3 - The Spectral Function of SI Spaces

Michigan Math. J.
51 (2003)
The Spectral Function of Shift-Invariant Spaces

M a rc i n Bow n i k & Z i e m ow i t R z e s z o t n i k
1. Introduction
The shift-invariant spaces are closed subspaces of L2 ( R n ) that are invariant under
all shifts (i.e., integer translations). The theory of shift-invariant subspaces of
L2 ( R n ) plays an important role in many areas, most notably in the theory of wave-
lets, spline systems, Gabor systems, and approximation theory [BMM; BDR1;
BDR2; BL; Bo1; HLPS; Ji; RS1; RS2; Rz2]. The study of analogous spaces for
L2 ( T, H) with values in a separable Hilbert space H in terms of range func-
tion, often called doubly invariant spaces, is quite classical and goes back to Hel-
son [He1].
The general structure of shift-invariant (SI) spaces was revealed in the work of
de Boor, DeVore, and Ron [BDR1] with the use of fiberization techniques based
on range function. In particular, conditions under which a finitely generated SI
space has a generating set satisfying some desirable properties (e.g., stability, or-
thogonality or quasi-orthogonality) were given. This has been further developed in
the work of Ron and Shen [RS1] with the introduction of the technique of Grami-
ans and dual Gramians. The general properties of SI spaces and shift-preserving
operators have also been studied by the first author [Bo1].
The contribution of this paper is a systematic study of yet another tool in SI
spaces, apparently overlooked in the previous research, which we call the spectral
function. This function was introduced by the second author in his Ph.D. the-
sis. It was motivated by [BDR1] and is similar to the multiplicity function studied
by Baggett, Medina, and Merrill [BMM]. More precisely, to every SI subspace
of L2 ( R n ) we associate a function on R n that contains much useful information
about that space.
Although [BDR1] and Helson’s range function [He1] is the origin of this ap-
proach, it is thanks to Weiss (see [WW]) that the spectral function has a very
elementary definition. Namely, for every SI space V ⊂ L2 ( R n ), there exists a
countable family of functions 8 whose integer shifts form a tight frame with con-
stant 1 for the space V, and the spectral function of V is defined as the sum of
the squares of the Fourier transforms of the elements of 8 (see Lemma 2.3). It
can be shown that such a function is well-defined, additive on orthogonal sums,
and bounded by 1. Moreover, it behaves nicely under dilations and modulations,
which makes it useful in the study of wavelet and Gabor systems. For example, it
Received March 26, 2002. Revision received January 30, 2003.
387
388 M a rc i n Bow n i k & Z i e m ow i t R z e s z o t n i k
has already been used to show a new characterization of wavelets conjectured by

Weiss (see [Rz1]), a result originally proved in [Bo2] by applying the techniques
of [RS1; RS2]. The present paper is organized as follows.
In Section 2 we show several equivalent methods of defining the spectral func-
tion, and we study its basic properties such as behavior under dilations and modu-
lations. We also show that the spectral function can be used to characterize the ap-
proximation order of SI spaces following [BDR2; BDR3]. In Section 3 we apply
the spectral function to give a complete characterization of dimension functions,
sometimes called multiplicity functions, associated to generalized multiresolution
analyses (GMRA) and refinable spaces. This extends the results of Baggett and
Medina [BM], who considered only locally integrable multiplicity functions, as
well as the results of Speegle and the authors [BRS] regarding the wavelet dimen-
sion function. In Section 4 we show an analogue of the Calderón reproducing
formula for GMRAs and give an explicit formula for the wavelet spectral func-
tion, whose periodization is a well-studied wavelet dimension function. Finally,
in Section 5 we present an elementary proof of Rieffel’s incompleteness theorem
for Gabor systems utilizing the spectral function.
In order to define the spectral function we need to recall a few basic facts about
shift-invariant spaces.
A closed subspace V ⊂ L2 ( R n ) is called shift-invariant (SI) if for every func-
tion f ∈ V we also have Tk f ∈ V when k ∈ Z n, where Ty f (x) = f (x − y) is the
translation by a vector y ∈ R n. For any subset 8 ⊂ L2 ( R n ), let
S(8) = span{Tk ϕ : ϕ ∈ 8, k ∈ Z n }
be the SI space generated by 8. A principal shift-invariant (PSI) space is a SI
space V generated by a single function ϕ ∈ L2 ( R n ), that is, V = S({ϕ}) = S(ϕ).
A range function is any mapping
J : T n → {closed subspaces of ` 2 ( Z n )},
where T n = R n/ Z n is identified with its fundamental domain [−1/2, 1/2) n. We say
that J is measurable if the associated orthogonal projections PJ (ξ) : ` 2 ( Z n ) →
J(ξ) are operator measurable; that is, ξ 7→ PJ (ξ)v is measurable for any v ∈
` 2 ( Z n ).
GivenSany subset E ⊂ R n, let E P be the periodization of E; in other words,
E = k∈ Z n (E + k). Let τ : R n → T n be the translation projection, τ (ξ) =
P
ξ + k, where k is a unique element of Z n such that ξ + k ∈ T n. Finally, let

T : L2 ( R n ) → L2 ( T n, ` 2 ( Z n )) be an isometric isomorphism defined for f ∈
L2 ( R n ) by
Tf : T n → ` 2 ( Z n ), Tf (ξ) = (fˆ(ξ + k))k∈ Z n ,
R
where fˆ(ξ) = f (x)e−2πihx, ξi dx.
The following proposition, due to Helson [He1, Thm. 8], plays an important
role in the theory of SI spaces in L2 ( R n ). A proof of Proposition 1.1 can be also
found in [Bo1, Prop. 1.5].
Proposition 1.1. A closed subspace V ⊂ L2 ( R n ) is SI if and only if
V = {f ∈ L2 ( R n ) : Tf (ξ) ∈ J(ξ) for a.e. ξ ∈ T n }, (1.1)
The Spectral Function of Shift-Invariant Spaces 389
where J is a measurable range function. The correspondence between V and J

is one-to-one under the convention that the range functions are identified if they
are equal a.e. Furthermore, if V = S(8) for some countable 8 ⊂ L2 ( R n ), then
J(ξ) = span{T ϕ(ξ) : ϕ ∈ 8}. (1.2)
The dimension function of a SI space V ⊂ L2 ( R n ) is a mapping dimV : R n →

N ∪ {0, ∞} given by dimV (ξ) = dim J(ξ), where J is the range function corre-
sponding to V. Alternatively, the dimension function of V can be introduced as
the multiplicity function of the projection-valued measure coming from the rep-
resentation of the lattice Z n on V via translations by Stone’s theorem; see [BM;
BMM]. The equivalence of the dimension function with the corresponding mul-
tiplicity function can then be easily deduced from [He2]. Note also that for V =
Ľ2 (E ), where E is a measurable subset of R n and
Ľ2 (E ) = {f ∈ L2 ( R n ) : supp fˆ ⊂ E}, (1.3)
its dimension function is given by
X
dimV (ξ) = 1E (ξ + k). (1.4)
k∈ Z n
Finally, we need to recall a few facts about dual Gramian analysis of SI systems
introduced by Ron and Shen [RS1]. Suppose 8 ⊂ L2 ( R n ) is a countable set of
functions such that
X
|ϕ̂(ξ)|2 < ∞ for a.e. ξ ∈ R n. (1.5)
ϕ ∈8
The dual Gramian of a SI system E(8), where

E(8) = {Tk ϕ : k ∈ Z n, ϕ ∈ 8}, (1.6)
is a map G̃ from the fundamental domain T n = (−1/2, 1/2] n into self-adjoint
infinite matrices (gk, l )k, l∈ Z n defined for a.e. ξ ∈ T n by
X
G̃(ξ)k, l := ϕ̂(ξ + k) ϕ̂(ξ + l ) for k, l ∈ Z n. (1.7)
ϕ ∈8
Recall that a matrix G̃(ξ) = (G̃(ξ)k, l )k, l∈ Z n is bounded on ` 2 ( Z n ) if G̃(ξ) (given

by hG̃(ξ)ek , el i = G̃(ξ)k, l , where (ek )k∈ Z n is the standard basis of ` 2 ( Z n )) de-
fines a bounded operator on ` 2 ( Z n ). It can be shown that, for any fixed ξ ∈ T n,
{T ϕ(ξ) : ϕ ∈ 8} ⊂ ` 2 ( Z n ) is a Bessel family if and only if G̃(ξ) is a bounded
operator in ` 2 ( Z n ). Furthermore, it follows from [RS1, Thm. 3.3.5] that (1.5) is a
necessary (but not sufficient) condition for E(8) to be a Bessel family.
The following result due to Ron and Shen [RS1] characterizes when the system
of translates of a given family of functions E(8) is a frame (or Bessel family if
a = 0) in terms of the dual Gramian. See also [Bo1, Thm. 2.5(ii)].
Theorem 1.2. Suppose that 8 ⊂ L2 ( R n ) is countable and that 8 satisfies (1.5).

Then the system E(8) is a frame for a SI space S(8) with frame bounds 0 ≤ a ≤
b < ∞, that is,
X X
akf k2 ≤ |hf, Tk ϕi|2 ≤ bkf k2 for all f ∈ S(8),
ϕ ∈8 k∈ Z n
if and only if the dual Gramian G̃(ξ) satisfies

akvk2 ≤ hG̃(ξ)v, vi ≤ bkvk2 for v ∈ J(ξ) and a.e. ξ ∈ T n,
where J(ξ) is the range function of S(8) given by (1.2).
2. The Spectral Function

In this section we introduce the notion of a spectral function associated to a shift-
invariant space and then show its basic properties. The spectral function, which
was introduced and investigated in [Rz, Sec. 1.4], contains much more informa-
tion about shift-invariant spaces than the dimension function. It is a very useful
tool that enables us to show many results that seem to be otherwise inaccessible
by using the properties of dimension function alone.
Definition 2.1. Suppose V ⊂ L2 ( R n ) is SI with the range function J(ξ) and

the corresponding projection PJ (ξ). The spectral function of V is a measurable
mapping σV : R n → [0, 1] given by
σV (ξ + k) = kPJ (ξ)ek k2 for ξ ∈ T n and k ∈ Z n, (2.1)
where {ek } k∈ Z n denotes the standard basis of ` 2 ( Z n ) and T n = [−1/2, 1/2) n.
Note that σV (ξ) is well-defined for a.e. ξ ∈ R n, since {k + T n : k ∈ Z n } is a parti-

tion of R n. Moreover, there is a simple relationship between the spectral and the
dimension function: X
dimV (ξ) = σV (ξ + k). (2.2)
k∈ Z n
P P
Indeed, k∈ Z n σV (ξ + k) = k∈ Z n kPJ (ξ)ek k2 = dim Ran(PJ (ξ)) = dimV (ξ).
Since our definition is rather abstract, we present a description of the spectral
function of a general SI space V in terms of the spectral function of orthogonal
PSI components of V.
Proposition 2.2. Let S be the set of all SI subspaces of L2 ( R n ). Then the spec-
tral function σV of V ∈ S is determined as the unique mapping
σ : S → L∞ ( R n )
that satisfies
¡P −1
|ϕ̂(ξ)|2 k∈ Z n |ϕ̂(ξ + k)|2 for ξ ∈ supp ϕ̂,
σS(ϕ)(ξ) = (2.3)
0 otherwise,
which is additive with respect to the orthogonal sums; that is,
M X
V = Vi for some Vi ∈ S H⇒ σV = σVi . (2.4)
i∈ N i∈ N
Proposition 2.2 is a consequence of yet another description of the spectral function

of SI space V in terms of the Fourier transform of a system of functions 8 whose
shifts form a tight frame for V. Lemma 2.3 can also serve as an alternative defini-
tion of the spectral function [Rz1; Rz2]. Indeed, a direct calculation involving the
standard Gabor orthonormal basis (e 2πihx, j i 1T n (x − k))j, k∈ Z n shows that formula
(2.5) is well-defined and independent of the choice 8 (see [Rz2, Thm. 1.8]).
Lemma 2.3. Suppose a SI space V ⊂ L2 ( R n ) is generated by the shifts of a

countable family 8 ⊂ V ; that is, let V = S(8). If E(8) = {Tk ϕ : k ∈ Z n, ϕ ∈ 8}
forms a tight frame with constant 1 for the space V, then
X
σV (ξ) = |ϕ̂(ξ)|2. (2.5)
ϕ ∈8
In particular, (2.5) does not depend on the choice of 8 as long as E(8) is a tight
frame with constant 1 for V.
Proof. Let J(ξ) be the range function of V and let PJ (ξ) be the corresponding or-
thogonal projection onto J(ξ). By [Bo1, Thm. 2.5(ii)], E(8) is a tight frame with
constant 1 for V if and only if {T ϕ(ξ) : ϕ ∈ 8} is a tight frame with constant 1 for
J(ξ) for a.e. ξ. Therefore, for a.e. ξ ∈ T n,
X
kvk2 = |hv, T ϕ(ξ)i|2 for all v ∈ J(ξ).
ϕ ∈8
Hence X
kP(ξ)vk2 = |hv, T ϕ(ξ)i|2 for all v ∈ ` 2 ( Z n )
ϕ ∈8
and, in particular, for any k ∈ Z n we have

X X
σV (ξ + k) = kP(ξ)ek k2 = |hek , T ϕ(ξ)i|2 = |ϕ̂(ξ + k)|2.
ϕ ∈8 ϕ ∈8
This shows (2.5) and completes the proof of Lemma 2.3.

Remark. Lemma 2.3 also shows that there is a close connection between spec-
tral functions and dual Gramians. Indeed, for a given SI space V and a family
8 ⊂ V such that E(8) is a tight frame with constant 1 for V, we can consider
the corresponding dual Gramian G̃(ξ) given by (1.7). By Theorem 1.2, G̃(ξ) re-
stricted to J(ξ) is an identity on J(ξ). Since the dual Gramian G̃(ξ) is self-adjoint
and since ker G̃(ξ) = J(ξ) ⊥ , it follows that G̃(ξ) is just the orthogonal projection
onto J(ξ), that is, G̃(ξ) = PJ (ξ). Therefore, by (1.7), the spectral function of V
represents the diagonal entries of the dual Gramian of E(8).
Proof of Proposition 2.2. It suffices to show that the spectral function σ satisfies
(2.3) and (2.4). Indeed, any mapping σ : S → L∞ ( R n ) satisfying
L (2.3) and (2.4)
is unique, since any SI space can be decomposed as V = ϕ ∈8 S(ϕ) for some
countable family 8 ⊂ L2 ( R n ) (see e.g. [Bo1, Thm. 3.3] and [Rz2, Thm. 1.2]). To
see (2.3), recall from [BL; BDR1] that if S(ϕ) is a PSI space then the function ϕ 0
given by
¡P
2 −1/2
|ϕ̂(ξ)| k∈ Z n |ϕ̂(ξ + k)| for ξ ∈ supp ϕ̂,
ϕ̂ 0 (ξ) =
0 otherwise,
is the quasi-orthogonal generator for S(ϕ), meaning that E(ϕ 0 ) is a tight frame
with constant 1 for S(ϕ).LHence, by Lemma 2.3, we have (2.3).
Next, suppose
L V = i∈ N Vi for some SI spaces Vi . We can decompose each
Vi as Vi = ϕ ∈8i S(ϕ) for some 8i ⊂ Vi such that E(8i ) forms a tight frame
with constant 1Sfor Vi , i ∈ N. Since E(8) forms a tight frame with constant 1 for
V, where 8 = i∈ N 8i , by Lemma 2.3 we have
X XX X
σV (ξ) = |ϕ̂(ξ)|2 = |ϕ̂(ξ)|2 = σVi (ξ).
ϕ ∈8 i∈ N ϕ ∈8i i∈ N
This completes the proof of Proposition 2.2.
Next we will show that the spectral function behaves nicely with respect to the
action of modulations and dilations. This will be relevant in our study of Gabor
systems and wavelets—or, more generally, framelets. Recall that the modulation
by a vector a ∈ R n of f ∈ L2 ( R n ) is given by
Ma (f )(x) = e 2πiha, xi f (x).
The dilation by an n × n nonsingular matrix A of f ∈ L2 ( R n ) is given by
DA f (x) = |det A|1/2f (Ax).
We restrict our attention to dilations A preserving the lattice Z n because this is ex-
actly when in general we can expect that DAV is SI (with respect to the action of
Z n ) if V is SI.
Theorem 2.4. Let V ⊂ L2 ( R n ) be SI and let A be an n × n integer matrix with

det A 6 = 0. Then DA(V ) is SI and
σDA(V )(ξ) = σV ((A∗ )−1ξ), (2.6)
where A∗ is the transpose of A. Likewise, for any a ∈ R n, Ma (V ) is SI and
σMa (V )(ξ) = σV (ξ − a). (2.7)
L
Proof. First we decompose V as the orthogonal sum V = i∈ N S(ϕ i ), where
ϕ i is a quasi-orthogonal generator
L of S(ϕ i ). Since D A is an unitary operator on
L2 ( R n ) we have DAV = i∈ N DA S(ϕ i ), and by Proposition 2.2 it suffices to
show that
σDA S(ϕ)(ξ) = σS(ϕ)((A∗ )−1ξ), (2.8)
where ϕ is a quasi-orthogonal generator of S(ϕ).
Let L be a set of |det A| representatives of different cosets of Z n/AZ n. For
l ∈ L, define 8l ∈ L2 ( T n, ` 2 ( Z n )) by
∗ )−1(ξ +k), li
8l (ξ) = T (DA Tl ϕ)(ξ) = (|det A|−1/2 ϕ̂((A∗ )−1(ξ + k))e−2πih(A )k∈ Z n .
Let D be any set of |det A| representatives of different cosets of Z n/A∗ Z n. For d ∈
D, define 9d ∈ L2 ( T n, ` 2 ( Z n )) by

ϕ̂((A∗ )−1(ξ + k)) if k ∈ d + A∗ Z n,
9d (ξ)(k) =
0 otherwise.
For l ∈ L, we have
∗ −1
X ∗ −1
8l (ξ) = e−2πih(A ) ξ, li |det A|−1/2 e−2πih(A ) d, li 9d (ξ).
d∈D
Because the |det A| × |det A| matrix

∗ )−1d, li
(|det A|−1/2 e−2πih(A )l∈L,d∈D
is unitary, by a simple calculation we have
X X
|hv, 8l (ξ)i|2 = |hv, 9d (ξ)i|2 for all v ∈ ` 2 ( Z n ). (2.9)
l∈L d∈D
Since 9d (ξ) ⊥ 9d 0 (ξ) for d 6= d 0 ∈ D and since k9d (ξ)k is either 0 or 1, the
system {9d (ξ) : d ∈ D} is a tight frame with constant 1 for its span. By (2.9),
{8l (ξ) : l ∈ L} is also a tight frame with constant 1 for the space J(ξ) =
span{8l (ξ) : l ∈ L} = span{9d (ξ) : d ∈ D}, where J(ξ) is the range func-
tion of DA S(ϕ). Therefore, for every k ∈ Z n,
X X
σDA S(ϕ)(ξ + k) = kPJ(ξ) ek k2 = |h8l (ξ), ek i|2 = |h9d (ξ), ek i|2
l∈L d∈D
= |ϕ̂((A ) (ξ + k))| = σS(ϕ)((A∗ )−1(ξ + k)).

∗ −1 2
This shows (2.8) and therefore (2.6).

The case of modulations is much easier, since by Proposition 2.2 it suffices to
show σMa S(ϕ)(ξ) = σS(ϕ)(ξ − a), where ϕ is a quasi-orthogonal generator of S(ϕ).
Since Ma ϕ is a quasi-orthogonal generator of S(Ma ϕ) = Ma S(ϕ), we have
[
σMa S(ϕ)(ξ) = |M a ϕ (ξ)| = |ϕ̂(ξ − a)| = σS(ϕ)(ξ − a),
2 2
which completes the proof of Theorem 2.4.
As an immediate consequence of (2.2) and Theorem 2.4, we obtain the following

corollary.
Corollary 2.5. Suppose that V ⊂ L2 ( R n ) is SI and that A is an n × n integer

matrix with det A 6 = 0. Then
X
dimDAV (ξ) = dimV ((A∗ )−1(ξ + d )), (2.10)
d∈D
where D is the set of |det A| representatives of different cosets of Z n/(A∗ Z n ).
Example 1. Given a measurable set E ⊂ R n, let V = Ľ2 (E ). Then σV =

1E . Indeed, consider the family of functions (ϕ k )k∈ Z n given by ϕ̂ k = 1E∩(k+ T n ) .
Clearly ϕ k is a quasi-orthogonal generator of S(ϕ k ), and S(ϕ k ) ⊥ S(ϕ k 0 ) for k 6 =

k 0 ∈ Z n. Now it suffices to invoke Lemma 2.3. Conversely, if V is any SI space
such that σV = 1E for some measurable E ⊂ R n, then necessarily V = Ľ2 (E ).
Indeed, let J(ξ) be the range function of V. Since

1 if ξ + k ∈ E,
kPJ (ξ)ek k2 = σV (ξ + k) = 1E (ξ + k) =
0 if ξ + k ∈ / E,
it follows that

ek if ξ + k ∈ E,
PJ (ξ)ek =
0 if ξ + k ∈
/ E,
and hence J(ξ) = span{ek : ξ + k ∈ E} is the range function of Ľ2 (E ). Therefore,
V = Ľ2 (E ).
Example 2. There exist distinct SI spaces that have identical spectral functions.
Indeed, let ϕ 0 , ϕ 1 ∈ L2 ( R) be given by ϕ̂ 0 = 2−1/2 (1 (0,1) + 1 (1, 2) ) and ϕ̂ 1 =
2−1/2 (1 (0,1) − 1 (1, 2) ). Then ϕ i is a quasi-orthogonal generator of Vi = S(ϕ i ); that
is, E(ϕ i ) is a tight frame of S(ϕ i ) with constant 1. Hence σV 0 = σV1 = 2−11 (0, 2) ,
but V0 ⊥ V1 .
We can now collect the main properties of the spectral function into a single
proposition.
Proposition 2.6. Let S be the set of all SI subspaces of L2 ( R n ). Then, for

V, W ∈ S, the spectral function satisfies the following properties:
(a) 0 ≤ σL V (ξ) ≤ 1; P
(b) V = i∈ N Vi (Vi ∈ S) H⇒ σV (ξ) = i∈ N σVi (ξ);
(c) V ⊂ W H⇒ σV (ξ) ≤ σW (ξ);
(d) V ⊂ W H⇒ (V = W ⇐⇒ σV (ξ) = σW (ξ));
(e) σV (ξ) = 1E (ξ) ⇐⇒ V = Ľ2 (E );
(f )σMa (V )(ξ) = σ(ξ − a), where Ma is a modulation by a ∈ R n ;
(g) σDAV (ξ) = σV ((A∗ )−1ξ), where DA is a dilation by nonsingular integer ma-
trix A; P
(h) dimV (ξ) = k∈ Z n σV (ξ + k).
We will also need the following approximation lemma.
Lemma 2.7. Let V be a SI space and let (Vj )j ∈ N be a sequence of SI spaces.

Suppose that P Vj → P V strongly as j → ∞, where P V denotes the orthogonal
projection onto V ; that is, for every f ∈ L2 ( R n ), let kP Vj f − P V f k2 → 0 as
j → ∞. Then, for any measurable set E ⊂ R n with finite Lebesgue measure,
Z
|σVj (ξ) − σV (ξ)| dξ → 0 as j → ∞. (2.11)
E
In particular, there exists a subsequence (j k )k∈ N such that

lim σVj k (ξ) = σV (ξ) for a.e. ξ ∈ R n. (2.12)
k→∞
Proof. Let J and Jj (j ∈ N ) be the range functions corresponding to V and Vj , re-

spectively. Denote the corresponding projections by PJ (ξ) and PJj (ξ). Then, for
any f ∈ L2 ( R n ),
T (P V f )(ξ) = PJ (ξ)(Tf (ξ)) for a.e. ξ ∈ T n (2.13)
by [He1, p. 58; Bo1, Lemma 1.4]. Because
kP Vj f − P V f k2 = kT (P Vj f − P V f )k2
Z
= kT (P Vj f )(ξ) − T (P V f )(ξ)k2` 2 dξ
Tn
Z
= kPJ (ξ)(Tf (ξ)) − PJj (ξ)(Tf (ξ))k2` 2 dξ,
Tn
it follows that, for any k ∈ Z n, we obtain
Z
0 = lim kPJ (ξ)(ek ) − PJj (ξ)(ek )k2` 2 dξ
j →∞ T n
Z
≥ lim |σV (ξ)1/2 − σVj (ξ)1/2 |2 dξ
j →∞ k+ T n
Z
1
≥ lim |σV (ξ) − σVj (ξ)|2 dξ
4 j →∞ k+ T n
Z 1/2
1
≥ lim |σV (ξ) − σVj (ξ)| dξ
4 j →∞ k+ T n
by taking f ∈ L2 ( R n ) such that Tf (ξ) = ek for all ξ ∈ T n. This shows (2.11).
Finally, (2.12) is a consequence of (2.11) and the standard diagonal subsequence
argument.
The following lemma provides another way of looking at the spectral function.
Lemma 2.8. Suppose V is SI and K ⊂ R n is a measurable set such that τ | K is
one-to-one. Then Z
kP V (1̌K )k2 = σV (ξ) dξ, (2.14)
K
where P V is an orthogonal projection onto V.
Proof. Let f = 1̌K . Then, for any ξ ∈ T n,

el if there is an l ∈ Z n such that ξ + l ∈ K,
Tf (ξ) =
0 otherwise.
Therefore, by (2.13),
Z XZ
kP V (f )k2 = kPJ (ξ)(Tf (ξ))k2 dξ = kPJ (ξ)(el )k2 1K (ξ + l ) dξ
Tn l∈ Z n Tn
XZ Z
= σV (ξ + l )1K (ξ + l ) dξ = σV (ξ) dξ,
l∈ Z n Tn K
where J is the range function of V.

As a consequence of Lemma 2.8 and the Lebesgue differentiation theorem, we

have another formula for the spectral function:
kP V (1̌B(ξ, r) )k2
σV (ξ) = lim+ for a.e. ξ ∈ R n, (2.15)
r→0 |B(ξ, r)|
where B(ξ, r) denotes the ball with center ξ and radius r.
Approximation Orders of SI Spaces. As an example of the utility of the spec-

tral function, we will show how it can be used to characterize approximation
orders of SI spaces. In order to state the main result we recall a few basic facts
from [BDR1; BDR2].
Suppose V ⊂ L2 ( R n ) is a closed subspace. For any h > 0, define the scaled
space
V h := {g(·/h) : g ∈ V } = Dh−1 (V ),
where Dh−1 is the dilation operator by a diagonal matrix h−1 Id. For any closed
subspace V ⊂ L2 ( R n ) and a function f ∈ L2 ( R n ), we define the approximation
error as
E(f, V ) := inf{kf − gk : g ∈ V }.
Given k > 0, we say that the space V provides approximation order k if there is
a constant C > 0 such that
E(f, V h ) ≤ Ch k kf kH k ( R n ) for all f ∈ H k( R n ), (2.16)
where H k( R n ) is the Sobolev space
H k( R n ) = {f ∈ L2 ( R n ) : kf kH k ( R n ) = k(1 + | · |) k fˆk < ∞}.
We say that V provides density order k if, for a given k ≥ 0, for every f ∈ H k( R n )
we have (in addition to (2.16))
E(f, V h ) = o(h k ) as h → 0. (2.17)
Given φ ∈ L2 ( R n ), we define the function 3 φ : T n → [0, 1] by
1/2
|φ̂(ξ)|2
3 φ (ξ) := 1 − P = (1 − σS(φ)(ξ))1/2, (2.18)
k∈ Z n | φ̂(ξ + k)|2
where 0/0 should be interpreted as 0. De Boor, DeVore, and Ron [BDR2, Thms.
1.6, 1.7] showed that the approximation and density orders of a PSI space S(φ)
are characterized by the behavior of the function ξ 7→ 3 φ (ξ)/|ξ| k at the origin.
Moreover, they also proved the following remarkable result showing that approx-
imation by arbitrary (closed) SI subspaces of L2 ( R n ) can be reduced to the case
of PSI spaces.
Theorem 2.9 [BDR2, Thm. 1.9]. Suppose that V ⊂ L2 ( R n ) is SI and that

k > 0. Then the following statements are equivalent:
(i) V provides approximation order k;
(ii) there exists a φ ∈ V such that S(φ) provides approximation order k,
(iii) S(P V (φ 0 )) provides approximation order k, where φ 0 is the sinc-function,

φ̂ 0 = 1T n , and P V is the orthogonal projection onto V.
Moreover, the same is true when k ≥ 0 and the term “approximation order” is
replaced by “density order”.
As a consequence of this result, we obtain the following characterization of ap-

proximation and density orders of SI spaces in terms of the spectral function.
Theorem 2.10. A SI space V ⊂ L2 ( R n ) provides approximation order k > 0 if

and only if there is a C > 0 such that
σV (ξ) ≥ 1 − C|ξ|2k for a.e. ξ ∈ T n. (2.19)
The space V provides density order k ≥ 0 if and only if (2.19) holds and
Z
1
lim n |ξ|−2k(1 − σV (ξ)) dξ = 0. (2.20)
h→0 h hT n
Q
Proof. Let φ 0 (x1, . . . , x n ) = ni=1 sin(πx i )/(πx i ) ∈ L2 ( R n ) be the sinc-function;
that is, φ̂ 0 = 1T n . Let φ 1 = P V (φ 0 ) be the orthogonal projection of φ 0 onto V.
Then, by Theorem 2.9, the approximation order and the density order provided by
V are the same as the approximation order and the density order provided by its
PSI subspace S(φ 1 ).
Note that
σS(φ 1 )(ξ) = σV (ξ) for a.e. ξ ∈ T n. (2.21)
Indeed, take any ξ ∈ T n. By (2.13),
T (φ 1 )(ξ) = T (P V (φ 0 ))(ξ) = PJ (ξ)(T (φ 0 )(ξ)) = PJ (ξ)(e 0 ).
Hence, if T (φ 1 )(ξ) 6 = 0 then by (2.3) we have
|φ̂ 1(ξ)|2 |hPJ (ξ)e 0 , e 0 i|2
σS(φ 1 )(ξ) = = = kPJ (ξ)(e 0 )k2 = σV (ξ).
kT (φ 1 )(ξ)k2 kPJ (ξ)(e 0 )k2
If T (φ 1 )(ξ) = 0 then clearly σS(φ 1 )(ξ) = σV (ξ) = 0. This shows (2.21).
By [BDR2, Thm. 1.6], S(φ 1 ) provides approximation order k > 0 if | · |−k3 φ 1
is in L∞ ( T n ). Combining this with (2.18) and (2.21) shows (2.19). Likewise, by
[BDR2, Thm. 1.7], S(φ 1 ) provides density order k ≥ 0 if | · |−k3 φ 1 is in L∞ ( T n )
and Z
lim h−n |ξ|−2k [3 φ 1 (ξ)] 2 dξ = 0.
h→0 hT n
Combining this with (2.18) and (2.21) shows (2.20), which completes the proof of
Theorem 2.10.
3. The Dimension Function of a GMRA

As one of the applications of the spectral function studied in Section 2, we extend
a result from [BM] by providing a characterization of dimension functions as-
sociated with an arbitrary generalized multiresolution analysis (GMRA). In other
words, we characterize all multiplicity functions associated with the core subspace
of a GMRA. This extends a result of Baggett and Merrill [BM], who considered
only locally integrable multiplicity functions. In this work we not only allow non-
integrable multiplicity functions, we also allow functions equal to ∞ on a set of
nonzero measure. As an immediate consequence, we also obtain a characteriza-
tion of wavelet dimension functions shown by Speegle and the authors in [BRS].
We start by recalling the notion of a GMRA, which has been studied by a num-
ber of authors [Ba2; BM; BMM; BL; Bo3; HLPS; LTW].
Definition 3.1. Let A be a fixed n × n integer expansive dilation matrix (i.e.,

for all eigenvalues λ of A, |λ| > 1). We say that a sequence of closed subspaces
(Vj )j ∈ Z ⊂ L2 ( R n ) is a generalized multiresolution analysis (GMRA) if
[ \
Vj ⊂ Vj +1, DAVj = Vj +1, Vj = L2 ( R n ), Vj = {0},
j ∈Z j ∈Z
and V0 is SI. The space V0 is often called a core space of (Vj )j ∈ Z . An (orthonor-
mal) wavelet is a collection 9 = {ψ 1, . . . , ψ L } such that the system
{DAj Tk ψ l : j ∈ Z n, k ∈ Z n, l = 1, . . . , L}
is an orthonormal basis of L2 ( R n ).
There is a close relationship between wavelets and GMRAs. For any orthonormal
wavelet 9 we can associate a GMRA (Vj )j ∈ Z by setting
Vj = span{DAi Tk ψ l : i < j, k ∈ Z n, l = 1, . . . , L}.

Conversely, [BMM] characterized GMRAs that can be generated by orthonor-
mal wavelets. These are precisely GMRAs (Vj )j ∈ Z such that the dimension func-
tion dimV 0 (ξ) of the core space V0 is finite for a.e. ξ and satisfies the consistency
equation
X
dimV 0 ((A∗ )−1(ξ + d )) = dimV 0 (ξ) + L for a.e. ξ,
d∈D
where D is the set of |det A| representatives of different cosets of Z n/A∗ Z n. Another

related result is the characterization in [LTW] of Riesz wavelets that generate
GMRAs.
A key ingredient of a GMRA is its core space V0 , which uniquely determines
the subspaces Vj = DAj V0 , j ∈ Z . Since the core space is a refinable space, our
main goal is to give a complete characterization of dimension functions of refin-
able spaces; see Theorem 3.2. The proof of this result will follow the ideas from
[BRS]. Recall that a SI space V ⊂ L2 ( R n ) is said to be refinable (with respect to
the expansive dilation A) if V ⊂ DAV.
Theorem 3.2. Suppose a SI space V ⊂ L2 ( R n ) is refinable, that is, V ⊂ DAV.

Then the dimension function of V, D(ξ) := dimV (ξ), satisfies:
D : R n → N ∪ {0, ∞} is a measurable Z n -periodic function;

(D1) P
−1 n
(D2) d∈D D(B (ξ + d )) ≥ D(ξ) for a.e. ξ ∈ R , where D is the set of repre-
sentatives of different cosets of Z /B Z (B = A∗ ); and
P
n n
n
(D3) k∈ Z n 11(ξ + k) ≥ D(ξ) for a.e. ξ ∈ R , where
1 = {ξ ∈ R n : D(B −j ξ) ≥ 1 for j ∈ N ∪ {0}}. (3.1)

Conversely, suppose that D satisfies (D1), (D2), and (D3). Then there exists
a refinable space V such that dimV (ξ) = D(ξ) for a.e. ξ ∈ R n. Furthermore, V
satisfies
\
DAj V = {0}, (3.2)
j ∈Z
[ [
DAj V = Ľ2 B j1 . (3.3)
j ∈Z j ∈Z
Proof of Theorem 3.2: Necessity of (D1)–(D3). Clearly, the dimension function

of any SI space V must satisfy (D1). Condition (D2) is a consequence of Corol-
lary 2.5, since V ⊂ DAV implies that dimV (ξ) ≤ dimDAV (ξ). To show (D3), note
that V ⊂ DAV implies that σV (ξ) ≤ σV (B −1ξ) by Proposition 2.6(c) and (g). By
Proposition 2.6(h), σV (ξ) 6 = 0 implies dimV (ξ) 6 = 0 and hence dimV (ξ) ≥ 1.
Thus
1 = {ξ ∈ R n : dimV (B −j ξ) ≥ 1 for j ∈ N ∪ {0}}
⊃ {ξ ∈ R n : σV (ξ) 6 = 0} = supp σV .
Therefore, by Proposition 2.6(a) and (h),
X X
11(ξ + k) ≥ σV (ξ + k) = dimV (ξ),
k∈ Z n k∈ Z n
which shows (D3).
The key role in the proof of the sufficiency of (D1)–(D3) in Theorem 3.2 is played
by Lemma 3.3.
Lemma 3.3. Suppose that a function D satisfies (D1), (D2), and (D3). Then
there exists a measurable set S ⊂ R n such that
X
D(ξ) = 1S (ξ + k), (3.4)
k∈ Z n
S ⊂ BS, (3.5)
\
B jS = ∅, (3.6)
j ∈Z
[ [
B jS = B j1, (3.7)
j ∈Z j ∈Z
where 1 is given by (3.1).

The proof of Lemma 3.3 is an adaptation of the proof of the characterization of di-
mension functions of wavelets that follows the constructive procedure described
in [BRS, Algo. 4.4]. The major difference with [BRS] is that we allow a func-
tion D to be infinite on some set with nonzero measure and that the consistency
equation [BRS, (D3) in Thm. 4.2] is replaced by our consistency inequality (D2).
Assuming that Lemma 3.3 holds, we can complete the proof of Theorem 3.2.
Proof of Theorem 3.2: Sufficiency of (D1)–(D3). Define a SI P space V = Ľ2 (S ),

where S is a set guaranteed by Lemma 3.3. Clearly, dimV (ξ) = k∈ Z n 1S (ξ +k) =
D(ξ) by (3.4). Hence V is refinable because V = Ľ2 (S ) ⊂ Ľ2 (BS ) = DAV.
Furthermore, V satisfies (3.2) and (3.3) by virtue of (3.6), (3.7), and the fact that
DAj V = Ľ2 (B jS ) for j ∈ Z .
To prove Lemma 3.3, we will need [BRS, Lemma 4.1] as follows.
Lemma 3.4. Let B be a dilation and let D satisfy (D1) and (D3). Let Aj =
k
{ξ ∈ T n : D(ξ) ≥ j } for j ∈ N and let {S i}i=1 (where k ∈ N is fixed ) be a col-
lection of measurable sets such that τ | S i is injective and onto A i for i = 1, ..., k.
Then
(i) there exists a measurable set G ⊂ 1 such thatS τ (G) = A1, and
k
(ii) there exists a measurable set H ⊂ 1 such that i=1 S i ∩ H = ∅ and τ (H ) =
A k+1.
Proof of Lemma 3.3 under {ξ ∈ T n : D(ξ) < ∞} with nonzero measure. Let
[
1̃ = B j1, (3.8)
j ∈Z
where 1 is given by (3.1). For any k ∈ N, let

A k = {ξ ∈ T n : D(ξ) ≥ k}. (3.9)
S idea of the proof is to construct a sequence of sets {S k } k∈ N such that S =

The
k∈ N S k satisfies (3.4)–(3.7). In particular, to guarantee (3.4), we will require
(among other things) that τ | S k be injective and onto A k for each k ∈ N.
Fix any measurable set Q ⊂ R n such that Q ⊂ BQ, τ | Q is injective, D(ξ) ≥ 1
for ξ ∈ Q, and
lim 1Q (B −j ξ) = 1 for a.e. ξ ∈ 1̃. (3.10)
j →∞
T
Since B is a dilation, Q = 1 ∩ ∞ j n
j =0 B T is an example of a set that satisfies
these properties.
We are now ready to define S1 . Let E1 = Q. For m ∈ N, define
/[ m
Ẽ m+1 = BE m Ei ∩ AP1
P
i=1
and let E m+1 ⊂ Ẽ m+1 be any measurable set such S that τ (E m+1) = τ (Ẽ m+1) and
τ | E m+1 is injective. We claim that the set S1 = m∈ N E m satisfies
Q ⊂ S1, S1 ⊂ BS1, τ | S1 is injective and onto A1 . (3.11)
This can been seen by repeating verbatim the arguments in the proof of [BRS,
Thm. 4.2] and then using Lemma 3.4(i). We continue defining sets {S k } k∈ N by
induction with the use of a technical Lemma 3.5.
Lemma 3.5. Suppose that there exist sets S1, ..., S k such that Q ⊂ S1 = P 1 and
[i
P i ⊂ BP i for i = 1, ..., k, where P i = Sj ; (3.12)
j =1
τ | S i is injective and onto A i for i = 1, ..., k; (3.13)

S i ∩ Sj = ∅ for i, j = 1, ..., k, i 6= j. (3.14)
Then there exist S k+1 such that S1, ..., S k+1 satisfy (3.12)–(3.14).
Proof. Given S1, . . . , S k satisfying (3.12)–(3.14), we define S k+1 by the inductive
procedure. Let F̃1 = (BP k \ P k ) ∩ APk+1 and
/[ m
F̃m+1 = BFm Fi ∩ APk+1,
P
(3.15)
i=1
where Fm ⊂ F̃m is any measurable set such that τ (Fm+1S ) = τ (F̃m+1) and τ | Fm+1
∞
is injective. We claim that S1, . . . , S k+1, where S k+1 = m=1 Fm satisfy (3.12)–
(3.14).
To see (3.12), it suffices to show that S k+1 ⊂ BP k+1. This follows from the fact
that F1 ⊂ BP k and Fm+1 ⊂ BFm ⊂ BS k+1 ⊂ BP k+1 for m ∈ N.
To see that τ | S k+1 is injective, suppose that ξ1, ξ 2 ∈ S n+1 and τ (ξ1 ) = τ (ξ 2 ).
Therefore, ξ1 ∈ Fj and ξ 2 ∈ Fk for some j, k ∈ N. If j < k (the case j > k is iden-
tical) then ξ 2 ∈/ FjP , which contradicts τ (ξ1 ) = τ (ξ 2 ). Thus j = k and ξ1 = ξ 2 ,
since τ | Fk is injective.
To see (3.14), it is enough to prove that S k+1 ∩ P k = ∅. Since F1 ⊂ BP k \ P k
and Fm ⊂ BFm−1 for m ≥ 2, by induction we obtain Fm ⊂ B mP k \ B m−1P k for all
m ∈ N. By (3.12) with i = k it follows that P k ⊂ B m−1P k for any m ∈ N. Hence
Fm ∩ P k = ∅, that is, S k+1 ∩ P k = ∅.
The proof of the remaining part of (3.13) (i.e., τ (S k+1) = A k+1) is much more
difficult. First note that, since τ (Fm ) ⊂ A k+1, we certainly have τ (S k+1) ⊂ A k+1.
For the reverse inclusion, we will find it useful to prove
APk+1 ∩ BS k+1
P P
⊂ S k+1. (3.16)
Indeed, if ξ ∈ APk+1 ∩ P
BS k+1, −1
then B ξ + l ∈ Fm for some m ∈ N and l ∈
Z n. Since ξ 0 := ξ + Bl ∈ BFm and ξ ∈ APk+1, we obtain ξ 0 ∈ BFm ∩ APk+1. If
Sm P P P
ξ0 ∈/ i=1 Fi , then by (3.15) we have ξ 0 ∈ F̃m+1 ⊂ S k+1 and hence ξ ∈ S k+1 .
0 P 0 P P
However, if ξ ∈ Fi for some i = 1, . . . , m, then ξ ∈ S k+1 and hence ξ ∈ S k+1 .
This shows (3.16).
Continuing with the proof of (3.13), it remains to show that A k+1 ⊂ τ (S k+1).
By Lemma 3.4(ii) there is a set H ⊂ 1 such that H ∩ P k = ∅, τ (H ) = A k+1,
and D(B −j ξ) ≥ 1 for every j ≥ 0 and ξ ∈ H. Therefore, all we have to prove is
P
τ (H ) ⊂ τ (S k+1), that is, H ⊂ S k+1 .
P
We split the proof of H ⊂ S k+1 into two cases. First, we consider all ξ ∈ H
that, for every j ≥ 0, D(B −j ξ) ≥ k + 1; that is, we consider the set R :=
such T
H∩ ∞ j P −j
j =0 B A k+1. For a.e. ξ ∈R, it follows from (3.10) that B ξ ∈ Q for some
j ≥ 1. Since Q ⊂ S1 ⊂ P k , we can consider j 0 = min{j ∈ N : B −j ξ ∈ P k }. Since
H ∩ P k = ∅, it follows that B −j 0 +1ξ ∈ BP k \ P k . Moreover, since B −j 0 +1ξ ∈
APk+1, we obtain
B −j 0 +1ξ ∈ (BP k \ P k ) ∩ APk+1 = F̃1 ⊂ S k+1
P
.
S∞ P
Hence R ⊂ j =0 B jS k+1. Therefore,
∞
[ ∞
[ j
\
R= (R ∩ B jS k+1
P
)⊂ H∩ B iAPk+1 ∩ B jS k+1
P
.
j =0 j =0 i=0
But (3.16) implies that, for j ≥ 0,

j
\
B iAPk+1 ∩ B jS k+1
P P
⊂ S k+1, (3.17)
i=0
P
so R ⊂ S k+1 .
T
Next we consider the set R̃ := H \ ∞ j P
j =0 B A k+1 and wish to show that R̃ ⊂
P P
S . Take any ξ ∈ R̃. Since ξ ∈ H ⊂ A k+1, we can find j 0 ≥ 0 such that ξ ∈
Tk+1
j0 j P
/ B j 0 +1APk+1. To prove that ξ ∈ S k+1
j =0 B A k+1 and ξ ∈
P
, it is enough to show that
j0 P P
ξ ∈ B S k+1 and then use (3.17) with j = j 0 . To see why ξ ∈ B j 0 S k+1 , observe
−j 0
that D(B ξ) ≥ k + 1. Hence, by the consistency inequality (D2),
X
k + 1 ≤ D(B −j 0 ξ) ≤ D(B −j 0 −1ξ + B −1d ), (3.18)
d∈D
where D is the set of representatives of different cosets of Z n/B Z n. Without loss

of generality we can assume that 0 ∈ D. For each d ∈ D we denote k(d ) =
T )
D(B −j 0 −1ξ + B −1d ). Then τ (B −j 0 −1ξ + B −1d ) ∈ k(d j =1 Aj . Moreover, since d =
−j 0 −1 j 0 +1 P
0 ∈ D, τ (B ξ) ∈ A k(0) , and ξ ∈ / B A k+1, we obtain k(0) ≤ k. Since ξ ∈
H, it follows that D(B −j 0 −1ξ) ≥ 1, and we obtain k(0) ≥ 1. By (3.13) we have
T
B −j 0 −1ξ ∈ k(0) P
j =1 Sj , that is, B
−j 0 −1
ξ +pj0 ∈ Sj , where pj0 ∈ Z n for j = 1, . . . , k(0)
are distinct by (3.14). For each d ∈ D \ {0} such that j(d ) 6 = 0, by using (3.13)
again we can find distinct pjd ∈ Z n such that B −j 0 −1ξ + pjd + B −1d ∈ Sj , where
j = 1, . . . , min(k(d ), k).
Thus, for each d ∈ D such that k(d ) 6 = 0, we have
B −j 0 ξ + Bpjd + d ∈ BSj ⊂ BP k for j = 1, . . . , min(k(d ), k). (3.19)
We claim that this gives us at least k + 1 distinct elements of BP k . Indeed, if
0
Bpjd + d = Bpjd0 + d 0 for some j = 1, . . . , min(k(d ), k) and j 0 = 1, . . . ,
min(k(d 0 ), k), then d − d 0 ∈ B Z n, hence d = d 0. Also, for fixed d ∈ D, we
have pjd 6 = pjd0 for j 6 = j 0. What remains to check is that the number of elements
P
in (3.19), which is equal to d∈D min(k(d ), k), is ≥ k + 1. This is indeed the
case by 1 ≤ k(0) ≤ k and (3.18).
By the induction hypothesis (3.13), at least one of the elements in (3.19) must lie
in the complement of P k . Hence there is a p ∈ Z n, and B −j 0 ξ + p ∈ BP k \ P k . In
addition, since B −j 0 ξ ∈ APk+1, we have B −j 0 ξ +p ∈ APk+1. Therefore, B −j 0 ξ +p ∈
P P P
F̃1 ⊂ S k+1 ; that is, ξ ∈ B j 0 S k+1 and hence ξ ∈ S k+1. Since both R and R̃ are con-
P P P
tained in S k+1, it follows that H ⊂ S k+1 and so A k+1 ⊂ S k+1 , which completes
the proof of (3.13) by the induction. We have thus shown the existence of the sets
{S k } k∈ N satisfying (3.12)–(3.14), which completes the proof of Lemma 3.5.
Finally, we are
S ∞ready to continue the proof of Lemma 3.3. By Lemma 3.5, define
the set S = k=1 S k . We claim that S satisfies (3.4)–(3.7). To see (3.4),P
it suffices
to show that the equality in (3.4) holds for a.e. ξ ∈ T n. By (3.14), 1S = ∞ k=1 1S k ,
and by (3.13),
X ∞ X
X
1S (ξ + l ) = 1S k (ξ + l )
l∈ Z n k=1 l∈ Z n
X∞ ∞
X
= 1τ (S k )(ξ) = 1A k (ξ) = D(ξ) for a.e. ξ ∈ T n.
k=1 k=1
For (3.5) it suffices to show that S k ⊂ BS for every k ∈ N, but this is immediate
from (3.12). For (3.7) it suffices to show that limj →∞ 1S (B −j ξ) = 1 for a.e. ξ ∈
1̃, which is an immediate consequence of Q ⊂ S and (3.10). T Finally, it remains
to show (3.6). By the contradiction, suppose that S̃ = j ∈ Z B jS has a nonzero
measure. Since S̃ = B S̃, we can partition S̃ into a countable family of subsets
{S̃ i}i∈ N such that S̃ i = B S̃ i and S̃ i has a nonzero (and hence infinite) measure for
all i ∈ N. Since B induces an ergodic endomorphism of T n (see [BRS, proof of
Thm. 5.11]), we have τ (S̃ i ) = T n. Therefore,
X X
D(ξ) = 1S (ξ + k) ≥ 1S̃ (ξ + k)
k∈ Z n k∈ Z n
XX
= 1S̃ i (ξ + k) = ∞ for a.e. ξ,
k∈ Z n i∈ N
which contradicts our initial assumption that {ξ ∈ T n : D(ξ) < ∞} has a nonzero
measure. This shows (3.6) and completes the proof of Lemma 3.3 under this
assumption.
Proof of Lemma 3.3 under D ≡ ∞. Finally, we need to construct a set S satisfy-

ing (3.4)–(3.7) for the case when the dimension function D(ξ) is constantly ∞.
Let S 0 be a set obtained from Lemma 3.3 when D ≡ 1. Any such set S 0 is called
a scaling set. By (3.5)–(3.7), the family {(B j +1S 0 ) \ (B jS 0 )}j ∈ Z forms a partition
of R n (modulo sets of measure zero), since 1 = 1̃ = R n. Define by induction a
sequence of sets {S i}∞i=0 such that, for all i = 0, 1, . . . ,
X
S i+1 ⊂ (BS i ) \ S i , 1S i (ξ + k) = 1 for a.e. ξ.
k∈ Z n
This is possible because
X X
1BS i (ξ + k) = 1S i (B −1ξ + B −1k)
k∈ Z n k∈ Z n
X X
= 1S i (B −1ξ + k + B −1d ) = |det B|.
k∈ Z n d∈D
S∞
Define the set S = i=0 S i . It remains to show (3.4)–(3.7). To see (3.4), notice
that S i ⊂ (B iS 0 ) \ (B i−1S 0 ) for i ∈ N. Hence the sets S i are pairwise disjoint and
X ∞
XX
1S (ξ + k) = 1S i (ξ + k) = ∞.
k∈ Z n k∈ Z n i=0
Observe that (3.5) follows from

∞
[ ∞
[
BS = BS 0 ∪ BS i ⊃ S 0 ∪ S i+1 = S,
i=0 i=0
S
while (3.7) is an immediate consequence of S 0 ⊂ S and j ∈ Z B jS 0 = R n. To see
T
(3.6), let c = j ∈ Z B jS ∩ ((BS 0 ) \ S 0 ). For any i ∈ N,
1 = |S i | = |((B iS 0 ) \ (B i−1S 0 ) ∩ S|
\

≥ ((B iS 0 ) \ (B i−1S 0 )) ∩ B jS = |det B|i−1c.
j ∈Z
By letting i → ∞, we conclude that c is zero and hence

\

((B iS 0 ) \ (B i−1S 0 )) ∩ j
B S = 0 for all i ∈ Z .

j ∈Z
This shows (3.6) and completes the proof of Lemma 3.3.
Finally, we are ready to show our main result.
Theorem 3.6. Suppose (Vj )j ∈ Z is a GMRA. Then the dimension function of the
core space V0 , D(ξ) := dimV 0 (ξ), satisfies (D1)–(D3) and
(D4) lim infj →∞ D(B −j ξ) ≥ 1 for a.e. ξ ∈ R n.
Conversely, if D satisfies (D1)–(D4) then there exists a GMRA (Vj )j ∈ Z such that
dimV 0 (ξ) = D(ξ) for a.e. ξ ∈ R n.
Proof. Suppose that (Vj )j ∈ Z is a GMRA. Then, by Theorem 3.2, D(ξ) = dimV 0 (ξ)
satisfies (D1)–(D3). To see (D4), by Proposition 2.6(g) and (h) it suffices to show
that
lim inf σVj (ξ) = lim inf σV 0 (B −j ξ) = 1 for a.e. ξ ∈ R n. (3.20)
j →∞ j →∞
However, by Lemma 2.7, there exists a sequence (j k )k∈ N such that

lim inf σVj k (ξ) = σL2 ( R n )(ξ) = 1 for a.e. ξ,
k→∞
because the Vj converge to L2 ( R n ) as j → ∞. Because (σVj (ξ))j ∈ Z is a non-

decreasing sequence, by Vj ⊂ Vj +1 and Proposition 2.6(c) we necessarily have
(3.20).
The converse
S is immediate by Theorem 3.2, (3.2), and (3.3), since (D4) is equiv-
alent to j ∈ Z B j1 = R n.
We conclude this section by giving an example of a GMRA whose dimension func-

tion of its core space is infinite on a set of nonzero (but not full) measure. Note
that the construction of a GMRA whose dimension function constantly equals ∞
is given in the proof of Lemma 3.3. More examples of dimension functions of
GMRAs associated with orthonormal wavelets can be found in [BRS, Sec. 5].
Example 3.7. For simplicity, assume that we work in the dimension n = 1 and
that the dilation factor A = 2. Given 0 < δ < 1/2, define a set S ⊂ R by
∞
[
S = [−1/2, 1/2] ∪ [2 j, 2 j + δ).
j =−1
T S
An elementary calculation shows that S ⊂ 2S, j ∈ Z 2 jS = ∅, and j ∈ Z 2 jS =
R . Hence the sequence (Vj )j ∈ Z , where Vj = Ľ2 (2 j S ), is a GMRA. Moreover, the
dimension function of its core space satisfies

∞ for τ (ξ) ∈ (0, δ),
dimV 0 (ξ) =
1 otherwise.
4. The Spectral Function of a GMRA
In this section we investigate properties of the spectral function associated to the

core space of a GMRA by showing the fundamental representation formula (4.1),
which is a close analogue to the Calderón reproducing formula in the theory of
wavelets. As an immediate consequence, we show an explicit formula for the
wavelet spectral function, which also gives the usual well-known formula for the
wavelet dimension function.
Theorem 4.1. Suppose (Vj )j ∈ Z is a GMRA such that {ξ ∈ R n : dimV 0 (ξ) < ∞}
has a nonzero (Lebesgue) measure. Then
∞
X
σW0 ((A∗ ) j ξ) = 1 for a.e. ξ, (4.1)
j =−∞
where W0 = V1 ª V0 . As a consequence, the spectral function of the core space

V0 can be represented as
∞
X
σV 0 (ξ) = σW0 ((A∗ ) j ξ) for a.e. ξ. (4.2)
j =1
It is intuitively clear that (4.1) should hold. Indeed, it appears

P that by Proposi-
tion 2.6 we have σWj (ξ) = σDA j W0 (ξ) = σW0 ((A∗ )−j ξ) and j ∈ Z σWj (ξ) = 1,
since M
Wj = L2 ( R n ), where Wj = Vj +1 ª Vj . (4.3)
j ∈Z
However, in general the spaces Wj are not SI (with respect to the standard lattice
Z n ) for j < 0. Indeed, for j < 0 we can be sure only that Wj is SI with respect to
a larger lattice A−j Z n. Nevertheless, this idea can be transformed into a rigorous
proof.
L ¡L ⊥
Proof of Theorem 4.1. Since V0 = j <0 Wj = j ≥0 Wj is SI and since V0 ⊕
L n
W
j ≥0 j = L2
( R ), by Proposition 2.6(b) and (g) we have
∞
X
1 = σL2 ( R n )(ξ) = σV 0 (ξ) + σWj (ξ)
j =0
∞
X
= σV 0 (ξ) + σW0 ((A∗ ) j ξ) for a.e. ξ. (4.4)
j =0
Pick any ξ 0 ∈ R n such that (4.4) holds for any ξ = Aiξ 0 , where i ∈ Z . By applying
(4.4) for ξ = Aiξ 0 and letting i → −∞, we obtain
∞
X
σW0 ((A∗ ) j ξ) ≤ 1 for a.e. ξ. (4.5)
j =−∞
Assume by way of contradiction

P ª (4.5), there is a δ > 0
that (4.1) fails. Then, by
such that E = ξ ∈ R n : j∞=−∞ σW0 ((A∗ ) j ξ) < 1 − δ has a nonzero measure.
Since A∗(E ) = E and A∗ is expansive, E must then have an infinite measure.
Clearly, we can partition E into a countable family of subsets {Ei}i∈ N such that
A∗(Ei ) = Ei and Ei has a nonzero (hence infinite) measure. Since A∗ induces an
ergodic endomorphism of T n (see [BRS, proof of Thm. 5.11]), we have τ (Ei ) =
T n. On the other hand, by (4.4), σV 0 (ξ) > δ for ξ ∈ E. Therefore, by Proposi-
tion 2.6(h) for a.e. ξ,
X XX X
dimV 0 (ξ) = σV 0 (ξ + k) ≥ 1Ei (ξ + k)σV 0 (ξ + k) ≥ δ = ∞.
k∈ Z n i∈ N k∈ Z n i∈ N
This contradicts our initial hypothesis and so proves (4.1). Equation (4.2) follows
from (4.1) and (4.4).
Theorem 4.2. Suppose that 9 = {ψ 1, ..., ψ L } ⊂ L2 ( R n ) is a semi-orthogonal

wavelet (i.e., that the affine system {DAj Tk ψ : j ∈ Z , k ∈ Z n, ψ ∈ 9} is a tight
frame with constant 1) and let Wi ⊥ Wj for i 6= j, where
Wi = span{DAi Tk ψ : k ∈ Z n, ψ ∈ 9} = DAi (S(9)).
L the spectral function of the core space V0 of the GMRA (Vj )j ∈ Z , Vj =

Then
i<j Wi , is given by
∞
XX
σV 0 (ξ) = |ψ̂((A∗ ) j ξ)|2. (4.6)
ψ∈9 j =1
Proof. By the semi-orthogonality condition, E(9) forms a tight frame with con-
P
stant 1 for W0 = S(9) and thus σW0 (ξ) = ψ∈9 |ψ̂(ξ)|2 by Lemma 2.3. By (4.4),
∞
X
σV 0 (ξ) = 1 − σW0 ((A∗ ) j ξ)
i=0
∞ X
X ∞
XX
= 1− |ψ̂((A∗ ) j ξ)|2 = |ψ̂((A∗ ) j ξ)|2,
j =0 ψ∈9 ψ∈9 j =1
P P∞
where we used the Calderón formula ψ∈9 j =−∞ |ψ̂((A∗ ) j ξ)|2 = 1 for a.e. ξ
(see e.g. [Bo2; HW]). This completes the proof of Theorem 4.2.
As a corollary to Proposition 2.6(h) and Theorem 4.2 we obtain the usual formula
for the wavelet dimension function [BRS; LTW; RS4; We].
Corollary 4.3. Suppose 9 = {ψ 1, ..., ψ L } ⊂ L2 ( R n ) is a semi-orthogonal

wavelet. Then the dimension function of the core space V0 of the GMRA (Vj )j ∈ Z
associated with 9 is given by
X ∞ X
XX
dimV 0 (ξ) = σV 0 (ξ + k) = |ψ̂((A∗ ) j(ξ + k))|2. (4.7)
k∈ Z n ψ∈9 j =1 k∈ Z n
Remark. It should be noted that neither Theorem 4.2 nor Corollary 4.3 are true
for general affine tight frame systems with constant 1. Indeed, Theorem 3.1 of
[PSWX] shows that a (dyadic) normalized tight frame wavelet ψ ∈ L2 ( R) is
P P
semi-orthogonal if and only if j∞=1 k∈ Z |ψ̂(2 j(ξ + k))|2 is integer-valued al-
most everywhere. Therefore, one can expect that the right-hand side of (4.7) is a
dimension function of V0 , and thus integer-valued, only if 9 is a semi-orthogonal
wavelet. As a result, Theorem 4.2 and Corollary 4.3 are not valid for more general
tight affine frame systems.
5. Rieffel’s Incompleteness Theorem for Gabor Systems

In this section we present an elementary proof of Rieffel’s incompleteness theo-
rem for Gabor systems that utilizes the spectral function introduced in Section 2.
We will start by giving some historical background information.
In 1986, Daubechies, Grossmann, and Meyer [DGM] constructed a Gabor
system
{g bl, am(x) = e 2πiblxg(x − am)}l, m∈ Z
forming a tight frame for L2 ( R) with g a compactly supported function in the

Schwartz class for every positive a and b such that ab < 1. Lyubarskii [Ly] and
independently Seip and Wallstén [SW] proved that the original choice of Gabor
[Ga] (i.e., g a Gaussian) gives rise to a frame of L2 ( R) if ab < 1 (this was con-
jectured by Daubechies and Grossmann in 1988).
Both of these results utilize the situation ab < 1, since for ab > 1 the Gabor
system {g bl, am }l, m∈ Z can never be complete in L2 ( R) for any choice of g ∈ L2 ( R).
This simple fact was communicated by Baggett [Ba1] and Daubechies [D], and it
follows from the work of Rieffel [Ri] on von Neumann algebras (see also [DLL]).
Even though the argument in [DLL] is rather simple, it is based completely on
[Ri], which requires a reasonable knowledge of the theory of C ∗ -algebras. This,
in turn, has motivated several researchers to find an “elementary” proof of the in-
completeness result.
Daubechies [D] was able to find a very elegant argument for the case when
ab > 1 is rational by constructing a function orthogonal to every element of
{g bl, am }l, m∈ Z . The problem of (explicitly) constructing such a function in the gen-
eral case ab > 1 is still open. In 1993 Landau [La] proved, under certain assump-
tions on the decay of g and ĝ, that {g bl, am }l, m∈ Z cannot be a frame for L2 ( R) if
ab > 1. Janssen [Ja] showed the same without any decay assumptions for a general
g ∈ L2 ( R n ); see also [CDH; RS3]. Furthermore, Ramanathan and Steger [RSt]
proved Landau’s result for irregular Gabor systems without any decay assump-
tions. Their methods allowed them to recover Rieffel’s incompleteness theorem.
They also conjectured that the incompleteness result can be extended to the case
of irregular sampling sets with a uniform density smaller than 1, which was later
disproved in [BHW] by exhibiting a counterexample based on Landau’s result
from 1960s. Another proof of Rieffel’s incompleteness theorem (again based on
von Neumann algebras) was recently given by Gabardo and Han [GH] (see also
[HWa, Thm. 3.3]).
We shall now show Rieffel’s incompleteness theorem in its full generality—that
is, for Gabor multi-systems in L2 ( R n ). The advantage of our approach is that we
do not use any results about von Neumann algebras; instead, our proof is based on
the spectral function introduced in Section 2. Moreover, we are not aware of any
other proofs of Theorem 5.1 in its full generality that do not use the machinery of
von Neumann algebras. For example, [RS3, Cor. 4.7] contains a proof of this re-
sult for compressible Gabor systems, which correspond in one dimension to the
case when ab is rational.
Theorem 5.1. If {g 1, ..., g L } ⊂ L2 ( R n ) and if A, B are n × n nonsingular ma-
trices, then the Gabor system
i n
{gBl, Am : l, m ∈ Z , i = 1, ..., L},
where
gBl, Am(x) = MBl TAm g = e 2πihBl, xig(x − Am), g ∈ L2 ( R n ), (5.1)
n
is incomplete in L ( R ) if |det A| · |det B| > L.
2
Proof. The standard dilation argument allows us to reduce the general case to
A = Id. Indeed, this follows immediately from DA(gBl, Am ) = (DA g)AT Bl, m ,
where l, m ∈ Z n. We present a proof by contradiction; that is, we shall assume that
i i=1, ..., L
the system {gBl , m } l, m∈ Z n is complete and then prove that |det B| ≤ L follows.
We start by constructing a multiresolution scheme for Gabor systems. Suppose
i i=1, ..., L
that {g 1, . . . , g L } ⊂ L2 ( R n ) and that {gBl , m } l, m∈ Z n is the corresponding Gabor
system. For every l ∈ Z n, define
i n L
Gl = span{gBl, m : m ∈ Z , i = 1, . . . , L} = S({MBl (g ), . . . , MBl (g )}).
1
(5.2)
n
We will use the lexicographic order on ( Z , <) defined by (k1, k 2 , . . . , k n ) <
(l1, l 2 , . . . , l n ) if there exist r = 1, . . . , n such that kj = lj for all j < r and k r <
lr . For any l ∈ Z n, let
]
Vl = span{gBk, m : k < l, k, m ∈ Z n } = Gk . (5.3)
k<l
n
Let S be the closed subspace of L ( R ) spanned by {gBl, m }l, m∈ Z n . Our discussion
2
is general and we do not yet assume that the system is complete. We would expect
to have the following properties of the “multiresolution analysis” (Vl )l∈ Z n :
(G1) Vk ⊂ Vl for any k < l ∈ Z n ;
n
(G2) M SBk (Vl ) = Vl+k for any k, l ∈ Z ;
(G3) Tl∈ Z n Vl is dense in S;
(G4) l∈ Z n Vl = {0};
(G5) V0 is shift-invariant, where 0 = (0, . . . , 0) ∈ Z n.
Indeed, all of these conditions except (G4) follow immediately from (5.3). How-
ever, a serious difficulty arises because (G4) does not hold in general (see Example
5.2). Nevertheless, if we assume in addition that the functions g 1, . . . , g L are band-
limited, then this easily implies that (G4) holds. Indeed, if supp ĝ 1, . . . , supp ĝ L ⊂
K for some bounded set K ⊂ R n, then Gl ⊂ Ľ2 (Bl +K) for any l ∈ Z n. Therefore,
] [
Vk = Gl ⊂ Ľ2 (Bl + K)
l<k l<k
and
\ \ [ \ [
Vk ⊂ Ľ2 (Bl + K) = Ľ2 B (l + B −1K) = {0},
k∈ Z n k∈ Z n l<k k∈ Z n l<k
and (G4) follows.

Finally, we will need an additional condition that, in general, is guaranteed to
hold only in the band-limited case. Recall that I ⊂ Z n is a beginning interval if,
for every k, l ∈ Z n, we have that k < l and l ∈ I together imply k ∈ I. It is not
hard to show that all nonempty beginning intervals of ( Z n, <) can be associated
with its ending, which consists of either
(i) elements of Z n or
(ii) elements of the form (k1, . . . , k r , ∞), where 0 ≤ r < n and k1, . . . , k r ∈ Z .
That is, for any beginning interval I ⊂ Z n, there exist r = 0, 1, . . . , n and k ∈ Z r

such that I = {l ∈ Z n : (l1, . . . , lr ) ≤ (k1, . . . , k r )}.
We claim that, for any beginning interval I ⊂ Z n of type (ii),
T S
(G6) l∈ Z n \I Vl = l∈I Vl .
The inclusion “⊃” in (G6) is obvious. To show “⊂”, suppose that a beginning in-
terval I is associated with (k1, . . . , k r , ∞) and that supp ĝ 1, . . . , supp ĝ L ⊂ K for
some bounded set K ⊂ R n. For any r > 0, consider an orthogonal projection Pr
of L2 ( R n ) onto Ľ2 ((−r, r) n ) defined by Pd ˆ
r f = 1 (−r, r) n f . For any j ∈ Z , consider
an index set Ij ,
Ij = {(k1, . . . , k r + 1, m1, . . . , m n−r ) ∈ Z n : m1 ≤ j, (m1, . . . , m n−r ) ∈ Z n−r }.
¡U
We now claim that, for sufficiently small j, Pr l∈Ij Gl = {0}. Indeed, if f ∈
U
l∈Ij Gl then
[
supp fˆ ∩ (−r, r) n ⊂ (Bl + K) ∩ (−r, r) n
l∈Ij
[
=B (l + B −1K) ∩ B −1(−r, r) n = ∅.
l∈Ij
Thus, for sufficiently small j,

\
Pr Vl ⊂ Pr (V(k1, ..., k r +1, j, 0, ..., 0) )
l∈ Z n \I
] ] ] [
⊂ Pr Gl ] Gl = Pr Gl = Pr Vl .
l∈I l∈Ij l∈I l∈I
Since r > 0 is arbitrary, the preceding formula yields (G6).

We shall use (G1)–(G6) in the same way one uses the multiresolution analysis
scheme to construct wavelets. This is exactly true in dimension n = 1, but an ad-
ditional argument based on (G6) is needed in dimensions n > 1. In our case, for
every l ∈ Z n we define a space
Wl = Vl+1 ª Vl , (5.4)
where 1 = (0, . . . , 0, 1) ∈ Z n. By (G2) we have
MBk (Wl ) = Wl+k for k ∈ Z n. (5.5)
We claim that from (G1)–(G6) it follows that
M
S= Wl . (5.6)
l∈ Z n
Indeed, by (G1), (G3), and (G4),

M \ [
S= Vl ª Vl , (5.7)
I l∈ Z n \I l∈I
where the orthogonal sum is taken over all nonempty beginning intervals I ⊂ Z n.
By (G6), the orthogonal sum in (5.7) is effectively taken only over the beginning
intervals of type (i), which clearly implies (5.6).
The decomposition (5.6) allows us to see that the spectral function of S is given
by a simple formula. In fact, we have
W0 = V1 ª V0 = (V0 ] G0 ) ª V0 .
By (5.2), G0 is a SI space with the dimension function dim G0 (ξ) ≤ L for a.e. ξ ∈
R n. Hence, by (G5), W0 is also a SI space with the dimension function that satisfies
dimW0 (ξ) ≤ dim G0 (ξ) ≤ L for a.e. ξ ∈ R n. Therefore, W0 has a quasi-orthogonal
basis 9 ⊂ W0 with cardinality at most L; that is, E(9) is a tight frame with con-
P
stant 1 for W0 (see [BDR1]). The spectral function of W0 is thus σW0 = ψ∈9 |ψ̂|2
by Lemma 2.3. Therefore, by Proposition 2.6, (5.5), and (5.6), we have
X X
σS (ξ) = σWl (ξ) = σMBl (W0 )(ξ)
l∈ Z n l∈ Z n
X X X
= σW0 (ξ − Bl ) = |ψ̂(ξ − Bl )|2 (5.8)
l∈ Z n l∈ Z n ψ∈9
for a.e. ξ ∈ R n.
The system {gBl, m }l, m∈ Z n is complete in L2 ( R n ) if and only if S = L2 ( R n ),
which by Proposition 2.6 is equivalent to having σS = 1 a.e. If we integrate this
equality over the cube B([0, 1] n ), then by (5.8) we have
Z X X
|det B| = |ψ̂(ξ − Bl )|2 dξ
B([0,1] n ) l∈ Z n ψ∈9
X XZ X
= |ψ̂(ξ)|2 dξ = kψk22.
ψ∈9 l∈ Z n B(l+[0,1] n ) ψ∈9
Now 9 is a quasi-orthogonal basis of W0 and so we must have kψk2 ≤ 1 for every

ψ ∈ 9, which proves that |det B| ≤ L.
To drop the assumption that g 1, . . . , g L are band-limited, we approximate each
g by a sequence of band-limited functions {g i, r }r∈ N ⊂ L2 ( R n ) in the L2 ( R n )-
i
norm. For example, we can define such a sequence by gd i, r = 1 bi

(−r, r) n g for r ∈ N,
and then
lim kg i − g i, r k2 = 0. (5.9)
r→∞
i, r i=1, ..., L
Let S r be the closure of the space spanned by the Gabor system {gBl , m } l, m∈ Z n
and let PS r be the orthogonal projection onto S r for r ∈ N. For every l, m ∈ Z n,
we have
i i, r i i, r
kgBl, m − gBl, m k2 = kg − g k2 .
i i=1, ..., L
Therefore, the completeness of the system {gBl , m } l, m∈ Z n and (5.9) imply that, for
n
every f ∈ L ( R ),
2
lim kPS r f − f k2 = 0.
r→∞
Thus, by Lemma 2.7 it follows that

lim σS r j (ξ) = 1 for a.e. ξ (5.10)

j →∞
for some subsequence {rj }j ∈ N . For every j ∈ N, the functions g 1, rj , . . . , g L, rj are

band-limited and so by (5.8) we have
X X
σS r j (ξ) = |ψ̂(ξ − Bl )|2,
l∈ Z n ψ∈9j
where 9j ⊂ L2 ( R n ) has cardinality ≤ L and kψk2 ≤ 1 for ψ ∈ 9j . This equality

together with (5.10) gives us
X X
lim |ψ̂(ξ − Bl )|2 = 1 for a.e. ξ ∈ R n.
j →∞
l∈ Z n ψ∈9j
Integrating this limit equality over the cube B([0, 1] n ) yields |det B| ≤ L. Indeed,
by Fatou’s lemma we have
Z X X
|det B| = lim |ψ̂(ξ − Bl )|2 dξ
B([0,1] n ) j →∞ l∈ Z n ψ∈9
j
Z X X X
≤ lim inf |ψ̂(ξ − Bl )|2 dξ = lim inf kψk22 ≤ L,
j →∞ B([0,1] n ) l∈ Z n ψ∈9 j →∞
j ψ∈9j
which concludes the proof of Theorem 5.1.
The following example shows that condition (G4) does not hold in general.
Example 5.2. We will construct a function g ∈ L2 ( R) such that the correspond-

ing Gabor system {gl, m(x) = e 2πlxg(x − m)}l, m∈ Z fails condition (G4)—that is,
such that
\
Vl 6 = {0}, where Vl = span{gk, m : k < l, k, m ∈ Z}. (5.11)
l∈ Z
Let g ∈ L2 ( R) be such that g(x) 6= 0 if and only if x ∈ (0, 1), and also let
R1
0 log|g(x)| dx = −∞. Then, by a classical result of Helson [He1, pp. 13, 21],
the system of functions {e 2πkxg(x) : k ∈ N} is complete in L2 (0, 1) := {f ∈
L2 ( R) : supp f ⊂ (0, 1)}. As an immediate consequence,
span{e 2πkxg(x) : k < l} = L2 (0, 1) for any l ∈ Z .
Therefore, Vl = L2 ( R) for any l ∈ Z , and (5.11) holds. This example shows that
we cannot expect (G4) to hold unless g satisfies some additional hypotheses (e.g.,
that g be band-limited).
References
[Ba1] L. W. Baggett, Processing a radar signal and representations of the discrete
Heisenberg group, Colloq. Math. 60/61 (1990), 195–203.
[Ba2] , An abstract interpretation of the wavelet dimension function using group
representations, J. Funct. Anal. 173 (2000), 1–20.
[BMM] L. W. Baggett, H. A. Medina, and K. D. Merrill, Generalized multi-resolution

analyses and a construction procedure for all wavelet sets in R n, J. Fourier Anal.
Appl. 5 (1999), 563–573.
[BM] L. W. Baggett and K. D. Merrill, Abstract harmonic analysis and wavelets in R n,
The functional and harmonic analysis of wavelets and frames (San Antonio, TX,
1999), pp. 17–27, Amer. Math. Soc., Providence, RI, 1999.
[BHW] J. J. Benedetto, C. Heil, and D. F. Walnut, Gabor systems and the Balian–Low
theorem, Gabor analysis and algorithms, pp. 85–122, Birkhäuser, Boston, 1998.
[BL] J. J. Benedetto and S. Li, The theory of multiresolution analysis frames and
applications to filter banks, Appl. Comput. Harmon. Anal. 5 (1998), 389–427.
[BDR1] C. de Boor, R. A. DeVore, and A. Ron, The structure of finitely generated shift-
invariant spaces in L 2 ( R d ), J. Funct. Anal. 119 (1994), 37–78.
[BDR2] , Approximation from shift-invariant subspaces of L 2 ( R d ), Trans. Amer.
Math. Soc. 341 (1994), 787–806.
[BDR3] , Approximation orders of FSI spaces in L 2 ( R d ), Constr. Approx. 14
(1998), 631–652.
[Bo1] M. Bownik, The structure of shift-invariant subspaces of L2 ( R n ), J. Funct.
Anal. 177 (2000), 282–309.
[Bo2] , On characterizations of multiwavelets in L2 ( R n ), Proc. Amer. Math.
Soc. 129 (2001), 3265–3274.
[Bo3] , Riesz wavelets and generalized multiresolution analyses, Appl. Comput.
Harmon. Anal. 14 (2003), 181–194.
[BRS] M. Bownik, Z. Rzeszotnik, and D. Speegle, A characterization of dimension
functions of wavelets, Appl. Comput. Harmon. Anal. 10 (2001), 71–92.
[CDH] O. Christensen, B. Deng, and C. Heil, Density of Gabor frames, Appl. Comput.
Harmon. Anal. 7 (1999), 292–304.
[D] I. Daubechies, Ten lectures on wavelets, Society for Industrial and Applied
Mathematics (SIAM), Philadelphia, 1992.
[DGM] I. Daubechies, A. Grossmann, and Y. Meyer, Painless nonorthogonal expansions,
J. Math. Phys. 27 (1986), 1271–1283.
[DLL] I. Daubechies, H. Landau, and Z. Landau, Gabor time-frequency lattices and the
Wexler–Raz identity, J. Fourier Anal. Appl. 1 (1995), 437–478.
[GH] L.-P. Gabardo and D. Han, Subspace Weyl–Heisenberg frames, J. Fourier Anal.
Appl. 7 (2001), 419–433.
[Ga] D. Gabor, Theory of communication, J. Inst. Electr. Eng., London (III) 93
(1946), 429–457.
[HLPS] D. Han, D. R. Larson, M. Papadakis, and Th. Stavropoulos, Multiresolution
analyses of abstract Hilbert spaces and wandering subspaces, The functional
and harmonic analysis of wavelets and frames (San Antonio, TX, 1999), pp.
259–284, Amer. Math. Soc., Providence, RI, 1999.
[HWa] D. Han and Y. Wang, Lattice tiling and the Weyl–Heisenberg frames, Geom.
Funct. Anal. 11 (2001), 742–758.
[He1] H. Helson, Lectures on invariant subspaces, Academic Press, New York, 1964.
[He2] , The spectral theorem, Springer-Verlag, Berlin, 1986.
[HW] E. Hernández and G. Weiss, A first course on wavelets, Studies in Advanced
Mathematics, CRC Press, Boca Raton, FL, 1996.
[Ja] A. J. E. M. Janssen, Signal analytic proofs of two basic results on lattice
expansions, Appl. Comput. Harmon. Anal. 1 (1994), 350–354.
[Ji] R.-Q. Jia, Shift-invariant spaces and linear operator equations, Israel J. Math.
103 (1998), 259–288.
[La] H. Landau, On the density of phase-space expansions, IEEE Trans. Inform.

Theory 39 (1993), 1152–1156.
[LTW] D. Larson, W.-S. Tang, and E. Weber, Riesz wavelets and multiresolution
structures, Proc. SPIE, 4478, pp. 254–262, SPIE, Bellingham, WA, 2001.
[Ly] Yu. I. Lyubarskiı̆, Frames in the Bargmann space of entire functions, Entire and
subharmonic functions, pp. 167–180, Amer. Math. Soc., Providence, RI, 1992.
[PSWX] M. Paluszyński, H. Šikić, G. Weiss, and S. Xiao, Tight frame wavelets, their
dimension functions, MRA tight frame wavelets and connectivity properties,
Adv. Comput. Math. 18 (2003), 297–327.
[RSt] J. Ramanathan and T. Steger, Incompleteness of sparse coherent states, Appl.
Comput. Harmon. Anal. 2 (1995), 148–153.
[Ri] M. A. Rieffel, Von Neumann algebras associated with pairs of lattices in Lie
groups, Math. Ann. 257 (1981), 403–418.
[RS1] A. Ron and Z. Shen, Frames and stable bases for shift-invariant subspaces of
L 2 ( R d ), Canad. J. Math. 47 (1995), 1051–1094.
[RS2] , Affine systems in L 2 ( R d ): the analysis of the analysis operator,
J. Funct. Anal. 148 (1997), 408–447.
[RS3] , Weyl–Heisenberg frames and Riesz bases in L 2 ( R d ), Duke Math. J. 89
(1997), 237–282.
[RS4] , The wavelet dimension function is the trace function of a shift-invariant
system, Proc. Amer. Math. Soc. 131 (2003), 1385–1398.
[Rz1] Z. Rzeszotnik, Characterization theorems in the theory of wavelets, Ph.D.
thesis, Washington University, St. Louis, 2000.
[Rz2] , Calderón’s condition and wavelets, Collect. Math. 52 (2001), 181–191.
[SW] K. Seip and R. Wallstén, Density theorems for sampling and interpolation in the
Bargmann–Fock space. II, J. Reine Angew. Math. 429 (1992), 107–113.
[We] E. Weber, Applications of the wavelet multiplicity function, The functional
and harmonic analysis of wavelets and frames (San Antonio, TX, 1999),
pp. 297–306, Amer. Math. Soc., Providence, RI, 1999.
[WW] G. Weiss and E. N. Wilson, The mathematical theory of wavelets, Twentieth
century harmonic analysis—a celebration (Il Ciocco, 2000), pp. 329–366,
Kluwer, Dordrecht, 2001.
M. Bownik Z. Rzeszotnik
Department of Mathematics Mathematical Institute
University of Michigan University of Wrocław
Ann Arbor, MI 48109 50-384 Wrocław
Poland
marbow@umich.edu
Current address
Department of Mathematics
University of Texas
Austin, TX 78712
zioma@mail.ma.utexas.edu

3 - The Spectral Function of SI Spaces

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

3 - The Spectral Function of SI Spaces

Uploaded by

Copyright:

Available Formats

Michigan Math. J.

The Spectral Function of Shift-Invariant Spaces

Received March 26, 2002. Revision received January 30, 2003.

has already been used to show a new characterization of wavelets conjectured by

ξ + k, where k is a unique element of Z n such that ξ + k ∈ T n. Finally, let

where J is a measurable range function. The correspondence between V and J

The dimension function of a SI space V ⊂ L2 ( R n ) is a mapping dimV : R n →

The dual Gramian of a SI system E(8), where

Recall that a matrix G̃(ξ) = (G̃(ξ)k, l )k, l∈ Z n is bounded on ` 2 ( Z n ) if G̃(ξ) (given

Theorem 1.2. Suppose that 8 ⊂ L2 ( R n ) is countable and that 8 satisfies (1.5).

if and only if the dual Gramian G̃(ξ) satisfies

2. The Spectral Function

Definition 2.1. Suppose V ⊂ L2 ( R n ) is SI with the range function J(ξ) and

Note that σV (ξ) is well-defined for a.e. ξ ∈ R n, since {k + T n : k ∈ Z n } is a parti-

Proposition 2.2 is a consequence of yet another description of the spectral function

Lemma 2.3. Suppose a SI space V ⊂ L2 ( R n ) is generated by the shifts of a

and, in particular, for any k ∈ Z n we have

This shows (2.5) and completes the proof of Lemma 2.3.

This completes the proof of Proposition 2.2.

Theorem 2.4. Let V ⊂ L2 ( R n ) be SI and let A be an n × n integer matrix with

Because the |det A| × |det A| matrix

= |ϕ̂((A ) (ξ + k))| = σS(ϕ)((A∗ )−1(ξ + k)).

This shows (2.8) and therefore (2.6).

which completes the proof of Theorem 2.4.

As an immediate consequence of (2.2) and Theorem 2.4, we obtain the following

Corollary 2.5. Suppose that V ⊂ L2 ( R n ) is SI and that A is an n × n integer

where D is the set of |det A| representatives of different cosets of Z n/(A∗ Z n ).

Example 1. Given a measurable set E ⊂ R n, let V = Ľ2 (E ). Then σV =

Clearly ϕ k is a quasi-orthogonal generator of S(ϕ k ), and S(ϕ k ) ⊥ S(ϕ k 0 ) for k 6 =

Proposition 2.6. Let S be the set of all SI subspaces of L2 ( R n ). Then, for

We will also need the following approximation lemma.

Lemma 2.7. Let V be a SI space and let (Vj )j ∈ N be a sequence of SI spaces.

In particular, there exists a subsequence (j k )k∈ N such that

Proof. Let J and Jj (j ∈ N ) be the range functions corresponding to V and Vj , re-

Proof. Let f = 1̌K . Then, for any ξ ∈ T n,

where J is the range function of V.

As a consequence of Lemma 2.8 and the Lebesgue differentiation theorem, we

Approximation Orders of SI Spaces. As an example of the utility of the spec-

Theorem 2.9 [BDR2, Thm. 1.9]. Suppose that V ⊂ L2 ( R n ) is SI and that

(iii) S(P V (φ 0 )) provides approximation order k, where φ 0 is the sinc-function,

As a consequence of this result, we obtain the following characterization of ap-

Theorem 2.10. A SI space V ⊂ L2 ( R n ) provides approximation order k > 0 if

3. The Dimension Function of a GMRA

Definition 3.1. Let A be a fixed n × n integer expansive dilation matrix (i.e.,

Vj = span{DAi Tk ψ l : i < j, k ∈ Z n, l = 1, . . . , L}.

where D is the set of |det A| representatives of different cosets of Z n/A∗ Z n. Another

Theorem 3.2. Suppose a SI space V ⊂ L2 ( R n ) is refinable, that is, V ⊂ DAV.

D : R n → N ∪ {0, ∞} is a measurable Z n -periodic function;

1 = {ξ ∈ R n : D(B −j ξ) ≥ 1 for j ∈ N ∪ {0}}. (3.1)

Proof of Theorem 3.2: Necessity of (D1)–(D3). Clearly, the dimension function

which shows (D3).

where 1 is given by (3.1).

Proof of Theorem 3.2: Sufficiency of (D1)–(D3). Define a SI P space V = Ľ2 (S ),

where 1 is given by (3.1). For any k ∈ N, let

S idea of the proof is to construct a sequence of sets {S k } k∈ N such that S =

τ | S i is injective and onto A i for i = 1, ..., k; (3.13)

But (3.16) implies that, for j ≥ 0,

where D is the set of representatives of different cosets of Z n/B Z n. Without loss

Proof of Lemma 3.3 under D ≡ ∞. Finally, we need to construct a set S satisfy-

Observe that (3.5) follows from

By letting i → ∞, we conclude that c is zero and hence

This shows (3.6) and completes the proof of Lemma 3.3.

Finally, we are ready to show our main result.

Assume by way of contradiction