J
OURNAL DE
T
HÉORIE DES
N
OMBRES DE
B
ORDEAUX
K
AE
I
NOUE
The metric simultaneous diophantine approximations
over formal power series
Journal de Théorie des Nombres de Bordeaux, tome
15, n
o1 (2003),
p. 151-161
<http://www.numdam.org/item?id=JTNB_2003__15_1_151_0>
© Université Bordeaux 1, 2003, tous droits réservés.
L’accès aux archives de la revue « Journal de Théorie des Nombres de Bordeaux » (http://jtnb.cedram.org/) implique l’accord avec les condi-tions générales d’utilisation (http://www.numdam.org/conditions). Toute uti-lisation commerciale ou impression systématique est constitutive d’une infraction pénale. Toute copie ou impression de ce fichier doit conte-nir la présente mention de copyright.
Article numérisé dans le cadre du programme Numérisation de documents anciens mathématiques
The metric simultaneous
Diophantine
approximations
overformal
power
series
par KAE INOUE
RÉSUMÉ. Nous étudions les
propriétés métriques
del’approxima-tion
diophantienne
simultanée dans le cas non archimédien. Nousprouvons d’abord une loi du 0 - 1 de type
Gallagher,
que nousutilisons ensuite pour obtenir un résultat de type Duffin- Schaeffer.
ABSTRACT. We discuss the metric
theory
of simultaneousDio-phantine approximations
in the non-archimedean case.First,
weshow a
Gallagher
type 0-1 law. Thenby
using
thistheorem,
weprove a Duffin-Schaeffer type theorem.
Introduction
In the
study
of metricDiophantine approximations,
one of thequestions
asks for conditions on a
non-negative function 1/J
whichguarantee
that theinequality
I I a" ’"
has
infinitely
manysolutions R
for almost every x. In the one-dimensional qcase it is known that if the
following
two conditions hold then(1)
hasinfinitely
many solutions for almost every x, Duffin-Schaeffer[1]:
where c is a
positive
constantand 0
is the Euler function.A similar result holds for s-dimensional
Diophantine approximations.
In theprevious
paper[4],
we discussed ananalogue
of Duffin-Schaeffer theorem for formal Laurent power series(one-dimensional case).
As its continuation we consider the samequestion
forhigher
dimensional case in this paper. To state our results we start with some fundamental definitions and notations.We use the
following
notations:F : a finite field with q
elements,
the set of
polynomials
withF-coefficients,
F(X) :
the fraction field ofIF((X-1)) :
the set of formal Laurent power series.We define the absolute value of
f
ElF((X-1 ))
by
If
-qdegf
andput
which is a
compact
abelian group with the metricd( f, g )
=f - 91.
Wedenote by
m the normalized Haar measure on L. Thisimplies
that for anyIn
[4],
we showed thathas either
infinitely
many solutions for m-a.e.f
E L oronly finitely
many solutions for m-a.e.f
E LHere, 0
is aIn
>0}
U101-
valued function and P andQ
have no non-trivial common factor. We call this"Gallagher
type
theorem" since thisproperty
wasproved by
Gallagher
in the classical case. Then, we showed "Duffin-Schaeffertype
theorem" :Let o
be a0}
U(0)-valued
function,
which satisfiesSuppose
for apositive
constantC,
there areinfinitely
manypositive integers
n such that
holds,
thenThe purpose of this paper is to show the s-dimensional version of these theorems for formal Laurent power series
(s > 2).
For agiven V)
asabove,
we consider thefollowing inequalities:
We shall
give
a sufficient conditionon 9
so that(3)
hasinfinitely
manysolutions
( pl , ... , p )
for almost every( f 1, ... ,
fs)
E ILS withrespect
tom~,
which is the s-foldproduct
of m. For this purpose, letfor a monic
polynomial Q
andwe first prove the
following
theorem in§1.
Theorem 1. For
either ms(E)
= 0 or 1holds, equivalently,
(3)
hasinfinitely
many solutionsof
(Q,
Pl, ... ,
P,)
for
m~-a.e.( fl, ... , f,)
E L~ or hasonly finitely
many solutionsfor
(fl,... ,
f,)
E L~.We call this the s-dimensional version of the
Gallagher
type
theorem.Secondly, making
use of thistheorem,
we can show the s-dimensionalver-sion of the D uffin- Schaeffer
type
theorem in§2.
Theorem 2. be a
01
U whichsatisfies
Suppose for
apositive
constantC,
there areinfinitely
manypositive integers
n such thatholds. Then
(3)
hasinfinitely
solutions~ , ... , p-
,for
( f p ... , f s ) E IL.~s .
Finally,
weexplain
a Duffin-Schaeffertype
conjecture
in formal Laurent1. Proof of Theorem 1
In the
sequel,
westudy
the metricproperty
of s-dimensional simultaneousapproximations
(3)
for s > 2. Forf
= EF((X-1))
(al
#
0),
weput
For
given hi
E such thatwe define the
cylinder
set ... ,his)
as follows:Then we see the
following,
...
Lemma 1. Let
I (hl k
h2 k ,
... ,
hs k) k
>1}
be a sequenceof cylinder
setsdefined
as above with11
be a sequenceof
measurable setsfor which Ek
C
( hi k ,
... ,hs k ) .
Suppose
there exists 6 > 0 such that >hsk))
for
any k>
1. ThenProof.
LetWe show that ml
(Ht)
= 0 forany l >
1,
whichimplies
the assertion of this lemma.Suppose
there exists aI~o
EZ+
such that > 0ko.
There is a naturalcorrespondence
betweencylinder
sets defined for L as in(2)
andq-adic
rationalintervals,
and so(L, m)
isisomorphic
to[0, 1]
with theLebesgue
measure.Similarly,
mS)
isisomorphic
to with theLebesgue
measure. Soby using cylinder
sets(hl, ... , hs)
C L~ instead ofh
x ... xIs
C[0,
we canapply
Lebesgue density
theorem. Then wefor some l~. On the other
hand,
From the
assumption
of thislemma,
That is
which contradicts
(4).
0Lemma 2. For any
polynomial hi
E0)
and gi E 1 i s, the map Tof
ILS ontoitself
defined
by
Proof.
It is easy to see that each mapis a Bernoulli transformation of L. In other
words,
if weput
1 ~
gives
a sequence ofindependent
andidentically
dis-tributed random variables. In
particular,
Ti
is weakmixing.
Since the s-foldproduct
of weakmixing
transformations isergodic
(see
[5]
Prop.
4.2.),
thisyields
the assertion of the lemma. 0Proof of
Theorem 1. In the case of formal powerseries,
we meanby
(Pi, Q)
= 1that Pi and Q
have no non-trivial common factor.If
then we can find a sequence of
polynomials Qi , Q2, Q3,
... and apositive
integer
l, such thatO(Qk)
>q-l
for any k > 1. In this case, for anyfi E
IL and asufficiently large I~,
we can findPi
(deg
Pi
deg
Qk),
such thatand Pi
andQk
have no non-trivial common factor.Indeed,
the number ofpolynomials
P
such thatis all such
polynomials P
are notrelatively prime
to~?~,
thenQk
has more thanqdeg Q k -l
factors,
which isimpossible
ifdeg Q k
issufficiently large.
Thisimplies
E
Now we show the assertion of the theorem when
For fixed
Q,1’1, ... ,
PS-1
andPs,
there existpolynomials
hl, ... , hs
suchthat Then
(6)
implies
deg h2
deg
tends
to oo. Thus we canapply
Lemma1 when
(6)
holds. Then we evaluate the measure ofUdeg Q>n
E’Q’
LetR be an irreducible
polynomial
and considerfor
We
put
and
Then we see
and
for j
=0,
1. From Lemma1,
we find thatfor
( fl, ... , fs)
E L .Suppose
(7),
we haveand see
Also,
we haveHere,
These
imply
for j
=0,1.
Hence from Lemma2,
we havefor j
=0,1.
Thus,
if eitherR))
orms(Ei(1 :
R))
> 0 for someirreducible
polynomial R,
then we havem (E)
= 1.Now we assume that
R))
=R))
= 0 forany irreducible
polynomial
R. Weput
where
(8)
refers to:Suppose (8),
we havefor any
polynomial
U with 0deg
Udeg
R.Here,
we seewhich
implies
that ifthen its measure is
qs deg R
Since
F(R)
is(.
+~)-invariant,
(0 deg U
degR)
and{(11, ... ,
1I.fI
deg f
i- deg R~
have the same measure. Hence we havewhich
implies
Suppose
ms(E)
>0,
since E =F(R)
UEo(I, R)
UEl(1,R),
we see thatmS(F(R))
> 0 for any irreducible R.By
thedensity theorem,
we havems (E)
=m8 (F (R))
= 1 where R is so chosen thatdeg R
issufficiently
large. Otherwise,
ms(E)
= 0. 02. Proof of Theorem 2
In the
sequel,
wealways
assume thatQ , Ql ,
Q’
andQ’
are monic. RecallFrom
(6),
if issufficiently small,
then we haveNow consider the measure of
EQ
flEQ, (deg Q’ deg Q).We
letN(Q, Q’)
be the number ofpairs
ofpolynomials
P and P’ which satisfiesfor
given Q
andQ’e
Then we show that the measure is bounded asfollows,
holds for some
polynomial
R and D =(Q,
Q’).
If D dividesR,
we may writeand have
If P* and P’* also
satisfy
(12),
thenFrom
(12),(13),
weget
From
(14),
we seewhich
implies
thatmust hold. Thus the
possible
number ofpolynomials
Psatisfying
(11)
fora
given
R is no more thanq deg D
. Since(10)
implies
and R is divisible
by D,
we find thatThen
Because we assume
for
sufficiently large deg Q.
From(4)
and(9),
we havefor
infinitely
many n. Then >(2 sC2)-l’
by
(15)
and Lemma 5 of[8] (pp.
17-18).
Finally using
to Theorem1,
wecomplete
theproof
of thistheorem. 0
Example.
PutThen,
from Theorem 2.2 of[7],
we haveand it is easy to see that
holds. Thus we see that there are
infinitely
many solutions(-~-,... ? ~f)
with irreducible of
for a.e.
( f 1, ... ,
eRemark. It is natural to ask whether we can
get
a necessary and suf-ficient condition instead of(4)
in Theorem 2. In this sense, wegive
the s-dimensional Duffin-Schaeffertype
conjecture
in thefollowing.
Conjecture.
(3)
hasinfinztely
many solutions( ~ , ...
for
mS -a. e.(11,... , fs) E
Uif
andanly if
diverges.
In the classical case, the s-dimensional Duffin-Schaeffer
conjecture
wasproved Pollington
andVaughan
[6]
for s > 2. We may also prove thisconjecture
for the s-dimensional formal powerseries,
s >2,
if we estimate the lower bound of We will discuss it in another occasion.References
[1] R. J. DUFFIN, A.C. SCHAEFFER, Khintchine’s problem in metric diophantine approximation. Duke Math. J. 8 (1941), 243-255.
[2] G. W. EFFINGER, D. R. HAYES, Additive Number Theory of Polynomials Over a Finite Field. Oxford University Press, New York, 1991.
[3] P. X. GALLAGHER, Approximation by reduced fractions. J. Math. Soc. Japan 13 (1961), 342-345.
[4] K. INOUE, H. NAKADA, On metric Diophantine approximation in positive characteristic,
preprint.
[5] M. G. NADKARNI, Basic Ergodic Theory. Birkäuser Verlag, Basel-Boston-Berlin, 1991.
[6] A. D. POLLINGTON, R. C. VAUGHAN, The k-dimensional Duffin and Shaeffer conjecture. Sém. Théor. Nombres Bordeaux 1 (1989), 81-87.
[7] M. ROSEN, Number Theory in Function Fields. Springer-Verlag, New York-Berlin-Heidelberg, 2001.
[8] V. G. SPRIND017BUK, Metric Theory of Diophantine Approximations. John Wiley & Sons, New
York-Toronto-London-Sydney, 1979. Kae INOUE Department of Mathematics Keio University 3-14-1 Hiyoshi, Kohoku-ku Yokohama, 223-0061 Japan kae@math.keio.ac.jp