and since

we get

Now do a substitution t = 2r. Then

16
Here Γ(s,x) is the incomplete gamma function [4]

Since ,

that if F ∈ Fα then we

However, from this expression it is not easy to draw any conclusions about the regularity of
ψ. With a little help from Stirlings formula we can unravel this regularity property.

Lemma 3.3.1.3.

Proof. By Stirling’s
formula [4] we have that

(3.3.7)

Using the fact that

the result follows after some algebraic manipulations.

Note that the notation ∼ is an asymptotic relation. If f(z) ∼ g(z),z → z0, then

We are now ready to give the main result which links the growth properties of F to the regularity of
ψ.

Theorem 3.3.1.4. Suppose

some ψ holomorphic in D− and ψ(∞) = 0, that is

17
If F ∈ Fα, then

(3.3.8)

that is {cnnα/2+1/4} ∈ `2(N) Consequently ψ is α/2 + 1/4-differentiable, i.e Dα/2+1/4 ∈ H2(D−).

From this we will have a Paley-Wiener type theorem that relates decay properties of the function F
with analytic properties of the Borel transform ψ. Setting α = −1/2 in (3.3.8) we have

which means the ψ belongs to the Hardy space H2(D−).

Theorem 3.3.1.5. If F ∈ F−1/2, then its Borel transform ψ belongs to H2(D−). Conversely, if ψ
belongs to H2(D−), then there exist a function F ∈ F−1/2 so that ψ is the Borel transform of F.

In a more abstract sense we can say that if B denotes the Borel transform operator, i.e BF = ψ, then
we have established that

BF−1/2 = H2(D−). (3.3.9)

In fact, we actually have a set of Paley-Wiener type theorems which relates the space Fα to the
holomorphic Sobolev spaces in D−.

3.3.2. Local Lipschitz regularity of the Borel transform

Now that we have seen that there is a relation between the decay properties of entire functions and
analytic properties of the corresponding Borel transform we want to make a more local result
concerning Lipschitz regularity of the Borel transform. The next two theorems relates the decay of
an entire function function and Lipschitz regularity of its Borel transform.

Theorem 3.3.2.1. Suppose ψ ∈ Lipα(1). If

18
x → +∞ along the real axis R

Proof. Since ψ ∈ Lipα(1) we have that there exist constants C,M > 0 and a polynomial
P1(ζ) of degree less than n = bαc, such that when |ζ −1| ≤ M, we have |ψ(ζ)−P1(ζ)| ≤ C|ζ − 1|α. Since

The latter integral is zero for k ≥ 0. To see this, observe that (ζ − 1)k is analytic inside the unit circle,
and hence by Cauchy’s theorem [2] the integral is zero. Hence, the integral on the left hand side is
also zero.

Since P1 is just a linear combination of

(ζ − 1)k with k ≥ 0, we see that

Thus, for z = x ∈ R we have

e−x F(x) =∫𝑇 (ψ(ζ) −

P1(ζ))e x(ζ−1) dζ.

Now we make the obvious parametrization of the circle, ζ = eiθ.

19
We take the absolute value, and use the Lipschitz condition

Using the estimate |eiθ − 1|α ≤ |θ|α for α > 0, we get

Consider the function ex(cosθ−1). Since cosθ − 1 < 0 for θ 6= 0 mod 2π we see that

ex(cosθ−1) → 0, as x → ∞,

when θ 6= 0 mod 2π. Consequently, if |θ| > δ for some δ > 0, then for all ε > 0 , we can find an M > 0
so that for x > M we have

Thus, for x > M

hence

(3.3.10)

Now if δ is small and |θ| ≤ δ we have by Taylors theorem that

and again by expanding the exponential in a Taylor series we obtain

Now the higher order terms in this series will only contribute to a constant factor in the integral
(3.3.10) since |θ| ≤ δ, and we can choose δ as small as we like.

Let C be a generic constant, that may change value from one line to the next. We have the following
estimate

20
The last equality follows from the fact that the integrand is even. Now for x > M we saw in (3.3.10)
that for large values of θ the integrand is approximately zero. Hence if we change the domain of
integration to a larger domain we will only add a constant factor to the integral, and we will not
change the asymptotic behaviour.

Now we make the change of variables, ω = θ2x/2,dω = θxdθ = (2ωx)1/2dθ. This gives

Hence as x → ∞ we have that

The next theorem is a converse to the previous one. The strategy used to prove this next theorem is
inspired by the proof of a theorem in [7] which relates Lipschitz regularity of a function to the decay
of its Wavelet transform. It is therefore reasons to expect that there are close relations between
these entire functions and the wavelet transform. First we need to have an expression for the
inverse Borel transform.

21
Lemma 3.3.2.2. Let F be and entire function and ψ the corresponding Borel transform,

Then

(3.3.11)

Proof. If

Theorem 3.3.2.3. If α > 0 is not an integer and there exist an α0 < α such that

for |ω| < β, for some β > 0. Then

if ψ is the Borel transform of F we have ψ(ζ) ∈ Lipα(1) for argζ < β.

Proof. We need to show that there exist a constant C > 0 and a polynomial P1(ζ) of with degP1 ≤ bαc
such that

|ψ(ζ) − P1(ζ)| ≤ C|ζ − 1|α,

whenever argζ < β.

The inverse Borel transform is given by (3.3.11)

Now we can decompose ψ(eiω) by dividing the interval of integration into dyadic intervals. That is,
let

(3.3.12)

Then

22
To prove that ψ is Lipschitz α at 1 we shall approximate ψ with the following polynomial

(3.3.13)

Here ∆k j (1) is the k0th derivative of ∆j(1) taken on the circle. It is easy to see that if ∆j(eiω) is given
by (3.3.12), then

(3.3.14)

Now, let us see if we can make an upper bound on the term ∆j(eiω). Let K be a generic constant that
may change value from one line to the next but that does not depend on j and ω.

Since |ω| < β we have, thus

Hence

(3.3.15)

Similarly we have

|∆j(1)| ≤ K2−jα. (3.3.16)

By same calculations we find that

(3.3.17)

and especially, we have that |∆k j (1)| ≤ K2−j(α−k). Our goal is to prove that

Pick J so that 2−J ≤ |eiω − 1| ≤ 2−J+1, and divide the sum in two

23
For j ≥ J we have

Since 2−J ≤ |eiω − 1| we have

For j < J we can use classical Taylor theorem to bound ∆j:

Inserting (3.3.17) gives

24
since 2−J ≤ |eiω − 1|. Altogether we have

for |ω| < β, which is want we wanted to prove. 

25