TADMOR: STANDARD PROBLEM IN SYSTEMS WITH SINGLE INPUT DELAY 385
Here and the bounded linear operators
are defined as
and . In the
framework of [5] and [49], the complete state of (10) is then the
pair
, which resides in the higher dimensional “ ”
state space
.
Our framework in (8), where the complete state is
,
is therefore a simplified version, as compared with the general
setting in [5] and [49]. This simplification is allowed, without
affecting the basic logic of the results in the cited papers, due
to the following specific features: 1) As defined above, the de-
pendence of both
and on the component of is
restricted to
. Particularly, the vector formed by
the first
entries of is . This allows us to replace
the component
in the complete state by , without losing
pertinent information. 2) The last
entries of are zero.
Thus, the subspace of complete-states
, where the last
entries of are zero, is an invariant subspace under
(10). Focusing on that subspace, the last
entries of
can be removed from the state without affecting the arguments
in [5], [49]. The resulting state is the current choice of
.
A standard practice (see, e.g., [31]) that we follow here is
to consider, along with the semigroup
, its restriction to
the dense submanifold
(embedded with the
topology) and an extension to the space (where gen-
erates the adjoint semigroup and
is also endowed with
the graph topology). The restriction is obvious. The purpose of
the following observations is to obtain a concrete, workable rep-
resentation of the extended semigroup. The reader is alerted to
the fact that the representation here is (slightly) different than
some other such representations that can be found in the litera-
ture, taking advantage of the relative simplicity of our system.
Invoking the definition of the adjoint of an operator over
the Hilbert space
, the following realization of the relation
is easily obtained:
else.
(11)
The infinitesimal generator of
is computed as the formal
adjoint of
, leading to
(12)
over the domain
and
.
By the representation of
, a member
is uniquely determined by, and therefore can be identified
with, the pair
. This representation
defines a linear homeomorphism between
, with the
-topology and the Hilbert space . In particular,
it endows
with a Hilbert space structure, including an
inner product and norm, inherited from the
structure and
consistent with the
-topology.
Using this homeomorphic representation of
by ,
the adjoint space,
can be identified with . Since
is a Hilbert space, we can identify , and elements
of the adjoint space must be represented by pairs
. That is, the pairing must be
defined via
(13)
where the element
is represented by the pair
. A Hilbert space structure in , con-
sistent with the adjoint-space topology, is therefore defined in
terms of the
inner products and norms of the representative
pairs
.
Relative to the representations and Hilbert space structures of
and , we introduce the notation of a continuous
injection
, to maintain the equality
(14)
for all
and . Thus defined, the
precise forms of the injection
, of its (unbounded, left) inverse
and of their adjoints, are computed by a
straightforward integration by parts, leading to
(15)
For notational simplicity, the extension of the semigroup
, defined over the entire , will still be denoted .
The extended semigroup is defined as a continuous extension
of
(combining (9) and (15)) from the dense
submanifold
. This readily leads to
else.
(16)
Noting the similarity to (9), the generator of the extended semi-
group, denoted
, is defined over the domain
, via
(17)
In particular,
is continuous over (relative to the
topology) and it satisfies for .
The following comments concern the definitions of the
input and output operators in (8), their adjoints, and their
extensions from
to . is
a bounded operator and it boundedly extends to a map-
ping into
, via . Its adjoint is