Abstract
We establish wavelet characterizations of homogeneous Besov spaces on stratified Lie groups, both in terms of continuous and discrete wavelet systems. We first introduce a notion of homogeneous Besov space in terms of a Littlewood-Paley-type decomposition, in analogy to the well-known characterization of the Euclidean case. Such decompositions can be defined via the spectral measure of a suitably chosen sub-Laplacian. We prove that the scale of Besov spaces is independent of the precise choice of Littlewood-Paley decomposition. In particular, different sub-Laplacians yield the same Besov spaces. We then turn to wavelet characterizations, first via continuous wavelet transforms (which can be viewed as continuous-scale Littlewood-Paley decompositions), then via discretely indexed systems. We prove the existence of wavelet frames and associated atomic decomposition formulas for all homogeneous Besov spaces with and .
1. Introduction
To a large extent, the success of wavelets in applications can be attributed to the realization that wavelet bases are universal unconditional bases for a large class of smoothness spaces, including all homogeneous Besov spaces. Given a wavelet orthonormal basis (consisting of sufficiently regular wavelets with vanishing moments) and , the expansion converges not only in , but also in any other Besov space norm , as soon as is contained in that space. Furthermore, the latter condition can be read off the decay behaviour of the wavelet coefficients associated to in a straightforward manner.
This observation provided important background and heuristics for many wavelet-based methods in applications such as denoising and data compression, but it was also of considerable theoretical interest, for example, for the study of operators. In this paper we provide similar results for simply connected stratified Lie groups. To our knowledge, studies of Besov spaces in this context have been largely restricted to the inhomogeneous cases. The definition of inhomogeneous Besov spaces on stratified Lie groups was introduced independently by Saka [1], and in a somewhat more general setting by Pesenson [2, 3]. Since then, the study of Besov spaces on Lie groups remained restricted to the inhomogeneous cases [4–8], with the notable exception of [9] which studied homogeneous Besov spaces on the Heisenberg group. A further highly influential source for the study of function spaces associated to the sub-Laplacian is Folland’s paper [10].
The first wavelet systems on stratified Lie groups (fulfilling certain technical assumptions) were constructed by Lemarié [11], by suitably adapting concepts from spline theory. Lemarié also indicated that the wavelet systems constructed by his approach were indeed unconditional bases of Saka’s inhomogeneous Besov spaces. Note that an adaptation, say, of the arguments in [12] for a proof of such a characterization requires a sampling theory for bandlimited functions on stratified groups, which was established only a few years ago by Pesenson [13]; see also [14].
More recent constructions of both continuous and discrete wavelet systems were based on the spectral theory of the sub-Laplacian [15]. Given the central role of the sub-Laplacian both in [8, 15], and in view of Lemarié’s remarks, it seemed quite natural to expect a wavelet characterization of homogeneous Besov spaces, and it is the aim of this paper to work out the necessary details. New results in this direction were recently published in [16–18].
The paper is structured as follows. After reviewing the basic notions concerning stratified Lie groups and their associated sub-Laplacians in Section 2, in Section 3 we introduce a Littlewood-Paley-type decomposition of functions and tempered discributions on . It is customary to employ the spectral calculus of a suitable sub-Laplacian for the definition of such decompositions, see, for example, [8], and this approach is also used here (Lemma 3.7). However, this raises the issue of consistency: the spaces should reflect properties of the group, not of the sub-Laplacian used for the construction of the decomposition. Using a somewhat more general notion than the -functions in [12] allows to establish that different choices of sub-Laplacian result in the same scale of Besov spaces (Theorem 3.11). In Section 4, we derive a characterization of Besov spaces in terms of continuous wavelet transform, with a wide variety of wavelets to choose from (Theorem 4.4). As a special case one obtains a characterization of homogeneous Besov spaces in terms of the heat semigroup. (See the remarks before Theorem 4.4.)
In Section 5, we study discrete characterizations of Besov spaces obtained by sampling the Calderón decomposition. For this purpose, we introduce the coefficient space . The chief result is Theorem 5.4, establishing that the wavelet coefficient sequence of lies in . Section 5 introduces our most important tool to bridge the gap between continuous and discrete decompositions, namely, oscillation estimates.
We then proceed to study wavelet synthesis and frame properties of the wavelet system. Our main result in this respect is that for all sufficiently dense regular sampling sets , the discrete wavelet system obtained by shifts from and dilations by powers of 2 is a universal Banach frame for all Besov spaces. In other words, the wavelet system allows the decomposition converging unconditionally in whenever , with coefficients depending linearly and boundedly on , and satisfying the norm equivalence
2. Preliminaries and Notation
Following the terminology in [19], we call a Lie group stratified if it is connected and simply connected, and its Lie algebra decomposes as a direct sum , with for and . Then is nilpotent of step and generated as a Lie algebra by . Euclidean spaces and the Heisenberg group are examples of stratified Lie groups.
If is stratified, its Lie algebra admits a canonical (natural) family of dilations, namely, which are Lie algebra automorphisms. We identify with through the exponential map. Hence is a Lie group with underlying manifold , for some , and the group product provided by the Campbell-Baker-Hausdorff formula. The dilations are then also group automorphisms of . Instead of writing for and , we simply use , whenever a confusion with the Lie group product is excluded. After choosing a basis of obtained as a union of bases of the , and a possible change of coordinates, one therefore has for and that for integers , according to .
Under our identification of with , polynomials on are polynomials on (with respect to any linear coordinate system on the latter). Polynomials on are written as where are the coefficients with respect to a suitable basis , and the monomials associated to the multi-indices . For a multi-index , define A polynomial of the type (2.3) is called of homogeneous degree if holds, for all multiindices with . We write for the space of polynomials of homogeneous degree .
We let denote the space of Schwartz functions on . By definition,. Let and denote the space of distributions and distributions modulo polynomials on , respectively. The duality between the spaces is denoted by the map . Most of the time, however, we will work with the sesquilinear version , for and .
Left Haar measure on is induced by Lebesgue measure on its Lie algebra, and it is also right-invariant. The number will be called the homogeneous dimension of . (For instance, for and we have and , respectively.) For any function on and , the -normalized dilation of is defined by Observe that this action preserves the -norm, that is, . We fix a homogeneous quasi-norm on which is smooth away from 0 with, for all , , for all , with if , and fulfilling a triangle inequality , with constant . Confer [19] for the construction of homogeneous norms, as well as further properties.
Moreover, by [19, Proposition 1.15], for any , there is a finite such that for all .
Our conventions for left-invariant operators on are as follows. We let denote a basis of , obtained as a union of bases of the . In particular, , for , is a basis of . Elements of the Lie algebra are identified in the usual manner with left-invariant differential operators on . Given a multi-index , we write for . A convenient characterization of Schwartz functions in terms of left-invariant operators states that if and only if, for all , , where In addition, the norms induce the topology of (see [19]).
The sub-Laplacian operator on can be viewed as the analog of the Laplacian operator on defined by . Using the above conventions for the choice of basis and , the sub-Laplacian is defined as . Note that a less restrictive notion of sub-Laplacians can also be found in the literature (e.g., any sum of squares of Lie algebra generators); we stress that the results in this paper crucially rely on the definition presented here. A linear differential operator on is called homogenous of degree if for any on . By choice of the for , these operators are homogeneous of degree 1; it follows that is homogenous of degree 2, and is homogenous of degree . Furthermore, any operator of the form is homogeneous of degree .
When restricted to , is formally self-adjoint: for any , . (For more see [15].) Its closure has domain , where we take in the sense of distributions. From this fact it quickly follows that this closure is self-adjoint and is in fact the unique self-adjoint extension of ; we denote this extension also by the symbol .
Suppose that has spectral resolution where is the projection measure. For a bounded Borel function on , the operator is a bounded integral operator on with a convolution distribution kernel in denoted by , and An important fact to be used later on is that for rapidly decaying smooth functions, , the kernel associated to is a Schwartz function. For a function on we define and . For , the adjoint of the convolution operator is provided by .
3. Homogeneous Besov Spaces on Stratified Lie Groups
In this section we define homogeneous Besov spaces on stratified Lie groups via Littlewood-Paley decompositions of distributions as
where is a dilated copy of a suitably chosen Schwartz function . In the Euclidean setting, it is customary to construct by picking a dyadic partition of unity on the Fourier transform side and applying Fourier inversion. The standard way of transferring this construction to stratified Lie groups consists in replacing the Fourier transform by the spectral decomposition of a sub-Laplacian , see Lemma 3.7. However, this approach raises the question to what extent the construction depends on the choice of . It turns out that the precise choice of sub-Laplacian obtained from a basis of is irrelevant. In order to prove this, we study Littlewood-Paley decompositions in somewhat different terms. The right setting for the study of such decompositions is the space of tempered distributions modulo polynomials, and the easiest approach to this convergence is via duality to a suitable space of Schwartz functions.
Definition 3.1. Let . A function has polynomial decay order if there exists a constant such that, for all , has vanishing moments of order , if one has with absolute convergence of the integral.
Under our identification of with , the inversion map is identical to the additive inversion map. That is, , and it follows that for all . Thus, if has vanishing moments of order , then for all that is, has vanishing moments of order as well.
Vanishing moments are central to most estimates in wavelet analysis, by the following principle: in a convolution product of the type , vanishing moments of one factor together with smoothness of the other result in decay. Later on, we will apply the lemma to Schwartz functions , where only the vanishing moment assumptions are nontrivial. The more general version given here is included for reference.
Lemma 3.2. Let be arbitrary.
(a)Let , such that is of decay order , for all with . Let have vanishing moments of order and decay order . Then there exists a constant, depending only on the decay of and , such that
In particular, if ,
(b)Now suppose that , with of decay order for all with . Let have vanishing moments of order and decay order . Then there exists a constant, depending only on the decay of and , such that
In particular, if ,
Proof. First, let us prove (a). Let . For , let denote the left Taylor polynomial of with homogeneous degree , see [19, Definition 1.44]. By that result,
with suitable positive constants and . We next use the homogeneity properties of the partial derivatives [19, page 21], together with the decay condition on to estimate for with
where the penultimate inequality used [19, 1.10], and the final estimate used . Thus,
Next, using vanishing moments of ,
and the integral is finite by [19, 1.15]. This proves (3.5), and (3.6) follows by
using .
For part (b), we first observe that
Our assumptions on allow to invoke part (a) with replacing , and (3.7) follows immediately. (3.8) is obtained from this by straightforward integration.
We let denote the space of Schwartz functions with all moments vanishing. We next consider properties of as a subspace of with the relative topology.
Lemma 3.3. is a closed subspace (in particular complete) of , with , as well as for all . The topological dual of , , can be canonically identified with the factor space .
Proof. By definition, is the intersection of kernels of a family of tempered distributions, hence a closed subspace. For and , one has by unimodularity of that , since is a polynomial. But then, for any and , one has for all polynomials on that since implies (translation on is polynomial). Thus . All further properties of follow from the corresponding statements concerning . For identification of with the quotient space , we first observe that a tempered distribution vanishes on if and only if its (Euclidean) Fourier transform is supported in , which is well known to be the case if and only if is a polynomial. Using this observation, we map to , where is a continuous extension of to all of ; such an extension exists by the Hahn-Banach theorem. The map is well defined because the difference between two extensions of annihilates and hence is a polynomial. Linearity follows from well-definedness. Furthermore, the inverse of the mapping is clearly obtained by assigning to the restriction .
In the following, we will usually not explicitly distinguish between and its equivalence class modulo polynomials, and we will occasionally write . The topology of is just the topology of pointwise convergence on the elements of . For any net , holds if and only if , for all . We next study convolution on .
Lemma 3.4. For every , the map is a well-defined and continuous operator . If , the associated convolution operator is a well-defined and continuous operator .
Proof. Note that . Hence induces a well-defined canonical map . Furthermore, is continuous on , as a consequence of [19, Proposition 1.47]. Therefore, for any net and any , the fact that allows to write
showing in .
For , the fact that makes the mapping well-defined modulo polynomials. The continuity statement is proved by (3.16), with assumptions on and switched.
The definition of homogeneous Besov spaces requires taking -norms of elements of . The following remark clarifies this.
Remark 3.5. Throughout this paper, we use the canonical embedding . For , this gives rise to an embedding , using that . Consequently, given , we let
assigning the value otherwise. Here the fact that guarantees that the decomposition is unique, and thus (3.17) well-defined.
By contrast, can only be defined on , if we assign the value to .
Note that with these definitions, the Hausdorff-Young inequality remains valid for all , and all (for ), respectively, (for ). For , this is clear. For , note that if , then with .
We now introduce a general Littlewood-Paley-type decomposition. For this purpose we define for ,
Definition 3.6. A function is called LP-admissible if for all , holds, with convergence in the Schwartz space topology. Duality entails the convergence for all .
The following lemma yields the chief construction of LP-admissible functions.
Lemma 3.7. Let be a function in with support in such that and on . Take . Thus, , with support in the interval , and Pick a sub-Laplacian , and let denote the convolution kernel associated to the bounded left-invariant operator . Then is LP-admissible, with .
Proof. Let us first comment on the properties of that are immediate from the construction via spectral calculus: follows from [20] and vanishing moments by [15, Proposition 1].
Now let . First note that 2-homogeneity of implies that the convolution kernel associated to coincides with . Then, by the spectral theorem and (3.21),
holds in -norm.
For any positive integer ,
where is the convolution kernel of . Since is a Schwartz function, it follows by [19, Proposition (1.49)] that , for , for all , with convergence in and a suitable constant .
We next show that in , as , for any . Fix a multi-index and with . Then left-invariance and homogeneity of yield
Here the first inequality is an application of (3.5); the constant can be estimated in terms of , for sufficiently large. But this proves in the Schwartz topology.
Summarizing, in , and in addition by (3.22), in , whence follows.
Note that an LP-admissible function as constructed in Lemma 3.7 fulfills the convenient relation which follows from .
Remark 3.8. By spectral calculus, we find that , with . In particular, the decomposition shows that is dense.
We now associate a scale of homogeneous Besov spaces to the function .
Definition 3.9. Let be LP-admissible, let , , and . The homogeneous Besov space associated to is defined as with associated norm
Remark 3.10. The definition relies on the conventions regarding -norms of distributions (modulo polynomials), as outlined in Remark 3.5. Definiteness of the Besov norm holds because of (3.20).
The combination of Lemma 3.7 with Definition 3.9 shows that we cover the homogeneous Besov spaces defined in the usual manner via the spectral calculus of sub-Laplacians. Hence the following theorem implies in particular that different sub-Laplacians yield the same homogeneous Besov spaces (at least within the range of sub-Laplacians that we consider).
Theorem 3.11. Let be LP-admissible. Let and . Then, , with equivalent norms.
Proof. It is sufficient to prove the norm equivalence, and here symmetry with respect to and immediately reduces the proof to showing, for a suitable constant ,
in the extended sense that the left-hand side is finite whenever the right-hand side is. Hence assume that ; otherwise, there is nothing to show. In the following, let ().
By LP-admissibility of ,
with convergence in . Accordingly,
where the convergence on the right-hand side holds in , by Lemma 3.4. We next show that the right-hand side also converges in . For this purpose, we observe that
where is a fixed integer. For , this follows directly from (3.8), using , and vanishing moments of , whereas for , the vanishing moments of allow to apply (3.6).
Using Young’s inequality, we estimate with from above that
Next observe that
By assumption, the sequence is in , in particular, bounded. Therefore, yields that (3.34) converges. But then the right-hand side of (3.31) converges unconditionally with respect to . This limit coincides with the -limit (which because of is even a -limit), yielding , with
Now an application of Young’s inequality for convolution over , again using , provides (3.29).
As a consequence, we write , for any LP-admissible . These spaces coincide with the homogeneous Besov spaces for the Heisenberg group in [9], and with the usual definitions in the case .
In the remainder of the section we note some functional-analytic properties of Besov spaces and Littlewood-Paley-decompositions for later use.
Lemma 3.12. For all and all , one has continuous inclusion maps , as well as , where the latter denotes the dual of . For , is dense.
Proof. We pick as in Lemma 3.7 and define for . For the inclusion , note that (3.6) and (3.8) allow to estimate for all and that
Here the constant is a suitable multiple of , for sufficiently large. But this implies that continuously.
For the other embedding, repeated applications of Hölder’s inequality yield the estimate
valid for all and . Here are the conjugate exponents of , respectively. But this estimate implies continuity of the embeddings and .
For the density statement, let , and . For convenience, we pick according to Lemma 3.7. Since , there exists such that
Next define
Let . By assumption on and Young’s inequality, , and since , there exists with . Let , then , and for ,
For , the construction of and implies that , whereas for , one has . As a consequence, one finds for
and for
For , one finds
with some constant depending only on . For instance, for ,
A straight forward application of triangle and Young’s inequality yields
Similar considerations applied to yield
Now summation over yields
as desired.
Remark 3.13. Let be as in Lemma 3.7. As a byproduct of the proof, we note that the space is dense in as well as , if . In , the decomposition holds with finitely many nonzero terms.
We next extend the Littlewood-Paley decomposition to the elements of the Besov space. For simplicity, we prove the result only for certain LP-admissible functions.
Proposition 3.14. Let , and let be an LP-admissible vector constructed via Lemma 3.7. Then the decomposition (3.19) converges for all in the Besov space norm.
Proof. Consider the operators , By suitably adapting the arguments proving the density statement of Lemma 3.12, it is easy to see that the family of operators is bounded in the operator norm. As noted in Remark 3.13, the strongly converges to the identity operator on a dense subspace. But then boundedness of the family implies strong convergence everywhere.
A further class of spaces for which the decomposition converges is .
Proposition 3.15. Let , and let be an LP-admissible vector constructed via Lemma 3.7. Then the decomposition (3.19) converges with respect to , for all .
Proof. Let the operator family be defined as in the previous proof. Then, , and Young’s inequality implies that the sequence of operators is norm-bounded. It therefore suffices to prove the desired convergence on the dense subspace . By [19, Proposition 1.49],. Furthermore, for , and thus Again by [19, Proposition 1.49],, in particular, Hence, , and the case yields .
Theorem 3.16. is a Banach space.
Proof. Completeness is the only issue here. Again, we pick as an LP-admissible vector via Lemma 3.7. Suppose that is a Cauchy sequence. As a consequence, one has in particular, for all , that is a Cauchy sequence, hence , for a suitable . Furthermore, the Cauchy property of implies that is a Cauchy sequence. On the other hand, the sequence converges pointwise to , whence We define Now, using (3.56) and , where are the conjugate exponents of , respectively, a straightforward calculation as in the proof of Lemma 3.12 shows that the sum defining converges in . Furthermore, (3.56) and (3.25) easily imply that . Finally, for the proof of , we employ (3.25) together with the equality , to show that Summarizing, the sequence is a Cauchy sequence, converging pointwise to 0. But then follows.
4. Characterization via Continuous Wavelet Transform
The following definition can be viewed as a continuous-scale analog of LP-admissibility.
Definition 4.1. is called -admissible, if for all , holds with convergence in the Schwartz topology.
The next theorem reveals a large class of -admissible wavelets. In fact, all the wavelets studied in [15] are also -admissible in the sense considered here. Its proof is an adaptation of the argument showing [15, Theorem 1].
We let
Theorem 4.2. Let , and let be the distribution kernel associated to the operator . Then is -admissible up to normalization.
Proof. The main idea of the proof is to write, for ,
with suitable . Once this is established, for follows by [19, Proposition (1.49)], with convergence in the Schwartz topology. Moreover, entails that in the Schwartz topology: given any and with associated left-invariant differential operator , we can employ (3.5) to estimate
which converges to zero for , as soon as and . But this implies in , by [19].
Thus it remains to construct . To this end, define
which is clearly in , and let denote the associated convolution kernel of . By the definition, . Let be in , and let denote the scalar-valued Borel measure associated to by the spectral measure. Then, by spectral calculus and the invariance properties of ,
as desired.
Hence, by [15, Corollary 1] we have the following.
Corollary 4.3.
(a) There exist -admissible .
(b) There exist -admissible with vanishing moments of arbitrary finite order.
Given a tempered distribution and a -admissible function , the continuous wavelet transform of is the family of convolution products. We will now prove a characterization of Besov spaces in terms of the continuous wavelet transform.
Another popular candidate for defining scales of Besov spaces is the heat semigroup; see for example, [1] for the inhomogeneous case on stratified groups, or rather [21] for the general treatment. In our setting, the heat semigroup associated to the sub-Laplacian is given by right convolution with , where is the kernel of with . Theorem 4.2 implies that is -admissible; it can be viewed as an analog of the well-known Mexican Hat wavelet. (For general stratified Lie groups, this class of wavelets was studied for the first time in [15].) The wavelet transform of associated to is then very closely related to the -fold time derivative of the solution to the heat equation with initial condition . By choice of , denotes the solution of the heat equation associated to , with initial condition . A formal calculation using left invariance of then yields Thus the following theorem also implies a characterization of Besov spaces in terms of the heat semigroup.
Theorem 4.4. Let be -admissible, with vanishing moments of order . Then, for all with , and all , , the following norm equivalence holds: Here the norm equivalence is understood in the extended sense that one side is finite if and only if the other side is. If , the equivalence is also valid for the case .
Proof. The strategy consists in adapting the proof of Theorem 3.11 to the setting where one summation over scales is replaced by integration. This time, however, we have to deal with both directions of the norm equivalence. In the following estimates, the symbol denotes a constant that may change from line to line, but in a way that is independent of .
Let us first assume that
for , , for a -admissible function with vanishing moments (, if ). Let be LP-admissible. Then, for all ,
holds in , by Lemma 3.4.
We next prove that the right-hand side of (4.11) converges in . For this purpose, introduce
We estimate
where we used that is scaling invariant. Note that the last inequality is Hölder’s inequality for . In this case, taking th powers and summing over yields
Using vanishing moments and Schwartz properties of and , we can now employ (3.6) and (3.8) to obtain
with a constant independent of . But then, since , we may proceed just as in the proof of Theorem 3.11 to estimate the integrand in (4.15) via
Summarizing, we obtain
In particular, . But then the right-hand side of (4.11) converges to in . The Minkowski inequality for integrals yields , and thus
as desired. In the case , (4.16) yields that
Thus, by (4.13),
The remainder of the argument is the same as for the case .
Next assume . Then, for all and ,
with convergence in ; for , convergence holds even in . As before,
Again, we have with a constant independent of . Hence, one concludes in the same fashion as in the proof of Theorem 3.11 that, for all ,
again with a constant independent of . In the case , this finishes the proof immediately, and for , we integrate the th power over and sum over to obtain the desired inequality.
Remark 4.5. Clearly, the proof of Theorem 4.4 can be adapted to consider discrete Littlewood-Paley decompositions based on integer powers of any instead of . Thus consistently replacing powers of 2 in Definitions 3.6 and 3.9 by powers of results in the same scale of Besov spaces.
As an application of the characterization via continuous wavelet transforms, we exhibit certain of the homogeneous Besov spaces as homogeneous Sobolev spaces, and we investigate the mapping properties of sub-Laplacians between Besov spaces of different smoothness exponents.
Lemma 4.6. , with equivalent norms.
Proof. Pick by Lemma 3.7. Then spectral calculus implies that for all Since is dense in both spaces, and both spaces are complete, it follows that .
The next lemma investigates the mapping properties of sub-Laplacians between Besov spaces of different smoothness exponents. Its proof is greatly facilitated by the characterization via continuous wavelet transforms.
Lemma 4.7. Let denote a sub-Laplacian. For all , , and , in the extended sense that one side is infinite if and only if the other side is. In particular, is a bijection, and it makes sense to extend the definition to negative . Thus, for all , is a topological isomorphism of Banach spaces.
Proof. Pick a nonzero real-valued , an integer and let denote the distribution kernel of . Hence is admissible by Theorem 4.2, with vanishing moments of order and . On , the convolution operator can be written as with a suitable function . For , spectral calculus implies where we employed left invariance to pull past in the convolution. Note that up to normalization, is admissible with vanishing moments of order . Thus, applying Theorem 4.4, we obtain Now assume that . Then, combining the density statements from Lemma 3.12 and Remark 3.8, we obtain a sequence with in ; thus also with convergence in . The norm equivalence and completeness of yield that , for suitable . Again, this implies convergence in . Since is continuous on that space, it follows that , establishing that . Since any distribution annihilated by is a polynomial, this finally yields , and follows by taking limits. A similar but simpler argument establishes the norm equivalence under the assumption that .
This observation shows that we can regard certain Besov spaces as homogeneous Sobolev spaces, or, more generally, as generalizations of Riesz potential spaces.
Corollary 4.8. For all : .
As a further corollary, we obtain the following interesting result relating two sub-Laplacians and . For all , the operator
is densely defined and has a bounded extension with bounded inverse. More general analogues involving more than two sub-Laplacians are also easily formulated. For the Euclidean case, this is easily derived using the Fourier transform, which can be viewed as a joint spectral decomposition of commuting operators. In the general, nonabelian case however, this tool is not readily available, and we are not aware of a direct proof of this observation, nor of a previous source containing it.
5. Characterization of Besov Spaces by Discrete Wavelet Systems
We next show that the Littlewood-Paley characterization of can be discretized by sampling the convolution products over a given discrete set . This is equivalent to the study of the analysis operator associated to a discrete wavelet system , defined by Throughout the rest of the paper, we assume that the wavelet has been chosen according to Lemma 3.7 and .
We first define the discrete coefficient spaces which will be instrumental in the characterization of the Besov spaces.
Definition 5.1. Fix a discrete set . For a family of complex numbers, we define The coefficient space associated to and is then defined as We simply write if is understood from the context.
We define the analysis operator associated to the function and , assigning each the family of coefficients . Note that the analysis operator is implicitly assumed to refer to the same set that is used in the definition of .
We next formulate properties of the sampling sets we intend to use in the following. We will focus on regular sampling, as specified in the next definition. Most of the results are obtainable for less regular sampling sets, at the cost of more intricate notation.
Definition 5.2. A subset is called regular sampling set, if there exists a relatively compact Borel neighborhood of the identity element of satisfying (up to a set of measure zero) as well as , for all distinct . Such a set is called a - tile. A regular sampling set is called -dense, for , if there exists a -tile .
Note that the definition of -dense used here is somewhat more restrictive than, for example, in [14]. A particular class of regular sampling sets is provided by lattices, that is, cocompact discrete subgroups . Here, -tiles are systems of representatives mod . However, not every stratified Lie group admits a lattice. By contrast, there always exist sufficiently dense regular sampling sets, as the following result shows.
Lemma 5.3. For every neighborhood of the identity, there exists a -dense regular sampling set.
Proof. By [14, Lemma 5.10], there exists and a relatively compact with nonempty open interior, such that tiles (up to sets of measure zero). Then is a -tile, for some point in the interior of . Finally, choosing sufficiently small ensures that , and is a -tile.
The chief result of this section is the following theorem which shows that the Besov norms can be expressed in terms of discrete coefficients. Note that the constants arising in the following norm equivalences may depend on the space, but the same sampling set is used simultaneously for all spaces.
Theorem 5.4. There exists a neighborhood of the identity, such that for all -dense regular sampling sets , and for all and all , the following implication holds: Furthermore, the induced coefficient operator is a topological embedding. In other words, on one has the norm equivalence with constants depending on , and .
Remark 5.5. As a byproduct of the discussion in this section, we will obtain that the tightness of the frame estimates approaches 1, as the density of the sampling set increases. That is, the wavelet frames are asymptotically tight.
For the proof of Theorem 5.4, we need to introduce some notations. In the following, we write
which is a space of smooth functions, as well as . Furthermore, let , and denote by the restriction operator.
In order to prove Theorem 5.4, it is enough to prove the following sampling result for the spaces ; the rest of the argument consists in summing over . In particular, note that the sampling set is independent of and , and the associated constants are independent of .
Lemma 5.6. There exists a neighborhood of the identity, such that for all -dense regular sampling sets , the implication holds. Furthermore, with suitable constants (for ), the inequalities hold for all and all .
Proof. Here we only show that the case implies the other cases; the rest will be established below. Hence assume (5.8) is known for . Let . For arbitrary , we have that , and thus Here , where the dilation action on distributions is defined in the usual manner by duality. The last equality follows from the fact that is a group homomorphism. Recall that for any and , , applying the case , we obtain for that which is the upper estimate for arbitrary . The lower estimate and the case follow by similar calculations.
For the remainder of this section, we will therefore be concerned with the case , which will be treated using ideas similar to the ones in [14], relying mainly on oscillation estimates. Given any function on and a set , we define the oscillation
We can then formulate the following result.
Proposition 5.7. Let be a space of continuous functions. Suppose that there exists such that, for all , holds pointwise. Define , for . Let , and, be a neighborhood of the unit element fulfilling . Then, for all -dense regular sampling sets , the following implication holds:
The restriction map induces a topological embedding .
More precisely, for ,
where denotes a -tile, and
Proof. We introduce the auxiliary operator defined by
with . Since the sets are pairwise disjoint, is a multiple of an isometry, . In particular, has a bounded inverse on its range, and implies for any sequence .
The equation implies the pointwise inequality
(see [14, page 185]). Now Young’s inequality provides for :
Since the ’s are disjoint, we may then estimate, for all ,
In particular, , whence . In addition, we obtain the upper bound of the sampling inequality for
The lower bound follows similarly by
Thus (5.13) and (5.12) are shown, for . For , we note that . Furthermore,
Now the remainder of the proof is easily adapted from the case .
It remains to check the conditions of the proposition for
Lemma 5.8. There exists a Schwartz function acting as a reproducing kernel for , that is, holds for all .
Proof. We pick a real-valued -function on that is identically 1 on the support of , and let be the associated distribution kernel to . Then , whence follows, for all .
Lemma 5.9. Let be a Schwartz function. For every , there exists a compact neighborhood of the unit element such that .
Proof. First observe that, by continuity, pointwise, as runs through a neighborhood base at the identity element. Thus by dominated convergence it suffices to prove , for some neighborhood .
Let . A straightforward application of the mean value theorem [19, Theorem 1.33] yields
Here and are constants depending on . The Sobolev estimate [22, (5.13)] for yields that for all with
where runs through all possible with , including the identity operator corresponding to . Furthermore, , and is a constant. Now integrating against Haar-measure (which is two-sided invariant) yields
and the last integral is finite because is a Schwartz function.
Now Lemma 5.6 is a direct consequence of Proposition 5.7 and Lemmas 5.8 and 5.9. Note that the tightness in Proposition 5.7 converges to 1, as runs through a neighborhood of the identity. This property is then inherited by the norm estimates in Theorem 5.4.
6. Banach Wavelet Frames for Besov Spaces
In Hilbert spaces a norm equivalence such as (5.5) would suffice to imply that the wavelet system is a frame, thus entailing a bounded reconstruction from the discrete coefficients. For Banach spaces one needs to use the extended definition of frames [23], that is, to show the invertibility of associated frame operator. In this section we will establish these statements for wavelet systems in Besov space. We retain the assumption that the wavelet was chosen according to Lemma 3.7.
We first prove that any linear combination of wavelet systems with coefficients in converges unconditionally in , compare [12, Theorem 3.1]. We then show that for all sufficiently dense choices of the sampling set , the wavelet system constitutes a Banach frame for .
Recall that the sampled convolution products studied in the previous sections can be read as scalar products
where denotes the wavelet of scale at position . In the following, the wavelet system is used for synthesis purposes, that is, we consider linear combinations of discrete wavelets. The next result can be viewed in parallel to synthesis results, for example, in [7]. It establishes synthesis for a large class of systems. Note in particular that the functions need not be obtained by dilation and shifts from a single function .
Theorem 6.1. Let be a regular sampling set. Let . (a)Suppose that one is given tempered distributions satisfying the following decay conditions: for all , there exist constants such that for all , , : Then for all , the sum converges unconditionally in the Besov norm, with for some constant independent of . In other words, the synthesis operator associated to the system is bounded. (b)The synthesis result in (a) holds in particular for
In order to motivate the following somewhat technical lemmas, let us give a short sketch of the proof strategy for the theorem. It suffices to show (6.4) for all finitely supported sequences; the rest follows by density arguments, using that is a Banach space. Hence, given a finitely supported coefficient sequence and , we need estimates for the -norms of These estimates are obtained by first looking at the summation over , with fixed, and then summing over . In both steps, we use the decay condition (6.2).
The following lemma shows that (6.2) is fulfilled for and thus allows to conclude part (b) of Theorem 6.1.
Lemma 6.2. There exists a constant such that for any , , , the following estimate holds:
Proof. We first compute In particular, (3.25) implies that the convolution vanishes if . For the other case, we observe that the convolution products , for are Schwartz functions, hence
For the convergence of the sums over , we will need the Schur test for boundedness of infinite matrices on -spaces.
Lemma 6.3. Let . Let be some countable set, and let denote a matrix of complex numbers. Assume that, for some finite constant , Then the operator is bounded on , with operator norm .
Lemma 6.4. Let , with and . Let be separated. Then for any , one has where the constant depends only on and .
Proof. By assumption, there exists an open set such that , for with . In addition, we may assume is relatively compact. Then, For , the triangle inequality of the quasi-norm yields with the last inequality due to . Accordingly, where the inequality used disjointness of the . For , the integral is finite.
The next lemma is an analog of [12, Lemma 3.4], which we will need for the proof of Theorem 6.1.
Lemma 6.5. Let and be fixed with . Suppose that is a regular sampling set. For any , let be a function on . Assume that the fulfill the decay estimate with a constant . Define , where . Then the series converges unconditionally in , with with a constant independent of , , and of the coefficient sequence.
Proof. To prove the assertion, let be a -tile. Then, On each integration patch , the triangle inequality of the quasi-norm yields the estimate compare the proof of Lemma 6.4, and thus the integrand can be estimated from above by the constant whence Here . Now Lemma 6.4 yields that the Schur test is fulfilled for the coefficients with (observe in particular that the right-hand side of the estimate above is independent of ), thus Lemma 6.3 yields as desired.
Proof of Theorem 6.1. We still need to prove part (a) of the theorem, and here it is sufficient to show the norm estimate for all finitely supported coefficient sequences . The full statement then follows by completeness of and from the fact that the Kronecker-s are an unconditional basis of (here we need ).
Repeated applications of the triangle inequality yield
Pick such that and . Define
For , assumption (6.2) yields
and thus, by Lemma 6.5,
But then
where denotes convolution over , and
By choice of , , and Young’s inequality allows to conclude that
For , assumption (6.2) provides the estimate
Here, Lemma 6.5 and straightforward calculation allow to conclude that
with
Hence, Young’s theorem applies again and yields
and we are done.
We conclude this section by showing that wavelets provide a simultaneous Banach frame for , for all and ; see [24] for an introduction to Banach frames. In the following, we consider the frame operator associated to a regular sampling set , given by By Theorems 6.1 and 5.4, is bounded, at least for sufficiently dense sampling sets . Our aim is to show that, for all sufficiently dense regular sampling sets, the operator is in fact invertible, showing that the wavelet system is a Banach frame for . The following lemma contains the main technical ingredient for the proof. Once again, we will rely on oscillation estimates.
Lemma 6.6. Let , with , such that , for some . For , there exists a neighborhood of the identity such that, for all -dense regular sampling sets and all -tiles , one has
Proof. We first consider the case . Let denote a -tile. We define the auxiliary function
using the notation of the proof of Proposition 5.7. By the triangle inequality,
Now Young’s inequality, together with the proof of Proposition 5.7, implies that for all sufficiently dense ,
For the second term in the right hand side of (6.37), we first observe that , and thus using the tiling ,
Since if and only if , it follows that
thus we can continue the estimate by
leading to
using as well as , both valid for sufficiently dense , by the proof of Proposition 5.7, and by Lemma 5.9, respectively.
Thus (6.35) is established for . The statement for general now follows by dilation, similar to the proof of Lemma 5.6. We write , where . Hence, for
we obtain that
Now
Thus, by the case ,
as desired.
Now, invertibility of the frame operator is easily established. In fact, we can even show the existence of a dual frame and an atomic decomposition for our homogeneous Besov spaces. Note however that the notation of the following theorem is somewhat deceptive. The dual wavelet frame might depend on the space , whereas the well-known result for wavelet bases in the Euclidean setting allows to take , regardless of the Besov space under consideration.
Theorem 6.7 (Atomic decomposition). Let . There exists a neighborhood of the identity such that, for all -dense regular sampling sets , the frame operator is an automorphism of .
In this case, there exists a dual wavelet family , such that for all , one has
and in addition
Proof. Fix , and choose the neighborhood according to the previous lemma, with replaced by
Let be a -dense regular sampling set, and let denote a -tile. Let , where is the dense subspace of functions for which
holds with finitely many nonzero terms; see Remark 3.13. For , we then obtain from (3.25) that
where the inequality used Lemma 6.6 and Young’s inequality. But then it follows that
Since is bounded, the estimate extends to all . Therefore, the operator is invertible by its Neumann series on , for all sufficiently dense quasi-lattices . In this case, we define the dual wavelet frame by
, since is bounded and .
Let . By Theorem 5.4, implies . Theorem 6.1 then implies that
with unconditional convergence in the Besov norm. Furthermore, Theorems 6.1 and 5.4 yield that
up to constants depending on , but not on . This completes the proof.
Remark 6.8. We wish to stress that an appropriate choice of provides a wavelet frame in , simultaneously valid for all and all . As the discussion in Section 5 shows, the tightness of the oscillation estimates converges to one with increasing density of the quasi-lattices. As a consequence, the tightness of the wavelet frame in converges to one also, at least when measured with respect to the Besov norm from Definition 3.9, applied to the same window . However, the tightness will depend on , and .
Remark 6.9. We expect to remove the restriction on and in our future work and prove the existence of (quasi)Banach frame for all homogeneous Besov spaces with and .
Remark 6.10. Our treatment of discretization problems via oscillation estimates is heavily influenced by the work of Feichtinger and Gröchenig on atomic decomposition, in particular the papers [23, 25] on coorbit spaces. A direct application of these results to our problem is difficult, since the representations underlying our wavelet transforms are not irreducible if the group is noncommutative, whereas irreducibility is an underlying assumption in [23, 25]. However, the recent extensions of coorbit theory, most notably [26], provide a unified approach to our results (see [27]).
Acknowledgment
The authors thank the referees for useful comments and additional references.