Infinite-time admissibility under compact perturbations
Jochen Schmid

TL;DR
This paper explores how infinite-time admissibility of semigroups is affected by compact perturbations, demonstrating through examples that it is not necessarily preserved even under strong stability.
Contribution
It provides the first examples showing that infinite-time admissibility can be lost under compact perturbations of the generator.
Findings
Infinite-time admissibility is not preserved under compact perturbations.
Both original and perturbed semigroups can be strongly stable.
Examples illustrate the non-preservation of admissibility.
Abstract
We investigate the behavior of infinite-time admissibility under compact perturbations. We show, by means of two completely different examples, that infinite-time admissibility is not preserved under compact perturbations of the underlying semigroup generator , even if and both generate strongly stable semigroups.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsQuantum chaos and dynamical systems · Mathematical Dynamics and Fractals · Stability and Controllability of Differential Equations
Infinite-time admissibility under compact perturbations
Jochen Schmid1,2
1 Institut für Mathematik
Universität Würzburg
97074 Würzburg
Germany
2 Fraunhofer Institute for Industrial Mathematics (ITWM)
67663 Kaiserslautern
Germany
Abstract
We investigate the behavior of infinite-time admissibility under compact perturbations. We show, by means of two completely different examples, that infinite-time admissibility is not preserved under compact perturbations of the underlying semigroup generator , even if and both generate strongly stable semigroups.
Index terms: infinite-time admissibility, compact perturbations, stabilization of collocated linear systems
1 Introduction
In this note, we investigate the behavior of infinite-time admissibility under compact perturbations of the underlying semigroup generator. So, we consider semigroup generators (with a Hilbert space) and possibly unbounded control operators (defined on another Hilbert space ) and we ask how the property of infinite-time admissibility of behaves under compact perturbations of the generator . Infinite-time admissibility of for means that for every control input the mild solution of the initial value problem
[TABLE]
is a bounded function from with values in . (A priori, the mild solution has values only in the extrapolation space of and, a fortiori, need not be bounded in the norm of , of course.)
It is well-known that (finite-time) admissibility is preserved under very general perturbations of the generator , in particular, under bounded perturbations. It is also clear that infinite-time admissibility, by contrast, is not preserved under bounded perturbations. Just think of a generator of an exponentially stable semigroup and a bounded perturbation (for example, a sufficiently large multiple of the identity) such that has spectral points in the right half-plane.
In this note, we will show by way of two completely different kinds of examples that infinite-time is also not preserved under compact perturbations which are such that both and generate stronly stable (but not exponentially stable) semigroups. So, in other words, we show that there exist semigroup generators and with being compact and a control operator such that
- •
the semigroups and are strongly stable but not exponentially stable
- •
is infinite-time admissible for but not infinite-time admissible for .
In our first – more elementary – example, we will use an old and well-known result from the 1970s, namely a stabilization result for collocated linear systems. In that example, the compact perturbation will be of rank and the control operator will be bounded. In particular, none of the technicalities coming along with unbounded control operators will bother us there. In our second – less elementary – example, we will use a more advanced result from the 1990s, namely a characterization of infinite-time admissibility for diagonal semigroup generators. In that example, the control operator will be unbounded and the compact perturbation will be of rank .
In the entire note, we will use the following notation.
[TABLE]
As usual, denotes the Banach space of bounded linear operators between two Banach spaces and and stands for the operator norm on . Also, denotes the norm of a square-integrable function with values in the Banach space . And finally, for a semigroup generator and bounded operators between appropriate spaces, the symbol will stand for the state-linear system [3]
[TABLE]
2 Some basic facts about admissibility and infinite-time admissibility
In this section, we briefly recall the definition of and some basic facts about admissibility and infinite-time admissibility. If is a semigroup generator on the Hilbert space and is the corresponding extrapolation space, then an operator (with another Hilbert space) is called control operator for . Also, is called a bounded control operator iff and an unbounded control operator iff . See [9] (Section 2.10) or [4] (Section II.5) for basic facts about extrapolation spaces.
Definition 2.1**.**
Suppose is a semigroup generator on and , where are both Hilbert spaces. Then is called admissible for iff for every
[TABLE]
is a function with values in , where is the generator of the continuous extension of the semigroup to .
Clearly, for a given semigroup generator every bounded control operator is admissible (because for ). It should also be noted that if is admissible for , then for every the linear operator defined in (2.1) is closed and thus continuous by the closed graph theorem. Consequently, is admissible for if and only if
[TABLE]
Definition 2.2**.**
Suppose is a semigroup generator on and , where are both Hilbert spaces. Then is called infinite-time admissible for iff for every
[TABLE]
is a function with values in that is bounded (in the norm of ), where is the generator of the continuous extension of the semigroup to .
Clearly, if is infinite-time admissible for a given semigroup generator , then it is also admissible for . It should also be noted that, by the uniform boundedness principle, is infinite-time admissible for if and only if
[TABLE]
Some authors [8], [2], [10] use the term input-stability for the system instead of infinite-time admissibility.
It is well-known that admissibility is preserved under bounded perturbations.
Proposition 2.3**.**
Suppose is a semigroup generator on and , where are both Hilbert spaces. Also, let . Then is admissible for if and only if is admissible for .
In fact, the conclusion of this proposition remains true for much more general perturbations , namely for perturbations of the (feedback) form , where is an admissible control operator for and with an arbitrary Hilbert space. See Corollary .5.5.1 from [9], for instance.
Proposition 2.4**.**
Suppose is the generator of an exponentially stable semigroup on and is admissible for . Then is even infinite-time admissible for .
See Proposition 4.4.5 in [9], for instance, and notice that for bounded control operators the above proposition is trivial. In view of that proposition, it is clear that infinite-time admissibility – unlike admissibility – is not preserved under bounded perturbations. Choose, for example, a bounded generator of an exponentially stable semigroup and let and (identity operator on ).
3 An example using a stabilization result for collocated linear systems
3.1 Stabilization of collocated linear systems
We will use the following well-known stabilization result for collocated systems, that is, systems of the form with a bounded control operator . It essentially goes back to [1] (Corollary 3.1) and, in the form below, can be found in [8] (Lemma 2.2.6), for instance. (Actually, for the more general version with the countability assumption on we have to refer to [10], but this more general version will not be used in the sequel.)
Theorem 3.1**.**
Suppose is a contraction semigroup generator on a Hilbert space with compact resolvent (or, more generally, with being countable). Suppose further that with another Hilbert space and that is approximately controllable in infinite time or approximately observable in infinite time. Then
- (i)
* is infinite-time admissible for , more precisely,*
[TABLE]
- (ii)
* is a strongly stable contraction semigroup on .*
A far-reaching generalization of this result to the case of unbounded control operators was obtained by Curtain and Weiss [10]. See Theorem 5.1 and 5.2 in conjunction with Proposition 1.5 from [10]. We also refer to [2] for a parallel result on exponential stabilization.
3.2 Infinite-time admissibility under compact perturbations
Example 3.2**.**
Set and let be defined by
[TABLE]
where and with
[TABLE]
Set and let be defined by
[TABLE]
where
[TABLE]
Clearly, and therefore . We now define
[TABLE]
and show, in various steps, that and are generators of strongly but not exponentially stable contraction semigroups on , that for a compact perturbation of rank one, and that is infinite-time admissible for but not infinite-time admissible for .
As a first step, we observe that with and that has rank one (because the same is true for ), whence is compact.
As a second step, we observe from
[TABLE]
that is the generator of a strongly stable but not exponentially stable contraction semigroup on .
As a third step, we show that is not infinite-time admissible for . In view of (2.4) we have to show that
[TABLE]
We first observe by Fatou’s lemma that
[TABLE]
for every and . Setting for and , we see that
[TABLE]
for every . Combining now (3.2), (3.3) and (3.4) we get
[TABLE]
Since \sup_{n\in\mathbb{N}}\big{(}n|b_{n}|^{2}\big{)}\geq\sup_{n\in I_{1}}\big{(}n|b_{n}|^{2}\big{)}=\infty, the desired relation (3.1) follows.
As a fourth step, we show that is infinite-time admissible for and that is the generator of a strongly stable contraction semigroup on . In order to do so, we apply the stablization theorem above (Theorem 3.1). Since
[TABLE]
we see that is a contraction semigroup generator on with compact resolvent, and since the eigenvalues of are pairwise distinct and for every , we see that the collocated linear system is approximately controllable and approximately observable in infinite time (Theorem 4.2.3 of [3]). So, by the stablization theorem above (Theorem 3.1), is infinite-time admissible for and is a strongly stable contraction semigroup on .
As a fifth and last step, we convince ourselves that the semigroup generated by is not exponentially stable. Assume the contrary. Then there exist and such that and
[TABLE]
So, since as , we conclude that
[TABLE]
We now observe that
[TABLE]
Combining (3.5) and (3.6) we arrive at
[TABLE]
Contradiction!
4 An example using an admissibility result for diagonal linear systems
4.1 Characterization of infinite-time admissibility
We will use the following well-known characterization of infinite-time admissibility for diagonal semigroup generators . It essentially goes back to [5] (Proposition 2.2) and can also be found in [9] (Theorem 5.3.9 in conjunction with Remark 4.6.5), for instance.
Theorem 4.1**.**
Suppose with a countable infinite index set and let be the diagonal operator given by
[TABLE]
where and for every . Suppose further that with , that is,
[TABLE]
for a uniqe sequence . Then the following statements are equivalent:
- (i)
* is infinite-time admissible for *
- (ii)
there exists a constant such that
[TABLE]
Clearly, in the situation of the above theorem the condition (ii) is equivalent to the existence of a constant such that
[TABLE]
A far-reaching generalization of the above theorem to the case of general contraction semigroup generators on a separable Hilbert space was obtained by Jacob and Partington [6]. See Theorem 1.3 from [6]. It states that for a contraction semigroup generator on a separable Hilbert space a control operator with is infinite-time admissible if and only if there is a constant such that the resolvent estimate (4.1) is satisfied. We also refer to [7] and [9] (Section 5.6) for an overview of many more admissibility results, for example, for infinite-dimensional input-value spaces .
4.2 Infinite-time admissibility under compact perturbations
Example 4.2**.**
Set and let and be defined by
[TABLE]
where and with
[TABLE]
Set and let be defined by
[TABLE]
where
[TABLE]
Clearly, and whence . And therefore
[TABLE]
We now show, in various steps, that and are generators of strongly but not exponentially stable contraction semigroups on , that for a compact perturbation of infinite rank, and that is infinite-time admissible for but not infinite-time admissible for .
As a first step, we observe from
[TABLE]
that and are generators of strongly stable but not exponentially stable contraction semigroups on .
As a second step, we observe that for a compact operator of infinite rank. Indeed, the operator defined by
[TABLE]
is a bounded operator on because is a bounded sequence. Also, is the limit in norm operator topology of the finite-rank operators defined by
[TABLE]
and therefore is compact, as desired.
As a third step, we show that is infinite-time admissible for . We have that
[TABLE]
for every and that
[TABLE]
So, by the admissibility theorem above (Theorem 4.1), the claimed infinite-time admissibility of for follows from (4.2) and (4.3).
As a fourth and last step, we show that is not infinite-time admissible for . We have that
[TABLE]
for every and . Choosing for , we see from (4.4) that
[TABLE]
So, by the admissibility theorem above (Theorem 4.1), is not infinite-time admissible for , as desired.
Acknowledgements
I would like to thank the German Research Foundation (DFG) financial support through the grant “Input-to-state stability and stabilization of distributed-parameter systems” (DA 767/7-1).
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[1] C. Benchimol: A note on weak stabilization of contraction semigroups. SIAM J. Contr. Optim. 16 (1978), 373-379
- 2[2] R. Curtain, G. Weiss: Exponential stabilization of well-posed systems by colocated feedback. SIAM J. Contr. Optim. 45 (2006), 273-297
- 3[3] R. Curtain, H. Zwart: An introduction to infinite-dimensional linear systems theory. 1st edition. Springer (1995)
- 4[4] K.-J. Engel, R. Nagel: One-parameter semigroups for linear evolution equations. Springer, 2000.
- 5[5] P. Grabowski: Admissibility of observation functionals. Int. J. Contr. 62 (1995), 1161-1173
- 6[6] B. Jacob, J. Partington: The Weiss conjecture on admissibility of observation operators for contraction semigroups. Integr. Equ. Oper. Theory 40 (2001), 231-243
- 7[7] B. Jacob, J. Partington: Admissibility of control and observation operators for semigroups: A survey. Operator Theory Adv. Appl., 149 (2004), 199–221
- 8[8] J. Oostveen: Strongly stabilizable distributed parameter systems. SIAM Frontiers in Applied Mathematics (2000)
