SERGE LANG’S ALGEBRA CHAPTER 4 SOLUTIONS

KELLER

Problem 1

(a) =⇒ (b): As k[x] is a PID, prime ideals are maximal.

(b) =⇒ (c): If (f (x)) is a maximal ideal, then, in particular f (x) is
a prime element. Since k[x] is an integral domain, any prime element
is irreducible.
(c) =⇒ (a): As k[x] is a PID, irreducible elements are prime so
that (f (x)) is a prime ideal.

Problem 2

a
(a). We want to prove the following statement: Let b
∈ Q with (a, b) =
1, a < b. If b = pα1 1 . . . pαnn . Then,
n
a X ai b i
=
b i=1
pαi i
for some ai , bi ∈ Z.

Proof. Let b = pα1 1 . . . pαnn . By the Chinese Remainder theorem, the

map
a + (b) 7→ (ai + (pαi i ))ni=1

has inverse
n
α
X Y
(ai + (pαi i ))ni=1 7→ ai b i pj j + (b)
i=1 i6=j

Date: January 14, 2020.

1
2 KELLER

Since we have assumed a < b and (a, b) = 1, we deduce

n
α
X Y
a= ai b i pj j
i=1 i6=j

Dividing both sides by b, we see

n
a X ai b i
= α
b p j
i=1 j

(b). The statement we want to prove is the following: Let a, b ∈ Z+ .

Then there exists ai ∈ Z+ such that

X
a= ai b i
i

Proof. We employ induction on a; for the base case when a = 1, we see

that

1 = 1 · b0

for any other b ∈ Z+ . Proceed inductively: if a < b, then a = a · b0 .

Assume now a > b. By the Euclidean algorithm, a = bq + r for some
r < b and q ∈ Z+ . Note that q < a, so by the inductive hypothesis,
P
q = i qi bi , so that
X 
a=b· q i bi + r
i
X
=⇒ a = qi bi+1 + r
i

Which completes the proof. 

 SERGE LANG’S ALGEBRA CHAPTER 4 SOLUTIONS 3

Problem 3
Pn
Suppose f (X) = i=0 ai X i . Then, we see:

n
X
f (X + Y ) = ai (X + Y )i
i=0
n X i  
X i
= ai Y k X i−k
i=0 k=0
k
n n  
X
k
X i i−k

= Y ai X
k=0 i=k
k
n n  
X
i
X i 
= f (X) + Y ai X i−k
i=1 i=k
k

Pn i


Setting φi (X) := i=k ai k
X i−k . In characteristic 0, we may rewrite
this as
n
X k!
φi (X) = ak X k−i
k=i
i!(k − i)!
n
1X
= ak Di (X k )
i! k=i
1 i
= D (f (x))
i!

As contended.

Problem 4

We employ multi-index notation. We may write any multivariable

polynomial in the form

X
f (x) := aα X α
|α|6n
4 KELLER

Consider:
X
f (X + Z) = aα (X + Z)α
|α|6n
X X α
= aα X α−β Y β
β
|α|6n β6α
 
X
β
 X α 
= Y aα X α−β
β
|β|6n β6α6n

Similar to the previous problem, we define

 
X α
φβ (X) := aα X α−β
β6α6n
β

And for characteristic 0,

 
X α
φβ (X) = aα X α−β
β6α6n
β
X α!
= aα X α−β
β6α6n
β!(α − β)!
1 X
= aα D β X α
β! β6α6n
Dβ f (X)
=
β!

Problem 5

(a). Consider:

(x + 1)4 + 1 = x4 + 4x3 + 6x2 + 4x + 2

Apply Eisenstein’s criterion for p = 2 to see that this is irreducible.

Similarly,

(x + 1)6 + (x + 1)3 + 1 = x6 + 6x5 + 15x4 + 21x3 + 18x2 + 9x + 3

Again, we employ Eisentein’s criterions for p = 3 to see that this is also

irreducible.
 SERGE LANG’S ALGEBRA CHAPTER 4 SOLUTIONS 5

(b). If deg f = 2 or 3 and f factors nontrivially, then f must have a

root. Consider now x3 − 5x2 + 1 and reduce this modulo 2. We see that
this becomes x3 + x2 + 1, which has no roots over Z/2, which implies
it must be irreducible over Z. By Gauss’ Lemma, this implies that the
above is orreducible over Q.

(c). Suppose that x2 + y 2 − 1 factors nontrivially over Q[x, y]. This is

equivalent to factoring over Q[x][y] by Gauss’ Lemma, so we see:

x2 + y 2 − 1 = (Q1 (x) + Q2 (x)y) · (Q01 (x) + Q02 (x)y)

= Q1 (x)Q01 (x) + Q01 (x)Q2 (x)

+ Q1 (x)Q02 (x) y + Q2 (x)Q02 (x)y 2

=⇒ Q1 (x)Q01 (x) = x2 − 1

Q01 (x)Q2 (x) = −Q1 (x)Q02 (x)

Q2 (x)Q02 (x) = 1

Where Qi , Q0i ∈ Z[x]. Thus the above immediately tells us Q2 (x) =

Q02 (x) = ±1 and Q1 (x) = −Q01 (x). Then we also see −Q21 (x) = x2 − 1,
implying 1 − x2 is a perfect square. But this is never true for any field
6= F2 , so the above is irreducible over Q.
To see that this is irreducible over C, we assume the above fac-
torization again. We instead see that Q2 (x)Q02 (x) = 1, so the these
polynomials are units. Since (C[x])× = C× , we know Q2 (x) = α,
Q02 (x) = 1/α, α ∈ C× . This tells us that

1 − x2 = (αQ1 (x))2

Which again is impossible, since 1 − x2 is not a perfect square over

C[x].
6 KELLER

Problem 6
Pn
Suppose f (x) = k=0 ak xk has a root a/b ∈ Q with (a, b) = 1,
ak ∈ Z and an = 1. Observe that upon substituting a/b into f (x), we
see:
an = −an−1 ban−1 − · · · − a0 bn

=⇒ b | an

=⇒ b = ±1

So that a/b = ±a ∈ Z, implying that we have integral closure.

Problem 7

(a). Consider 1 − f (X)q−1 . Counting degrees, we see that this has

degree d(q − 1). Also,

(1 − X1q−1 ) . . . (1 − Xnq−1 )

has degree n(q − 1). Suppose that the only root of f (X) is 0; then
1 − f (X)q−1 has n(q − 1) roots. Since d < n, this is a contradiction.

(b). Suppose first q − 1 | i. By standard group theoretic facts, xi = 1

for all x ∈ k × . This implies
X
= q − 1 = −1
x∈k

Suppose now q − 1 6| i. Choose a ∈ k × ; since the map x 7→ ax is an

automorphism of k, we have:
X X
xi = (ax)i
x∈k x∈k
X 
= ai xi
x∈k
 SERGE LANG’S ALGEBRA CHAPTER 4 SOLUTIONS 7

Since ai 6= 0, we must have xi = 0, as desired. Since f has N

P
x∈k

roots, we know
(
1, if x is a root of f
1 − f (x)q−1 =
0, else
So that summing over all x ∈ k,
X
N= (1 − f (x)q−1 )
x∈kn

Now, observe:
X X X
xi11 . . . xinn = xi11 xi22 . . . xinn
x∈kn x1 ∈k x∈kn−1
X
= ψ(i1 ) · xi22 . . . xinn
x∈kn−1

Proceeding inductively, we see

X
xi11 . . . xinn = ψ(i1 ) . . . ψ(in )
x∈kn
P
Now let I = (i1 , . . . , in ) denote a multi-index; we may rewrite x∈kn (1−

f (x)q−1 ) = x∈kn |I|6(q−1)d xi11 . . . xinn , so that:

P P
X X X
N= xi11 . . . xinn = xi11 . . . xinn
x∈kn |I|6(q−1)d |I|6d(q−1)
X
= ψ(i1 ) . . . ψ(in )
|I|6d(q−1)

≡0 mod q
So that N ≡ 0 mod q, as contended.

(c). Consider the product

Y
(1 − fi (x)q−1 )
i
P
This has degree (q − 1) · i di < (q − 1) · n by assumption. By identical
reasoning as in part (a0, if the fi had no other zeroes, the above must
have at least (q − 1)n zeroes, a contradiction.
8 KELLER

(d). Let us first find a function characteristic on 0. Define

Y
f (X) := (1 − Xiq−1 )
i

We see that f (X) = 0 for all X 6= 0, and f (0) = 1. Then, shifting the
above, we see:
Y
1 − (Xi − Yi )q−1


g(X) := f (Y ) ·
i

is such that g(X) = 0 unless X = Y , in which case g(Y ) = f (Y ).

Extending this idea, define
X Y
1 − (Xi − Yi )q−1


g(X) := f (Y )
Y ∈kn i

We see that g(X) = f (X) for all x ∈ k n . Since k is a finite field, the
above is a finite sum, whence f (X) can be written as a polynomial.

Problem 8

Let X 7→ aX + b and extend as a homomorphism that restricts to

the identity on A. Then, the above has inverse map X 7→ a−1 X − a−1 b,
in which case the above map is an automorphism.

Problem 9

Suppose φ is an automorphism with φ|A ≡ id. Observe that φ is

completely determined by its value on X, which

n
φ(X n ) = φ(X)

We also deduce that φ must leave degree invariant in order to be an

automorphism. Thus, φ(X) = aX + b for some constants a, b ∈ A.
Since φ is invertible, we know that φ−1 (X) = cX +d for other constants
c, d ∈ A. By definition, φ−1 ◦ φ(X) = X, so that ac = 1. Thus a is a
unit, and we are done.
 SERGE LANG’S ALGEBRA CHAPTER 4 SOLUTIONS 9

Problem 10

Suppose that f (X) ∈ k(X) is an automorphism of our field. By

h(X)
definition of k(X), f (X) = g(X)
for polynomials h(X), g(X) ∈ k[X].
Consider now the polynomial

P (Y ) := h(Y ) − f (X)g(Y ) ∈ K(f )[Y ]

By construction, this has root X. Indeed, we want to show

mX,k(f ) (Y ) = P (Y )

To this end, we only need to prove irreducibility. By Gauss’ Lemma, be-

ing irreducible in K(f )[Y ] is equivalent to being irreducible in K[f ][Y ] =
K[f, Y ]. However, this is certainly irreducible in K[f, Y ] as it is a lin-
ear polynomial in f . Thus, P (Y ) is the minimal polynomial of X over
K(f ), and

[K(X) : K(f )] = degY P (Y )

We see however that degY P (Y ) = max{degY g(Y ), h(Y )}, since the
multiple of f on g(Y ) makes it impossible for any higher powers of g
and h to cancel out.
Using this, any automorphism f must be such that [K(X) : K(f )] =
1. By the above, this is possible if and only if deg g and deg h are both
less than or equal to 1. But then we see that f is of the form

aX + b
f (X) = , a, b, c, d ∈ K
cX + d

and, to avoid any cancellations, ad − bc 6= 0.

10 KELLER

Problem 11

(a). Observe that

0 = D(1)

= D(x · 1/x)
D(x)
= + xD(1/x)
x
D(x)
=⇒ D(1/x) = − 2
x
Thus, using the property of derivations,
D(x) xD(y)
D(x/y) = −
y y2
yD(x) − xD(y)
=
y2
(b). We see:
D(xy)
L(xy) =
xy
yD(x) xD(y)
= +
xy xy
= L(x) + L(y)

(c). Observe:

m1 (x − α1 )m1 −1
L (x − α1 )m1 =
(x − α1 )m1
m1
=
(x − α1 )
Then by part (b) combined with the above,
X mi
L(R) =
i
(x − αi )

Problem 12

(a). In the monic case, we know that X 2 +b/aX +c/a has discriminant
(b2 −4ac)/a2 . Multiplying the above by a, this rescales the discriminant
by a2 . Thus aX 2 + bX + c has discriminant b2 − 4ac.
 SERGE LANG’S ALGEBRA CHAPTER 4 SOLUTIONS 11
P
(b). This problem uses symmetric sum notation ( sym ). By
definition of discriminant,
∆ = (x1 − x2 ) · (x1 − x3 ) · (x2 − x3 )

= x21 x2 − x21 x3 − x22 x1 + x22 x3 + x23 + x23 x2 + x23 x1

Thus
X X X
∆2 = x41 x22 − 2 x41 x2 x3 − 2 x31 x32
sym sym sym
X
+2 x31 x22 x3 − 6x21 x22 x23
sym

Let σi denote the degree i elementary symmetric polynomials on 3

variables. We see:
X X X
σ12 σ22 = x41 x22 + 2 x41 x2 x3 + 2 x31 x32
sym sym sym
X
+8 x31 x22 x3 + 15x21 x22 x23
sym
X X
σ13 σ3 = x41 x2 x3 + 3 x31 x32 x3 + 6x21 x22 x33
sym sym
X X
σ23 = x31 x32 + 3 x21 x22 x3 + 6x21 x22 x23
sym sym
X
σ1 σ2 σ3 = x31 x22 x3 + 3x21 x22 x23
sym

σ32 = x21 x22 x23

Then, comparing the above with the expression for ∆2 , we see

∆2 = σ1 σ22 − 4σ13 σ3 − 4σ23 + 18σ1 σ2 σ3 − 27σ32

By Vieté’s formulas,

σ1 = −a, σ2 = a2 , σ3 = −a3

And since ∆2 is homogeneous of degree 4 with respect to the coefficient

of our polynomial, we insert the appropriate power of a0 to homogenize
the discriminant. All in all, we see:

∆2 = a21 a22 − 4a0 a32 − 4a31 a3 − 27a31 a3 − 27a20 a23 + 18a0 a1 a2 a3

12 KELLER

(c). Using part (c) of Problem 11, one sees that

Y
f 0 (ti ) = (ti − tj )
i6=j

So that
n
Y n Y
Y
0
f (ti ) = (ti − tj )
i=1 i=1 i6=j

Looking at Df explicitly,
Y
Df = (ti tj )2
i<j
Y Y
= (ti tj ) (ti − tj )
i<j j>i
Y Y
= (−1)n(n−1)/2 (ti − tj ) (tj − ti )
i<j i<j
n Y
Y
n(n−1)/2
= (−1) (ti − tj )
i=1 i6=j
Yn
n(n−1)/2 0
= (−1) f (ti )
i=1

So that Df = (−1)n(n−1)/2 ni=1 f 0 (ti ).

Q

Problem 13

(a). Set h := f 3 − g 2 . By Mason’s Theorem, we have the inequalities:

3 deg(f ) 6 deg(f ) + deg(g) + deg(h) − 1

2 deg(g) 6 deg(f ) + deg(g) + deg(h) − 1

Add these equations together to find

3 deg(f ) 6 2 deg(f ) + 2 deg(h) − 2

whence
deg(f ) 6 2 deg(h) − 2

As desired.
 SERGE LANG’S ALGEBRA CHAPTER 4 SOLUTIONS 13

(b). Find d(x) such that

A(x)f (x) + B(x)g(x) = d(x)

Then by definition, A(x)f (x)/d(x) and B(x)g(x)/d(x) are coprime. We

apply part (a) to these two polynomials:
 
Af 3  Bg 2
deg(f ) + deg(A) − deg(d) 6 2 deg − −2
d d
6 2 deg (Af )3 − (Bg)2 − 4 deg(d) − 2

6 2 deg f 3 − g 2 − 4 deg(d) − 2

=⇒ deg(f ) 6 2 deg(f 3 − g 2 ) − 2
Which yields the general case.

(c). By part (b), we may assume without loss of generality that f and
g are coprime. By Mason’s Theorem,

m deg(f ) 6 deg(f ) + deg(g) + deg(h) − 1

n deg(g) 6 deg(f ) + deg(g) + deg(h) − 1

Then, with the above equations we find:

(n − 1)(m − 1) deg(f ) 6 (n − 1) deg(g) + (n − 1) deg(h) − (n − 1)

(n − 1) deg(g) 6 deg(f ) + deg(h) − 1

Adding these together, we find the general version:

(n − 1)(m − 1) − 1 deg(f ) 6 n deg(h) − n

Problem 14

Let our Frey polynomial be x(x − 3a)(x + 3b). We may complete the
cube to find

u = 3(a2 + ab + b2 ), v = (a − b)(2a2 + 2b2 + 5ab)

14 KELLER

We see that
u1/2 = O(a + b)

v 1/3 = O(a + b)

So that Szpiro’s conjecture says

u1/2 = O N0 (D)1+/2

v 1/3 = O(N0 (D)1+/3

And as D = O(abc), we combine all of the above to find:

|a + b| = O N0 (abc)1+/2

Which yields the abc-conjecture.

Problem 15

Let
S := {p | 2p−1 6≡ 1 mod p2 }

We want to use the abc conjecture to show that S must be infinite.

Suppose that p is such that 2n ≡ 1 mod p and 2n 6≡ 1 mod p2 . We

×
want to show that in fact p ∈ S. Let d denote the order of 2 in Z/p ;
by definition, d | p − 1, whence d | n. Since 2n 6≡ 1 mod p2 , we deduce
that 2d 6≡ 1 mod p2 as well.
Write p − 1 = dm with (m, p) = 1, and 2d = 1 + pk, (p, k) = 1. Then,

m
2p−1 = 2d

= (1 + pk)m

≡ (1 + pmk) mod p2

6≡ 1 mod p2
So that by definition, p ∈ S. Now, suppose for sake of contradiction
that S is a finite set. Factor 2n − 1 = un vn , where un is a product of
 SERGE LANG’S ALGEBRA CHAPTER 4 SOLUTIONS 15

primes in S and vn is a product of primes not in S. By assumption,

the correspondence n 7→ un is bounded. If p | vn , the above argument
implies that p2 | 2n − 1, whence p2 | vn .
Set a = 2n −1, b = 1, so that a+b = 2n . Applying the abc conjecture
to this triple, we find:

2n − 1 = un vn

1+
 N0 2n (2n − 1)

 (un vn1/2 )1+

 vn(1+)/2

This implies that the correspondence n 7→ vn is also bounded, so that

S c is also bounded. But this implies that there are only finitely many
primes, which is a contradiction. Thus there must be infinitely many
primes with 2p−1 6≡ 1 mod p2 .

Problem 16

Let M (F ) denote the Mahler measure:

deg(F )
Y
M (F ) := |F0 | · max{1, |αi |}
i=1

where |F0 | is the constant term, |αi | are roots of F . By the assumption,
|F |, |G| > 1 and

0
M (F ) 6 |F0 | · |F |d , M (G) 6 |G0 | · |G|d
16 KELLER

So that
0 0
|R| 6 2dd · M (F )d M (G)d (Hadamard’s inequality)

dd0 0
= 2|F | · |G| · |F0 |d |G0 |d
0 0
6 (|F |2 + |G|2 )dd |F0 |d |G0 |d
0 0 0
6 cd+d · |F (w)| + |G(w)| · |F0 |d |G0 |d (d + d0 )d+d

0 0 0
6 cd+d |F (w)| + |G(w)|) · |F |d |G|d (d + d0 )d+d

Problem 17

(a). Set g(x) := (x2 + a)(x − β1 ) · · · (x − βd−2 ) where a > 0, and let
p
gn (x) = g(x) +
pdn
We see that g(x) has d − 2 real roots and gn (x) → g(x) uniformly. Let
 > 0; we may find N such that for all n > N , |gn (x) − g(x)| <  for
all x ∈ R. Choosing  > 0 sufficiently small, we see than gn (x) will be
positive and negative on the same number of intervals as g(x), in which
case the intermediate value theorem yields that gn (x) has precisely d−2
roots for all n > N .

(b). Choose a and bi , i = 1, . . . , d − 2 to all be even integers. Consider

the function
2p
gn (x) = g(x) +
pn
where p is any odd prime. We see:

pn gn (x) = pn g(x) + 2p

Employing Eisenstein’s criterion for p = 2, we see that the above are

irreducible. Dividing both sides by pn , gn (x) remains irreducible. Using
part (a), we may take n sufficiently large to conclude that gn (x) has
 SERGE LANG’S ALGEBRA CHAPTER 4 SOLUTIONS 17

precisely d − 2 real roots and 1 pair of complex conjugate roots, and is

also irreducible.

Problem 18

(a). Observe that n(n − 1) ≡ 0 mod 2, so that

n2 n
− ∈Z
2 2
for all n ∈ Z. However, we see that the above polynomial does not
have coefficients in Z.
x


(b). Note first that if f (x) = , then
i
(
x


i−1
, i>0
(∆f )(x + 1) =
0, i = 0
The polynomials      
x x x
, ..., ,
r 1 0
form a basis over Q for polynomials of degree r. Given f ∈ Z[x] of
degree d, there exists constants k0 , . . . , kr ∈ Q such that
   
x x
f (x) = k0 + · · · + kr−1 + kr
r 1
Observe that f (1) = kr , whence kr ∈ Z. Similarly, δf (1) = kr−1 ∈ Z;
proceeding inductively, ∆i f (1) = kr − i ∈ Z, so that all coefficients are
itnegral valued as contended.

(c). Suppose ∆f (n) = Q(n). By part (b),

X n
∆f (n) = ci
i
i
Now define
X x + 1
0
p (x) := ci
i
i+1
We see that ∆p0 (x) = ∆f (x), implying p0 (x) − f (x) = C for some
constant C. If we set p(x) := p0 (x) − c, we conclude that p(x) = f (x).
18 KELLER

Problem 19

(a). Suppose our polynomial f (X1 , . . . , Xn ) is homogeneous with deg(f ) >

n(n − 1). For each monomial term, the exponent of some Xi is at least
n. Recall now that by definition of the symmetric polynomials,

(t − X1 ) . . . (t − Xn ) = tn − s1 tn−1 ± · · · ± sn

Setting t = Xi , the above implies

Xin = s1 Xin−1 ± · · · ± sn ∈ (s1 , . . . , sn )

Thus if the degree is > n(n − 1), every monomial term is contained in
the ideal (s1 , . . . , sn ). By definition, this tells us that f ∈ (s1 , . . . , sn )
as desired.

(b). Observe first that Z[s1 , . . . , sn ][X1 ] has basis 1, X1 , . . . , X1n−1 over
Z[s1 , . . . , sn ]. Similarly, Z[s1 , . . . , sn ][X1 ][X2 ] has basis 1, X2 , . . . , X2n−2
over Z[s1 , . . . , sn ][X1 ].
Continuing inductively, Z[s1 , . . . , sn ][X1 ] . . . [Xi ] has basis 1, Xi , . . . , Xin−i .
To get a free basis over Z[s1 , . . . , sn ], we just take all possible products
over the above generators. We then find that

X (r) := X1r1 . . . Xnrn , 0 6 ri 6 n − i

is a Z[s1 , . . . , sn ] basis over Z[s1 , . . . , sn ][X1 ] . . . [Xn ]; note however that

Z[s1 , . . . , sn ][X1 ] . . . [Xn ] = Z[X1 , . . . , Xn ]

In which case Z[X1 , . . . , Xn ] is a free Z[s1 , . . . , sn ]-module with free

basis as above. Moreover, the rank of the above over Z[s1 , . . . , sn ] is
n!.
 SERGE LANG’S ALGEBRA CHAPTER 4 SOLUTIONS 19

(c). We employ the same technique as in part (b). Since Z[X] is a Z[S]-
module, we proceed inductively to see that Z[X, S 0 ][Y1 ] . . . [Yi ] has basis

1, X (r) , Yi , . . . , Yin−i

overZ[S, S 0 ][Y1 ] . . . [Yi−1 ]. Taking all possible products, we find that

Z[X, Y ] is a free Z[S, S 0 ]-module with basis X (r) Y (q) satisfying the con-
ditions 0 6 ri 6 n − i, 0 6 qj 6 m − j.

(d). Formally we have that J = IZ[X, Y ]. If we consider J ∩ Z[S, S 0 ],

recall that f ∈ J ∩ Z[S, S 0 ] is of the form
X
f (X, Y ) g(r),(q) X (r) Y (q)
(r), (q)

for g(r),(q) ∈ I. Since f ∈ Z[S, S 0 ] by assumption, the monomial terms

X (r) Y (q) are stmmetric for all multi-indices, implying that by the fun-
damental theorem on elementary symmetric polynomials,

X (r) Y (q) ∈ Z[S, S 0 ]

IZ[S, S 0 ] = I, so that J ∩ Z[S, S 0 ] = I.

P
But then f ∈ i

Problem 20

Let e1 , . . . , en denote the roots of f and e01 , . . . , e0m denote the roots of
g. Treating the above as variables and letting si denote the elementary
symmetric polynomial of degree i, we see that

ai = si (e1 , . . . , en )

bi = si (e01 , . . . , e0m )

Viewing ai and bi as polynomials in e1 , . . . , en and e01 , . . . , e0m , respec-

tively, we see that deg(ai ) = deg(bi ) = i. By the assumption f (t)g(t) =
1, we deduce that ki=0 ai bk−i = 0 for all k > 1.
P
20 KELLER

By definition of symmetric polynomials, note that we have

k
X
si (e1 , . . . , en )sk−i (e01 , . . . e0m ) = sk (e1 , . . . , en , e01 , . . . , e0m )
i=0

whence sk (e1 , . . . , en , e01 , . . . , e0m ) = 0 for all k < n. If ar11 . . . arnn has
degree greater than (m + n)(m + n − 1) viewed as a polynomial in
Z[e1 , . . . , en , e01 , . . . , e0m ], Problem 19 part (a) implies that ar11 . . . arnn ∈
(s1 , . . . , sn ). As all si vanish for i < n, we deduce that ar11 . . . arnn ∈
(sn ) = Z. However, the above product has no constant term, whence
the only possibility is that ar11 . . . arnn = 0 identically.
Moreover, counting degrees, we see that this condition holds when-
ever
X
jaj > (m + n)(m + n − 1)
j

Which completes the problem.

Problem 21

We show that the map x 7→ λt (x) is a homomorphism:

∞
X
λt (x + y) = λi (x + y)ti
i=0
∞ X
X i
= λk (x)λi−k (x)ti
i=0 k=0
X ∞ ∞
 X 
i i
= λ (x)t λi (y)ti
i=0 i=0
= λt (x)λt (y)

As desired. Conversely, given any homomorphism φ of the form stated,

we may define λi (x) to be the coefficient of ti of the image of x under
this map.
 SERGE LANG’S ALGEBRA CHAPTER 4 SOLUTIONS 21

By the homomorphism property and definition of product of formal

power series,

k
X
φ(x)φ(y) = φ(x + y) =⇒ λk (x + y) = λi (x)λk−i (y)
i=0

So that this is a λ-operation.

Problem 22

We shall employ multi-index notation for simplicity. Note first:

∞
X m ∞
X X 
k
ak t = tk aα
k=1 k=m |α|=k

Whence
∞
X ∞
X m
t= bm ak tk
m=0 k=1
X∞ ∞
X X 
= bm tk aα
m=0 k=m |α|=k
∞
X  k
X  X 
= tk bm aα
k=0 m=0 |α|=m

Comparing coefficients in the above, we immediately see that b0 = 0.

Next, we have a1 b1 = 1 so that b1 = a−1
1 . All powers > 2 must all

vanish; this then implies

k−1
X  X 
bk = −bk1 bm aα
m=1 |α|=m

Defining our bk inductively by the above, this defines our desired in-
verse.
22 KELLER

Problem 23

(a). Since x 7→ γt (x) is a homomorphism, we compare coefficients of

the power series representations to find
k
X
k
γt (x)γt (y) = γt (x + y) =⇒ γ (x + y) = γ i (x)γ k−i (y)
i=0

(b). Recall that λi (1) = 0 for all i whenever we have a λ-operation.

Thus,
t (1)
γt (1) = λ 1−t
t
=1+
1−t
1
=
1−t

1
(c). By the homomorphism property, γt (−1) = γt (1)
. Thus by part (b),

γt (−1) = 1 − t

Problem 24

(a). Using Problem 23 part (c), we see:

γt (u − 1) = γt (u)γt (−1)
 ut 
= 1+ (1 − t)
1−t
= 1 + (u − 1)t

(b). By part (a),

1
γt (1 − u) =
1 − (1 − u)t
X∞
= (1 − u)i ti
i=0
 SERGE LANG’S ALGEBRA CHAPTER 4 SOLUTIONS 23

Problem 25

(a). We have:
t
B0 = lim =1
t→0 et −1
−1
B1 = F 0 (0) =
2
1
B2 = F 00 (0) =
6

(b). Observe:
−t
F (−t) =
e−t
−1
tet
= t
e −1
t
=t+ t
e −1
= t + F (t)
Using this, we see that F (−t) − F (t) = t; expanding as power series:
∞ 
X  tk
(−1)k Bk − Bk =t
k=0
k!

Comparing coefficients, we see that for k > 1 odd, −Bk − Bk = 0 =⇒

Bk = 0.

Problem 26

(a). We see:
tetX
B0 (X) = lim =1
t→0 et − 1

∂F 1
B1 (X) = (0, X) = X −
∂t 2
∂ 2F 1
B2 (X) = 2
(0, X) = X 2 − X +
∂t 6
24 KELLER

(b). We compute:
∞
X Bk (X) tk
= N F (t/N, X)
k=0
N k−1 k!
tetX/N
=N·
N (et/N − 1)
tetX/N et − 1
= t · t/N
e −1 e −1
N −1
tetX/N X at/N
= t · e
e − 1 a=0
N −1
X te(X+a)t/N
=
a=0
et − 1
N
X −1  X + a
= F t,
a=0
N
∞ N
X X −1  X = a  tk
= Bk ·
k=0 a=0
N k!
Comparing coefficients in the above, we conclude:
N
X −1 X + a
k−1
Bk (X) = N Bk
a=0
N
As contended.

(c). By definition we have:

∞
X tk ∞
t X tk
= Bk , etX = Xk
et − 1 k=0 k! k=0
k!
Then, upon multiplication of the above:
∞
X tk tetX
Bk (X) = t
k=0
k! e −1
∞ ∞
X tk   X k tk 
= Bk · X
k=0
k! k=0
k!
∞ X k    tk
X k
= Bj X k−j ·
k=0 j=0
j k!
 SERGE LANG’S ALGEBRA CHAPTER 4 SOLUTIONS 25

Comparing coefficients,
k  
X j
Bk (X) = Bj X k−j
j=0
k
−1
Since B0 = 1 and B1 = 2
, we always have:
k
Bk (X) = X k − X k−1 + other terms
2
as desired.

(d) and (e). Observe first:

tetX+t tetX
F (t, X + 1) − F (t, X) = −
et − 1 − 1 et − 1
= tetX
Furthermore,
∞
tX
X tk
te =t· Xk
k=0
k!
∞ k
k−1 t
X
= kX
k=1
k!

Comparing coefficients, we deduce that Bk (X + 1) − Bk (X) = kX k−1 .

Problem 27

(a). By definition,
N −1
X f (a)e(a+X+k)t
Ff (t, X + k) =
a=0
eN t − 1
N −1
X f (a)e(a+X)t
= ekt ·
a=0
eN t − 1
kt
= e Ff (t, X)

(b). By part (a),

Ff (t, X + N ) − Ff (t, X) = eN t Ff (t, X) − Ff (t, X)

= (eN t − 1)Ff (t, X)

26 KELLER

(c). First, we compute:

N
X −1
Nt
(e − 1)Ff (t, X) = f (a)te(a+X)t
a=0
N −1 ∞
X X tk
= f (a)t · (a + X)k
a=0 k=0
k!
N −1 ∞
X X tk
= f (a) k(X + a)k−1
a=0 k=1
k!
∞ N
X X −1  tk
k−1
= f (a)k(X + a)
k=1 a=0
k!
By part (b), we also have:
∞ 
X  tk
Nt
(e − 1)F (t, X) = Bk,f (X + N ) − Bk,f (X)
k=0
k!
So that upon comparing coefficients,
 N −1
1 X
Bk,f (X + N ) − Bk,f (X) = f (a)(X + a)k−1
k a=0

(d). Observe that by part (a):

Ff (t, X) = etX Ff (t, 0)
∞ k X ∞
X
kt tk 
= X · Bk,f
k=0
k! k=0
k!
∞ X k    tk
X k
= Bj,f X k−j
k=0 j=0
j k!
Comparing coefficients in the above, we see:
k  
X k
Bk,f (X) = Bi,f X k−i
i=0
i
As contended.