Math 250A, G.

Bergman, 2002

A principal ideal domain that is not Euclidean

developed as a series of exercises

Let α denote the complex number (1 + 0---

–19) ⁄ 2, and R the ring Z[α ]. We shall show that R is a
principal ideal domain, but not a Euclidean ring. This is Exercise III.3.8 of Hungerford’s Algebra (2nd
edition), but no hints are given there; the proof outlined here was sketched for me by H. W. Lenstra, Jr..

(1) Verify that α 2 – α + 5 = 0, that R = {m +n α ! m, n ∈Z} = {m +n α- ! m, n ∈Z}, where the bar

denotes complex conjugation, and that the map x → |x| 2 = x x- is nonnegative integer valued and respects
multiplication.

(2) Deduce that |x| 2 = 1 for all units of R, and using a lower bound on the absolute value of the
imaginary part of any non-real member of R, conclude that the only units of R are ±1.

(3) Assuming R has a Euclidean function ϕ , let x be a nonzero nonunit of R minimizing ϕ (x).
Show that R ⁄ xR consists of the images in this ring of 0 and the units of R, hence has cardinality at
most 3. What nonzero rings are there of such cardinalities? Show that α 2 – α + 5 = 0 has no solution
in any of these rings, and deduce a contradiction, showing that R is not Euclidean.

We shall now show that R is a principal ideal domain. To do this, let I be any nonzero ideal of R,
and x a nonzero element of I of least absolute value, i.e., minimizing the integer x x- . We shall prove
I = xR. (Thus, we are using the function x → x x- as a substitute for a Euclidean function, even though it
does not have all the properties of one.)
For convenience, let us ‘‘normalize’’ our problem by taking J = x –1 I. Thus, J is an R-submodule of
C, containing R and having no nonzero element of absolute value < 1. We shall show from these
properties that J – R = ∅, i.e., that J = R.

(4) Show that any element of J that has distance less than 1 from some element of R must belong to
R. Deduce that in any element of J – R, the imaginary part must differ from any integral multiple of
0--
19 ⁄ 2 by at least 0-
3 ⁄ 2. (Suggestion: draw a picture showing the set of complex numbers which the
preceding observation excludes. However, unless you are told the contrary, this picture does not replace a
proof; it is merely to help you find a proof.)

(5) Deduce that if J – R is nonempty, it must contain an element y with imaginary part in the range
[0-3 ⁄ 2, 0--
19 ⁄ 2 – 0-
3 ⁄ 2], and real part in the range (–1 ⁄ 2, 1 ⁄ 2].

(6) Show that for such a y, the element 2y will have imaginary part too close to 0--
19 ⁄ 2 to lie in
J – R. Deduce that y = α ⁄ 2 or – α- ⁄ 2, and hence that α α- ⁄ 2 ∈J.

(7) Compute α α- ⁄ 2, and obtain a contradiction. Conclude that R is a principal ideal domain.

(8) What goes wrong with these arguments if we replace 19 throughout by 17? By 23?
 Math 250A, G. Bergman Lüroth, p.1

Lüroth’s Theorem, and some related results,

developed as a series of exercises
A simple transcendental extension of a field k means an extension of the form k(t) where t is
transcendental over k; in other words, an extension isomorphic to the field of rational functions in one
indeterminate over k. Throughout this set of exercises, k(t) will denote such a field.
We shall prove that every subextension of k(t), other than k itself, is also a simple transcendental
extension of k. We shall need to consider algebraic relations satisfied by elements of k(t) over subfields,
hence we shall work with the polynomial ring k(t)[x]. In calculating with elements of this ring we shall
make use of the subring k[t, x], which is a polynomial ring over k in two indeterminates.
(1) Show that every nonzero element f ∈k(t)[x] can be written uniquely f = P(t, x) ⁄ Q(t) where
P(t, x) ∈k[t, x], Q(t) ∈k[t], P and Q are relatively prime in the unique factorization domain k[t, x],
and Q is monic as a polynomial in t.
From now on, whenever we write an element f ∈k(t)[x] as P(t, x) ⁄ Q(t), we shall understand P and
Q to have the above properties. In this situation, let us define the height of f to be the nonnegative
integer ht( f ) = max(deg t (P(t, x)), deg t (Q(t))), where deg t means ‘‘degree as a polynomial in t’’.
In particular, if u ∈k(t) and we write u = P(t) ⁄ Q(t), then the above assumptions on P and Q, and
the above definition of the height ht(u), will be understood to apply.
(2) Suppose f = P(t, x) ⁄ Q(t) is monic as a polynomial in x over k(t). Show that ht( f ) = deg t (P(t, x)),
and that P(t, x) is not divisible by any nonunit element of k[t].
(3) Deduce that if f, g ∈k(t)[x] – {0} are both monic as polynomials in x, then ht( fg) = ht( f) + ht(g).
(4) If u is any element of k(t) not in k, show that there is an element u′, of the same height as u,
such that k(u′) = k(u), and with the properties that when u′ is written P′(t) ⁄ Q′(t) we have
deg t (P′) > deg t (Q′), and that P′ (as well as Q′) is monic.
(You will find that u′ can be taken to be of the form α u or α ⁄(u – β ), where α , β ∈k.)
The next observation shows that t, though transcendental over k, is algebraic over any nontrivial
intermediate field.
(5) Given any u = P(t) ⁄ Q(t) ∈k(t) – k, verify that t is a root of the polynomial P(x) – uQ(x) ∈k(u)[x].
Show further that if deg t (P) > deg t (Q), and P is monic, then the above polynomial is monic.
Now suppose L is any intermediate field: k ⊂ L ⊆ k(t). Let u = P(t) ⁄ Q(t) be an element of L – k
chosen to minimize n = ht(u), and, using (4), also taken so that deg t (P) > deg t (Q) and P is monic.
These conditions will be assumed in the next three steps.
(6) Show that every polynomial f ∈L[x] has height either 0 or ≥ n, and that P(x) – uQ(x) has
height exactly n; deduce from (3) that P(x) – uQ(x) is either irreducible in L[x], or divisible by a
nonunit element of k[x].
(7) Show that if P(x) – uQ(x) is divisible in L[x] by an element of k[x], then this element must
divide both P(x) and Q(x). Deduce that this element must be a unit. (Suggestion for the first assertion:
extend {1, u} to a basis of L over k. You might prove that assertion in the general context of
divisibility of a polynomial over any field L by a polynomial over a subfield k. The second assertion
depends on our assumptions on P and Q, specific to our present situation.)
(8) Conclude that P(x) – uQ(x) is the minimal polynomial of t over L. Deduce that P(x) – uQ(x) is
also the minimal polynomial of t over k(u) ⊆ L, and from this that L = k(u).
 Math 250A, G. Bergman Lüroth, p.2

(9) Deduce:
Lüroth’s Theorem. If k(t) is a simple transcendental extension of a field k, and k ⊂ L ⊆ k(t), then
L is also a simple transcendental extension of k, and is generated over k by any element u ∈L of
minimal positive height. Moreover, [k(t) : L] = ht(u).
To tie things up, we would like to be able to conclude that for every element u ∈k(t) – k, [k(t):k(u)] =
ht(u). Unfortunately, we have only proved (8) for elements u having minimal height in an intermediate
field L, and nothing we have proved excludes the possibility that if we start with a random element u,
k(u) – k might contain an element of smaller height than u. The only way I see to get the desired result is
to give a second proof of the irreducibility of P(x) – uQ(x), with the hypothesis of minimal height
replaced by the hypothesis L = k(u).
(10) Show for arbitrary u = P(t) ⁄ Q(t) ∈k(t) – k that P(x) – uQ(x) is irreducible in k(u)[x]. (Hint:
Apply Gauss’s Lemma to k[u].)
(11) Deduce that in the above situation, [k(t):k(u)] = ht(u), and conclude that u is of minimal height in
k(u) – k.
(12) Deduce the following result, including the ‘‘that is’’ clause. (This is essentially Exercise V.2.6(d) of
Hungerford; cf. Exercise IV.10 of Lang),
Corollary. The simple generators of k(t) as an extension field of k are precisely the elements of
height 1, that is, the elements of the form (at +b) ⁄ (ct +d) with ad – bc ≠ 0. Hence every automorphism
of k(t) over k takes t to an element of this form.
Lüroth’s Theorem is analogous to the result that a nontrivial subgroup of the free abelian group Z of
rank 1 is free of rank 1, is generated by any element of minimal absolute value, and has index in Z
equal to this absolute value. However, analogous results do not hold in all situations:
(13) Show that the subalgebra k[x 2, x 3] ⊆ k[x] is not a polynomial ring over k. (Suggestion: show it
is not a unique factorization domain.)
If one looks at subfields of fields of rational functions in more than one indeterminate, the analog of
Lüroth’s Theorem fails – these need not, in general, be isomorphic to rational function fields. However, to
give counterexamples, and to develop the partial positive results that do hold, we would need the theory of
transcendence degree, and the concept of separability for transcendental extensions (Hungerford,
Chapter VI; Lang, Chapter VIII).
Fall 2002 Math 250A, G. Bergman quad.recip., p.1

Quadratic Reciprocity
developed from the theory of finite fields, as a series of exercises

If p is a prime, then an element x ∈Z ⁄ pZ – {0} is called a quadratic residue if

(∃ y ∈Z ⁄ pZ – {0}) x = y 2. One also, loosely, calls an integer q not divisible by p a ‘‘quadratic residue
modulo p’’ if q- ∈Z ⁄ pZ is a quadratic residue.
Given p, we can clearly determine with a finite amount of work which elements of Z ⁄ pZ are
quadratic residues, and thus which elements of Z are quadratic residues modulo p; the latter set will be
periodic with period p. But given q ∈Z, it is not evident whether there will be any pattern in the set of
primes p modulo which q is a quadratic residue. We shall discover below that there is.
We begin the study of this question with a reduction:
(1) Let p be an odd prime, (Z ⁄ pZ) ∗ the group of units of Z ⁄ pZ, and (Z ⁄ pZ) ∗2 the subgroup {x 2 !
x ∈(Z ⁄ pZ) ∗ }. (i) Find [(Z ⁄ pZ) ∗ : (Z ⁄ pZ) ∗2]. (ii) Show from this that one can determine, given two
integers r and s relatively prime to p, whether rs is a quadratic residue modulo p, if one knows
whether r is and whether s is. (iii) Deduce that one can describe the set of primes modulo which an
arbitrary integer q is a quadratic residue if one can do this whenever q is a prime, and also when
q = –1.
Until further mention, p and q will be distinct odd primes. We shall establish below a relation
between the conditions that q be a quadratic residue modulo p, and that p be a quadratic residue
modulo q, known as the Law of Quadratic Reciprocity, by studying in two ways the question
Does the Galois group of the polynomial x q – 1 over Z ⁄ pZ
(∗)
act by even permutations of the roots of that polynomial?
and comparing the answers that we get. As noted in (1), we also need to investigate the cases q = –1 and
q = 2. These will be dealt with subsequently.
The first of our approaches to (∗) will be based on the fact that the Galois group is generated by the
map a → a p ; the second, on the use of the discriminant. For the first approach we begin with a general
result on cyclic groups:
(2) If x is a finite cyclic group, and y ∈ x , show that the map z → zy is an even permutation of the
elements of x if and only if y is a square in x .
For p, q still as above, let (Z ⁄ pZ)(ζ q ) be the extension of Z ⁄ pZ by a primitive qth root of
unity, ζ q .
(3) Show that the following conditions are equivalent:
(a) The pth-power map acts by an even permutation on ζ q ⊆ Z ⁄ pZ(ζ q )∗.
(b) Multiplication by p acts as an even permutation on the elements of Z ⁄ qZ.
(c) p is a quadratic residue modulo q.
We have now related something that happens in characteristic p and something that happens in
characteristic q. But (a) is formulated, not in Z ⁄ pZ, but (Z ⁄ pZ)(ζ q ). The next step is to reduce this to
a computation in Z ⁄ pZ.

(4) Compute the discriminant of x q – 1 in Z. (Suggestion: Consider successively the relation among
the following elements: d = Π i <j (ζ qi – ζ qj) 2, d′ = Π i ≠ j (ζ qi – ζ qj), d′ ′ = Π i ≠ j (1 – ζ qj –i ), E(1) =
Π i ≠ 0 (1 – ζ qi), E(x) = Π i ≠ 0 (x – ζ qi ). If you do it this way, be more precise than I have been about the
ranges of i and j. And be careful!)
(5) Obtain from (4) a necessary and sufficient condition for the Galois group of (Z ⁄ pZ)(ζ q ) to act by
even permutations on the roots of x q – 1. In putting this criterion in the simplest possible form, recall that
for nonzero elements a and b of a field, a is a square if and only if ab 2 is a square.
Fall 2002 Math 250A, G. Bergman quad.recip., p.2

(6) Comparing the results of (3) and (5), get a relation between p being a quadratic residue modulo q
and q being a quadratic residue modulo p, which holds whenever q ≡ 1 (mod 4). Reversing the roles of
p and q, show that the same relation holds when p ≡ 1 (mod 4).
To handle the remaining case, where both p and q are ≡ 3 (mod 4), we discover that we need to
know when –1 is a quadratic residue. This is something we wanted to find out anyway, so
(7) Show that –1 is a quadratic residue modulo p if and only if the group Z p∗ has an element of order
4. Determine conditions for this to happen in terms of the order of this group, and hence in terms of p.
(8) Complete the incomplete result of (6). Note: this result, the Law of Quadratic Reciprocity, is often
n
stated in terms of the Legendre Symbol ( – ), which is defined, for p a prime, and n an integer not
p
divisible by p, to be +1 if n is a quadratic residue modulo p, –1 if not. The special situation when
p, q ≡ 3 (mod 4) is expressed by bringing into the statement of the Law a term (–1) (( p –1) ⁄ 2)((q –1) ⁄ 2).
As a practical application:
(9) Characterize the set of odd primes p such that –39 is a quadratic residue modulo p as those
belonging to an explict list of congruence classes modulo some integer. Give all such primes < 100.
(You need not find square roots of –39 modulo each!)
It remains to handle the case q = 2. Note that all fields Z ⁄ pZ ( p odd) have a primitive square root
of unity, namely –1, so the Galois group of (Z ⁄ pZ)(ζ2 ) is trivial; nevertheless, 2 is a quadratic residue
in some fields Z ⁄ pZ but not in others. It turns out that, just as the case q = –1 involved the study of
primitive 4th roots of 1, so the case q = 2 can be solved by considering primitive 8th roots of 1, as is
suggested by the formula ζ8 = (1+i) ⁄ 0-2 .
(10) Without assuming the above formula, show that in any field K containing a primitive 8th root of
unity ζ8 , the element ζ8 + ζ8–1 is a square root of 2. Taking K = (Z ⁄ pZ)(ζ8 ), deduce that Z ⁄ pZ
contains a square root of 2 if and only if ζ8 + ζ8–1 is fixed under the pth-power automorphism.
Determine for what values of p this holds. (Hint: given any field K with a primitive 8th root of unity
–1 n –n
ζ8 , and an odd integer n, compare ζ8 + ζ8 and ζ8 + ζ8 .)

This completes the task we set ourselves to begin with. But one may generalize our question, and ask
when an element of a larger field F p n is a square in that field. This is answered in
(11) Letting F p n denote the field of p n elements, show that the norm map: F p n → Z ⁄ pZ is surjective.
(This is essentially Hungerford’s Exercise V.7.2.) Deduce that an element a ∈F p n is a square in this field
if and only if its norm in Z ⁄ pZ is a quadratic residue.
There are other related questions which do not have such neat answers. For instance, given a prime p
one may ask whether neither, one or both of 4 ± 0-3 are squares in (Z ⁄ pZ)(0-3 ).
(12) Show (with the help of the preceding results) how to answer the above question in terms of
congruences satisfied by p, in all cases except when 3 and (4 + 0-3 )(4 – 0-3 ) = 13 are both quadratic
residues modulo p.
I am told by a number-theorist that in the case excluded above, the answer is not determined by
congruences satisfied by p; however, from the fact that 3 is a quadratic residue mod p, it follows that
there is a factorization p = (a + 0-3 b)(a – 0-3 b) in Z[0-3 ], and the question can, it seems, be answered in
terms of congruences satisfied by the coefficients a and b in this factorization.
We end with an exercise which is tangential to the above results, and uses only (1) above:
(13) If m and n are integers, show that the polynomial (x 2 – m)(x 2 – n)(x 2 – mn) has a root in every
field Z ⁄ pZ. By appropriate choice of m and n, get an example of a polynomial in Z[x] having a root
in every field of finite characteristic, but no root in Q.
Fall 2002 Math 250A, G. Bergman sol’n in radicals, p.1

Solution in radicals of polynomials of degree ≤ 4

developed as a series of exercises

The idea of these exercises can be motivated as follows. Suppose u and are the roots of a
2
quadratic equation x + ax + b = 0. The quadratic formula expresses these roots as u = A 0 + A1 , =
A 0 − A1 , where A 0 is a polynomial in the coefficients of the given equation, and A1 is the square root
of such a polynomial; this is clearly consistent with the fact that A 0 is the same in both expressions,
while A1 changes sign. And in fact, if we solve the above pair of equations for A 0 and A1 and then
write out the resulting expressions for A 0 and A12 and compare with the expressions for the coefficients
a and b of the given quadratic equation in terms of u and , we immediately see that we can express
A 0 and A12 in terms of a and b, allowing us to rederive the quadratic formula.
Below, we will seek to generalize this idea to get cubic and quartic formulas as well.
Throughout these exercises:
n will denote one of the values 2, 3, 4 (the degree of the equation we seek to solve),
F will be a field which contains primitive mth roots of unity for all m ≤ n, hence has
characteristic either 0 or > n,
u1 , ... , un will be elements of F (which we will later take to be roots of a polynomial; but to
begin with, they will just be arbitrary elements on which we will perform some interesting
manipulations)
ω (when used) will denote a primitive cube root of unity assumed to lie in F,
i (when used) will denote a primitive fourth root of unity assumed to lie in F.
(1) For each of the systems of equations (a)-(d) below, assuming the the existence in F of the roots of
unity occurring in those equations, verify that u1 , ... , un can be written in the form shown, for unique
A 0 , ... , An − 1 ∈F. You may assume results of undergraduate linear algebra. (Suggestion: Write each of
these systems of equations as a matrix equation, and compute the square of the matrix in question. You
can simplify your computation by noting certain properties of the rows and columns which force most
columns to have zero product with most rows. Also recall the condition on the characteristic, mentioned in
the definition of F above.)

(a) n=2 u1 = A 0 + A1
u 2 = A 0 − A1 .

(b) n=3 u1 = A 0 + A1 + A 2
u 2 = A 0 + ω A1 + ω 2 A 2
u 3 = A 0 + ω 2 A1 + ω A 2 .

(c) n=4 u1 = A 0 + A1 + A 2 + A 3
u 2 = A 0 + i A1 − A 2 − i A 3
u 3 = A 0 − A1 + A 2 − A 3
u 4 = A 0 − i A1 − A 2 + i A 3 .

(d) n=4 u1 = A 0 + A1 + A 2 + A 3
u 2 = A 0 + A1 − A 2 − A 3
u 3 = A 0 − A1 + A 2 − A 3
u 4 = A 0 − A1 − A 2 + A 3 .
Fall 2002 Math 250A, G. Bergman sol’n in radicals, p.2

(2) Suppose we perform a permutation σ on the elements u1 , ... , un , writing ui′ = uσ (i) , and obtain a
representation of the new elements ui′ in terms of elements Ai′ ∈F, using the same systems of equations
as above. Show that in all four of cases (a-d), we get A0′ = A 0 , and that in three of the four cases, but
not in the remaining one, we can say that the elements A1′ , ... , An′ − 1 are obtained from A1 , ... , An − 1 by
a combination of a permutation, and multiplication by certain mth roots of unity, for some m ≤ n.
(Suggestion: either verify the result for a certain small number of permutations σ ∈Sn and show that the
whole result follows, or note certain conditions characterizing the set of expressions giving the A’s, i.e.,
the rows of your inverse matrix, verify that any permutation of the u’s, i.e., the columns, will
approximately preserve those conditions, and deduce that such a permutation will have the asserted affect
on the A’s.)
(3) Deduce that in each of the three ‘‘good’’ cases, the n − 1 elementary symmetric polynomials in
A1m, ... , Anm− 1 can be expressed as polynomials in the n elementary symmetric polynomials in
u1 , ... , un . Here m, which depends on n, is as in the preceding question.
(4) Deduce from this that for 2 ≤ n ≤ 4, we can find formulas for the solution of polynomial equations
of degree n if we have such formulas for polynomial equations of degree n − 1, and can take mth roots
(for the same m as in the last two questions). More precisely, you should show that under those
assumptions, we can get formulas associating to a polynomial equation of degree n a finite set of
elements, possibly more than n of them, among which the n roots must lie.
In the next two parts, you get to put the above theory into practice! Note that though (6) deals with a
harder case than (5), it asks for less: only that you go through the reduction, and not that you grind out an
explicit formula.
Before doing either or both of these, you might look ahead at (7), and see whether you wish to
incorporate the idea thereof into your calculations. Alternatively, you could do (7) afterwards and show
retroactively how it can be used to sharpen the results you get in (5) and/or (6).
(5) Carry out the procedure outlined in (1-4) for n = 3, culminating in a formula for solving the general
monic cubic equation x 3 + ax 2 + bx + c = 0.
(6) Carry out the reduction outlined above for n = 4. That is, derive formulas which can be used to solve
a quartic equation x 4 + ax 3 + bx 2 + cx + d = 0 in terms of the roots of an appropriate cubic.
Suggestion: begin by showing how you can reduce to the case u1 + u 2 + u 3 + u 4 = 0.
We now return to the problem noted in (4), that our procedure gives us a set of ‘‘candidate roots’’,
which may be larger than the set of actual roots of our polynomial. You might prepare for the next
problem by determining how many ‘‘candidate roots’’ it gives in each case. Be careful; some cases that
may look at first as though they are different will turn out to be the same.
(7) Verify that in the ‘‘good’’ cases of (2) arising in the solution of the cubic and quartic equations, one
has A1′ ... An′ − 1 = A1 ... An − 1 . Translate this into a statement about symmetric polynomials, and show
how it allows us to modify our procedures so as to get exactly the set of roots of the given equation.
(8) (Assumes the student has seen a presentation of the resolvant cubic of a quartic polynomial; e.g.,
Hungerford, Lemmas V.4.9 and V.4.10.) In the n = 4 case, suppose the coefficients of the given monic
quartic equation lie in a field K ⊆ F. Show that for A1 , A 2 , A 3 defined as in (1)(d), the extension
fields K(A12), K(A22), K(A32) are the fields generated over K by the three roots of the resolvant cubic of
the given quartic.