11.08.2013 Views

Composition Operators on the Fock Space - Mathematics - San ...

Composition Operators on the Fock Space - Mathematics - San ...

Composition Operators on the Fock Space - Mathematics - San ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

<str<strong>on</strong>g>Compositi<strong>on</strong></str<strong>on</strong>g> <str<strong>on</strong>g>Operators</str<strong>on</strong>g> <strong>on</strong> <strong>the</strong> <strong>Fock</strong> <strong>Space</strong><br />

Brent Carswell Barbara D. MacCluer Alex Schuster<br />

Abstract<br />

We determine <strong>the</strong> holomorphic mappings of C n that induce bounded compositi<strong>on</strong><br />

operators <strong>on</strong> <strong>the</strong> <strong>Fock</strong> space in C n . Fur<strong>the</strong>rmore, we determine which<br />

of <strong>the</strong>se compositi<strong>on</strong> operators are compact, and we compute <strong>the</strong> operator<br />

norm of all bounded compositi<strong>on</strong> operators in this setting. We also c<strong>on</strong>sider<br />

extensi<strong>on</strong>s of <strong>the</strong>se results in various generalizati<strong>on</strong>s of <strong>the</strong> <strong>Fock</strong> space.<br />

1 Introducti<strong>on</strong><br />

<str<strong>on</strong>g>Compositi<strong>on</strong></str<strong>on</strong>g> operators <strong>on</strong> spaces of analytic functi<strong>on</strong>s have been studied in<br />

many settings. Much has been written about <strong>the</strong> properties of <strong>the</strong>se operators<br />

<strong>on</strong> <strong>the</strong> Hardy, Bergman, and Bloch spaces <strong>on</strong> <strong>the</strong> unit disk in <strong>the</strong> complex<br />

plane, or unit ball in Cn (see, for example [3], [10] or [7]). Our purpose here is<br />

to study compositi<strong>on</strong> operators <strong>on</strong> <strong>the</strong> <strong>Fock</strong> space in Cn . We will determine<br />

which compositi<strong>on</strong> operators are bounded, and which are compact. We will<br />

also compute <strong>the</strong> norm of every bounded compositi<strong>on</strong> operator in this setting,<br />

an interesting result since finding <strong>the</strong> norm of such an operator <strong>on</strong> <strong>the</strong> Hardy<br />

and Bergman spaces is still largely an open problem (see [1], [2] or [4] for<br />

recent norm results in <strong>the</strong> Hardy space setting). We remark that while we<br />

will be doing our analysis in <strong>the</strong> several variable setting, we will make every<br />

effort to interpret our results and methods in <strong>the</strong> case n = 1 in order to<br />

clarify our ideas.<br />

The <strong>Fock</strong> space F 2 n is <strong>the</strong> Hilbert space of all holomorphic functi<strong>on</strong>s <strong>on</strong><br />

Cn with inner product<br />

〈f, g〉 ≡ 1<br />

(2π) n<br />

<br />

n<br />

1<br />

−<br />

f(z)g(z)e 2 |z|2<br />

dν(z).<br />

1


Here ν denotes Lebesgue measure <strong>on</strong> C n . To simplify notati<strong>on</strong> we will often<br />

use F instead of F 2 n, and we will denote by f <strong>the</strong> norm of f.<br />

The reproducing kernel functi<strong>on</strong>s for <strong>the</strong> <strong>Fock</strong> space are given by<br />

kw(z) = e 〈z,w〉/2 ,<br />

where 〈z, w〉 = n<br />

1 zjwj. Note that <strong>the</strong> substituti<strong>on</strong> f = kw into <strong>the</strong> reproducing<br />

formula 〈f, kw〉 = f(w), which holds for all f ∈ F and w ∈ C n , leads<br />

to <strong>the</strong> identity kw = exp(|w| 2 /4).<br />

For a given holomorphic mapping ϕ : C n → C n , <strong>the</strong> compositi<strong>on</strong> operator<br />

Cϕ : F → F is given by Cϕ(f) = f ◦ϕ. Our first main result (see Theorem 1)<br />

will show that if <strong>the</strong> operator Cϕ is bounded, <strong>the</strong>n ϕ must be of <strong>the</strong> form<br />

ϕ(z) = Az + B, where A is an n × n matrix and B is an n × 1 vector.<br />

Fur<strong>the</strong>rmore, it will follow that A ≤ 1 for bounded Cϕ, and that B will be<br />

restricted by <strong>the</strong> c<strong>on</strong>diti<strong>on</strong> that 〈Aζ, B〉 = 0 for any ζ in C n with |Aζ| = |ζ|.<br />

Theorem 1 will also show that if Cϕ is compact, <strong>the</strong>n A < 1 with no<br />

restricti<strong>on</strong> <strong>on</strong> B. Our sec<strong>on</strong>d result will be Theorem 2, which gives <strong>the</strong><br />

c<strong>on</strong>verse to Theorem 1. In our third main result, Theorem 4, we determine<br />

<strong>the</strong> operator norm of Cϕ (where ϕ(z) = Az + B is chosen so that Cϕ is<br />

bounded) in terms of A and B. Finally, in <strong>the</strong> last secti<strong>on</strong>, we will outline<br />

some ideas about how to extend our results to various generalizati<strong>on</strong>s of <strong>the</strong><br />

<strong>Fock</strong> space.<br />

We thank Christopher Hamm<strong>on</strong>d for several helpful remarks <strong>on</strong> <strong>the</strong> proof<br />

of Theorem 4.<br />

2 Main results<br />

Our first result c<strong>on</strong>cerns <strong>the</strong> boundedness and compactness of compositi<strong>on</strong><br />

operators <strong>on</strong> F.<br />

Theorem 1. Suppose ϕ : C n → C n is a holomorphic mapping.<br />

(a) If Cϕ is bounded <strong>on</strong> F, <strong>the</strong>n ϕ(z) = Az +B, where A is an n×n matrix<br />

and B is an n × 1 vector. Fur<strong>the</strong>rmore, A ≤ 1, and if |Aζ| = |ζ| for<br />

some ζ in C n , <strong>the</strong>n 〈Aζ, B〉 = 0.<br />

(b) If Cϕ is compact <strong>on</strong> F, <strong>the</strong>n ϕ(z) = Az + B, where A < 1.<br />

2


Note that when n = 1, Theorem 1(a) simply says that if Cϕ is bounded<br />

<strong>on</strong> F, <strong>the</strong>n ϕ(z) = az + b, where |a| ≤ 1, and if |a| = 1, <strong>the</strong>n b = 0.<br />

In <strong>the</strong> proof of Theorem 1 we will use <strong>the</strong> fact that every entire functi<strong>on</strong><br />

<strong>on</strong> Cn has a multi-index power series expansi<strong>on</strong> of <strong>the</strong> form <br />

α c(α)zα , where<br />

α denotes a multi-index (α1, α2, · · · , αn) of n<strong>on</strong>-negative integers, and z α =<br />

z α1<br />

1 z α2<br />

2 · · · z αn<br />

n . We use <strong>the</strong> standard notati<strong>on</strong> |α| = α1 + · · · + αn and α! =<br />

α1! · · · αn!. The quantity |α| is referred to as <strong>the</strong> total order of <strong>the</strong> multi-index<br />

α.<br />

Proof of Theorem 1. If Cϕ is bounded <strong>on</strong> F, <strong>the</strong>n<br />

sup<br />

w∈n C∗ ϕ(kw)<br />

kw<br />

= sup<br />

w∈n kϕ(w)<br />

kw<br />

= sup<br />

w∈n <br />

1 2 2<br />

exp |ϕ(w)| − |w|<br />

4<br />

<br />

< ∞, (1)<br />

where <strong>the</strong> first equality follows from <strong>the</strong> easily verified property C ∗ ϕ(kw) =<br />

kϕ(w). From (1) it follows that<br />

lim sup<br />

|w|→∞<br />

|ϕ(w)|<br />

|w|<br />

≤ 1. (2)<br />

For each ζ ∈ ∂Bn and 1 ≤ j ≤ n, we define <strong>the</strong> analytic functi<strong>on</strong> ϕ j<br />

ζ by <strong>the</strong><br />

equati<strong>on</strong> ϕ j<br />

ζ (λ) = ϕj(λζ) for λ ∈ C, where ϕj is <strong>the</strong> jth coordinate functi<strong>on</strong><br />

of ϕ. By (2) we must have<br />

lim sup<br />

|λ|→∞<br />

|ϕ j<br />

ζ (λ)|<br />

|λ|<br />

≤ 1.<br />

If ϕj(z) = <br />

α c(α)zα has homogeneous expansi<strong>on</strong> ∞ s=0 Fs(z) (defined by<br />

Fs(z) = <br />

|α|=s c(α)zα ), <strong>the</strong>n since ϕ j<br />

ζ (λ) = λsFs(ζ), we must have Fs(ζ) =<br />

0 for all s ≥ 2 and ζ ∈ ∂Bn; that is Fs ≡ 0 for s ≥ 2 and each coordinate<br />

functi<strong>on</strong> ϕj is linear. This proves that ϕ(z) = Az+B as desired. If |Aζ| > |ζ|<br />

for some ζ of norm 1, <strong>the</strong>n setting w = tζ, t > 0 in (2) and letting t → ∞ we<br />

obtain a c<strong>on</strong>tradicti<strong>on</strong>. Thus we must have A ≤ 1.<br />

Next we show that if |Aζ| = |ζ|, <strong>the</strong>n 〈Aζ, B〉 = 0. As a special case<br />

of this, suppose Aζ = λζ where λ is a complex number of modulus 1. If<br />

〈Aζ, B〉 = 0, we may choose ρ ∈ C, |ρ| = 1 so that ρ〈Aζ, B〉 > 0. C<strong>on</strong>sidering<br />

w = tρζ as t → ∞, we obtain a c<strong>on</strong>tradicti<strong>on</strong> to (1). Now suppose Aζ = η,<br />

where |ζ| = |η| = 1. Let U be a unitary map of Cn such that Uη = ζ. Then<br />

for τ(z) ≡ ϕ ◦ U(z) = A ◦ U(z) + B, we have Cτ bounded <strong>on</strong> F and, since<br />

3


A(U(η)) = η, by <strong>the</strong> special case just c<strong>on</strong>sidered, we have 〈AUη, B〉 = 0 or<br />

〈Aζ, B〉 = 0 as desired. This completes <strong>the</strong> proof of (a).<br />

To prove (b) we need <strong>on</strong>ly show that if ϕ(z) = Az + B induces a compact<br />

compositi<strong>on</strong> operator <strong>on</strong> F, <strong>the</strong>n A < 1. Since <strong>the</strong> normalized reproducing<br />

kernel functi<strong>on</strong>s kw/kw tend to 0 weakly as |w| → ∞, we have<br />

<br />

1<br />

Cϕe ≥ lim sup exp<br />

|w|→∞ 4 (|ϕ(w)|2 − |w| 2 <br />

) ,<br />

where for an operator T , T e = inf{T −Q : Q is compact} is <strong>the</strong> essential<br />

norm of T . (c.f. Propositi<strong>on</strong> 3.13 in [3] for <strong>the</strong> proof of a similar fact in<br />

o<strong>the</strong>r spaces). Now suppose that A = 1 and that ζ = 0 satisfies |Aζ| = |ζ|.<br />

Let w = λζ, where λ ∈ C and λ〈Aζ, B〉 ≥ 0. Letting |λ| → ∞, we see that<br />

Cϕe = 0, and so A is not compact. Thus A < 1.<br />

The c<strong>on</strong>verse to Theorem 1 holds as well.<br />

Theorem 2. Suppose ϕ(z) = Az + B, where A is an n × n matrix with<br />

A ≤ 1, and B is an n × 1 vector.<br />

(a) If 〈Aζ, B〉 = 0 whenever |Aζ| = |ζ|, <strong>the</strong>n Cϕ is bounded <strong>on</strong> F.<br />

(b) If A < 1, <strong>the</strong>n Cϕ is compact <strong>on</strong> F.<br />

Before we give <strong>the</strong> proof of this <strong>the</strong>orem, we need several preliminary<br />

results that will be of use throughout <strong>the</strong> rest of this secti<strong>on</strong>. First we recall<br />

<strong>the</strong> noti<strong>on</strong> of <strong>the</strong> singular value decompositi<strong>on</strong> of an n × n matrix.<br />

Theorem 3. If A is an n × n matrix of rank k, <strong>the</strong>n A can be written as<br />

A = V ΣW , where V, W are n × n unitary matrices, and Σ is a diag<strong>on</strong>al<br />

matrix (σij) with σ11 ≥ σ22 ≥ · · · ≥ σkk > σk+1,k+1 = · · · = σnn = 0. The<br />

σii are <strong>the</strong> n<strong>on</strong>-negative square roots of <strong>the</strong> eigenvalues of AA ∗ ; if we require<br />

that <strong>the</strong>y be listed in decreasing order, <strong>the</strong>n Σ is uniquely determined from A.<br />

A proof can be found in [6]. For brevity of notati<strong>on</strong> we will write σi for<br />

σii, <strong>the</strong> i th diag<strong>on</strong>al entry of Σ.<br />

Note that if A ≤ 1, <strong>the</strong>n <strong>the</strong> σi will all be less than or equal to 1, and<br />

at least <strong>on</strong>e will equal 1 if A = 1. We make two notati<strong>on</strong>al definiti<strong>on</strong>s that<br />

will be in force for <strong>the</strong> rest of this secti<strong>on</strong>. Set<br />

j = max{r : σr = 1}, (3)<br />

4


(where j = 0 in <strong>the</strong> case that no σr = 1), and<br />

k = max{r : σr > 0}, (4)<br />

so that k = rank A.<br />

The singular value decompositi<strong>on</strong> will allow us to perform a useful normalizati<strong>on</strong><br />

in proving Theorem 2. This normalizati<strong>on</strong> will also play a role<br />

later in this secti<strong>on</strong> when we compute <strong>the</strong> norm of Cϕ <strong>on</strong> F. We remark<br />

that in <strong>the</strong> <strong>on</strong>e variable setting this normalizati<strong>on</strong> is not needed, though its<br />

presence <strong>the</strong>re does no harm.<br />

Propositi<strong>on</strong> 1. Suppose ϕ(z) = Az +B, where A and B are as described in<br />

(a) of Theorem 1. Let ψ(z) = Σz+B ′ , where <strong>the</strong> singular value decompositi<strong>on</strong><br />

of A is V ΣW , and B ′ = V ∗ B. Then <strong>on</strong> F, Cϕ = CW CψCV , where CW and<br />

CV are <strong>the</strong> unitary operators given by CW (f) = f ◦ W and CV (f) = f ◦ V .<br />

Proof. We have<br />

CW CψCV (f)(z) = (f ◦ V ◦ ψ ◦ W )(z) = f(V ΣW (z) + V B ′ )<br />

= f(Az + B) = Cϕ(f)(z)<br />

as desired, where we have used <strong>the</strong> relati<strong>on</strong>ship V B ′ = B.<br />

Corollary 1. The operator Cϕ is bounded (resp., compact) <strong>on</strong> F if and <strong>on</strong>ly<br />

if Cψ is bounded (resp., compact) <strong>on</strong> F. Moreover, <strong>the</strong> norm of Cϕ is equal<br />

to <strong>the</strong> norm of Cψ.<br />

If ϕ(z) = Az + B and ψ(z) = Σz + B ′ , where Σ and B ′ are as given in<br />

Propositi<strong>on</strong> 1, we call ψ a normalizati<strong>on</strong> of ϕ. The following lemma gives<br />

<strong>the</strong> reformulati<strong>on</strong> of <strong>the</strong> hypo<strong>the</strong>sis “〈Aζ, B〉 = 0 whenever |Aζ| = |ζ|” from<br />

Theorem 2 as applied to <strong>the</strong> normalizati<strong>on</strong> ψ.<br />

Lemma 1. Suppose that ϕ(z) = Az + B satisfies <strong>the</strong> hypo<strong>the</strong>sis of (a) of<br />

Theorem 2, and suppose ψ = Σz + B ′ is a normalizati<strong>on</strong> of ϕ. Then <strong>the</strong> first<br />

j coordinates of B ′ are 0.<br />

Proof. Let A have singular value decompositi<strong>on</strong> A = V ΣW as in Theorem 3.<br />

Since |Aζ| = |ζ| implies 〈Aζ, B〉 = 0, we have<br />

0 = 〈ΣW ζ, V ∗ B〉 = 〈ΣW ζ, B ′ 〉 (5)<br />

5


whenever |Aζ| = |ζ|. If ζ is chosen so that W ζ = (0, · · · , 1, · · · , 0) t (where <strong>the</strong><br />

1 appears in <strong>the</strong> m th positi<strong>on</strong>, 1 ≤ m ≤ j), <strong>the</strong>n |Aζ| = |ζ| and ΣW ζ = W ζ,<br />

so that (5) implies that <strong>the</strong> m th coordinate of B ′ is 0.<br />

Proof of Theorem 2. By Corollary 1 and <strong>the</strong> remarks following it, to<br />

prove (a) we need <strong>on</strong>ly show that if ψ(z) = Σz + B ′ , where Σ is as described<br />

in Propositi<strong>on</strong> 1 and <strong>the</strong> first j coordinates of B ′ are 0 (where j is defined<br />

by (3)), <strong>the</strong>n Cψ is bounded <strong>on</strong> F. If Σ is invertible, this will follow from a<br />

change of variables argument. We have<br />

<br />

|f ◦ ψ(z)| 2 1<br />

−<br />

e 2 |z|2<br />

<br />

dν(z) = c |f(w)| 2 exp<br />

n<br />

= c ′<br />

<br />

n<br />

n<br />

<br />

− 1<br />

<br />

<br />

<br />

−1 ′<br />

Σ (w − B ) 2<br />

| dν<br />

2<br />

|f(w)| 2 <br />

exp − 1<br />

2 (|Σ−1w| 2 − |w| 2 − 2Re〈Σ −1 w, Σ −1 B ′ 〉)<br />

<br />

1<br />

−<br />

e 2 |w|2<br />

dν<br />

where c = (det Σ) −2 and c ′ = c exp(−|Σ −1 B ′ | 2 /2). It suffices to show that<br />

|Σ −1 w| 2 − |w| 2 − 2Re〈Σ −1 w, Σ −1 B ′ 〉 (6)<br />

is bounded away from −∞ as w ranges over C n . By Lemma 1 we see that<br />

<strong>the</strong> expressi<strong>on</strong> in (6) attains its minimum at points w ∈ C n satisfying<br />

|wm| = |b′ m|<br />

1 − σ 2 m<br />

and wmb ′ m ≥ 0<br />

for j + 1 ≤ m ≤ n. This gives <strong>the</strong> desired result.<br />

If Σ is not invertible, <strong>the</strong>n we have σm = 0 for k < m ≤ n, where k is<br />

as defined in (4). Let Σ1 = Σ and Σ2 be <strong>the</strong> diag<strong>on</strong>al matrix with Σ1 = Σ2<br />

except in <strong>the</strong> last n − k positi<strong>on</strong>s al<strong>on</strong>g <strong>the</strong> diag<strong>on</strong>al, where Σ1 has entries<br />

equal to 0 and Σ2 has entries equal to 1<br />

2<br />

. By <strong>the</strong> argument just finished for<br />

<strong>the</strong> invertible case, Cψ2 is bounded <strong>on</strong> F, where ψ2(z) = Σ2z + B ′ . Thus if<br />

f = c(α)z α is in F, so is f ◦ ψ2. Notice we can write f ◦ ψ2 as g + h, where<br />

g is<br />

<br />

α<br />

c(α)z α1<br />

1 · · · z αj<br />

j (σj+1zj+1 + b ′ j+1) αj+1 · · · (σkzk + b ′ k) αk (b ′ k+1) αk+1 · · · (b ′ n) αn<br />

and h = f ◦ ψ2 − g. The terms of h involve at least <strong>on</strong>e of <strong>the</strong> variables<br />

zk+1, · · · , zn, but <strong>the</strong> terms of g do not. The orthog<strong>on</strong>ality of <strong>the</strong> m<strong>on</strong>omials<br />

6


in F <strong>the</strong>n implies that f ◦ ψ2 ≥ g. But g = f ◦ ψ, and so f ◦ ψ ∈ F. By<br />

<strong>the</strong> closed graph <strong>the</strong>orem we c<strong>on</strong>clude that Cψ is bounded <strong>on</strong> F.<br />

Finally we prove (b). Again first suppose that Σ is invertible. To show<br />

Cψ is compact, it suffices to show that a sequence {fn} that is bounded in F<br />

and c<strong>on</strong>verges to 0 uniformly <strong>on</strong> compact subsets of C n has its image under<br />

Cψ c<strong>on</strong>verge to 0 in norm. Changing variables as above, we have<br />

fn ◦ ψ 2 <br />

= c<br />

n |fn(w)| 2 1<br />

−<br />

e 2 |Σ−1w−Σ−1B ′ | 2<br />

e 1<br />

2 |w|2<br />

1<br />

−<br />

e 2 |w|2<br />

dν(w).<br />

Since σi < 1 for all 1 ≤ i ≤ n, a calculati<strong>on</strong> shows that<br />

1<br />

−<br />

e 2 |Σ−1w−Σ−1B ′ | 2<br />

e 1<br />

2 |w|2<br />

is as small as desired off a compact subset of C n . Since fn c<strong>on</strong>verges to<br />

0 uniformly <strong>on</strong> compact sets, and fn is bounded, this guarantees that<br />

fn ◦ ψ → 0. In <strong>the</strong> case Σ is not invertible, we again compare Cψ to Cψ2,<br />

as defined above. Then fn ◦ ψ → 0, since fn ◦ ψ2 → 0 and fn ◦ ψ2 ≥<br />

fn ◦ ψ.<br />

We turn now to <strong>the</strong> issue of norm calculati<strong>on</strong>s.<br />

Theorem 4. Suppose ϕ(z) = Az + B, where ei<strong>the</strong>r A < 1 and B is<br />

arbitrary, or A = 1 and 〈Aζ, B〉 = 0 whenever |Aζ| = |ζ|. Then <strong>on</strong> F we<br />

have<br />

<br />

1 <br />

Cϕ = exp |w0|<br />

4<br />

2 − |Aw0| 2 + |B| 2<br />

, (7)<br />

where w0 is any soluti<strong>on</strong> to (I − A ∗ A)w = A ∗ B.<br />

Before turning to <strong>the</strong> proof of Theorem 4, we make several remarks. First<br />

note that Theorem 4 is trivial in <strong>the</strong> case that A is unitary, so henceforth we<br />

will assume A is not unitary.<br />

Next note that in <strong>on</strong>e variable, <strong>the</strong> equati<strong>on</strong> (1 − aa)w = ab has <strong>the</strong><br />

unique soluti<strong>on</strong> w = ab/(1 − |a| 2 ) if |a| < 1, so that<br />

<br />

1 |b|<br />

Cϕ = exp<br />

4<br />

2<br />

1 − |a| 2<br />

<br />

.<br />

7


In several variables it is natural to ask whe<strong>the</strong>r <strong>the</strong> equati<strong>on</strong> (I−A ∗ A)w =<br />

A ∗ B always has a soluti<strong>on</strong> under our hypo<strong>the</strong>sis <strong>on</strong> ϕ, and if so, whe<strong>the</strong>r <strong>the</strong><br />

soluti<strong>on</strong> uniquely determines <strong>the</strong> expressi<strong>on</strong> in (7). Indeed, since A ∗ B is<br />

orthog<strong>on</strong>al to <strong>the</strong> kernel of I − A ∗ A, a soluti<strong>on</strong> will always exist. To verify<br />

that <strong>the</strong> expressi<strong>on</strong> given in Equati<strong>on</strong> 7 is c<strong>on</strong>stant <strong>on</strong> <strong>the</strong> soluti<strong>on</strong> set of<br />

(I − A ∗ A)w = A ∗ B, note that if<br />

(I − A ∗ A)w0 = A ∗ B = (I − A ∗ A)w1<br />

<strong>the</strong>n w0 − w1 ∈ ker (I − A ∗ A), and thus 〈w0 − w1, A ∗ B〉 = 0, so that<br />

|w0| 2 −|Aw0| 2 = 〈w0, w0−A ∗ Aw0〉 = 〈w0, A ∗ B〉 = 〈w1, A ∗ B〉 = |w1| 2 −|Aw1| 2 .<br />

In order to prove Theorem 4, we will use a normalizati<strong>on</strong> of ϕ as we<br />

did in <strong>the</strong> proof of Theorem 2. Indeed, let ψ = Σz + B ′ where Σ =<br />

diag {σi} with σ1 = · · · = σj = 1 and σj+1, · · · , σn < 1, and with B ′ =<br />

(0, · · · , 0, b ′ j+1, · · · , b ′ n) t . We first observe that if (I − A ∗ A)w0 = A ∗ B and<br />

(I − Σ ∗ Σ)w = Σ ∗ B ′ , <strong>the</strong>n<br />

|w0| 2 − |Aw0| 2 + |B| 2 = |w| 2 − |Σw| 2 + |B ′ | 2 .<br />

This, toge<strong>the</strong>r with <strong>the</strong> fact that Cϕ = Cψ, implies that in order to prove<br />

Theorem 4 it suffices to prove it for <strong>the</strong> normalizati<strong>on</strong> ψ(z) = Σz + B ′ .<br />

We next observe that soluti<strong>on</strong>s to (I − Σ ∗ Σ)w = Σ ∗ B ′ are easily seen to<br />

be<br />

wm =<br />

Fur<strong>the</strong>rmore, for such w we have<br />

σmb ′ m/(1 − σ 2 m) for m ≥ j + 1<br />

arbitrary for m ≤ j.<br />

|w| 2 − |Σw| 2 + |B ′ | 2 =<br />

n<br />

m=j+1<br />

|b ′ m| 2<br />

1 − σ2 .<br />

m<br />

This implies that in order to prove Theorem 4 we need <strong>on</strong>ly show that<br />

′<br />

1 |b j+1|<br />

Cψ = exp<br />

4<br />

2<br />

1 − σ2 j+1<br />

+ · · · + |b′ n| 2<br />

1 − σ2 <br />

,<br />

n<br />

(8)<br />

which we will turn to shortly.<br />

The proof of Theorem 4 will require a representati<strong>on</strong> for <strong>the</strong> adjoint of<br />

Cϕ, which we obtain in <strong>the</strong> following result.<br />

8


Lemma 2. If ϕ(z) = Az + B where A, B satisfy <strong>the</strong> hypo<strong>the</strong>sis of Theorem<br />

1(a), so that Cϕ is bounded <strong>on</strong> F, <strong>the</strong>n C∗ ϕ = MkBCτ, where τ(z) = A∗z and<br />

is multiplicati<strong>on</strong> by <strong>the</strong> kernel functi<strong>on</strong> kB.<br />

MkB<br />

Proof. The proof follows by checking that C ∗ ϕ(kw) = exp(〈z, Aw + B〉/2) =<br />

MkB Cτ(kw) for all w ∈ C n . Since <strong>the</strong> span of <strong>the</strong> kernel functi<strong>on</strong>s is dense in<br />

F, <strong>the</strong> result follows.<br />

The next result gives a lower bound for Cψ, where ψ is a normalizati<strong>on</strong><br />

of ϕ.<br />

Lemma 3. Suppose ϕ(z) = Az + B satisfies <strong>the</strong> hypo<strong>the</strong>sis of Theorem 4<br />

and let ψ be a normalizati<strong>on</strong>. We have<br />

Cψ ≥ sup<br />

w∈n C∗ ψ (kw)<br />

′<br />

1 |b j+1|<br />

= exp<br />

kw 4<br />

2<br />

1 − σ2 + · · · +<br />

j+1<br />

|b′ n| 2<br />

1 − σ2 <br />

.<br />

n<br />

Proof. We have<br />

C∗ 2<br />

ψ (kw)<br />

kw2 2<br />

kψ(w) 1 2 2<br />

= = exp |ψ(w)| − |w|<br />

kw2 2<br />

<br />

.<br />

Since Σ = diag(σi) with σi = 1 for i ≤ j, where j defined is as in Equati<strong>on</strong> 3,<br />

it is easy to check that |ψ(w)| 2 − |w| 2 attains its maximum at points w ∈ C n<br />

which satisfy, for j + 1 ≤ m ≤ n,<br />

and<br />

|wm| = σm|b ′ m|<br />

1 − σ 2 m<br />

arg wm chosen so that wmb ′ m ≥ 0.<br />

Hence <strong>the</strong> maximum value of |ψ(w)| 2 − |w| 2 is <strong>the</strong>n<br />

|B ′ | 2 +<br />

This gives <strong>the</strong> desired result.<br />

n<br />

m=j+1<br />

σ 2 m|b ′ m| 2<br />

1 − σ 2 m<br />

=<br />

n<br />

m=j+1<br />

|b ′ m| 2<br />

1 − σ2 .<br />

m<br />

For <strong>the</strong>se normalized maps ψ we will compute Cψ by identifying a<br />

reducing subspace M for Cψ and computing Cψ|M and Cψ|N, where N<br />

is <strong>the</strong> orthog<strong>on</strong>al complement of M. The next lemma identifies this subspace<br />

M.<br />

9


Lemma 4. Let ψ(z) = Σz + B ′ be a normalizati<strong>on</strong> of ϕ, where ϕ satisfies<br />

<strong>the</strong> hypo<strong>the</strong>sis of Theorem 4. Define j, possibly 0, by Equati<strong>on</strong> (3). Let<br />

M = {f ∈ F : f depends <strong>on</strong> zj+1, zj+2, · · · , zn <strong>on</strong>ly}.<br />

Then M is a reducing subspace for Cψ and <strong>the</strong> restricti<strong>on</strong> of Cψ to M is<br />

compact.<br />

Proof. If f is in M, <strong>the</strong>n <strong>the</strong> power series expansi<strong>on</strong> of f about 0 has<br />

<strong>the</strong> form c(α)zα where <strong>the</strong> sum is over multi-indices α = (α1, · · · , αn)<br />

satisfying α1 = · · · = αj = 0. That Cψ(M) ⊂ M is <strong>the</strong>n immediate from <strong>the</strong><br />

form of Σ.<br />

To see that C∗ ψ (M) ⊂ M, let f ∈ M and use Lemma 2 to write<br />

C ∗ ψ(f)(z) = kB ′(z)f(Σz) = exp(〈z, B′ 〉/2)f(Σz).<br />

Since <strong>the</strong> first j coordinates of B ′ are 0, this is a functi<strong>on</strong> in M. Thus M is<br />

a reducing subspace for Cψ.<br />

To see that Cψ restricted to M is compact, c<strong>on</strong>sider <strong>the</strong> map ˜ ψ obtained<br />

from ψ by replacing any 1 ′ s al<strong>on</strong>g <strong>the</strong> diag<strong>on</strong>al of Σ by 1/2 ′ s. By Theorem<br />

2(b), C ˜ ψ is compact <strong>on</strong> F. Since Cψ and C ˜ ψ agree <strong>on</strong> <strong>the</strong> subspace M,<br />

and <strong>the</strong> restricti<strong>on</strong> of a compact operator to a closed subspace is compact,<br />

<strong>the</strong> result follows.<br />

Corollary 2. Let ψ and M be as in Lemma 4. Then<br />

<br />

1<br />

Cψ|M ≥ exp<br />

4<br />

n |b ′ m| 2<br />

1 − σ2 <br />

.<br />

m<br />

Proof. When w ∈ C N satisfies<br />

and<br />

m=j+1<br />

wm = 0 for 1 ≤ m ≤ j<br />

|wm| = σm|b ′ m|<br />

1 − σ 2 m<br />

for j + 1 ≤ m ≤ n<br />

wmb ′ m ≥ 0<br />

10


<strong>the</strong>n kw ∈ M, and by <strong>the</strong> calculati<strong>on</strong>s in <strong>the</strong> proof of Lemma 3 we have<br />

Cψ|M ≥ C∗ ψ (kw)<br />

<br />

n 1 |b<br />

= exp<br />

kw 4<br />

′ m| 2<br />

1 − σ2 <br />

.<br />

m<br />

m=j+1<br />

In <strong>the</strong> next lemma we compute <strong>the</strong> norm of Cψ restricted to M. We<br />

remark that in <strong>on</strong>e variable this lemma provides <strong>the</strong> proof of Theorem 4,<br />

since <strong>the</strong>n M = F if |a| < 1.<br />

Lemma 5. Let T be <strong>the</strong> restricti<strong>on</strong> of Cψ to M, for ψ and M as given in<br />

Lemma 4. Then <strong>the</strong> norm of T is given by<br />

′<br />

1 |b j+1|<br />

T = exp<br />

4<br />

2<br />

1 − σ2 + · · · +<br />

j+1<br />

|b′ n| 2<br />

1 − σ2 <br />

.<br />

n<br />

Proof. Since T T ∗ is a positive, compact, self-adjoint operator, T T ∗ =<br />

T 2 is an eigenvalue for T T ∗ and <strong>the</strong>re exists F ∈ M such that (T T ∗ )F =<br />

T 2 F . By Lemma 2 we have<br />

kB ′(Σz + B′ )F (Σ(Σz + B ′ )) = T 2 F (z). (9)<br />

Evaluate both sides of this equati<strong>on</strong> at any point z0 with (I −Σ ∗ Σ)z0 = Σ ∗ B ′ .<br />

If F (z0) = 0 we obtain immediately that T 2 = kB ′(Σz0 + B ′ ). Since <strong>the</strong><br />

c<strong>on</strong>diti<strong>on</strong> (I − Σ ∗ Σ)z0 = Σ ∗ B ′ implies that (1 − σ 2 i )(z0)i = σib ′ i for 1 ≤ i ≤ n<br />

and b ′ i = 0 for 1 ≤ i ≤ j we obtain<br />

T 2 = exp<br />

<br />

1<br />

2<br />

n<br />

m=j+1<br />

|b ′ m| 2<br />

1 − σ 2 m<br />

as desired.<br />

Thus it just remains to handle <strong>the</strong> case that F (z0) = 0. For motivati<strong>on</strong><br />

we first show how to handle this case in <strong>the</strong> <strong>on</strong>e variable setting, with ψ(z) =<br />

σz + b ′ . If F has a zero of order k at z0, <strong>the</strong>n c<strong>on</strong>sider <strong>the</strong> equati<strong>on</strong><br />

F (a(az + b))<br />

kb(ψ(z))<br />

F (z)<br />

<br />

= Cψ 2<br />

and take limits as z → z0. Using L’Hopitals’ rule we get<br />

<br />

exp<br />

|σ| 2k = Cψ 2 .<br />

|b ′ | 2<br />

2(1 − |σ| 2 )<br />

11


This cannot be correct, since by Lemma 3<br />

<br />

Cψ ≥ exp<br />

|b ′ | 2<br />

4(1 − |σ| 2 )<br />

and |σ| < 1.<br />

For F a functi<strong>on</strong> of several complex variables, we say F has a zero of<br />

order k at z = z0 if in <strong>the</strong> power series expansi<strong>on</strong> for F about z0, F (z) =<br />

<br />

α c(α)(z − z0) α , <strong>the</strong>re is a multi-index α of total order k such that c(α) =<br />

0, but for all multi-indices of order less than k, c(α) = 0. Write z0 =<br />

(z 1 0, z 2 0, · · · , z n 0 ) and c<strong>on</strong>sider z ′ s of <strong>the</strong> form (z 1 0 + γζ1, z 2 0 + γζ2, · · · , z n 0 + γζn)<br />

where (ζ1, ζ2, · · · , ζn) is a fixed point of ∂Bn and γ varies over <strong>the</strong> complex<br />

plane. Substitute such z ′ s into <strong>the</strong> equati<strong>on</strong><br />

kB ′(Σz + B′ ) F (Σ(Σz + B′ ))<br />

F (z)<br />

<br />

= T 2<br />

and let γ → 0. L’Hopital’s rule (differentiating k times with respect to γ)<br />

yields<br />

kB ′(Σz0 + B ′ <br />

|α|=k<br />

)<br />

c(α)σ2α1 1 · · · σ2αn n ζ α1<br />

1 · · · ζαn n<br />

= T 2<br />

(10)<br />

<br />

|α|=k c(α)ζα1 1 · · · ζαn n<br />

for any choice of (ζ1, · · · , ζn) for which <strong>the</strong> denominator in this expressi<strong>on</strong><br />

is not 0. The quotient <strong>on</strong> <strong>the</strong> left side of Equati<strong>on</strong> (10) must be some positive<br />

c<strong>on</strong>stant, say µ, since T 2 and kB ′(Σz0 + B ′ ) are positive. Thinking of<br />

ζ1, · · · , ζn as variables for <strong>the</strong> moment, this says <strong>the</strong> homogeneous polynomi-<br />

als <br />

and<br />

|α|=k<br />

c(α)σ 2α1<br />

1<br />

µ <br />

|α|=k<br />

· · · σ 2αn<br />

n ζ α1<br />

1 · · · ζ αn<br />

n<br />

c(α)ζ α1<br />

1 · · · ζ αn<br />

n<br />

must agree and hence have <strong>the</strong> same coefficients. Thus<br />

c(α)σ 2α1<br />

1<br />

· · · σ 2αn<br />

n<br />

= µc(α)<br />

for every multi-index α of total order k. Since for some such α, c(α) = 0,<br />

= µ; this forces µ to be at most 1. But µ < 1<br />

we must have σ 2α1<br />

1<br />

· · · σ 2αn<br />

n<br />

12


yields a c<strong>on</strong>tradicti<strong>on</strong> to Corollary 2, so we are left with µ = 1. Hence<br />

kB ′(Σz0 + B ′ ) = T 2 , as desired.<br />

The final matter before we complete <strong>the</strong> proof of Theorem 4 in <strong>the</strong> several<br />

variable case is to estimate <strong>the</strong> norm of Cψ restricted to <strong>the</strong> orthog<strong>on</strong>al complement<br />

of our reducing subspace M. We address this issue next. Without<br />

loss of generality we may assume 1 ≤ j < n, where n is <strong>the</strong> dimensi<strong>on</strong> and j<br />

is defined by Equati<strong>on</strong> (3).<br />

It is a straightforward calculati<strong>on</strong> (see, e.g., Secti<strong>on</strong> 1.4.9 of [9], where<br />

a slightly different normalizati<strong>on</strong> is used) to see that for a multi-index α,<br />

z α 2 = 2 |α| α!. In particular<br />

z k1<br />

1 · · · z kj<br />

j zkj+1 j+1 · · · zkn n 2 = 2 k1+···+kjk1! · · · kj!z kj+1<br />

j+1 · · · zkn n 2 .<br />

Lemma 6. If f ∈ F depends <strong>on</strong> zj+1, · · · , zn <strong>on</strong>ly <strong>the</strong>n<br />

z k1<br />

1 · · · z kj<br />

j f2 = 2 k1+···+kj k1! · · · kj!f 2<br />

for any n<strong>on</strong>-negative integers k1, · · · , kj.<br />

Proof. Write f = c(α)z α where <strong>the</strong> sum is over multi-indices α with<br />

α1 = α2 = · · · = αj = 0. Then<br />

z k1<br />

1 · · · z kj<br />

j f2 = <br />

<br />

α=(0,··· ,0,αj+1··· ,αn)<br />

= |c(α)| 2 z k1<br />

1 · · · z kj<br />

j zα 2<br />

z k1<br />

1 · · · z kj<br />

j c(α)zα 2<br />

= 2 k1+···+kj k1! · · · kj! |c(α)| 2 2 |α| αj+1! · · · αn!<br />

= 2 k1+···+kj k1! · · · kj!f 2<br />

We next obtain an estimate <strong>on</strong> <strong>the</strong> norm of Cψ restricted to <strong>the</strong> orthog<strong>on</strong>al<br />

complement of <strong>the</strong> subspace M. To do this we will take a functi<strong>on</strong> g in M ⊥<br />

and write Cψ(g) as a sum of terms each of which is a functi<strong>on</strong> in Cψ multiplied<br />

by a m<strong>on</strong>omial of <strong>the</strong> form z α1<br />

1 · · · z αj<br />

j . The desired estimate will <strong>the</strong>n follow<br />

by orthog<strong>on</strong>ality and Lemma 6.<br />

Propositi<strong>on</strong> 2. Suppose ψ(z) = Σz+B ′ and M are as described in Lemma 4.<br />

Let g ∈ M ⊥ . Then g ◦ ψ 2 ≤ T 2 g 2 , where T denotes <strong>the</strong> restricti<strong>on</strong> of<br />

Cψ to M.<br />

13


Proof. Write g as a sum of pairwise orthog<strong>on</strong>al functi<strong>on</strong>s of <strong>the</strong> form zα′ gα ′<br />

where α ′ = (α1, · · · , αj, 0 ′ ) with α1, · · · , αj not all zero and gα ′ is a functi<strong>on</strong><br />

depending <strong>on</strong> zj+1, · · · , zn <strong>on</strong>ly. This is accomplished by writing g is<br />

its power series expansi<strong>on</strong> about 0 and grouping toge<strong>the</strong>r all terms of <strong>the</strong><br />

form c(α)z α1<br />

1 · · · z αj<br />

j (z′′ ) β , where α1, · · · , αj are fixed and not all zero and z ′′<br />

denotes (zj+1, · · · , zn). Note that<br />

〈z α′<br />

gα ′, z ˜ α ′<br />

gα ˜′〉 = 0<br />

if α ′ = ˜α ′ . Using Lemma 6 this orthog<strong>on</strong>ality implies<br />

g 2 = z α′<br />

gα ′2 = 2 |α′ | α1! · · · αj!gα ′2<br />

where α ′ = (α1, · · · , αj, 0 ′ ). We have g ◦ ψ = zα′ gα ′ ◦ ψ and, by <strong>the</strong><br />

form of ψ, <strong>the</strong> terms of this sum are pairwise orthog<strong>on</strong>al as α ′ ranges over<br />

multi-indices of <strong>the</strong> form α ′ = (α1, · · · , αj, 0 ′ ). Thus<br />

g ◦ ψ 2 = z α′<br />

gα ′ ◦ ψ2<br />

= 2 |α′ |<br />

α1! · · · αj! gα ′ ◦ ψ2<br />

≤ 2 |α′ | α1! · · · αj! T 2 gα ′2 = T 2 g 2<br />

where <strong>the</strong> inequality holds since gα ′ is in M.<br />

Finally, we combine <strong>the</strong> above results to obtain Theorem 4.<br />

Proof of Theorem 4. As previously discussed (see Equati<strong>on</strong> (8), it<br />

suffices to show<br />

Cψ = exp<br />

′<br />

1 |b j+1|<br />

4<br />

2<br />

1 − σ2 + · · · +<br />

j+1<br />

|b′ n| 2<br />

1 − σ2 <br />

n<br />

(11)<br />

where ψ is a normalizati<strong>on</strong> of ϕ. By Lemma 5 we know that <strong>the</strong> expressi<strong>on</strong><br />

<strong>on</strong> <strong>the</strong> right hand side of Equati<strong>on</strong> (11) is <strong>the</strong> norm of <strong>the</strong> restricti<strong>on</strong> of Cψ to<br />

<strong>the</strong> reducing subspace M, and by Propositi<strong>on</strong> 2 <strong>the</strong> norm of <strong>the</strong> restricti<strong>on</strong><br />

of Cψ to <strong>the</strong> orthog<strong>on</strong>al complement of M is bounded above by this same<br />

expressi<strong>on</strong>. This gives <strong>the</strong> result.<br />

14


3 C<strong>on</strong>cluding remarks<br />

In this final secti<strong>on</strong> we c<strong>on</strong>sider <strong>the</strong> possibility of extending <strong>the</strong> results of <strong>the</strong><br />

previous secti<strong>on</strong> to various generalizati<strong>on</strong>s of F. For simplicity we restrict<br />

to <strong>the</strong> case n = 1. First c<strong>on</strong>sider <strong>the</strong>, in general, n<strong>on</strong>-Hilbert <strong>Fock</strong> space F p<br />

defined, for 0 < p < ∞, to be <strong>the</strong> space of all entire functi<strong>on</strong>s f <strong>on</strong> C for<br />

which <strong>the</strong> norm<br />

f p p ≡ 1<br />

<br />

|f(z)|<br />

2π C<br />

p 1<br />

−<br />

e 2 |z|2<br />

dA(z) < ∞<br />

where dA denotes Lebesgue measure in C.<br />

The boundedness and compactness of Cϕ can be determined by <strong>the</strong> Carles<strong>on</strong><br />

properties of a certain measure associated with ϕ. In particular, recall<br />

that a finite measure µ <strong>on</strong> C is a (p, q)-Carles<strong>on</strong> measure if <strong>the</strong> space F p is<br />

a subset of L q (µ). By <strong>the</strong> closed graph <strong>the</strong>orem, this is equivalent to <strong>the</strong><br />

existence of a positive c<strong>on</strong>stant C such that<br />

fL q (µ) ≤ Cfp<br />

(12)<br />

for all f ∈ F p . The inequality (12) is equivalent to <strong>the</strong> boundedness of <strong>the</strong><br />

inclusi<strong>on</strong> map i : F p → Lq (µ). If <strong>the</strong> map i is compact, we say that µ is a<br />

vanishing (p, q)-Carles<strong>on</strong> measure. (A linear map is compact if <strong>the</strong> image of<br />

<strong>the</strong> closed unit ball has compact closure.)<br />

For a given entire functi<strong>on</strong> ϕ, <strong>the</strong> weighted pullback measure µϕ <strong>on</strong> C is<br />

given by<br />

<br />

µϕ(E) =<br />

ϕ −1 (E)<br />

1<br />

−<br />

e 2 |z|2<br />

dA(z)<br />

for every Borel subset E of C. Since <strong>on</strong>e can easily see that Cϕ(f)q =<br />

fL q (µϕ) for every f in F p , <strong>the</strong> operator Cϕ : F p → F q is bounded (resp.,<br />

compact) if and <strong>on</strong>ly if <strong>the</strong> pullback measure µϕ is (p, q)-Carles<strong>on</strong> (resp.,<br />

vanishing (p, q)-Carles<strong>on</strong>).<br />

The next two results give a characterizati<strong>on</strong> of <strong>the</strong> Carles<strong>on</strong> and <strong>the</strong><br />

vanishing Carles<strong>on</strong> measures. We will not give <strong>the</strong> proofs of <strong>the</strong>se results since<br />

<strong>the</strong>y are slight improvements of results c<strong>on</strong>tained in a paper of Ortega [8].<br />

Here ∆(ζ, r) denotes <strong>the</strong> Euclidean disk with center ζ and radius r.<br />

Theorem 5. Let 0 < p ≤ q < ∞. The following statements are equivalent:<br />

(i) The measure µ is a (p, q)-Carles<strong>on</strong> measure.<br />

15


(ii) For every r > 0 <strong>the</strong>re is a c<strong>on</strong>stant C for which<br />

<br />

∆(ζ,r)<br />

e αq|z|2<br />

p dµ(z) ≤ C (13)<br />

for all ζ ∈ C.<br />

(iii) There exists r > 0 and a c<strong>on</strong>stant C for which (13) holds for all<br />

ζ ∈ C.<br />

For vanishing Carles<strong>on</strong> measures, we have <strong>the</strong> following.<br />

Theorem 6. Let 0 < p ≤ q < ∞. The following statements are equivalent:<br />

(i) The measure µ is a vanishing (p, q)-Carles<strong>on</strong> measure.<br />

(ii) For every r > 0<br />

<br />

∆(ζ,r)<br />

as |ζ| → ∞.<br />

(iii) There exists r > 0 for which (14) holds.<br />

e αq|z|2<br />

p dµ(z) → 0 (14)<br />

From <strong>the</strong> previous results, we can easily obtain Propositi<strong>on</strong> 3 by replacing<br />

<strong>the</strong> measure µ with <strong>the</strong> measure µϕ.<br />

Propositi<strong>on</strong> 3. C<strong>on</strong>sider <strong>the</strong> operator Cϕ : F p → F q for 0 < p ≤ q < ∞.<br />

(a) If Cϕ is bounded, <strong>the</strong>n ϕ(z) = az + b. Moreover,<br />

<br />

q<br />

|a| = 1, <strong>the</strong>n b = 0.<br />

p<br />

<br />

q<br />

(b) If Cϕ is compact, <strong>the</strong>n |a| < 1.<br />

p<br />

C<strong>on</strong>versely, suppose that ϕ(z) = az + b.<br />

q<br />

(c) If p |a| = 1 and b = 0, <strong>the</strong>n Cϕ is bounded.<br />

<br />

q<br />

(d) If p |a| < 1, <strong>the</strong>n Cϕ is compact.<br />

<br />

q<br />

|a| ≤ 1, and if<br />

p<br />

It is natural to ask whe<strong>the</strong>r <strong>on</strong>e can compute <strong>the</strong> operator norm Cϕ :<br />

F p → F q for 0 < p ≤ q < ∞ when at least <strong>on</strong>e of p and q is different than<br />

2. When p = q an obvious c<strong>on</strong>jecture for this norm is<br />

exp<br />

1<br />

2q<br />

|b| 2<br />

1 − |a| 2<br />

16


when |a| < 1. When p < q a lower bound for <strong>the</strong> norm is<br />

1<br />

2π<br />

1/p−1/q<br />

exp<br />

1<br />

2q<br />

(q/p)|b| 2<br />

1 − (q/p)|a| 2<br />

These results follow an argument similar to that in Lemma 3, using <strong>the</strong> point<br />

evaluati<strong>on</strong> functi<strong>on</strong>als <strong>on</strong> F p and F q .<br />

Returning to a Hilbert space setting, <strong>on</strong>e might also c<strong>on</strong>sider generalizati<strong>on</strong>s<br />

to <strong>the</strong> <strong>Fock</strong> space obtained by defining FW to c<strong>on</strong>sist of those entire<br />

functi<strong>on</strong>s <strong>on</strong> C satisfying<br />

<br />

|f(z)| 2 exp(−W (z))dA(z) < ∞<br />

C<br />

for some positive weight functi<strong>on</strong> W (z) = W (|z|) satisfying appropriate hypo<strong>the</strong>ses.<br />

To <strong>the</strong> extent that <strong>on</strong>e can obtain good asymptotic informati<strong>on</strong><br />

about kw, as |w| → ∞, (where, as usual, kw denotes <strong>the</strong> reproducing kernel<br />

functi<strong>on</strong> for evaluati<strong>on</strong> at w in FW ) <strong>the</strong> methods of Theorem 1 will apply.<br />

For example, if W (z) = |z| 2c , c > 0, <strong>the</strong>n a <strong>the</strong>orem of Folland and Rochberg<br />

(see, e.g. Theorem 4.6 in [?]) shows that<br />

<br />

.<br />

kw 2 = exp(|w| 2c )|w| 2c−2 c 2 (1 + O(c 2 |w| 2c ) −γ )<br />

for some γ > 0, as |w| → ∞. From this it follows, as in <strong>the</strong> proof of<br />

Theorem 1, that if Cϕ is bounded <strong>on</strong> FW , <strong>the</strong>n<br />

lim sup<br />

|z|→∞<br />

|ϕ(z)|<br />

|z|<br />

and hence that ϕ must be linear.<br />

The analogue of Theorem 2 also holds for W (z) = exp(−|z| 2c ), c > 0 with<br />

a similar proof, involving a change of variable argument, and elementary<br />

estimates <strong>on</strong><br />

|w| 2c (1 − |a| −2c |1 − b<br />

w |2c )<br />

for |w| large.<br />

It seems unlikely that a precise calculati<strong>on</strong> of Cϕ can be made in this<br />

setting, however it would be reas<strong>on</strong>able to expect that<br />

CϕFW<br />

≤ 1<br />

C<br />

= sup<br />

w∈C<br />

∗ ϕ(kw)<br />

kw<br />

might c<strong>on</strong>tinue to hold when W (z) = exp(−|z| 2c ), c > 0.<br />

17


References<br />

[1] M. Appel, P. Bourd<strong>on</strong>, and J. Thrall, “Norms of compositi<strong>on</strong> operators<br />

<strong>on</strong> <strong>the</strong> Hardy space”, Experiment. Math., 5 (1996), 111-117.<br />

[2] P. Bourd<strong>on</strong> and D. Retsek, “Reproducing kernels and norms of compositi<strong>on</strong><br />

operators”, Acta Sci. Math. (Szeged), 67, (2001), 387-394.<br />

[3] C. Cowen and B. MacCluer, <str<strong>on</strong>g>Compositi<strong>on</strong></str<strong>on</strong>g> <str<strong>on</strong>g>Operators</str<strong>on</strong>g> <strong>on</strong> <strong>Space</strong>s of Analytic<br />

Functi<strong>on</strong>s, CRC Press, Boca Rat<strong>on</strong>, FL, 1995.<br />

[4] C. Hamm<strong>on</strong>d, On <strong>the</strong> norm of a compositi<strong>on</strong> operator with linear fracti<strong>on</strong>al<br />

symbol, preprint.<br />

[5] F. Holland and R. Rochberg, “Bergman kernels and Hankel forms <strong>on</strong><br />

generalized <strong>Fock</strong> spaces”, C<strong>on</strong>temp. Math. 232, (1999), 189-200.<br />

[6] R. Horn and C. Johns<strong>on</strong>, Matrix Analysis, Cambridge University Press,<br />

Cambridge, 1990.<br />

[7] F. Jafari, B. MacCluer, C. Cowen and D. Porter, eds. Studies <strong>on</strong> <str<strong>on</strong>g>Compositi<strong>on</strong></str<strong>on</strong>g><br />

<str<strong>on</strong>g>Operators</str<strong>on</strong>g>, C<strong>on</strong>temp. Math. 213, Amer. Math. Soc., 1998.<br />

[8] J. Ortega-Cerdà, “Sampling Measures,” Publ. Mat. 42 (1998), no. 2,<br />

559-566.<br />

[9] W. Rudin, Functi<strong>on</strong> Theory in <strong>the</strong> Unit Ball of C n , Springer-Verlag,<br />

New York, 1980.<br />

[10] J. Shapiro, <str<strong>on</strong>g>Compositi<strong>on</strong></str<strong>on</strong>g> <str<strong>on</strong>g>Operators</str<strong>on</strong>g> and Classical Functi<strong>on</strong> Theory,<br />

Springer, New York, 1993.<br />

The University of Michigan<br />

Department of Ma<strong>the</strong>matics<br />

East Hall, 525 East University Ave<br />

Ann Arbor, MI 48109-1109<br />

Email: carswell@umich.edu<br />

Department of Ma<strong>the</strong>matics<br />

PO Box 400137 Kerchof Hall<br />

University of Virginia<br />

18


Charlottesville, VA 22904-4137<br />

Email: bdm3f@weyl.math.virginia.edu<br />

<strong>San</strong> Francisco State University<br />

Department of Ma<strong>the</strong>matics<br />

<strong>San</strong> Francisco, CA 94132<br />

Email: schuster@sfsu.edu<br />

19

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!