This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.Lecture 4 Covariant Transform
A general group-theoretical
construction [279, 95, 173, 5, 106, 63, 220]
of wavelets (or coherent state) starts from an
irreducible square integrable representation—in the
proper sense or modulo a subgroup. Then a mother wavelet is chosen to
be admissible. This leads to a wavelet transform which is an
isometry to L2 space with respect to the Haar measure
on the group or (quasi)invariant measure on a homogeneous space.
The importance of the above situation shall not be diminished, however
an exclusive restriction to such a setup is not necessary, in fact.
Here is a classical example from complex analysis: the Hardy space
H2(T) on the unit circle and Bergman spaces
B2n(ⅅ), n≥ 2 in the unit disk produce wavelets
associated with representations ρ1 and ρn of the group
SL2(ℝ) respectively [170]. While representations ρn, n≥ 2
are from square integrable discrete series, the mock discrete series
representation ρ1 is not square
integrable [240]*§ VI.5
[321]*§ 8.4. However it would be natural to treat the
Hardy space in the same framework as Bergman ones. Some more examples
will be presented below.
4.1 Extending Wavelet Transform
To make a sharp but still natural generalisation of wavelets we give the following
definition.
Definition 1 [192]
Let be a representation of
a group G in a space V and F be an operator from V to a space
U. We define a covariant transform
W from V to the space L(
G,
U)
of
U-valued functions on G by the formula:
W: v↦ v(g) = F((g−1) v),
v∈ V, g∈ G.
(1) |
Operator F will be called fiducial operator
in this context.
We borrow the name for operator F from fiducial vectors of
Klauder and Skagerstam [220].
Remark 2
We do not require that fiducial operator F shall be linear.
Sometimes the positive homogeneity, i.e. F(
t v)=
tF(
v)
for
t>0
, alone can be already sufficient, see
Example 15.
Remark 3
Usefulness of the covariant transform is in the reverse proportion
to the dimensionality of the space U. The covariant transform
encodes properties of v in a function Wv on G.
For a low dimensional U this function can be ultimately
investigated by means of harmonic analysis. Thus dim
U=1
(scalar-valued functions) is the ideal case, however, it is
unattainable sometimes, see Example 12 below. We
may have to use a higher dimensions of U if the given group
G is not rich enough.Moreover, the relation between the dimensionality of U and
usefulness of the covariant transform shall not be taking
dogmatically. Paper [201] gives an important example of
covariant transform which provides a simplification even in the case
of dimU =dimV.
As we will see below covariant transform is a close relative of
wavelet transform. The name is chosen due to the following common
property of both transformations.
Theorem 4
The covariant transform (1)
intertwines and the left regular representation
Λ
on L(
G,
U)
:
Here Λ
is defined as usual by:
Λ(g): f(h) ↦ f(g−1h).
(2) |
Proof.
We have a calculation similar to wavelet
transform [173]*Prop. 2.6. Take u=(g) v and
calculate its covariant transform:
[W( (g) v)] (h) | = | [W((g) v)] (h)=F((h−1) (g) v ) |
| = | F(((g−1h)−1) v) |
| = | [Wv] (g−1h) |
| = | Λ(g) [Wv] (h).
|
|
The next result follows immediately:
Corollary 5
The image space W(
V)
is invariant under the
left shifts on G.
Remark 6
A further generalisation of the covariant transform can be obtained
if we relax the group structure. Consider, for example, a
cancellative semigroup
ℤ
+ of non-negative
integers. It has a linear presentation on the space of polynomials
in a variable t defined by the action m:
tn ↦
tm+n
on the monomials. Application of a linear functional l, e.g.
defined by an integration over a measure on the real line, produces
umbral calculus
l(
tn)=
cn, which has a magic efficiency
in many areas, notably in combinatorics [238, 174]. In
this direction we also find fruitful to expand the notion of an
intertwining operator to a
token
[178].
4.2 Examples of Covariant Transform
In this Subsection we will provide several examples of covariant
transforms. Some of them will be expanded in subsequent sections,
however a detailed study of all aspects will not fit into the present
work. We start from the classical example of the group-theoretical
wavelet transform:
Example 7
[279, 95, 173, 5, 220, 95]
Let V be a Hilbert space with an inner product
⟨ ·,·
⟩
and be a unitary
representation of a group G in the space V. Let F:
V
→ ℂ
be the functional v↦
⟨
v,
v0
⟩
defined by a vector v0∈
V. The vector
v0 is often called the mother wavelet
in areas related
to signal processing, the vacuum state
in the quantum
framework, etc.In this set-up, the transformation (1) is
the well-known expression for a wavelet transform [5]*(7.48) (or
representation coefficients):
W: v↦ v(g) = ⟨ (g−1)v,v0
⟩ =
⟨ v,(g)v0
⟩,
v∈ V, g∈ G.
(3) |
The family of vectors vg=(g)v0 is called
wavelets or coherent states. In this case we obtain
scalar valued functions on G, cf. Rem. 3.
This scheme is typically carried out for a square integrable
representation with v0 being an
admissible vector [279, 95, 5, 106, 63, 87].
In this case the wavelet (covariant) transform is a map into the
square integrable functions [87] with respect to the left
Haar measure on G. The map becomes an isometry if v0 is
properly scaled. Moreover, we are able to recover the input v from its
wavelet transform through the reconstruction formula, which requires an
admissible vector as well, see Example 28 below.
The most popularised case of the above scheme is
as follows.
Example 8
An isometric representation of the ax+
b group on
V=
Lp(ℝ)
is given by the formula:
[p(a,b) f](x)= a | | f | ⎛
⎜
⎜
⎝ | | ⎞
⎟
⎟
⎠ | .
(4) |
The representation (7) is square integrable for
p=2
. Any function v0, such that its Fourier transform
v(ξ)
satisfy to
is admissible [5]*§ 12.2. The continuous
wavelet transform is generated by the
representation (7) acting on an admissible vector
v0
in the expression (3). The image of a function
from L2(ℝ)
is a function on the upper
half-plane square integrable with respect to the measure
a−2 da db. For sufficiently regular v
0 the
admissibility (5) of v0 follows from a
weaker condition
However, square integrable representations and admissible vectors do
not cover all interesting cases.
Example 9
For the above G=
Aff and the
representation (7), we consider the operators
F±:
Lp(ℝ) → ℂ
defined by:
Then the covariant transform (1) is the
Cauchy integral from Lp(ℝ)
to the space of
functions f(
a,
b)
such that
a−1/pf(
a,
b)
is in the Hardy space in the
upper/lower half-plane Hp(ℝ
±2)
.
Although the representation (7) is square
integrable for p=2
, the function 1/
x±
i used
in (7) is not an admissible vacuum vector. Thus the
complex analysis become decoupled from the traditional wavelet
theory. As a result the application of wavelet theory shall relay on
an extraneous mother wavelets [146].Many important objects in complex analysis are generated by
inadmissible mother wavelets like (7). For example, if
F:L2(ℝ) → ℂ is defined
by F: f ↦ F+ f − F−f then the covariant
transform (1) reduces to the Poisson
integral. If F:L2(ℝ) →
ℂ2 is defined by F: f ↦( F+ f, F−f) then the
covariant transform (1) represents a
function f on the real line as a jump:
f(z)=f+(z)−f−(z), f±(z)∈ Hp(ℝ±2)
(8) |
between functions analytic in the upper and the lower half-planes.
This makes a decomposition of L2(ℝ) into
irreducible components of the representation (7).
Another interesting but non-admissible vector is the Gaussian
e−x2.
Example 10
For the group G=
SL2(ℝ)
[240] let us consider the unitary
representation (9) on the space of
square integrable function L2(ℝ
+2)
on the
upper half-plane through the Möbius transformations (1):
(g): f(z) ↦ | |
f | ⎛
⎜
⎜
⎝ | | ⎞
⎟
⎟
⎠ | , g−1=
| | .
(9) |
This is a representation from the discrete series and
L2(ⅅ)
and irreducible invariant subspaces
are parametrised by integers. Let Fk be the functional
L2(ℝ
+2)→ ℂ
of
pairing with the lowest/highest k-weight vector in the
corresponding irreducible component (Bergman space)
Bk(ℝ
±2)
, k≥ 2
of the discrete
series [240]*Ch. VI. Then we can build an operator F
from various Fk similarly to the previous Example. In
particular, the jump representation (8) on the real
line generalises to the representation of a square integrable
function f on the upper half-plane as a sum
f(z)= | | ak fk(z), fk∈Bn(ℝ±2)
|
for prescribed coefficients ak and analytic functions fk in
question from different irreducible subspaces.Covariant transform is also meaningful for principal and
complementary series of representations of the group
SL2(ℝ), which are not square integrable [170].
Example 11
Let G=
SU(2)×
Aff be the Cartesian product
of the groups SU(2)
of unitary rotations of
ℂ
2 and the ax+
b group Aff. This group
has a unitary linear representation on the space
L2(ℝ,ℂ
2)
of square-integrable (vector)
ℂ
2-valued functions by the formula:
(g)
| | = | ⎛
⎜
⎝ | α f1(at+b)+ β f2(at+b) |
γ f1(at+b)+δ f2(at+b)
|
| ⎞
⎟
⎠ |
| ,
|
where g=
(
)× (
a,
b)∈
SU(2)×
Aff. It
is obvious that the vector Hardy space, that is functions with both
components being analytic, is invariant under such action of G.As a fiducial operator F: L2(ℝ,ℂ2)
→ ℂ we can take, cf. (7):
Thus the image of the associated covariant transform is a subspace
of scalar valued bounded functions on G. In this way we can
transform (without a loss of information) vector-valued problems, e.g.
matrix Wiener–Hopf factorisation [47],
to scalar question of harmonic analysis on the group G.
Example 12
A straightforward generalisation of Ex. 7 is
obtained if V is a Banach space and F:
V →
ℂ
is an element of V*. Then the
covariant transform coincides with the construction of wavelets in
Banach spaces [173].
Example 13
The next stage of generalisation is achieved if V is a
Banach space and F:
V → ℂ
n is a linear
operator. Then the corresponding covariant transform is a map
W:
V →
L(
G,ℂ
n)
. This is
closely related to M.G. Krein’s works on directing
functionals
[236], see also multiresolution
wavelet analysis
[51],
Clifford-valued
Fock–Segal–Bargmann spaces [66]
and [5]*Thm. 7.3.1.
Example 14
Let F be a projector Lp(ℝ)→
Lp(ℝ)
defined by the relation (
Ff)
(λ )=χ(λ)f(λ)
, where the hat denotes the Fourier
transform and χ(λ)
is the characteristic function of
the set [−2,−1]∪[1,2]
. Then the covariant transform
Lp(ℝ)→
C(
Aff,
Lp(ℝ))
generated by the
representation (7) of the affine group from
F contains all information provided by the Littlewood–Paley
operator
[114]*§ 5.1.1.
Example 15
A step in a different direction is a consideration of
non-linear operators. Take again the “ax+
b” group and its
representation (7).
We define F to be a homogeneous but non-linear functional
V→ ℝ
+:
Fm (f) = | | | | ⎪
⎪ | f(x) | ⎪
⎪ | dx.
(11) |
The covariant transform (1) becomes:
[Wp f](a,b) = F(p(a,b) f)
= | | |
| ⎪
⎪
⎪ | a | | f | ⎛
⎝ | ax+b | ⎞
⎠ | ⎪
⎪
⎪ | dx
= a | | | |
| ⎪
⎪ | f | ⎛
⎝ | x | ⎞
⎠ | ⎪
⎪ | dx.
(12) |
We will see its connections with the Hardy–Littlewood maximal
functions in
Example 41.
Since linearity has clear advantages, we may prefer to reformulate the
last example through linear covariant transforms. The idea is similar
to the representation of a convex function as an envelope of linear
ones, cf. [108]*Ch. I, Lem. 6.1. To this end, we take a
collection F of linear fiducial functionals and, for a
given function f, consider the set of all covariant transforms
WF f, F∈F.
Example 16
Let us return to the setup of the previous Example for
G=
Aff and its
representation (7). Consider the unit ball B
in L∞[−1,1]
. Then, any ω∈
B defines a
bounded linear functional Fω on L1(ℝ)
:
Fω(f)= | | | f(x) ω(x) dx
= | | ∫ | | f(x) ω(x) dx.
|
Of course, sup
ω∈ B Fω(
f)=
Fm(
f)
with Fm
from (11) and for all
f∈
L1(ℝ)
. Then, for the non-linear
covariant transform (12)
we have the following expression in terms of the linear covariant transforms
generated by Fω:
[W1m f](a,b) = | | [W1ω
f](a,b).
(13) |
The presence of suprimum is the price to pay for such a “linearisation”.
Remark 17
The above construction is not much different to the
grand maximal function
[314]*§ III.1.2.
Although, it may look like a generalisation of covariant transform,
grand maximal function can be realised as a particular case of
Defn. 8. Indeed, let M(
V)
be a
subgroup of the group of all invertible isometries of a metric space
V. If represents a group G by isometries of V
then we can consider the group G′
generated by all finite products of
M(
V)
and (
g)
, g∈
G with the straightforward
action [2]
on V. The grand maximal functions
is produced by the covariant transform for the representation
[2]
of G′
.
Example 18
Let V=
Lc(ℝ
2)
be the space of
compactly supported bounded functions on the plane. We take F
be the linear operator V→ ℂ
of integration
over the real line:
F: f(x,y)↦ F(f)= | ∫ | | f(x,0) dx.
|
Let G be the group of Euclidean motions of the plane
represented by on V by a change of variables. Then
the wavelet transform F((
g)
f)
is the Radon
transform
[127].
4.3 Symbolic Calculi
There is a very important class of the covariant transforms which maps
operators to functions. Among numerous sources we wish to single out
works of Berezin [34, 35]. We start from the
Berezin covariant symbol.
Example 19
Let a representation of a group G act on a
space X. Then there is an associated representation
B of G on a space V=
B(
X,
Y)
of
linear operators X→
Y defined by the
identity [35, 173]:
(B(g) A)x=A((g−1)x), x∈ X, g∈ G, A
∈ B(X,Y).
(14) |
Following the Remark 3 we take F to be a
functional V→ℂ
, for example F can be
defined from a pair x∈
X, l∈
Y* by the expression
F:
A↦ ⟨
Ax,
l
⟩
. Then the covariant
transform is:
This is an example of covariant calculus
[173, 34].
There are several variants of the last Example which are of a separate
interest.
Example 20
A modification of the previous construction is obtained if we
have two groups G1 and G2 represented by 1
and 2
on X and Y* respectively. Then we have a covariant
transform B(
X,
Y)→
L(
G1×
G2, ℂ)
defined by the formula:
W: A ↦ Â(g1,g2)=⟨ A1(g1)x,2(g2)l
⟩.
|
This generalises the above Berezin covariant calculi
[173].
Example 21
Let us restrict the previous example to the case when X=
Y is a
Hilbert space, 1=2=
and x=
l with
||
x||=1
. Than the range of the covariant transform:
W: A ↦ Â(g)=⟨ A(g)x,(g)x
⟩
|
is a subset of the numerical range
of the operator A. As
a function on a group Â(
g)
provides a better description of
A than the set of its values—numerical range.
Example 22
The group SU(1,1)≃
SL2(ℝ)
consists of
2× 2
matrices of the form (
)
with the unit
determinant [240]*§ IX.1. Let T be an operator
with the spectral radius less than 1
. Then the associated
Möbius transformation
g: T ↦ g· T = | | , where
g=
| | ∈ SL2(ℝ),
(15) |
produces a well-defined operator with the spectral radius less than
1
as well. Thus we have a representation of SU(1,1)
.Let us introduce the defect operators DT=(I−T*T)1/2 and
DT*=(I−TT*)1/2. For the fiducial operator F=DT*
the covariant transform is, cf. [318]*§ VI.1,
(1.2):
[W T](g)=F(g· T)=−eiφ ΘT(z) DT,
for
g= | |
| | ,
|
where the characteristic function
ΘT(z) [318]*§ VI.1, (1.1) is:
ΘT(z) = −T+DT* (I−zT*)−1 z DT.
|
Thus we approached the functional model of operators from the
covariant transform. In accordance with Remark 3
the model is most fruitful for the case of operator
F=DT* being one-dimensional.
The intertwining property in the previous examples was obtained as a
consequence of the general Prop. 6 about the covariant
transform. However it may be worth to select it as a separate definition:
Definition 23
A covariant calculus
, also known as
symbolic calculus
, is a map from operators to functions,
which intertwines two representations of the same group in the
respective spaces.
There is a dual class of covariant transforms acting in the opposite
direction: from functions to operators. The prominent examples are the
Berezin contravariant symbol [34, 173] and symbols
of a pseudodifferential operators
(PDO) [139, 173].
Example 24
The classical Riesz–Dunford functional calculus
[88]*§ VII.3
[266]*§ IV.2 maps analytical functions on the
unit disk to the linear operators, it is defined through the
Cauchy-type formula with the resolvent. The calculus is an
intertwining operator [182] between the Möbius
transformations of the unit disk, cf. (23), and the
actions (15) on operators from the
Example 22. This topic will be developed in
Subsection 1.2.
In line with the Defn. 23 we can
directly define the corresponding calculus through the intertwining
property [168, 182]:
Definition 25
A contravariant calculus
, also know as functional
calculus
, is a map from functions to operators,
which intertwines two representations of the same group in the
respective spaces.
The duality between co- and contravariant calculi is the particular
case of the duality between covariant transform and the
contravariant transform defined in the next Subsection. In many cases a
proper choice of spaces makes covariant and/or contravariant calculus
a bijection between functions and operators. Subsequently only one
form of calculus, either co- or contravariant, is defined explicitly,
although both of them are there in fact.
4.4 Contravariant Transform
An object invariant under the left action
Λ (2) is called left invariant.
For example, let L and L′ be two left invariant spaces of
functions on G. We say that a pairing ⟨ ·,·
⟩:
L× L′ → ℂ is left invariant if
⟨ Λ(g)f,Λ(g) f′
⟩= ⟨ f,f′
⟩, for all
f∈ L, f′∈ L′.
(16) |
Remark 26
-
We do not require the pairing to be linear in general.
- If the pairing is invariant on space L× L′ it is not
necessarily invariant (or even defined) on the whole
C(G)× C(G).
- In a more general setting we shall study an invariant pairing
on a homogeneous spaces instead of the group. However due to length
constraints we cannot consider it here beyond the Example 29.
- An invariant pairing on G can be obtained from an invariant
functional l by the formula ⟨ f1,f2
⟩=l(f1f2).
For a representation of G in V and w0∈ V,
we construct a function w(g)=(g)w0. We assume that the pairing
can be extended in its second component to this V-valued
functions. For example, such an extension can be defined in the weak
sense.
Definition 27
Let ⟨ ·,·
⟩
be a left invariant pairing on
L×
L′
as above, let be a representation of
G in a space V, we define the function
w(
g)=(
g)
v0 for v0∈
V. The contravariant
transform
M is a map L →
V
defined by the pairing:
Mw0: f ↦ ⟨ f,w
⟩,
where f∈ L.
(17) |
We can drop out sup/subscripts in Mw0 as we
are doing for WF.
Example 28 (Haar paring)
The most well-known example of an invariant pairing on
L2(
G,
dµ)×
L2(
G,
dµ)
is integration over the Haar measure:
⟨ f1,f2
⟩= | ∫ | | f1(g) f2(g) dg.
(18) |
If is a unitary square integrable representation of G and w0 is an
admissible vector, see Example 7, then this
pairing can be extended to w(
g)=(
g)
w0. The
contravariant transform is known in this setup as the
reconstruction formula
, cf. [5]*(8.19):
Mw0 f = | ∫ | | f(g) w(g) dg,
where w(g)=(g) w0.
(19) |
It is possible to use different admissible vectors v0 and
w0 for the wavelet
transform (3) and the reconstruction
formula (19), respectively, cf.
Example 40.
Example 29
Let ρ
be a square integrable representation of G modulo a subgroup
H⊂
G and let X=
G/
H be the corresponding homogeneous
space with a quasi-invariant measure dx. Then integration over
dx with an appropriate weight produces an invariant pairing.
The contravariant transform is a more general
version [5]*(7.52) of the reconstruction
formula
mentioned in the previous example.
If the invariant pairing is defined by integration over the Haar
measure, cf. Example 28, then we can show an
intertwining property for the contravariant transform as well.
Proposition 30 ([173]*Prop. 2.9)
The inverse wavelet transform
Mw0 (19) intertwines the
left regular representation Λ
(2)
on L2(
G)
and :
Corollary 31
The image Mw0(
L(
G))⊂
V of a left
invariant space L(
G)
under the inverse wavelet
transform Mw0 is invariant under the representation
.
Remark 32
It is an important observation, that the above intertwining property
is also true for some contravariant transforms which are not based
on pairing (18). For example, in the case of the
affine group all pairings (21),
(27) and
(non-linear!) (23) satisfy
to (20) for the respective representation
p (7).
Let
- be not a square integrable representation (even modulo
a subgroup); or
- w0 be an inadmissible vector of a square integrable
representation .
A suitable invariant pairing in this case is not associated with
integration over the Haar measure on G. In this
case we speak about a Hardy pairing. The following example explains
the name.
Example 33 (Hardy pairing)
Let G be the “ax+
b” group and its representation
(7) from Ex. 8. An
invariant pairing on G, which is not generated by the Haar
measure a−2da db, is:
⟨ f1,f2
⟩H=
| | | |
f1(a,b) f2(a,b) | | .
(21) |
For this pairing we can consider functions 1/2π
i
(
x+
i)
or e−x2, which are not admissible vectors in the
sense of square integrable representations. Then the contravariant transform
provides an integral resolutions
of the identity. For example, for
v0=1/2π
i (
x+
i)
we obtain:
In other words, it expresses the boundary values at a=0
of the Cauchy integral
−[
Cf])(
a,
x+
ai)
.
Similar pairings can be defined for other semi-direct products of two
groups. We can also extend a Hardy pairing to a group, which has a
subgroup with such a pairing.
Example 34
Let G be the group SL2(ℝ)
from the Ex. 10. Then
the “ax+
b” group is a subgroup of SL2(ℝ)
, moreover we can
parametrise SL2(ℝ)
by triples (
a,
b,θ)
,
θ∈(−π,π]
with the respective Haar
measure [240]*III.1(3). Then the Hardy
pairing
⟨ f1,f2
⟩= | | |
f1(a,b,θ) f2(a,b,θ) db dθ.
(22) |
is invariant on SL2(ℝ)
as well. The corresponding
contravariant transform provides even a finer resolution of the identity
which is invariant under conformal mappings of the Lobachevsky
half-plane.
Here is an important example of non-linear pairing.
Example 35
Let G=
Aff, an invariant homogeneous functional
on G is given by the L∞ version of the Haar
functional (18):
⟨ f1,f2
⟩∞= | | ⎪
⎪ | f1(g)f2(g) | ⎪
⎪ | .
(23) |
Define the following two functions on ℝ
:
v0+(t)= | ⎧
⎨
⎩ | |
| and
v0*(t)= | ⎧
⎪
⎨
⎪
⎩ | |
|
(24) |
The respective contravariant transforms are generated by
the representation ∞
(7) are:
|
[ Mv0+f](t) | = | f+(t)=⟨ f(a,b),∞(a,b)
v0+(t)
⟩∞=
| | ⎪
⎪ | f(a,t) | ⎪
⎪ | , |
| (25) |
[ Mv0*f](t) | = | f*(t)=⟨ f(a,b),∞(a,b)
v0*(t)
⟩∞=
| | ⎪
⎪ | f(a,b) | ⎪
⎪ | .
|
| (26) |
|
The transforms (25)
and (26) are the vertical
and
non-tangential maximal
functions
[231]*§ VIII.C.2, respectively.
Example 36
Consider again G=
Aff equipped now with an
invariant linear functional, which is a Hardy-type modification
(cf. (21)) of
L∞-functional (23):
⟨ f1,f2
⟩ | | = | | | | (f1(a,b) f2(a,b)),
(27) |
where lim is the limit superior. Then, the covariant
transform MH for this pairing from functions v+ and
v* (24) becomes:
|
[ Mv0+Hf](t) | = | ⟨ f(a,b),∞(a,b)
v0+(t)
⟩ | | =
| | f(a,t), |
| (28) |
[ Mv0*Hf](t) | = | ⟨ f(a,b),∞(a,b)
v0*(t)
⟩ | | =
| | f(a,b).
|
| (29) |
|
They are the normal
and
non-tangential
limits superior from the upper-half plane to
the real line, respectively.
There is the obvious inequality ⟨ f1,f2
⟩∞ ≥
⟨ f1,f2
⟩∞H between
pairings (23) and (27),
which produces the corresponding relation between respective
contravariant transforms.
There is an explicit duality between the covariant transform and the
contravariant transform. Discussion of the grand maximal function in the
Rem. 17 shows usefulness of the covariant
transform over the a family of fiducial functional. Thus, we shall not
be surprised by the contravariant transform over a family of
reconstructing vectors as well.
Definition 37
Let w:
Aff →
L1(ℝ)
be a function. We define a new function 1
w on
Aff with values in L1(ℝ)
through the point-wise action [1
w](
g)=1(
g)
w(
g)
of
∞
(7). If
sup
g||
w(
g)||
1< ∞
, then, for
f∈
L1(
Aff)
, we define the extended
contravariant transform
by:
[Mw f](x)= | ∫ | | f(g) [1 w](g) dg.
(30) |
Note, that (30) reduces to the
contravariant transform (19) if we start from
the constant function w(g)=w0.
Definition 38
We call a function r on ℝ
a nucleus
if:
-
r is supported in [−1,1],
- | r |<1/2 almost everywhere, and
- ∫ℝ r(x) dx=0, cf. (6).
Clearly, for a nucleus r, the function s=1(a,b) r has properties:
- s is supported in a ball centred at b and radius a,
- | s |<1/2a almost everywhere, and
- ∫ℝ s(x) dx=0.
In other words, s=1(a,b) r is an atom, cf. [314]*§ III.2.2 and any atom may be
obtained in this way from some nucleus and certain (a,b)∈Aff.
Example 39
Let f(
g)=∑
j λ
j δ
gj(
g)
with ∑
j
| λ
j |<∞
be a countable sum of point masses on
Aff. If all values of w(
gj)
are nucleuses,
then (30) becomes:
[Mw f](x)= | ∫ | | f(g) [1 w](g) dg
= | | λj sj,
(31) |
where sj=1(
gj)
w(
gj)
are atoms.
The right-hand side of (31) is known as an atomic
decomposition
of a function h(
x)=[
Mw
f](
x)
, see [314]*§ III.2.2.
4.5 Composing the Co- and Contravariant Transforms
In the case of classical wavelets,
the relation between the wavelet transform (3)
and the inverse wavelet transform (19) is
suggested by their names.
Example 40
For a square integrable representation and admissible vectors
v0 and w0, there is the
relation [5]*(8.52):
where the constant k depends on v0, w0 and the
Duflo–Moore operator [87]
[5]*§ 8.2.
It is of interest, that two different vectors can be used as analysing
vector in (3) and for the reconstructing
formula (19). Even a greater variety
can be achieved if we use additional fiducial operators and invariant
pairings.
Example 41
The composition of the contravariant transform
Mv0* (26) with the
covariant transform W∞ (12) is:
|
[Mv0* W∞f](t) | = |
| | ⎧
⎪
⎨
⎪
⎩ | | | |
| ⎪
⎪ | f | ⎛
⎝ | x | ⎞
⎠ | ⎪
⎪ | dx | ⎫
⎪
⎬
⎪
⎭ |
| (32) |
| = |
| | | ⎧
⎪
⎨
⎪
⎩ | | | |
| ⎪
⎪ | f | ⎛
⎝ | x | ⎞
⎠ | ⎪
⎪ | dx | ⎫
⎪
⎬
⎪
⎭ | .
|
| |
|
Thus, Mv0* W∞f coincides with the
Hardy–Littlewood maximal function
fM [231]*§ VIII.B.1, which contains important
information on the original function f. Combining
Props. 6 and 30 (through
Rem. 32), we deduce that the operator M:
f↦
fM intertwines p with itself:
pM=
M p (yet, M is non-linear).
Example 42
Let the mother wavelet v0(
x)=δ(
x)
be the Dirac delta
function, then the wavelet transform Wδ generated
by ∞
(7) on
C(ℝ)
is [
Wδf](
a,
b)=
f(
b)
.
Take the reconstruction vector w0(
x)=(1−χ
[−1,1](
x))/
x and
consider the respective inverse wavelet transform Mw0
produced by the Hardy pairing (21). Then, the
composition of both maps is:
The last expression is a singular integral operator
(SIO) [314]*§ I.5
[257]*§ 2.6 defined through the principal
value (in the sense of Cauchy).
Example 43
Let W be a covariant transfrom generated either by the
functional F± (7) (i.e. the Cauchy integral)
or (
F+ −
F−)
(i.e. the Poisson integral) from the
Example 9. Then, for the contravariant
transform Mv0+H (25)
the composition Mv0+H W becomes the normal
boundary value of the Cauchy/Poisson integral, respectively. The
similar composition Mv0*H W for the
reconstructing vector v0* (24) turns to be
the non-tangential limit of the Cauchy/Poisson integrals.
It is the classical question of harmonic analysis to identify a class
of functions on the real line such that Mv0*H
W becomes the identity operator on it. Combining intertwining
properties of the covariant and contravariant transforms
(Props. 6, 30 and
Rem. 32) we conclude that
Mv0*H W will intertwine the representation
with itself. If we restrict our attention to
-irreducible subspace, then a sort of Schur’s lemma
suggests that such an operator is a (possible zero) multiple of the
identity operator. This motivates the following template definition,
cf. [203]*§ 1.
Definition 44
For a representation of a group G in a space V,
a generalised Hardy space
H is an
-irreducible subspace of V.
Example 45
Let G=
Aff and the representation p is
defined in V=
Lp(ℝ)
by (7). Then the classical Hardy spaces
Hp(ℝ)
are p-irreducible, thus are provided
by the above definition.
We illustrate the group-theoretical technique by the following statement.
Proposition 46
Let B
be the spaces of bounded uniformly continuous functions
on the real line. Let F:
B→ ℝ
be
a fiducial functional such that:
| | F(
∞(1/a,0) f )= 0, for all f∈
B such that f(0)=0
(33) |
and F(∞(1,
b)
f)
is a continuous function of
b∈ℝ
for a given f∈
B.Then, Mv0*H∘ WF is a constant times the
identity operator on B.
Proof.
First of all we note that Mv0+H WF is a
bounded operator on B. Let
v(a,b)*=∞(a,b) v*. Obviously,
v(a,b)*(0)=v*(−b/a) is an eigenfunction for
operators Λ(a′,0), a′∈ℝ+ of the left
regular representation of Aff:
Λ(a′,0) v(a,b)*(0)= v(a,b)*(0).
(34) |
This and the left invariance of the
pairing (30) imply
Mv0*H∘ Λ (1/a,0)=Mv0*H for
any (a,0)∈Aff. Then, applying the intertwining
properties (44) we obtain:
[Mv0*H ∘ WF f](0) | = | [Mv0*H∘ Λ (1/a,0)∘ WF f](0) |
| = | [Mv0*H∘ WF ∘ ∞(1/a,0)f](0)
|
|
Using the limit a→ 0 (33)
and the continuity of F∘ ∞(1,b) we conclude that
the linear functional l:f↦ [Mv0*H∘ WF
f](0) vanishes for any f∈B such that f(0)=0.
Take a function f1∈B such that f1(0)=1 and
define c=l(f1). From linearity of l, for any f∈
B we have:
l(f)=l(f−f(0)f1+f(0)f1)=l(f−f(0)f1)+f(0)l(f1)=cf(0).
|
Furthermore, using the intertwining
properties (44) and (20):
[Mv0*H∘ WF
f](t) | = | [∞(1,−t) ∘ Mv0*H ∘ WF
f](0) |
| = | [Mv0*H ∘ WF ∘ ∞(1,−t) f](0) |
| = | l ( ∞(1,−t) f) |
| = | c[ ∞(1,−t) f](0) |
| = | cf(t).
|
|
This finishes the proof.
To get the classical statement we need the following lemma.
Lemma 47
For w(
t)∈
L1(ℝ)
, define
the fiducial functional on B:
F(f)= | ∫ | | f(t) w(t) dt.
(35) |
Then F satisfies to the condition (and thus conclusions) of
Prop. 46.
Proof.
Let f be a continuous bounded function such that
f(0)=0. For ε>0 chose
- δ>0 such that | f(t) |<ε for
all | t |<δ;
- M>0 such that
∫| t |>M| w(t) | dt<ε.
Then, for a<δ/M, we have the estimation:
| = | | ⎪
⎪
⎪
⎪
⎪
⎪ | ∫ | | f | ⎛
⎝ | at | ⎞
⎠ | w(t) dt | ⎪
⎪
⎪
⎪
⎪
⎪ |
|
| ≤ | ⎪
⎪
⎪
⎪
⎪
⎪ | ∫ | | f | ⎛
⎝ | at | ⎞
⎠ | w(t) dt | ⎪
⎪
⎪
⎪
⎪
⎪ |
+ | ⎪
⎪
⎪
⎪
⎪
⎪ | ∫ | | f | ⎛
⎝ | at | ⎞
⎠ | w(t) dt | ⎪
⎪
⎪
⎪
⎪
⎪ |
|
| ≤ | ε ( | ⎪⎪
⎪⎪ | w | ⎪⎪
⎪⎪ | 1 + | ⎪⎪
⎪⎪ | f | ⎪⎪
⎪⎪ | ∞) .
|
|
|
Finally, for a uniformly continuous function g: for
ε>0 there is δ>0 such that
| g(t+b)−g(t) |<ε for all b<δ and
t∈ℝ. Then:
| ⎪
⎪ | F(∞(1,b) g )−F(g) | ⎪
⎪ | = | ⎪
⎪
⎪
⎪
⎪
⎪ | ∫ | |
(g(t+b)−g(t)) w(t) dt | ⎪
⎪
⎪
⎪
⎪
⎪ | ≤ ε | ⎪⎪
⎪⎪ | w | ⎪⎪
⎪⎪ | 1.
|
That demonstrates the continuity of F(∞(1,b) g ) at
b=0 and, by the group property, at any other point as well.
Remark 48
A direct evaluation shows, that the constant c=
l(
f1)
from the
proof of Prop. 46 for the fiducial
functional (35) is equal to
c=∫
ℝ w(
t)
dt. Of course, for non-trivial
boundary values we need c≠ 0
. On the other hand, the
admissibility condition (6) requires
c=0
. In this sense, the classical harmonic analysis and the
traditional wavelet construction are two orthogonal parts of
the same covariant transform theory.
The table integral ∫ℝ dx/x2+1=π tells
that the “wavelet” p(t)=1/π1/1+t2 is in
L1(ℝ) with c=1, the corresponding wavelet
transform is the Poisson integral. Its boundary behaviour from
Prop. 46 is the classical result,
cf. [108]*Ch. I, Cor. 3.2.
The comparison between our demonstrations and the traditional proofs, e.g.
in [108], does not reveal any significant distinctions. We
simply made an explicit usage of the relevant group structure, which
is implicitly employed in traditional texts anyway.
4.5.1 Real and Complex Technique in Harmonic Analysis
There are two main approaches in harmonic analysis on the real line.
The real variables technique uses various maximal functions, dyadic
cubes and, occasionally, the Poisson integral [314]. The
complex variable technique is based on the Cauchy integral and fine
properties of analytic functions [264, 265].
Both methods seem to have clear advantages. The real variable
technique:
- does not require an introduction of the imaginary unit for a
study of real-valued harmonic functions of a real variable (Occam’s
Razor);
- allows a straightforward generalisation to several dimensions.
By contrast, access to the beauty and power of analytic functions
(e.g., Möbius transformations, factorisation of zeroes,
etc. [231]) is the main reason to use the complex variable
technique. A posteriori, a multidimensional analytic version was also
discovered [252], it is based on the monogenic
Clifford-valued functions [49].
Therefore, propensity for either techniques becomes a personal choice
of a researcher. Some of them prefer the real variable method,
explicitly cleaning out any reference to analytic or harmonic
functions [314]*Ch. III, p. 88. Others,
e.g. [233], happily combine the both techniques. However,
the reasons for switching beetween two methds at particular places may
look mysterious.
We demonstrated above that both—real and complex—techniques in
harmonic analysis have the same group-theoretical origin. Moreover,
they are complemented by the wavelet construction. Therefore, there is
no any confrontation between these approaches. In other words, the
binary opposition of the real and complex methods resolves into Kant’s
triad thesis-antithesis-synthesis: complex-real-covariant.
Last modified: October 28, 2024.