Proceedings of the Institute of Mathematics and Mechanics,

National Academy of Sciences of Azerbaijan

Volume 46, Number 1, 2020, Pages 79–93
https://doi.org/10.29228/proc.19

CHARACTERIZATION OF CONTINUOUS g-FRAMES VIA

OPERATORS

MORTEZA RAHMANI

Abstract. In this paper we introduce and show some new notions and
results on cg-frames of Hilbert spaces. We define cg-orthonormal bases
for a Hilbert space H and verify their properties and relations with cg-
frames. Actually, we present that every cg-frame can be represented as
a composition of a cg-orthonormal basis and an operator under some
conditions. Also, we find for any cg-frame an induced c-frame and study
their properties and relations. Moreover, we show that every cg-frame
can be written as addition of two Parseval cg-frames. In addition, we
show each cg-frame as a sum of a cg-orthonormal basis and a cg-Riesz
basis.

1. Introduction
Frames (discrete frames) in Hilbert spaces were introduced by Duffin and Scha-
effer [7] in 1952 to study some deep problems in nonharmonic Fourier series. Af-
ter the illustrious paper [6] by Daubechies, Grossmann and Meyer, frame theory
popularized immensely.
A frame for a Hilbert space allows each vector in the space to be written as a
linear combination of the elements in the frame, but linear independence between
the frame elements is not required. Intuitively, a frame can be thought as a basis
to which one has added more elements.
Generally, frames have been used in signal processing, image processing, data
compression and sampling theory. Later, motivated by the theory of coherent
states, this concept was generalized to families indexed by some locally compact
space endowed with a Radon measure. This approach leads to the notion of
continuous frames [2, 3, 11, 13]. Some results about continuous frames and their
generalizations can be found in [8, 9, 10, 15].
In this paper, inspired by [16] and [12], we generalize some results to cg-frames.
The paper is organized as follows. In Section 2, we introduce the concept of cg-
orthonormal bases for Hilbert spaces and discuss about their characteristics and
their relations with cg-frames and c-frames. Our aim in Section 3 is describing
every continuous g-frame as a sum of two Parseval continuous g-frames. We also
present that every continuous g-frame can be written as a linear combination of
an cg-orthonormal basis and a cg-Riesz basis.
2010 Mathematics Subject Classification. 42C15, 46C05.
Key words and phrases. Hilbert space, c-frame, cg-frame, cg-orthonormal basis.
79
80 MORTEZA RAHMANI

Throughout this paper, H is a separable Hilbert space, (Ω, µ) is a measure

space with positive measure µ and {Hω }ω∈Ω is a family of separable Hilbert
spaces.
We first review the definitions of continuous frames and continuous g-frames.
Definition 1.1. ([15]) Suppose that (Ω, µ) is a measure space with positive
measure µ. A mapping f : Ω −→ H is called a continuous frame, or simply a
c-frame, with respect to (Ω, µ) for H, if:
(i) for each h ∈ H, ω 7−→ hh, f (ω)i is a measurable function,
(ii) there exist positive constants A and B such that
Z
Akhk2 ≤ |hh, f (ω)i|2 dµ(ω) ≤ Bkhk2 , h ∈ H. (1.1)
Ω
The constants A, B are called c-frame bounds. If A, B can be chosen such that
A = B, then f is called a tight c-frame and if A = B = 1, it is called a Parseval
c-frame. A mapping f is called c-Bessel mapping if the second inequality in (1.1)
holds. In this case, B is called the Bessel bound.
Some operators associated to c-Bessel mappings can be useful to characterize
them.
Proposition 1.1. ([15]) Let (Ω, µ) be a measure space and f : Ω −→ H be a
c-Bessel mapping for H. Then the operator Tf : L2 (Ω, µ) −→ H, weakly defined
by
Z
hTf ϕ, hi = ϕ(ω)hf (ω), hidµ(ω), h ∈ H, (1.2)
Ω
is well defined, linear, bounded, and its adjoint is given by
Tf∗ : H −→ L2 (Ω, µ), Tf∗ h(ω) = hh, f (ω)i, ω ∈ Ω. (1.3)
The operator Tf is called synthesis operator and Tf∗ is called the analysis
operator of f .
If f is a c-Bessel mapping with respect to (Ω, µ) for H, then the operator
Sf : H −→ H defined by Sf = Tf Tf∗ , is called frame operator of f . Thus
Z
hSf h, ki = hh, f (ω)ihf (ω), kidµ(ω), h, k ∈ H.
Ω
If f is a c-frame for H, then S is invertible.
The converse of above proposition holds when µ is σ-finite in the measure space
(Ω, µ).
Theorem 1.1. ([15]) Suppose that (Ω, µ) is a measure space where µ is σ-finite.
Let f : Ω −→ H be a mapping such that for each h ∈ H, ω 7−→ hh, f (ω)i is
measurable. The mapping f is a c-frame with respect to (Ω, µ) for H if and
only if the operator Tf : L2 (Ω, µ) −→ H defined by (1.2), is a bounded and onto
operator.
Definition 1.2. Let ϕ ∈ Πω∈Ω Hω . We say that ϕ is strongly measurable if ϕ as
a mapping of Ω to ⊕ω∈Ω Hω is measurable, where

Πω∈Ω Hω = f : Ω −→ ∪ω∈Ω Hω ; f (ω) ∈ Hω .
Now, we review the definition of continuous g-frames.
 CHARACTERIZATION OF CONTINUOUS g-FRAMES VIA OPERATORS 81

Definition 1.3. We call {Λω ∈ B(H, Hω ) : ω ∈ Ω} a continuous generalized

frame, or simply a cg-frame, for H with respect to {Hω }ω∈Ω , if:
(i) for each f ∈ H, {Λω f }ω∈Ω is strongly measurable,
(ii) there are two positive constants A and B such that
Z
2
Akf k ≤ kΛω f k2 dµ(ω) ≤ Bkf k2 , f ∈ H. (1.4)
Ω

We call A and B the lower and upper cg-frame bounds, respectively. If A, B

can be chosen such that A = B, then {Λω }ω∈Ω is called a tight cg-frame and if
A = B = 1, it is called a Parseval cg-frame. A family {Λω }ω∈Ω is called cg-Bessel
family if the second inequality in (1.4) holds.


Now, let the space ⊕ω∈Ω Hω , µ L2 ⊆ Πω∈Ω Hω be defined as follows:
Z
⊕ω∈Ω Hω , µ L2 = ϕ| ϕ is strongly measurable, kϕ(ω)k2 dµ(ω) < ∞ .


Ω


The space ⊕ω∈Ω Hω , µ L2 is a Hilbert space with inner product
Z
hϕ, ψi = hϕ(ω), ψ(ω)idµ(ω).
Ω

Proposition 1.2. ([1]) Let {Λω }ω∈Ω be a cg-Bessel family for H with respect

to
{Hω }ω∈Ω with Bessel bound B. Then the mapping T of ⊕ω∈Ω Hω , µ L2 to H
defined by
Z
hT ϕ, hi = hΛ∗ω ϕ(ω), hidµ(ω), ϕ ∈ ⊕ω∈Ω Hω , µ L2 , h ∈ H,


(1.5)
Ω
√
is linear and bounded with kT k ≤ B. Furthermore for each h ∈ H and ω ∈ Ω,

T ∗ (h)(ω) = Λω h. (1.6)

The operators T and T ∗ are called synthesis and analysis operators of cg-Bessel
family {Λω }ω∈Ω , respectively.
Let {Λω }ω∈Ω be a cg-frame for H with respect to {Hω }ω∈Ω with frame bounds
A, B. The operator S : H −→ H defined by
Z
hSf, gi = hf, Λ∗ω Λω gidµ(ω), f, g ∈ H, (1.7)
Ω

is called the frame operator of {Λω }ω∈Ω which is a positive and invertible operator.
Now, we state a known result that is helpful in proving some results.

Proposition 1.3. ([4]) Let K : H −→ H be a bounded linear operator. Then the

following hold.
(i) K = α(U1 + U2 + U3 ), where each Uj , j = 1, 2, 3, is a unitary operator and α
is a constant.
(ii) If K is onto, then it can be written as a linear combination of two unitary
operators if and only if K is invertible.
82 MORTEZA RAHMANI

2. cg-Orthonormal bases
Similar to the continuous frames, we want to generalize orthonormal bases.
Indeed, our purpose here is to define a mapping f : Ω −→ H that has similar
properties to an orthonormal basis of H.
Definition 2.1. Suppose that (Ω, µ) is a measure space. A mapping f : Ω −→ H
is called a c-orthonormal basis with respect to (Ω, µ) for H, if:
(i) for each h ∈ H, ω 7−→ hh, f (ω)i is measurable,
(ii) for almost all ν ∈ Ω,
Z
hf (ω), f (ν)idµ(ω) = 1,
Ω

2 = khk2 .
R
(iii) for each h ∈ H, Ω |hh, f (ω)i| dµ(ω)

Now we define the generalization of orthonormal basis in case of operators.

Definition 2.2. Assume (Ω, µ) is a measure space. A family of operators Λ =
{Λω ∈ B(H, Hω ) : ω ∈ Ω} is called a continuous g-orthonormal basis or simply a
cg-orthonormal basis, for H with respect to {Hω }ω∈Ω , whenever:
(i) for each h ∈ H, {Λω h}ω∈Ω is strongly measurable,
(ii) for almost all ν ∈ Ω,
Z
hΛ∗ω fω , Λ∗ν gν idµ(ω) = hfν , gν i, {fω }ω∈Ω ∈ (⊕ω∈Ω Hω , µ)L2 , gν ∈ Hν ,
Ω

2 dµ(ω) = khk2 .
R
(iii) for each h ∈ H, Ω kΛω hk
If only conditions (i) and (ii) hold, Λ = {Λω ∈ B(H, Hω ); ω ∈ Ω}ω∈Ω is called
a cg-orthonormal system for H with respect to {Hω }ω∈Ω .

Example 2.1. Suppose that Ω = {a, b, c}, Σ = ∅, {a, b}, {c}, Ω and µ : Σ −→
[0, ∞] is a measure such that µ(∅) = 0, µ({a, b}) = 1, µ({c}) = 1 and µ(Ω) = 2.
Let H be a 2 dimensional Hilbert space with an orthonormal basis {e1 , e2 }. We
define
f : Ω −→ H
by f = e1 χ{a,b} + e2 χ{c} . So for each h ∈ H,
hf (ω), hi = he1 , hiχ{a,b} (ω) + he2 , hiχ{c} (ω), ω ∈ Ω,
hence ω 7−→ hh, f (ω)i is measurable. Now, for each ω ∈ Ω, we define
Λω : H −→ C
Λω (h) = hh, f (ω)i.
Actually, we consider for each ω ∈ Ω, Hω = C. By an easy calculation, we have
Λ∗ω (z) = f (ω)z, z ∈ C.
For any ν ∈ Ω, xν ∈ C and any {zω }ω∈Ω ∈ (⊕ω∈Ω Hω , µ)L2 = L2 (Ω, µ), due to
the Example 4.2 in [16], we have
Z Z
∗ ∗
hΛω zω , Λν xν idµ(ω) = zω xν hf (ω), f (ν)idµ(ω) = zν xν .
Ω Ω
 CHARACTERIZATION OF CONTINUOUS g-FRAMES VIA OPERATORS 83

Also for each h ∈ H,

Z Z
2
kΛω hk dµ(ω) = |hh, f (ω)i|2 dµ(ω) = khk2 .
Ω Ω
Therefore {Λω }ω∈Ω is a cg-orthonormal basis for H with respect to {Hω }ω∈Ω ,
where for each ω ∈ Ω, Hω = C.
We present some equal conditions for cg-orthonormal bases.
Theorem 2.1. Let {Λω }ω∈Ω be a cg-orthonormal R system for H with respect to
{Hω }ω∈Ω . Also assume that for each h ∈ H, Ω kΛω hk2 dµ(ω) < ∞. Then the
following conditions are equivalent:
(i) {Λω }ω∈Ω is a cg-orthonormal basis for H with respect to {Hω }ω∈Ω .
(ii) For each h, k ∈ H,
Z
hh, ki = hΛω h, Λω kidµ(ω).
Ω

(iii) If Λω h = 0, a.e. [µ], then h = 0.

(υ) For each zero measure set Ω0 ⊆ Ω, H = span{Λ∗ω (Hω )}ω∈Ω\Ω0 .
Proof. (i) ⇔ (ii) Since {Λω }ω∈Ω is a Parseval cg-frame for H, so its frame oper-
ator SΛ = I. Hence (ii) is obvious. The converse side clearly holds.
(ii) ⇒ (iii) If Λω h = 0, a.e. [µ], then for every k ∈ H,
Z
hh, ki = hΛω h, Λω kidµ(ω) = 0.
Ω
Therefore h = 0.
(iii) ⇒ (υ) Suppose that Ω0 ⊆ Ω and h ⊥ span{Λ∗ω (Hω )}ω∈Ω\Ω0 , so for almost
all ω ∈ Ω, hΛ∗ω Λω h, hi = 0. Then kΛω hk2 = 0, a.e. [µ], which implies h = 0.
(υ) ⇒ (ii) Let k ∈ H. Assume that
n Z o
Hk = h ∈ H : hh, ki = hΛω h, Λω kidµ(ω) .
Ω
Hk is a subspace of H. Also, it is closed, since if limn→∞ hn = h, where hn ’s
belong to Hk , then
Z
hh, ki = lim hhn , ki = lim hΛω hn , Λω kidµ(ω).
n→∞ n→∞ Ω

According to assumption,
Z Z 1  Z 1
2 2 2 2
|hΛω h, Λω ki|dµ(ω) ≤ kΛω hk dµ(ω) kΛω gk dµ(ω) < ∞.
Ω Ω Ω
So by Lebesgue’s Dominated Convergence Theorem,
Z Z
lim hΛω hn , Λω kidµ(ω) = hΛω h, Λω kidµ(ω),
n→∞ Ω Ω
which means h ∈ Hk . For almost all ν ∈ Ω and each f ∈ H, we have
Z Z
hΛω Λν Λν f, Λω kidµ(ω) = hΛ∗ν Λν f, Λ∗ω Λω kidµ(ω)
∗
Ω Ω
= hΛν f, Λν ki = hΛ∗ν Λν f, ki,
84 MORTEZA RAHMANI

therefore Λ∗ν Λν f ∈ Hk . Assume f ⊥ Hk , then for almost all ν ∈ Ω,

0 = hΛ∗ν Λν f, f i = kΛν f k2 ,
which gives Λν f = 0. For almost all ν ∈ Ω and any gν ∈ Hν ,
hf, Λ∗ν gν i = hΛν f, gν i = 0.
So f ⊥ span{Λ∗ω (Hω )}ω∈Ω\Ω0 , where Ω0 is a zero measure subset of Ω. By
condition (υ), span{Λ∗ω (Hω )}ω∈Ω\Ω0 = H. Thus f = 0 and Hk = H. Therefore
Z
hh, ki = hΛω h, Λω kidµ(ω), h, k ∈ H.
Ω


In the following of this section, suppose that there exists a cg-orthonormal

basis for H.
Proposition 2.1. Suppose that {Θω }ω∈Ω is a cg-orthonormal basis for H with
respect to {Hω }ω∈Ω and {Λω ∈ B(H, Hω ) : ω ∈ Ω} is a family such that for
each h ∈ H, {Λω h}ω∈Ω is strongly measurable. Then {Λω }ω∈Ω is a Parseval
cg-frame for H if and only if there exists a unique isometry V ∈ B(H) such that
Λω = Θω V , a.e. [µ].
Proof. Let {Λω }ω∈Ω is a Parseval cg-frame for H. We define the operator V
weakly by Z
hV f, hi = hΘ∗ω Λω f, hidµ(ω), f, h ∈ H.
Ω
For each f, h ∈ H, we have
Z 1  Z 1
2 2 2
|hV f, hi| ≤ kΛω f k dµ(ω) kΘω hk2 dµ(ω) ≤ kf kkhk.
Ω Ω
So V is well-defined and bounded. For almost all ν ∈ Ω and each f ∈ H, hν ∈ Hν ,
Z
hΘν V f, hν i = hV f, Θν hν i = hΘ∗ω Λω f, Θ∗ν hν idµ(ω) = hΛν f, hν i,
∗
Ω

since {Θω }ω∈Ω is a cg-orthonormal basis. Thus Λω = Θω V , a.e. [µ].

For each f ∈ H,
Z Z
2 ∗
kV f k = hV f, V f i = hΘω Λω f, V f idµ(ω) = hΛω f, Θω V f idµ(ω)
ZΩ Ω

= hΛω f, Λω f idµ(ω) = kf k2 .
Ω
Therefore V is an isometry.
Now, let V1 and V2 be two isometries such that Λω = Θω V1 , a.e. [µ] and
Λω = Θω V2 , a.e. [µ]. Then for each f ∈ H, Θω ((V1 − V2 )f ) = 0, a.e. [µ], which
implies Z
0= kΘω (V1 − V2 )f k2 dµ(ω) = k(V1 − V2 )f k2 ,
Ω
so V1 f = V2 f .
 CHARACTERIZATION OF CONTINUOUS g-FRAMES VIA OPERATORS 85

Conversely, let V ∈ B(H) be a unique isometry such that Λω = Θω V , a.e. [µ].

For any f ∈ H,
Z Z
2
kΛω f k dµ(ω) = kΘω V f k2 dµ(ω) = kV f k2 = kf k2 .
Ω Ω

Hence {Λω }ω∈Ω is a Parseval cg-frame for H. 

Theorem 2.2. Assume that {Θω }ω∈Ω is a cg-orthonormal basis for H with re-
spect to {Hω }ω∈Ω and {Λω ∈ B(H, Hω ) : ω ∈ Ω} is a family such that for each
h ∈ H, {Λω h}ω∈Ω is strongly measurable. Then {Λω }ω∈Ω is a cg-frame for H
with bounds A and B if and only if there exists a unique V ∈ B(H) such that
Λω = Θω V , a.e. [µ] and AI ≤ V ∗ V ≤ BI.
Proof. Let {Λω }ω∈Ω be a cg-frame for H with bounds A, B. Similar to the proof
of Proposition 2.1, the operator V weakly defined by
Z
hV f, hi = hΘ∗ω Λω f, hidµ(ω), f, h ∈ H,
Ω

is a one-to-one and bounded operator such that Λω = Θω V , a.e. [µ]. Also for
each f ∈ H,
Z Z
2 ∗
kV f k = hV f, V f i = hΘω Λω , V f idµ(ω) = hΛω f, Θω V f idµ(ω)
ZΩ Ω

= kΛω f k2 dµ(ω).
Ω
Therefore
Ahf, f i ≤ hV ∗ V f, f i ≤ Bhf, f i,
which implies AI ≤ V V ∗ ≤ BI.
The opposite implication is similar to the Proposition 2.1. 
Theorem 2.3. Suppose that {Λω }ω∈Ω is a cg-orthonormal basis for H with re-
spect to {Hω }ω∈Ω and V ∈ B(H). Then {Λω V }ω∈Ω is a cg-orthonormal basis for
H with respect to {Hω }ω∈Ω if and only if V is unitary.
Proof. Assume that {Λω V }ω∈Ω is a cg-orthonormal basis for H. Since {Λω }ω∈Ω
is also a cg-orthonormal basis for H, for each f ∈ H, we have
Z
2
kV f k = kΛω V f k2 dµ(ω) = kf k2 .
Ω
Hence V is an isometry and V ∗V = I. Considering Θω = Λω V, ω ∈ Ω, in The-
orem 2.1, there exists a unique isometry U ∈ B(H) such that Λω = Λω V U ,
a.e. [µ]. Let TΛ and TΛV U be the synthesis operators of Parseval cg-frames
{Λω }ω∈Ω and {Λω V U }ω∈Ω , respectively. Then TΛ = (V U )∗ TΛ = U ∗ V ∗ TΛ . We
deduce TΛ TΛ∗ = U ∗ V ∗ TΛ TΛ∗ or SΛ = U ∗ V ∗ SΛ , where SΛ is the frame operator of
{Λω }ω∈Ω . Since SΛ = I, so I = U ∗ V ∗ or equivalently V U = I. This implies that
V is onto. Also V is one-to-one, so V is invertible and V −1 = V ∗ , which means
V is a unitary.
For the reverse implication, suppose that V is a unitary operator. Now, we
show that {Λω V }ω∈Ω is a cg-orthonormal basis for H. For almost all ν ∈ Ω, each
86 MORTEZA RAHMANI

gν ∈ Hν and each {fω }ω∈Ω ∈ (⊕ω∈Ω Hω , µ)L2 , we have

Z Z
h(Λω V ) fω , (Λν V ) gν idµ(ω) = hV ∗ Λ∗ω fω , V ∗ Λ∗ν gν idµ(ω)
∗ ∗
Ω
ZΩ
= hΛ∗ω fω , Λ∗ν gν idµ(ω) = hfν , gν i.
Ω
Also for each f ∈ H,
Z
kΛω V f k2 dµ(ω) = kV f k2 = kf k2 .
Ω
Therefore {Λω V }ω∈Ω is a cg-orthonormal basis for H with respect to {Hω }ω∈Ω .

Concerning to cg-Riesz bases which are defined in [14], we have next result.
Theorem 2.4. Let (Ω, µ) be a measure space where µ is σ-finite. Suppose that
{Λω }ω∈Ω is a cg-Riesz basis for H and V ∈ B(H). Then {Λω V }ω∈Ω is a cg-Riesz
basis for H if and only if V is invertible.
Proof. Let {Λω V }ω∈Ω be a cg-Riesz basis for H. By definition of a cg-Riesz basis,
the operator TΛV weakly defined by
Z
hTΛV ϕ, hi = h(Λω V )∗ ϕ(ω), hidµ(ω), ϕ ∈ (⊕ω∈Ω Hω , µ)L2 , h ∈ H,
Ω
is well-defined and there exist positive constants A and B such that
Akϕk ≤ kTΛV ϕk ≤ Bkϕk, ϕ ∈ (⊕ω∈Ω Hω , µ)L2 .
An easy calculation shows TΛV = V ∗T ,
where TΛ is defined similarly for {Λω }ω∈Ω .
Λ
By Lemma 3.2 (i) in [14], TΛV and TΛ both are invertible. So V ∗ = TΛV TΛ−1 is
invertible and V is invertible.
Conversely, let V ∈ B(H) be invertible. If Λω V f = 0, a.e. [µ], then V f = 0 and
it implies f = 0. The operator TΛV given by
Z
hTΛV ϕ, hi = h(Λω V )∗ ϕ(ω), hidµ(ω), ϕ ∈ (⊕ω∈Ω Hω , µ)L2 , h ∈ H,
Ω
is well-defined, bounded and TΛV = V ∗ TΛ , where TΛ is defined similar to TΛV .
Also, there are positive constants A and B such that
Akϕk ≤ kTΛ ϕk ≤ Bkϕk, ϕ ∈ (⊕ω∈Ω Hω , µ)L2 .
For each ϕ ∈ (⊕ω∈Ω Hω , µ)L2 , kTΛV ϕk = kV ∗ TΛ ϕk ≤ BkV ∗ kkϕk, and
1 A
kTΛV ϕk = kV ∗ TΛ ϕk ≥ −1
kTΛ ϕk ≥ kϕk.
kV k kV −1 k
This shows that {Λω V }ω∈Ω is a cg-Riesz basis for H. 
Now, we define cg-complete families as follows:
Definition 2.3. A family {Λω ∈ B(H, Hω ) : ω ∈ Ω} is called a cg-complete
family for H with respect to {Hω }ω∈Ω , if:

h : Λω h = 0, a.e. [µ] = {0}.
Lemma 2.1. {Λω }ω∈Ω is a cg-complete family for H and V ∈ B(H). Then
{Λω V }ω∈Ω is a cg-complete family for H if and only if V is one-to-one.
 CHARACTERIZATION OF CONTINUOUS g-FRAMES VIA OPERATORS 87

Proof. Let {Λω V }ω∈Ω be a cg-complete family. If V h = 0, then

Λω V h = 0, ω ∈ Ω,
so h = 0 and V is one-to-one.
Now, suppose V is one-to-one and Λω V h = 0, a.e. [µ]. Since {Λω }ω∈Ω is cg-
complete, V h = 0. Hence h = 0 , which implies {Λω V }ω∈Ω is cg-complete. 
Proposition 2.2. Let (Ω, µ) be a measure space where µ is σ-finite and {Λω }ω∈Ω
be a cg-Bessel family for H. Then {Λω }ω∈Ω is cg-complete if and only if R(TΛ ) =
H, where TΛ is the synthesis operator of {Λω }ω∈Ω .
Proof. Assume that {Λω }ω∈Ω is cg-complete. To show that R(TΛ ) = H, it is
enough to prove that if f ∈ H and f ⊥R(TΛ ), then f = 0. Let f ∈ H and
f ⊥R(TΛ ), so for each F ∈ (⊕ω∈Ω Hω , µ)L2 ,
Z
0 = hTΛ F, f i = hΛ∗ω F (ω), f idµ(ω).
Ω
Since (Ω, µ) is σ-finite, there
S∞exists a family {Ωn }∞
n=1 of disjoint measurable sub-
sets of Ω, such that Ω = n=1 Ωn and µ(Ωn ) < ∞, n ≥ 1. For each n ≥ 1,
set 
Λω f, ω ∈ Ωn
Fn (ω) = ,
0, otherwise
then Z
hTΛ Fn , f i = hFn (ω), Λω f idµ(ω) = kΛω f k2 µ(Ωn ) = 0.
Ω
Thus Λω f = 0, a.e. [µ], which implies f = 0. So R(TΛ ) = H.
For the opposite implication, suppose that R(TΛ ) = H and there exists a f 6= 0
such that
Λω f = 0, a.e. [µ].
There exists a sequence {Fn }∞
n=1 ⊆ (⊕ω∈Ω Hω , µ)L2 such that limn→∞ TΛ Fn = f .
Then
kf k2 = hf, f i = h lim TΛ Fn , f i = lim hTΛ Fn , f i
Z n→∞ n→∞

= lim hΛ∗ω Fn (ω), f idµ(ω)

n→∞ Ω
Z
= lim hFn (ω), Λω f idµ(ω) = 0,
n→∞ Ω

which is a contradiction. 
Remark 2.1. Let (Ω, µ) be a measure space and consider the family {Λω ∈
B(H, Hω ) : ω ∈ Ω}. Also, suppose that {eω,k }ω∈Ω,k∈Kω is an orthonormal basis
for Hilbert space ⊕ω∈Ω Hω such that for each ω ∈ Ω, {eω,k }k∈Kω is an orthonormal
basis of Hω and for each h ∈ H, the mapping
Ω × K −→ C
(ω, k) 7−→ hh, eω,k i
S
is measurable, where K = ω∈Ω Kω .
88 MORTEZA RAHMANI

The mapping h 7−→ hΛω h, eω,k i, ω ∈ Ω, k ∈ Kω , defines a bounded linear

functional on H. So there exist some uω,k ∈ H such that
hh, uω,k i = hΛω h, eω,k i, h ∈ H, ω ∈ Ω, k ∈ Kω .
P
Therefore Λω h = k∈Kω hh, uω,k ieω,k , h ∈ H. Since
X
|hh, uω,k ieω,k |2 = kΛω hk2 ≤ kΛω k2 khk2 ,
k∈Kω

so for each ω ∈ Ω, {uω,k }k∈Kω is a c-Bessel family for Hω . Also

uω,k = Λ∗ eω,k , ω ∈ Ω, k ∈ Kω . (2.1)
The family {uω,k }ω∈Ω,k∈Kω is called the family induced by {Λω }ω∈Ω with respect
to {eω,k }ω∈Ω,k∈Kω .
Consider the mapping u(ω, k) : Ω × K −→ H defined by

uω,k , k ∈ Kω
u(ω, k) = ,
0, otherwise
S
where K = ω∈Ω Kω .
For each h ∈ H, (ω, k) 7−→ huω,k , hi is measurable and
Z Z X
2
kΛω hk dµ(ω) = |hh, uω,k i|2 dµ(ω)
Ω Ω k∈K
ω
Z Z
|hh, uω,k i|2 dl(k) dµ(ω),


= (2.2)
Ω K
where l : K −→ K is the counting measure on K. If {Λω }ω∈Ω is a cg-frame for H
with respect to {Hω }ω∈Ω , then u is a c-frame for H with respect to (Ω × K, µ × l)
and with the same bounds of {Λω }ω∈Ω .
The converse of above statement is true, too; if {uω,k }ω∈Ω,k∈Kω is a c-frame for
H, then {Λω }ω∈Ω is a cg-frame for H with the same bounds of {uω,k }k∈Kω .
Theorem 2.5. Let (Ω, µ) be a measure space where µ is σ-finite. Consider the
family {Λω ∈ B(H, Hω ); ω ∈ Ω} and let {uω,k }ω∈Ω,k∈Kω be defined as in (2.1).
Then {Λω }ω∈Ω is a cg-frame (respectively cg-Bessel family, tight cg-frame, cg-
Riesz basis, cg-orthonormal basis) for H if and only if {uω,k }ω∈Ω,k∈Kω is a c-frame
(respectively c-Bessel family, tight c-frame, c-Riesz basis, c-orthonormal basis) for
H.
Proof. We see from (2.2) that
Z Z Z
2
|hh, uω,k i|2 dl(k) dµ(ω),


kΛω hk dµ(ω) = h ∈ H.
Ω Ω K
Hence {Λω }ω∈Ω is a cg-frame (respectively cg-Bessel family, tight cg-frame) for
H if and only if {uω,k }ω∈Ω,k∈Kω is a c-frame (respectively c-Bessel family, tight
c-frame).
Now, assume that {Λω }ω∈Ω is a cg-Riesz basis for H. So there are constants
A, B > 0 such that the operator TΛ : (⊕ω∈Ω Hω , µ)L2 −→ H defined by
Z
hTΛ F, hi = hΛ∗ω F (ω), hidµ(ω), F ∈ (⊕ω∈Ω Hω , µ)L2 , h ∈ H,
Ω
 CHARACTERIZATION OF CONTINUOUS g-FRAMES VIA OPERATORS 89

satisfies in
AkF k ≤ kTΛ F k ≤ BkF k, F ∈ (⊕ω∈Ω Hω , µ)L2 .

Consider the operator T : L2 (Ω × K) −→ H which is defined by

Z Z
hTϕ, hi = ϕ(ω, k)huω,k , hidl(k)dµ(ω)
ZΩ XK

= ϕ(ω, k)huω,k , hidµ(ω), ϕ ∈ L2 (Ω × K), h ∈ H.

Ω k∈K
ω

To show that {uω,k }k∈Kω is a c-Riesz basis for H, it is enough to show that T is
one-to-one (by Theorem 2.1 in [16]). If Tϕ = 0, then for each h ∈ H,
Z X
0 = hTϕ, hi = ϕ(ω, k)hΛ∗ω eω,k , hidµ(ω)
Ω k∈K
ω
Z X
= h ϕ(ω, k)eω,k , Λω hidµ(ω)
Ω k∈K
ω
Z
= hΛ∗ω ψ, hidµ(ω) = hTΛ ψ, hi = 0,
Ω
P
where ψ(ω) = k∈Kω ϕ(ω, k)eω,k , ω ∈ Ω. So TΛ ψ = 0. Since TΛ is bounded
below, ψ = 0. But kψk = kϕk, so ϕ = 0. Hence T is one-to-one and it implies
{uω,k }k∈Kω is a c-Riesz basis for H.
Now, let {uω,k }ω∈Ω,k∈Kω be a c-Riesz basis for H. By Theorem 3.3 in [14], it
suffices to show that TΛ is one-to-one. If TΛ φ = 0, then for each h ∈ H,
Z Z X
0 = hTΛ φ, hi = hΛω φ(ω), hidµ(ω) = hΛ∗ω (
∗
hφ(ω), eω,k ieω,k ), hidµ(ω)
Ω Ω k∈Kω
Z X
= hφ(ω), eω,k ihΛ∗ω eω,k , hidµ(ω) = (∗),
Ω k∈K
ω

set ϕ(ω, k) = hφ(ω), eω,k i, ω ∈ Ω, k ∈ Kω , then

Z X Z
2
|hφ(ω), eω,k i| dµ(ω) = kφ(ω)k2 dµ(ω) = kφk2 .
Ω k∈K Ω
ω

So ϕ ∈ L2 (Ω, K) and kϕk = kφk. Also

Z Z
(∗) = ϕ(ω, k)hΛ∗ω eω,k , hidl(k)dµ(ω).
Ω K

So for each h ∈ H, 0 = hTϕ, hi = 0, hence Tϕ = 0. Since {uω,k }ω∈Ω,k∈Kω is a

c-Riesz basis for H, then T is invertible, which implies ϕ = 0 and φ = 0. Thus
TΛ is one-to-one.
Now, suppose that {Λω }ω∈Ω is a cg-orthonormal basis for H. For almost all ν ∈ Ω
90 MORTEZA RAHMANI

and all m ∈ K,
Z Z Z Z
huω,k , uν,m idl(k)dµ(ω) = hΛ∗ω eω,k , Λ∗ν eν,m idl(k)dµ(ω)
Ω K Ω K
Z Z
= hΛ∗ω eω,k , Λ∗ν eν,m idµ(ω)dl(k)
ZK Ω
= heν,k , eν,m idl(k) = 1.
K
Also, for each h ∈ H,
Z Z Z
2
kΛω hk dµ(ω) = |hh, uω,k i|2 dl(k)dµ(ω) = khk2 .
Ω Ω K
So {uω,k }ω∈Ω,k∈Kω is a c-orthonormal basis.
The opposite implications are similar. 

3. cg-Orthonormal bases and cg-frames

At first, we present some result on cg-frames which are constructed by com-
posing with operators.
Proposition 3.1. Let {Λω }ω∈Ω be a cg-frame for H with bounds A and B and
V ∈ B(H). Then {Λω V }ω∈Ω is a cg-frame for H if and only if there exists a
positive constat α such that
kV f k2 ≥ αkf k2 , f ∈ H.
Proof. Suppose {Λω V }ω∈Ω is a cg-frame for H with bounds C and D. For each
f ∈ H, Z
Ckf k2 ≤ kΛω V f k2 dµ(ω) ≤ BkV f k2 ,
Ω
C C
so kV f k2 ≥ B kf k2 . Set α = B , then the proof is done.
Conversely, let α be such that
kV f k2 ≥ αkf k2 , f ∈ H.
For each f ∈ H,
Z
kΛω V f k2 dµ(ω) ≤ BkV f k2 ≤ BkV k2 kf k2 ,
Ω
and Z
kΛω V f k2 dµ(ω) ≥ AkV f k2 ≥ Aαkf k2 .
Ω
Hence {Λω V }ω∈Ω is a cg-frame for H. 
Corollary 3.1. Let {Λω }ω∈Ω be a cg-frame for H and V ∈ B(H). Then
{Λω V }ω∈Ω is a cg-frame for H if and only if V ∗ is onto.
Proof. By Lemma 2.4.1 (iii) in [5], it is obvious. 
Corollary 3.2. Let M be a close subspace of H and {Λω }ω∈Ω be a cg-frame for
H and V ∈ B(H, M ). Then {Λω V ∗ }ω∈Ω is a cg-frame for M if and only if there
exists a positive constat α such that
kV ∗ f k2 ≥ αkf k2 , f ∈ M.
 CHARACTERIZATION OF CONTINUOUS g-FRAMES VIA OPERATORS 91

Corollary 3.3. Let {Λω }ω∈Ω be a tight cg-frame for H with frame bound A and
V ∈ B(H). Then {Λω V }ω∈Ω is a tight cg-frame for H with frame bound α if and
only if
α
kV f k2 = kf k2 , f ∈ H.
A
Proposition 3.2. Suppose that {Θω }ω∈Ω is a cg-orthonormal basis for H with
respect to {Hω }ω∈Ω and {Λω }ω∈Ω is a cg-frame for H with respect to {Hω }ω∈Ω .
Then there exists a bounded and one-to-one operator V on H such that Λω =
Θω V , a.e. [µ]. Furthermore, V is invertible if {Λω }ω∈Ω is a cg-Riesz basis for H
and V is unitary if {Λω }ω∈Ω is a cg-orthonormal basis for H.
Proof. By the proof of Theorem 2.2, the first part is obvious. If {Λω }ω∈Ω is a cg-
Riesz basis for H, then by definition of V in the proof of Theorem 2.2, V = TΘ TΛ∗ .
Theorem 3.3 in [14] implies that TΛ∗ is onto, So V is onto and consequently V is
invertible. If {Λω }ω∈Ω is a cg-orthonormal for H, then Theorem 2.3 implies the
result. 
Proposition 3.3. Suppose that {Θω }ω∈Ω is a cg-orthonormal basis for H with
respect to {Hω }ω∈Ω and {Λω }ω∈Ω is a cg-frame for H with respect to {Hω }ω∈Ω .
Then there exist cg-orthonormal bases {Ψω }ω∈Ω , {Γω }ω∈Ω and {Φω }ω∈Ω for H
with respect to {Hω }ω∈Ω and a constant α such that
Λω = α(Ψω + Γω + Φω ), a.e. [µ].
Proof. Due to Proposition 3.2 and Proposition 1.3, we have an operator V ∈
B(H) so that V = α(U1 + U2 + U3 ), where each Uj , j = 1, 2, 3, is a unitary
operator and α is a constant. Then Λω = Θω V = α(Θω U1 + Θω U2 + Θω U3 ),
a.e. [µ]. It is obvious that every {Θω Uj }ω∈Ω , j = 1, 2, 3, is a cg-orthonormal basis
for H. Assuming Ψω = Θω U1 , a.e. [µ], Γω = Θω U2 , a.e. [µ] and Φω = Θω U3 ,
a.e. [µ], the proof is completed. 
Proposition 3.4. Consider {Θω }ω∈Ω as a cg-orthonormal basis for H with re-
spect to {Hω }ω∈Ω . If {Λω }ω∈Ω is a cg-Riesz basis for H, then {Λω }ω∈Ω is sum
of two cg-orthonormal bases for H with respect to {Hω }ω∈Ω .
Proof. Let {Λω }ω∈Ω is a cg-Riesz basis for H. Proposition 3.2 implies that there
exists an invertible V ∈ B(H) such that Λω = Θω V , a.e. [µ]. Via Proposition
1.3, V can be written as V = aU1 + bU2 , where U1 and U2 are unitary operators.
The rest of proof is similar to the proof of Proposition 3.3. 
Composing of a cg-orthonormal basis and an isometry, gives us a Parseval
cg-frame.
Proposition 3.5. If V ∈ B(H) is an isometry and {Θω }ω∈Ω is a cg-orthonormal
basis for H, then {Θω Λ}ω∈Ω is a Parseval cg-frame for H.
Proof. A straightforward calculation gives the proof. 
Every bounded operator V on H has a representation in the form V = U |V |
(called the polar decomposition of √V ), where U is a partial isometry, |V | is a
positive operator defined by |V | = V ∗ V and kerU = kerV .
Also, every positive operator P on H with kP k ≤ 1 can be written in the form
92 MORTEZA RAHMANI
√
P = 12 (W + W ∗ ), where W = P + i 1 − P 2 is unitary.

Next theorem shows that we can represent a cg-frame by some Parseval cg-
frames.
Theorem 3.1. Suppose that {Θω }ω∈Ω is a cg-orthonormal basis for H. Every
cg-frame for H can be written as a linear combination of two Parseval cg-frames.
Proof. By Proposition 3.2, there exists a bounded and one-to-one operator V ∈
B(H) such that Λω = Θω V , a.e. [µ]. By above note, V can be written as
V = 12 (U W +U W ∗ ), where U is an isometry and W is unitary. So U W and U W ∗
are isometries. Proposition 3.5 implies that {Θω U W }ω∈Ω and {Θω U W ∗ }ω∈Ω are
Parseval cg-frames for H. 
Now, we can show each cg-frame as a combination of a cg-orthonormal basis
and a cg-Riesz basis of H.
Theorem 3.2. Assuming {Θω }ω∈Ω as a cg-orthonormal basis for H, Every cg-
frame for H is sum of a cg-orthonormal basis for H and a cg-Riesz basis for
H.
Proof. The proof is similar to the proof of Theorem 4.2 in [16].


References
[1] M. R. Abdollahpour and M. H. Faroughi, Continuous G-Frames in Hilbert spaces,
Southeast Asian Bull. Math., 32 (2008), 1–19.
[2] S. T. Ali, J. P. Antoine and J. P. Gazeau, Coherent States, Wavelets and their
Generalizations, Springer-Verlag, 2000.
[3] S. T. Ali, J. P. Antoine and J. P. Gazeau, Continuous frames in Hilbert space, Ann.
Phys., 222 (1993), 1–37.
[4] P. G. Casazza, Every frame is a sum of three (but not two) orthonormal bases and
other frame representations, J. Fourier Anal. Appl., 4(6) (1998), 727–732.
[5] O. Christensen, Frames and Bases: An Introductory Course, Birkhauser, Boston,
2008.
[6] I. Daubechies, A. Grossmann and Y. Meyer, Painless nonorthogonal Expansions, J.
Math. Phys., 27(5) (1986), 1271–1283.
[7] R. J. Duffin and A. C. Schaeffer, A class of nonharmonic Fourier series, Trans. Amer.
Math. Soc., 72 (1952), 341–366.
[8] M. H. Faroughi, R. Ahmadi and M. Rahmani, Bochner (p, Y )-operator frames, In-
volve, 5:3 (2012), 283–293.
[9] M. H. Faroughi and M. Rahmani, Bochner pg-frames, J. Inequal. Appl., 2012
(2012), 196.
[10] M. H. Faroughi and M. Rahmani, Bochner p-Frame of Subspaces, Southeast Asian
Bull. Math., 37 (2013), 181–192.
[11] J-P. Gabardo, D. Han, Frames associated with measurable spaces, Adv. Comput.
Math., 18(2-4) (2003), 127–147.
[12] X. Guo, Operator parameterizations of g-frames, Taiwanese J. Math., 1 (18) (2014),
313–328.
[13] G. Kaiser, A Friendly Guide to Wavelets, Birkhauser, Boston, 1994.
[14] M. Madadian and M. Rahmani, g-Frame Sequence Operators, cg-Riesz Bases and
Sum of cg-Frames, Int. Math. Forum, 6 (68) (2011), 335–3369.
 CHARACTERIZATION OF CONTINUOUS g-FRAMES VIA OPERATORS 93

[15] A. Rahimi, A. Najati and Y. N. Dehghan, Continuous frames in Hilbert spaces,

Methods Funct. Anal. Topology, 12(2) (2006), 170–182.
[16] M. Rahmani, Sums of c-Frames, c-Riesz bases and Orthonormal mappings, U.P.B.
Sci. Bull., Series A, 77(3) (2015), 3–14.

Morteza Rahmani
Young Researchers and Elite Club, Ilkhchi Branch, Islamic Azad University,
Ilkhchi, Iran.
E-mail address: [email protected]

Received: May 30, 2019; Revised: July 24, 2019; Accepted: January 16, 2020