SERGE LANG’S ALGEBRA CHAPTER 5 SOLUTIONS

KELLER

Problem 1

We see:
(α2 + α + 1)(α2 + α) = α(α3 + 2α2 + 2α + 1)

= α(α3 + α2 + α + 2 + α2 + α − 1)

= α3 + α2 − α

= 2α − 2
Using the Euclidean algorithm, we have:

(α − 1)(−α2 − 2α − 3) + (α3 + α2 + α + 2) = 5

Taking this relation modulo (α3 + α2 + α + 2), we find that

−1 2 2 3
(α + 1)−1 = α − α−
5 5 5

Problem 2

Consider f (x) := x2 − α2 ∈ F (α2 )[x]. This has root α, so that

[E : F (α2 )] 6 2

Since we are assuming this extension is odd, the only odd integer 6 2
is 1. Hence,

[E : F (α2 )] = 1 =⇒ F (α) = F (α2 )

Date: January 14, 2020.

1
2 KELLER

Problem 3

Define deg(f ) := n, deg(g) = m, and consider [F (α, β) : F ]. Using

the Tower law,

[F (α, β) : F ] = [F (α, β) : F (α)] · [F (α) : F ]

[F (α, β) : F ] = [F (α, β) : F (β)] · [F (β) : F ]

The above implies that n and m both divide [F (α, β) : F ]. As n and

m are coprime, we have that nm | [F (α, β) : F ]. We then find that
m | [F (α, β) : F (α)], so that in particular m 6 [F (α, β) : F (α)].
On the other hand, g ∈ F (α)[X] has degree m with root β, so
that [F (α, β) : F (α)] 6 m as well. Thus, [F (α, β) : F (α)] = m
identically, in which case g must be the minimal polynomial of β over
F (α), implying g is irreducible.

Problem 4

Has α has minimal polynomial x4 − 2, we know that [Q(α) : Q] = 4.

From this, we know that any other nontrivial extension has degree 2.
Certainly [Q(α) : Q(α2 )] = 2, and indeed since α is a real root, the
above intermediate extension is the only one.

Problem 5

Observe that Φ9 (X) = X 6 + X 3 + 1, where Φ9 denotes the 9th cyclo-

tomic polynomial. This implies that any homomoprhism σ : Q(α) → C
must carry α to another primitive 9th root of unity. Thus, the only
 SERGE LANG’S ALGEBRA CHAPTER 5 SOLUTIONS 3

homomorphism are of the form

σi
Q(α) / C

α / αi
where i is coprime to 9.

Problem 6
√ √ √ √ √
Observe first that ( 2 + 3)2 = 5 + 2 6, and we see that 2 + 3
is a root of
f (X) = X 4 − 10X + 1

And no strictly smaller degree works, whence this element is algebraic

of degree 4.

Problem 7

Set n := [E : k]. By definition, there exist {α1 , . . . , αn } ⊂ k such

that E = Spank {α1 , . . . , αn }. Consider now

SpanF {α1 , . . . , αn }

This span contains both E and F , hence EF , and is generated over

F . This then tells us that [EF : F ] 6 [E : k] by definition of degree.
Hence,
[EF : k] = [EF : F ] · [F : k]

6 [E : k] · [F : k]
Furthermore, if [E : k] and [F : k] are coprime, the tower law tells us
that both [E : k] and [F : k] divide [EF : k], so that [E : k] · [F : k] |
[EF : k]. Thus

[E : k] · [F : k] 6 [EF : k] 6 [E : k] · [F : k]
4 KELLER

=⇒ [EF : k] = [E : k] · [F : k]

Problem 8

Observe that E is formed by adjoining finitely many αi ; suppose first

that f is irreducible. If α1 is any root of f , by definition

[F (α1 ) : F ] = n

Then, f factors as (X − α1 )g(X) over F (α1 ), where g(X) is a not

necessarily irreducible polynomial in F (α1 )[X]. Then, deg(g) | n − 1,
so that adjoining another root α2 , we have that

[F (α1 , α2 ) : F (α1 )] | n − 1

Continuing inductively,

[E : k] = [F (α1 , . . . , αn ) : F (α1 , . . . , αn−1 )] · · · [F (α1 ) : F ]

And we see that the above must divide n!. For the general case, note
that f factors into irreducibles of degree n1 , . . . , nk such that n1 + · · · +
nk = n. Then, by the first case, we see that [E : k] divides n1 ! . . . nk !.
Moreover, we know that n1 ! . . . nk ! | n!, in which case

[E : k] | n!

as desired.

Problem 9
8 8
We see that xp − 1 = (x − 1)p , so the splitting field is Fp .

Problem 10

(a). Note that iα2 has minimal polynomial x2 + 5, so that [Q(iα2 ) :

Q] = 2. Every degree 2 extension is normal, so this is a normal exten-
sion.
 SERGE LANG’S ALGEBRA CHAPTER 5 SOLUTIONS 5

(b). We see that (α + iα)2 = 2iα2 , so that α + iα has minimal polyno-

mial X 2 − 2iα2 over Q(iα2 ). Whence

[Q(α + iα) : Q(iα2 )] = 2

and we again deduce as in (a) that this is a normal extension. Then,

by the Tower law we find [Q(α + iα) : Q] = 4. However, α + iα has
minimal polynomial x4 + 20 over Q; we observe that this polynomial
has splitting field Q(α, i), which is a degree 8 extension. Thus Q(α+iα)
cannot be a normal extension.

Problem 11
√ √
(a). The splitting field is Q( 2), and [Q( 2) : Q] = 2.

(b). We see that X 2 − 1 = (X − 1)(X + 1), so this splits over Q,

implying the splitting field is Q itself. This is a degree 1 extension.

(c). The splitting field is Q(21/3 , ζ3 ), where ζ3 is a primitive 3rd root

of unity and 21/3 is the real root of X 3 − 2. Then,

[Q[21/3 , ζ3 ) : Q] = 6
√
(d). The splitting field is Q( 2, 21/3 , ζ3 ), and the Tower law yields
√
[Q( 2, 21/3 , ζ3 ) : Q] = 12
√
(e). We have discriminant −3, so the splitting field is Q( 3, i) and
√
[Q( 3, i) : Q] = 4.

(f ). The splitting field is Q(ζ9 ), where ζ9 is a primitive 9th root of

unity. This is a normal and separable extension, and we see that

[Q(ζ9 ) : Q] = 6
6 KELLER

(g). The splitting field is Q(71/5 , ζ5 ), where 71/5 is a real root of X 5 − 7

and ζ5 is a primitive 5th root of unity. By the Tower law, we also have:

[Q(71/5 , ζ4 ) : Q] = 20

Problem 12

Note that k must have characteristic p > 0. Let φ denote the Frobe-
nius endomorphism α 7→ αp . Suppose that αp = β p ; by definition,
(α − β)p = 0, so that α = β which gives injectivity. As φ is injective
on a finite set, we must also have surjectivity. Thus for all ` ∈ k, there
exists some α ∈ k such that αp = `.

Problem 13

Since there are finitely many roots, this is a finite field with prime
n
characteristic. Thus deg(f ) = pn for some n ∈ N, and since xp − x has
n
the same roots, we conclude that f (x) = xp − x since f is assumed
monic.

Problem 14

We know that L = K(α1 , . . . , αk ) for some elements αi in an alge-

braic closure of K. Let fi denote the minimal polynomial of αi over
K and suppose some αi is not separable; this implies that Dfi (X) is
divisible by fi . As fi has strictly larger degree, we must have Dfi ≡ 0,
so that p | deg(fi ). Then p | [L : K], which contradicts our assumption.

Problem 15

Let β be a pth root of a in some algebraic closure of K. Then,

n n−1
X p − a = (X p − β)p
 SERGE LANG’S ALGEBRA CHAPTER 5 SOLUTIONS 7

Assume the minimal polynomial of β over K has degree strictly smaller.

This implies
n−1
(γ p − β)i = 0
n−1
for some i < p. Then by definition, (X p − β)i must have coefficients
in K. Observe however that X i−1 has coefficients iβ, implying β ∈ K,
n
a contradiction. We conclude that the only possibility is that X p − a
n
is the minimal polynomial of γ over K, so that X p − a is irreducible.

Problem 16

” =⇒ ” Assume K(α) is separable. Observe that in K(αp )[X], α is a

n
root of X p − αp , so that the minimal polynomial of f of α over K(αp )
n n
divides X p − αp .
n n n
Notice that f | X p − αp = (X − α)p . Since f must be a separable
polynomials, the only possibility is that f (X) = X − α. Thus α ∈
n
K(αp ).
” ⇐= ” Assume K(α) = K(αp ); let f (X) denote the minimal poly-
nomial of α over K. If f is not separable, then Df = 0, implying
f (X) = g(X p ) for some g(X) ∈ K[X]. By definition, f (α) = g(αp ) =
0, so that [K(αp ) : K] = deg(g) < deg(f ).
By assumption, [K(αp ) : K] = [K(α) : K] = deg(f ), which leads to
a contradiction. This implies that K(α) must be a separable extension.

Problem 17

Suppose (b) holds. If the characteristic of k is 0, let fi denote the

minimal polynomial of each αi , algebraic over k. We have that fi0 (αi ) 6=
0 since the degree is strictly smaller, so that each αi is separable. Thus
k(α1 , . . . , αn ) is separable.
8 KELLER

Now, we consider the characteristic p case. Suppose some α is insep-

arable over k. The minimal polynomial f (X) ∈ k[X] of α over k then
has 0 derivative, implying further that f (X) ∈ k[X p ]. If all elements of
k were pth powers, we may rewrite the coefficients of f as pth power;
this implies
f (X) = (g(X))p

for some g(X) ∈ k[X]. This contradicts irreducibility of the minimal

polynomial.
Suppose now that k has characteristic p and not every element of
k is a power of p. Let α denote such an element; choose β in some
algebraic closure such that β p = α. Then f (X) := X p − α is the
minimal polynomial of β over k by Problem 15. We also see that
f 0 (X) ≡ 0, whence k(β) is an inseparable extension as desired.

Problem 18

Recall that (Fpn )× is cyclic. Let x denote a generator; then x2 gen-

pn −1 pn +1
erates all perfect squares of (Fpn )× , implying Fp has 2
+1 = 2

perfect squares.
Now, let y ∈ Fpn and consider the set
pn − 1
S := {y − x2n | n = 1, . . . , } ∪ {0}
2
pn +1
Observe that |S| = 2
. Counting elements, there are pn total el-
ements in Fpn and pn + 1 elements in both sets. By the Pigeonhole
principle, S contains a perfect square; that is, y − w2 = z 2 for w,
z ∈ Fpn . Rearranging,
y = w2 + z 2

So that y is a sum of 2 squares.

 SERGE LANG’S ALGEBRA CHAPTER 5 SOLUTIONS 9

Problem 19

Let R be a subring of K containing F . In particular, R is a vector

space over F ; let r ∈ R and consider the map φr : R → R mapping
x 7→ rx. This is linear and since R is contained in a field, R is an
integral domain.
Thus, φr is an injective map between vector spaces of equal dimen-
sion, hence surjective as well. This implies that there exists y ∈ R such
that ry = 1; that is, r is a unit. As r ∈ R was arbitrary, we conclude
that R is a field.
Now, let x be any transcendental element over Q and consider the
containments
Q ⊂ Q[x] ⊂ Q(x)

Both Q and Q(x) are fields, but Q[x] is not a field.

Problem 20

(a). Observe that F (x) is precisely the function field of F . Let F ⊂ K

f (x)
be a proper containment; K contains some y of the form g(x)
, where
f, g ∈ F [x].
Consider m(T ) := f (T )+yg(T ) ∈ F (y)[T ]. By construction, m(x) =
0, so that F (x) is an algebraic extension of F (y), and since F (y) ⊂ K ⊂
F (x), we deduce that F (x) is algebraic over K.

(b). We first show that m(T ) as above is the minimal polynomial of

x over F (y). To see this, it suffices to prove irreducibility. By Gauss’
Lemma, irreducibility of m(T ) over F (y)[T ] is the same as irreducibility
over F [y][T ] = F [y, T ]. We see that m(T ) is linear as a polynomial in
y, so it is certainly irreducible over F [y, T ].
10 KELLER

This tells us that [F (x) : F (y)] = deg(m). It remains to find this

degree. Observe that the factor of y on g(T ) makes is so that no higher
power can cancel with any terms of f (T ); counting degrees, this tells
us explicitly that degT (m) = max{deg(f ), deg(g)}. Thus [F (x), F (y)]
has the desired degree.

Problem 21

Recall that if µ denotes the Möbius function, then Problem 12 of

Chapter 2 is such that
(
0, n > 1
(µ ∗ 1)(n) =
1, n = 1
P
Suppose now that f (n) = d|n g(d) = (φ1 ∗ g)(n). Convolving with
our Möbius function,
(µ ∗ f )(n) = µ ∗ (φ1 ∗ g)(n)

= g(n)
Which yields the Möbius inversion formula.

Problem 22
n
Note first that X q − X is irreducible and separable over Fq . Its
splitting field is of course Fqn . Let α be a root of P . Since P is
irreducible,
[Fq (α) : Fq ] = d

But, then
n = [Fqn : Fq ] = [Fqn : Fq (α)] · [Fq (α) : Fq ]

= [Fqn : Fq (α)] · d

=⇒ d | n
To prove the converse, we will need the following:
 SERGE LANG’S ALGEBRA CHAPTER 5 SOLUTIONS 11

Claim: d | n ⇐⇒ q d − 1 | q n − 1. To prove this, assume first that

d | n so that n = kd for some integer k. Then,

q n − 1 = q kd − 1

= (q d − 1) · (q kd−d + · · · + 1)

=⇒ q d − 1 | q n − 1

Conversely, assume that q d − 1 | q n − 1, we may use the Euclidean

algorithm to write n = kd + r with r < d. We then see

q n − 1 = q r · (q kd − 1) + q r − 1

Since q d − 1 | q kd − 1 and q d | q n − 1, we deduce that q d − 1 | q r − 1.

However, since r < d, this is possible only if r = 0. But then n = kd,
so that d | n, as desired.
Now, back to the converse of the problem at hand. Since d | n, we
know that q d −1 | q n −1; employing the above result again, this implies

d −1 n −1 d n
Xq − 1 | Xq − 1 =⇒ X q − X | X q − X

d
Then, X q − x has splitting field Fqd , whence Fqd ⊂ Fqn . Choose some
α ∈ Fqn and let mα,Fq (X) denote its minimal polynomial over Fq . Since
[Fqd : Fq ] = d, this minimal polynomial has degree d. By definition of
d
minimal polynomial, it is irreducible, and since α satisfies αq − α = 0,
d
we know that the minimal polynomial divides X q − X.
By the above observations, this then implies that mα,Fq (X) also di-
n
vides X q − X. Thus, we have found an irreducible polynomial of
n
degree d dividing X q − X, as desired.
n
Now, every factor of X q − X is an irreducible polynomial whose
degree divides n. Taking the product of all such polynomials, we must
12 KELLER
n
get X q − X; that is
n
Y Y
Xq − X = fd (X)
d|n fd irred

Counting degrees on each side of the above equality, we see that

X
dψ(d) = q n
d|n

By the Möbius inversion formula of Problem 21, we have that

X
nψ(n) = µ(n/d)q d
d|n

where µ denotes the Möbius function. Dividing by n in the above, we

see:
1X
ψ(n) = µ(n/d)q d
n
d|n

as desired.

Problem 23

(a). Using the notation of Problem 22, we may rewrite:

∞
Y
Z(t) = (1 − td )−ψ(d)
d=1

Taking the natural log, we compute:

∞
X
log(Z(t)) = −ψ(d) log(1 − td )
d=1
∞ ∞ `d
X X t
= −ψ(d)
d=1 `=1
`
∞ n
X t X
=− dψ(d)
n=1
n
d|n
∞ n n
X t q
=
n=1
n
= − log(1 − qt)
 SERGE LANG’S ALGEBRA CHAPTER 5 SOLUTIONS 13

Exponentiating the above, we find

1
Z(t) =
1 − qt
(b). To do.

Problem 24

By the Tower law,

[k(t, u) : k(tp , up )] = [k(t, u) : k(tp , u)] · [k(tp , u) : k(tp , up )]

Consider the degree of the extension k(t, u) over k(tp , u). We see that
t has minimal polynomial xp − tp over k(tp , u) which is degree p. By
definition we see that

[k(t, u) : k(tp , u)] = p

Similarly, we deduce that [k(tp , u) : k(tp , up )] = p as u similarly has

minimal polynomial xp − up . Overall,

[k(t, u) : k(tp , up )] = p2

as contended.

(b). By the primitive element theorem, it suffices to show that k(t, u) is

not a simple extension of k(tp , up ). We claim that any simple extension
k(f (t, u)) over k(tp , up ) has degree less than or equal to p.
To see this, consider the monic polynomial

xp − f (tp , up ) ∈ k(tp , up )[x]

This has root f (t, u), since f (t, u)p = f (tp , up ). Thus the degree of the
minimal polynomial of f is 6 p, implying [k(f (t, u)) : k(tp , up )] 6 p.
Combining this fact with the above, we see that k(t, u) cannot be a
simple extension since it has degree p2 > p over k(tp , up ), as desired.
14 KELLER

Problem 25

Assume first that we have a purely inseparable extension. By defi-

s
nition, we may choose β ∈ E such that β p ∈
/ k for any s < r. By the
tower law,
2 r−1
[k(β) : k] = [k(β) : k(β p )] · [k(β p ) : k(β p )] · · · [k(β p ) : k] = pr

Since the extension is purely inseparable, the inseparable degree is pre-

cisely the regular degree of the extension. Thus the above shows we
must have E = k(β).
Now we consider the general case. Let E0 denote the compositum
of all intermediate separable extensions of k. We have that E/E0 is a
purely inseparable extension, whence E = E0 (β) by the above para-
graph for some β ∈ E. As E0 is a finite separable extension, the prim-
itive element theorem guarantees that E0 = k(α) for some α ∈ E0 .
k
Thus E = k(α, β); furthermore, β p = ` + mα for `, m ∈ k. Raising
r−k
this equality to the pr−k power, we find that αp ∈ k; by assumption
however the smallest power for which this can hold is pr . Thus, we
must have that β = ` + mα ∈ k(α) so that in fact E = k(α), and we
are done.

Problem 26

Let K denote our normal extension. Suppose that f (x) factors into
irreducibles as f (x) = g(x) · h(x) in K[x]; let K be an algebraic closure
of K so that over K[x],

g(x) = (x − α1 ) . . . (x − αk ), h(x) = (x − β1 ) . . . (x − βj )

Define σ : K → K by sending α1 7→ β1 . By normality, restricting

σ to K induces an automorphism of K. Under this automorphism,
 SERGE LANG’S ALGEBRA CHAPTER 5 SOLUTIONS 15

σ(α1 ) = β1 . Since g and h are assumed irreducible, we see that σ(α1 )

has minimal polynomial h. But then all σ(αi ) must have minimal
polynomial h, whence g σ = h as contended.
Moreover, the above shows that if a polynomial factors into irre-
ducibles in a normal extension, as irreducible factors will have the same
degree. For a counterexample in the non-normal case, we consider Q(α)
where α = 21/3 . Then

x3 − 2 = (x − α) · (x2 + αx + α2 )

And both of these factors are irreducible. However, as their degrees do

not match, there is no automorphism between these polynomials since
it would induce a bijection between their roots.

Problem 27

Suppose f (X) = g(X)h(X) with X = (X1 , . . . , Xn+1 ). Then, we

may write
d
X
g(X) = φi (X 0 )Xn+1
i

i=0

where X 0 = (X1 , . . . , Xn ). Observe that

g(X)
0
∈ k(X 0 )[Xn+1 ]
φ(X )
and this would induce a nontrivial factorization of the minimal poly-
nomial P (Xn+1 ), contradicting the definition of P . Thus f must be
irreducible in k[X1 , . . . , Xn+1 ].
Conversely, if f ((X1 , . . . , Xn+1 ) is irreducible in k[X1 , . . . , Xn+1 ], then
by Gauss’ lemma, f (x1 , . . . , xn , Xn+1 ) is irreducible in k(x1 , . . . , xn )[Xn+1 ]
as contended.
16 KELLER

Problem 28

Observe that if f has no nontrivial zero, then it must be irreducible

since any nontrivial factorization has a linear term. Suppose that α ∈ F
is a root of f with F an odd degree extension. By the tower law,

[F : k] = [F : k(α)] · [k(α) : k]

If α ∈
/ k, then [k(α) : k] = 2, implying [F : k] is even. Thus we must
have that [k(α) : k] = 1, so α ∈ k is a nontrivial zero.
Similarly, if f is degree 3 we again see that f is irreducible. Let F
be an extension with [F : k] = 2 and α ∈ F a nontrivial root. Again
by the tower law,

[F : k] = [F : k(α)] · [k(α) : k]

so that if α ∈
/ k, 3 | [F : k]; this is again a contradiction, so we conclude
that α ∈ k.

Problem 29

Problem 30

(Artin Conjecture)

Problem 31

(a). Assume that α satisfies the characteristic polynomial. By con-

struction we have that

αn = u1 αn−1 + · · · + ud αn−d

So that by definition, xn := αn satisfies the desired recurrence relation.

 SERGE LANG’S ALGEBRA CHAPTER 5 SOLUTIONS 17

(b). Observe that we have a total of d degrees of freedom for choosing

the first d coordinate of the infinite vector (x0 , x1 , . . . ). The rest of the
entries are determined by the recurrence relations. This tells us that
the solution space is completely determined by the maximum amount
of linearly independent vectors in k ⊕d , which is of course d since we are
dealing with a vector space.
Note as well that the solution space is actually a vector space because
of the linearity of our recurrence relation.

(c). We only need prove that the vectors (αin )n∈N are linearly indepen-
dent. Indeed it suffices to show that the first d terms of each vector
are linearly independent, since as remarked in part (b), the rest of the
entries are determined by satisfying the recurrence relation.
Linear independence is equivalent to having the matrix
 
1 α1 . . . α1d−1
 .. . . .. 
. . . 
d−1
1 αd . . . αd

be nonsingular, which in turn is equivalent to having nonzero determi-

nant. As the determinant of the above is the standard Vandermonde
determinant, we know the above determinant is

Y
(αi − αj ) = (−1)n(n−1)/2 ∆
i6=j

where ∆ is the discriminant of the polynomial associated to the roots

αi . As we assumed this polynomial is separable, the above discriminant
is nonzero. This yields the contention.
18 KELLER

(d). Suppose xn := b1 α1n +· · ·+bd αdn . We have the system of equations:

x 0 = b1 + · · · + bd

x 1 = b1 α 1 + · · · + bd α d
..
.

xd−1 = b1 α1d−1 + · · · + bd αdd−1

In matrix form this may be rewritten
   
1 1 ... 1 b1 x0

..   ..  =  .. 
 α1 . . .

.  . .
d−1 d−1
α1 ... αd b d x d−1

The left most matrix is the transpose of the Vandermonde matrix,

which remains invertible. Thus we may invert the above relation to
solve for our bi explicitly or use Cramer’s rule.

(e). Set
X
f (T ) := xn T n
n>0
By the assumption of part (d),
d
XX
f (T ) = bi αin · T n
n>0 i=1
n
X X
= bi αin T n
i=1 n>0
d
X bi
=
i=1
1 − αi T
b1 bd
= + ··· +
1 − α1 T 1 − αd T
Which is the desired rational function.

Problem 32

Suppose that

1 − uX + vtX 2 = (1 − α1 X)(1 − α2 X)
 SERGE LANG’S ALGEBRA CHAPTER 5 SOLUTIONS 19

Which α1 6= α2 . We compute:
X
F (X) = an X n
n>0
X
= a0 + a1 + (uan−1 − vtan−2 − tn w)X n
n>2
X
= a0 + a1 (1 − u) + u · an X n+1
n>0
X X
− vt · an X n+2 − X 2 t2 w · tn X n
n>0 n>0

X 2 t2 w
= a0 + a1 (1 − u) = uXF (X) − vtX 2 F (X) −
1 − tX
2 2
X tw
=⇒ (1 − uX + vtX 2 )F (X) = a0 + a1 (1 − u) −
1 − tX
a0 + a1 (1 − u) X 2 t2 w
=⇒ F (X) = −
(1 − α1 X)(1 − α2 X) (1 − tX)(1 − α1 X)(1 − α2 X)

Employing partial fraction decomposition for the above, we deduce that

there exist constants A, B¡ and C such that

A B C
F (X) = + +
1 − α1 X 1 − α2 X 1 − tX

(b). Expanding the right hand side of the above partial fraction de-
composition as a power series, we find:

X 
F (X) = Aα1n + Bα2n + Ctn X n
n>0

whence comparing coefficients with the generating function F (X), we

conclude

xn = Aα1n + Bα2n + Ctn

as desired.
20 KELLER

Problem 33

(a). For n > d, we see:

xn = xn−d · xd

= xn−d a0 + · · · + ad−1 xd−1

= xn−d a0 + · · · + ad−1 xn−1

Continuing inductively, we see that all powers of x that are > d can be
reduced to a relation purely of powers strictly less than d. That is, we
may write

xn = a0,n + a1,n x + · · · + ad−1,n xd−1

for any integer n.

Pn i
(b). Suppose f (T ) = i=0 bi T . Substituting, we see that f (x) =
Pn i
i=0 bi x ; by part (a), every power may be rewritten as a linear com-

bination with respect to the power basis {1, x, . . . , xd−1 }. Letting ai (F )

denote the constants used in this linear combination, we ahve:

f (x) = a0 (F ) + a1 (F )x + · · · + ad−1 (F )xd−1

Pm i
Linearity also follows since if G(T ) = i=0 ci T ,
max{n,m}
X
F (T ) + G(T ) = (bi + ci )T i
i=0

By uniqueness of the representation of elements with respect to the

span of a basis, we see that ai (F + G) = ai (F ) + ai (G).

(c). Set
 
1 x1 . . . xd−1 
 1 
∆ := v(x1 , . . . , xd ) :=  ... . . . .. 

. 
1 xd . . . xd−1

d

 SERGE LANG’S ALGEBRA CHAPTER 5 SOLUTIONS 21

and assume g(T ) factors:

d
Y
g(T ) = (T − xi )
i=1

Then for each root of g, part (b) yields the matrix equation
 
1 x1 . . . xd−1 a0 (g) g(x1 )
  
1
 .. . . ..   ..  =  .. 
. . .  . .
d−1
1 xd . . . xd a d−1 (g) g(x d)

Use the notation  

1 x1 . . . g(x1 ) . . . xd−1 
 1 
∆j (F ) :=  ... . . . .. .. 

. . 
d−1

1 xd . . . g(xd ) . . . xd 
By Cramer’s rule we have

∆ · aj (g) = ∆j (g)

Assuming ∆ 6= 0 (that is, g is separable) we find that

∆j (g)
aj (g) =
∆
Problem 34

Define
d
Y
kT
f (T ) := e h(T − Xi )
i=1