Algebra Qualifying Exam Solutions

Fall 2016 - Spring 2019

Yizhou Chen

Preface:
This is a little solution manual of the 6 most recent past algebra quals I have written up. If
you want to see the problems, you can find them on the our math department website. If
you are interested in older problems and their solutions, two other graduate students Ian
Coley and Yacoub Kureh have put their solutions on the internet. These solutions are
made possible by the help and support of my dear friends and colleagues Liao Wang, Weiyi
Liu, Yudong Qiu, and other algebra enthusiasts on Stackexchange. I have attached some
links as references. If you find them incomplete or want to add some other resources, please
let me know. Sometime I tend to explain things in a fast and confusing manner. So please
don’t hesitate to contact me when you find some ideas baffling or questionable. Good luck
to all who reasons in an algebraic way!

Useful Links:
Past Exams:
https://secure.math.ucla.edu/gradquals/hbquals.php
Ian’s Solutions:
https://www.math.ucla.edu/~iacoley/hw/algqual.pdf
Yacoub’s Solutions:
https:
//www.math.ucla.edu/~ykureh/Algebra_Qualifying__Quals__Exams_Solutions.pdf

Department of Mathematics
University of California, Los Angeles
U.S
Algebra Quals Solutions Yizhou Chen September 8, 2019

Fall 2016:

Solution 1. That’s just free product of two copies of Z/2

Solution 2. I will just calculate the normalizers by brute force.

So let {1, x} be a basis of K over Q. We have a relation x2 + ax + b = 0, a, b ∈ Q. Let M
be an element of the normalizer. Suppose

M : 1 7→ u1 + u2 x

x 7→ u3 + u4 x
It must commute with the operation of multiplying with x. Thus the following two maps
must have the same result:
M ·x
1 7→ u1 + u2 x 7→ u1 x − au2 x − u2 b
·x M
1 7→ x 7→ u3 + u4 x
Thus we must have u3 + u4 x = (u1 − au2 )x − bu2 ⇒ u3 = u1 − au2 , u4 = −bu2 . But then M
acting on x would be exactly multiplying x with u1 + u2 x. Thus M ∈ ρ(K × ) ⇒ the index
is 1.

Solution 3. This solution may be an overkill.

Lemma: Let M be A module. Then M is projective if and only if ∃{mi }i ⊂ M and
fi : M → A homomorphism of A-module such that Σi fi (m)mi makes sense and Σi fi (m)mi =
m, ∀m ∈ M .
Proof : First we can take {mi } a set of generators of M . Then there is map g : F → M
such that ei 7→ mi where F is a free module and {ei } is a set of basis of F . Then M is
projective if and only if ∃f : M → F such that gf = id. Notice the data of this f is exactly
fj : M → A given by mi 7→ aij where f (mi ) = Σj aij ej . Thus the existence of f and the
existence of the described {fj } are equivalent.
⇐: Now we know ab = A ⇒ ∃a1 , ......an ∈ a, b1 , ......bn ∈ b such that Σai bi = 1. We know
that ∀a ∈ a, b ∈ b, ab ⊂ ab = A ⇒ ab ∈ A. Thus ∀x ∈ a, x = x(Σai bi ) = Σ(bi x)ai ⇒ a =
(a1 , ......an ) as A-module hence finitely generated. For the projectiveness we use the lemma
and define fj : a → A given by a 7→ aj bj a which makes sense since ai bi ∈ A.
⇒: Let a be generated by a1 , ......an . By the lemma we have fj : a → A such that Σj fj (ai ) =
a
ai . Let fj (ai ) = ai,j . Define b be the submodule of K generated by { aii,j } . Then we are
aj j
done.


Reference: http://www.math.uchicago.edu/∼ may/MISC/Dedekind.pdf

2
Algebra Quals Solutions Yizhou Chen September 8, 2019

Solution 4 (Stackexchange). Let the group be hr, s, rp = s2 = 1, srs = r−1 i. There 2

kinds of 1 dim complex representations as below:
r 7→ 1, s 7→ ±1
2p−2
There are 4
different kinds of 2 dimensional representations as shown below:
 2πk
cos p −sin 2πk
  
p 0 1
r 7→ , s 7→
sin 2πk
p
cos 2πk
p
1 0
where k ∈ Z, 2p−2
4
≥ k ≥ 1. You can check to see if the dimension is right.

Solution 5. By the fundamental theorem of Galois theory, it suffices to find a subgroup
of Gal(K/F ) of order p. This is guaranteed by the generalized first Sylow theorem, since
p|[K : F ]

Solution 6. ∀y ∈ Fp , (x + y)p − a = xp − a ⇒ If x0 is a root of the given polynomial, then
x0 + y is also a root.

1
Solution 7. Suppose for contradiction that ∃ζ primitive nth root of unity such that 2 ∈ 4
1
Q(ζ). Then we consider Gal(Q(ζ)/Q(2 4 )) ⊂ Gal(Q(ζ)/Q), the latter of which is cyclic. So
the first group is normal subgroup of Gal(Q(ζ)/Q). By the fundamental theorem of Galois
1
theory, Q(2 4 )/Q is Galois extension. Now we have contradiction because x4 − 2 have root
1
over Q(2 4 ) but does not split.

Solution 8. Let the left adjoint of F be G. Let {∗} be the set of one object. Then we know
for B ∈ Ob(C), F B = M orSets ({∗}, F B) = M orC (G{∗}, B) ⇒ F = M orC (G{∗}, )

Solution 9. The functor is given by M orF −commalg (F [X, Y ]/(XY − 1), )

Solution 10 (by Qiu). Consider B as a A-mod by left multiplication. Then A simple
⇒ B is semisimple as A-mod. We have the following lemma:
Lemma: A has only one isomorphism class of simple module.
Proof : Since A is of finite dimensional hence artinian, we can take I, a minimal left ideal
of A. ∀a ∈ A, Ia is a left ideal of A. By the minimality of I the map I → Ia given by
x 7→ xa is isomorphism or 0 map. Now by the simpleness of A (no two sided ideals) we
have A = IA = Σa∈A Ia. Thus A as a A-module must be a sum of copies of I’s. Any simple
module of A as a quotient of A must be isomorphic to I.
Now back to the problem, let B = I k , A = I n . We consider B as a A, A bimodule. Then
we have B = B ⊗A A = B ⊗A (I n ) = (B ⊗A I)n . Let B ⊗A I = I l then B = (I n )l = Al as
A-module. Thus dimF (A)|dimF (B)


3
Algebra Quals Solutions Yizhou Chen September 8, 2019

Spring 2017:

Solution 1. There are three cases:

1. We have multiples of I, the identity
 matrix.

λ 1
2. We have matrices which look like .
0 λ
3. The last case we must have characteristic polynomial and minimal polynomial are the
same. So you
 can write out the rational canonical form (or companion matrix?) which look
0 a
like such that the polynomial x2 + ax + b does not have multiple roots.
1 b

Solution 2. There are five conjugacy classes in this group: |h1i| = 1, |hyi| = 4, |hxi| =
|hx2 i| = |hx3 i| = 5. So we want to fill out the following chart:
G h1i, 1 hyi, 4 hxi, 5 hx2 i, 5 hx3 i, 5

χ1 1 1 1 1 1

Let’s start by calculating the one dimensional representations. Let ζ, ξ be primitive 4th,
5th root of unity respectively. Suppose φ is a one dimensional representation. Then
φ(y) = φ(y 2 ) ⇒ φ(y) = 1 or 0. But y 5 v = 1 ⇒ φ(y) = 1. But we can have φ(x) = ζ i , i ∈ Z.
So we can have some extra rows. Use Schur’s orthogonality (mentioned in the second to last
solution in this whole set) we know that all other entries for hxi, hx2 i, hx3 i are 0. By the first
Schur’s orthogonality we know that there is only one row left starting with 4. So below is
our chart:
G h1i, 1 hyi, 4 hxi, 5 hx2 i, 5 hx3 i, 5

χ1 1 1 1 1 1

χ2 1 1 i −1 −i

χ3 1 1 −i −1 i

χ4 1 1 −1 1 −1

χ5 4 −1 0 0 0

Solution 3. One way of doing this is that suppose H ⊂ F2 a subgroup of index 3. Then
F2 acting on the left cosets gives us a map F2 → S3 . The map is uniquely determined by

4
Algebra Quals Solutions Yizhou Chen September 8, 2019

the image of u, v. So there are in total 36 possible maps. Notice the map correspond to a
previous subgroup if and only if its image is a transitive subgroup. So we have 26 choice
left. Notice if you have cosets H.xH.yH; H, yH, xH correspond to the same subgroup H but
the second order correspond to a different map. So our answer would be to divide 26 by 2,
which gives us 13.
Another way of thinking about this problem is to use algebraic topology. You have to
consider possible choices of 3 sheeted coverings of two circles wedged together. Namely, if
you have three points and 2 lines connecting each pair of two, what orientation can you put
on the lines under certain rules?

Solution 4. I am not so good with Dedekind domains. So feel free to ignore what I will
say. One characterisation of Dedekind domain is noetherian integrally closed domain of
Krull dimension 1. Noetherianness is easy. Krull dimension 1 is by Krull’s Hauptidealsatz.
Domain because of the irreducibility of the polynomial. The only thing left is the integrally
closedness. A noetherian domain is integrally closed if and only if the ring is intersection of
all its localization at prime ideals of height 1 and all the previous localizations are discrete
valuation rings. The first condition is easy, because as long as the denominator is nonunit,
you can include it in a maximal ideal. The second condition, a DVR is the same as a local
PID. height 1 restricts to the number of generator to 1. So we are good.

Solution 5. This is elementary algebraic geometry. You can do it.

Solution 6. Let {x1 , ......xn } be a set of generators of M over R with minimal cardinality.
Then we want to reduce its cardinality. Take u ∈ M , since M = JM , we have u = Σs js us
for us ∈ M, js ∈ J. Then we can write us = Σl rs,l xl . By the fact that the Jacobson
radical is two sided, we have u = Σs ks xs for ks ∈ J. Now replace u with x1 we have
(1 − k1 )x1 = Σs>1 js xs . Now by the characterisation J = {r ∈ R, 1 + xr ∈ R× , ∀x ∈ R}, we
have 1 − x1 ∈ R× ⇒ x1 ∈ Rx2 + ......Rxn ⇒ contradiction to the minimality of the previous
set of generator.


√ √
Solution 7. The
√ roots of the polynomial
√ √ √ are ζ ·3 2 ± 3 where ζ is the third root of unity.
So K = Q(ζ,3 2, 3) = Q(i,3 2, 3) ⇒ [K : Q] = 12

Solution 8. The functor is representable, it is M orRingsop (EndZ (M ), )

5
Algebra Quals Solutions Yizhou Chen September 8, 2019

Solution 9. The maps f : Rn → Rm , g : Rm → Rn that are inverses of each other corre-

spond to matrices A, B with entries in R such that AB = Im , BA = In . You can construct
right module maps with the same matrices. Thus they are isomorphic as right R modules.

Solution 10. The automorphism of K H over F still extends to K, so our answer would be
Gal(K H /F ) = NH (G)/H

6
Algebra Quals Solutions Yizhou Chen September 8, 2019

Fall 2017:

Solution 1. So we have S ⊂ CG (X) ⇒ gSg −1 ⊂ CG (gXg −1 ) = CG (Y ). Also we have

S ⊂ CG (Y ). S, gSg −1 are Sylow subgroups of CG (Y ). Thus ∃a ∈ CG (Y ), aSa−1 = gSg −1 ⇒
a−1 g ∈ N . Let n = a−1 g then nxn−1 = a−1 gxg −1 a = gxg −1 .


Reference: https://math.stackexchange.com/questions/3327435/stabiliser-of-a-subset-of-center

Solution 2. This problem is called Burnside’s basis theorem and our Φ(G) is called the
Frattini subgroup. So first we notice that every maximal proper subgroup of G is of index
p. This is because any group of order p2 is either Z/p2 or some semidirect product of two
Z/p hence must have a proper subgroup. Then for a maximal proper subgroup N , we have
G/N = Z/p. Then the quotient is abelian ⇒ all the commutators are inside N . Every
element is of order p inside the left coset ⇒ all elements of the form g p for g ∈ G is also
inside N . So now we know Φ(G) ⊂ ∩Hmaximal proper subgroup H. Now for the equal sign, notice
that G/Φ(G) is a characteristic subgroup of G with each element of order p. Thus we can
consider it as a Z/p-vector space. And each maximal subgroup correspond to a hyperplane.
Since the intersection of all hyperplanes is the origin, we have the desired result.

Solution 3. We know that M/JM is a A/J module and A/J is semisimple ring ⇒ M/JM
is semisimple as A-module. We also know JM ⊂ ker(t) ⇒ JM = 0 ⇒ M is semisimple.
Also M is left ideal because ∀x ∈ J, y ∈ A, a ∈ M, t(xya) = 0 since J is a two sided ideal.
Now suppose ∃N semisimple left ideal of A then JN = 0 ⇒ JN ⊂ ker(t) ⇒ t(xb) = 0, ∀x ∈
J, ∀b ∈ N ⇒ N ⊂ M . Thus M is largest sesmisimple left ideal of A.


Reference: https://math.stackexchange.com/questions/3324395/characterisation-of-largest-
semisimple-left-ideal

Solution 4. The proof is essentially the same as the proof of Artin-Tate Lemma. We let
x1 , ......xn be the generators of A over R. Let y1 , ......yk be the generators of A over B. Let
xi = Σj bij yj . Let yi yj = Σk bijk yk . Let B0 be the subalgebra of B generated by bij , bijk over
R. Since B ⊂ Z(A), we can show that A is a finitely generated module over B0 . B0 is a
finitely generated commutative algebra over noetherian ring R. By Hilbert’s basis theorem
B0 is noetherian. Thus A is noetherian B0 -module. Thus B as a submodule of A is finitely
generated B0 -module. Thus B is finitely generated algebra over R


Reference: https://math.stackexchange.com/questions/3328253/finitely-generated-algebra-
over-commutative-ring

Solution 5. This is the density theorem. You can check page 49 of the reference.

7
Algebra Quals Solutions Yizhou Chen September 8, 2019


Reference: https://lkempf.github.io/AlgebraStroppel/algebraII.pdf

Solution 6. ⇒: Suppose M torsion free but Mp is not Rp -torsion free. r ∈ R, s, s0 ∈ / p, x ∈

M, sr0 · xs = 0 ⇒ ∃s00 ∈
/ p such that s00 rx = 0 and s00 r 6= 0 ⇒ x is torsion, contradiction.
⇐: Suppose x ∈ M is a nontrivial torsion of R then ∃r ∈ R, rx = 0. We know r is not unit
because otherwise x = 0. Thus take m be a maximal ideal containing r. Notice that if x1 is Rm -
torsion in Mm then ∃s ∈ / m, sx = 0. Then ∃a, b such that ar + sb = 1 ⇒ x = (ar + sb)x = 0,
x
contradiction. Thus 1 is Rm -torsion in Mm , contradiction.

Solution 7. a. Some nice tricks: Start with the polynomial x4 −α4 ∈ F [x]. Let the minimal
polynomial of α be f . We know f divides the previous polynomial. Moreover, x2 − α4 is the
minimal polynomial of α2 . So we have [F (α) : F ] = [F (α2 ) : F ] = 2. So f is of degree 2.
4 4
√ to choose the other root of f from the roots of x − α . Our choices
We need are ±α, ±iα
(i = −1). We can eliminate the first two because α ∈ / F . So f looks like x −(i+1)αx+iα2
2

(or you substitute the i’s with −i, not really affect the result, so let’s put up with this). So
(i + 1)α ∈ F, iα2 ∈ F, i ∈
/ F . But that implies i ∈ F (α) ⇒ F (α) = F (i). So the only choice
for F (α) is F (i).
b. The Galois group is the dihedral group of order 8.

Solution 8. We can define the polynomial h(x) = fg (g(x)) − f (x) ∈ F ( fg )[x]. We have
h(x) = 0 and degh = d. So we need to make sure it is the minimal polynomial. Suppose
we have l(x) ∈ F ( fg )[x] such that degl < d, l(x) = 0. Each coefficient of l(x) looks like
bi,ni y ni +......+bi,−mi y −mi where y = fg . Without loss of generality we can assume degf ≤ degg
(otherwise do the later procedures in an inverted manner, namely, eliminate the positive
powers and inver the whole thing).Then multiply l with a power of y such that each mi = 0.
Then we can consider l as an element in F (x)[y]. So the new l looks like p(y) := am y m +
am−1 y m−1 + ...... + a0 where each ai ∈ F [x] and has degree less than d. Now suppose m ≥ d,
multiply p( fg ) with g m . Since p( fg ) = 0 and f, g coprime we have g m |am f m + am−1 f m−1 g +
...... + a1 f g m−1 . Because the degree on the right side is strictly less than the degree of the
left side, this is not going to happen unless the right side is 0, which means am = 0. So
l = 0, contradiction.

Solution 9. Let’s start with the short exact sequence

0 → L → Rk → A → 0

where L is finitely generated over R. Apply the left exact functor Hom( , B) we have

0 → Hom(A, B) → Hom(Rk , B) → Hom(L, B)

8
Algebra Quals Solutions Yizhou Chen September 8, 2019

The flatness of C implies that the functor ⊗R C is exact. Thus we have the following exact
sequence
0 → Hom(A, B) ⊗ C → Hom(Rk , B) ⊗ C → Hom(L, B) ⊗ C
We replace the B in the second sequence with B ⊗ C then we have

0 → Hom(A, B ⊗ C) → Hom(Rk , B ⊗ C) → Hom(L, B ⊗ C)

Put the last two sequences together we have the following commutative diagram:
0 Hom(A, B) ⊗ C Hom(Rk , B) ⊗ C Hom(L, B) ⊗ C

ρ1 ρ2 ρ3

0 Hom(A, B ⊗ C) Hom(Rk , B ⊗ C) Hom(L, B ⊗ C)

where ρi is the giving by the following general pattern: namely there is always a map
φ : Hom(X, Y ) ⊗ Z → Hom(X, Y ⊗ Z) given by f ⊗ z 7→ (x 7→ (f (x) ⊗ z)). You can check
as an exercise that this map is isomorphism when X is free of finite rank. By some diagram
chasing, if we can show ρ3 is injective then ρ1 is isomorphism. So we can first show that ρ1
is injective if ρ2 is isomorphism. Then by substituting A with L we get ρ3 is injective,


→ →
Solution 10. a. So F is given by (A, B) 7→ A × B on objects and (A f A0 , B g B 0 ) 7→ h
where h is the unique arrow in the following commutative diagram:

A A×B B

f h g

A0 A0 × B 0 B0
A natural candidate of left adjoint of F is the functor G : C → C 2 given by A 7→ (A, A) on
objects and (A → B) 7→ (A → B, A → B) on the morphism. To show G is the left adjoint,
it suffices to show that there is the canonical isomorphism M or(GX, Y ) ∼ = M or(X, F Y ).
Let Y = Y1 , Y2 . Then the canonical isomorphism is given by the following diagram:
X X ×X X

Y1 Y1 × Y2 Y2

I admit it is not crystal clear explanation. So please tell me when you get confused.
b. The category of abelian groups is abelian category. So the idea of product and coproduct
are the same. So the previous G functor becomes right adjoint of F by inverting all the
arrows in the last diagram (and change product to coproduct).

9
Algebra Quals Solutions Yizhou Chen September 8, 2019

Spring 2018:

n n
Solution 1. Suppose x = [Q(α) Q:k Q],nyi = [Q(α) : Q(α )], z = [Q(α ) : Q]. By the given
condition we know: x = yz, x = i=1 pi where each pi is prime then pi > n, ∀i. But we also
know y ≤ n ⇒ y = 1 ⇒ z = x ⇒ Q(αn ) = Q(α).


√ √
Solution 2. a. Suppose 3 3 ∈ Q(ζ). By Galois theory, Q(3 3) ⊂ Q(ζ) correspond to a
subgroup of Gal(Q(ζ)/Q) = Z/6 of order 2, which is apparently a copy of Z/2 generated by
the conjugation map. Now we have a problem: Z/2 ⊂ Z/6 is normal since the latter group is
abelian,
√ thus Z/2 should correspond to a normal extension. We have a contradiction because
3
Q( 3)/Q is not a splitting field of polynomial. (third root of unity not included).
b. Suppose α has a Q sube root in√Q(ζ, α), then the polynomial X 9 −3 would split over Q(α, ζ)
(we can write it as 9i=1 (X −3 αζ i )). But [Q(α, ζ) : Q] = 18 and √ [K : Q] > 18 where K is
splitting field of X 9 − 3, since X 6 + X 3 + 1 does not split over Q(9 3)

Solution 3. This is a good problem for lunch or dinner, if you find it too boring sitting
alone. So the answer is that the quotient is given by gcd of all entries.
Here is the solution if you want to give up the fun: Let ei be the usual vector with 1 at
the ith entry and 0 otherwise. Suppose we start with a matrix M , put Ei (1 on the i, i
th entry and 0 otherwise) on the left and ej on the right you get Mi,j ej . Now you can
shift the second j around by placing the shifting matrices on the further left. So you get
Mi,j ek , ∀k. Now with mI, m ∈ Z, I the identity matrix, you can put coefficients before
Mi,j ek and add them up together. So what you end up with is (αZ)n where α is the gcd
of {Mi,j ∃M ∈ I such that Mi,j is an entry of it}. So the quotient would be a finite abelian
group.

Solution 4. (Qiu) D ⊗k K is central simple over K. So it must be in the form Ma×a (L)
where L is division algebra over K. By degree reasons if a > 1, they must be p. Thus it
suffices to show that D ⊗k K is not division algebra. We can see this by showing that K ⊗k K
is not division algebra. Notice that K = k[x]/(f ) where f is the minimal polynomial of a.
So K ⊗k K = K[x]/(f ), but in this case f is not irreducible over K ⇒ K[x]/(f ) is not a
domain hence not a division algebra.

Solution 5. a. ⇒: Suppose f is not surjective but epimorphism. We know f (M ) 6= N .

Then consider N/f (M ). We can define two maps g, h : N → Z given by the following:
g = 0, h is 0 on f (M ) by maps some a ∈ N/f (M ) − {0} to 1 ∈ Z (this is a bit abuse of
notation, you should think a as one of its preimage in N ). Then we have g 6= h, g ◦ f = h ◦ f ,

10
Algebra Quals Solutions Yizhou Chen September 8, 2019

contradiction.
⇐: Suppose ∃g, h such that g ◦ f = h ◦ f ⇒ ∀a ∈ N, ∃m ∈ M such that f (m) = a ⇒ g(a) =
(g ◦ f )(m) = (h ◦ f )(m) = h(a) ⇒ g = h ⇒ f is epimorphism.
i
b. A nice example is the natural injection Z ,−→ Q. If we have g, h : Q → A such that gi = hi,
p g(p) h(p) p
we have that g( q ) = g(q) = h(q) = h( q ). This is an epimorphism which is not surjection.

Solution 6. We need to work on the following uncompleted character table:

G h1i, 1 hyi, 3 hxi, 4 hx2 i, 4

χ1 1 1 1 1

where the number near the conjugacy class is its size. Now I want to first figure out all the
one dimensional irreducible representations: Let χ be such a representation and ζ the third
root of 1. We know that χ(y) = χ(z) = χ(yz) and χ(y) ∈ {±1} since y is of order 2. Thus
we must have χ(y) = χ(z) = 1. So to make χ nontrivial, we have χ(x) ∈ {ζ, ζ 2 }. So we
have two more rows: (1 1 ζ ζ 2 ), (1 1 ζ 2 ζ). Use Schur’s orthogonality (see problem 11 of fall
2018), the last row should be (3 − 1 0 0). So the complete character table is:
G h1i hxi hx2 i hyi

χ1 1 1 1 1

χ2 1 ω ω2 1

χ3 1 ω2 ω 1

χ4 3 0 0 −1

The group is actually A4

Solution 7. We know that the I is a finitely generated B-module. Hence I k /I k+1 is finitely
generated B/I-module, ∀k ∈ N. Let x1 , ......xm be the generators of I and ai such that
xai i = 0 ⇒ I a = 0 where a = Σi ai + 1. We consider the filtration I, I 2 , ......I a = 0.
Each successive quotient is finitely generated B/I-module hence finitely generated A-module.
Then by induction and short exact sequences of the form 0 → I k+1 → I k → I k /I k+1 → 0
we can conclude that I is finitely generated A-mod. Then by SES 0 → I → B → B/I → 0
we conclude B is finitely generated A-mod.

11
Algebra Quals Solutions Yizhou Chen September 8, 2019

Solution 8. In the case that i ∈

/ F then F ⊗R C = F [i], which is a field. Otherwise there is
an idempotent in F ⊗R C given by 12 (1 ⊗ 1 + i ⊗ i), which shows us that the tensor algebra
is a product of two fields.

Solution 9. Another classical application of Sylow theorems:

Let ni be the number of Sylow-i groups. If some ni = 1 then we are done. First notice that
616 = 8 × 7 × 11. So we need to calculate n2 , n7 , n11 . By Sylow theorems, if none of the
previous numbers is 1, we have: n2 = 7 or 11 or 77, n7 = 22 or 8, n11 = 56. So the order of
the group must be at least 56 · (11 − 1) + 8 · (7 − 1) + (8 − 1) · 7 = 657 > 616, contradiction.

Solution 10. We know that the localization of a noetherian intergal domain is still a noethe-
rian integral domain. Now it suffices to show that R[S −1 ] is integrally closed and every prime
ideal is maximal.
For integrally closedness, the fraction field of R[S −1 ] is the same as that of R. Take an element
there which is integral over R[S −1 ], let it be a. Then a would satisfy some polynomial like
X n + asn−1
n−1
X n−1 +......+ as00 = 0 where ai ∈ R, si ∈ S. Then we have san +bn−1 an−1 +......b0 = 0
for s ∈ S, bi ∈ R. Then (sa)n + cn−1 (sa)n−1 + ...... + c0 = 0 for ci ∈ R ⇒ sa integral over
R ⇒ sa ∈ R ⇒ a ∈ R[S −1 ] ⇒ R[S −1 ] integrally closed.
For the second part, we know Spec(R[S −1 ]) ⊂ Spec(R). Any prime ideal in R[S −1 ] is a
maximal ideal in R and thus should be a maximal ideal in R[S −1 ].

12
Algebra Quals Solutions Yizhou Chen September 8, 2019

Fall 2018:

Solution 1. a. Let G be a nontrivial subgroup of Q8 which does not contain −1. Then we
have G ∩ Q8 − {±1} =6 ∅. Any element in the intersection squared would give us −1.
b. Suppose we have the embedding
→ S7
Q8 ,−
−1 7→ a
i 7→ x
j 7→ y
k 7→ z
If we write out x, y, z as product of disjoint cycles, then each of them would contain a 4-cycle,
because these three elements should be of order 4. In the product there may or may not be a
2-cycle and that’s it (otherwise either the order goes wrong or we run out of objects). Thus
we can let x = x1 x2 , y = y1 y2 , z = z1 z2 where x1 , y1 , z1 are 4-cycles. x2 , y2 , z2 are 2-cycles or
identities. Now if x1 , y1 are the same 4-cycle then z = xy does not have a 4-cycle (instead,
a bunch of 2-cycles). Then x1 6= y1 . Thus x1 , y1 , z1 = x1 y1 are three distinct 4-cycles.
Now this is where it goes wrong:
Without loss of generality let x1 = (1 2 3 4) ⇒ a = x2 contains (1 3)(2 4). In fact
a = (1 3)(2 4) because x22 = 1. Then we need to find three different 4-cycles, the square of
which are the same. This is impossible: suppose the 4-cycle is (1 x y z), in order to have
(1 3) in the square, we must have y = 3. Then we are left with only two choices.

Solution 2. Let H ⊂ G be the subgroup of finite index n.

Let’s start with a remark: G acts on the left cosets G/H by left multiplication, which gives
us a map G → Sn . Notice that this map uniquely determines H because H is exactly the
elements in G the image of which under the previous map fixes the first object. Thus we
have an injection {subgroup of G with index n} ,− → {maps from G to Sn }. Now notice that
the latter set is finite in the case G is finite. The reason is that there are only finitely many
choices for the images of the generators. Thus G only has finitely many subgroups of index
n.
For the second part, notice that for a automorphism φ of G, φ(H) is still a subgroup
of G of index n. Since there are only finitely many subgroups of index n, the group
∩φan automorphism of G φ(H), a proper characteristic subgroup of G, as an intersection of finitely
many subgroups of index n, is of finite index.

Solution 3. This one is very similar to the proof of the primitive element theorem:
In the case F is finite, F = Fpn for some prime p. Then the generator of the cyclic group
K ∗ would give us the primitive element.

13
Algebra Quals Solutions Yizhou Chen September 8, 2019

If F is infinite, assume K is not primitive extension. Pick w ∈ K such that [F (w) : F ]

is maximal among all F (a) for a ∈ K. Now suppose F (w) 6= K. Pick b ∈ K, b ∈ / F (w).
Consider the family of extensions {F (w + λb)λ ∈ F }. This must be a finite set since there
are only finitely many intermediate fields and w + λb ∈/ F (by linear independence of two
elements). Now ∃λ1 6= λ2 such that F (w + λ1 b) = F (w + λ2 b) ⇒ b, w ∈ F (w + λ1 b) ⇒
[F (w + λ1 b) : F ] > [F (w) : F ], contradiction.

Solution 4. Let K 0 be the splitting field of f . Since everything is separable, it suffices

to calculate [K 0 : K]. We can write f (x) = (x − a)(x − b)(x − c)(x − c) where a, b real,
c not. In the case b ∈ / K(a), [K 0 : K] = 24 and the only choice of Galois group is S4 .
In the case b ∈ K(a), (x − c)(x − c) does not split over K(a) since K(a) ⊂ R. Thus
[K 0 : K] = [K(a) : K] · 2 = 8.

Solution 5. a. First we have 0 ∈ / S, otherwise R = S and we have nothing to do here. Then

it suffices for x ∈
/ S, we find a prime ideal px such that px ∩ S = ∅ and x ∈ px . We do this by
first pick I an maximal element of the set W := {ideals in R{whichcontain}x and does not intersect S}.
Notice that W is nonempty because xR ∈ W and such a maximal element exists, because if
C is an ascending chain of ideals in W , then ∪J∈C J is still an element in W . Suppose I is
not prime then ∃c, d ∈ R such that cd ∈ I, c ∈ / I, d ∈ / I ⇒ cd ∈ / S. Without loss of generality
assume c ∈ / S ⇒ I + cR ∈ W and strictly larger than I. Hence contradiction.
b. ⇒: Q Let p be a prime ideal. Pick x ∈ p, x 6= 0 and of course, x is not a unit, Then let
x = ki=1 pi where each pi is irreducible ⇒ ∃j, pj ∈ p ⇒ (pj ) ⊂ p and (pj ) is out pricnipal
prime ideal.
⇐: First note that there are irreducible elements in R as generators of principal prime
ideals, because there exists prime ideals in R (such as maximal ideals). Then let S =
{product of irreducible elements with units}.Qn It is easy to see that S is a multiplicative
set. For saturatedness, suppose xyQ= i=1 ai , ai distinct
Q irreducibles. Then we have
I, J, I ∪ J = {1, 2, ......n} such that i∈I ai · u = x, i∈J ai · v = y where u, v are units.
Thus x, y ∈ S. If there are powers on ai , we can do an easy induction on the power by
dividing both sides of the equation (on appropriates letters of course) by ai .
Now we use result from part a to see that there can’t be any nonzero element outside of
S ⇒ R − S = {0} ⇒ R is a UFD by definition.

Solution 6. a. We can start by showing that trace is nonzero: take x ∈ K, x ∈ / F , by

the transitivity of trace we may assume that K = F (x). Let the solutions of the minimal
polynomial of x to be x1 , ......xn . Then trace of xk is Σni=1 xki . Then by a corollary of Artin’s
theorem on linear independence of characters the previous sum cannot vanish all the time.
Now we can define a pairing T r : K × K → F given by (x, y) 7→ T r(xy). Notice that
pairing is nondegenerate: if ∃x 6= 0, T r(xy) = 0∀y ⇒ T r = 0 contradiction. Then the map
x 7→ T r(x · ) gives us an isomorphism of F -vector spaces K ∼ = K V where K V is the dual

14
Algebra Quals Solutions Yizhou Chen September 8, 2019

of K, because the map is an injective homomorphism between two finite dimensional vector
spaces of equal dimensions. Now for the given basis {xi }, we pick the dual basis in K V and
then take their inverse image in K under the previous map and obtain the desired {yj }.
b. We can start by choosing a basis of K over F which is inside B: Let {xi } be a basis of K
over F . Then xi is algebraic over F . Multiply the coefficients with common denominators in
minimal polynomial of xi over F we know that xi is algebraic over A. Let the polynomial of xi
over A be an X n + ...... + a0 = 0. Then ann X n + an−1
n an−1 X
n−1
+ ...... + a0 an−1
n = 0 ⇒ an x i ∈ B
then we can substitute xi with an xi . After the substitution {xi } is still a basis of K over F
because they are only under scaling from elements in F .
Now define B V := {x ∈ K, T r(xb) ∈ A, ∀b ∈ B}. Apparently ∀b ∈ B, T r(b) as a coefficients
in the minimal polynomial of b is inside A ⇒ B ⊂ B V . Now choose basis of K over
F, {xi } ⊂ B and its dual {yj } as described in part a. Now we know that B V ⊂ ΣAyj
because for x = Σaj yj ∈ B V , aj ∈ F, T r(xxi ) = ai ∈ A. Now A noetherian ⇒ ΣAyj
noetherian ⇒ B ⊂ ΣAyj finitely generated as A-module.

Solution 7. This problem is mostly about to find a way out of the maze in definitions of
category theory.
For teminologies, let η : idC → GF,  : F G → idD be the unit and counit respectively.
⇒:
We have the isomorphism M or(X, X) ∼ = M or(F X, F X) ∼= M or(X, GF X) where ηX lives in
the last morphism set. If we pull it back in M or(F X, F X) and show that it’s an isomorphism
then we are done. This is exactly what we will do:
The last isomorpohism is given by

M or(X, GF X) ∼
= M or(F X, F X)
g 7→ F X ◦ F (g)
Under this isomorphism we have ηX 7→ F X ◦ F (ηX ) = idF X ⇒ ηX is isomorphism.
⇐:
Basically we want to make the following equation work:
−1
f 7→G(f )◦ηX h7→ηY ◦h
M or(F X, F Y ) ∼
= M or(X, GF Y ) ∼
= M or(X, Y )

The first one works by adjunction and second by the fact η is natural isomorphism.
Let the previous composition of maps be φ : M or(F X, F Y ) → M or(X, Y ). Then φ(F f ) =
ηY−1 ◦ GF f ◦ ηX = f by the fact that η is natural transformation.
Now the only thing left to show is F (ηY−1 ◦Gf ◦ηX ) = f . This need a bit more work. Consider
the following diagram:
F (ηX ) F X
FX F GF X FX

f F Gf f
F (ηY ) F Y
FY F GF Y FY

15
Algebra Quals Solutions Yizhou Chen September 8, 2019

The goal is to show that the left square commute. We already know that the right square
commutes and the whole big square commutes. Thus the two paths on the left composing
F Y on the left would be the same. We know F Y is isomorphism because it is one side
inverse of an isomorphism F (ηY ). Thus we can throw F Y away.

Solution 8. A coproduct of two copies of k[x] would do the trick.

Solution 9. This calculation is inspired by analysis:

We have
1
= 1 + gf + (gf )2 + ......
1 − gf

1
= 1 + f g + (f g)2 + (f g)3 + ......
1 − fg
Let x = 1 + gf + (gf )2 + ...... then we would naturally have 1 + f g + (f g)2 + (f g)3 + ...... =
1 + f xg. So our candidate would be 1 + f xg.
Now formally: let x = (1 − gf )−1 , y = 1 + f xg. We have (1 − f g)(1 + f xg) = 1 + f xg −
f g − f gf xg.
Plug in x−xgf = 1 = x−gf x we have: 1+f xg −f g −f gf xg = 1+f xg −f g −f xg +f g = 1.
You can show that it is the left side inverse with the same calculation.

Solution 10. a. Let φ be the map. The map makes sense. The work is to show that the
map is bijective.
Injectiveness:
Suppose ∃a, b such that a(1 + x) − by = ay + b(1 + x) = 0. Combine the two equations we
have a(y 2 + x2 + 2x + 1) = 0. But the left side is a(2 + 2x) ⇒ a = 0 ⇒ b = 0.
Surjectiveness:
I suffices to show (y, 0) ∈ Im(φ) and (0, y) ∈ Im(y). Indeed, (y, 0) = φ( y2 , x−1
2
) and (0, y) =
1−x y
φ( 2 , 2 ).
m1 m2
b. The inverse can be explicitly written out: m1 ⊗ m2 7→ 2(1+x)

Solution 11. By the given problem we have the following uncompletetd chart and we want
to complete it:

16
Algebra Quals Solutions Yizhou Chen September 8, 2019

G h1i hg1 i hg12 i hg2 i

χ1 1 1 1 1

χ2 1 ω ω2 1

χ3 1 ω2 ω 1

χ4 ......

Where χ3 comes from the symmetry that if we have an irreducible representation, we can
do a conjugate version of it and it is still an irreducible representation.
Now here are the two orthogonality relations that we are going to use:

(
|G| i = j
Σg∈G χi (g)χj (g) =
0 otherwise

|G|
Σi χi (g) · χi (g) =
c

where c is the cardinality of hgi, Now by the given data, multiplication by g1 gives an injective
map from hg1 i into hg12 i. Similarly, multiplication by g12 gives an injective map from hg12 i
into hg1 i. From the two maps we have |hg1 i| = |hg12 i|. Let a = |hg1 i|. Then multiplication
by g1 gives an injective map from hg12 i into {1} ∪ hg2 i, multiplication by g1 gives injective
map from {1} ∪ hg2 i into hg1 ‘i. From these two maps we have |hg2 i| = a − 1. Let M be
the matrix of entries from the character table and let Mij be the jth entry of χi . By second
orthogonality we have Mi2 = Mi3 = 0 for i > 3. For i > 3, apply the first orthogonality
to the first row and ith row we have Mi1 + Mi4 = 0 ⇒ Mi4 ∈ Z − {0}. Apply the second
3a
orthogonality to the last column we have 3 ≤ 1 + 1 + 1 + Σi>3 ||Mi4 ||2 = a−1 . a = 1 doesn’t
work. If a = 2 then there should be three more rows looking like 1 0 0 − 1, there are too
3a 3a
few choices. If a = 3, a−1 is not an integer. If a > 4, a−1 < 4. Thus our only choice is that
a = 4 and the extra row is 3 0 0 − 1. So the completed chart is:

17
Algebra Quals Solutions Yizhou Chen September 8, 2019

G h1i hg1 i hg12 i hg2 i

χ1 1 1 1 1

χ2 1 ω ω2 1

χ3 1 ω2 ω 1

χ4 3 0 0 −1

This is the character table of A4 . It has cardinality 12. We have g1 = (1 2 3), g2 = (1 2)(3 4)

Solution 12. Notice that K contains all the elements in F which are algebraic over F ,
because any algebraic number over F is inside some Fpn , which can be obtained by adjoining
a root of unity to F . Suppose D is a division algebra of finite dimension over K. Take d ∈ D
and choose a basis {d1 , ......dn } of D over K. We can associate a matrix under the previous
basis with coefficients in F by letting it acts on D by left multiplication.The
Qk matrixt has a
mk
Jordan canonical form over F and we have a polynomial i=1 (d − xk ) = 0 where xk is
algebraic over F and with slight abuse of notation I denote the matrix associated to d by d.
We know that every division algebra does not have zero divisors. Hence we have d = xk for
some k. By previous argument we have xk ∈ K ⇒ d ∈ K ⇒ D = K. By Artin Wedderborne
theorem all the simple finite dimensional algebras over K are in the form Mn×n (D). Thus
all the finite dimensional simple algebras are Mn×n (K)

18
Algebra Quals Solutions Yizhou Chen September 8, 2019

Spring 2019:

Solution 1. We can start by showing that N is abelian: for n1 , n2 ∈ N, g ∈ G, gn1 n2 n−1 −1 −1

1 n2 g =
gn1 g −1 gn2 g −1 gn−1
1 g −1
gn−1 −1
2 g ⇒ [N, N ] / G. By the minimality of N we have [N, N ] = {1}
or N . The second case is impossible because if so, N would not be solvable. But as a
subgroup of a solvable group N is always solvable.
Now we have N is abelian. Suppose p a prime and p||N |. Let S be the Sylow p subgroup
of N . We know P / N . ∀g ∈ G, we know gP g −1 is a Sylow p subgroup of the group
gN g −1 ⇒ gP g −1 = P ⇒ P / G ⇒ P = N .
Now by the characterisation of all the finite abelian groups, the last step is to show that
every element in N is of order p. For x ∈ N, g ∈ G, gxp g −1 = (gxg −1 )p ⇒ pN / G. Also
pN ( N because there are nontrivial elements in N of order p. Then by the minimality of
N, pN = 0.

Solution 2. Here is a sketch of the proof. I am not 100 percent sure if it works (on some set
theory level). Define W := {C ⊂ B, ∀x ∈ C, Zx ∩ A = ∅}. Let A0 = {x ∈ B, Zx ∩ A 6= ∅}.
Then by Zorn’s lemma we can pick a maximal element C from W (if A0 6= B) and show
B = C ⊕ A0 . Then we can define the map to be 0 on the map and the on A0 it is x 7→ f (nx)n
where nx ∈ A, n ∈ Z, f is the given homomorphism.

Solution 3. The basic √ idea is that we can use the idea of module. We define |||| in the
following way: ||a + b −d|| := a2 + b2 d. Then we can prove x|y ⇒ ||x|||||y||. Thus x|2 ⇒
||x|||4 ⇒ ||x|| =√1 or 2 or 4. The first two possibilities implies x is ±1. The last one says
x is ±2 or 1 ± −3 in case d = 3. That does not √ divide 2 of course.
√ So 2 is irreducible.
2
But (2)
√ is not irreducible.
√ In case 2|d,
√ we have ( −d) ∈ (2) but −d ∈
/ (2). Otherwise
(1 + −d)(1 − −d) ∈ (2) but 1 + −d ∈ / (2).

Solution 4. We can take P = Rx1 + ......Rxk for xi ∈ P where k is minimal. Then

we have P ⊕ Q = Rk for some Q. Let m be the maximal ideal of R. Then we have
Rk ⊗ R/m = R/m ⊗ (P ⊕ Q) ⇒ (R/m)k = P/mP ⊕ Q/mQ. Then by dimension reasons
Q/mQ = 0. Then by Nakayama’s lemma Q = 0 ⇒ P = Rk hence free.

Solution 5. First suppose a has order k mod p. Then we have p|Φn (a) ⇒ p|an − 1 ⇒
k|n ⇒ k = n or k < n.
In the case k = n, then we know by Fermat’s little theorem that ap−1 ≡ 1modp ⇒ n|p − 1 ⇒
p ≡ 1modn.
In the case k < n, we have X n − 1 = m|n Φm (X). By the fact k|n, (X − a)|Φk (X), (X −
Q

19
Algebra Quals Solutions Yizhou Chen September 8, 2019

a)|Φn (X) ⇒ (X − a)2 |X n − 1. Now we let X n − 1 = (X − a)2 f (X). Let X = Y + a then we

have (Y +a)n −1 = Y 2 f (Y +a). By observation on the right side we know that the coefficient
of the first degree term on the left side is 0 ⇒ nan−1 ≡ 0modp. We know that a 6≡ 0modp
(otherwise p 6 |Φn (a)). Thus p|n. This case should not be taken into consideration.

Solution 6. This problem is the same as #6, Fall 2016

Solution 7. a. I use the following characterization of the Jacobson radical: J(A) = {a ∈
A, ∀x ∈ A, 1 + xa ∈ A× }. Let  a = (ai,j )  be an element in J. The last row of a must be
0 0 0
zero, otherwise we have I + 0 0
 0  · a has the last row zero hence not invertible.
0 0 −a−1 3,3  
0 u v
For the same reason the other two diagonals must also be 0. Now a looks like w 0 x.
  0 0 0
0 0 0
−1
If u 6= 0. Then I + 0 −u
 0 · a has a row of zero in the middle hence not invertible.
0 0 0
Thus u = 0. For the same reasons v = w = x = 0 ⇒ a = 0 ⇒ J = 0.
b. Yes of course because it is 0.

Solution 8. We know Sn−1 ⊂ GLn−1 (C). So the problem is to embed a group of order n
into Sn−1 . By taking the group action of left multiplication on left cosets of a subgroup,
this is always possible unless the group does not have nontrivial subgroup. By the Sylow
theorems, this happens only if the group is Z/p, you can just embed it into C by sending
the generator to a primitive pth root of unity.

Solution 9. a. R = C[x, y], a = x, b = y
b. I asked the question on stackexchange, You can check it out here:
https://math.stackexchange.com/questions/3324366/showing-a-functor-is-not-representable
There is also a specific way of doing this problem in this case: Take the R as described in
part a. By Yoneda’s lemma, assume the functor is representable, Then ∃(x, y) ∈ A2 with
the following universal property: For any ring B, (z1 , z2 ) ∈ B 2 then ∃!f : A → B ring
map such that f (x) = z1 , f (y) = z2 . In fact, this universal element is given by our iso-
morphism in M or(Hom(A, ), F ) under the iso M or(Hom(A, ), F ) ∼ = F (A). Now we have
morphism f : A → R[ a ] such that f : x 7→ a, y 7→ b. Also we have g : A → R[ 1b ] such that
1

g : x 7→ a, y 7→ b. Now we consider f, g as maps from A to F rac(R). Let h = f − g. Then

h(x) = h(y) = 0 ⇒ h(1) = h(x)h(c1 ) + h(y)h(c2 ) = 0 for some c1 , c2 ∈ A ⇒ h = 0 ⇒ f = g.
Thus f, g must take values in R[ a1 ]∩R[ 1b ] = R ⇒ we can take a map f : A → R sending (x, y)
to (a, b), which is inside Hom(A, R). Then we have contradiction because (a, b) ∈ / F (R)

20
Algebra Quals Solutions Yizhou Chen September 8, 2019

Solution 10. An object A is projective ⇔ every SES 0 → X → Y → A → 0 splits. An

object B is injective ⇔ every SES 0 → X → X → Y → 0 splits. Thus every object is
projective ⇔ every SES splits ⇔ every object is injective.

21