FOURIER REPRESENTATIONS IN BERGMAN SPACES
Pranav Upadrashta
A thesis submitted in partial fulfillment of the requirements for the degree of Master of Arts
Department of Mathematics
Central Michigan University Mount Pleasant, Michigan April 2018 ACKNOWLEDGEMENTS
I am glad to have Prof. Debraj Chakrabarti supervising my Master’s thesis, who, along with being a brilliant teacher, is also a wonderful professional mentor. I thoroughly enjoyed discussing math (and a host of other topics) with him, and still do. I would like to thank Prof. David Barrett and Prof. Sivaram Narayan for their careful ex- amination of this thesis and for providing insightful remarks on the subject matter. I am fortunate to have them on my thesis defense committee. I’d also like to thank Prof. Meera Mainkar for having me at dinner on occasions where I lost track of time working on this thesis - she was a godsend!
ii ABSTRACT
FOURIER REPRESENTATIONS IN BERGMAN SPACES
by Pranav Upadrashta
It is a well known classical theorem of Paley and Wiener that every function in the Hardy space of the upper half plane is the holomorphic Fourier transform of a function square integrable on the positive real line. This correspondence establishes an isometric isomorphism between the Hardy space of the upper half plane and the Hilbert space of square integrable functions on the positive real line. Under this isometric isomorphism, the action of the group of translation symmetries of the upper half plane on the Hardy space corresponds to multiplication by a function in the Hilbert space of square integrable functions. In this thesis we obtain similar results for Bergman spaces of certain domains which we call polynomial half spaces. The upper half plane and Siegel upper half space in higher dimen- sions are familiar examples of polynomial half spaces. Like the upper half plane, these domains have three one-parameter subgroups of symmetries, namely rotations, translations and scalings. Corresponding to each of the one-parameter subgroup of symmetries of a polynomial half space above, we obtain an isometric isomorphism between the Bergman space of the polynomial half space and a Hilbert space of square integrable functions. This isometric isomorphism respects the action of the corresponding one-parameter subgroup of symmetries on the Bergman space of polynomial half space analogous to the classical Paley-Wiener theorem. We give complete and detailed proofs of the three Fourier representations. Some similar results can be found in literature for the translation group, though perhaps not for the scaling group. Finally we use the Fourier representations of Bergman space of polynomial half spaces to obtain integral representations of the Bergman kernel of these domains. For the Siegel upper half space, the classical formula for its Bergman kernel is recaptured using our method.
iii TABLE OF CONTENTS
LIST OF FIGURES ...... vi
LIST OF SYMBOLS ...... vii
CHAPTER
I. INTRODUCTION ...... 1 I.1. Motivation ...... 1 I.1.1. Motivating Example ...... 1 I.1.2. Relation with the Translation Group ...... 1 I.2. A Class of Domains with Large Automorphism Group ...... 2 I.2.1. Bergman Space of a Domain ...... 2 I.2.2. Polynomial Half Space ...... 3 I.3. Automorphisms of Polynomial Half Spaces...... 4 I.3.1. Rotation Subgroup ...... 4 I.3.2. Translation Subgroup ...... 5 I.3.3. Scaling Subgroup ...... 6 I.4. Results ...... 6 I.4.1. Rotation Subgroup ...... 7 I.4.2. Translation Subgroup ...... 9 I.4.3. Scaling Subgroup ...... 10 I.4.4. Bergman Kernel of Polynomial Half Spaces ...... 12 I.5. Outline of this Thesis ...... 14 I.6. Historical Note ...... 15
II. PRELIMINARIES ...... 16 II.1. Bergman Spaces ...... 16 II.2. Geometry of Polynomial Half Spaces ...... 18 II.2.1. Siegel Upper Half Space ...... 21 II.2.2. Bedford-Pinchuk Domains...... 21 II.3. Direct Integral Representations ...... 22 II.3.1. Direct Integral Representation of Hp ...... 23 II.3.2. Direct Integral Representation of Xp ...... 27
III. FOURIER REPRESENTATIONS ...... 35 III.1. Maximal Compact Subgroup ...... 35 2 III.1.1. A Fourier Representation of A (Ep) ...... 36 III.2. Translation Subgroup ...... 39 III.2.1. The case of the strip S(a,b) ...... 39 2 III.2.2. A Fourier Representation of A (Up) ...... 47 III.3. Scaling Subgroup ...... 51
iv III.3.1. The case of the sector V(a,b) ...... 51 2 III.3.2. Another Fourier Representation of A (Up) ...... 55
IV. BERGMAN KERNEL OF POLYNOMIAL HALF SPACES ...... 61 IV.1. Reproducing Kernel Hilbert Spaces ...... 61 IV.1.1. Reproducing Kernel Hilbert Spaces ...... 61 IV.1.2. Computing Reproducing Kernels ...... 62 IV.2. Asymptotic Behavior of the Bergman Kernel ...... 66 IV.3. A Formula for the Bergman Kernel of Ep ...... 70 IV.4. A Formula for the Bergman Kernel of Up ...... 72 IV.5. Another Formula for the Bergman Kernel of Up ...... 76 IV.6. An Application: the Siegel Upper Half Space ...... 78
REFERENCES ...... 82
v LIST OF FIGURES
FIGURE PAGE
1. Action of the Rotation Subgroup on D ...... 5
2. Action of the Translation Subgroup on U ...... 6
3. Action of the Scaling Subgroup on U ...... 7
4. The Rectangle SN ...... 45
5. Conformal Equivalence of a Sector with a Strip ...... 51
vi LIST OF SYMBOLS
SYMBOL PAGE
m ...... 3
p ...... 3
Up ...... 3
Ep ...... 3
Bp ...... 7
Wp(k) ...... 7
Yp ...... 8
Hp ...... 9
Xp ...... 10
λ(p(ζ),t) ...... 10
ρbγ ...... 11 1/µ ...... 11
M ...... 23
Sp(t) ...... 24
δζ ...... 28
Qp(t) ...... 30
S(a,b) ...... 39
ωa,b ...... 39
LTrans(p) ...... 47
V(a,b) ...... 51
Cp ...... 53
LScal(p) ...... 55
vii RH ...... 61
viii CHAPTER I INTRODUCTION
I.1. Motivation
I.1.1. Motivating Example
Let U = {z ∈ C|Imz > 0} denote the upper half plane in C. Recall that the Hardy space H2(U) of the upper half plane consists of functions which are holomorphic in U and uniformly square integrable on horizontal lines in U. More precisely, H2(U) is the Banach space of functions F ∈ O(U) such that Z 2 2 kFkH2(U) := sup |F(x + iy)| dx < ∞. y>0 R A classical theorem of Paley and Wiener (see [21, Theorem 19.2]) states that every func- tion in H2(U) arises as the holomorphic Fourier transform of a square integrable function on the positive real line (0,∞). More precisely, the map T : L2(0,∞) → H2(U) given by
Z ∞ T f (z) = f (t)ei2πzt dt (I.1) 0 is an isometric isomorphism of Banach spaces. Consequently, H2(U) is a Hilbert space.
I.1.2. Relation with the Translation Group
For θ ∈ R and F ∈ H2(U), let γ(θ)F ∈ H2(U) be given by (γ(θ)F)(z) = F(z+θ). Since the translation z 7→ z + θ is a biholomorphic automorphism of U, θ 7→ γ(θ) is a unitary action of
R on H2(U). i.e., 2 γ : R → U (H (U) is a unitary representation of R in the Hilbert Space H2(U). Thus, we have in (I.1) ∞ Z (γ(θ)F)(z) = F(z + θ) = ei2πθt f (t)ei2πzt dt = T(ei2πθ(·) f )(z), 0
1 for all t ∈ (0,∞). That is, for each θ ∈ R, the following diagram commutes
L2(0,∞) T H2(U)
χθ γ(θ) L2(0,∞) T H2(U)
2 i2πθ(·) where χθ denotes the operator which multiplies a function in L (0,∞) by e . We think of T as a linear change of co-ordinates in the vector space H2(U) which diagonalizes all the linear op- erators {γ(θ)|θ ∈ R}. When such simultaneous diagonalization holds, we say that the isometric isomorphism T : L2(0,∞) → H2(U) in (I.1) a Fourier representation of H2(U) corresponding to the translation group, the subgroup of Aut(U) consisting of maps of the form z 7→ z + θ for real θ.
I.2. A Class of Domains with Large Automorphism Group The goal of this thesis is to study some analogs of (I.1) for Bergman spaces of certain
n domains in C admitting large groups of automorphisms. We first recall the notion of Bergman space of a domain, and then introduce the class of domains we want to study.
I.2.1. Bergman Space of a Domain
n Let Ω be a domain in C , and let λ be a positive continuous function on Ω. The space L2(Ω,λ) consists of all measurable complex valued functions f on Ω such that
Z 2 2 k f kL2(Ω,λ) := | f (z)| λ(z) dV(z) < ∞, Ω where we identify functions that are almost everywhere equal. This is a Hilbert space of L2 functions on Ω with the natural inner product. The closed subspace A2(Ω,λ) of L2(Ω,λ) (see Proposition II.2) consisting of functions holomorphic in Ω is called the weighted Bergman space of Ω with weight λ. When λ ≡ 1, we denote A2(Ω,1) = A2(Ω). The Hilbert space A2(Ω) is called the Bergman space of Ω.
2 I.2.2. Polynomial Half Space
n We consider a class of domains in C which generalize the upper half plane U ⊂ C. Let n m = (m1,··· ,mn) be a given tuple of positive integers. Given a tuple α ∈ N of nonnegative integers define the weight of α with respect to m to be
n αi wtm(α) := ∑ . i=1 2mi
n A real polynomial p : C → R is then called a weighted homogeneous balanced polynomial (with n respect to the tuple m ∈ N ) if p is of the form
α β p(w1,··· ,wn) = ∑ Cα,β w w . (I.2) wtm(α)=wtm(β)=1/2
n n Let p : C → R be a weighted homogeneous balanced polynomial such that p ≥ 0 on C . n+1 The polynomial half space Up defined by p is the unbounded domain in C
n Up = {(z,w) ∈ C × C | Im z > p(w)}. (I.3)
We show in Lemma II.6 that a polynomial half space Up is biholomorphically equivalent to the
n+1 bounded domain Ep ⊂ C , where
n 2 Ep = {(z,w) ∈ C × C | |z| + p(w) < 1}. (I.4)
We call the domain Ep a polynomial ellipsoid. In the trivial case n = 0, the unit disk D = {z ∈ C| |z| < 1} is the only example of a polynomial ellipsoid corresponding to the only weighted ho- mogeneous polynomial p ≡ 0 on the 0-dimensional space C0 = {0}.The domain D is biholomor- phically equivalent to the upper half plane U ⊂ C, which is a polynomial half space corresponding n to the polynomial p = 0. The unit ball in C is another familiar example of polynomial ellipsoid (see Section II.2), which is biholomorphically equivalent to a polynomial half space called the Siegel upper half space.
3 Henceforth, m = (m1,...,mn) will denote a tuple of positive integers and p will be a nonnegative weighted homogeneous balanced polynomial with respect to m, as in (I.2). It is very natural to consider polynomial half spaces in the problem of finding Fourier representations of Bergman spaces, thanks to the following result.
Theorem 1 ([2]). Let Ω be a bounded convex domain with smooth boundary, and of finite type in sense of D’Angelo. Then Aut(Ω) is noncompact if and only if Ω is biholomorphic to a polynomial ellipsoid.
The type of a bounded domain Ω in sense of D’Angelo (see [6, Pg 118]) at a point p ∈ ∂Ω is the maximum order of contact of ambient one-dimensional complex varieties with ∂Ω at p. The automorphism group of polynomial half spaces (and therefore polynomial ellipsoids) are noncompact, and contain one-parameter subgroups (see Section I.3 below) with respect to which we can obtain Fourier representations of Bergman spaces of these domains. This explains our interest in these domains.
I.3. Automorphisms of Polynomial Half Spaces.
In this section, we describe the subgroups of Aut(Up), the group of biholomorphic auto- morphisms of Up that are of interest to us. These subgroups described below do not necessarily generate the group Aut(Up).
I.3.1. Rotation Subgroup
In case of the upper half plane U ⊂ C, the action of the rotation subgroup of Aut(U) is easier to understand in the bounded model D, the unit disk in C. Similarly, in the case of polynomial half space Up, the action of the compact one-parameter subgroup of automorphisms is easier to understand in the bounded model Ep, which we now describe.
It is apparent that for θ ∈ R, the map σθ : Ep → Ep given by
iθ n σθ (z,w) = (e z,w), for z ∈ C,w ∈ C such that (z,w) ∈ Ep (I.5)
4 n 2 is an automorphism of Ep = {(z,w) ∈ C × C ||z| + p(w) < 1}. We call such an automorphism a rotation of Ep. The rotations of Ep form a compact one-parameter subgroup of the group Aut(Ep) of biholomorphic automorphisms of Ep, isomorphic to the circle group. Since there is a bi- holomorphic equivalence Λ : Up → Ep (see Lemma II.6), therefore, for each θ ∈ R, the map −1 Λ ◦ σθ ◦ Λ : Up → Up is an automorphism of the polynomial half space Up.
Figure 1. Action of the Rotation Subgroup on D
I.3.2. Translation Subgroup
We next describe a noncompact one-parameter subgroup of automorphism group Aut(Up) of the polynomial half space Up.
For θ ∈ R, the map τθ : Up → Up given by
n τθ (z,w) = (z + θ,w), for z ∈ C,w ∈ C such that (z,w) ∈ Up (I.6)
n is an automorphism of Up = {(z,w) ∈ C × C | Imz > p(w)}, since Im(z + θ) = Imz > p(w).
We call τθ a translation of Up. The translations form a one-parameter subgroup of Aut(Up) isomorphic to R.
5 Figure 2. Action of the Translation Subgroup on U
I.3.3. Scaling Subgroup We now describe another non-compact one-parameter subgroup of automorphism group
Aut(Up) of a polynomial half space Up.
Recall that m = (m1,...,mn) is the tuple of positive integers with respect to which p is a n n weighted homogeneous balanced polynomial. For θ > 0, let ρbθ : C → C be given by 1/2m1 1/2mn n ρbθ (w) = θ w1,...,θ wn , for w ∈ C . (I.7)
Then the map ρθ : Up → Up given by
1/2m1 1/2mn ρθ (z,w) = (θz,ρbθ (w)) = θz,θ w1,··· ,θ wn for (z,w) ∈ Up (I.8) can be shown to be an automorphism of Up (see Lemma II.5). We call such an automorphism a scaling of Up. The scalings of Up form a one-parameter subgroup of Aut(Up) isomorphic to the multiplicative group of positive real numbers.
Note that, in case of the upper half plane U ⊂ C, these consist of dilations z 7→ θz, where θ > 0.
I.4. Results In this thesis, we obtain Fourier representations analogous to (I.1) corresponding to the action of the rotation, translation and scaling groups on Bergman spaces of polynomial half spaces.
6 Figure 3. Action of the Scaling Subgroup on U
n ∗ Recall that for a domain Ω ⊂ C , if φ : Ω → Ω is an automorphism then the map φ from A2(Ω) to itself given by
φ ∗ f (z) = f (φ(z)) · detφ 0(z), for all z ∈ Ω, f ∈ A2(Ω) (I.9) is an isometric isomorphism of A2(Ω) with itself (see Lemma II.3). In other words, the map
φ 7→ φ ∗ is a unitary representation of Aut(Ω) in A2(Ω). The rotation group being compact, in this case (Theorem 2) we obtain a Fourier repre-
2 sentation of A (Ep) with respect to this group as an infinite series instead of as an integral as in Theorems 3 and 4. We include this theorem here for completeness. In Theorems 3 and 4 we
2 2 n obtain Fourier representations of A (Up) as a Hilbert space of L functions on a subset of R×C , such that the functions in the Hilbert space are holomorphic in the complex parameter.
I.4.1. Rotation Subgroup
n Let Bp ⊂ C be the domain given by
n Bp = {w ∈ C | p(w) < 1}. (I.10)
For k ∈ N, let Wp(k) be the weighted Bergman space on Bp with respect to the weight w 7→ (1 − p(w))k+1, i.e.,
2 k+1 Wp(k) = A Bp,(1 − p)
7 When n = 0, the constant polynomial p = 0 is the only weighted homogeneous balanced polyno-
0 k+1 mial, so Bp = C = {0}. Then the weight w 7→ (1 − p(w)) is identically 1 for all k ∈ N, and 0 we endow C with the counting measure. Then we interpret Wp(k) as the space of all maps from 0 0 C to C that are square integrable on C , i.e., Wp(k) = C for all k ∈ N. ∞ Let Yp be the Hilbert space of sequences a = (ak)k=0 where for each k ∈ N, ak ∈ Wp(k) and ∞ 1 kak2 := ka k2 < ∞. (I.11) Yp π ∑ k Wp(k) k=0 k + 1 A convenient language to express constructions like (I.11) is that of direct integrals (see [9, Sec
7.4]). Let µ be the measure on N which assigns to each nonnegative integer k, the mass µ(k) = π k+1 . Notice that the function a ∈ Yp is such that for each k ∈ N, ak ∈ Wp(k). Then the space Yp is the direct integral Z ⊕ Yp = Wp(k) dµ(k). (I.12)
Notice that in this case the direct integral reduces to a weighted direct sum of Hilbert spaces.
2 Theorem 2. The map T : Yp → A (Ep) given by
∞ k Ta(z,w) = ∑ ak(w)z , for all (z,w) ∈ Ep (I.13) k=0 is an isometric isomorphism of Hilbert spaces.
Note that when n = 0, this reduces to a familiar result about Taylor expansions of a function in Bergman space A2(D) of the unit disc D (see Proposition III.1). 2 Let σθ be a rotation of Ep as in (I.5), when θ ∈ R. Then for each F ∈ A (Ep), the map ∗ 2 ∗ F 7→ σθ F determines an action of the rotation subgroup of Aut(Ep) on A (Ep), where σθ F is defined as in (I.9). By Theorem 2, there is a sequence a = (a0,a1,...) ∈ Yp such that F = Ta, where T is as in (I.13). Thus, for (z,w) ∈ Ep, we have
∞ ! ∗ 0 iθ k iθ σθ F(z,w) = F(σθ (z,w))detσθ (z,w) = ∑ ak(w)(e z) · e k=0
8 iθ i2θ i3θ = T(a0e ,e a1,e a2,...). (I.14)
2 Therefore T is a Fourier representation of A (Ep) corresponding to the rotation subgroup of Aut(Ep). It follows from equation (I.14) that the isometric isomorphism T in Theorem 2 diag-
2 onalizes the action of rotation subgroup of Aut(Ep) on A (Ep).
I.4.2. Translation Subgroup
2 2 We now introduce Hp, which is a space of L functions, such that functions in A (Up) are holomorphic Fourier transforms of functions in Hp (see Theorem 3). n Let Hp be the Hilbert space of measurable functions g on (0,∞) × C such that
Z Z ∞ e−4π p(w)t kgk2 := |g(t,w)|2 dV(w) dt < , (I.15) Hp ∞ Cn 0 4πt and ∂g = 0 in the sense of distributions, 1 ≤ j ≤ n, ∂w j n where w1,...,wn are co-ordinates of C . In other words, functions in Hp are square integrable on n − p(w)t (0,∞)×C with respect to the weight (t,w) 7→ e 4π /4πt, and are holomorphic in the variable n w ∈ C .
2 Theorem 3. The map TS : Hp → A (Up) given by
Z ∞ i2πzt TS f (z,w) = f (t,w)e dt, for all (z,w) ∈ Up (I.16) 0 is an isometric isomorphism of Hilbert spaces.
2 For θ ∈ R, let τθ be a translation of Up as in (I.6). Then for each F ∈ A (Up) the map ∗ 2 ∗ F 7→ τθ F determines an action of the translation subgroup of Aut(Up) on A (Up), where τθ F is defined analogous to (I.9). By Theorem 3, there is an f ∈ Hp such that F = TS f , where TS is as in
(I.16). Thus, for (z,w) ∈ Up we have
Z ∞ ∗ 0 i2π(z+θ)t i2πθ(·) τθ F(z,w) = F(τθ (z,w)) · detτθ (z,w) = f (t,w)e dt · 1 = TS(e f ). 0 9 2 Therefore (I.16) is a Fourier representation of A (Up) corresponding to the translation subgroup of
Aut(Up) as the isometric isomorphism TS in Theorem 3 diagonalizes the action of the translation 2 subgroup of Aut(Up) on A (Up).
I.4.3. Scaling Subgroup
To study Fourier representations with respect to the scaling group we introduce Xp, another
2 Hilbert space isometrically isomorphic to A (Up) (see Theorem 4). For s ∈ (−1,1) and t ∈ R let λ be given by 1 −4π sin−1(s)t −4π(π−sin−1(s))t e − e , if t 6= 0 λ(s,t) = 4πt (I.17) −1 π − 2sin s, if t = 0.
−1 Since limt→0 λ(s,t) = π −2sin (s), we see that for all s ∈ (−1,1), the function λ(s,·) is contin- uous on R. n Recall from (I.10) that Bp = {w ∈ C | p(w) < 1}. Let Xp be the Hilbert space consisting of measurable functions f on R × Bp such that
Z Z k f k2 := | f (t, )|2 (p( ),t) dV( ) dt < , Xp ζ λ ζ ζ ∞ R Bp and such that ∂ f = 0 in the sense of distributions, 1 ≤ j ≤ n, ∂w j where w1,...,wn are co-ordinates of Bp. That is, Xp is the space of measurable functions on
R × Bp which are holomorphic in the variable ζ ∈ Bp and square integrable with respect to the weight (t,ζ) 7→ λ(p(ζ),t), where for ζ ∈ Bp, ! 1 −1 −1 exp −4πt sin p(ζ) − exp − 4πt π − sin p(ζ) , if t 6= 0 4πt λ (p(ζ),t) = −1 π − 2sin (p(ζ)), if t = 0. (I.18)
10 n Let γ ∈ C, and suppose that γ does not lie on the negative real axis. Let the map ρbγ : C → n C be given by
1/2m1 1/2mn n ρbγ (w1,...,wn) = γ w1,...,γ wn for all w ∈ C . (I.19) where the powers of γ are defined using a branch of the logarithm which coincides with the natural logarithm on the positive real line. Note that this extends the map (I.7) on page 8 to the complex plane.
2 Theorem 4. Let 1/µ = 1/m1 + ··· + 1/mn. For (z,w) ∈ Up, the map TV : Xp → A (Up) given by
Z zi2πt T g(z,w) = g t,ρ (w) dt (I.20) V b1/z 1+1/2µ R z Z w w zi2πt = g t, 1 ,··· , n dt 1/2m 1/2m 1+1/2µ R z 1 z n z is an isometric isomorphism of Hilbert spaces.
2 Let ρθ ∈ Aut(Up) be as in (I.8), for θ > 0. Then for each F ∈ A (Up), the map F 7→ ∗ 2 ∗ ρθ F determines the action of the scaling subgroup of Aut(Up) on A (Up), where ρθ F is defined analogous to (I.9). By Theorem 4, there is an f ∈ Xp such that F = TV f , where TV is as in (I.20).
Thus, for (z,w) ∈ Up we have
∗ 0 ρθ F(z,w) = F(ρθ (z,w)) · detρθ (z,w) Z (θz)i2πt = f t,ρ (ρ (w)) dt · θ 1+1/2µ b1/θz bθ 1+1/2µ R (θz) i2π(·) = TV θ f .
2 Thus, TV is a Fourier representation of A (Up) corresponding to the scaling subgroup of Aut(Up) as the isometric isomorphism TV in Theorem 4 diagonalizes the action of the scaling subgroup of
2 Aut(Up) on A (Up).
11 I.4.4. Bergman Kernel of Polynomial Half Spaces
Recall that associated to a weighted Bergman space A2(Ω,λ) is a function K : Ω×Ω → C such that for all f ∈ A2(Ω,λ), we have
Z f (z) = f (w)K(z,w)λ(w) dV(w) for all z ∈ Ω. Ω
The reproducing kernel K for the weighted Bergman space A2(Ω,λ) (see Proposition IV.1) is the unique function with this property that is holomorphic in the variable z and anti-holomorphic in the variable w. Also, recall that the reproducing kernel for A2(Ω) is called the Bergman kernel of Ω (see [15, Chapter 1]). The Bergman kernel of a domain Ω can sometimes be determined explicitly when Ω has a large group of symmetries, as in the case of the ball. For the polynomial half space Up, the asymptotic behavior of the Bergman kernel on the diagonal is determined by the symmetries of
Up, as we show in Theorem 5 below. This result is a special case of a more general theorem in [4].
Theorem 5. Let Up be a polynomial half space and for c > 0, let Γc be the subset of Up,
Γc = {(z,w) ∈ Up | |z| > c|w|}. (I.21)
Let K be the Bergman kernel for Up and for (z,w) ∈ Up, let
Kdiag(z,w) = K(z,w;z,w) be the Bergman kernel of Up on the diagonal of Up × Up. Then we have for each c > 0,
2+1/µ lim Kdiag(z,w)d(z,w) = Kdiag(i,0), (I.22) (z,w)→(0,0) (z,w)∈Γc
n where 1/µ = ∑ j=1 1/m j and d(z,w) is the distance of a point (z,w) ∈ Up to ∂Up.
Remark. The limit in the left hand side of equation (I.22) corresponds to the nontangential convergence of (z,w) ∈ Up to the point (0,0) ∈ ∂Up.
12 Recall from (I.12) that Yp admits the following direct integral representation
Z ⊕ Yp = Wp(k) dµ(k),
k+1 where for k ∈ N, Wp(k) is the weighted Bergman space on Bp with weight w 7→ (1 − p(w)) . 2 Using this together with the fact that A (Ep) is isometrically isomorphic to Yp allows us to express the Bergman kernel of Ep in terms of the reproducing kernels for Wp(k) below.
Theorem 6. For k ∈ N, let Yp(k;·,·) be the reproducing kernel for the weighted Bergman space n n Wp(k) on the domain Bp ⊂ C . Then for z,Z ∈ C and w,W ∈ C such that (z,w),(Z,W) ∈ Ep, the
Bergman kernel B of Ep is given by
∞ k + 1 k k B(z,w;Z,W) = ∑ Yp(k;w,W)z Z . k=0 π
n −4π p(w)t For t > 0, let Sp(t) be the weighted Bergman space on C with weight w 7→ e , i.e.,
2 n −4π pt Sp(t) = A C ,e . (I.23)
Then we have the following direct integral representation (see Proposition II.7, Section II.3) of
Hp Z ⊕ dt Hp = Sp(t) , (I.24) 4πt where the integral is taken with respect to the Haar measure dt/4πt on (0,∞). Then we may
2 represent the Bergman kernel for A (Up) in terms of reproducing kernel for Sp(t) as below.
Theorem 7. Let the Bergman kernel for Sp(t) be denoted by Hp(t;·,·). Then for (z,w),(Z,W) ∈
Up, the Bergman kernel K of Up is given by
Z ∞ i2π(z−Z)t K(z,w;Z,W) = 4π tHp(t;w,W)e dt (I.25) 0
13 For t ∈ R let Qp(t) be the weighted Bergman space on Bp with weight w 7→ λ(p(w),t), where λ is as in equation (I.18) i.e.,
2 Qp(t) = A (Bp,λ(p,t)). (I.26)
Then the Hilbert space Xp admits the following direct integral representation (see Proposition II.9, Section II.3) Z ⊕ Xp(t) = Qp(t) dt, (I.27)
2 where the integral is taken over R. Then we may represent the Bergman kernel for A (Up) in terms of reproducing kernel for Qp(t) as below.
Theorem 8. Let the reproducing kernel for Qp(t) be denoted by Xp(t;·,·). Then for (z,w),(Z,W) ∈
Up, the Bergman kernel K of Up is given by
Z z2πitZ−2πit K(z,w;Z,W) = X (t;w,W) dt. (I.28) p 1+1/2µ R (zZ)
I.5. Outline of this Thesis In Chapter 2, we recall basic facts about Bergman spaces and present proofs of proposi- tions that we will need to prove our main results in Chapter 3 and Chapter 4. We will also discuss geometric properties of polynomial ellipsoids and give a few examples. Finally, we discuss the direct integral representation of Hilbert spaces Hp and Xp. In Chapter 3, we present proofs of Theorems 2, 3, and 4. We first prove analogous results in one variable, and use these in the proofs of Theorems 2, 3, and 4. Finally in Chapter 4, we recall notions about Hilbert spaces that are necessary to obtain integral representations of Bergman kernel of Up as in Theorems 7 and 8, and complete the proofs of Theorems 6, 7 and 8.
14 I.6. Historical Note The classical Paley-Wiener theorem was obtained in [17] for the Hardy space of the upper
n half plane. A Fourier representation for Hardy space of tube domains in C was obtained by S. Bochner ([5]); tube domains are a different set of generalizations of the upper half plane in C to higher dimensions. A Fourier representation of Bergman space of tube domains over self dual cones was obtained by O. Rothaus ([20]) and this was generalized to Bergman space of tubes over more general cones by A. Koranyi ([13]). The Fourier type representation of Bergman space of the upper half plane with respect to the translation group has been rediscovered many times (see [10, 22]), most recently by P. Duren, et. al ([7]). A. Koranyi along with E. Stein also obtained a Fourier representation for Hardy spaces of generalized half spaces ([14]). R. Ogden and S. Vagi´ ([16]) obtained Fourier representations of functions in Hardy space of Siegel domains of type II, with respect to the translation group.
n For Bergman space of the Siegel upper half space Un ⊂ C , M. Peloso et al., ([1]) ob- tained Fourier representations of Bergman space of Un with respect to the Heisenberg group. N. Vasilevski et al., ([18]) obtained representations of Bergman space of Siegel upper half space
n 2 Un ⊂ C as L spaces, with respect to maximal abelian subgroups of Aut(Un). Their representa- tion however is not holomorphic in the complex parameter. Integral representation of the Bergman kernel of polynomial half space as in (I.25) was already obtained by F. Haslinger ([12]), by differentiating a similar formula for the Szego¨ kernel. It seems that Fourier representations of Bergman spaces of polynomial half spaces (even for the upper half plane) with respect to the scaling group have not been noted before.
15 CHAPTER II PRELIMINARIES In this chapter we state and prove familiar properties of the weighted Bergman space
n A2(Ω,λ) of a domain Ω ⊂ C . We first show that the weighted Bergman space A2(Ω,λ) of Ω 2 2 2 is a closed subspace of L (Ω,λ). Then we show that Bergman spaces A (Ω1) and A (Ω2) of n biholomorphically equivalent domains Ω1 and Ω2 in C are isometrically isomorphic. We also show that a polynomial half space Up is biholomorphically equivalent to a polynomial ellipsoid
Ep. Then we give a few examples of polynomial half spaces. Finally we obtain the direct integral representations (II.5) and (II.19) of the Hilbert spaces Hp and Xp introduced in Theorems 3 and 4 respectively.
II.1. Bergman Spaces
Proposition II.1 (Bergman Inequality). Let A2(Ω,λ) be the weighted Bergman space of Ω with weight λ. Then, for each compact set K ⊂ Ω, there is a constant CK > 0 such that
sup| f (z)| ≤ CK k f kA2(Ω,λ) . (II.1) z∈K
n n Proof. Define the number r > 0 in the following way. If Ω = C , let r = 1, and if Ω 6= C , 1 r = 2 (dist(K,∂Ω)). Let n Kr = {z ∈ C | dist(z,K) ≤ r}.
Then Kr ⊂ Ω and for all z ∈ K, B(z,r) ⊂ Kr. Here B(z,r) is the closed ball of radius r and center z. Then, by mean value property of harmonic functions applied to real and imaginary parts of f ∈ A2(Ω,λ), we have 1 Z f (z) = f (ζ) dV(ζ). Vol (B(z,r)) B(z,r) Consequently, we have
1 Z | f (z)| = f (ζ) dV(ζ) Vol (B(z,r)) B(z,r)
16 1 Z ≤ | f (z)| dV(ζ) Vol (B(z,r)) B(z,r) 1 Z dV(ζ)1/2 Z 1/2 ≤ | f (z)|2 λ(ζ) dV(ζ) Vol (B(z,r)) B(z,r) λ(ζ) B(z,r) 1/2 1 1 ≤ k f k 2 1/2 L (B(z,r),λ) (Vol B(z,r)) λ B(z,r) 1/2 1 1 ≤ k f k 2 1/2 A (Ω,λ) (Vol B(z,r)) λ Kr 1/2 1 1 = k f k 2 1/2 n A (Ω,λ) (Vol B(0,1)) r λ Kr
:= CK k f kA2(Ω,λ) .
Here C > 0 is finite, because K is compact and is strictly positive on so k1/ k , the K r λ Ω λ Kr minimum value of λ on Kr is a positive quantity.
n Corollary II.2. Let Ω be a domain in C . The weighted Bergman space A2(Ω,λ) of Ω with weight λ is a closed subspace of L2(Ω,λ).
∞ 2 2 Proof. Let { fn}n=1 be a sequence of functions in A (Ω,λ), which converges in L (Ω,λ). Then by ∞ Bergman inequality (II.1) it follows that the sequence of holomorphic functions { fn}n=1 converges uniformly on compact sets of Ω, and consequently the limiting function is holomorphic in Ω.
Thus, the limiting function is in A2(Ω,λ) which shows that A2(Ω,λ) is closed in L2(Ω,λ).
Lemma II.3 (Biholomorphic Invariance of Bergman Space). Let Φ : Ω1 → Ω2 be a biholomorphic n map between two domains Ω1 and Ω2 in C . The map Φ induces an isometric isomorphism ∗ 2 2 2 Φ : A (Ω2) → A (Ω1) between the Bergman spaces of Ω2 and Ω1, where for all F ∈ A (Ω2) and all z ∈ Ω1 we have (Φ∗F)(z) = F (Φ(z)) · detΦ0(z).
2 Proof. Let F ∈ A (Ω2). Then we have
Z 2 2 kFk 2 = |F(ζ)| dV(ζ). A (Ω2) Ω2 17 Applying the change of variables formula to the map Φ,
Z 2 2 0 2 kFk 2 = |F(Φ(z))| detΦ (z) dV(z) A (Ω2) Ω1 Z 0 2 = F(Φ(z))detΦ (z) dV(z) Ω1 0 2 = (F ◦ Φ) · detΦ 2 . A (Ω1)
2 0 2 ∗ This shows that F ∈ A (Ω2) if and only if (F ◦Φ)·detΦ ∈ A (Ω1). This also shows that Φ is an 2 isometry and injective. Let Ψ : Ω2 → Ω1 be the inverse of Φ. Given F ∈ A (Ω1), let G : Ω2 → C be given by
G(z) = F(Ψ(z)) · detΨ0(z).
Then we have
G ◦ Φ(z) · detΦ0(z) = F(Ψ(Φ(z))) · detΨ0(Φ(z)) · detΦ0(z) = F(z).
0 2 2 ∗ This shows that (G ◦ Φ) · detΦ ∈ A (Ω1). Thus, it follows that G ∈ A (Ω2) and Φ G = F. Con- sequently, Φ∗ is surjective and an isometric isomorphism as claimed.
II.2. Geometry of Polynomial Half Spaces
n Lemma II.4. Let γ ∈ C \{z ∈ C|Imz = 0,Rez ≤ 0}, and let p : C → R be a weighted homoge- n n neous balanced polynomial. Recall from (I.19) that ρbγ : C → C is given by 1/2m1 1/2mn n ρbγ (ζ1,...,ζn) = γ ζ1,...,γ ζn , for ζ ∈ C .
Then we have p(ρbγ (ζ)) = |γ| p(ζ).
Proof. The function L : C \{z ∈ C|Imz = 0,Rez ≤ 0} → C given by
L(γ) = log|γ| + iargγ, −π < argγ < π
18 is a well defined branch of logarithm, using which we define the powers γ1/2m1 ,...,γ1/2mn . Con- sequently, for a multi-index α = (α1,...,αn) we have
α1 αn α 1/2m1 1/2mn ρbγ (ζ) = γ ζ1 ... γ ζn
n ∑ j=1 α j/2m j α1 αn = γ ζ1 ···ζn
= γwtm(α)ζ α .
Thus, for a weighted homogeneous balanced polynomial p we have