The divisor group of a fir
Abstract
Cohn, P. M.
Full text
Pub
.
Ma
.
UAB
Vol
.
26 N°
1
Ma a
1982
In oduc ion
THE
DIVISOR
GROUP
OF A FIR
P
.M
.
COHN
Bed o dCollege
Regen 's
Pa k
London,
NW1
4NS
Rebu
el
30
d'oc ub e
del
1981
The e
ha e been
many
a emp s
o
de ine
de e minan s
on
non-commu a i e
ings
(c
.e.g
.[13]
and
.
he
e e ences
quo-
ed
he e),
o
which
pe haps
he
mos
success ul
is
he
de ini
ion o
Dieudonné
[10],
leading
o any skew
ield
K
and
any
n
>
1
(excep
when
n=2 and K=F
2
)
o
an
isomo phism
(1)
GLn(K)
ab
-
K
*ab
.
Suppose
now ha
K
is
ob ained
om
a
ing
R
by
in
e ing
ce ain
ma ices
o e
R,
o ming
a
se
E
.
The
way
in
which
he
elemen s
o
K
a e
ob ained
om
R
and
E
was
desc ibed
in Ch
.7
o
[3], and we
mayask
whe he
GL
n
(K)
ab
can
be
desc ibed
di ec ly
in
e ms
o
R
and
E
.
Since
(1)
is
an
isomo phism
o
all
n, we
can
limi
ou sel es
o
a
sin
gle
alue
o
n, o
we
maysimply
ake
he
limi
GL(K)=1im
GL
n
(K)
.
Ou aim he e
is o
desc ibe
he
Whi ehead
g oup
K
1
(K)~
GL(K)
ab in
e ms
o E
;
his
can
be
done
unde
ai ly
gene al
condi ions,
hough
o
mo e
p ecise esul s
we
need
o
ake
R
o
be
a
i
and
K
i s
uni e sal ield
o
ac ions
.
In
pa i
cula ,
by
aking
R
o be
a
ee
associa i e
algeb a
we
ob-
ain
an
explici
exp ession
o
de e minan e
o e
a
ee
ield
(Th
.5
.2)
.
To
s a e
he
esul e,
le
:R
->
K
be
a
homomo -
phism
o
any
ings
and
suppose
ha
e e yelemen
o
K
can
be
ob ained
om
he
en ies
o he o mal
in e ses
o
he
ma i
ces
om
a
se E
,
which
is
mul iplica i e
(as
de ined
below,
c
.
aleo[3],
p
.249),
hen
i
is
no ha d
o
show ha
in-
duces
an
epimo phism
o
abelian
g oups
(2)
*
:
Eab
->
K1(K),
whe e
E
ab
is
he
uni e sal
abelian
g oup
o
E
(Th
.2
.2
and
Co
.)
.
In
gene al
he e
is no
eason
o *
o be
injec i e,
bu
when
K
is
he
uni e sal
ield
o
ac ions
o
a
Syl es e
domain
R
and
E he
se
o
al]
ull
ma ices
o e
R,
hen
(2)
is an
isomo phism
.
This
is
p o ed
( o
he
sligh ly
la ge
clase
o
pseudo-Syl es e
domains)
in
Th
.3 .1
by
cons uc ing
an
in e se
ma)pping
o
*
.
Fo
a
somewha
di e en
ea men
o
he
same p obl
em
see
[121
and
al
so
[61
.
Fo
Syl es e
domains
i
is
di icul
o
say
mo e
be-
cause
li le
is
known
abou
ac o iza ion
in
such
ings
.
Bu
when
we ha e
a
i
R
,
o
.mo e
gene ally
a
ully
a omic
semi
i
(i
.e
.
one
in
which
e e y
ull
ma ix
can
be
exp essed
as
a
p oduc
o
a oms)
hen
a
mo e
p ecise
s a emen
is
possible
.
In
R
de ine
apn,íme
as
a
class
o
s ably
associa ed
a oms
and
he
d
.í
.í,&wc
g oup
D(R)
as
he
ee
abelian
g oup
on
al] he
p imes,
and le
U
be he
uni e sal
ieldo
ac ions
o
R
,
henwe
p o e
in
Th
.4 .4
ha
K 1
(U)°-
U*ab-
D(R)
x
[GL(R)/GL(R)nGL(U)']
.
In
pa icula ,
when
R
= k
<
X
>
is
a
ee
algeb a,
his be-
comes
U*ab
=
D
(R)
x
k*
(c
.
Th
.5.2
.)
.
These
esul s
ha e
also
been
ob ained
by
G
.Ré-
ész
[12] by
a
di e en me hod
.
Ou
second
main
esul
is
conce ned
wi h
localiza ion
o
i s
.
Le
R
be
a
ully
a omic
semi i
and
E
a
mul ipli-
ca i e
se o
ma ices
such ha
R
.
is
again
a
semi i ,
hen
R,,
is
also
ully
a omic
and he
di iso
g oupo
R
..
is
iso-
mo phic
o
D~(R),
he
subg oup
o
D(R)
gene a ed
by
he
p imes
which
su i e
in
R,,
(Th
.6
.3)
.
I
am
indeb ed
o
G
.M
.
Be gman
o
his
ex ensi e
com-
men s
on an
ea lie
e sion,
and
o
G
.
Ré ész
o
se e al
help-
ul
ema ks
.
1
.
No a ion
and
gene al
backg ound
Le
R
be
any
ing
;
we
w i e
m
Rn
o
he se o
al]
m x
n
ma ices
o e
R
and also pu
MR
o
M R
1
and
R
n
o
1
R
n
.
The
chanac enía íc
o an
m x
n
ma ix
is
de ined
as
n -
m
.
I
a
ma ix
is
exp essed
in
block
o m
(P), we
o en
w i e
his
as
P,
T
(
Q)
o
sa e
space
;he e
T
indica es
ha he
blocks
P, Q
a e
o be
w i en
as
a
column,
bu a e
no
hemsel es ansposed
.
Simila ly
o
mo e
han wo
blocks
.
The
d iagonal
sum o wo
ma ices
A,
B
is
de ined
as
A +
B =
(A
B)
.
The
se
o
all
in e ible
n
x n
ma ices
o e
R
is
deno ed
by GL
n
(R)
and we
embed
GL
n
(R)
in
GL
n+1
(R)
by
he
ule
A
1
-
=
A
+
1
.
Fu he
we
pu
GL(R)
_
lim
GL
n
(R)
.
As
usual,
a
ma ix
is
said o
be
elemen a y
i i
di e s
om
he
uni
ma ix
in
a
mos
oneo -diagonal
en-
y
;
he
g oup
gene a ed
by
all
elemen a y
n
x n
ma ices
is
w i en
E
n
(R)
and as
be o e
we
pu
E(R)
=
lim`En(R)
.
Fo
any
A
E
m Rn
,
he
leas
in ege
such
ha
A =
PQ,
whe e
P
E
M
R
,
Q
E
R
n
is
called
he
inneh
hank
and
-a
ma ix
is
said
o
be
jul'l
i i
is
squa e,
say
n x n
and
o
inne
ank
n
.
In
gene al,
i
A
is
ull,
i
need
no
be
he
case ha
A+
1
is
ull
;
i
A +
I
is
ull
o any uni
ma ix
I,
A is
calleds ably
bul'l
and he se
o
al]
s ably
ull
n
x n
ma ices
o e
R
is
w i en
F
n
(R)
;
we
embed
F n
(R)
in F
n+1
(R)
as o GL(R) and
w i e
F(R)
=
lim
F
n
(R)
.
Some imes
we
shall need
a
gene aliza ion
This
limi
always
exis sand
is
an
in ege
o
-
-,
bu
i
I
n
is
ull
o all
n,
he
inne
ank
is
ac ually
non-nega i
e
(c
.
[81)
.
'We
no e ha
an
n
x n
ma ix
is
s ably
ull
p ecisely
when
i
has
s able
ank
n
.
Two
ma ices
A,
B
a e
said
o be abeoc ia edi
A =
PBQ o some
in e ible
ma ices
P,Q
.
I
P, Q
can
be
aken
in
E n
(R),
we
call
A
and
B
E-assoc ia ed
.
I
A +
I
is
(E-)
associa ed
o
B
+
1
(whe e
he
uni
ma ices
need
o he
inne
ank
.
The
zxabl'e
nank
o
a
ma ix
A
is
de ined
as
lim
{ k(A
+
I
n
)
-
n},
whe e
k
deno es
he
inne
ank
.
no
be
o
he
same
size),
henwe say
ha
A
and
B
a e
e ably
(E-)
associa ed
.
Twos ably
associa ed
ma ices
a e
no
necessa ily
o
he same size, bu
hey
ha e he same
cha-
ac e is ic,
p o ided
ha
R
is
a
ing wi h
in a ian
basis
numbe
(i
.e
.
all
in e ible
ma ices
a e
squa e)
.
By
a
bíel'd
we
unde s and
a
no
necessa ily
commu a i-
e
di ision
ing
;some imes
we
use
he
p e ix
'skew'
o em-
phasis
.
I
G
is
any
g oup,
i s
de i ed
g oup
is
deno ed
by
G'
and
we
w i e
G
ab
=
G/G'
o
he
abelianiza ion
o
G
.
Fo
a
ield
K
we
w i e
K* o he
g oup
o
i s
non-ze o
elemen s
and
by
abuse
o
no a ion
we
simply
w i e
K
ab
o
K*
ab
.
I
R'
is
any
ing,
an
R-6 ield
is
a
ield
K
wi h
a
homomo phism
R
'
K
;
i
K
is
gene a ed,
as
a
ield,
by
he
image
o
R, we
speak
o an
epic
R- ield
.
An
m
x
n
ma ix
o e
a
ield
K,
o
ank
,
is
said
o
ha e
l'eb
nullí u
m- and
nígh
nul'l'íxy
n-
.
These
nulli ies
a e
only
de ined
o e
a
ield,bu
i
A is
a
ma
ix
o e
R
and
K
is
any
R- ield,we
can
conside
he nu-
lli ies
o
A
o e
K
; heya e
simply
he
nulli ies
o he
image
o
A
in
K
.
We shall
no
epea
he
de ini ions
o
i and
semi i
(c
.
[31,
Ch
.
1
o
[41,
Ch
.
4)
.
We
me ely
ecall
ha
e e y
semi i
R
has
a
uni e sal
ieldo
ac ions
U,
ob ained
by
o mally
in e ine
all
ull
ma ices
o e
R
(c
.
[31,
p
.
282
.)
.
Mo e
gene ally,
i
R
is
any
ingandE
a
se o
squa e
ma ices
o e
R,
hen
a
map
:
R
=
Sis
called
a
E-ín eh ing
ep imonphíbm
i
is
a
homomo phismmapping
each
ma ix
o E
o an
in e ible
ma ix
o e
S
and
S
consis s
o
he
en ies
o
in e ses
o he
ma ices
A
,
AEE
.
I is
easily
seen ha his
is
in
ac
an
epimo phism
in
he
ca ego y
o
ings
.
In
wha ollows,
Ewill
gene ally
be
mul
ipl' ica í e,
i
.e
.
1
E
and
i
A,
B
E
=
-
S,
hen
(
A
B)
E E
o all
C
o
app op ia e
size
.
Fo
any ing
R
and se E
o ull
ma ices
o e
R,
he
localíza íon
Rz
:
o
R
by
E
is
de ined
as
he ing
ob ained
om
R
by
o mally
in e ing
all
he
ma ices
in
E
.
Then
in
any
E-in e ing
epimo phism
:
R
'
S,
S
is
clea ly
a
homomo phic
image
o
RE
.
Suppo
se
now
ha
s
is
mul iplica i e
; hen
each
elemen
o
S
may
be
ob ained
as he las
componen
u
n
o he
solu ion
o
a
ma
ix
equa ion
(1)
A
u
=
0,
A
=
(Ao,A1,
.
.
.,An)
E
n R
n+1
,
whe e
(A
l
.. .
.,A
n
)
EE
and
u
=
(1,u1,
.
.
.,un)T
.
I
p
= u
n
,
we
say
ha
(1)
is an
S-admissible
sys em
o
p
and
call
(Ao,A1,
.
.
.,An-1)
he
numenazon,
(A
1
.
.,A
n
)
he
denom ina on
o
p
(c
.
[5],
§
4)
.
I
is
o en
con enien
o
pu
(Al$
. .
.,An-1)
=
A
*
and
A
n
= A
.), hen he
nume a o
will
be
(A 0
,A
*
)
and he
denomina o
(A
*
,
A.)
.
2
.
The
calcula ion
o
K
1
o
a
localiza ion
I is
a
well
known
ac ha
o any
ing
R,
E(R)
=
GL(R)'
(c
.
[1],
V
.1
.5,
p
.223),
so
ha
GL(R)
ab
=
GL(R)/E(R)
;
by
de ini ion
his
is
he
Whi ehead
g oup
K 1
(R)
.
I we
ha e
a
skew
ield
K,
hen o
any
non-singula
ma ix
A
o e
K
he e
exis s
a E
K*
such ha
A
--_
a
(mod
E
(K)),
and
he e a
is
de e mined
mod
K*'
.
The
esidue
class
o
a
in K
ab
is
called
he
I)íeudonné
de enmínan
o
w i en
de
A
.
We
no e
ha
by
(1),
o
any
skew
ield
K
.
Suppose
now
ha
R,
S
a e any ings, E
is a
mul i
plica i e
se o
ma ices
o e
R
and
:R
'
S
is
E-in e
ing
epimo phism
.
Ou
objec
is o
exp ess
K
1
(S)
in
e ms
o
R
and
E
.
In
o de
o
do hiswe
need
o
exp ess
ma i-
ces
o e
S
as
solu ions
o
sys ems
o
equa ions
o e
R
(as
was
done
in
(1)
o
§ 1
o
elem
.
en s)
.
P oposi ion
2
.1
Le
R,
S
be
any
nínga,
E
a
mu.l íplíea í-
e
ee
o6
ma
. níeee
o en
R,
andle
:R
--
=
S
be
a
E-ín en
.íng
epímonphísm
.
Gí en
any
P
E
m S
n
,
he e
exíb s
an
ínxegen
'>
0
and
ma níces
(A
o
,AA
)
E
+m
R
n+ +m
,
u
=
(u
u*,u,)
T
E
n+ +m
S
n
(3)
A =
,
*,
~
o
,
whene
zhe
numbena
ob
eolumne
oj
Ao,A*,A
.
and
2í zewíbe
he
numbenh
oj
nows
oj
uo,u*,u,o
ane
n, ,m,
nespee í ely,
eueh
xha
(A
*,A
)
CE,
a
ab
A,
one
non-ze o
en y,
since
he
gene al
case
may
be
ob ained
by
adding
such
ma ices
.
By ow and
column
ans o ma ions
we
can
educe
e e y hing
o he case
P =
(p
o),
p
E
S
.
Le
A
u =
0
is
a
sys em
o
P,
as
equi ed
.
and
(5)
u
o
=
I,
u
.
=
P
.
Moneo en,
u
ís
he
uníque
elemen
o5
n+ +m
S
n
sa ís1yíng
(4),
(5)
hon
he
gí en
ma níx
A
.
We
shall
cal]
A
an
S-admíssíble
sys em
o
P
.
P oo
.
The
uniqueness
o
u
ollows
om
(4), (5),
since
any
ma ix
in
E
is
in e ible
o e
S
.
To
p o e
he
main
asse ion
we
no e
ha
i
i
holds
o wo
ma ices
P',
P"
E
m
S
n
,
hen
i
also
holds
o
P
=
P'
+
P"
.
Indeed,
i
P',
P"
a e
de e mined
by
sys ems
Al
u'
=
0,
A"
u"
=
0,
analogous
o
A
abo e,
hen
(as
in
he
case
o
elemen s,
c
.
[3],
p
.250),
P
is
gi en
by
he
sys em
A¿
A*
A,~
0 0
Í
(I,u*,P',u*,P)T
=0
.
A"
0
-A"
A" A"
1
o
00
*
00
Hence
i
su ices
o
p o e
he
esul
o
ma ices
wi h
only
0
be an
S-admissible
sys em
o
p,
hen
A
o0 A
*
A
.
0
1
0
~0
1
n-1
0 0 0 0
1
m-1
u * 0
=
p 0
0 0
The
equa ion
(4)
may
again
be
w i en
in a
o mo
C ame 's
ule
(131,
p
.251
and
[51,
§ 4)
lde
shall
again
cal]
(A
*
,
A
.) he
denomina o
o he
sys em
A,
bu
de ine
he
nume a oA
as
(A
*
,
-A
0
)
.
This
is
igh
associa ed
o
he
nume a o
as
de ined
in
[51
(and
ecalled
in
§
1),
so
he
change
has
no
e ec
on
he
conside a ions
o
[51
excep
no a ionally
.
F om
(6)
we
see
ha
P
is
s ably
associa ed
o e
S
o
i s
nume a o ,
in
pa icula
i
has sa-
me
cha ac e is ic,
and
i
is
in e ible
i
he
nume a o
is
in e ible
o e
S,
i
.e
.
i
he
sys em
can
be
chosen
so
as
o
ha e
a
nume a o
in
E
.
Mo eo e ,
when
S
is
a
ield,
he
le
and
igh
nulli ies
o
P
o e
S
ag ee
wi h
hose
o
i s
nume a o
.
Le
R, S
be
any
ings,
E
a
mul iplica i e
se
o
ull
ma ices
o e
R
and
:R
'
S
a
E-in e ing
epimo -
phism
.
Since
E
is
mul iplica i e,
i s
ma ices
a e
e en
s a
bly
ull
and we
may
embed
E
in
F(R)
by
he
ule
A
F->
A
+I
;
his
allows
us o
ega d
E
as a
submonoid
o
F(R),
wi h
he
mul iplica i n
AB
.
Now
conside
he
uni e sal
abelian
g oup
Eab
o E
;
his
is
de ined
asan
abelian
g oup
,ab
wi h
a
homomo phism
E
'
sab
which
is
uni e sal
o
all
homomo phisms
o E
in o
abelian
g oups
.
To
desc ibe
,ab
explici ly,
le
us
deno e
by
[A1
o
[Al
E
when
con u-
sion
is
possible,
he
class
o
A
E
E
unde
s able
E-associa-
e minan s
o
ma ices
o e
pseudo-Syl es e
domains
:
Co olla y
.
Le
R
be
a pseudo-Sy
.C ee e
domaín
and
U
í h
uní enba
.C
(íeld
oj
bnae
. íona,
hen
bo
any
z ab1y
6u
.Cl
ma níx
A
o en
R
we
ha e
whene
íz
hemap
(1)
índueed
by
:R
-
U
andse
íh
alzen
o en
U
.
Fo
o e
U, A
is
s ably
E-associa ed
o
a E
U,
such
ha
a
ab
=
se
A
.
Hence
[Al
=
aab
=
se
A
.
Fo
Syl es e
domains
Th
.3
.1
has also been
ob ained
by
G
.Ré ész
[121,
by
ano he
me hod,
based
on
he
abo e
Co-
olla y
( o
he
case
o i s
he e
is
ye
ano he
p oo
in
[61
)
.
4
.
The
di iso
g oup
o
a
ully
a omic
semi i
In
o de
o
in es iga e
he
s uc u e
o
U
ab
mo e
ully
we need
o
assume
he
exis ence
o
comple e
ac o iza-
ions
in
ou ing
R
.
We ecall
ha
a
squa e
ma ix
A
is
calles
an
a om
i i is
a
non-uni
and
canno
be
w i en
as'a
p oduc
o wo
(squa e)
non-uni
ma ices
;
i
is
clea
om
his
ha
an
a om
is
necessa ily
ull
.
A
ing
is
said
o
be
bully
a
.íomíe
i
e e y
ull
ma ix
can be
w i en
as
a
p oduc
o
a
ini e
numbe
o
a oms,
o
is a
uni
.
In
pa icula ,
e e y
elemen
no
ze o
o
a
uni
hen
has
a
comple e
ac o i-
za ion
in o a oms
.
Le
R
be
a
semi i
and
U
i s
uni e sal
ield
o
ac ions
.
By
a
Z- al'ue
on
R
we
shall
unde s and
a
homomo -
phism
:GL(U)
->
Z
such
ha
(A)
>
0
o
all
A
E
F(R)
.
To
gi e
an
example,
le
us
assume
ha
R is a
ully
a omic
semi i
and
ecall
om
(3],p
.201
he
unique
ac o i-
za ion
p ope y
:
E e y
ull
ma i
.
x
o e
R is
ei he
a
uni
o
has
a
ac o iza ion
in o
a omswhich
is
unique
up o
s able
associa ion
and he
o de
o
he
ac o s
.
Now
le
P
be an
a om
and o any
A
E
F(R)
de ine
(A)
=
i
in
any
comple
e
ac o iza ion
o
A
he
numbe
o
a omic
ac o s
s ably
associa ed
o
P is
jus
.
By
unique
ac o iza ion
his
is
well-de ined
and we
ob ain
a
Z- alue
on
R
by
pu ing
(
A]
.F
-
[B]
F)
=
(A)
-
(B)
.
This
is
called
he
6ímp
. e
Z- alue
associa ed
wi h
he
a om
P
.
P oposi ion
4 .1
.
Le
R
be a
bull'y
a omíc
semílín
and
le
be any
Z- alle
on
R
.
Then
(i)
(P)=0
bo
PEGL(R),
and
(ii)
(A)= (A')
whene en
A,
A'
a e
6 abl?y
aaeoeía ed
.
P oo
.
(i)
Le
P
E
GL(R),
hen
(P)
>
0,
(P
-1
)
>
0,
bu
(P)
+
(P
-1
)
=
(I)
=
0,
hence
(P)
= 0
.
(ii)
Le
A, A'
be
s ably
associa ed,
say
(A +
I)U
=
V(A'
+
I),
U,V
EGL(R)
;
since
(U)
=
(V)
=
0,
we ha e (A)
=
(A')
as
claimed
.
Le
us
de ine
a
p íme
o
R
as
a
classo
s ably
asso
cia ed
a omic
ma ices
.
Wi h each
p ime
p
i
he e
is
associa-
ed
a
simple
Z- alue
i
.
Mo e
gene ally,
picic
an
in ege
n
i
>
0
o
each
p ime
p i
,
hen
w
=
En
i
i
is
a
Z- alue,
o
i
is
de ined
on
each
ull
ma ix
A
:
w(A)
=
En
i
i
(A),
whe e
he sum
on
he
igh
is
ini e
because
i
(A)
=
0
o
almos
all
i .
We
obse e
ha
e e y
Z- alue
a ises
in
his
way
;
o i w
is
a
Z- alue
on
R,
le
P
i
be an
a om
in
he
class
p
i
and pu
n
i
=
w(P
i
),
hen
w
and En
i
i
ha-
e
he same
alue
on
eacha om
and
hence
on
al]
o
F(R)ab,
Theo em
4
.2
.
be
he
s
ímple
Fon,
any
lam
.í
.Cy
Z- alue,
and
eon en
.bely,
e e y
Z- alue
on
R
íz o6
h
.í~s
6onm
.
We
ema k
ha wi h
e e y
ull
ma ix
associa ed
a
Z- alue
w
A
which
is
simple
i
is
an
a om,
iz
.
wA
=
En
i
i ,
whe e
he
i
ple
Z- alues
and
n
i
=
i(A)
.
We can
also
use
Z- alues
o
semi i s
:
P oposi ion
4
.3
.
Lex
R
be
a sem
.í6
.íA,
hen
m,íe
.í6
and
on1y
.í6
Zhene
.íe
a
Z- alue
w on
w(A)
=
0
pnee
.íeelN
¡ohen
A
.íó
a
un
.í
.
P oo
.
I
R
is
a
ully
a omic
semi i
and
i
ple
Z- alues
co esponding
o
he
di e en
p imes
o
R,
hen
w =
E
i
has
he
desi ed
p ope y
.
Con e sely,
when
w
exis s,
ake
any
ac o iza ion
A E
F(R)
and
ac o ize
i
in o
non-uni s
in
anyway
:
This
p o es
Le
R
be a
6ully
a om
.íe
zem
.í6
.ín
andle
( i)
Z- aluea
eon
.n
.ezpond
.íng
o
he
pn
.ímeb
01
R
.
(n
i )
o6
non-nega
.í e
.ín egena,
In
i
i
íh
a
A he e
is
and
only
i
A
a e
al]
he sim-
cha ac e ize
ully
a omic
R .íz
6ully
a o
R
such
ha
a e he sim-
(2)
A
=
P
1
-
.P
.
Since
w(P
i
)
>
1
by
hypo hesis,
we ha e
w(A)
=
Ew(Pi)
>
,
and his
p o ides
a
bound
on
he
numbe
o
ac o s
in
(2)
.
By
aking
a
ac o iza ion
wi h
maximal
we
ob ain
a
comple-
e
ac o iza ion
o
A
.
This
comple es
he
p oo
.
Now
ake
a
ully
a omic
semi i
R
and le
pi (i
E
I)
be
he
amily
o
all
p imes
.
Fo
each
p
i
we
ha e
a
homomo
phism
:
i:
F(R)ab
-
=
Z,
and
combining
al]
hese
maps,
we
ha
e
a
homomo phism
and hence,
by
Th
.3
.1,
F(R)ab
,.
ZI
.
Bu
each
ull
ma ix
maps
o
0
in
almos
al]
ac o s
o
Z
I ,
hence
he
image
lies
in
he
weak
di ec
powe
Z(
I
)
.
Le
us
w i e
D =
D(R) o he ee
abelian
g oup
on
he
p
i
(w i-
en
addi i ely)
.
hen
we
ha e
a
homomo phism
X
:
F(R)ab
->D
K
1
(U)
-
D(R)
.
F om
i s
cons uc ion
he map X
is
su jec i e,
hence
so is
(3)
.
We
claim
ha
i s
ke nel
is
GL(R)/(GL(R)
n
E(U))
.
Fo
any
A
E
GL(R)
sa is ies
i
(A)
= 0
o al]
i,
hence
A
E
ke
X
*
.
Con e sely,
i
([
A]
-
[
B]
)
x*
=
0,
hen
~
=
B~
,
hence
A,
B
ha e
he same
a omic
ac o s,
up
o
o de
and
s able
associa ion
.
Le
A
=
P
1
-
P
be
a
comple e
ac o i-
za ion
and le
B
be
he
p oduc
(in
some
o de )
o
Q
1
, . .
.,
Q
,
whe e
Qi
i .
s
s ably
associa ed
o P
i .
Replacing
A,
B
by
A +
I,
B
+
I
o
sui ably
la ge
I,
we
mayassume
Q
i
o
be
associa ed
o
P i
,
say
P
i
=
U
¡
Q
¡
V
i ,
whe e
U
i ,
V
i
EGL(R)
.
Then
excep
o he
o de
o he
ac o s
we can
w i e
A
=
Q
1
.
.
.Q
U 1
. .
.U
V
1
. .
.V
=
BF,
whe e
FEGL(R)
.
Hence
A
=
BF
(mod
GL(U)')
and
so
[Al
-
[B]
_
[
F]EGL(R)
.GL(U)'
.
I
ollows
ha ke
X =
GL(R)
.GL(U)'/GL(U)
-
GL(R)/GL(R)
n
GL(U)'
.
He
e
we may
eplace
GL(U)'
by
E(U)
;
mo eo e ,
since
D
is
ee
abelian,X
is
spli
by
D
o e
i s
ke neland
we
ob-
ain
Theo em
4
.4
.
Ley
Rbe
a
bul
.2y
a
. amíe
eemí,Iín
w
.ízh
uní enha2
6íeld
ob
gnae
. íanó U
and
d
.wíeangnaup
D(R),
hen
(4)
K
1
(U)
=
U
ab
-
D
xJ
[GL(R)/(GL(R)
n
E(U))]
.
The
di iso
g oup
D
inhe i s
a
pa ial
o de ing
om
R, by
w i ing
i >
0
whene e
n
is
posi i e
on
R
.
Howe e ,
he
o
de ing
on
D is
no
enough
o
de ine
R
wi hin
U,
as is
shown
by
he
ac
ha he
de e minan
o
a
ma ix
o e
R
is
usually
a
p ope
ac ion
(¡
.e
.
has
no
ep esen a i e
in
R)
.
I
is
also o
in e es
o
compa e
Z- alues
wi h
al a
ions
(c
.
[11])
.
Clea ly
a
Z- alue
will
be
a
alua ion
i
and
only
i
Le
A
=
(A
0
,
A*
,
A
.)
be
an
admissible
ma ix
o
p,
hen
an
admissible
ma ix
o
p-1
is
(Ao+A
.,A*,A
.),
so
he
condi-
ion
(5)
becomes,
a e
a
sligh
ea angemen ,
(
6)
(A
o + A
w
,A
*
)
>
min
{ (A
o
,A
* ),
(A,',
A
*
)}
.
We
ecall
ha when wo
ma ices
di e
in
only
one
column,
say he
i s
:
A
=
(A1,
.
.
.,An),
B
=
(B 1
,A
2
,-
,A
n
),
hen
he
ma ix
ob ained
by
adding
he
i s
columns
and
lea ing
he
o he columns
unchanged
is
called
he
de enmínan al
sum and
is
w i en
A
.
B
=
(A
1
+
B
1
,A
2
-
.-
A
n
) .
Wi h
.his
no a ion
we
see
ha
is
a
alua ion
i
and
only
i
(7)
(A
B)
>
min
{ (A), (B)},
whene e
he
de e minan al
sum
is
de ined
(c
.
[111)
.
In
gene
al
his
condi ion
need
no hold,
e .g
.
in
k
<x,y'>
consi-
de he
simple
Z- alue
associa edwi h
x
.
We
ha e
(xy)
=
(yx)
= 1,
bu
(xy
-
yx)-=
0
.
Ne e heless he e
is a
alua ion
on
he
uni e sal ield
o
ac ions
U
asso-
cia ed
wi h
x
;
o
ob ain
i
we
w i e
U
as
a
skew
unc ion
ield K(x
;a),
whe e
K
is
he
uni e sal
ieldo
ac ions
o
k
<y
¡
¡
i
E
Z>
and
a
is
he
shi
au omo phism
y
i
I
'
Y¡
+i ( hus
yi
is
ealized
as
x -i
yx
i).
0n
K(x
;a)
he
o de
in
x
is
he
equi ed
alua ion
.
In
e ms
o Z- a
lues his
alua ion
is
ob ained
as
he
sum
o
ce ain
simple
Z- alues,
bu
his
is
no
a
e y
e icien
way
o cons uc ing
his
alua ion
.
5
.
The
case
o
ee
algeb as
To
illus a e
Th.4.4
we
shall
conside
he
case
o
ee
algeb as,
whe e
i
is
possible
o
compu e
he
second
ac-
o on he
igh
o
(4)
o
§4
.
We
i s
p o e
a
lemma
.
Lemma
5
.
1
.
Le
k
be
a
commu a
. í e
bíeld
and
U=k
-~
X'~
he
uní enaal
6íe
.Cd
01
b ae íona
ob
he
linee
k-algeb a
k<
X
>
,
xhen
E
(U)
nGL
1
(k)
=
1
.
P oo
.
Le
A=a+
lEE(U),
whe e
a
E
-
=
k
;
we ha e
o
show
ha
a
=1
.
W i e
A
as
a
p oduc
o
elemen a y
ma ices
o e
U
and
le
P
be
he
diagonal
sum o
al]
he
denomina o s
o he
en ies
occu ing
in
hese
ma ices
.
Ou plan
will
be o
ind
a
k- ield
K
such
ha
we can
specialize
X
o
alues
in
K
so
ha
P
emains
in e ible
and A
maps
o
I .
Fo
each
n
no
di isible
by
X,
he
cha ac e is ic
o
k,
we
adjoin
a
p imi i e
n h
oo o
1,
co
n
say, o
k
and
de ine
(1)
K(n)
=
k(x,y
I
yx
=
w
n
xy)
.
I
is
easily
seen
ha
K(n)
is
hen
a
skew
ield,
in
ac
a
di ision
algeb a
o
index
n
.
Le
K
be an
ul ap oduc
o
he
K(n)
wi h
a
non-p incipal
ul a il e ,
anddeno e
by
x',
y',
w'
he
elemen s
o
K
whose
componen e
a e
all
x,
y,
w
n
espec i ely,
hen
y'x'
=
w'x'y'
and
w'
n
$
1
o
all
n
.
I
ollows
ha
K
is
in ini e-dimensional
o e i s
cen e
.
We
now
apply
he
specializa ion
lemma
om [41,
p
.141
.
Clea ly
he'cen e
o
K, C
say,
is
in ini e
and
k
<~X
:~
is
embedded
in
KC
,`X'~
so we
can
specialize
X
o
a-
lues
in
K
so
ha
P
emains
in e ible
.
I
ollows
ha
o al]
bu
ini ely
many
n
no
di isible
by
x
we ha e
a
specializa ion
om
X
o
K(n)
making
P
in e ible
.
In
each
o
hese
ields
K(n)
he
educed
no mmaps
each
ma ix
in
E(U)
o
I,
hence
a n =
1
o
all
bu
ini ely
many
n
no
di isible
by
X
.
This s ill
lea es
in ini ely
many
alues
o
n
and
so
is
impossible
unless
a
=1
.
Fo he
ee
algeb a
k
<X>
=
R,
e e y
in e ible
ma
ix
is
a
p oduc
o
elemen a y
and
diagonal
ma ices,
i
.e
.
GL(R)
=
E(R)
.k
(by
P op
.2 .7
.2 o
[3],p
.95),
hence
GL(R)nE(U)
=
E(R)
.k'nE(U)
=
E(R)
(k
nE(U))
=
E(R),
by
he
le
mma
.
The e o e
GL(R)/GL(R)nE(U)
=
E(R)
.k
/E(R)
-
k
/k
nE(R)
k
,
and
so
we
ind
Theo em
5 .2
.
Le
.
R =
k
<X>
be
xhe
{,Lee
k-alpebna
on
a
se
X
and
U
=
k
K
X
",~
í b
6íe
.Cd
u
Ó
l nac ía
nb
and
D
(R)
í ó
dí í-
,son
gnoup,
.
. Nen
U
ab
-
D(R)
xk
*
This
sol es
Exe cise
7
.6
.10
o
[3]
.
In
many
cases
i is
ue ha
E(U)nGL(R)
=
E(R),
as in
he
caseo
k
<X>
,
bu
by no
means
always
.
_Fo
a
s udy
o
he
gene alcase
we
e e
o
Ré ész
[121
A
he
o he
ex eme,
le
K
be
a
skew
ield
in
which
e e y
non-ze o
elemen
is a
commu a o
(c
.[6]),
le
C
be
i s
cen e
and
conside
he ee
K- ing
K
C
<X>
and
i s un¡
e sal
ield
o
ac ions
U =
K
C
-i X
:~
.
The ing
R
has
a
weak
algo i hm
(c
.
[3],
p
.78),
hence
GL(R)
=
GL1(R)
.E(R),
and
so
GL(R)/GL(R)nE(U)
=
GL(R)
.E(U)/E(U)
=
GL1(R)
.E(U)/E(U)
.
Now
G1
1
(R)
=K
*
nE(U),
hence
he
second ac o
on hé
igh
o
(4)
in
§4
is
i ial
and
so
KC
<_
X~
ab
_
D,
whe e
D is
ee
abelian
o
coun able
ank
(o
o
ank
IXI
i
his
is
la ge )
.
6
.
Localiza ion
Le
R
be
a
semi i
and E any se o
squa ema i-
ces
o e
R
;
i
is
na u al
o .ask
unde
wha
condi ions
he
lo
caliza ion
R
is
again
a
semi i
.
This
has
been
answe ed
in
E
171,
whe e
i
is
shown
ha
R
E
is
a
semi i
i
and
only
i
E
i
s
6acxon
.
comple e,
i .
e
.
whene e
ABEYE,
hen
he e
exis s
a
ma ix
C
o e
R
E
such ha
(B,C)
is
in e ible
o e
R
E
.
We
shall
show
ha
0
when
R
is
ully
a omic,
hen
so is
R
E
and
ou
aim
will
be o
s udy
he
ela ion
be ween
he
di iso
g oups
o
R
and
R
E
in
ha case
.
An
a om
A
in
R
and
also
he
associa ed
simple
Z- alue
is
called
E-ínnele an
-
i
A
becomes
a
uni
in R~
.
and
E-nele an
o he wise
.
Theo em
6
.1
.
Le
R
be
a
jullu
a omíc
semí6íx
and
le
.
E
be
a
6acxon
comple e
he
og
ma íce~s
o en
R,
hen
R
E
íó
aga,Ln
a
~ull(!
a omíc
zemílín,
and
e eny
a om o en
R
eí hen
becomes
a
uní
on
nema
.Ln~s
an a om
o en
R
E
.
P oo
.
We
begin
by
p o ing
he las pa
.
Le
A
be an
a om
o e
R
and
suppose
ha
o e
R
E
we
na e
A
=
B
1
B 2
,
whe e
he
B
i
a e
non-uni s
.
Then by
C ame 's
ule,
U
i
(B
i
+
I)V
i
=
C
i
(i
=
1,2),
whe e
C
i
is
a
ma ix
o e
R
and
U
i ,
V
i
E
Hence
Le
be
he
simple
Z- alue
de ined
by
A,
ake
comple e
ac o iza ions
o
C
i ,
C
2
o e
R
and
le
w
1
,
w
2
be
he
Z- alues
co esponding
o
C
1
,
C
2
bu
coun ingonly
E- ele-
an
a oms
.
Them
by
(1),
=w
1
+ w
2
.
Bu
w
i
(C
i )
:
l
and
so
2
s
w
1
(C 1
)
+
w
2
(C
2
)
=
(A)
=
1
a
con adic ion,
an'd
his
shows
ha
A
is
an
a om
o
a
uni
o e
R
E
.
Now le
P
be
any
ull
ma ix
o e
R
E and
w i e
(2)
U(P
+
I)V
= A,
whe e
AEF(R),
U,
VEGL(R
E
) .
We can
w i e
A
as
a
p oduc
o
a oms
say,
o e
R
;
each
will be
ei he
an
a om
o
a
uni
o e
R
E
,
hence
P
can
be
w i en
as
a
p oduc
o
a
mos
a oms
o e
R
E and his
shows
R
E
o
be
ully
a omic
.
The
ac ha
R
E
is
ully
a omic
may also
be
p o ed
as
ollows
:
Deno e
by
w
he sum
o
all
E- ele an
simple
Z- alues
on
R,
henw
is a
Z- alue
on
RE
and w(A)
= 0
o
AEF(R
E
)
only
i
A
is
in e ible
(by
C ame 's
ule),
hence
he
c i e ion
o
P op
.4 .3 is
sa is ied
.
By
P op
.6
.1
we can
de ine
he
di iso
g oups
o
bo h
R
and
R
E
;
o
desc ibe
he
mapping
be ween
hemwe need
P oposi ion
6 .2
.
Le
R
be a
jul' q
a omíe
eemíbí
and E
a
jae on
comple e
he
oj
ma íceh
o en
R,
ao
ha
R
E
agaín
jully
a omíe
.
Then
(i)
any
wo
a omz
o en
R
ha
a eno
s ablyaasocía ed
o en
R
a e
no
e ably
azsocía ed
0k)e
R
£
,
unle,sn
bo
. h
.
beeome
un
. ó,
(i
i )
e e y
ma íx
P
Theo em
6
.6
.
Lex
R
=k<
X
>
be
he
linee
al'gebna
on an
in6 i-
ní e
b
e
X,
andle
Xo
be
a
sub,6
ex
o6
X
wí h
an
ínjíníze
complemen
.
Deno e
b
y
E
=E
(X o
)
hehez
o
6
all
6ull
ma níceb
o al?l
y
copníme
o
Xo
,
hen
R
E
íb
a
símple
pnínc
.Lpal
ideal
domaín
.
In
pa icula ,
aking
.
X
o
o
consis
o
a
single
elemen ,
we
ob ain
a
simple
PID
wi h
a
single
a om,
bu no
a
local ing
.
Re e ences
1
.
H
.Bass,
Algeb aic
K- heo y,
Benjamin
(New
Yo k
1968)
2
.
G.M
.Be gman
and
W
.Dicks,
Uni e sal
de i a ions
and
uni e -
sal
ing
cons uc ions,
Paci
.
J
.
Ma h
.79(1978)
293-337
3
.
P.M
.Cohn,
F ee
ings
and
hei
ela ions,
.LMS
monog aphs
No
.2
Academic
P ess
(London,
New
Yo k
1971)
4
.
P.M
.Cohn,
Skew
ield
cons uc ions,
LMS
Lec u eNo es
No
.27
Camb idge
Uni e si y
P ess
(Camb idge
1977)
5
.
P.M
.
Cohn,
The
uni e sal
ield
o
ac ions
o
a
semi i
I
.
Nume a o s
and
denomina o s,
P oc
.
London
Ma h
.
Soc
.
(3)44(1982)
1-32
.
6
.
P.M
.Cohn,
De e minan s
on ee
ields,
o
appea
7
.
P
.M
.Cohn
and
W
.Dicks,
Localiza ion
in
semi i s
II
.
J
.
Lon-
don
Ma h
.
Soc
.
(2)
13
(1976)
411-418
8
.
P
.M
.Cohn
and
A
.H
.Scho ield,
On
he law
o
nulli y
Ma h
.
P oc
.
Camb idge
Phil .Soc
.
9
.
W
.Dicks
and
E
.D
.Son ag,
Syl es e
domains,
J
.
Pú e
appl
.
Algeb a
13
(1978)
243-275
10
.
J
.
Dieudonné,
les
dé e minan s
su
un
co ps
non-commu a i
Bull
.
Soc
.
Ma h
.
F ance
71
(1943)
27-45
11
.
M
.Mahda i-Heza ehi,
Ma ix
alua ions
on
ings
and
hei
associa ed
Skew
ields,
Resul a e
d
.
Ma h
.
12
.
G
.
Ré ész,
On
he
abelianized
g oup
o
uni e sal
ields
o
ac ions,
o
appea
13
.
A .R
.Richa dson,Simul aneous
linea
equa ions
o e
a
di i-
sion
algeb a,
P oc
.
London
Ma h
.
Soc
.(2)18 (1928)
395-420