The p-adic Measure on the Orbit of an Element of C
p.
V. ALEXANDRU(*) - N. POPESCU(**) - M. VAÃJAÃITU(**) - A. ZAHARESCU(***)
ABSTRACT- Given a prime numberpand the Galois orbitO(x) of an elementxofCp, the topological completion of the algebraic closure of the field ofp-adic numbers, we study functionals on the algebraC(O(x);Cp) with values in a subfield ofCp.
Introduction.
Let pbe a prime number,Qpthe field of p-adic numbers,Qp a fixed algebraic closure ofQp, andCpthe completion ofQpwith respect to thep- adic valuation. LetO(x) denote the orbit of an elementx2Cp, with respect to the Galois groupGGalcont(Cp=Qp). We are interested in the behavior of rigid analytic functions defined on E(x)(Cp[ f1g)nO(x), the com- plement ofO(x). In the present paper we provide several results concerned with functionals defined on the algebraC(O(x);Cp) with values in a suitable subfield ofCp. This investigation is needed in the more general attack on the problem of explicit description of rigid analytic functions onE(x), since, as we shall see below, there is a close relationship between these func- tionals and certain classes of rigid analytic functions. The paper consists of seven sections. The first one contains notations and some basic results. The
(*) Indirizzo dell'A.: Department of Mathematics, University of Bucharest, Romania.
E-mail: vralex@k.ro
Partially supported by Grant no. 27.694/2005, cod CNCSIS 849.
(**) Indirizzo degli A.: Institute of Mathematics of the Romanian Academy, P.O. Box 1-764, RO-014700, Bucharest, Romania.
E-mail: Nicolae.Popescu@imar.ro Marian.Vajaitu@imar.ro
Partially supported by CEEX Program of the Romanian Ministry of Education and Research, Contract No. CEx 05-D11-11/2005.
(***) Indirizzo dell'A.: Department of Mathematics, University of Illinois at Urbana-Champaign, Altgeld Hall, 1409 W. Green Street, Urbana, IL, 61801, USA.
E-mail: zaharesc@math.uiuc.edu
2000 Mathematics Subject Classification: 11S99.
second section is concerned with linear functions and functionals on var- ious fields. Section 4 studies p-adic measures and p-adic equivariant measures on O(x). Theorem 1 shows that the study of all functionals on C(O(x);Cp) can be reduced to the study of all functionals onCG(O(x);Cp).
Then, using results from the previous sections, the functionals on C(O(x);Cp) are closely related to the trace. Theorems 2 and 3 are useful complements to Theorem 1. In order to further clarify the relation between functionals and rigid analytic functions, in Section 5 we investigate the Cauchy transform of a function with respect to a measure. In the last section we present an analogue at an important theorem of Barsky [B], which relates the measures onO(x) with a suitable class of rigid analytic functions on the complement ofO(x). We remark that some of the results of this paper can be extended to a wider class of compact subsets ofCp.
1. Notations and basic results.
1. Letpbe a prime number,Qpthe field ofp-adic numbers,Qpa fixed algebraic closure of Qp, andCp the completion of Qp (see [Ar], [APZ1], [APZ2]). Denote by G the Galois group Gal(Qp=Qp) endowed with the Krull topology. One knows that G is canonically isomorphic to GGalcont(Cp=Qp), the group of all continuous automorphisms ofCp. We shall identify these two groups.
For any closed subgroupHofGdenoteFix(H) fx2Cp:s(x)xfor all s2Hg. Then Fix(H) is a closed subfield of Cp. If x2Cp, denote H(x) fs2G:s(x)xg. ThenH(x) is a subgroup ofG, andFix(H(x))
Qgp[x] is the closure of the polynomial ringQp[x] inCp. We say thatxis a topological generic elementofQgp[x]. Moreover, by [APZ1] one knows that any closed subfieldK ofCp has a topological generic element, i.e. there existsx2Ksuch thatK Qgp[x].
2. Let x2Cp. Denote O(x) fs(x):s2Gg the orbit of x. The map s?s(x) from Gto O(x) is continuous, and it defines a homeomorphism fromG=H(x) (endowed with the quotient topology) toO(x) (endowed with the induced topology fromCp) (see [APZ1]). In such a way,O(x) is a closed compact and totally disconnected subspace of Cp, and the group G acts continuously onO(x): ifs2G,t(x)2O(x) thens?t(x)(st)(x). One has the following result:
PROPOSITION1. 1)The subfieldQgp[x]is canonically isomorphic to the set of all equivariant continuous functionsf :O(x)!Cp, i.e. the contin-
uous functions which verify the condition:f(s? y)s(f(y))for alls2G andy2O(x).
2)There exists a familyfMn(x)gn0of polynomials inQp[x]such that i) degMn(x)n for all n0,
ii) 1
p<jMn(x)j 1,
iii) Any element f 2Qgp[x] can be written uniquely in the form:
f P
n0anMn(x)where fangn is a sequence of elements in Qp such that limn an0. Moreover one has:jfj sup
n0janMn(x)j.
3)If KxQgp[x]\Qp, thenKfxQgp[x]and Gal(Qp=Kx)is canonically isomorphic to H(x).
2. Linear functions and functionals.
1. Let K=k be a (not necessarily finite) algebraic extension, and fKngn0,K0k, KnKn1 for anyn, a family of subfields of K, finite overksuch that[nKnK. Denote byL(K=k) the set of allk-linear maps ofKintok. ThenL(K=k) is in a canonical way aK-vector space. Namely if W2 L(K=k) andx2KthenxWis the linear map defined by (xW)(a)W(xa) for alla2K.
Now assume thatkis of zero characteristic and for alla2Kdenote by TrK=k(a) or simply by Tr(a) the element of k defined as follows: if kK0K is a finite intermediate extension and a2K0, then Tr(a)
1
[K0:k]trK0=k(a), wheretrK0=k(a) denotes the usual relative trace ofaover k. It is clear thatTr(a) is independent of the choice ofK0.
If K=k is a finite extension, then for any W2 L(K=k) there exists a uniquea2Ksuch thatWaTr. Now assume [K:k] 1and letWnbe the restriction of W to Kn. Then Wn2 L(Kn=k) and so there exists a unique elementan 2Knsuch that for alla2Kn one has:
Wn(a)Tr(ana):
1
Thus for any a2Kn, one has: Wn1(a)Wn(a), and so by (1) one has:
Tr(an1a)Tr(ana). This means that 1
[Kn1 :Kn]trKn1=Kn(an1)an; n0:
2
Conversely, letA fangnbe a sequence of elements ofKsuch thatan 2Kn
for alln0, and that condition (2) is accomplished. Then for any n0 denote byWAn :Kn!kthe mapWAn(a)Tr(ana),a2Kn. By (2) there re- sults that one can define a linear mapWA:K!ksuch that the restriction of WAtoKnis justWAn. DenotePlim
Kn; 1
[Kn1:Kn]trKn1=Kn
n0. Then by the above considerations there results that the map A?WAdefines a k- isomorphism between thek-vector spacePandL(K=k).
2. Now assume that K is endowed with an ultrametric absolute value j j:K!R. We consider the functionals W:K!k, i.e. allk-linear maps which are continuous.
A linear mapW:K !kis continuous if and only if there exists a positive real numberMsuch thatjW(a)j Mjaj, for anya2K. SinceWis defined uniquely by a sequence A fangn, an2K for all n0, which verify condition (2), thenWis continuous if and only if there exists a positive real numberMsuch that
jTr(any)j Mjyj;
for alln, and ally2Kn. Now one can writey x
an, wherex2Kn, and so the above condition can be written as:
jTr(x)j janj Mjxj;
for alln0 and anyx2Kn.
LetLKn, be fixed. We want to relate sup
x2Kn
tr(x)
[L:k]jxjand DL
[L:k], where DLis the different ofLwith respect tok(see [Ar]).
REMARK 1. Any x2Kn can be represented as xplxa, where 1 jaj>1
pandlxis integer. Then one has: max
x2Kn
jtr(x)j
jxj max
x2Kn;p1<jxj1
jtr(x)j jxj , and the quotient between this number and MKn maxn
jtr(x)j:x2 2Kn;1
p<jxj 1o
is a real number whose module belongs to the real intervalh
1;1 p
.
If a2Kn then jaj jDKnj 1 if and only if jtr(aOKn)j 1. (Here jtr(aOKn)j maxfjtr(ax)j:x2OKn; the ring of integers of Kng.) Let us denote by mKn the integer part of log1=pjDKnj. If ap mKn, then jtr(p mKnOKn)j 1, or equivalently jtr(OKn)j (1=p)mKn. This means that MKn(1=p)mKn.
Letap (mKn1). Then there existsy2OKn such thatjtrp (mKn1)yj>1 or equivalently log1=pjtryj<mKn1. Therefore one obtains
log1=p(MKn)<mKn1 or1 p
mKn1
<MKn:
Thus we have mKnlog1=p(MKn)<mKn1. In conclusion, one has the following result:
PROPOSITION2. The linear mappingW:K!k,WWA, is continuous if and only if there exists a positive real numberM such that:
jDKnjjanj
[Kn:k] M; for all n1:
3
REMARK 2. Letx2Cp such thatTr(x) is defined (see [APZ2]). One has the following result. LetKQgp[x]\Qp, andK [nKnbe the union of a filtered family of finite extensions ofQp. The following assertions are equivalent:
1) The sequencefjTr(Mn(x))jgnis bounded (see Proposition 1).
2) The sequencen jDKn=Qpj j[Kn :Qp]j
o
n is bounded.
3) The linear mapTr:K!Qpis continuous.
These results are related to some results from [APP].
3. Denote byAthe set of all elementsA2Pwhich verify condition (3) for a suitableM>0. It is easy to see thatAis ak-vector subspace ofP.
Denote by K0 thek-vector space of all k-functionals onK. By the above considerations one has:
PROPOSITION3. The mapping A?WA defines an isomorphism ofk- vector spaces betweenAandK0.
Now letA fangn be an element ofAsuch thatan60 for alln0.
Letc2K0,cWB,B fbngn2 A. Denoteunbn
an,n0. We assert that the sequencefungnis bounded, i.e. there exists a real numberM>0 such thatjunj Mfor alln0. Indeed, assume that there exists a subsequence fuqngsuch thatjuqnj ! 1. Thenj1=uqnj !0 and so lim
n c(1=uqn)0. But c(1=uqn)Tr(aqn), and thusWA(1)0, a contradiction.
If for anyn0, one denotesWnunWA, then one has:c(x)lim
n Wn(x) for any x2K. Indeed, for n large enough one has: Wn(x)(unWA)(x)
WA(unx)Tr(anunx)Tr(bnx)c(x). This shows that the sequence fWngn converges pointwise to c in K0 and so by the Banach-Steinhaus Theorem (see [R]) one hasclim
n Wn lim
n (unWA).
PROPOSITION4. Notations and hypotheses are as above. LetWWA. Denote bySAthe family of all sequencesu fungnsuch that:
1)un2Kn for all n1.
2)The sequence of real numbersjunjis bounded.
3)One has: 1
[Kn1:Kn]trKn1=Kn(un1)un, n0.
For any u fungn2SA, denote by uW the mapping defined by:
(uW)(x)lim
n W(unx). Then uW2K0and ifW60, then for anyc2K0there exists a unique u2SAsuch that uWc.
We leave the details to the reader.
3. Integration onC(X;K).
1. Let K be a field which is complete with respect to an ultrametric absolute valuej j, and letX be a compact ultrametric space. Denote by C(X;K) theK-algebra of all continuous functions fromXtoK(see [Sch]).
Also, denote byV(X) the collection of all the open compact subspaces ofX.
By aK-valued measureonXwe mean a functionm:V(X)!Ksuch that:
(i) If U;V 2V(X); U\V ;, then m(U[V)m(U)m(V) (ad- ditivity).
(ii) kmk supfjm(U)j:U2V(X)g<1(boundedness).
The K-valued measures on K form a normed vector space M(X;K) under the obvious operations and with the normk kdefined by (ii).
The following statement (see [Sch]) can be viewed as the ultrametric analog of Riesz representation theorem:
PROPOSITION5. For eachW2C(X; K)0(the space of all functionals on C(X; K)), the mappingU?W(jU)mW(U),U 2V(X), is a measuremWon X(herejU denotes the characteristic function ofU). The mappingW?mW is aK-linear isometry ofC(X; K)0ontoM(X; K).
4. Equivariant measures onO(x).
1. Let x2Cp. The subset of C(O(x);Cp) consisting of all equivariant elements (see Proposition 1) is denoted by CG(O(x);Cp). The mapping f?f(x) defines an isomorphism betweenCG(O(x);Cp) andQgp[x]. We shall identify theseQp-algebras via this isomorphism.
Letf 2C(O(x);Cp) ands2G. Denote bys? fthe function defined by:
(s?f)(y)f(s 1(y)) for all y2O(x). In this way the group G acts con- tinuously on the algebraC(O(x);Cp).
2. LetKbe a closed and normal subfield ofCp(i.e. for anys2G, one hass(K)K). By anequivariant K-functionalonC(O(x);Cp) we mean a linear and continuous map W:C(O(x);Cp)!K such that W(s?f)
s(W(f)) for alls2Gand allf 2C(O(x);Cp). It is easy to see that by the correspondence stated in Proposition 5, the equivariant K-functionals on C(O(x);Cp) are in one to one correspondence with the so calledequivariant measures onO(x), i.e. the elementsm2M(O(x);Cp), such that m(sU)
s(m(U)) for allU2V(O(x)) ands2G. In this paper we consider mainly the caseKQpand shall investigate the above correspondence between functionals and measures.
THEOREM 1. Any Qp-functional W:CG(O(x);Cp)!Qp can be un- iquely extended to a Cp-functional eW:C(O(x);Cp)!Cp and this func- tional provides us with a measure mW on O(x) with values in Qp. The measuremWis equivariant.
Next, we recall an important criterion for the existence of a measure with given properties.
PROPOSITION6 (The abstract Kummer congruences, [Ka]). LetXbe a compact ultrametric space, letOpbe the ring of integers ofCpand letffig be a system of continuous functions fromC(X; Op). If theCp-linear span of ffigis dense inC(X;Cp)andfligis an arbitrary system of elements ofOp, then the following assertions are equivalent:
a) There ism2 M(X;Op)with the propertyR
X fidmli.
b) For an arbitrary choice of elements gi2Cp almost all of which vanish,
X
i
gifi(x)2pnOp for all x2 X implies X
i
gili2pnOp:
Now, let fangn0 be a bounded sequence of Qp. Let us define W:CG(O(x);Cp)!Qp such thatW(Mn(x))an. As we know from Propo- sition 1 any f 2CG(O(x);Cp) can be written as f P
n0anMn(x), with an!0. We define W(f): P
n0anan. Using Theorem 1 we have a similar result of abstract Kummer congruences. More precisely we have:
PROPOSITION7. Letfangn0be a bounded sequence ofQp. There exists a unique functional W:C(O(x);Cp)!Cp such that W(Mn(x))an, for anyn0.
THEOREM2. Ifmis an equivariant measure on O(x)with values inQp then the mapping f? R
O(x)f(t)dm(t)is a functional on CG(O(x);Cp) with values inQp.
PROOF OFTHEOREM1. We use the notations from Proposition 1. For any s0 let us denote AsW(Ms(x)). Then for anyuP
s asMs(x), one has W(u)P
s asAs. Since the equality is true for any u2Qgp[x], there results that the sequencefAsgs ofp-adic numbers is bounded and so the set of all Qp-functionals on Qgp[x]CG(O(x);Cp) is in one to one corre- spondence with the set of bounded sequencesfAsgsofp-adic numbers.
Now let U be an open ball on O(x), x2U, and denote H(U) fs2G:sUUg. ThenH(U) is a subgroup ofG, and denote by KU the subfield of Qp fixed by H(U). Since H(x)H(U), then KU KxQgp[x]\Qp. It is clear that [G:H(U)]<1 and let
Sn fs1e;s2;. . .;sngbe a system of representatives for the right co-
sets ofGwith respect toH(U). Also choosea2Qpsuch thatKuQp(a). It is clear that the elements fs1(a);. . .;sn(a)g are distinct, and the balls fsi(U)g1in are pair-wise disjoint and coverO(x). Let us put
f0(U) P
si2Sn
si(1)jsi(U)jO(T) f1(U) P
si2Sn
si(a)jsi(U) . . . . fn(U)1 P
si2Sn
si(an 1)jsi(U): 8>
>>
>>
>>
>>
<
>>
>>
>>
>>
>: 4
We assert that the function fi(U) is equivariant for all i, 0i<n, i.e.
fi(U)(s(y))s(fi(U)(y)) for all y2O(x) and any s2G. For that it is enough to assume i1. Firstly, we remark that one has: fi(U)(y)
si(y) fory2si(U); 1in:Now we can write:fs;1(U)sf1(U)(s 1(y))
ssi(a) ifs 1(y)2siU(or equivalentlyx2ssi(U)).
It is clear that the following permutations
s1U s2U . . . snU
ss1U ss2U . . . ssnU
; s1(a) s2(a) . . . sn(a)
ss1(a) ss2(a) . . . ssn(a)
5
coincide, so the applications
s1U s2U . . . snU
s1(a) s2(a) . . . sn(a)
; ss1U ss2U . . . ssnU
ss1(a) ss2(a) . . . ssn(a)
also coincide. This shows thatf1(U)fs;1(U), i.e.f1is equivariant.
Furthermore (4) is a Cramer system whose determinant is different from zero. It follows that
Xn 1
i0
Qp(a)fi(U)Xn
i1
Qp(a)jsiU; 6
and this sum is direct. This shows that the functionalW can be extended uniquely to a functional
e
W:CG(O(x);Cp)QpQp!Qp:
According to ([Sch], page 273), the Cp-algebra C(O(x);Cp) has an orthonormal basis consisting of characteristic functions of balls. Then the Qp-subalgebraCG(O(T);Cp)QpQpis dense in theCp-algebraC(O(T);Cp).
But theneWcan be extended uniquely to aCp-functional:
e
W:C(O(T);Cp)!Cp:
Then by Riesz's Theorem (Proposition 5) there exists a unique measuremWon O(x) associated toeW. It is clear that for any ballUinO(x),mW(U)eW(jU), can be obtained from (4) by applying the functionaleW. We must show that the measuremWis equivariant. For that let as aboveU be an open ball ofO(x) which contains x. Denote A(aij) the nn matrix, where aijsi(aj), 1in; 0j<n 1. Then by (4) there results:
js1(U) jsn...(U) 0 B@
1 CAA 1
f0(U) ...
fn(U)1 0 BB
@ 1 CC A
and so by applyingeWone has:
mW(s1(U)) mW(sn...(U)) 0
B@
1 CAA 1
W(f0(U)) ...
W(fn(U)1) 0 BB
@
1 CC A:
Now fors2Gdenote byAsthe matrix obtained fromAby applyingsto all the entries ofA. Then (As) 1As1.
Sincefi(U)are equivariant functions, one has:fi(s(U))fi(U)for alls2G.
Then one obtainsmW(sU)smW(U), for alls2G, as claimed. p PROOF OFTHEOREM2. Letmbe an equivariant measure onO(x) with values inQp. Denote byW:C(O(x);Cp)!Cpthe functional associated tom.
We must show that for anyf 2 CG(O(x);Cp) one hasW(f)2Qp. Since any elementf 2 CG(O(x);Cp)Qgp[x] is a limit of a sequencefangnof elements ofKx(see [APZ1]) it is enough to show that fora2Kx, one hasW(a)2Qp. But this follows by the proof of Proposition 5 (see [Sch]) sincemis equivariant.
Some details are left to the reader (see also the proof of Theorem 3). p 3. By the above considerations there results that ifW:CG(O(x);Cp)!Qp
is a functional then the associated measuremW(Proposition 5) takes values in KxQgp[x]\Qp. At this point we describe the equivariant measures onO(x) with values inQp.
THEOREM 3. For the element x2Cp the following assertions are equivalent:
1) There exists a functional W:CG(O(x);Cp)!Qp such that for any open ball U of O(x)one has:mW(U)2Qp.
2)The function Tr:Kx!Qpdefined by Tr(a) 1
degatrQp(a)=Qp(a)is continuous and one has:W(a)Tr(a)for alla2Kx.
PROOF. Leta2Kx, andas1(a); ;sn(a) be all the conjugates ofa over Qp, ndeg(a). Since a2Qgp[x], denote a:O(x)!Cp the local constant function defined bya(s(x))s(a). One has:W(a)W(a). The ele- mentssi(x), 1inare all distinct and lete>0 be a real number such that all the balls B(si(x);e) are pairwise disjoint. Denote H(x;e)
fs2G:js(x) xj<eg. ThenH(x;e) is a subgroup ofGof finite index. If N[G:H(x;e)] andftig1iNis a set of right representatives for the co- sets ofGwith respect toH(x;e), then the ballsfB(ti(x);e)g1iNcoverO(x)
and are any two disjoint. One has B(ti(x);e)ti(B(x;e)), and so by hy- pothesis there results thatmW(B(ti(x);e))mW(B(x;e)). SinceW(1)1, one has:
mW(s(x))1X
i
mW(B(ti(x);e))NmW(B(x;e)):
According to our choice ofe, there results that on any ballB(ti(x);e) the functionais constant andNkdeg (a), wherekis an integer. Then
W(a)X
i
a(ti(x))mW(B(ti(x);e)):
Since for exactlykballsB(ti(x);e) the functionatakes the same valuessj(a), by the above considerations one further obtains
W(a)W(a)Xn
j1
ksj(a)mW(B(x;e))nkmW(B(x;e)1 n
Xn
j1
sj(a)
!
Tr(a):
The implication from 2) to 1) is left to the reader. p REMARK 3. Let W be as in Theorem 3. By the above considerations there results that for any e>0, one has: mW(B(x;e)) 1
N, where N[G:H(x;e)]. This shows that thep-adic measuremWcoincides with the p-adic Haar measurepxdefined in [APZ2].
Under the hypothesis of Theorem 3 there results that if QpK1 Kn Kxis a tower of finite extensions ofQp such that [nKn Kx, then by considerations from Section 3, one hasWWA, whenAn 1
[Kn :Qp] o
n1.
5. Cauchy Transforms onO(x).
1. Let x2Cp. For any real number e>0 denote B(x;e)
fy2Cp:jx yj<eg and B[x;e] fy2Cp:jx yj eg. Also denote E(x;e) fy2Cp:jy tj e; for allt2O(x)g. The complement ofE(x;e) in Cp[ f1g is denoted byV(x;e). Both setsE(x;e) andV(x;e) are open and closed, and one has: \eV(x;e)O(x). Denote E(x) [eE(x;e)
Cp[ f1g nO(x).
For anyx2Cpande>0 denoteH(x;e) fs2G:js(x) xj<egand H[x;e] fs2G:js(x) xj eg. Let Se (respectively Se) be a complete
system of representatives for the right cosets ofGwith respect toH(x;e) (respectivelyH[x;e]). ThenE(x;e) [s2SeB(s(x);e).
2. LetW:C(O(x);Cp)!Kbe a functional, whereKis a suitable closed subfield ofCp. Also denote bymthe measure associated toWaccording to Proposition 5. Then for anyf 2C(O(x);Cp) one has:
W(f) Z
O(x)
f(t)dm(t):
For anyz2E(x) and anyf 2C(O(x);Cp), the functionT(f;z):O(x)!Cp defined by
T(f;z)(t) f(t)
z t; t2O(x)
belongs toC(O(x);Cp). Hence for anyz2E(x) one can define the element FK(m;f;z)
Z
O(x)
f(t) z tdm(t);
called theCauchy transform of f with respect tom(orW).
Now letz02E(x). Then for anyz2E(x) andt2O(x) one can write:
f(t)
z0 t f(t)
z tf(t)(z z0)
(z t)2 f(t)(z z0)n (z t)n1 : 7
Generally the series (7) does not converge, but it converges for a suitable choice of z. Indeed, let jz0 tj e for all t2O(x), and let a be a real number such that 0<e
a<1. Then for any z2B(z0;e2=a), the series (7) converges since one has:z z0
z t ne a
n 1
e<1, for anyt2O(x). Hence for anyz2B(z0;e2=a) one can write:
Z
O(x)
f(t)
z0 tdm(t) Z
O(x)
f(t)
z tdm(t) Z
O(x)
f(t)(z z0)n
(z t)n1 dm(t) 8
and this series is also convergent (since W is continuous) for all z2B(z0;e2=a).
3. Letfengnbe a strictly decreasing sequence of positive real numbers with limit zero. One can assume that SnSn1 for all n0, where SnSen
For anyn0, choose an elementan2Kxsuch thatjx anj<en. Then by ([Sch], page 276) it follows that
FK(m;f;z)lim
n
X
s2Sn
m(B(s(x);en)) f(s(x)) z s(x): By this equality it also follows that
FK(m;f;z)lim
n
X
s2Sn
m(B(s(x);en)) f(s(an)) z s(an): 9
By the above considerations one obtains the following result.
THEOREM4. Let x2Cp, and letW:C(O(x);Cp)!K be a functional, where K is a closed subfield ofCp. Then for any f 2C(O(x);Cp)the function FK(m;f;z):E(x)!Cpdefined by(8)is a rigid analytic function on E(x), and lim
z!1FK(m;f;z)0.
4. If the field K and the functional Ware fixed we shall write simply F(f;z) instead ofFK(m;f;z).
6. Equivariant Cauchy Transforms.
1. Forx2Cpconsider an equivariant functionalW:CG(O(x);Cp)!Qp. By Theorem 1 the associate measuremmWis equivariant. On has the fol- lowing result:
THEOREM5. For any nonzero element f 2CG(O(x);Cp), the function F(f;z), the Cauchy transform of f with respect toW, is an equivariant rigid analytic function defined on E(x). Any element of O(x)is a singular point for F(f;z), if F(f;z)60.
PROOF. We observe that for any z2E(x) and any s2G, one has s(z)2E(x). Recall that F(f;z) is equivariant if for any s2G one has:
F(f;s(z))s(F(f;z)). Now this equality is true sinceWis equivariant (see (8)). Furthermore ifF(f;z)60, it has a singular point (a pole) which must belong to O(x). Then by equivariance all elements ofO(x) are poles for
F(f;z). p
COROLLARY 1. For an element x2Cp the following assertions are equivalent:
a) O(x)is a finite set.
b) x2Qp.
c) There exists an element f 2CG(O(x);Cp), f 60 such that F(f;z)2Qp(z).
2. Forf 2CG(O(x);Cp) the elementf(x) belongs toQgp[x], and so with notations as in Proposition 1 one has: f(x)P
n anMn(x). Then for any t2O(x) it followsf(t)P
n anMn(t). Then by (8) one can write:
F(f;z)X
n0
Z
O(x)
f(t)(z z0)n
(z t)n1 dm(t)X
n0
X
m0
Z
O(x)
amMm(t)(z z0)n (z t)n1 dm(t):
7. Cauchy Transforms of measures.
1. In what follows we shall prove that Barsky's Theorem (see [B]) is valid forE(x),x2Cp. Namely, we shall prove that there exists a bijective mapping between the measures on O(x) and rigid analytic functions on E(x) with residue zero at infinity and which verify some boundary condi- tions.
LetF:E(x)!Cpbe a rigid analytic function such thatF(1)0 and that the set of real numbers fekFkE(x;e)ge>0 is bounded. (HerekFkE(x;e)
sup
z2E(x;e)jF(z)j.)
One knows that E(x;e)Cp[ f1g nV(x;e), and V(x;e)
[s2SeB(s(x);e). By the Mittag-Leffler Theorem (see [FV]) one can write:
F(z)X
s2Se
X
n1
a(e)n;s
(z s(x))n; z2E(x;e); a(e)n;s2Cp: One also has
F(z)X
s2Se
F(e)s (z);
where Fs(e)(z) P
n1
a(e)n;s
(z s(x))n,ja(e)n;sj
en !0. Then by Cauchy's inequalities, one obtains
ja(e)n;sj enkFkE(x;e); n1:
10
By the unicity of Mittag-Leffler's conditions for any 0<e0<eone has F(e)s (z)X
t2Se0
^ t^s
F(et0)(z); z2E(x;e)
(here^t^smeanst2sH(x;e)) and so X
n1
a(e)n;s
(z s(x))nX
t2Se0
^ t^s
X
m1
a(em;t0) (z t(x))m: 11
Since jz s(x)j e, and js(x) t(x)j<e, it follows that jz t(x)j
jz s(x)jand so:
a(em;t0)
(z t(x))m a(em;s0) (z s(x))m
1 t(x)z s(x)s(x)m
a(em;t0) (z s(x))m
X
k0
mk 1 k
ht(x) s(x) z s(x)
ik : 12
If we denotemkn, then by identifying the coefficients of the terms of degreenin (11) and (12) one obtains:
a(e)n;sX
t2Se0
^ t^s
Xn 1
k0
n 1 k
a(en k;t0) (t(x) s(x))k; 13
wheren1.
Now for anyn1 one defines a sequencefmn;egn;eof measures onO(x) by the equality
mn;eX
s2Se
a(e)n;sds(x); 14
wheredydenotes the Dirac measure concentrated aty2Cp. By (13) one obtains, forn1:
a(e)1;sX
t2Se0
^ t^s
a(e1;t0); 0<e0e:
This equality further implies that for any ballBof radiusd,ed, one has m1;e(B)m1;e0(B) whereas e0e. Then by (10) and the Banach-Steinhaus Theorem (see [R]) there results that the mapping
B?m1(B)lim
e m1;e(B)
(whereBruns over all the open balls ofO(x)) defines ap-adic measure on O(x). One says thatmm1is themeasure associatedto the rigid analytic functionF.
Furthermore, forn2, by (13) one obtains a(e)2;sX
t2Se0
^ t^s
[a(e2;t0)a(e1;t0)(t(x) s(x))]:
IfBis an open ball ofO(x) of radiusd, by the previous equality and (14) there results that
m2;e m2;e0 X
s(x)2Bs2Se
X
t2Se0
^ t^s
a(e1;t0)(t(x) s(x))
and so by (10) one has:
jm2;e m2;e0j ee0j jj jF E(x;e0)eM whereMsup
e>0 ej jj jF E(x;e)<1, by hypothesis. Then by (10) and the Banach- Steinhaus Theorem, there exists a measurem2onO(x) defined by
m2(B)lim
e m2;e(B);
for all the open ballsBofO(x). In the same manner for alln3 one can define a measuremnonO(x) by:
mn(B)lim
e mn;e(B):
Next, by an easy computation it follows that for all n2, one has mn;e
Men 1, and somn 0 for alln2. In what follows we shall prove that
F(z) Z
O(x)
1
z tdm(t); z2E(x):
15
Indeed, with the above notations and using (10) one has:
F(z)
Z
O(x)
1
z tdm1;e(t)
F(z) X
s2Se
a(e)1;s z s(x)
e2j jj jF E(x;e)Me;
forjzjsufficiently large. It is clear, by the definition ofmm1, that (see (14))
one has: Z
O(x)
1
z tdm(t)lim
e!0
Z
O(x)
1
z tdm1;e(t)
and so by the last inequality one obtains (15) forjzjsufficiently large. Be- causeE(x) is infra-connected, by analytic continuation one obtains (15) for allz2E(x). Finally one has the following result:
THEOREM6. For any x2Cp, there exists a bijective mapping between the p-adic measures on O(x)and the functions F:E(x)!Cpwhich verify the following conditions:
i) F is rigid analytic on E(x), and F(1)0.
ii) The set of real numbersfej jj jF E(x;e)ge>0 is bounded.
Moreover, by this bijective mapping the rigid analytic and equivar- iant functions are in one to one correspondence with equivariant mea- sures on O(x).
8. Cauchy Transforms of Lipschitzian Distributions.
LetXbe an arbitrary compact subset ofCpwithout isolated points. A mapf :X!Cpis said to bel-Lipschitzian provided there exists a positive real number lsuch thatjf(x) f(y)j ljx yj, for any x;y2X. A map f :X!Cpis said to be Lipschitzian provided there exists a real numberl for which f isl-Lipschitzian. Let us remark that Lip(X;Cp), the set of Lipschitzian functions, is aCp-vector space. Moreover, it is a Banach space with the norm
f j j j j sup
x6y
jf(x) f(y)j jx yj sup
x2Xjf(x)j:
Let us denote byV(X) the set of open compact subsets ofX. It is clear that any element ofV(X) is a disjoint finite union of open balls ofX.
A distributionmonXwith values inCpis a mapm:V(X)!Cpwhich is finitely additive, thus if D [n
i1Di with Di2V(X) for 1in and Di\Dj ;for 1i6jn, thenm(D)Pn
i1m(Di).
We call a distributionm:V(X)!CpLipschitzian provided that for any e>0 there is a de>0 such that for any 0<dde and any x2X, djm(B(x;d))j e.
One knows (see [VZ, Theorem 1]) that any Lipschitzian function f :X!Cp is integrable with respect to any Lipschitzian distribution m:V(X)!Cp. Moreover, by the proof of this result, for anye>0 there existsde>0 such that if
S(m;f;B1;. . .;Bn;x1;. . .;xn)Xn
i1
f(xi)m(Bi)
is an arbitrary Riemann sum withxi2BiandBiopen ball of radiusdide
for 1in, the following inequality holds:
S(m;f;B1;. . .;Bn;x1;. . .;xn) Z
X
f(t)dm(t)emaxf1;2lg;
16
wherelis the Lipschitzianity constant with respect tof.
Let us consider now an element z2P1(Cp)nX, and f 2Lip(X;Cp).
DefineT(f;z):X!Cpby
T(f;z)(t) f(t)
z t; t2X:
17
It is easy to see that T(f;z) is well defined and Lipschitzian. In fact, if dd(z;X) is the distance fromztoX, andt1;t22Xwe have
jT(f;z)(t1) T(f;z)(t2)j
f(t1) z t1
f(t2) z t2
Ajt1 t2j;
18
whereAA(z;f;X) 1
d2maxfljzj; lsup
t2X jtj; sup
t2X jf(t)jg.
We can integrate this function with respect to any Lipschitzian dis- tribution, so the map
FX(m;f;z) Z
X
T(f;z)(t)dm(t)
is well defined. It is the Cauchy transform of the Lipschitzian function f with respect to the Lipschitzian distributionm. Using a similar argument as in the proof of Theorem 6.1 from [APZ2] one obtains the following result.
THEOREM7. Let X be a compact subset ofCpwithout isolated points,m a Lipschitzian distribution on X, and f 2Lip(X;Cp). Then
FX(m;f;z):
Z
X
f(t) z tdm(t) 19
is well defined, rigid analytic on P1(Cp)nX, and vanishes at infinity.
Moreover, any element of X is a singular point of FX(m;f;z).
Next, let f 2Lip(X;Cp) and z2P1(Cp)nX. From (16), with T(f;z) instead off andSfor the corresponding Riemann sum, one has
jS FX(m;f;z)j emaxf1;2Ag;
20
whereAis defined above. If (Bi)1inis a covering ofXwith disjoint open balls of radiusde, we have from the definition ofmthat
jSj eM dde; 21
whereMsup
x2Xjf(x)j. From (20) and (21) one has jFX(m;f;z)j emax
1;2A;M dde
; 22
and so
d2jFX(m;f;z)j emax
d2;2d2A;Md de
: 23
Denoting byEd(X) the complement inP1(Cp) of ad-neighborhood ofX, and using the definition ofA, from (23) we obtain
limd!0d2jjFX(m;f;z)jjEd(X)Be;
24
whereBis an absolute constant that does not depend ond. Lettinge!0, we obtain the following result.
THEOREM8. Let X be a compact subset ofCpwithout isolated points.
Let f 2Lip(X;Cp)and letmbe a Lipschitzian distribution on X. To any pair(m;f)as above, we can associate a rigid analytic function FX(m;f;z)in such a way that it vanishes at infinity, and satisfies the boundary condi- tion:
d!0limd2jjFX(m;f;z)jjEd(X)0:
25
If X,mand f are equivariant then FX(m;f;z)is also equivariant.
A natural question that arises is to provide a converse to the statement of this theorem.
REFERENCES
[A] V. ALEXANDRU,On the transcendence of the trace function, Proceedings of the Romanian Academy, vol.6, no. 1 (2005), pp. 11-16.
[APP] V. ALEXANDRU- E.L. POPESCU - N. POPESCU, On the continuity of the trace, Proceedings of the Romanian Academy, Series A, vol.5, no. 2 (2004), pp. 117-122.
[APZ1] V. ALEXANDRU- N. POPESCU- A. ZAHARESCU,On closed subfields ofCp, J.
Number Theory,68, 2 (1998), pp. 131-150.
[APZ2] V. ALEXANDRU- N. POPESCU- A. ZAHARESCU,Trace onCp, J. Number Theory,88, 1 (2001), pp. 13-48.
[APZ3] V. ALEXANDRU- N. POPESCU- A. ZAHARESCU,A representation theorem for a class of rigid analytic functions, Journal de Theories des Nombres de Bordeaux,15(2003), pp. 639-650.
[Ar] E. ARTIN, Algebraic Numbers and Algebraic Functions, Gordon and Breach, N.Y. 1967.
[B] D. BARSKY, Transformation de Cauchy p-adique et algebre d'Iwasawa, Math. Ann.,232(1978), pp. 255-266.
[Br] F. BRUHAT,Integration p-adique, Seminaire Bourbaki, 14e annee, 1961/62, no. 229.
[FV] J. FRESNEL- M. VAN DERPUT,Geometrie Analitique Rigide et Applica- tions, Birkhauser, Basel, 1981.
[Ka] N. M. KATZ,p-adic L-functions for CM-fields, Invent. Math.,48(1978), pp.
199-297.
[R] W. RUDIN, Functional Analysis, Mc. Graw Hill Book Company, N.Y., 1973.
[Sch] W. H. SCHIKHOF,Ultrametric calculus. An introduction to p-adic Analy- sis, Cambridge Univ. Press, Cambridge, U.K., 1984.
[VZ] M. VAÃJAÃITU- A. ZAHARESCU,Lipschitzian Distributions, to appear in Rev.
Roumaine Math. Pures Appl.
Manoscritto pervenuto in redazione il 12 settembre 2006.