Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Algebra Exam – Spring 2007

Alexander J. Wertheim
Last Updated: June 27, 2017

Contents
1 Groups 2
1.1 Problem 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Problem 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Problem 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2 Rings 3
2.1 Problem 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Problem 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3 Problem 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

3 Fields 6
3.1 Problem 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2 Problem 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.3 Problem 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

1
1 Groups
1.1 Problem 1
Let G be a simple group containing an element of order 21, and let H be a proper subgroup
of G with index n := [G : H]. Then G acts by left multiplication on the set G/H of cosets
of H in G. One can view this action as a homomorphism ϕ : G → S(G/H) ∼ = Sn , where
ϕ(g 0 )(gH) = g 0 gH for any g 0 ∈ G, gH ∈ G/H. Since ker(ϕ) is a normal subgroup of G,
it must either be trivial or all of G. For any g ∈ / H (such a g exists since H is a proper
subgroup of G), ϕ(g)(H) = gH, so ker(ϕ) cannot be all of G. Hence, ϕ is injective, so G is
isomorphic to a subgroup of Sn .
Since G has an element of order 21, Sn must also have an element σ of order 21. The
order of any element of Sn is the least common multiple of the cycle lengths in its disjoint
cycle decomposition. The nontrivial factors of 21 are 3, 7 and 21, so the disjoint cycle
decomposition of σ must contain a cycle whose length is a multiple of 3 and a cycle whose
length is a multiple of 7. If this is the same cycle, then the length of this cycle is a multiple
of 21, so n > 21; if these are two different (disjoint) cycles, then n > 3 + 7 = 10, so in either
case, n > 10.

1.2 Problem 2
Consider instead the more general problem of counting the subgroups of Zn of index p
for some prime p. Since Zn is abelian, every subgroup of Zn is normal. Hence, let G
be a (normal) subgroup of Zn of index p. Then the quotient Zn /G is a finite group of
order p, hence cyclic of order p, so choosing an isomorphism Zn /G → Z/pZ, the quotient
map Zn → Zn /G determines a surjective homomorphism Zn → Z/pZ with kernel G by
composition. If Ωp denotes the set of surjective homomorphisms Zn → Z/pZ, and Γp denotes
the set of subgroups of Zn of index p, this shows there is a surjective map Φ : Ωp → Γp defined
by Φ(α) = ker(α).
Now, let H = Aut(Z/pZ) ∼ = (Z/pZ)× ∼ = Z/(p − 1)Z. For any α ∈ Ωp and f ∈ H, the
morphism f ◦ α is surjective, since it is the composition of two surjective maps, so f ◦ α ∈ Ωp .
Hence, H acts on Ωp by postcomposition, so we can consider the orbit space Ωp /H, which
consists of equivalence classes of Ωp under the action of H (where two elements of Ωp are
equivalent if they belong to the same orbit under the action of H). Since every f ∈ H is
injective, ker(f ◦α) = ker(α) for every α ∈ Ωp . Hence, Φ descends to a well-defined surjective
map Φ̃ : Ωp /H → Γp . I claim that Φ̃ is injective, hence a bijection.
Indeed, suppose α, β ∈ Ωp such that ker(α) = Φ(α) = Φ(β) = ker(β); we need to show that
α = f ◦ β for some f ∈ H. Indeed, put G = ker(α) = ker(β), and let πG : Zn → Zn /G
be the canonical quotient morphism. By the universal property of the quotient, there are
isomorphisms σ, τ : Zn /G → Z/pZ such that α = σ ◦ πG , β = τ ◦ πG . Then α = σ ◦ τ −1 ◦ β,
so putting f := σ ◦ τ −1 ∈ H proves the claim.
Hence, Φ̃ is a bijection, whence |Γp | = |Ωp /H|. The orbit of any element α ∈ Ωp has exactly
|H| elements, so |Ωp /H| · |H| = |Ωp |, i.e. |Ωp /H| = |Ωp |/|H|. Since Zn is a free abelian group
with Z-basis given by the standard idempotents ei = (0, . . . , 0, |{z} 1 , 0, . . . , 0), any group
ith place

2
homomorphism α : Zn → Z/pZ is completely determined by the images α(e1 ), . . . , α(en ) ∈
Z/pZ. Hence, Hom(Zn , Z/pZ) ∼ = (Z/pZ)n as sets, so | Hom(Zn , Z/pZ)| = pn . Moreover,
since every nonzero element of Z/pZ is a generator, a homomorphism α : Zn → Z/pZ is
not surjective if and only if α(ei ) = [0]p for each i = 1, . . . , n, i.e. if and only if α is
the zero map. Hence, every nonzero element of Hom(Zn , Z/pZ) is surjective, so |Ωp | =
| Hom(Zn , Z/pZ)| − 1 = pn − 1. Since |H| = p − 1, this gives

|Ωp | pn − 1
|Γp | = =
|H| p−1

In particular, we were originally asked to find |Γ5 | = (5n − 1)/4.

1.3 Problem 3
This is proved in F06.G1.

2 Rings
2.1 Problem 1
First, some preliminary observations. Let D be a division ring, and let Mn (D) denote the
ring of n × n matrices with entries in D. Let Sn denote the symmetric group on {1, 2, . . . , n},
and consider a permutation τ ∈ Sn . For each i = 1, . . . , n, let vτ,i be the column vector of
length n whose entries are all zero except for the τ (i)th entry, which is a 1. Let Aτ ∈ Mn (D)
be the matrix with columns vτ,1 , . . . , vτ,n . Of course, this is just the regular representation
Sn → GLn (D) induced by the natural action of Sn on Dn .
For any X ∈ Mn (D), it is easy to see that left multiplication by Aτ permutes the n rows of
X according to τ , and right multiplication by Aτ permutes the n columns of X according
to τ . As an example, consider n = 2, and let τ = (1, 2) ∈ S2 . Then
 
0 1
Aτ =
1 0

so for X = (xij ) ∈ M2 (D), we see


    
0 1 x11 x12 x21 x22
Aτ X = =
1 0 x21 x22 x11 x12

and     
x11 x12 0 1 x12 x11
XAτ = =
x21 x22 1 0 x22 x21
as expected. Now, let J ⊂ Mn (D) be a two-sided ideal; I claim that J = 0 or J = Mn (D).
It suffices to show that if J is nonzero, then J contains the identity matrix In of Mn (D).
Hence, suppose J contains a nonzero element A, with entries aij for 1 6 i, j 6 n.
By left and right-multiplying by the appropriate permutation matrices, we can assume that
the entry a11 of A is nonzero, since A is nonzero. Then let X = (xij ) ∈ Mn (D) be the matrix

3
with x11 = a−1 2
11 , xij = 0 if i 6= 1 or j 6= 1, and for any (i, j) ∈ {1, . . . , n} , let Eij ∈ Mn (D)
be the matrix whose i, j entry is 1 and all other entries are 0. Then it is straightforward to
see that XAE11 = E11 . Indeed, for any 1 6 i, j 6 n
n
(
X a−1
11 a1j if i = 1
(XA)ij = xik akj =
k=1
0 otherwise
so
n
(
X (XA)i1 if j = 1
(XAE11 )ij = (XA)ik (E11 )kj =
k=1
0 otherwise
(
1 if i = 1, j = 1
=
0 otherwise

Hence, E11 ∈ J. Left/right-multiplying by the appropriate permutation matrices, we can


obtain Eij for any pair (i, j) ∈ {1, 2, . . . , n}2 , so in particular Eii ∈ J for all i = 1, . . . , n.
Hence, In = E11 + · · · Enn ∈ J, as desired.
N.B.: one can also approach this problem by showing that for any ring R, the two-sided
ideals of Mn (R) are of the form Mn (I) for some (unique) two-sided ideal I of R. Having
proved this, the result is immediate: the two-sided ideals of any division ring D are only {0}
and D. For this approach, see F10.NCR1; the main ideas of the proof are (unsurprisingly)
the same.

2.2 Problem 2
Let R = End(V ), where V is an infinite-dimensional C-vector space with countable basis
{e1 , e2 , . . .}. For any ϕ ∈ R, define ϕ0 , ϕ1 ∈ R on basis vectors by ϕ0 (ei ) = ϕ(e2i ), ϕ1 (ei ) =
ϕ(e2i−1 ) for all i ∈ N. Then there is a map ∆ : R → R⊕R given by ∆(ϕ) = (ϕ0 , ϕ1 ), and it is
straightforward to check that ∆ is a morphism of left R-modules. Suppose ∆(ϕ) = ∆(ρ) for
some ϕ, ρ ∈ R. Then ϕ(e2i ) = ρ(e2i ) and ϕ(e2i−1 ) = ρ(e2i−1 ) for all i ∈ N, so ϕ(ei ) = ρ(ei )
for all i ∈ N. Hence, ϕ and ρ agree on the chosen basis for V , and hence must be equal, so
∆ is injective. On the other hand, let (α, β) ∈ R ⊕ R. Define ϕ ∈ R on basis elements of V
by (
α(ei/2 ) if i is even
ϕ(ei ) =
β(e(i+1)/2 ) if i is odd
Then by construction, ∆(ϕ) = (α, β), whence ∆ is surjective, hence an isomorphism.

2.3 Problem 3
(a) It’s a bit unclear what this question wants, since it says one can use the Nullstellensatz,
but one formulation of the Nullstellensatz (among many) is precisely the description of
maximal ideals of F [X1 , . . . , Xn ] for F an algebraically closed field! Nevertheless, maybe
here is a fair response.
First, let F be an arbitrary field. I claim that for any tuple (a1 , . . . , an ) ∈ F n , the ideal

4
M := hX1 − a1 , . . . , Xn − an i ⊂ F [X1 , . . . , Xn ] is maximal, with F [X1 , . . . , Xn ]/M ∼
= F.
In particular, M is the kernel of the surjective morphism of F -algebras ϕ : F [X1 , . . . , Xn ] →
F defined on generators by Xi 7→ ai .
We proceed by induction on n, so suppose n = 1. Clearly, M ⊆ ker(ϕ), so it suffices to
show that ker(ϕ) ⊂ M . Suppose g ∈ ker(ϕ). Since F [X1 ] is a Euclidean domain, we can
perform Euclidean division on g by X1 − a1 to write

g(X1 ) = q(X1 )(X1 − a1 ) + r(X1 )

for some q(X1 ), r(X1 ) ∈ F [X1 ] with deg(r(X1 )) < deg(X1 − a1 ) = 1, so in fact r(X1 ) ∈
F ⊂ F [X1 ]. Since
ϕ(g) = ϕ(r) = r(a1 ) = 0
we must have r = 0, whence g ∈ M , so M = ker(ϕ), and F [X1 ]/M ∼ = F , as desired.
Now suppose n > 1, letting M = hX1 − a1 , . . . , Xn − an i ⊂ F [X1 , . . . , Xn ]. Consider
the surjective morphism of F -algebras ϕn : F [X1 , . . . , Xn ] → F [Y1 , . . . , Yn−1 ] defined on
generators by Xi 7→ Yi for 1 6 i < n, and Xn 7→ an . Then as above, hXn −an i ⊆ ker(ϕn ),
so suppose g ∈ ker(ϕn ). Though F [X1 , . . . , Xn ] is not a Euclidean domain for n > 1,
we can use a neat trick: via the isomorphism F [X1 , . . . , Xn ] ∼ = (F [X1 , . . . , Xn−1 ])[Xn ],
consider g as a polynomial in Xn with coefficients in F [X1 , . . . , Xn−1 ]. Since the leading
coefficient of Xn −an is a unit in F [X1 , . . . , Xn−1 ], we can perform the standard Euclidean
division procedure on g by Xn − an to write

g(X1 , . . . , Xn ) = q(X1 , . . . , Xn )(Xn − an ) + r(X1 , . . . , Xn )

for some q(X), r(X) ∈ F [X1 , . . . , Xn ] such that the degree of r in Xn is less than
degXn (Xn − an ) = 1, so in fact r ∈ F [X1 , . . . , Xn−1 ] ⊂ F [X1 , . . . , Xn ]. Since

ϕn (g) = ϕn (r) = r(Y1 , . . . , Yn ) = 0

we must have r = 0, whence g ∈ hXn − an i, so hXn − an i = ker(ϕn ). Hence, ϕn descends


to an isomorphism of F -algebras ϕ̃n : F [X1 , . . . , Xn ]/hXn − an i → F [Y1 , . . . , Yn−1 ] with
ϕ̃n (Xi ) = Yi for each 1 6 i < n. Now, only fussy details remain. We have a sequence of
isomorphisms

F [X1 , . . . , Xn ]/M ∼
= (F [X1 , . . . , Xn ]/hXn − an i)/hX1 − a1 , . . . , Xn−1 − an−1 i

= F [Y1 , . . . , Yn−1 ]/hY1 − a1 , . . . , Yn−1 − an−1 i

By induction, the surjective morphism of F -algebras ρ : F [Y1 , . . . , Yn−1 ] → F defined on


generators by Yi 7→ ai has kernel hY1 −a1 , . . . , Yn−1 −an−1 i, and so descends to an isomor-
phism ρ̃ : F [Y1 , . . . , Yn−1 ]/hY1 − a1 , . . . , Yn−1 − an−1 i → F satisfying ρ̃(Yi ) = ai . Compos-
ing ρ̃ with the isomorphism F [X1 , . . . , Xn ]/M → F [Y1 , . . . , Yn−1 ]/hY1 − a1 , . . . , Yn−1 −
an−1 i indicated above, we obtain an isomorphism ϕ̃ : F [X1 , . . . , Xn ]/M → F satisfy-
ing ϕ̃(Xi ) = ai . Let π : F [X1 , . . . , Xn ] → F [X1 , . . . , Xn ]/M denote the canonical quo-
tient morphism. Then ϕ̃ ◦ π = ϕ, since they agree on the generators X1 , . . . , Xn , so
ker(ϕ) = ker(ϕ̃ ◦ π) = ker(π) = M , since ϕ̃ is an isomorphism.
With this out of the way, we can characterize the maximal ideals of F [X1 , . . . , Xn ] for

5
any algebraically closed field F . First, we recall one precursor to the Nullstellensatz,
namely Zariski’s lemma. (At some point, maybe I’ll include a proof of Zariski’s lemma
here, but for now, I’ll just list a standard reference: see, e.g., Atiyah-Macdonald, Theo-
rem 7.9). Zariski’s lemma says that if F is a field and E is a finitely generated F -algebra
which is also a field, then E is a finite extension of F .
Hence, let F be an algebraically closed field, and let M ⊂ F [X1 , . . . , Xn ] be a maximal
ideal. Applying Zariski’s lemma to the finitely generated F -algebra E = F [X1 , . . . , Xn ]/M ,
we deduce that E is a finite field extension of F , whence E ∼ = F since F is algebraically
closed. Let α : F [X1 , . . . , Xn ]/M → F be an isomorphism, and put ai := α(Xi ); then
the same argument as above shows that M = hX1 − a1 , . . . , Xn − an i. Hence, every
maximal ideal of F [X1 , . . . , Xn ] is of the form hX1 − a1 , . . . , Xn − an i for some tuple
(a1 , . . . , an ) ∈ F n . Since we proved above that hX1 − a1 , . . . , Xn − an i is a maximal
ideal for any tuple (a1 , . . . , an ) ∈ F n , this completely characterizes maximal ideals of
F [X1 , . . . , Xn ].
To actually answer the question, take F = C, n = 2. The above shows the maximal
ideals of C[X, Y ] are the ideals of the form hX − a, Y − bi for any a, b ∈ C.
(b) Let M = hX 2 − Y, Y 2 − 5i ⊂ Q[X, Y ]. Consider the surjective morphism of Q-algebras
α : Q[X, Y ] → Q[Z] defined on generators by α(X) = Z, α(Y ) = Z 2 . Clearly hX 2 −Y i ⊂
ker(α); I claim ker(α) = hX 2 − Y i. Indeed, let g(X, Y ) ∈ ker(α). Via the isomorphism
Q[X, Y ] ∼
= (Q[X])[Y ], we can view g as a polynomial in Y with coefficients in Q[X].
Since the leading coefficient of X 2 − Y is a unit in (Q[X])[Y ], we can perform Euclidean
division on g by X 2 − Y to write
g(X, Y ) = q(X, Y )(X 2 − Y ) + r(X, Y )
for q, r ∈ Q[X, Y ], where the degree of r in Y is less than degY (X 2 −Y ) = 1, i.e r ∈ Q[X].
Then as g ∈ ker(α),
α(g) = α(q)α(X 2 − Y ) + α(r) = α(r) = r(Z) = 0
Hence, r = 0, so g ∈ hX 2 −Y i. Then α descends to an isomorphism Q[X, Y ]/hX 2 −Y i →
Q[Z] which sends the equivalence class Y to Z 2 . In particular, we have a series of
isomorphisms
Q[X, Y ]/M ∼ = (Q[X, Y ]/hX 2 − Y i)/hY 2 − 5i ∼
= Q[Z]/hZ 4 − 5i
Since Z 4 − 5 is irreducible over Q[Z] by Eisenstein at 5, Q[Z]/hZ 4 − 5i is a field, and
thus Q[X, Y ]/M is a field as well. Hence, M is maximal.

3 Fields
3.1 Problem 1
Let F = Q(ζ), where ζ is a primitive 5th root of unity, and let E/F be a cyclic Galois
extension of degree 5. Let G := Gal(E/F ) = hσi, and let β ∈ E \ F . Put
4
X
α := ζ −i σ i (β) ∈ E
i=0

6
Since σ fixes F , σ(ζ k ) = ζ k for any k ∈ Z, so
4
X 4
X 4
X
−i i+1 −(i+1) i+1
σ(α) = ζ σ (β) = ζ ζ σ (β) = ζ ζ −j σ j (β) = ζα
i=0 i=0 j=0

where the second to last equality is obtained because ζ 5 = 1, σ 5 = IdE . Thus, σ i (α) = ζ i α
for each i = 0, 1, . . . , 4, and ζ i α 6= ζ j α for any i 6= j ∈ {0, 1, . . . , 4}, so the Galois orbit S of α
is given by the set {α, ζα, . . . , ζ 4 α} of 5 distinct elements. The minimal polynomial mα (X)
of α over F is then given by

Y 4
Y
mα (X) = (X − ξ) = (X − ζ i α) = X 5 − α5 ∈ F [X]
ξ∈Gα i=0

so F (α) is a subfield E of degree 5 over F , and hence must be equal to E. Put η = α5 ∈ F .


√ √
Then 5 η is (by convention) some root of mα (X) = X 5 − η, so 5 η = ζ i α for some i ∈

{0, 1, . . . , 4}, and F ( 5 η) = F (ζ i α) = F (α) = E.

3.2 Problem 2
√ √ 2
Let K = Q( 3, 7 5). Note that√ X − 3 ∈ Q[X] is irreducible
√ by Eisenstein at 3,√so X 2 − 3 is
√ polynomial of 3 over√
the minimal Q, whence [Q( 3) : Q] = 2. Likewise, [Q( 7 5) : Q] = 7;
since [Q( 3) : Q] | [K : Q] and [Q( 7 5)√: Q] | [K : Q],√we must have gcd(2, 7) = 14 | [K : Q].
Since [K : Q] is certainly less than [Q( 7 5) : Q] · [Q( 3) : Q], we must have [K : Q] = 14.

(a) Certainly, K contains the subfield Q( 3), which as noted above √ is quadratic over √ Q.
Suppose F ⊂ K is a subfield with [F : Q] = 2 distinct from Q( 3). Since F ∩ Q( 3) is
a subfield of√(say) F , it must have
√ degree dividing 2 over Q; if it has degree √ 2 over Q,
then F ∩ Q( 3)√ = F , i.e. F = Q( 3), a contradiction, so we must have F ∩√Q( 3) = Q.
Since F and Q( 3) √ are both Galois extensions of Q, the compositum F Q( 3) ⊂ K has
degree [F : Q][Q( 3) : Q] = 4 over Q. This does not divide [K : Q] = 14, a contradic-
tion.
N.B.: we can argue more directly without appealing to facts about the degree of composi- √
tums of Galois extensions. Indeed, if F is a√ quadratic extension
√ of Q,
√ then F = Q( d)
for some squarefree d ∈ Z. Since √ F√ 6= Q( 3), we have d ∈ / Q( 3) by degree con-
siderations, as √
above. Then √ Q( 3, d) is a subfield √ of K of√degree 4 over Q, because
the degree of d over Q( 3) is 1 if and only if d ∈ Q( 3). This gives the same
contradiction as above.

(b) Every nontrivial subfield of K over Q has degree properly dividing [K : Q] = 14. In part
(a), we noted that K has a unique subfield of degree 2 over √Q, so it remains to classify
all subfields of K of degree 7 over Q. As noted above, Q( 7 5) is one such subfield; I
claim this is the unique such subfield. √ √
Suppose F √⊂ K with [F : Q] = √ 7 and F 6= Q( 7 5). Then 7 5 ∈ / F , or
√ else F would
7 7 7
contain Q( 5), hence equal Q( 5) by degree considerations. Then F ( 5) is a proper
extension of F contained in K, and so by degree considerations must equal K. In

7
√ √
particular, 2 = [K : F ] = [F ( 7 5) : F ], so the minimal polynomial f (X) of 7 5 over√F
has degree 2. Further, f (X) must divide X 7 − 5 ∈ Q[X] ⊆ F [X], since X 7 − 5 has 7 5
as a root. Over C, X 7 − 5 factors as
6
5
Y √
(X − ξ i 5)
7
X −7=
i=0

where ξ is a primitive 7th root of unity. Hence, since f (X) has 7 5 as a root, and f
divides X 7 − 5, it must factor over C as
√ √ √ √ 2
f (X) = (X − 5)(X − ξ i 5) = X 2 − (1 + ξ i ) 5X + ξ i 5 ∈ F [X]
7 7 7 7

for some √ 1 6 i 6 6. √Since all of the coefficients of f belong to F , we must have


(1 + ξ i ) 7 5 ∈ F . Since 7 5 ∈ K, this implies 1 + ξ i ∈ K, i.e. ξ i ∈ K. Since gcd(i, 7) = 1,
there exist r, s ∈ Z such that 7r + si = 1, so ξ = ξ 7r+si = (ξ i )s , i.e. ξ ∈ K. But
[Q(ξ) : Q] = ϕ(7) = 6, which does not divide [K : Q] = 14, a contradiction.
√ √
(c) Let u = 3 + 7 5 ∈ K. I claim that K = Q(u); to show this, it suffices to prove
that
√7
u√does not belong to any proper subfield√
7
√ Certainly, u ∈
of K. / Q; if it were, then
5√= 3 + q for √ some q ∈ Q, whence Q( 5) = Q( 3). This is a √ contradiction, because
Q( 7 5) and Q( 3) have √ distinct
√ degrees over Q. Note that u ∈
/ Q( 3), because if it were,
7
then we√would have 5 ∈ Q( 3), √ a contradiction
√ by degree considerations. Likewise,
u∈ / Q( 7 5), since we would have 3 ∈ Q( 7 5). By parts (a) and (b), these are all proper
subfields of K over Q, so this proves our claim.
√ i√ j
(d) Note that K has a Q-basis given by { 3 7 5 } for 0 6 i 6 1, 0 6 j 6 6, so every element
of Q can be described by a sequence of coefficients (ai,j ) ∈ Q14 for 0 6 i 6 1, 0 6 j 6 6.
We can classify the elements of K which generate K in terms of this list of coefficients.
Indeed, let u ∈ K, and let (ai,j ) √ be the coefficients of u with respect to the prescribed
Q-basis of K. Note that√u ∈ Q( 3) if and only if ai,j = 0 for all pairs (i, j) such that
j > 1; likewise, u ∈ Q( 7 5) if and only if ai,j = 0 for all pairs (i, j) with i = 1. Most
obviously, u ∈ Q if and only if ai,j = 0 for all pairs (i, j) with i > 1 or j > 1. Otherwise,
as reasoned in part (c), u must generate K.

3.3 Problem 3
Let F = Z/3Z, and put f (X) = X 2 − 2, g(X) = X 2 − 2X − 1 ∈ F [X]. It is easy to check
that f, g have no roots over F , whence f, g are irreducible over F since deg(f ) = deg(g) = 2.
Since F [X] is a PID, the ideals hf (X)i, hg(X)i ⊂ F [X] are maximal. Hence, F [X]/hf (X)i
is a field, with F -basis {1, X}, and likewise the same can be said for F [X]/hg(X)i. Then
F [X]/hf (X)i and F [X]/hg(X)i are both fields with 9 elements. Since there is one finite field
of a given order up to isomorphism, we thus conclude F [X]/hf (X)i ∼ = F [X]/hg(X)i.
To exhibit an explicit isomorphism, consider the morphism of F -algebras ϕ : F [X] → F [X]
defined on generators by ϕ(X) = X − 1. Then ϕ(f (X)) = (X − 1)2 − 2 = X 2 − 2X + 1 − 2 =
X 2 −2X −1 = g(X). Hence, composing ϕ with the canonical quotient morphism πg : F [X] →
F [X]/hg(X)i gives a surjective morphism πg ◦ϕ : F [X] → F [X]/hg(X)i whose kernel contains

8
hf (X)i. Since hf (X)i is a maximal ideal of F [X] and πg ◦ϕ is nonzero, the kernel of πg ◦ϕ must
be exactly hf (X)i, whence πg ◦ ϕ descends to an isomorphism F [X]/hf (X)i → F [X]/hg(X)i
of F -algebras. Explicitly, the isomorphism sends the generator X to X − 1.

You might also like