Try   HackMD
tags: study notes number theory

Filling out details in Baker's Comprehensive Course

Theorem 11.2

"Then, by symmetry, we have "

We give two proofs of Baker's claim that for

fOK[X]
F=σ:KCfσZ[X]
Baker claims that this follows "by symmetry". One proof was provided to me in a comment by user @reuns and relies on the Primitive Element Theorem and the Symmetric Function Theorem; we give a second more elementary proof based on Lagrange interpolation of
F
, but it is less in the "by symmetry" spirit.

Both proofs can be found on math.stackexchange.

Theorem 12.2

Why λ_1 ⨉ ··· ⨉ λ_n = |det(A)| is the right condition

We denote by

L() the Lebesgue measure on
Rn
. Consider the set
S={xRn | Axi=1n[λi,+λi]}
Then
A(S)=[λ1,+λ1]××[λn,+λn]
and thus
|det(A)|L(S)=2nλ1λn
To apply the refinement of Minkovski's theorem with the lattice
Zn
and the symmetric convex set
S
, we need that
L(S)2n1
that is
λ1λn|det(A)|.

Theorem 12.3 (Dirichlet's Unit Theorem)

Properly applying Minkovski's theorem

Minor discrepancies

Let

n=dimQK. We recall (corollary of the primitive element theorem) that there are precisely
n
ring (field) homomorphisms
σi:KC
. We assume that the
σ1,,σs
are real and that the remaining
2t
are complex with
σs+t+j=σs+j
for
j=1,,t
.

Applying Minkovski's theorem requires

  • an
    n
    -dimensional real vector space
    V
    ,
  • n
    linear
    R
    -linear forms
    L1,,Ln:VR
    ,
  • which have to be linearly independent.

This is not quite the setup we have here. For several reasons:

  • K
    is a rational vector space,
  • the
    σi
    are
    Q
    -linear,
  • and they are are complex valued.

Fixing these discrepancies is easy enough. I highlighted them merely to pin point precisely what made me uneasy.

Linear independence of the field embeddings

The most important point, and which is quite independent from the other points, is the linear independence of the

σiLinQ(K,C). This follows from the fact that
αβK:=TrK/Q(αβ)
defines a (symmetric) nondegenerate bilinear form on
K
. Indeed,
TrK/Q(1)=n
so that
αK0
,
αα1K=n0
.

Lemma. The bilinear pairing

αβK:=TrK/Q(αβ) is nondegenerate.

Now if

θ=(θ1,,θn) is some
Q
-basis of
K
, the matrix
S=(σj(θi))1i,jn
is invertible since
SSt=Matθ(K)GLn(Q).
In conclusion,

Lemma. The field embeddings

σ1,,σn:KC are
Q
-linearly independent.

Realification

When applying Minkovski we will implicitely apply it to

V=KQR and to the maps
{Li=σi(i=1,,s)Ls+2j1=Re(σs+j)(j=1,,t)Ls+2j=Im(σs+j)(j=1,,t)
(we use the same notation for the
R
-linear extensions of
Q
-linear maps). We also notice that simple line operations convert the matrix
(Li(θj))  (σi(θj))
at the cost of multiplying the determinant by
(2i)t
. Since the matrix on the right is invertible, we get that the matrix on the left is aswell. This is the "
(aij)
"-matrix in Theorem 12.2.

Covolume of ring of integers = discriminant of number field

In this subsection we investigate why when applying Minkovski's theorem in Dirichlet's Unit Theorem, one requires

λ1λn=disc(K) rather than
λ1λn=covol(OK)
. The answer is, unsurprisingly, that
covol(OK)=disc(K)
relative to a natural volume form associated to the nondegenerate bilinear form
K
.

Proposition. Let

Q be a nondegenerate quadratic form on
V
, a finite dimensional real vector space. Then there exists a unique translation invariant measure
LQ
on
V
such that for any
Q
-orthonormal basis
LQ(i=1n[0,1]ei)=1.

The is easy:

LQ is simply the Lebesgue measure on
V
identified with
Rn
by choosing a
Q
-orthonormal basis. This is independent on the choice of
Q
-orthonormal basis. Indeed, for any base change matrix
P=MatB(C)
,
MatC(Q))=tPMatB(Q)P
Therefore, if
B
and
C
are both
Q
-orthonormal bases,
|det(P)|=1
, and
LQ
is well-defined.

Proposition. Let

Λ be a (full rank, discrete) lattice in
V
; let
θ=(θ1,,θn)
be a
Z
-basis of
Λ
. Then setting
covolQ(Λ):=LQ(i=1n[0,1]θi)
, we have
|det(Matθ(Q))|=covolQ(Λ)2

Let

θ be as above and let
e=(e1,,en)
be a
Q
-orhtonormal basis of
V
. We have
Matθ(Q))=tPMate(Q)P=tPDiag(±1)P
where
P=Mate(θ)
is the matrix of the vectors of
θ
relative to the basis
e
. Also, if we define
gGL(V)
by
g(ei)=θi
, then
g(i=1n[0,1]ei)=i=1n[0,1]θi
and so
covolQ(Λ)=|det(g)|
. Since
Mate(g)=P
this imposes
covolQ(Λ)=|det(P)|.

Corollary. If

K is a number field,
covolQ(OK)=|disc(K)|
.

The real vector space

V=KQR inherits the nondegenerate bilinear form
K
from
K
. The statement above is w.r.t. the associated volume form. Recall that the discriminant of
K
is defined as the determinant
disc(K)=det(TrK/Q(θiθj))=det(Matθ(Q))=det(P)2
of any integral basis
θ=(θ1,,θn)
of
OK
. That is,
disc(K)=covolQ(Λ)2.

Kernel of the Log map

By construction, the kernel of the Log map

LogK:{OK0Rrx(ln|σ1(x)|,,ln|σs(x)|,ln|σs+1(x)|,,ln|σr(x)|) where
r=s+t1
. Its kernel is comprised of those algebraic integers of
K
all of whose "conjugates" lie on the unit circle.

Lemma. Let

ξOK have all its conjugates of modulus
1
, then
ξ
is a root of unity, i.e. the kernel of
K
's Log map is the group of roots of unity in
K
.

Clearly, every root of unity in

K is an algebraic integer and lies in the kernel of the Log map:
μKker(LogK)
(
μK
is the group of all roots of unity in
K
). Conversely, if
xOK
has all its conjugates of modulus
1
, then its minimal polynomial lies in a finite set of integer polynomials by Kummer's argument. Therefore
ker(LogK)
is a finite subgroup of
U(K)
. If
ξ
is in the kernel, then so are all its powers, and thus
ξ
is a root of unity. If
N=#ker(LogK)
, then if follows that
ker(LogK)μN
is a subgroup of the group of
N
-th roots of unity. Equality follows from cardinality.

Note. The lemma is wrong if one changes "

ξOK" to "
ξK
": the point
35+45i
and its (sole) conjugate
3545i
have both modulus
1
yet aren't roots of unity: if they were they would have to be integers in
Q(i)=Q(1)
which they are not since their coordinates aren't integers (note
13 [4]
).

Note. The Log map clearly makes sense as a map

K×Rr where its defines a group homomorphism. For this note we only care about it along the subset
OK0
.

How and where "r = s + t - 1" is used in the proof

"r = s + t - 1" is enough

This is simply the observation that for any unit

uUK,
1=|NK/Q(u)|=|σCσ(u)|=i=1s|σi(u)|j=1t|σs+j(u)|2
so that
0=i=1sln|σi(u)|+2j=1tln|σs+j(u)|
In other words, the more natural "full Log map"
LogK(u)=(ln|σi(u)|)1is+t
has its image in the hyperplane
1isxi+2s<js+txj=0
. The "missing modulus"
ln|σr+1(u)|=ln|σs+t(u)|
is redundant information.

How the precise value "r = s + t - 1" is used

The beginning of the proof works without modification for "shorter Log maps" (i.e. by including fewer than

r of the
σi
). The fact that
r=s+t1
isn't used until one invokes the proposition below.

Let

η1,,ηrUK be the family of units constructed in Dirichlet's Unit Theorem (DUT). By construction, the vectors
LogK(ηi)
,
i=1,,r
, are linearly independent in
Rr
, and
η1,,ηr
[1] is a free abelian group of rank
r=s+t1
within
UK
.

Proposition. The quotient group

UK/η1,,ηr is finite.

Proof. This is the proposition that is used in the proof of the DUT. Its proof is basically that of corollary 12.1. Let

uUK:

  • u
    is equivalent (up to an element in
    η1,,ηr
    ) to a unit
    v
    such that for all
    σ:KC
    ,
    |σ(v)|C
    for some absolute constant
    C>0
    ;
  • v
    's minimal polynomial thus belongs to a finite set of integer polynomialls;
  • thus
    v
    belongs to the finite set of the
    K
    -roots of these polynomials,
  • and thus
    UK/η1,,ηr
    is finite.

The unit

v is constructed as in Baker's proof by taking a nearby vector
Log(η1m1ηrmr)
to
Log(u)
in the (discrete, full rank) lattice
Log(η1,,ηr)Rr
and setting
v=uη1m1ηrmr.
Here "nearby" means w.r.t. the
-norm on
Rr
. We are not necessarily interested in the closest lattice point, merely in the fact that there is a uniform error bound
δ>0
on the coordinates
|ln|σi(v)||
. This is clear, for by construction, if
v
was constructed by

  1. first: expressing
    Log(u)
    in the
    R
    -basis
    (Log(η1),,Log(ηr))
    as
    Log(u)=i=1ruiLog(ηi)
  2. defining
    mi=ui
  3. setting
    v=uη1m1ηrmr
    ,

then for

i=1,,r:
|ln|σi(v)||rmaxj=1,,r|ln|σi(ηj)||rmaxi=1,,rmaxj=1,,r|ln|σi(ηj)||=δ

The only noteworthy point is how one deduces the boundedness of the length of the "single missing conjugate"

|σr+1(v)|=|σs+t(v)|. We get it from the previously made observation:
0=i=1sln|σi(v)|+2j=1tln|σs+j(v)|thus|ln|σs+t(v)||i=1r|ln|σi(v)||rδ
(not bothering with the
12
factors one would expect on the first
s
terms allows us not to have to differentiate between the case
t>0
and
t=0
.) This (trivial) computation is precisely where we use the fact that we used
r=s+t1
: had we chosen a value lower, we would not have been able to untangle the remaining moduli. Taking
r=s+t1
allows us to have a single modulus to dominate.

The remaining points are easy consequences of the boundedness of the conjugates.

The conclusion of the proof

Once Baker knows that the quotient is finite, the proof is nearly done. We put

N=#(UK/η1,,ηr). Then
UKNη1,,ηr
[2], where we set
UKN={ϵNϵUK}
. We note that
LogK(UKN)
includes the full rank sugroup
LogK(η1N),,LogK(ηrN)=NLogK(η1),,LogK(ηr)
of
LogK(η1),,LogK(ηr)
, hence has full rank itself. Using Lemma 11.3, there exists a basis
(LogK(ϵ1N),,LogK(ϵrN))
of
LogK(UKN)
such that its matrix w.r.t. the basis
(LogK(η1),,LogK(ηr))
is upper triangular. I don't believe the "upper triangular" property is important here, though.

If

uUK is a unit, then there exist
j1,,jrZ
such that
LogK(uN)=k=1rjkLogK(ϵkN)
therefore
uϵ1jrϵrjr  some root of unity
(indeed, raising the LHS to the
N
-th power lands it in the kernel of the Log map; the kernel of the Log map is the set of roots of unity in
K
; the LHS is thus a root of unity in
K
) i.e.
u=ρϵ1jrϵrjr
for some
N
-th root of unity
ρUK
.

Wrapping up

In conclusion, we get a split short exact sequence since its final term is free abelian:

0μKUKLog(UK)Zr0 and so
UKμK×Zr
.


  1. We use here the notation

    g1,,gr for
    g1,,grG
    to denote the subgroup of a group
    G
    generated by the
    gi
    . This should not be confused with the ideal
    η1,,ηrOK
    generated by the
    ηi
    . Since ↩︎

  2. In this subsection we write

    v1,,vn for the subgroup of some group
    Λ
    generated by elements
    v1,,vnΛ
    . Here
    Λ=Log(UK)Rr
    is actually a lattice. There should be no possible confusion with the analogous notation for ideals. ↩︎