiply by a nonzero constant u in k.
ii) translation x->x+l, y->y+m, l,m in k.
iii) linear transform (x,y)^t -> A (x,y)^t for any A in GL(2,k).
2) Any irreducible conic f(x,y) can be transformed into one of the
following two cases using only above three operations:
i) y-x^2 if D(f)=0 (parabolic)
ii) xy-1 if D(f)!=0 (elliptic)
3) For irreducible conic f(x,y), its affine coordinate ring k[x,y]/(f) is
stable (up to isomorphism) under the operations in i).

proof of 1); i) D(uf)=(bu)^2 - (au)(cu)=D(f)u^2.

ii) Let f' be the transformed conic. Note that, when
calculating D(f'), only coefficients of x^2, xy, y^2 matter. But,
translation does not change these coefficients. Hence, D(f')=D(f).
iii) linear transforms preserve degrees, so, we may assume that
f is homogeneous of degree 2. Note that f=ax^2 +2bxy + cy^2 = (x
y)([[a,b],[b,c]])(x,y)^t and D(f)= -det([[a,b],[b,c]]). So by (x,y)^t ->
A(x,y)^t, we obtain (x,y)^t A^t ([[a,b],[b,c]])A(x,y)^t, so D(f')=-detA^t
D(f) detA. But, A is in GL(2,k), so, D(f')=nonzero constant multiple of
D(f)./

proof of 2); i) D(f)=b^2-ac=0. Then, f=ax^2 + 2bxy + cy^2 + dx+ey+g =

(sqrt(a)x + sqrt(c)y)^2 + dx+ey+g (k: algebraically closed => sqrt(a),
sqrt(c), sqrt(-1) exist).
Take transform x<- sqrt(a)/sqrt(-1) x + sqrt(c)/sqrt(-1) y
y<- dx + ey + g
This transform is a composition of operation ii), iii).
Then, f-> y-x^2.
 ii) D(f)=b^2-ac!=0. Then, f(x,y)=ax^2+2bxy+cy^2+dx+ey+g =
a{(x+b/a y)^2 + c/a y^2 - b^2/a^2 y^2} + dx+ey +g = a(x+b/ay)^2 + (ac-b^2)/a
y^2 + dx+ey+g = (sqrt(a)x + b/sqrt(a) y)^2 + (sqrt((ac-b^2)/a) y)^2 +
dx+ey+g = (sqrt(a)x + (b/sqrt(a) + sqrt((ac-b^2)/a) sqrt(-1) y ) (sqrt(a)x +
(b/sqrt(a) - sqrt((ac-b^2)/a)sqrt(-1) y) + dx+ey+g.
Take transform x<- sqrt(a)x + (b/sqrt(a) + sqrt((ac-b^2)/a)sqrt(-1) y
y<- sqrt(a)x + (b/sqrt(a) - sqrt((ac-b^2)/a)sqrt(-1) y
which is operation iii).
=> f -> xy + d'x + e'y + g for some d', e' in k.
= (x+e')(y+d') + g-e'd'
Take transform x<- x+e'
y<- y+d'
which is operation ii).
=> f -> xy + g-e'd'
Since f is irreducible, g-e'd' is not zero, so there is u in k s.t
u(g-e'd')=-1.
Then, operation i): f-> (ux)y -1
operation iii) : x<-ux, y<-y
=> f -> xy-1.

proof of 3); i) multiplication by u in k^* results in,

k[x,y]/(f') = k[x,y]/(uf) = k[x,y]/(f) so stable.
ii) translation x-> x+l , y-> y+m results in,
k[x,y]/(f') = k[x,y]/(f(x+l,y+m)) -> k[x,y]/(f)
g(x,y) + (f') -> g(x-l,y-m) + (f)
is an isomorphism
iii) linear transform (x,y)^t -> A (x,y)^t results in,
k[x,y]/(f') = k[x,y]/(f(A(x,y)^t)^t) -> k[x,y]/(f)
g(x,y) + (f') -> g((A^-1 (x,y)^t ))^t) + (f)
is an isomorphism./

So, any irreducible conic f can be transformed into

y-x^2 if D(f)=0,
xy-1 if D(f)!=0.
Under this transformation, coordinate ring is stable upto isomorphism. Hence
for any irreducible conic f in k[x,y], A(W)=k[x,y]/(f) is isomorphic to
k[x]=A(Y) or k[x,y]/(y-x^2) = A(Z).//

1.2. Clearly, A(Y)=k[x,y,z]/(z-x^3, y-x^2). Note that A(Y)==

k[x,x^2,x^3]=k[x]: ID, so Y is irreducible, and Y is an affine variety. and
dim Y = dim A(Y) = dim k[x] = 1 and I(Y) = (z-x^3,y-x^2).//

1.3. Y= Z(x^2-yz, xz-x)= Z(x^2-yz, x(z-1)). Let x^2-yz=0 be (i), x(z-1)=0 be

(ii). From (ii), if x=0 => y=0 or z=0 in (i). If x!=0 => z=1 and x^2-y=0.
Hence we have three cases: {x^2-y=0 and z=1}, {x=0,y=0},{x=0,z=0}. Hence
Y=Z(x^2-y, z-1) union Z(x,y) union Z(x,z).
Note that A(Z(x^2-y,z-1))=k[x,y,z]/(x^2-y,z-1) == k[x,x^2,1]=k[x] : ID
A(Z(x,y))==k[z] : ID
A(Z(x,z))==k[y] : ID
So, Z(x^2-y, z-1), Z(x,y), Z(x,z) are all irreducible and, above expression
of Y is the irreducible decomposition.//

1.4. Consider Z(x-y) in A^2. It is closed in A^2. But, in A^1 x A^1, closed
sets are finite union or arbitrary intersection of V1 x V2, V1,V2 : closed
sets of A^1. Since V1,V2 are empty or A^1 or finite sets, V1xV2 must be
empty or finite set or {finite}xA^1, A^1x{finite} or A^1 x A^1. Z(x-y) is
not any of above form. So, A^2 is not homeomorphic to A^1 x A^1.//

1.5. (=>) Assume that B== k[x1,...,xn]/I(X) for some X. Clearly, B is then
finitely generated. Assume that f+I(X) satisfies (f+I(X))^m = 0 for some m
in N. Then, f^m is in I(X) => f^m (x) = 0 for all x in X => f is in I(X) =>
f(x)+I(X)=0. Hence, there is no nilpotent element./
(<=) Let a1,a2,...,an be generators of B is a k-algebra. Then, phi :
k[x1,...,xn] -> B mapping xi to ai is a surjection. Then,
k[x1,...,xn]/ker(phi) = B. So, we have to prove that ker(phi) is a radical
ideal. Let f be in sqrt(ker(phi)). => f^m is in ker(phi). => phi(f^m) =
(phi(f))^m = 0 but, B does not have nilpotent elements, so phi(f)=0. i.e. f
is in ker(phi). Hence ker(phi) is a radical ideal, and so, ker(phi)=I(X) for
some X. //

1.6. Let U be nonempty open in X, X:irreducible.

Claim: U is dense and irreducible.
If U is not dense, there is a nonempty open set V in X s.t. U doesnot meet
V. Then U^c union V^c is X, contradicting the irreducibility of X. Hence U
is dense.
If U is not irreducible, there is a nonempty proper closed sets W_1,W_2 in U
s.t. W_1 union W_2 = U. Since W_1=U intersection V_1, W_2=U intersection V_2
for some closed sets V_1,V_2 of X. Then, (V_1 union V_2) union U^c =X
contradicting the irreducibility of X. Hence is U is irreducible.

Claim: When Y is irreducible, so is Y~.

Lemma: If every nonempty open set U of Z is dense, then Z is irreducible.
pf) If not, V_1 U V_2 = Z, V_1,V_2;nonempty proper closed. Then, (V_1)^c
intersection (V_2)^c = empty, contradicting denseness of (V_1)^c./
Hence ETS: Any nonempty open set U in Y~ is dense. Let V be any nonempty
open set of Y~. Then, U intersection Y, V intersection Y are nonempty open
sets of Y(because Y is dense in Y~). Hence intersection of all U,V,Y =
nonempty. => intersection of U,V is = nonempty.//

1.7. (a) (i) => (ii) Let S be a nonempty collection of closed subsets. On
elements of S, give a partial order '<=' as F1<=F2 if F1 contains F2. Let
{F_i} be a chain in S. Since X is a noetherian space, {F_i} is actually a
finite set, so there is a maximal element. Hence by Zorn's lemma, there is a
maximal element in S with respect to '<=', i.e. a minimal element in S with
respect to the inclusion./

(ii) => (i) Let F1 \contain F2 \contain F3 .... be a descending chain of

closed subsets of X. Then, S= {Fi| i in N} has a minimal element by (ii),
Let F_N be the minimal element. Then, of course, this chain is stationary
beyond Nth./

(iii)=>(iv) and (iv)=>(iii) can be done in the same way.

(i) => (iii) Let G1 \contained_in G2 \contained_in G3 ..... be an ascending

chain of open sets. Then, G1^c \contain G2^c \contain G3^c ..... is a
descedning chain of closed sets. By the DCC, there is N s.t. Gn^c = GN^c for
all n>=N, i.e. Gn = GN. Hence, ACC for open sets hold./

(iii)=>(i) In the same way as in (i)=>(iii).

Hence, (i), (ii), (iii), (iv) are all equivalent.//

(b) Assume not, i.e. let I be a collection of open sets of a noetherian

space X s.t. any finite subcollection of I cannot cover X. -------(*)
Choose any nonempty U_1 in I. By (*), U_1 is properly in X. Let X_1 =
(U_1)^c;nonempty closed. Since I covers X, there is nonempty U_2 in I such
that U_2 meets X_1. By (*) again, U_1 U U_2 is properly in X, so, let X_2 =
(U_1 U U_2)^c; nonempty closed, and X_1 properly contain X_2.
Assume U_1,...,U_n in I, X_1,...,X_n are given s.t. X_i = (Union of
U_1,...,U_n)^c;nonempty closed and {X_i,i=1,...,n} is strictly decreasing.
Since I covers X, there is a nonempty U_n+1 in X, so, let X_n+1 = (Union of
U_1,...,U_n+1)^c;nonempty closed, and, then, {X_i,i=1,...,n+1} is strictl
decreaing. Hence by induction, we can obtain an infinite properly descending
chian of closed subsets of X=> contradiction because X is noetherian.//

(c) Let Y be in X, and X is a noetherian space. If Y is not noetherian,

there is a sequence of strictly decreasing closed subsets of Y:
{Y_i,i=1,2,.....}. Since Y has induced topology, Y_i = Y intersection F_i
for closed subsets F_i of X. Since Y_i contains Y_i+1, Y_i+1 = Y_i
intersection Y_i+1 = Y intersection (F_i intersection F_i+1). Hence, by
replacing F_i by intersection of F1,...,F_i, WMA {F_i,i=1,2,..} is an
decreasing sequence of closed sets of X. Since X is noetherian,
F_N=F_N+1=.... => Y_N=Y_N+1=..... (contradiction)//

(d) If V is an irreducible closed set, V is a single point. (If not, two

points x,y have disjoint open sets, which are dense.) Since X is noetherian,
it is a finite union of irreducible closed sets, i.e. finite union of
points. Hence X has only finitely many points.//

1.8. Let Y=Z(p), H=Z(f) where p is a prime ideal of k[x1,...,xn], f: a

nonconstant polynomial. Since Y \not_contained_in H, f is not in p, and Y
\intersection H = Z(p,f).
Consider A(Y)= k[x1,...,xn]/p. Then, f+p in A(Y) is not zero, not unit, not
a zero-divisor. since A(Y) is a noetherian space, there is a primary
decomposition (f+p) = \intersection_{i=1 to m} q_i with r(q_i)=p_i; minimal
prime ideals of (f+p).
Then, by the Krull's Hauptidealsatz, ht(p_i)=1, and every component of Y
\intersection H corresponds to p_i.
Here, ht(p_i) + dim A(Y)/p_i = k[x1,...,xn]/(p,p_i) : coordinate ring of
each irreducible component => dim(component) = dim A(Y)/p_i = dim A(Y) -
ht(p_i) = r-1.//

1.9. Let a = (f_1,f_2,...,f_r)

=> 0 \contained_in (f_1) \contained_in (f_2) .... \contained_in
(f_1,f_2,...,f_r)=a.
=> A^n \contains Z(f_1) \contains Z(f_2) ..... \contains
Z(f_1,f_2,...,f_r)=Z(a).
so, note that dim Z(f_1,f_2,...,f_i) >= dim Z(f_1,f_2,...,f_i+1) for each i.
If f_i+1 is not a zero-divisor in A(Z(f_1,f_2,...,f_i)), there is a minimal
prime ideal p in A(Z(f_1,f_2,...,f_i)) containing the image of f_i+1 with
ht(p)=1. So, ht(p) + dim A(Z(f_1,...,f_i))/p = dim A(Z(f_1,...,f_i)), so,
dim Z(f_1,...,f_i+1) = dim A(Z(f_1,...,f_i+1) >=dim A(Z(f_1,...,f_i))/p =
dim Z(f_1,...,f_i))-1.
Otherwise, Z(f_1,...,f_i)=Z(f_1,...,f_i+1).
Hence, at each step, dimension decreses by 0 or 1. Because there are r such
steps, hence dimension of Z(f_1,...,f_r) >=n-r.//

1.10. (a) Any closed subset of Y is of the form : Y \intersection F, F:

closed in X.
For any chain Y \intersection F_0 \strictly_in Y \intersection F_1 .... of
irreducible closed sets, cl(Y \intersection F_0) \strictly_in cl(Y
\intersection F_1) .... : irreducible closed sets of X. Hence dim Y <= dim
X.

(b) dim U_i <= dim X is clear by (a) for all i. Hence sup dim U_i <= dim X.
Conversely, for any chain of irreducible closed subsets F_0 \strictly_in F_1
\strictly_in ..... F_n, choose an open set U_0 in {U_i} s.t. F_0
\intersection U_0 !=\empty. Then, since F_0 \strictly_in F_1, F_1
\intersection U_0 !=\empty. Since F_1 is irreducible and F_1 \intersection
U0 is a nonempty open subset, it must be dense in F_1. THen, F_1 - F_0 :
nonempty open subset of F_1 => (F_1 - F_0 ) \intersection U_0 != \empty.
Hence, F_0 \intersection U_0 \strictly_in F_1 \intersection U_0. But, using
same argument, we can construct a strict chain {F_i \intersection U_0}
i=0,1,...,n. Hence dim U_0 >=dim X. Hence sup dim U_i = dim X.//

(c) Consider X={0,1} with topology T={\empty, {0}, {0,1}}. Then, {0} is
dense open subset, because arbitrary open set intersects it. But, dim{0} =
0, because {0} doesnot contain any nonempty closed subsets other than
itself. But, {1} \strictly_in {1,2} is the maximal chain of irreducible
closed sets of X, so dim X=1. Hence dim U<dim X.//

(d) Assume not, i.e. Y \strictly_in X. Choose a maximal chain of irreducible

closed subsets of Y with the maximal length: Y_0 \strictly_in Y_1 ....
\strictly_in Y_n. Then, by adding X, we obtain a chain with length n+1, so
dim X>=n+1 > n=dimY. contradiction .//
(e) (From Atiyah-Macdonald)
Let A=k[x_1,...,x_n,....], m_1,m_2, ... : increading sequence of natural
numbers, s.t. m_i+1 - m_i > m_i - m_i-1 for all i>1. Let p_i = (x_{m_i +1} ,
..., x_{m_{i+1}}), S= (Union of all P_i)^c. Let X = Spec(S^-1 A) => X is
noetherian because S^-1 A is a noetherian ring (Chapter 7, Ex.9 of
Atiyah-Macdonald) but,, S^-1 p_i has height m_{i+1} - m_i : dim S^-1 A =
infinity.//

1.11. Define phi: k[x,y,z] -> k[t^3, t^4, t^5] by phi(f(x,y,z)) =

f(t^3,t^4,t^5). Clearly phi is surjective and f is in ker(phi) iff
f(t^3,t^4,t^5)=0, i.e. f is in I(Y). Hence ker(phi) = I(Y). Note that
k[t^3,t^4,t^4] is in k[t], so it is an ID, hence I(Y) is a prime ideal.
Now, ht(I(Y)) + dim (k[x,y,z]/I(Y)) = dim (k[x,y,z]).
Note dim k[x,y,z]/I(Y) = dim k[t^3,t^4,t^5] = dim k[t] (because k[t] is an
integral extension of k[t^3,t^4,t^5]) = 1. Hence, ht(I(Y))= 3-1=2.
Now we show that I(Y) cannot be generated by 2 elements. Let's search for
the elements of I(Y) with "minimal degree, next minimal degree, etc."
Let f(x,y,z) be in I(Y), f(x,y,z)= sum_{i,j,k} b(i,j,k) x^i y^j z^k,
b(i,j,k) is in k.
Then, f(t^3,t^4,t^5)=0 implies sum_{i,j,k} b(i,j,k) t^(3i+4j+5k) = 0. Let's
collect terms with same degree with respect to t. Let n=3i+4j+5k
If n=0, (i,j,k)=(0,0,0) => b(0,0,0)=0.
n=1, (i,j,k) does not exist
n=2, (i,j,k) does not exist
n=3, (i,j,k)=(1,0,0) => b(1,0,0)=0.
n=4, (i,j,k)=(0,1,0) => b(0,1,0)=0.
n=5, (i,j,k)=(0,0,1) => b(0,0,1)=0.
n=6, (i,j,k)=(2,0,0) => b(2,0,0)=0.
n=7, (i,j,k)=(1,1,0) => b(1,1,0)=0.
n=8, (i,j,k)=(1,0,1) and (0,2,0) => b(1,0,1) + b(0,2,0) = 0 i.e. we find
xz-y^2 = f_1.
n=9, (i,j,k)=(3,0,0) and (0,1,0) => b(3,0,0) + b(0,1,1) = 0 i.e. we find x^3
- yz = f_2.
n=10, (i,j,k)=(2,1,0) and (0,0,2) => b(2,1,0) + b(0,0,2) = 0 i.e. we find
x^2 y - z^2 = f_3.
n=11, (i,j,k)=(1,2,0) and (2,0,1) => b(1,2,0) + b(2,0,1) = 0 i.e. xy^2 -
x^2z = x(y^2-xz)=xf_1.
n>=12, if there are solutions (i,j,k), i+j+k>=3. So, polynomials obtained
from now must have degree>=3 for every term.
So, we have (f_1, f_2, f_3) \contained_in I(Y).
Note that f_1 is not in (f_2, f_3) (degf_1 = 2, deg of minimal degree term
of f_2,f_3 = 2)
f_2 is not in (f_1,f_3) (f_2 has yz but, deg 2 terms of f_1 and f_3 are
xz,y^2,z^2)
f_3 is not in (f_1,f_2) (f_3 has z^2 but, deg 2 terms of f_1 and f_2 are
xz,y^2,yz)
Note that if we find all other generators f_4,f_5,... of I(Y), each term of
f_i (i>=4) must have degree >=3 and f_1, f_2, f_3 are the only generators
those who have terms of degree 2. If (g_1 , g_2 ) = I(Y), i.e. I(Y) can be
generated by only two elements, (g_1, g_2) must contain f_1,f_2,f_3. By the
minimality of the degrees of f_1,f_2,f_3 in I(Y), g_1, g_2 must be constant
multiples of f_1,f_2,f_3. But, it is not possible. Hence I(Y) must have at
least 3 generators./

Remark: We did not prove that (xz-y^2, x^3-yz , x^2 y -z^2) = I(Y). It
requires more work.

1.12. Let f(x,y) = (x^2 -1)^2 + y^2 = x^4 -2x^2 +1 + y^2. It is irreducible.
(f has a factorization in C[x,y]: (x^2 -1 + iy)(x^2-1 -iy) in to
irreducibles. If f factors in R[x,y], since both R[x,y] and C[x,y] are UFDs
and R[x,y] is in C[x,y], the factorization must be equal to (x^2 -1 +
iy)(x^2-1-iy), but it is not possible in R[x,y].)
But, Z(f) = {(1,0),(-1,0)} = Z(x-1,y) \union Z(z+1,y). which is obviously
reducible.//
 Robin Hartshorne’s Algebraic Geometry Solutions
by Jinhyun Park

Chapter II Section 2 Schemes

2.1. Let A be a ring, let X = Spec(A), let f ∈ A and let D(f ) ⊂ X be the open
complement of V ((f )). Show that the locally ringed space (D(f ), OX |D(f ) ) is
isomorphic to Spec(Af ).
Proof. From a basic commutative algebra, we know that prime ideals in AS , for a multi-
plicative set S of A, correspond to prime ideals of A which do not intersect S. In particular,
Af = AS for S = {1, f, f 2 , · · · , } so that prime ideals of Af correspond to prime ideals of
A not containing f . This shows that the underlying topological spaces are homeomorphic.
For the morphism of structure sheaves, Prop. 2.2 -(b) gives the answer. This proves the
assertion. 
2.2. Let (X, OX ) be a scheme, and let U ⊂ X be any open subset. Show that
(U, OX |U ) is a scheme. We call this the induced scheme structure on the open set
U , and we refer to (U, OX |U ) as an open subscheme of X.
Proof. By the remark on p. 71 above the Prop. 2.2., affine subschemes of X form a basis
for the topology of X. Thus, for any open U ⊂ X there is an affine open subscheme Y ⊂ U ,
thus, by definition, (U, OX |U ) is a scheme. 
2.3. Reduced Schemes. A scheme (X, OX ) is reduced if for every open set U ⊂ X,
the ring OX (U ) has no nilpotent elements.
(a) Show that (X, OX ) is reduced if and only if for every P ∈ X, the local ring
OX,P has no nilpotent elements.
Proof. (⇒) Assume not, i.e. there is P ∈ X and 0 6= f ∈ OX,P such that f m = 0
for some m ∈ N. Then there is an open set V 3 P and g ∈ OX (V ) which represents
f . But, then g m = 0 which is a contradiction.
(⇐) Assume that for some open V ⊂ X, there is nonzero g ∈ OX (V ) such that
g m = 0. Then, there is P ∈ V for which the image f ∈ OX,P of g is nonzero and
f m = 0, which is a contradiction. 
(b) Let (X, OX ) be a scheme. Let (OX )red be the sheaf associated to the
presheaf U 7→ OX (U )red , where for any ring A, we denote by Ared the
quotient of A by its ideal of nilpotent elements. Show that (X, (OX )red )
is a scheme. We call it the reduced scheme associated to X, and denote it
by Xred . How that there is a morphism of schemes Xred → X, which is a
homeomorphism on the underlying topological spaces.
Claim. (X, (OX )red ) is a scheme.
Proof. For any affine schemes V ⊂ U ⊂ X, (OX |U )red (V ) = (OX )red |U (V ), so, the
rest is obvious. 
Claim. There is a morphism of schemes Xred → X which is a homeomorphism on
the underlying spaces.
Proof. Just define f : Xred → X to be the identity map on the underlying spaces.
We define f ] : OX → f∗ (OX )red to be
f ] (U ) : OX (U ) → OX (U )/nilrad(OX (U ))
for any open subset U ⊂ X. 
1
2

(c) Let f : X → Y be a morphism of schemes, and assume that X is reduced.

Show that there is a unique morphism g : X → Yred such that f is obtained
by composing g with the natural map Yred → Y .

Proof. Define g : X → Yred as follows. As a map on underlying spaces, g = f .

As a morphism of sheaves, g ] : (OY )red → g∗ OX = g∗ (OX )red is defined from
f ] : OY → f∗ OX by taking g ] = (f ] )red . This is possible because a nilpotent is sent
to a nilpotent so that a nilradical is sent to a nilradical. 

2.4. Let A be a ring and let (X, OX ) be a scheme. Given a morphism f : X →

Spec(A), we have an associated map on sheaves f ] : OSpec(A) → f∗ OX . Taking
global sections we obtaion a homomorphism A → Γ(X, OX ). Thus there is a
natural map
α : HomSch (X, Spec(A)) → HomRings (A, Γ(X, OX )).
Show that α is bijective (cf. (I, 3.5) for an analogous statement about varieties).

Proof. Let φ : A → Γ(X, OX ) be a ring homomorphism. We want to construct a natural

morphism of schemes which corresponds to φ.
φ X ρU
Notice that for any affine open U ⊂ X, we have A → Γ(X, OX ) → Γ(U, OX ) from which
we can obtain φ∗U = Spec(ρU X ◦ φ) : U ' Spec(Γ(U, O X )) → Spec(A). The question is
whether they glue together nicely so that we can we can actually obtain a map from X to
Spec(A). But, this is easy: for two affine open sets U and V and any affine open subset
W ⊂ U ∩ V , the restiction maps ρ are transitive so that
ρW W U W W
X = ρU ◦ ρX = ρV ◦ ρX

and Spec is contravariant functorial so that the morphism of schemes

φ∗U |W = Spec(ρU W W U W ∗
X ◦ φ) ◦ Spec(ρU ) = Spec(ρU ◦ ρX ◦ φ) = Spec(ρX ◦ φ) = φW

and by symmetry, φ∗V |W = φ∗W . Thus, by collecting {φ∗U }U ⊂X , we have φ∗ : X → Spec(A).

That these two procedures are inverse to each other is obvious. 

2.5. Describe Spec(Z), and show that it is a final object for the category of
schemes, i.e., each scheme X admits a unique morphism to Spec(Z).

Proof. Spec(Z) = {(0)} ∪ {(p)|p:prime number} with (0), not closed and (p) are closed
points. This is a dimension 1 scheme. On the other hand, take A = Z in Ex. 2.4. Then,
HomRings (Z, Γ(X, OX )) has only one element, namely, the ring homomorphism sending 1
to 1. This corresponds to a unique morphism of schemes X → Spec(Z), thus, it is a final
object for the category of schemes. 

2.6. Describe the spectrum of the zero ring, and show that it is an initial object
for the category of schemes. (According to our conventions, all ring homo-
morphisms must take 1 to 1. Since 0 = 1 in the zero ring, we see that each
ring R admits a unique homomorphism to the zero ring, but that there is no
homomorphism from the zero ring to R unless 0 = 1 in R.)

Proof. For A = 0, Spec(A) = φ. On the other hand, for any scheme X, any ring homo-
morphism Γ(X, OX ) → 0 is 0. Hence, by Ex. 2.4, there is a unique morphism of schemes
Spec(0) → X, namely, the inclusion of empty set to X. Hence, Spec(0) is an initial object
in the category of schemes. 
 3

2.7. Let X be a scheme. For any x ∈ X, let Ox be the local ring at x, and mx its
maximal ideal. We define the residue field of x on X to be the field k(x) = Ox /mx .
Now let K be any field. Show that to give a morphism of Spec(K) to X is
equivalent to give a point x ∈ X and an inclusion map k(x) → K.
Proof. (⇒) Let (η, η ] ) : Spec(K) → X be a morphism of schemes. As a map on topological
spaces, since Spec(K) consists of a single point {∗}, there is a unique point x ∈ X with
x := η(∗).
Now, from η ] , we obtain a local homomorphism η∗] : OX,x → OSpec(K),∗ = K, thus, the
map of their residue fields
OX,x OSpec(K),∗ K
η∗] : k(x) = → = = K.
mX,x mSpec(K),∗ 0
This is injective because k(x) is a field.
(⇐) Conversely, suppose that x ∈ X and an embedding k(x) ,→ K are given. We have the
obvious map on topological spaces η : Spec(K) → X defined to be ∗ 7→ x, thus, we need to
construct η ] : OX → η∗ OSpec(K) . But, this is easy:
If x ∈ U ⊂ X, then, η ] (U ) : OX (U ) → η∗ OSpec(K) (U ) = K is defined to be the


composition of maps
OX (U ) → OX,x → OX,x /mX,x = k(x) ,→ K.
If x 6∈ U ⊂ X, we let η ] (U ) = 0, where the target is the zero ring.
Thus, we constructed the desired morphism of schemes (η, η ] ) : Spec(K) → X. This
finishes the proof. 

2.8. Let X be a scheme. For any point x ∈ X, we define the Zariski tangent space
Tx to X at x to be the dual of the k(x)-vector space mx /m2x . Now assume that X
is a scheme over a field k, and let k[]/2 be
the ring of dual numbers over k. Show
2
that to give a k-morphism of Spec k[]/ to X is equivalent to giving a point
x ∈ X, rational over k (i.e., such that k(x) = k), and an element of Tx .
Proof. Notice first that as a topological space, Spec k[]/2 is a single point {∗} with

residue field k(∗) = k.

(⇒) Let (η, η ] ) ∈ Mork−sch Spec k[]/2 , X be given. Let x = η(∗). Since η is a k-

morphism, and k(∗) = k, we must have k(x) = k and x is a rational point.

On the other hand, we have a k-algebra local homomorphism η∗] : OX,x → OSpec(k[]/2 ),∗ =
k[]/2 =: k[], thus, η∗] (mX,x ) ⊂ (). But, since (2 ) = 0, we have η∗] (m2X,x ) ⊂ (2 ) = 0, thus
we get a k-vector space homomorphism
η∗] : mX,x /m2X,x → () ' k,
where the last map is an isomorphism of k-vector spaces.
Thus, we obtained a k-rational point x ∈ X and η∗] ∈ Homk mx /m2x , k = Tx as desired.

(⇐) Conversely,
suppose that we have a k-rational point x ∈ X and a k-linear map ξ ∈

Homk mx /m2x , k . Out of this data, we will define an element (η, η ] ) ∈ Mork−sch Spec k[]/2 , X .

First, as a map of topological spaces, just define η(∗) = x. Let’s define η ] .

If x 6∈ U ⊂ X, define η ] (U ) : OX (U ) → η∗ OSpec(k[]/2 ) (U ) = 0 to be 0.


If x ∈ U ⊂ X, notice that since x is a k-rational point, we first have a decomposition
OX,x = k ⊕ mX,x . Then, using this, define η ] (U ) as the composition of maps
α
OX (U ) → OX,x = k ⊕ mX,x → k[]/2 = η∗ OSpec(k[]/2 ) (U )


4

where the second map α sends (a, b) 7→ a + ξ(b), where b denotes its image in
mx /m2x . This proves the assertion.

2.9. If X is a topological space, and Z an irreducible closed subset of X, a generic
point for Z is a point ζ such that Z = {ζ}− . If X is a scheme, show that every
(nonempty) irreducible closed subset has a unique generic point.
Proof. Choose an affine open subset V ⊂ X, V ' Spec(A), with V ∩ Z 6= φ.
Claim (1). Z = V ∩ Z where the closure is taken in X.
Set theoretically, Z = V ∩ Z ∪ (Z ∩ (X − V )). But, since Z is irreducible and Z ∩ (X − V )
is a proper subset of Z, this claim is true.
Claim (2). V ∩ Z is irreducible.
If not, there are two proper closed subsets F1 , F2 of Z such that V ∩Z = (V ∩F1 )∪(V ∩F2 )
so that Z = (Z ∩ F1 ) ∪ (Z ∩ F2 ) ∪ (Z ∩ (X − V )) which contradicts the irreducibility of Z.
Thus, V ∩ Z is an irreducible closed subset of an affine variety V , i.e. there is a point
x corresponding to a prime ideal of A such that V ∩ Z = {x}− , where the closure here is
taken in V . Hence, by extending the closure in X, by Claim (1), Z = V ∩ Z = {x}− , which
shows the existence of a generic point.
If there are two generic points x1 , x2 , then, x1 ∈ {x2 }− . Thus, if we choose an affine
open subset V containing x2 , x1 must lie in V as well, and for two prime ideals p1 , p2
corresponding to x1 , x2 , p1 ⊃ p2 . But, by interchanging the roles of x1 and x2 , we also have
p1 ⊂ p2 , which means, x1 = x2 . Hence there is a unique generic point. 
2.10. Describe Spec(R[x]). How does its topological space compare to the set R?
to C?
Proof. See my solutions for Atiyah-MacDonald’s Introduction to commutative algebra Chap-
ter 1. 
2.11. Let k = Fp be the finite field with p elements. Describe k[x]. What are the
residue fields of its points? How many points are there with a given residue
field?
Proof. First of all, what are Spec(k[x])? (0) is the generic point and (f ) are closed points,
when f are nonzero irreducible polynomials. It is in general not very easy to enumerate all
irreducible polynomials. But, we can count the number of them, which will be done in the
sequel.
When ξ = (0), k[x]ξ = k(x) and mξ = 0, thus, the residue field is same as the fraction
field so that k(ξ) = nk(x). When
o ξ =n(f ), where o f is an irreducible polynomial of degree
h h
n ≥ 1, then, k[x]ξ = g |f 6 |g , mξ = g |f 6 |g, f |h so that
k[x]ξ /mξ ' (k[x]/(f ))ξ ' k[x]/(f ) ' Fpn .
So, Fp [x] and Fpn , n ≥ 1 are all possible residue fields. Obviously only the generic point
can have k[x] as the residue field.
To compute the number of points which have a specific Fpn as its residue field is equivalent
to count the number of monic irreducible polynomials of degree n over Fp . To do so, we
will use the collection of all maps from Spec (Fn := Fpn ) to Spec(k[x]).
If ξ = (f ) 6= 0 with deg f = m is a Fn -rational point, then it means, the image of
f ∈ k(ξ) = k[x]/(f ) ,→ Fn is an element of Fn . In particular, m|n and there are m distinct
 5

embeddings coming from various conjugates. Conversely, each nonzero element of Fn is a

root of a unique monic irreducible polynomial of degree m dividing n. Hence each irreducible
monic polynomial of degree m, m|n determines m elements of Fn and each element of Fn
is also determined by an irreducible monic polynomial.
So, let Sn be the number of monic irreducible polynomials. Let Tn = nSn . Then,
X
T m = pn .
m|n
P
To solve this equation, we use the Möbius inversion formulae: if g(n) = d|n f (d), then,
f (n) = d|n µ(d)g nd where
P


 1 n=1
µ(n) = 0 n is not square free. .
(−1)k n = p1 · · · pk : distinct primes


(See any reasonable number theory book.)

Hence,
1X  n
Sn = µ(d) p d
n
d|n

is the number of monic irreducible polynomials of degree n over Fp , which is equal to the
number of points of Spec(k[x]) whose residue field is exactly Fpn . 

2.12. Glueing Lemma. Generalize the glueing procedure described in the text
(2.3.5) as follows. Let {Xi } be a family of schemes (possibly infinite). For
each i 6= j, suppose given an open subset Uij ⊂ Xi , and let it have the induced
scheme structure (Ex. 2.2). Suppose also given for each i 6= j an isomorphism
of schemes φij : Uij → Uji such that (1) for each i, j, φji = φ−1 ij , and (2) for each
i, j, k, φij (Uij ∩ Uik ) = Uji ∩ Ujk , and φik = φjk ◦ φij on Uij ∩ Uik . Then show that there
is a scheme X, together with morphisms ψi : Xi → X for each i, such that (1)
ψi is an isomorphism of Xi onto an open subscheme of X, (2) the ψi (Xi ) cover
X, (3) ψi (Uij ) = ψi (Xi ) ∩ ψj (Xj ) and (4) ψi = ψj ◦ φij on Uij . We say that X is
obtained by glueing the schemes Xi along the isomorphisms φij . An interesting
special case is when the family Xi is arbitrary, but the Uij and φij are all ` empty.
Then the scheme X is called the disjoint union of the Xi , and is denoted Xi .

Proof. Obvious. 

2.13. A topological space is quasi-compact if every open cover has a finite sub-
cover.
(a) Show that a topological space is noetherian (I, §1) if and only if every
open subset is quasi-compact.

Proof. (⇒) By Ex. I-1.7-(c), any open subset is noetherian, hence, by Ex. I-1.7-(b),
it is quasi-compact. S
(⇐) let U1 ⊂ U2 · · · be an ascending chain of open
Sr subsets of X. Let U = i Ui .
By assumption, U is quasi-compact so that U = i=1 Ui for some r. Then, Ur =
Ur+1 = · · · so that X is noetherian. 

(b) If X is an affine scheme, show that sp(X) is quasi-compact, but not in

general noetherian. We say that X is quasi-compact is sp(X) is.
6

Proof. Let X = Spec(A). We know that for g ∈ A, D(g) ' Spec(AP g ) form
 a basis
S
for X. Hence, Spec(A) = g∈A D(g) which means V (1) = V g∈A (g) , which
P Pr
means 1 ∈ g∈A (g), thus, 1 = i=1 ci gi for some ci ∈ A and gi ∈ A. But, then it
means Spec(A) = ri=1 D(gi ). Hence Spec(A) is quasi-compact.
S
For A = k[x1 , x2 , · · · ], Spec(A) is not noetherian. 
(c) If A is a noetherian ring, show that sp(Spec(A)) is a noetherian topological
space.
Proof. Let V (a1 ) ⊃ V (a2 ) ⊃ · · · be a descending chain of closed subsets of Spec(A).
√ √ √
Then, a1 ⊂ a2 ⊂ · · · . Since A is noetherian, for all sufficiently large N , aN =
√ √
aN +1 = · · · . Hence, by applying V ( ) again and noting that V (ai ) = V ( ai ),
V (aN ) = V (aN +1 ) = · · · . Hence Spec(A) is noetherian. 
(d) Give an example to show that sp(Spec(A)) can be noetherian even when
A is not.
Proof. ? 
2.14.
2.15.
2.16.
2.17.
2.18.
2.19.
 Robin Hartshorne’s Algebraic Geometry Solutions
by Jinhyun Park

Chapter II Section 7, Projective Morphisms

7.1.

7.9. Let r + 1 be the rank of E.

(a). There are several ways to prove it.

Proof 1 We assume the following result from Chow group theory: (See Appendix A section2
A11 and section 3. The group A(X) is here CH(X).)
r
!
X
CH ∗ (P(E)) ' Z[ξ]/ (−1)i ci (E)ξ r−i ⊗Z CH ∗ (X)
i=0
as graded rings. If we look at the grade 1 part, as Z-modules,
CH 1 (P(E)) ' (Z ⊗Z CH 0 (X)) ⊕ (Z ⊗Z CH 1 (X))
and CH 1 (−) = Pic(−) so that Pic(P(E)) ' Z ⊕ Pic(X) as desired.
Proof 2 We can use the Grothendieck groups, i.e. K-theory to do so. Note that
r
!
X
K(P(E)) ' Z[ξ]/ (−1)i ci (E)ξ r−i ⊗Z K(X)
i=0

as rings. For the detail, see Yuri Manin Lectures on the K-functor in Algebraic
Geometry, Russian Mathematical Surveys, 24 (1969) 1-90, in particular, p. 44, from
Prop (10.2) to Cor. (10.5).
Proof 3 Here we give a direct proof. In fact, it adapts a way from Proof 1. It can also use
the method from Proof 2. Totally your choice.
Define a map φ : Z ⊕ Pic(X) → Pic(P(E)) by (n, L) 7→ (π ∗ L)(n) := (π ∗ L) ⊗
OP(E) (n).
Claim. This map is injective.
Assume that φ(n, L) = OP(E) , i.e. π ∗ L ⊗ OP(E) (n) ' OP(E) . Apply π∗ to it. From
II (7.11), recall thet

 0 n<0
π∗ (OP(E) (n)) = OX n=0 .
n
Sym (E) n > 0


So, by applying the projection formula (Ex. II (5.1)-(d)), we obtain, L⊗π∗ OP(E) (n) '
OX , i.e.
π∗ OP(E) ' L−1 .
Note that it is a line bundle and rk(Symn (E)) ≥ r + 1 ≥ 2 if n > 0 by the given
assumption, so that the only possible choice for n is n = 0. Then, it implies that
L ' OX . Hence φ is injective.
Claim. This map is surjective.
In case E is a trivial bundle, then P(E) ' X × Pr so that we already know the
result.
1
2

In general, choose an open subset U ⊂ X over which E is trivial and let Z =

X − U . Then, we have a closed immersion P(E|Z ) ,→ P(E) and an open immersion
P(EU ) ,→ P(E|U ) ' U × Pr . Let m = dim X. Then we have
CHm+r−1 (P(E|Z )) / Pic(P(E)) / Pic(P(E|U )) /0.
O O O
φZ φX φU

Z ⊕ CHm−1 (Z) / Z ⊕ PicX / Z ⊕ PicU /0

By induction on the dimension, φZ is surjective and we already know that φU is

an isomorphism. Hence, by a simple diagram chasing, we have the surjectivity of
φX .
(b). Let π : P(E) → X, π 0 : P(E 0 ) → X be the structure morphisms and let φ : P(E) → P(E)
be the given isomorphism over X:
P(E) .
φ xx
x;
x
xx π
xx 0 
P(E 0 )
π /X

φ∗ OP(E) (1) is an invertible sheaf on P(E) so that by part (a), we have

∗
(1) : φ∗ OP(E) (1) ' π 0 L0 ⊗ OP(E 0 ) (n0 )
for some L0 ∈ PicX and n0 ∈ Z. Similarly, φ−1 being a morphism, we have
∗
(2) : φ−1 OP(E 0 ) (1) ' π ∗ L ⊗ OP(E) (n)
for some L ∈ PicX and n ∈ Z. By applying φ∗ to (2), we have
∗ ∗
⊗n
OP(E 0 ) (1) ' φ∗ φ−1 OP(E 0 ) (1) ' φ∗ π ∗ L ⊗ φ∗ OP(E) (n) ' π 0 L ⊗ φ∗ OP(E) (1)
 
0∗ 0∗ 0 0 ⊗n 0∗ 0 ⊗n
⊗ OP(E 0 ) (nn0 ).


' π L ⊗ π L ⊗ OP(E) (n ) 'π L⊗L
Recall that 

 0 m<0
π∗0 PP(E 0 ) (n) = OX m=0
Symm E 0 m > 0

so that if we apply π∗0 to the above, then by the projection formula, we will have
⊗n
OX ' L ⊗ L0 ⊗ π∗0 OP(E 0 ) (nn0 ) .

Since OX , L ⊗ L0 ⊗n are invertible sheaves, it makes sense only when nn0 = 1. Hence we
have either (n, n0 ) = (−1, −1) or (n, n0 ) = (1, 1).
∗
If (n, n0 ) = (−1, −1), then, we have L ' L0 and (2) becomes φ−1 OP(E 0 ) (1) ' π ∗ L ⊗
∗ ∗
OP(E) (−1). φ being an isomorphism, φ−1 = φ∗ , so that π∗0 = π∗ φ∗ = π∗ φ−1 and the
projection formula gives E 0 ' L ⊗ 0 ' 0 which is not possible. Hence (n, n0 ) = (1, 1).
∗ ∗
Hence, we have (2): φ−1 OP(E 0 ) (1) ' π ∗ L⊗OP(E) (1), and as above, noting that φ−1 = φ∗ ,
applying π∗ and using the projection formula, we will have E 0 ' L ⊗ E as desired.
 Robin Hartshorne’s Algebraic Geometry Solutions
by Jinhyun Park

Chapter II Section 8, Differentials

8.1.

(a).

(b).

(c).

(d).

8.2.

8.3.

(a).

(b).

(c).

8.4.

(a).

(b).

(c).

(d).

(e).

(f ).

(g).

8.5.

(a).

(b).

8.6.
1
2

(a). Here, we assume that there exists at least one lefting g : A → B 0 . We prove all the
required propositions.
Claim. I has a natural structure of B-module.
Let b ∈ B, x ∈ I. Let b0 ∈ B 0 be a lifting of b under the given surjection p : B 0 → B.
Define b · x = b0 x ∈ I. If b00 ∈ B 0 is another lifting of b, then p(b00 − b0 ) = 0 implies b00 − b0 ∈ I.
Hence, b00 x − b0 x = (b00 − b0 )x ∈ I 2 = 0, i.e. b · x is well defined. It proves the claim.
Since we have a k-algebra homomorphism f : A → B and g : A → B 0 is a lifting, in fact,
b · x = g(b)x by above claim for any lifting g.
If g 0 : A → B 0 is another such lifting, then obviously the image of θ = g − g 0 lies in I.
Claim. θ : A → I is a k-derivation.
Obviously, it is additive because g, g 0 are. For a ∈ k, since g(1) = g 0 (1), θ(a) = g(a) −
g 0 (a)
= ag(1) − ag 0 (1) = 0. We now need to prove that for a, b ∈ A, θ(ab) = aθ(b) + bθ(a),
i.e.
g(ab) − g 0 (ab) = a(g(b) − g 0 (b)) + b(g(a) − g 0 (a)).
Recall how the action of A was defined on I. Hence,
RHS = g(a)(g(b) − g 0 (b)) + g 0 (b)(g(a) − g 0 (a)) = g(ab) − g(a)g 0 (b) + g 0 (b)g(a) − g 0 (ab)
= g(ab) − g 0 (ab) = LHS
so that θ is a k-derivation, i.e. θ ∈ Derk (A, I) = HomA (ΩA/k , I). It proves the claim.

Now, conversely, let θ ∈ HomA (ΩA/k , I) = Derk (A, I).

Claim. g 0 := g + θ is another lifting of f .
Since θ is additive, so is g 0 . Now,
g 0 (ab) = g(ab) + θ(ab) = g(ab) + aθ(b) + bθ(a)
= g(a)g(b) + g(a)θ(b) + g(b)θ(a) + θ(a)θ(b)
= (g(a) + θ(a))(g(b) + θ(b)) = g 0 (a)g 0 (b)
so that g 0 is multiplicative.
If a ∈ k, then θ is a k-derivation so that θ(a) = 0. Hence g 0 (a) = g(a) = ag(1) = a. Hence
g 0 is a k-algebra homomorphism. Now, (p◦g 0 )(a) = p(g(a)+θ(a)) = p◦g(a)+p(θ(a)) = f (a)
because θ(a) ∈ I and p(I) = 0. Hence g 0 is another lifting of f .
(b). For each i, choose bi ∈ B 0 such that p(bi ) = f (x̄i ). Define h : P = k[x1 , · · · , xn ] → B 0 be
the k-algebra homomorphism determined by h(xi ) := bi . Obviously, the diagram commutes
by construction.
Let q : P → A be the given surjection. If j ∈ J, then since q(j) = 0, we have f (q(j)) =
p(h(j)) = 0 i.e. h(j) ∈ I. Hence we have h|J : J → I. But I 2 = 0 implies that we have a
k-homomorphism h̄ : J/J 2 → I.
Claim. This map is even A-linear.
First, we note that J/J 2 has a canonical A-action. Let a ∈ A, [j] ∈ J/J 2 . Choose any
lifting a0 ∈ P of a and define a·[j] = [a0 j]. If we have another lifting a00 of a, then a00 −a0 ∈ J
so that (a00 − a0 )j ∈ J 2 , i.e. [a0 j] = [a00 j] so, this action is well-defined.
In part (a), we noted that the action of A on I is well-defined. To show that h̄ : J/J 2 → I
is A-equivariant, it is enough to show that the action of A is preserved. This is easy: Let
a ∈ A and choose a lifting a0 ∈ P . Then by the commutativity of the diagram, h(a0 ) is a
lifting of f (a) so that for [j] ∈ J/J 2 ,
h̄(a · [j]) = h̄([a0 j]) = h(a0 j) = h(a0 )h(j) = a · h(j) = a · j̄([j]).
 3

It proves the required A-linearity.

(c). By the hypothesis, SpecA ,→ Ank is a nonsingular subvariety. Hence by (8.17), we have
an exact sequence
0 → J/J 2 → ΩP/k ⊗ A → ΩA/k → 0.

A being nonsingular, ΩA/k is projective (because the sheaf ΩSpecA/k is locally free). Hence,
above sequence splits and so by applying HomA (−I), we obtain

0 / HomA (ΩA/k , I) / HomA (ΩP/k ⊗ A, I) / HomA (J/J 2 , I) /0.

'

HomP (ΩP/k , I) = / Derk (P, I)

Let θ ∈ HomP (ΩP/k , I) be an element mapped to h̄ ∈ HomA (J/J 2 , I) defined in part (b).
Regard θ as a k-derivation of P to B 0 ⊃ I. Let h0 = h − θ.

Claim. h0 : P → B 0 is a k-homomorphism such that h0 (J) = 0.

Obviously, θ being a k-derivation, h0 (a) = a for a ∈ k and h0 is additive. If a, b ∈ P , then

h0 (ab) = h(ab) − θ(ab) = h(ab) − bθ(a) − aθ(b) + θ(a)θ(b)

= (h(a) − θ(a))(h(b) − θ(b)) = h0 (a)h0 (b).

If j ∈ J, θ(j) = h̄(j) = h(j) so that h(j) = h(j) − θ(j) = 0. Hence h0 gives a rise to a
k-homomorphism g : A → B 0 . Since h was a lifting of f from P to B 0 , obviously, g is indeed
a required lifting.

8.7. As an application of the infinitesimal lifting property, we consider the fol-

lowing general problem. Let X be a scheme of finite type over k, and let
F be a coherent sheaf on X. We seek to classify schemes X 0 over k, which
have a sheaf of ideals I such that I 2 = 0 and (X 0 , OX 0 /I) ' (X, OX ), and such
that I with its resulting structure of OX -module is isomorphic to the given
sheaf F. Such a pair X 0 , F we call an infinitesimal extension of the scheme X
bye the sheaf F. One such extension, the trivial one, is obtained as follows.
Take OX 0 = OX ⊕ F as sheaves of abelian groups, and define multiplication by
(a ⊕ f ) · (a0 ⊕ f 0 ) = aa0 ⊕ (af 0 + a0 f ). Then the topological space X with the sheaf
of rings OX 0 is an infinitesimal extension of X by F.
The general problem of classifying extensions of X by F can be quite com-
plicated. So for now, just prove the following special case: if X is affine and
nonsingular, then any extension of X by a coherent sheaf F is isomorphic to the
trivial one. See (III, Ex. 4.10) for another case.

Proof. Suppose that we have an infinitesimal extension

α
0 → I → A0 → A → 0
4

defined by a ring A0 and its square-zero ideal I with I 2 = 0. By the infinitesimal lifting
property we have a lifting f , that is a k-algebra homomorphism, of the identity map of A:
0


I


0
>~ A
f ~~
~~~ α
~ id 
A /A


0
and it gives a splitting of A0 ' A⊕I as k-modules. We show that it is in fact an isomorphism
of k-algebras, where A ⊕ I is seen as given the structure of the trivial extension as in the
statement of the problem.
For each x, y ∈ A0 , we have x − f (α(x)), y − f (α(y)) ∈ I. Since I 2 = 0 we have
(x − f (α(x)))(y − f (α(y))) = 0
that gives xy = −f (α(x))f (α(y)) + xf (α(y)) + f (α(x))y. Thus,
xy − f (α(xy)) = xy − f (α(x))f (α(y)) = −2f (α(x))f (α(y)) + xf (α(y)) + f (α(x))y
= (x − f (α(x)))f (α(y)) + f (α(x))(x − f (α(y))).
This immediately implies that, when we identify x ∈ A0 with the pair (f (α(x)), x − f (α(x)))
of A ⊕ I, the product structure of A0 is identical to that of A ⊕ I, as desired. Thus there is
only one extension up to isomorphism. 
8.8.
 Robin Hartshorne’s Algebraic Geometry Solutions
by Jinhyun Park

Chapter III Cohomology Section 4 Cohomology of Affine spaces

4.1.

4.2.

4.3.

4.4.

4.5.

4.6.

4.7.

4.8.

4.9.

4.10. Let X be a nonsingular variety over an algebraically closed field k, and let
F be a coherent sheaf on X. Show that there is a one-to-one correspondence
between the set of infinitesimal extensions of X by F (II, Ex. 8.7) up to iso-
morphism, and the group H 1 (X, F ⊗ T ), where T is the tangent sheaf of X (II,
§8). [Hint: Use (II, Ex. 8.6) and (4.5)]
Proof. Let U = {Ui }i∈I be an affine open cover of X. Note that since X is nonsingular, so
is each Ui . Let (X 0 , OX 0 ) be an infinitesimal extension of X by F, that is, there is a scheme
X 0 with an ideal sheaf I ' F as OX -modules, such that I 2 = 0, (X, OX 0 /I) ' (X, OX ) as
ringed spaces, and a short exact sequence of OX -modules
0 → I → OX 0 → OX → 0.
Since each Ui is nonsingular and affine, by Ex.II-8.7, the above short exact sequence
restricts to a split exact sequence on Ui , where the splitting is given by a lifting αi :
OX |Ui → OX 0 |Ui .
On each Uij = Ui ∩ Uj , that is affine since X is separated, we have two liftings αi |Uij ,
αj |Uij : OX |Uij → OX 0 |Uij , and they differs by a section βij in Derk (OX (Uij ), I(Uij )) so
that on Uij we have
αi − αj = βij .
Notice that βij can be seen as a section in (F ⊗ T )(Uij ) via isomorphisms
Derk (OX (Uij ), I(Uij )) ' HomOX/k (Uij ) (ΩX/k (Uij ), I(Uij )) ' (F⊗Ω∗X/k )(Uij ) ' (F⊗T )(Uij ).
Restricting all the above sections onto Uijk = Ui ∩ Uj ∩ Uk , we thus obtain
βij + βjk + βki = (αi − αj ) + (αj − αk ) + (αk − αi ) = 0,
and {βij } gives a cocycle of the Čech complex C̆ • (U, F ⊗ T ) in degree 1.
For a different choice of liftings µi : OX |Ui → OX 0 |Ui for each Ui , as above we have the
corresponding sections βij 0 of Der (O (U ), I(U )) with µ − µ = β 0 on U , and with
k X ij ij i j ij ij
0 + β 0 + β 0 = 0 on U .
βij jk ki ijk
1
2

Applying the Ex. II-8.6- (a) again to the pair of liftings αi and µi on Ui , we have sections
ξi of Derk (OX (Ui ), I(Ui )) for each Ui with αi − µ = ξi , that can also be seen as a section of
F ⊗ T on Ui . Then, on Uij we have
0
βij − βij = (αi − αj ) − (µi − µj ) = ξi − ξj ,
0 } give the same cohomology class in H̆ 1 (U, F ⊗ T ). This
thus the cocycles {βij } and {βij
last group is isomorphic to H 1 (X, F ⊗ T ) by (4.5). The converse is easy. This finishes the
proof. 
4.11.
 Robin Hartshorne’s Algebraic Geometry Solutions
by Jinhyun Park

Chapter III Cohomology Section 5 The Cohomology of Projective Space

5.1.
5.2.
5.3.
5.4.
5.5.
5.6.
(a) Identify Q = P1 × P1 and let |Y | = P1 × ∗, |Z| = ∗ × P1 . First, observe that
Q is birational to P2 and h1 (X, OX ) is a birational invariant, so, h1 (Q, OQ ) =
h1 (P2 , OP2 ) = 0.
Claim (1). Let p > 0. Then, H 1 (Q, OQ (p, 0)) = 0.
Proof. Let Y = P1 × {p-points}. Then, we have a short exact sequence 0 →
OQ (−p, 0) → OQ → OY → 0. Tensor it by OQ (p, 0) then, we obtain 0 → OQ →
OQ (p, 0) → OY (p, 0) → 0. Then, from the cohomology long exact sequence, we
obtain
M
0 = H 1 (Q, OQ ) → H 1 (Q, OQ (p, 0)) → H 1 (P1 , OP1 (p|Y |2 )) → 0,
p

|2
but, |Y = 0, so, by Serre duality, H 1 (P1 , O 0 1 ∗
P1 ) ' H (P , OP1 (−2)) = 0. Hence,
1
H (Q, OQ (p, 0)) = 0 for p > 0. This finishes the proof of Claim 1. 
By symmetry, we also have H 1 (Q, OQ (0, q)) = 0 for q > 0.
Claim (2). For all p ≥ 0, q ≥ 0, H 1 (Q, OQ (p, q)) = 0.
Proof. If (p, q) = (0, 0) or p = 0 or q = 0, then, we already know this result, so,
assume that p, q > 0. Tensor the sequence 0 → OQ (−p, 0) → OQ → OY → 0 with
OQ (p, q) to obtain a short exact sequence 0 → OQ (0, q) → OQ (p, q) → OY (p, q) → 0.
Then, from the cohomology long exact sequence we have
M
H 1 (Q, OQ (0, q)) → H 1 (Q, OQ (p, q)) → H 1 (P1 , OP1 (p|Y |2 + q|Y |.|Z|))
p

but p|Y |2 + q|Y |.|Z| = q and by Serre duality, H 1 (P1 , OP1 (q)) ' H 0 (P1 , OP1 (−q −
2))∗ = 0 as −q − 2 < 0. By Claim 1, we know that H 1 (Q, OQ (0, q)) = 0 so,
H 1 (Q, OQ (p, q)) = 0 consequently. This proves the result. 
Claim (3). For any p ∈ Z, H 1 (Q, OQ (p, −1)) ' H 1 (Q, OQ (0, −1)).
Proof. If p = 0, it is obvious. First consider the case when p > 0. From the
sequence 0 → OQ (−p, 0) → OQ → OY → 0, by tensoring with OQ (p, −1), we
obtain 0 → OQ (0, −1) → OQ (p, −1) → OY (p, −1) → 0. Hence, the long exact
sequence gives
M M
H 0 (P1 , OP1 (p|Y |2 +(−1)|Y |.|Z| = −1) → H 1 (Q, OQ (0, −1)) → H 1 (Q, OQ (p, −1)) → H 1 (P1 , OP1 (−1)).
p p
1
2

Then, H 0 (P1 , OP1 (−1)) = 0 and H 1 (P1 , OP1 (−1)) ' H 0 (P1 , OP1 (−1))∗ = 0, so,
H 1 (Q, OQ (p, −1)) ' H 1 (Q, OQ (0, −1)) indeed.
Now consider the case when p < 0. let p0 = −p > 0 and let Y 0 = P1 × {p0 -points}.
Then we have 0 → OQ (−p0 , 0) → OQ → OY 0 → 0 and by tensoring with OQ (0, −1),
we obtain 0 → OQ (−p0 , −1) → OQ (0, −1) → OY 0 (0, −1) → 0. Hence, the long exact
sequence gives us
M M
H 0 (P1 , OP1 (−1)) → H 1 (Q, OQ (−p0 , −1)) → H 1 (Q, OQ (0, −1)) → H 1 (P1 , OP1 (−1))
p0 p0

and H 0 (P1 , OP1 (−1)) = H 1 (P1 , OP1 (−1)) = 0. This shows that H 1 (Q, OQ (p, −1)) '
H 1 (Q, PQ (0, −1)) for p < 0. 
Claim (4). (i) H 1 (Q, OQ (0, q)) 6= 0 if q ≤ −2.
(ii) H 1 (Q, OQ (0, −1)) = 0.
Proof. Let p > 0. From 0 → OQ (−p, 0) → OQ → OY → 0, by tensoring with
OQ (0, q), we obtain 0 → OQ (−p, q) → OQ (0, q) → OY (0, q) → 0 so that the long
exact sequence gives
M M
H 0 (P1 , OP1 (q)) → H 1 (Q, OQ (−p, q)) → H 1 (Q, OQ (0, q)) → H 1 (P1 , OP1 (q)) → 0.
p p

When q ≤ −2, H 0 (P1 , OP1 (−1)) = 0 and h1 (P1 , O 0 1

P1 (q)) = h (P , OP1 (−q − 2)) >
0 so that H 1 (0, q) 6= 0. This proves (i) and by symmetry we also have H 1 (p, 0) 6= 0
if p ≤ −2. This proves (3).
When p = 1, q = 0, we have
'
k = H 0 (Q, OQ ) → H 0 (P1 , OP1 ) = k → H 1 (Q, OQ (−1, 0)) → H 1 (Q, OQ ) = 0
so that H 1 (Q, OQ (−1, 0)) = 0. This proves (ii) and similarly we have H 1 (Q, OQ (0, −1)) =
0. 
Now, we prove (2). From V. 1.4.4, the canonical line bundle K ' OQ (−2, −2),
so, when a, b < 0, by Serre duality,
H 1 (Q, OQ (a, b)) ' H 1 (Q, OQ (−a − 2, −b − 2))∗ .
If a, b ≤ −2, then, by Claim 2, this group vanishes.
In case (a, b) = (0, −1), (−1, 0), (−2, −1), (−1, −2), (−1, −1), the previous claims
already show it. Hence, it is 0 for any a, b < 0. This proves (2). (1) is trivial once
we have (2) and the previous claims.
(b) (1) For Y , 0 → OQ (−Y ) → OQ → OY → 0 is exact and OQ (−Y ) ' OQ (−a, −b).
Thus,
0 → H 0 (Q, PQ (−a, −b)) → H 0 (Q, OQ ) → H 0 (Y, OY ) → H 1 (Q, OQ (−a, −b)) → 0
so, H 0 (Y, OY ) ' H 0 (Q, OQ ) ' k. Hence Y has only 1 connected component,
i.e. connected.
(2) Let L be a line bundle on Q of type (a, b) with a > 0, b > 0. Then, by II.
7.6.2, L is very ample so that it gives an embedding of Q into a projective
space PN . Then, by Bertini’s theorem (II, 8.18), there is a hyperplane H ⊂ PN
whose intersection with Q is a nonsingular projective curve Y and this OQ (Y )
is isomorphic to L, i.e. Y is of type (a, b).
(3) By Ex. II 5.14-(d), X ⊂ PrA is projectively normal if and only if it is normal
and for all n ≥ 0, the natural map Γ(Pr , OPr (n)) → Γ(X, OX (n)) is surjective.
We will use this.
 3

Since we have a sequence of closed embeddings Y ,→ Q ,→ P3 , it gives a

commutative diagram

Γ(P3 , OP3 (n)) / Γ(Y, OY (n))

QQQ nn7
QQQ
QQQ nnnnn
QQQ nn
( nnn
Γ(Q, OQ (n))

so, if, Γ(Q, OQ (n)) → Γ(Y, OY (n)) is not surjective, then, Γ(P3 , OP3 (n)) →
Γ(Y, OY (n)) cannot be surjective.
On the other hand, since Q = V (xy − zw) ⊂ P3 , the ideal sheaf of Q IQ '
OP3 (−2) so that the sequence 0 → OP3 (−2) → OP3 → OQ → 0 is exact. Hence,
by tensoring with OP3 (n), we have 0 → OP3 (n − 2) → OP3 (n) → PQ (n) → 0
whose cohomology long exact sequence gives
H 0 (P3 , OP3 (n)) → H 0 (Q, OQ (n)) → H 1 (P3 , OP3 (n − 2)) = 0.
Consequently, the map Γ(P3 , OP3 (n)) → Γ(Q, OQ (n)) is always surjective and it
implies that Γ(P3 , OP3 (n)) → Γ(Y, OY (n)) is surjective if and only if Γ(Q, OQ (n)) →
Γ(Y, OY (n))) is surjective if and only if Y ⊂ P3 is projectively normal, because
being nonsingular, Y is already normal.
Hence, it remains to show that Γ(Q, OQ (n)) → Γ(Y, OY (n)) is surjective if and
only if |a − b| ≤ 1.
(⇐) Suppose that |a − b| ≤ 1. Then, from 0 → OQ (−a, −b) → OQ → OY → 0,
we obtain 0 → OQ (n − a, n − b) → OQ (n, n) → OY (n) → 0 which gives us
Γ(Q, OQ (n)) → Γ(Y, OY (n)) → H 1 (Q, OQ (n − a, n − b).
But, |a−b| ≤ 1 means |(n−a)−(n−b)| ≤ 1 so, by part (a) - (1), H 1 (Q, OQ (n−
a, n − b)) vanishes and the natural map is surjective.
(⇒) Conversely, suppose that the natural map is surjective for all n ≥ 0. Then,
the same sequence gives
Γ(Q, OQ (n)) → Γ(Y, OY (n)) → H 1 (Q, OQ (n − a, n − b)) → H 1 (Q, OQ (n, n))
where the last one is 0 by Claim 2 of (a) and the first map is surjective. Hence,
we must have H 1 (Q, OQ (n − a, n − b)) = 0 for all n ≥ 0.
Toward contradiction, so, suppose that |a − b| ≥ 2, i.e. a ≥ b + 2 or b ≥ a + 2.
For the first case, when n = b, n − a ≤ −2 so that by (a)- (3), we have
H 1 (Q, OQ (n − a, n − b)) 6= 0, which is a contradiction. For the second case, we
will have the same contradiction. Hence |a − b| ≤ 1.
Hence, a nonsingular Y ⊂ Q of type (a, b) with a, b > 0 is projectively normal
in P3 if and only if |a − b| ≤ 1.
(c) First, we reduce this problem to a nonsingular Y . By part (b)-(2), Y is linearly
(hence rationally) equivalent to a nonsingular projective curve lying on Q and this
new curve has the same bidegree. Also, since this is a rational equivalence, they
belong to the same flat family, so, the arithmetic genera are unchanged (which
are defined to be h1 (Y, OY )). Hence, we may replace Y by its linearly equivalent
nonsingular Y . Then, for this Y , the arithmetic genus pa (Y ) = pg (Y ), the geometric
genus, and we can compute it in terms of a, b as follows: OQ (Y ) = OQ (a, b) and the
first Chern class c1 (NQ/Y ) = degY (NQ/Y ) = Y.Y = (ah+bk)2 = ab(h.k)+ba(k.h) =
2ab where h, k are generators of PicQ ' Z⊕Z with intersection product h2 = k 2 = 0,
h.k = k.h = 1.
4
 ∗
On the other hand, TP1 ×P1 = Ω1P1 ×P1 implies that c1 (TP1 ×P1 ) = c1 ∧2 TP1 ×P1 =


 
c1 KP∗1 ×P1 = −(c1 (KP1 ), c1 (KP1 )) = −(2 · 0 − 2, 2 · 0 − 2) = (2, 2) = 2h + 2k and
so, c1 (TQ |Y ) = degY TQ ⊗OQ OC = degY ∧2 TQ ⊗OQ OC = [KQ ∗ ] · (ah + bk) =

(2h + 2k) · (ah + bk) = 2(a + b).

Of course, c1 (TY ) = − degY (KY ) = −(2g − 2) = 2 − 2g. Hence the short exact
sequence
0 → TY → TQ |Y → NQ/Y → 0
gives c1 (TY ) + c1 (NQ/Y ) = c1 (TQ |Y ) and it is equivalent to 2(a + b) = 2 − 2g + 2ab,
i.e. g = ab − a − b + 1 = (a − 1)(b − 1). This proves the result.
5.7.
5.8.
5.9.
5.10.
 Robin Hartshorne’s Algebraic Geometry Solutions
by Jinhyun Park

Chapter III Section 9 Flat Morphisms

9.1.
9.2.
9.3.
9.4.
9.5.
9.6.
9.7.
*9.8. Let A be a finitely generated k-algebra. Write A as a quotient of a poly-
nomial ring P over k, and let J be the kernel:
0 → J → P → A → 0.
Consider the exact sequence of (II, 8.4A)
J/J 2 → ΩP/k ⊗P A → ΩA/k → 0.
Apply the functor HomA (·, A), and let T 1 (A) be the cokernel:
HomA (ΩP/k ⊗ A, A) → HomA (J/J 2 , A) → T 1 (A) → 0.
Now use the construction of (II, Ex. 8.6) to show that T 1 (A) classifies infinites-
imal deformations of A, i.e., algebras A0 flat over D = k[t]/t2 , with A0 ⊗D k ' A. It
follows that T 1 (A) is independent of the given representation of A as a quotient
of a polynomial ring P . (For more details, see Lichtenbaum and Schlessinger
[1].)
Proof. Suppose that P = k[x1 , · · · , xn ] is a polynomial k-algebra of which A is a quotient
with the kernel J. Let P2 := k[x1 , · · · , xn , y1 , · · · , yn ].
For each infinitesimal deformation A0 of A, we can define a k-algebra homomorphism
f : P2 → A0 so that we obtain the following commutative diagram with exact rows and
columns:
0 0 0

  
0 /J t /K /J /0

  
0 /P t / P2 /P /0

f
  
0 /A t / A0 /A /0

  
0 0 0
where K is an ideal of P − 2.
1
2

Notice that to give a k-algebra A0 with the required properties is equivalent to give an
ideal K, and the ambiguity is given by the choice of the k-algebra homomorphism f . Thus,
the set of equivalence classes of infinitesimal deformations A0 of A is equal to
{ choices of an ideal K}
.
{ choices of f }
We will identify the numerator and the denominator.
Claim. { choices of an ideal K} ' HomP (J, A) as sets.
Notice that the middle row splits via the natural inclusion P → P2 of the right hand side
P . So that as modules, P2 = P ⊕ tP .
Suppose an ideal K was chosen. For each x ∈ J, lift it to x̃ ∈ K. Since P2 = P ⊕ tP ⊃ K,
x̃ = x + t(y) for some y ∈ P . Two liftings of x differ by an image of tz for some z ∈ I, thus,
y ∈ P is not uniquely determined by x, but ȳ ∈ A is uniquely determined. Thus, it defines
a map in HomP (J, A) that sends x 7→ ȳ.
Conversely, suppose that φ ∈ HomP (J, A). Define an ideal K of P2 by
K = {x + ty|x ∈ J, y ∈ P such that ȳ = φ(x) in A}.
It is easy to see that K is an ideal of P2 , and the image of K in P is J so that
0→J →K→J →0
is exact. It defines A0 := P2 /K, and here f is the canonical quotient map. Thus, it shows
the claim.
Claim. { choices of f } ' Derk (P, A) as sets.
A choice of f : P2 → A0 gives after composing with t : P → P2 , a lifting of P → A to
P → A0 . Thus, Ex. II-8.6-(a) shows the assertion. This proves the claim.
Hence, the obvious identities
HomP (J, A) ' HomA (J/J 2 , A), and

Derk (P, A) ' HomP (ΩP/k , A)

show that the set of isomorphism classes of infinitesimal deformations are in one-to-one
correspondence with the coker HomP (ΩP/k , A) → HomA (J/J 2 , A) , which is by definition

T 1 (A). This finishes the proof. 

Remark. In fact, via a natural map
T1 (A) ⊃ Ext1A (ΩA/k , A),
where the natural map will be apparent from the following discussion.
For the exact sequence
J/J 2 → ΩP/k ⊗P A → ΩA/k → 0,
let L be the kernel of the second map so that we have a natural projection J/J 2 → L and
a commutative diagram

J/J 2
KKK
KKK
KKK
 K%
0 /L / ΩP/k ⊗P A / ΩA/k / 0.
 3

Then, it induces a commutative diagram with exact rows

HomP (ΩP/k , A) / HomP (J/J 2 , A) / T 1 (A) /0
O O
=

HomP (ΩP/k , A) / HomP (L, A) → Ext1A (ΩA/k , A) / Ext1 (ΩP/k , A)

P

First of all, since P is smooth over k, Ext1P (ΩP/k , A) ' 0, ΩP/k being projective. Hence,
by diagram chasing we can define a map
Ext1A (ΩA/k , A) → T 1 (A)
and furthermore, the diagram implies that it must be injective.
It is known that this map becomes an isomorphism when
(1) k is a perfect field, and
(2) A is a reduced k-algebra of finite type,
according to Lichtenbaum and Schlessinger.
9.9.
9.10.
9.11.
 Robin Hartshorne’s Algebraic Geometry Solutions
by Jinhyun Park

Chapter III Section 10 Smooth morphisms

10.1. Over a nonperfect field, smooth and regular are not equivalent. For ex-
ample, let k0 be a field of characteristic p > 0, let k = k0 (t), and let X ⊂ A2k be
the curve defined by y 2 = xp − t. Show that every local ring of X is a regular
local ring, but X is not smooth over k.
Proof. We need to suppose that!char(k0 ) = p > 2. Let f = y 2 − xp + t ∈ k[x, y]. Then
∂f  
∂f ∂f ∂x 0
∂x = 0, ∂y = 2y so that rk ∂f = rk = 1 everywhere on X because on X y 6= 0.
∂y
2y
Indeed, if y = 0, then xp = t over k = k0 (t), which is not possible. Hence X is regular
everywhere and every local ring of X is a regular local ring.
Let’s now prove that X → Spec(k) is not smooth.

Toward contradiction, suppose that
it
is smooth. Then by base change via Spec k → Spec (k), the morphism Xk → Spec k is
1 p
 
smooth. But, this is not true: Xk ⊂ Ak2 is defined by the equation y 2 = xp − t = x − t p
 1 
over k and the point (x, y) = t p , 0 on Xk has multiplicity 2 so that it is not regular at
this point. Contradiction. Hence X → Spec(k) is not smooth. 
10.2. Let f : X → Y be a proper, flat morphism of varieties over k. Suppose for
some point y ∈ Y that the fibre Xy is smooth over k(y). Then show that there
is an open neighborhood U of y in Y such that f : f −1 (U ) → U is smooth.
Proof. Let n be the relative dimension of the flat morphism f : X → Y . Since Xy →
Spec (k(y)) is smooth, ΩXy /k(y) is a locally free coherent sheaf of rank n on Xy . That is, for



each x ∈ Xy , dimk(x) ΩX/Y ⊗ k(x) = dimk(x) ΩXy /k(y) ⊗ k(x) = n. But, ΩX/Y ⊗ k(x) =



ΩX/Y x /mx ΩX/Y x so that by Nakayama’s lemma, there exist sections s1 , · · · , sn of



ΩX/Y over a neighborhood Ux of x whose images in ΩX/Y ⊗ k(x) = ΩX/Y x /mx ΩX/Y x
form a k(x)-basis and they generate ΩX/Y over Ux . This implies that for all z ∈ Ux ,



dimk(z) ΩX/Y ⊗ k(z) ≤ n. But, by Theorem II-8.6A, dimk(z) ΩX/Y ⊗ k(z) ≥ n so that


dimk(z) ΩX/Y ⊗ k(z) = n for all z ∈ Ux , i.e. ΩX/Y |Ux is locally free of rank n. Since
x ∈ Xy was arbitrary, by collecting
S all such Ux , we see that {Ux }x∈Xy is a cover of Xy such
that ΩX/Y is locally free on x∈Xy Ux .
The remaining point is to find a kind of tubular neighborhood of Xy . Since f : X → Y
is proper, by base change Spec (k(y)) → Y , Xy → Spec (k(y)) is also proper. Thus, in
particular, Xy is quasi-compact and there are finitely many points x1 , · · · , xm ∈ Xy such
that Ux1 , · · · , Uxm cover Xy .T Since f is flat, it is an open map so that f (Uxi ⊂ Y is
open containing y. Let U = m i=1 f (Uxi ). This is then an open subset of Y containing y.
Obviously, f −1 (U ) ⊂ x∈Xy Ux and thus ΩX/Y |f −1 (U ) is locally free on f −1 (U ). Because
S

flatness is stable under base change, that ΩX/Y |f −1 (U ) is locally free of rank n on f −1 (U ) is
equivalent to that f −1 (U ) → U is smooth. This finishes the proof. 
10.3. A morphism f : X → Y of schemes of finite type over k is étale if it is smooth
of relative dimension 0. It is unramified if for every x ∈ X, letting y = f (x), we
have my · Ox = mx , and k(x) is a separable algebraic extension of k(y). Show that
the following conditions are equivalent:
(i) f is étale;
1
2

(ii) f is flat, and ΩX/Y = 0;

(iii) f is flat and unramified.

Proof. (i) ⇔ (ii) is obvious by definition. (ii) ⇔(iii) is a direct consequence of Theorem
II-8.6A. 

10.4. Show that a morphism f : X → Y of schemes of finite type over k is étale

if and only if the following condition is satisfied: for each x ∈ X, let y = f (x).
Let Obx and Oby be the completions of the local rings at x and y. Choose fields
of representatives (II, 8.25A) k(x) ⊂ Obx and k(y) ⊂ O by so that k(y) ⊂ k(x) via
the natural map O by → Obx . Then our condition is that for every x ∈ X, k(x) is a
separable algebraic extension of k(y), and the natural map

O
by ⊗k(y) k(x) → O
bx

is an isomorphism.

Proof. By definition, f : X → Y is unramified if and only if for all x ∈ X with y = f (x),

k(x) is separable over k(y) and my · Ox = mx . Also, Ex III-10.3 shows that f is étale if and
only if f is flat and unramified. Thus, it is enough to show that f is flat if and only if for
' b
all x ∈ X with y = f (x), O by ⊗k(y) k(x) → Ox .
Since flatness is a local condition, it follows from the following three statements. (All
rings are supposed to be noetherian.)

Claim (1). A is a ring and I √ ⊂ A is an ideal. Then, the I-adic completion A → A

b is
faithfully flat if and only if I ⊂ A. (It is always flat.)

Proof. See SGA 1, IV-Cor 3.2 

Claim (2). Let (A, m), (B, n) be local rings with a local homomorphism A → B. Then,

grB ' grA ⊗A B ⇔ B

b'A
b ⊗A B.

Proof. Easy. 

Claim (3). Let (A, m), (B, n), A → B be as above. Then, A → B is flat if and only if
b'A
B b ⊗A B.

Proof. (⇒) Since 0 → mr+1 → mr → mr /mr+1 → 0 is exact and A → B is flat, 0 →

mr+1 ⊗A B → mr ⊗A B → mr /mr+1 ⊗A B → 0 is exact. Since mB = n and mr /mr+1 ⊗A B =
mr B/mr+1 B = (mB)r /(mB)r+1 , we immediately obtain that grB ' grA⊗A B which implies
that B b'A b ⊗A B by Claim (2).
(⇐) Let M → N be an injective A-module homomorphism. We want to show that
M ⊗A B  → N ⊗A B is an injection.
 Since
 A→A b is flat, B⊗A A
b is also a flat A-module. Thus,
 
M ⊗A B ⊗A A b → N ⊗A B ⊗A A b is an injection But, −⊗A B ⊗A A b ' (− ⊗A B)⊗A A b
is an injection and since A → A b is faithfully flat by Claim (1), M ⊗A B → N ⊗A B is injective
as desired. This finishes the proof. 

Thus, taking A = Oy , B = Ox gives the desired result because, when L = k(y), k = k(x),
we have L = B ⊗A k so taht − ⊗k L = − ⊗k k ⊗A B = − ⊗A B. 
 3

10.5. If x is a point of a scheme X, we define an étale neighborhood of x to be an

étale morphism f : U → X, together with a point x0 ∈ U such that f (x0 ) = x.
As an example of the use of étale neighborhoods, prove the following: if F
is a coherent sheaf on X, and if every point of X has an étale neighborhood
f : U → X for which f ∗ F is a free OU -module, then F is locally free on X.
Proof. The question being local, we may suppose that both U and X are affine. Further-
more, by localizing them at x and x0 , we reduce the problem to show the following:
A → B is an étale local homomorphism of local rings and M is an A-module such that
M ⊗A B ' B n . Then, M ' An .
But, this is easy: M ⊗A B ' B n ' An ⊗A B and since B 6= 0 and B is A-flat, B is a
faithfully flat A-module (SGA1 IV-cor 2.2) so that M ' An . This finishes the proof. 
 Algebraic Geometry by Robin Hartshorne
Exercises solutions by Jinhyun Park

Warning!!!!!!
1)This material is not for sale.
2)This is just for personal use only.
3)No commercial purpose please.
4)Never Never Ever read this solution unless you tried problems quite long time and
gave up. Doing so may impair your ability to think and solve problems.

Chapter 4.Curves, Section 1.Riemann-Roch Theorem.

1. Choose Q ∈ C. Choose n big enough so that deg n(2P − Q) > 2g − 2, g, 1.
⇒ h0 (n(2P − Q)) = 1 − g + n(2P − Q) > 1 ⇒ ∃ effective divisor D ∈ |n(2P − Q)| ⇒
∃f ∈ K(C) such that D + nQ − 2nP = (f ). Since deg D = n, so D cannot cancel
−2nP i.e. f has a pole only at P .//
2. Let F = {P1 , · · · , Pr }. Multiplying functions of Ex.IV.1.1 might give cancellation
of poles and zeros, so we need slightly different approach.
Let Q ∈ C − F . Consider D0 = n(P1 + · · · Pr − (r − 1)Q). Choose n > 2g − 2, g.
Then, ∃D ∈ |D0 | i.e. ∃f ∈ K(c) such that D + (r − 1)Q − nP1 − · · · − nPr = (f ).
Note that deg D = n. Each Pi occurs with order −n so, if Pi ∈ SuppD, then either
(i) ordPi D < n or (ii) D = nPi for some i, in this case, WMA i = 1 WLOG.
For (i) there is no problem.
For (ii) f has ples only at P2 , · · · , Pr not at P1 . By Ex.IV.1.1, ∃g ∈
K(C) which has a pole only at P1 . Let ordPi g = ni 2 ≤ i ≤ r. Then, if
we choose m > 1, n2 , · · · , nr , then, f m g has poles at and only at F .//
3. Proof 1) By I-(6.10), there is a projective nonsingular curve X̄ over k such that X
is an open subset of X̄, i.e. X̄ − X is a finite set, say, {P1 , · · · , Pr } =
6 φ because X is
not proper.
Then, by Ex.IV.1.2, ∃f ∈ k(X̄) = k(X) such that f has poles only at P1 , · · · , Pr .
We can consider f ∈ k(X̄) as a morphism f : X̄ → P1 .Then, f −1 (An ) = X, so,
g = f |X : X → A1 is a morphism.
f (¯(X)) is proper over k(because X̄ is proper) and irreducible and, f (X̄) 6=a point.
Hence, f (X̄) = P1 . And by (II-6.8), f is a finite morphism, in particular, affine
morphism. Hence, f −1 (A1 ) = X is affine.

Proof 2) As above, let F = {P1 , · · · , Pr } = X̄ − X. Choose m such that mr > 2g

Then, D = m(P1 + · · · Pr has a degree > 2g, so by (3), it is very ample. Then, it
gives an embedding of X̄ into a projective space PN for some N , and D = X̄.H for
a hyper plane H of PN . Then, X̄ − F = a closed subscheme of AN = PN − H which
is affine, so, X = X̄ − F is also affine.//
1
 Robin Hartshorne’s Algebraic Geometry Solutions
by Jinhyun Park

Chapter V Section 5 Birational Transformations

5.8. A surface singularity.Let k be an algebraically closed field, and let X be the
surface in A3k defined by the equation x2 +y 3 +z 5 = 0. It has an isolated singularity
at the origin P = (0, 0, 0).
(a). Show that the affine ring A = k[x, y, z]/(x2 + y 3 + z 5 ) of X is a unique factor-
ization domain, as follows. Let t = z −1 ; u = t3 x, and v = t2 y. Show that z is
irreducible in A; t ∈ k[u, v], and A[z −1 ] = k[u, v, t−1 ]. Conclude that A is a UFD.
Claim. z is irreducible in A.
Proof. Notice that
z is irreducible in A.
⇔ (z) is a prime ideal in A.
⇔ A/(z), which is k[x, y]/(x2 + y 3 ), is an integral domain.
⇔ x2 + y 3 is irreducible in k[x, y].
We prove the last statement. Suppose that for some f, g in k[x, y], we have
f g = x2 + y 3 .
(1) (case 1) Assume that degx (f ), the degree of f in x, is zero. Then f is a polynomial
in y, and we can write g = cx2 + ax + b for some c in k × and a, b in k[y]. Thus,
x2 + y 3 = f g = cf x2 + f ax + f b.
This implies that f = 1/c, which is a unit in k[x, y].
(2) (case 2) Assume that degx f = 1. Then, by multiplying a suitable constant in k × ,
we may assume that f = x + a and g = x + b for some a, b in k[y]. Then,

x2 + y 3 = f g = x2 + (a + b)x + ab
so that a + b = 0 and ab = y 3 . Then, b2 = −y 3 and since y is irreducible in k[y],
b = yb0 for some b0 in k[y]. Hence y 2 (b0 )2 = −y 3 , thus (b0 )2 = −y, which is impossible
because we then have 2 degy (b0 ) = 1.
(3) (case 3) Assume that degx (f ) = 2. Then, by symmetry, (case 1) shows that g must
be a unit.
Hence x2 + y 3 is irreducible in k[x, y], and thus z is irreducible in A. 
Claim. t ∈ k[u, v].
Proof. The equality x2 + y 3 + z 5 = 0 implies that in the fraction field we have −x2 /z 6 −
t3 /z 6 = 1/z. This is equivalent to t = −u2 − v 3 . 
Claim. A[z −1 ] = k[u, v, t−1 ].
Proof. The equalities
 x = (t−1 )3 t3 x = (t−1 )3 u,


y = (t−1 )2 t2 y = (t−1 )2 v,
z = t−1


show that A ⊂ k[u, v, t−1 ]. By the previous claim z −1 = t ∈ k[u, v], thus A[z −1 ] ⊂
k[u, v, t−1 ].
1
2

Conversely, we have
 u = (z −1 )3 x ∈ A[z −1 ],


v = (z −1 )2 y ∈ A[z −1 ],
 −1
t =z∈A
so that A[z −1 ] ⊃ k[u, v, t−1 ]. This finishes the proof. 
Claim. A is a UFD.
Notice that, being a polynomial ring, A[z −1 ] = k[u, v, t−1 ] is a UFD.
Lemma (1). Let f be an irreducible element in A, that is not in (z). Then, f is irreducible
as an element in A[z −1 ].
Proof. Suppose that (g/z m )(h/z n ) = f /1 for some integers m, n ≥ 0 and g, h in A, so that
gh = z m+n f . If m > 0 or n > 0, then since the quantity gh = z m+n f is in the ideal (z),
either g ∈ (z) or h ∈ (z). By canceling a suitable number of z’s if necessary, we may assume
that m = n = 0. Thus, gh = f in A. But, since f is irreducible in A, g or h must be a unit
in A. Hence g/1 or h/1 is a unit in A[z −1 ], thus, f /1 is irreducible in A[z −1 ]. 
Lemma (2). If a nonzero element f /1 ∈ A[z −1 ] is irreducible for some f in A, then
f = z m g for some integer m ≥ 0 and an irreducible element g in A, where g 6∈ (z).
Proof. Since f is nonzero, we can write f = z m g for some integer m ≥ 0 and an element
g ∈ A that is not in the ideal (z). We need to check that this g is irreducible in A.
If not, then for some nonunits p, q in A, the equality g = pq holds. Thus, f /1 =
(z m /1)(g/1) = (z m /1)(p/1)(q/1). Since f /1 is irreducible in A[z −1 ] and z m is a unit, one
of p/1 and q/1 must be a unit element in A[z −1 ], say p/1, without loss of generality. Thus,
for some r in A and an integer n ≥ 0, we have (p/1)(r/z n ) = 1, that is, pr = z n . Thus,
pr ∈ (z), and z being irreducible either p ∈ (z) or r ∈ (z). But since g 6∈ (z) and g = pq, the
element p must not be in (z). Hence r ∈ (z). Thus, by repeating this argument, we may
assume that n = 0. Then, we have the equality pr = 1 in A, contradicting the assumption
that p is not a unit in A. 
We now prove that A is a UFD. For any nonzero f ∈ A, since the ring A[z −1 ] is a UFD,
we have a factorization of f /1
f u f1 fn
= m m1 · · · mn
1 z z z
for some nonnegative integers m, m1 , · · · , mn , a unit u in A, and f1 , · · · , fn in A, where
fi /z mi are irreducible in A[z −1 ]. Since each z mi is a unit, by replacing m + m1 + · · · + mn
by m, we may simplify the above equation as
f u f1 fn
= m ···
1 z 1 1
where fi are irreducible in A[z −1 ]. Thus, z m f = uf1 · · · fn in A. By the Lemma (2), each
fi = z ri gi for some integer ri ≥ 0 and an irreducible element gi ∈ A, where gi 6∈ (z), so that
z m f = uz r1 +···+rn g1 · · · gn .
Note that we must have m ≥ r1 · · · rn since all gi is not in (z). Thus, f = uz s g1 · · · gn , where
s = r1 + · · · + rn − m ≥ 0, gives a factorization of f into a product of irreducible elements
of A.
To show that this factorization is unique, suppose that we have two such factorizations
g = ug1 · · · gn = vh1 · · · hm ,
where u, v in A are units, and gi , hj in A, 1 ≤ i ≤ n, 1 ≤ j ≤ m, are irreducible. Since
f = ug1 · · · gn is in the ideal (h1 ), for some i, the element gi must be in (h1 ). We may
 3

assume that g1 ∈ (h1 ) so that g1 = h1 k for some k in A. Since g1 is irreducible and h1 is

not a unit (being irreducible), k must be a unit in A. Hence, continuing in this way, by
suitably renumbering them if necessary, we must have m = n and each irreducible element
gi is a unit multiple of hi . This finishes the proof.
 Robin Hartshorne’s Algebraic Geometry Solutions
by Jinhyun Park

Appendix C; The Weil Conjectures

` P
Exercise 5.1. Let X = i Xi . Obviously, then, Nr (X) = i Nr (Xi ) so that
∞ ∞ X
X tr X tr
Z(X, t) = exp( Nr (X) ) = exp( Nr (Xi ) )
r r
r=1 r=1 i
∞ ∞
XX tr Y X tr Y
= exp( Nr (Xi ) ) = exp( Nr (Xi ) ) = Z(Xi , t).
r r
i r=1 i r=1 i

Exercise 5.2. The point is to compute the number of kr = Fqr -rational points, i.e. to
compute Nr in X̄ = Pnk̄ . We can consider the following stratification of Pn :
Pn = An ∪ Pn−1 = · · · = An ∪ An−1 ∪ · · · ∪ A1 ∪ {∗}.
Hence for a field kr of q r elements, Pnkr has Nr = 1 + q r + q 2r + · · · + q nr points. Hence,
∞ r n ∞
n
X
r nr t
Y X (q i t)r
Z(P , t) = exp( {1 + q + · · · + q } ) = exp( )
r r
r=1 i=0 r=1
n n
Y Y 1 1
= exp(− log(1 − q i t)) = i
= .
1−q t (1 − t)(1 − qt) · · · (1 − q n t)
i=0 i=0
Obviously, this is a rational function so that (1.1) is true. Also, we have Podd (t) = 1,
2i
P2i (t) = 1 − q i t so that (1.3) is true, because |q i | = q 2 , indeed.
Let’s find the self intersection number E of ∆ in Pn × Pn . Note that CH ∗ (Pn × Pn ) '
CH ∗ (Pn ) ⊗ CH ∗ (Pn ) so that in particular, we have
CH n (Pn × Pn ) ' ⊕ Zsi tj ,
i+j=n,0≤i,j≤n

where si tj corresponds to an n-cycle Pi × Pj in Pn × Pn . Hence if ∆ = i,j aij si tj , then if

P
0 0
we look at the intersection product, (∆.(Pi × Pj )) = 1. Also, (si tj ).(si tj ) = 1 iff i + i0 = n
and j + j 0 = n and otherwise it is 0. Hence each aij = 1. That is,
X X 0 0
E = (∆.∆) = ( si tj )( si tj ) = n + 1.
i,j i0 ,j 0

Now,
1 1 (q n t)(q n−1 t) · · · (t)
Z(Pn , )= 1 1 =
qnt (1 − q n t )(1 − q n−1 t
) · · · (1 − 1t ) (q n t − 1)(q n−1 t − 1) · · · (t − 1)
n(n+1
n+1 q 2 tn+1 nE
= (−1) n
= (−1)n+1 q 2 tE Z(Pn , t),
(1 − t)(1 − qt) · · · (1 − q t)
so that we have (1.2).
Now, obviously, from Z(Pn , t), we see that Bi = 0 if i is odd and Bi =1 if i is even. Hence
Z i : even
E = n + 1 = 2n i n i n
P
i=0 (−1) Bi , indeed. Also, for PC , we had H (PC , Z) = 0 i : odd
, so
that Bi indeed is rkH i (PnC , Z). This one shows (1.4). Hence Pn satisfies the Weil conjectures.
1
2

Exercise 5.3. Obviously, Nr (X × A1 ) = q r Nr (X). Hence,

∞ ∞
X tr X tr
Z(X × A1 , t) = exp( Nr (X × A1 ) ) = exp( q r Nr (X) )
r r
r=1 r=1
∞
X (qt)r
= exp( Nr (X) ) = Z(X, qt).
r
r=1

Exercise 5.4. Let Nr (x) be the contribution of the closed point x ∈ X to Nr in X. A Fqr -
rational point is determined by the number of morphisms SpecFqr → X, which is the same
as to give a Fq -homomorphism k(x) → Fqr and Nr (x) is the number of all such embeddings.
This is possible iff deg x = [k(x) : Fq ]|r, and, the number of all such embeddings is just
deg x. (If you are confused, consider, say deg x = 1. How many Fq -linear maps are there?
Just 1!.)
Thus, in fact, N (x) = q deg x so that
∞
Y 1 Y 1 Y 1 XX (q −s deg x )r
ζX (x) = −s
= deg x −s
= −s deg x
= exp( )
1 − N (x) 1 − (q ) 1 − (q ) r
x∈X x∈X x∈X x∈X r=1
∞ ∞ ∞
X X (q −s )r deg x X X (deg x)(q −s )r deg x X X Nn (x)(q −s )n
= exp( ) = exp( ) = exp( )
r r deg x n
x∈X r=1 x∈X r=1 x∈X n=1
∞
X (q −s )n
= exp( Nn = Z(X, q −s ).
n
n=1

Exercise 5.5. By the Weil conjectures, since dim H 1 (X, Ql ) = B1 = deg P1 (t) = 2g, for
some αi , we have
Q2g
P1 (t) (1 − αi t)
Z(X, t) = = i=1
(1 − t)(1 − qt) (1 − t)(1 − qt)
∞ 2g ∞
X tr q r tr X r tr X tr
= exp( ( + − αi )) = exp( Nr ).
r r r r
r=1 i=1 r=1
Hence, Nr = 1 + q r − 2g r
P
i=1 αi for all r ≥ 1.
Now, from the functional equation, we have
1
) = ±q 1−g t2−2g Z(t).
Z(
qt
The left hand side of the equation is,
Q2g αi 2−2g q 1−g
Q2g √ αi
t i=1 ( qt − q)
(1 − ) √
i=1 qt
1 = .
(1 − qt )(1 − 1t ) (1 − t)(1 − qt)
Hence, by comparing terms, we have
2g 2g
Y Y √ αi
P1 (t) = (1 − αi t) = ± ( qt − √ ).
q
i=1 i=1
1 −1 b b 2 2
Recall that =
−a+bt so that log(−a + bt) = −( ab t + ( ab )2 t2 + · · · ).
a (1 + a t + ( a ) t + · · · )
Hence, if we replace P1 (t) be the right hand side of above,
∞ 2g ∞ 2g
X tr q r tr X q r tr X
r r
X 1 tr
Z(X, t) = exp( ( + + r )) = exp( (1 + q + q )).
r r αi r αir r
r=1 i=1 r=1 i=1
 3

P2g
Hence, Nr = 1 + q r + q r as well, which was also 1 + q r − 2g
1 r
P
i=1 αri i=1 αi . Since we know
P2g r
N1 , N2 , · · · , Ng , we hence know all of i=1 αi for −g ≤ r ≤ g. Using some combinatorial
argument and Nowton’s P r identity on symmetric polynomials, above
P r information is enough
to determine all r
αi for r > g as well. Hence Nr = 1 + q − αi is determined as well.

Exercise 5.6. From IV, Exercise 4.16, Nr = q r − (f r + f˘r ) + 1. Hence,

∞ ∞
X tr X tr
Z(, t) = exp( Nr ) = exp( (q r − (f r + f˘r ) + 1) )
r r
r=1 r=1

1 1 1 − at + qt2
= (1 − f t)(1 − f˘t) = ,
1 − qt 1 − t (1 − t)(1 − qt)
since f f˘ = q and a = f + f˘ ∈ Z.
Now from the functional equation, we will have

f √ f˘
P1 (t) = (1 − f t)(1 − f˘t) = ±( (q)t − √ )( qt − √ )
p
q q
so that
√
|a| = |f + f˘| ≤ 2g ⇔ |f | = |f˘| = q.
(See Exercise 5.7, (b) and (c).)

Exercise 5.7.

(a). We have
Q2g
P1 (t) (1 − αi t)
Z(X, t) = = i=1
(1 − t)(1 − qt) (1 − t)(1 − qt)
∞
(1 − 2g r r r
P
i=1 αi + q )t
X
= exp( ).
r
r=1
P2g r
P2g
Hence, Nr = 1 − i=1 αi + q r = 1 − ar + q r so that ar = r
i=1 (αi ) .

√
(b). (⇐) If |αi | ≤ q for all i, then,
2g 2g
X
r
X √ √
|αi | ≤ |αi | ≤ q r = 2g q r .
i=1 i=1

(⇒) Consider the following easy power series expansion:

2g ∞
X αi t X
= ar tr .
1 − αi t
i=1 r=1
√ √
Since |ar | ≤ 2g q r , the RHS is holomorphic in |t| < √1 . If |αi | > q for some i, then,
q
the LHS has a pole of order 1 in |t| < √1 at t = 1
which is hence a contradiction.
q αi
4

(c). By using the functional equations, as in Ex (5.5), we have

2g 2g
Y Y √ αi
P1 (t) = (1 − αi t) = ± ( qt − √ ).
q
i=1 i=1
Hence,
α1 α2g
{α1−1 , · · · , α2g
−1
}={,··· , }
q q
α √
so that for all j, there is a unique number i(j) with αj−1 = i(j) q , so that |αi | ≤ q then
−1 |αi | 1 √ √
implies |αj | = q ≤ √q i.e. |αj | ≥ q for all j. This proves that |αi | = q.