rings
Citation for published version (APA):
Praagman, C. (1985). Roots and logarithms of automorphisms of complete local rings. (Memorandum COSOR;
Vol. 8511). Technische Hogeschool Eindhoven.
Document status and date:
Published: 01/01/1985
Document Version:
Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)
Please check the document version of this publication:
• A submitted manuscript is the version of the article upon submission and before peer-review. There can be
important differences between the submitted version and the official published version of record. People
interested in the research are advised to contact the author for the final version of the publication, or visit the
DOI to the publisher's website.
• The final author version and the galley proof are versions of the publication after peer review.
• The final published version features the final layout of the paper including the volume, issue and page
numbers.
Link to publication
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal.
If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please follow below link for the End User Agreement:
www.tue.nl/taverne Take down policy
If you believe that this document breaches copyright please contact us at: openaccess@tue.nl
providing details and we will investigate your claim.
Memorandum COSOR 85 - 11
Roots and Logarithms of Automorphisms of complete local rings
by
C. Praagman
Eindhoven, The Netherlands June 1985
INTRODUCTION.
Let k ~,m orf'
then the following observation can readily be made :LEMMA 1.
There exist a
XE
Hom (k,k*)such that
xE x(k)
if and only if
for aU
h E :INthere exists a
z E ksuch that
zh=
x.Let G be a Lie group,
9
its Lie algebra, and exp :9
-+ G the exponential map. Then one may pose the question whether an equivalent of LEMMA 1 holds does exp9
equal the set of elements of G which have roots of arbitrary order?
In this paper the question is answered in the affirmative for groups of automorphisms of complete local rings with the complex numbers as coeffi-cient field. As an immediate consequence a number of results on iterations of automorphisms, partly well known (LEWIS [3J, REICH [8J, REICH-SCHWAIGER [9J, PRAAGMAN [4,5]), partly new, follows.
§
1PRELIMINARIES.
NOTATION.
R will denote a complete local (noetherian) ring with maximal ideal m and coefficient field¢,
i.e. R/m=
¢
~ R.R~ = R/m~+l
has finite dimension as f-linear space. EndowR~
with the ordi-nary topology. Since Rx + x if and only if x
n n
lim R~ this induces a topology on R
~+1
-+ x mod m for all ~
E
:IN •Let
A
be the group of ¢-automorphisms of R andV
the Lie-algebra of ¢-derivations of R intom.
A~ resp. V~ denote the corresponding entities for R~.C
and C~ denote the group of all t-linear continuous maps of R resp. R~ which map mk into itself for all k.c~ has a natural topology induced by the topology on R~. Since
C
this induces a topology onC,
in whichA
andV
are closed.JORDAN DECOMPOSITION.
Every element of A~ or V~ allows a decomposition into a semisimple and a unipotent resp_ nilpotent part (HUMPHREYS [2], THEOREM 15.3). This property transfers toA
andV.
LetL
= A
u
V.
Definition.
Let LE
L.
Thena) L is
topologically semisimple
if every closed subspace V of R, invariant under L, has an invariant complement.b) L is
topologically nilpotent
if for allk E
~ there exists a hE
~ such that LhR cmk.
c) L is called
topologically unipotent
if L - I is topologically nilpotent (here I denotes the identity inA).
LEMMA
2.
LE
L is topologicalZy semisimple (unipotent, niZpotent) if and only
if
L~,the map induced on
R~!is semisimple (unipotent, nilpotent) f01' aU
~ E ~ •LEMMA 3 a. L E
L is topologically semisimple if and only if there exist
x
1, ••. ,xm
generating
m
(as ideal in R)such that
LXi A . ~ ~ x. ·for A. E ~ ¢:, i=l, ••• ,m.LEMMA 3 b. L
E
A(V)
is topoZogicaZly unipotent (nilpotent) i f and only i f
Ll
is unipotent (nilpotent).
These lemmata, as well the next theorem are proven for regular R in PRAAGMAN [4],
§
2. The proofs stay valid for arbitrary complete local rings. Let [A,B] = AB-BA as usual then:THEOREM l.a.
Let
F EA,
then there exists a topologically semisimple
sF E A,
and a topologically unipotent
u F E Asuch that
F=
sFuFand
[sF,uF ] O.MOT'eoVer
sFand
u Fare unique and satisfy
(sF) £,THEOREM l.b.
Let
0E
V.
Then there exists a topologically semisimple
SoE
V
and a topologically nilpotent
no EV
such that
0
=
so+noand
[so,no] =O •
..
~V
Sod
no .and
sat~s~y (sO) t~o~eoer
an
are unt-que
~ J:. £,
n
(O~) •
00 k
EXPONENTIALS AND LOGARITHMS.
Define for 0 EV
exp 0 \' L. _D • Clearly k=O k!exp 0 E
C,
and in fact exp D E A, which is proved in PRAAGMAN [4J, THEOREM 4 for regular R, but the proof does not depend on the regularity.Denoting the set of topologically semisimple automorphisms (derivations) by SA (
Sv
) and the set of topologically unipotent automorphisms (nilpotent deri-vations) by uA(nV) one proves without any problem:LEMMA 4.a. exp
4.b. exp
nV
~ ~ uA,
and
~'s ~a t-Jec t.on.
b .. t'
4.c.
Let
D EV,
then
s(exp D)=
exp So, u (exp D)00
n exp O.
For F E uA, define log F (I_F)k k • Clearly log FEe, and in fact
log F E nV (PRAAGMAN [4J, THEOREM 4), and log is the inverse of exp on
nV.
Clearly [D,LJ=
0 implies [exp D,L] = 0, and i f F E uA :
[F,L]o
if and only if [log F,L]=
o.
SCHAUDER BASES.
Letxl
f ' " ,xm generate m, then
{xu
I
u
E IN~
} with the usual convention on multi-indices generate R as ¢-li~ear space.m {xu! E }
'l'here exists a subset S c :IN 0 such (1 S constitute a Schauder basis for R : if y E R then there exist unique y E t such that y =
I
y xU.u
uEs a
In particular xS A.x., then
l l
I
e(a,S) Aaxa implies that e(a,S)aEs
a
S
c(a,S) A x for all a and
S
and hence c(a,S)
~
0 implies AU AS. Similarly if D ESv
with D x. lthen c(o,S) ~ 0 implies
<
u,~>
COI"Y'1UTATION.
Let F EsA ,
and let x l ' ... ,xm be a set of eigenvectors of F generating
m,
F x. = A.x .• Let LEe be defined by Lx~
=L
l l l
uEs
Then [F,L] if and only if
~(o,B) ~
0 implies AU =AB.
a
R,( 0 , B) x •
Similarly, let D E
Sv
be defined by D x.=
~.x.,
then [D,L]=
0 if and l l lonly if tea,S)
#
0 implies<
a,~>
=<
S,~>.
Both statements may be proven analogously to the statement on the e(o,S).§
2THE TORSION-FREE DECOMPOSITION,
It will turn out that the decomposition in (topologically) semisimple and unipotent (nilpotent) parts does not provide a sufficient simplication of the problem. Therefore a further decomposition of the semisimple part will be introduced in this section. I shall start with a generalization of a theorem of LEWIS ([3J, section 5). The proof and its consequences will provide motivation for the introduction of characteristic subgroups and the
torsion-free decomposition.
CONCERNING A THEOREM OF LEWIS.
LE"I'IIS ([3J, section 5) proved that, in case of R is regular, every automorphism has a power which can be embedded in a complex analytic iteration group. I shall prove this, or rather an equiva-lent statement (see PRAAGMAN [5 ], THEOREH 2) for arbitrary R.THEOREM 2.
Let
PEA,then there exists an
h E :INsuch that
phhas a logarithm.
Fr·oof.
I shall construct an E sV
such that exp h D s= (
s P) , and hA.
x..A,
the multiplica-~ multiplica-~tive group generated by
{A
1, •••
,A
m} is a finitely generated subgroup ofi
r*
and hence
A
=
AO
~A
l, where
AO
is a finite cyclic group andA1
is free. Letvo
generateAO
andv
1/ " ' Vr
A
l, and 0 ; LZ m + LZ r be defined by "Ct notation). and let 6 0 ; 0(0.) o(x)
Vo
0v
LZ m + LZ / h LZ , h =I
AOI ,
(\lith a small abuse ofLet sD be defined by
<
<5 (a), p>
X Ct , aES, where = logv.
for some ~s
choice of the dermination of log. D is topologically semisimple and satis-fies ;
s ct 13
D(x .x ) SD (
I
c(y,ct+S)XY)yEs
L
c(y,a+S)
<
o(y), P>
xy yESI
c(y,ct+S)<
o(a) + 0(13), P>
x\(since c(y,ct+S) " 0 implies ).,YyES
a+S
<
o{a) +0(13),
p>
xHence sD E
SV.
Further exp{h D)X, S1 U
F
13
=
I
a Finally, i f x f(a,S)x ,aES
0(0.) -F 0(13) and hence if
<
o(a),Remarks.
hence o(a) +
0(13)
o
(y) )h h S
: 0 A, x,
,
sinceVo
1, so exp(h D)=(1 1
then f(a,S) = 0 i f Act " AS, hence if
\l> -F
<
6(13),
\l>.
So [ D, F] s u=
o.
AU+S and
h
1. Since [sD,sF] = 0 and in virtue of [sF,logUF ] = 0 i t is clear that [D,F] O. This will be important in the next subsections.
2. As an immediate corollary one finds that F has a logarithm if 11\01 1.
The following example shows that this condition is not necessary Let R
=
C[[x]J, Fx=
-x, then F == exp D, with Dx=
nix.*
A DECOt<lPOSITION.
Let F E A, then II(F) will be the subgroup oft
generated by the eigenvalues of .Let tA be the subset of SA for which A(F) is finite,f
Clnd
A
be tJw subset for whi.ch II (1") is free.THEOREM 3.
Let
F E A,then there
ex1,-S
. t t F EA
,
E
fA
and
u F E uA
such
that
[ F, FJ '" t U [ F, FJ t f =: [fF/upJ =: 0, F == tFfFuFand
if
L EC
then
Proof·
Let
s Dbe as
m
the proof of theorem 2. Then
f
F
exp
s
DE fA, and
(fF
}-l =tF EtA. The first part of the theorem now follows from remark
1"above. Let LEe then [L,F]
= 0 i fand only if [L,sF]
:::[L,uF ]
= 0as
follows immediately from the well known corresponding property for spaces
of finite dimension.
,sF]
=
0if and only if
~(a,S) ~
0implies
Aa
=
ASor
equivalenty aO(a)
=
00(S} and a(a} ::: a(S} which yields the second
of the
theorem.
Remapks.
3.
Note that [L,fF]
=
0if and only if
I SD]
=
O.
4.
The problem of finding a root or a logarithm of F has been reduced to
finding one of tF which commutes with fF and uF or equivalently with F.
t
5.
F and
are not uniquely determined by F : first of all \(F) is not,
and secondly the choice of
Vo
is significant. In the next subsection
Iwill therefore introduce characteristic subgroups which do not depend on
the chosen decomposition in
atorsion part
, a f.ree part
f Fan
da unl-
.u
potent part F.
CHARACTERISTIC SUBGROUPS.
Let F
E
A,
and let
60
and
6
be as defined in the
o
proof of THEOREM 2. Let
L
(F)
ker 6
m 00
c z:; •Then
L(F) does not depend on
the choice of
va; and neither on the choice of
x1, •••
,x
m
m
Let L E
C
then L(L) is the subgroup of
z:;generated by the set
{n-r~ E ilL. m c(a·;I3)
cf
0or
£«(X,r{) I O},where
o!:i u";ual LXr5I
aEs
LEMMA 5.
Let
FE
tA
and
LE
C.
Then
[F,L]L
(L) cLa
(F) •Proof.
[F,L]
°
if and only if
~(a,S)
#
°
implies A
a
=AS or equivalantly
a-SEker 00
EO(F);henee if and only if ElL)
C EO(F).(Note that eta,S)
~
°
also implies AU
=A
P
!).§
3ROOTS.
In this section I shall derive a criterion for the existence of an
n-th root
of an automorphism F, that is an automorphism GThis criterion will be formulated in terms of the characteristic subgroups of
F.
ThereforeI
start with an investigation into the structure of these groups.THE STRUCTURE OF THE CHARACTERISTIC SUBGROUPS,
Since 00 (Zl m) is a finite group there exists an eOE
Zlm and a subgroupV
C Zlm such that ZlmZl eO til V and
r:
0 (F) h Zl eO til V. Let p be any number. I f p divides h then EO(FP ) (hlp) ZleO
tilV,
and if p and h are relativelythen EO (FP ) 1:0 (F) •
And the other way around, if G
P
==F,
andP
divides h then necessarily (G) phZleO til V, while if p does not divide h one might have 1:°(G)
On the other hand if F and G are automorphisms, then
a FG x
L
9 (i3 ,
Ct) Fxi3
==GEs
So 1:(FG) == E(F) + 1:{G), and in g(S,a)L
yEs y fey,S) x . F.o
1: (F).If F
~
Gn, then E(F} == E(G) C EO(G), which is a sublattice of [,0(1") of finiteindex, the exact index on the prime factorizations of hand n. To avoid these complications in the formulation of the theorem define }; (F)
~
{a E Zl m i n a E I:(F)}. Then the condition E{F) c 1:0 (G) which isn
necessary for F == Gn, translates to
r:
(F) C EO (F) •LEMMA 6.
Let
Fand
GE A,
then
Z(F) C (G)implies that
L (F') C Z (G ).o
nn
fI, z::; eO $
v,
then £ (g.c.d(hrn» == h. Let a E E (F) satisfy (l E h k eO ffi V for some k E Z::;. Letf3
E En (F), thennp E hke
O $ V, hence hkE n z::; I so n
I
(g.c.d(h,n» k and henceB E
£ z::; eO $ V. So L (F) C LO (G)AUTOMORPHISMS HAVING ROOTS.
It turns out that the condition L (F) C IO(F) not nonly is necessary for the existence of an n-th root of F, but also sufficient
THEOREM
4.
Let
FC
A.
Fan
root
if
t
(F') c: Lo
(F).n
So assume L (F) C ZO{F). Choose a torsion-free decomposition F n
t f u
F. F. Ft 2nik/h as constructed in THEOREM 3. Let
Va
be a generator of Ao(F),Va
== e letn
P
m(p) ben = the decomposition of n in factors, and let pEp
h
n
plhpm(p) , and choose q E z::; such that q
n
plhm(p)
p 1 mod i . This
is possible since
£
and hn/£ areThen h 7.l prime. Define tG
E C
by 1, and6
0 (V) tFx .
affi
V
for some eO andV.
o.
Let
a
E Z(F) thena
E mh eO $ V. If g.c.d(mh,n} j 1 then mh/jE
h z::; hencem E j Z::;. Bu t g. c • d ( j sh , n ) j implies that g.c.d1h,n) divides j, hence hj
E
fI, z::; 1 and hence aE
£::z
eO ffi V.t t
This implies in the first place that G E
A,
and secondly that ,F] ~O.
In view of remark 4 this proves the theorem.§
4LOGARITHMS.
A criterion for the existence of a logarithm may be formulated by a
similar condition as for the existence of a root. As in the preceding section, I start with deriving a necessary condition, and prove afterwards that this is also sufficient.
t h
LOGARITH~1S
OF
ROOTS
OF UNITY.
Let FE
A,
and F = I ('"I
AO (F)I
=
h) 1. thenany D such that exp D
=
F much have its eigenvalues in (2ni/h)~. Let D be such a derivation and let TO : tzm +~
be defined by Dxa=
(2'1fi"C(a)/h)xa• If exp D=
F, then 60(0',)
=
TO(a} mod h, so "CO is onto and hence ker TO has rank m-1. Since [L,D]=
0 if and only if l:(L} c ker TO' one can easily prove that to have a logari thm~
(F) C V wherer.o
(F) h LZ eO Ell V for some V.Again i t turns out that this condition is also sufficient. Again some notation {a E
~
mI
3 n E IN such that na E l:(F)} is the smallest directsummand of
~
m containing l:(F).AUT~ORPHISMS
HAVING A LOGARITHM.
THEOREM 5.
Let
F EA.
Then F has a logarithm
if
and only
if
l: (F) C l:°{F}.00
Proof.
Let D be a logarithm of F, then [sD, == 0. Let the subgroup of¢
generated by the eigenvalues of sD "'1
n
2niQThen [ ,F]
m
o
I let 60 '" (2ni/h) ~, and "CO : ~ -+ 2Z as above.
o
implies L(F} C ker TO' and since ker TO is a direct summand of tzm, (F) cker TO.ker TO C ker 6
0 = E?F}.
On the other hand since 6
0 factors over TO' clearly So E (F) C l:°(F) which proves the only if part.
On the other hand, if 1:"" (F)
c: LO(F), then there exist eOEZZ; m and V
c:ZZ; m
m 0
such that ZZ;
==eO ZZ;
+ V , E(F)
==h eO ZZ;
(j) Vand E(F)
c:v.
2nik/h
tn
e
generate AO(F), and define
t a
O. Then (exp n)x
==0, and hence the theorem is proved.
a x (T
o
(a)2'lfik/h)
ax , where
ax , and 1:(F)
c:ker TO implies
Remark
6.Note that this criterion does not refer to a particular choice of
the logarithms of A(F), as does the 'smooth additional monomial' condition
(see
REICH[6J).
§
5ROOTS, LOGAR ITHIVlS AND ITERATIONS.
In this section I establish the connection with iterations of auto-morphisms. But first a simple consequence of the preceding sections will be proved.
ROOTS AND LOGARITHMS.
Since ~oo(F) = UnEJN
readily follows from THEOREM 4 and 5.
~ (F) the following theorem now n
THEOREM 6.
Let
FE
A,
then
Fhas a logarithm if and only if
FhaD an n-th
root
for aU n
E IN •ITERATIONS.
A complex (real, rational, fractional) iteration of FE A
is a 1group homomorphism from
¢
(JR,f2, -
Z;:;) toA
such that 1 -+ F. nIf F
=
exp D, t -+ exp t D is an iteration of F, which is even analytic in t. Analytic iterations have been studied by LEWIS [3J, REICH and SCHWAIGER [9J, by BUCHER [1J in connections with continuous iterations, by REICH andKRAUTER [7,8J in relation to fractional iterations. A good survey is REICH [6J. In PRAAGMAN [4,5J I studied the connection between iterations and logarithms.
As a direct consequence of THEOREM 6 a large number of statements on iterations follows. For if F has a logarithm D, then exp t D is an iteration, complex and analytic, of F. On the other hand if F
t is a rational iteration of F then F
1/n is an n-th root of F. Hence the question posed by REICH and SCHWAIGER ([9J. page 617) is answered, as well as the question whether the existence of a rational iteration implies the existence of an analytic one
(REICH [6J, §6, PRAAGMAN [5J).