Modulus of continuity on parts of the boundary and solid modulus of continuity
Salla Franz´ en
U.U.D.M. Project Report 2002:P7
Examensarbete i matematik, 20 po¨ang Handledare och examinator: Burglind Juhl-J¨oricke
Augusti 2002
Department of Mathematics
Our starting point is a theorem of Hardy and Littlewood, which asserts that an analytic function in the open unit disc D satisfies a H¨older condition of order α ∈ (0, 1), if its boundary values on the unit circle ∂D have this property. More precisely, denote by A(D) the algebra of analytic functions in the unit disc which are continuous in its closure D . The following theorem holds:
Theorem 1 Let f ∈ A(D). Suppose for a number α ∈ (0, 1] and a positive constant C
|f (w) − f (z)| ≤ C|w − z|α for all w, z ∈ ∂D.
Then
|f (w) − f (z)| ≤ C|w − z|α for all w, z ∈ D.
Here we stated the theorem also for α = 1. This case was proved later. For completeness we will give here a proof which follows [R-S-T] and works for α ∈ (0, 1]. We start with the following lemma which works for arbitrary domains G in Cn. Denote by A(G) the algebra of analytic functions in G which are continuous in G.
Lemma 2 Let G be a domain in Cn and let f ∈ A(G). Then for every δ > 0 the following supremum
sup{|f (w) − f (z)| : w, z ∈ G, |w − z| < δ}
is attained when one of the points w or z is contained in the boundary.
Proof of Lemma 2 Denote w = z + h. Both points z and z + h are contained in G iff z ∈ GT(G − h). Fix the complex number h and apply the maximum principle to the analytic function f (z + h) − f (z) in GT(G − h).
Since the boundary of the set GT(G − h) is contained in the union of the boundaries ∂G and ∂(G − h), sup{|f (z + h) − f (z)| : z ∈ GT(G − h)} is attained if either z ∈ ∂G or z ∈ ∂(G − h), i.e. w = z + h ∈ ∂G.
Proof of Theorem 1 By Lemma 2 we have to prove that for w ∈ ∂D and z ∈ D, |w − z| < δ, the inequality
|f (w) − f (z)| ≤ C|w − z|α
holds. We may assume that z is in the open unit disc. Consider a holomor- phic branch of φw(z) = (w−z)1 α, z ∈ D. This function is in the Hardy space Hp for each p ∈ (0,α1]. Indeed, if w = eiψ∈ ∂D and z = reiϕ∈ D, then
1
|w − z|pα = 1
|eiψ− reiϕ|pα ≤ 2
|eiψ− eiϕ|pα
and for a fixed ψ the latter function defines an integrable function on ∂D.
It follows that the function gw(z) = (f (w) − f (z))φw(z) is in Hp and is continuous on D\{w}. Since f is H¨older continuous on ∂D, |gw(z)| ≤ C for all z ∈ ∂D\{w}. Hence |gw(z)| ≤ C on D\{w}. This is exactly the inequality we wanted to prove.
For our purposes we need the corresponding statement for arbitrary sim- ply connected (maybe unbounded) domains in C (see, e.g. [R-S-T]).
Theorem 3 Let G ⊂ C be a simply connected domain and α ∈ (0, 1].
Suppose f ∈ A(G) and for some number C
|f (w) − f (z)| ≤ C|w − z|α for all w,z ∈ ∂G. (1) Then
|f (w) − f (z)| ≤ AC|w − z|α for all w,z ∈ G. (2) Here A is a constant which does not depend on G or on f . For a proof we refer to [R-S-T].
Theorem 3 can be proven for a very general class of domains in Cn, n > 1. However, one can expect, that for some domains one can replace the boundary by the Shilov boundary of the domain. The Shilov boundary of a domain G ⊂ Cn is the smallest closed subset S(G) of G such that the maximum principle
|f (z)| ≤ max
S(G)
|f | for all z ∈ G and all f ∈ A(G) holds. Such a set always exists.
In [J¨o] a class of domains G is described for which (1) with ∂G replaced by the Shilov boundary S(G) of G implies (2). On the other hand examples are given of domains G for which this is not the case. It is not known if, e.g., for bounded pseudoconvex domains of holomorphy, the inequality (1) for ∂G replaced by an arbitrary open subset of ∂G which contains the Shilov boundary implies (2).
Here we will consider bounded complete Reinhardt domains in C2 which are pseudoconvex and have C∞boundary. Recall that a complete Reinhardt domain G ⊂ Cn is a domain for which z = (z1, ..., zn) ∈ G implies that (z1ζ1, ..., znζn) ∈ G for arbitrary ζj ∈ D, j = 1, ..., n. A domain G has C∞ boundary, if there is a C∞ function ρ in Cn (called defining function of G) such that ρ < 0 on G, ρ > 0 on Cn\G, ρ = 0 on ∂G and dρ 6= 0 on ∂G.
Such a domain is pseudoconvex, if for each p ∈ ∂G the Levi form Lpρ of ρ at the point p is nonnegative in complex tangent directions:
Lpρ(w) =
n
X
j,k=1
( ∂2
∂zj∂zkρ)(p)wjwk ≥ 0
for all w ∈ Cn such that Pnj=1∂z∂ρ
j(p)wj = 0.
Note that a domain G with C∞ boundary satisfies this condition iff it is a domain of holomorphy, i.e. iff there exists an analytic function in G which does not extend analytically across the boundary. For a domain G ⊂ Cn with C2 boundary a point p ∈ ∂G is called a strictly pseudoconvex boundary point if the Levi form Lpρ(w) > 0 for all w ∈ Cn\{0} such that Pn
j=1
∂ρ
∂zj(p)wj = 0. The conditions do not depend on the choice of the defining function ρ.
The following theorem of Basener and Pflug describes the Shilov boundary for smoothly bounded pseudoconvex domains.
Theorem 4 [Ba], [Pf ]. Let G be a bounded pseudoconvex domain in Cn with C∞ boundary. Then the Shilov boundary S(G) is equal to the closure of the set of strictly pseudoconvex boundary points of G.
To understand the properties of a complete Reinhardt domain D ⊂ C2 it is useful to associate with D the domain D∗ in R2 defined in the following way: D∗= {(ξ1, ξ2) ∈ R2 : (eξ1, eξ2) ∈ D}. By the definition of a complete Reinhardt domain, D is the union of all bidiscs of the form
{(z1, z2) ∈ C2 : |z1| < eξ1, |z2| < eξ2} for some (ξ1, ξ2) ∈ D∗.
A complete Reinhardt domain D in C2 with C∞ boundary is pseudoconvex iff D∗ is convex. A point z = (z1, z2) ∈ ∂D is a strictly pseudoconvex boundary point iff (ξ1, ξ2) = (log|z1|, log|z2|) is a strictly convex boundary point of D∗.
Let now D be a bounded complete pseudoconvex Reinhardt domain in C2 with C∞boundary. Denote by P the set of all strictly pseudoconvex bound- ary points of D∗. Then
∂D∗\P =[
k
ωk∗
for an at most countable number of open line segments ω∗k in R2. Moreover
∂D\S(D) =[
k
ωk,
where ωk is the connected relatively open subset of ∂D which corresponds to ωk∗. If ωk∗ is bounded this correspondence has the following form:
ωk= {(z1, z2) ∈ C2 : log|z1| = ξ1, log|z2| = ξ2, (ξ1, ξ2) ∈ ω∗k} (3) Note that in this case
∂ωk= {(z1, z2) ∈ C2 : (ξ1, ξ2) = (log|z1|, log|z2) ∈ ∂ω∗k}, (4)
and this set is contained in the Shilov boundary. We will consider here only the case when the ωk∗ are uniformly bounded. This is equivalent to the following condition:
For z = (z1, z2) ∈ ∂D\S(D), cj ≤ |zj| ≤ Cj (5) for j=1,2 and some constants cj, Cj, 0 < cj < Cj.
The main goal of this work is to prove the following theorem.
Theorem 5 Let D be a bounded complete pseudoconvex Reinhardt domain in C2 with smooth boundary for which condition (5) holds. Let α ∈ (0, 1].
Then there is a constant A depending only on D such that for any function f ∈ A(D) the inequality
|f (w) − f (z)| ≤ C|w − z|α for all w, z ∈ S(D) (6) implies
|f (w) − f (z)| ≤ CA|w − z|α for all w, z ∈ D. (7) We think that condition (5) can be removed.
Before proving theorem 5 we state and prove the version of it for the bidisc, which will be used later. Let
Dn= {z ∈ Cn: |zj| < rj, j = 1, ..., n}
be a polydisc. The Shilov boundary of Dn is the direct product of the boundaries of the onedimensional discs {|zj| = rj}, i.e.
S(Dn) = {z ∈ Cn: |zj| = rj, j = 1, ..., n}.
Proposition 6 Let Dn be an arbitrary polydisc in Cn and f ∈ A(Dn).
Suppose
|f (w) − f (z)| ≤ C|w − z|α if w, z ∈ S(Dn) Then
|f (w) − f (z)| ≤ Cγ|w − z|α for all w, z ∈ Dn. γ is a constant not depending on f nor on the radii rj.
Proof of Proposition 6 Write z = (z1, ..., zn) ∈ Cn, w = (w1, ..., wn) ∈ Cn. Denote z0 = (z2, ..., zn), D0 = {z0 : |z2| < r2, ..., |zn| < rn}. The function
fz0(ζ) = f (ζ, z0), |ζ| < r1,
is analytic for any fixed z0 with |z2| ≤ r2, ..., |zn| ≤ rn. Indeed, the function is the uniform limit of the functions
frz0(ζ) = f (ζ, rz0), r ∈ (0, 1), r → 1.
Here rz0 = (rz2, ..., rzn). Analyticity of the function frz0 follows from the fact that the point (ζ, rz0) ∈ D for r ∈ (0, 1), |ζ| < r1 and, moreover, the function fz0 ∈ A(|ζ| < r1) for each z0 ∈ D0. By the same argument for each z1, w1, |z1| ≤ r1, |w1| ≤ r1, the function z0 → f (w1, z0) − f (z1, z0) is in A(D0). Hence, the maximum of this function is obtained for z0 ∈ S(D0).
Now theorem 1 gives
|f (w1, z0) − f (z1, z0)| ≤ C|w1− z1|α. Applying the same argument to the function
ζ → f (w1, ..., wj−1, ζ, zj+1, ..., zn), j = 2, ..., n, we get the assertion.
Proof of Theorem 5 The first step of the proof is the following Lemma.
Lemma 7 (6) implies
|f (w) − f (z)| ≤ Cb|w − z|α for w, z ∈ ∂D (8) for a constant b which depends only on D.
The second step of the proof consists of the following Lemma.
Lemma 8 (8) implies (7).
Lemma 8 is known for a large class of domains. For completeness in the end of the paper we give a short argument which works in our situation.
Lemma 9 For each k,
|f (w) − f (z)| ≤ CB|w − z|α for z, w ∈ ωk for a constant B depending only on D.
Lemma 9 implies Lemma 7. This is a consequence of the fact that D has smooth boundary. We will not give the arguments here.
Proof of Lemma 9 The set ωk∗ is an open line segment in R2, say
ωk∗= {(ξ1, ξ2) ∈ R2 : ξ2 = akξ1+ bk, ξ1 ∈ (σ1, σ2)} (9) (or with the roles of ξ1 and ξ2 interchanged). Here ak, bk, σ1 and σ2 are real numbers. We may assume that |ak| ≤ 1.
Let w, z ∈ ωk, |w − z| < δ for a small positive number δ. (10)
Using condition (5) we get
|1 − wj zj
| < a0δ, j = 1, 2, (11)
where a0= max(c1
1,c1
2) depends only on D.
Write z in the form z = (z1, z2) = (elog|z1|+iϕ1, elog|z2|+iϕ2) = (eζ1, eζ2). Since z ∈ ωk, we have Reζ2 = akReζ1 + bk, see (3). Hence (after the branches of the arguments ϕ1 and ϕ2 are chosen) there is a uniquely defined β ∈ R, β = β(ζ), such that ζ2 = akζ1+ bk+ iβ, namely β = ϕ2− akϕ1.
By (11) we can find (the branches of) arguments ϕ01 and ϕ02 of w1, w2 such that
w = (w1, w2) = (|w1|eiϕ01), |w2|eiϕ02)) = (eζ10, eζ20) and (12)
|ϕ01− ϕ1| ≤ λ1δ, |ϕ02− ϕ2| ≤ λ2δ (13) for constants λ1, λ2 depending only on D. As above ζ20 = akζ10 + bk+ iβ0, where β0 = ϕ02− akϕ01. Inequalities (10) and (13) give
|ζ − ζ0| ≤ c0δ (14)
|β − β0| ≤ c00δ (15)
For β ∈ R consider the analytic function
φβ(ζ) = akζ + bk+ iβ, Reζ ∈ (σ1, σ2).
So, z = (eζ, eφβ(ζ)) for some ζ, Reζ ∈ (σ1, σ2), (16) w = (eζ0, eφβ0(ζ0)) for some ζ0, Reζ0 ∈ (σ1, σ2),
with |ζ − ζ0| ≤ c0δ, |β − β0| ≤ c00δ.
The following Lemmas hold.
Lemma 10 There is a constant M depending only on D such that for the function gβ(ζ) = f (eζ, eφβ(ζ)), Reζ ∈ (σ1, σ2) the following inequality holds
|gβ(ζ) − gβ( ˜ζ)| ≤ M |ζ − ˜ζ| for all ζ, ˜ζ ∈ C, Reζ, Re ˜ζ ∈ (σ1, σ2).
Lemma 11 For all ζ ∈ C, Reζ ∈ (σ1, σ2),
|gβ(ζ) − gβ0(ζ)| ≤ mδα for a constant m depending only on D.
We postpone the proof of the Lemmas 10 and 11 and finish the proof of Lemma 9.
Proof of Lemma 9 continued. Write the points z and w as in (16), with
|ζ − ζ0| ≤ c0δ and |β − β0| ≤ c00δ. By Lemma 10 we have
|f (eζ, eφβ(ζ)) − f (eζ0, eφβ(ζ0))| = |gβ(ζ) − gβ(ζ0)| ≤ M (c0)αδα. By Lemma 11 we have
|f (eζ0, eφβ(ζ0)) − f (eζ0, eφβ0(ζ0))| ≤ mδα. The two inequalities prove Lemma 9.
Proof of Lemma 10 The function gβ(ζ), Reζ ∈ (σ1, σ2), is a bounded analytic function as the uniform limit of the analytic functions
grβ(ζ) = f (reζ, reφβ(ζ)), Reζ ∈ (σ1, σ2). Here r ∈ (0, 1), r → 1.
By the same reason it is continuous in Reζ ∈ [σ1, σ2]. By Theorem 3 it is enough to prove that
|gβ(ζ) − gβ( ˜ζ)| ≤ M0|ζ − ˜ζ|α,
if ζ, ˜ζ ∈ C and Reζ equals σ1 or σ2, Re ˜ζ equals σ1 or σ2. Here M0 depends only on D. This follows from (6) and the following two facts:
|(eζ, eφβ(ζ)) − (eζ˜, eφβ( ˜ζ))| ≤ ˜a|ζ − ˜ζ|
for Reζ, Re ˜ζ ∈ [σ1, σ2] and a constant ˜a not depending on β (this follows from the fact, that φβ has uniformly bounded derivative for Reζ ∈ [σ1, σ2]).
Moreover, if Reζ equals σ1 or σ2, then (eζ, eφβ(ζ)) ∈ S(D).
Proof of Lemma 11 If Reζ ∈ [σ1, σ2], then
|(eζ, eφβ(ζ)) − (eζ, eφβ0(ζ))| < b0|β − β0| ≤ b0c00δ.
Since for Reζ equal to σ1 or σ2 both points (eζ, eφβ(ζ)) and (eζ, eφβ0(ζ)) are in S(D) the maximum principle applied to the analytic function gβ − gβ0, gives
|gβ(ζ) − gβ0(ζ)| = |f (eζ, eφβ(ζ)) − f (eζ, eφβ0(ζ))| ≤ C(b0c00δ)α = mδα. Proof of Lemma 8 Let δ > 0 and z = (z1, z2), w = (w1, w2) ∈ D such that |w − z| < δ. Say |z2| ≤ |w2|. Let rz, rw > 0 be such that (z1, rz), (w1, rw) ∈ ∂D. Then by Proposition 6, f is H¨older continuous of order α on the closure of the bidiscs Dz and Dw, where:
Dz = {ζ ∈ C : |ζ1| < |z1|, |ζ2| < rz}, Dw = {ζ ∈ C : |ζ1| < |w1|, |ζ2| < rw}.
We have two cases. If |z1| ≤ |w1|, then (z1, z2) ∈ Dw, so |f (w) − f (z)| ≤ C|w − z|α. If |z1| ≥ |w1| then by (18), (w1, z2) ∈ DzT
Dw and we have
|f (w) − f (z)| ≤ |f (w) − f (w1, z2)| + |f (w1, z2) − f (z)|, from which (7) now follows.
Bibliography
[H-L] G.H.Hardy and J.E.Littlewood, ”Some properties of fractional inte- grals”, II, Math.Zeit.34(1931), 403-439.
[R-S-T] L.A.Rubel, A.L.Shields and B.A.Taylor, ”Mergelyan Sets and Modulus of Continuity of Analytic Functions”, J.Appr.Th.15(1975), 23-40.
[J¨o] B.J¨oricke, ”The Relation between the Solid Modulus of Continuity and the Modulus of Continuity along the Shilov Boundary for An- alytic Functions of Several Variables”, Math.USSR.Sbornik Vol.50, No.2(1985), 495-511.
[Ba] R.F.Basener, ”Peak Points, Barriers and Pseudoconvex Boundary Points”, Proc.Am.Math.Soc.65(1977), 89-92.
[Pf ] P.Pflug, ” ¨Uber Polynomiale Funktionen auf Holomorphiegebieten”, Math Z.139(1974), 133-139.