Home Luenberger compensator theory for heat-Kelvin-Voigt-damped-structure interaction models with interface/boundary feedback controls
Article Open Access

Luenberger compensator theory for heat-Kelvin-Voigt-damped-structure interaction models with interface/boundary feedback controls

  • Roberto Triggiani and Xiang Wan EMAIL logo
Published/Copyright: June 22, 2023

Abstract

An optimal, complete, continuous theory of the Luenberger dynamic compensator (or state estimator or state observer) is obtained for the recently studied class of heat-structure interaction partial differential equation (PDE) models, with structure subject to high Kelvin-Voigt damping, and feedback control exercised either at the interface between the two media or else at the external boundary of the physical domain in three different settings. It is a first, full investigation that opens the door to numerous and far reaching subsequent work. They will include physically relevant fluid-structure models, with wave- or plate-structures, possibly without Kelvin-Voigt damping, as explicitly noted in the text, all the way to achieving the ultimate discrete numerical theory, so critical in applications. While the general setting is functional analytic, delicate PDE-energy estimates dictate how to define the interface/boundary feedback control in each of the three cases.

MSC 2010: 93C20; 35Q93; 37C50

1 Introduction

The present article was conceived in response to the intents of the special issue of Open Mathematics. In fact, it provides a new, optimal, rather complete and comprehensive theory on a control theory topic of long standing relevance to applications, with the focus on a recently introduced, by necessity special, class of coupled partial differential equation (PDE) models: heat-structure interaction (HSI) models with high, Kelvin-Voigt damping for the wave-structure, subject to feedback control exercised on the boundary.

Homogeneous (uncontrolled) model. Throughout, Ω f R n , n = 2 , or 3, will denote the bounded domain on which the heat component of the coupled PDE system evolves. Its boundary will be denoted here as Ω f = Γ s Γ f , Γ s Γ f = , with each boundary piece being sufficiently smooth. Moreover, the geometry Ω s , immersed within Ω f , will be the domain on which the structural component evolves with time. As configured then, the coupling between the two distinct heat (fluid) and elastic dynamics occurs across boundary interface Γ s = Ω s ; see Figure 1. In addition, the unit normal vector ν ( x ) will be directed away from Ω f , and so toward Ω s . (This specification of the direction of ν will influence the computations to be done in the following text.)

Figure 1 
               The fluid-structure interaction.
Figure 1

The fluid-structure interaction.

On this geometry in Figure 1, we thus consider the following heat-structure PDE model in solution variables u = [ u 1 ( t , x ) , u 2 ( t , x ) , , u n ( t , x ) ] (the heat component here replacing the usual velocity field as a first step in this new investigation), and w = [ w 1 ( t , x ) , w 2 ( t , x ) , , w n ( t , x ) ] (the structural displacement field) with boundary conditions (BC) and initial conditions (IC):

( PDE ) u t Δ u = 0 in ( 0 , T ] × Ω f ; (1.1a) w t t Δ w Δ w t + b w = 0 in ( 0 , T ] × Ω s ; (1.1b)

( BC ) u Γ f = 0 on ( 0 , T ] × Γ f ; (1.1c) u = w t on ( 0 , T ] × Γ s ; (1.1d) ( w + w t ) ν = u ν on ( 0 , T ] × Γ s ; (1.1e)

(1.1f) (IC) [ w ( 0 , ) , w t ( 0 , ) , u ( 0 , ) ] = [ w 0 , w 1 , u 0 ] H b .

The constant b in (1.1b) will take up either the value b = 0 , or else the value b = 1 , as explained below. Accordingly, the space of well-posedness is taken to be the finite energy space

H b = H 1 ( Ω s ) \ R × L 2 ( Ω s ) × L 2 ( Ω f ) , b = 0 ; (1.2a) H 1 ( Ω s ) × L 2 ( Ω s ) × L 2 ( Ω f ) , b = 1 , (1.2b)

for the variable [ w , w t , u ] . (We are using the common notation H s [ H s ] n .) H b is a Hilbert space with the following norm inducing inner product, where ( f , g ) Ω Ω f g ¯ d Ω :

(1.2c) v 1 v 2 f , v ˜ 1 v ˜ 2 f ˜ H b = ( v 1 , v ˜ 1 ) Ω s + ( v 2 , v ˜ 2 ) Ω s + ( f , f ˜ ) Ω f , b = 0 ; ( v 1 , v ˜ 1 ) Ω s + ( v 1 , v ˜ 1 ) Ω s + ( v 2 , v ˜ 2 ) Ω s + ( f , f ˜ ) Ω f , b = 1 .

In (1.2a), the space H 1 ( Ω s ) \ R = H 1 ( Ω s ) \ const is endowed with the gradient norm. Relevant properties of this model, obtained in [1], will be reviewed in Section 1.3.

In particular, it was shown in [1] and reproduced in Section 1.3 that homogeneous problems (1.1a)–(1.1f) can be rewritten as the abstract model

(1.3) y = A y , y = [ w 1 , w 2 , u ]

with A = A b the generator of a s.c. contraction semigroup e A t , which is uniformly stable and analytic on H b ̲ .

Three controlled systems. We consider three cases of boundary/interface control g , as applied to systems (1.1a)–(1.1f).

  • CASE 1 (Section 2): the control g acts on the matching of the stresses condition (2.1e) at the interface between the two media, thus as a Neumann control as follows:

    (1.1eN) ( w + w t ) ν = u ν + g on ( 0 , T ] × Γ s .

    See the entire system (2.1a)–(2.1f) in Section 2.

  • CASE 2 (Section 4): the control g acts this time as a Dirichlet control on the matching of the “velocity” condition (1.1d) also at the interface between the two media as follows:

    (1.1dD) u = w t + g on ( 0 , T ] × Γ s .

    See the entire system (4.1a)–(4.1f) in Section 4.

  • CASE 3 (Section 5): the control g acts, still as a Dirichlet control, but this time as a boundary control on the external boundary of the heat domain in (1.1c) as follows:

    (1.1cD) u Γ f = g on ( 0 , T ] × Γ f .

    See the entire system (5.1a)–(5.1f) in Section 5.

In each case, the control g introduces a control operator , highly unbounded. Thus, the resulting model is now

y = A y + g ( 1.3-controlled )

with the highly unbounded control operator depending on the three cases, see equation (2.2) for CASE 1; equation (4.2) for CASE 2; and equation (5.13) for CASE 3. The selection of the operators entering into the development of the Luenberger theory, as described in Section 1.2.2, depends on the three cases.

Objective of the present article, as a template for future work. The initial objective in the present work is to offer a “continuous theory” of the long-standing, highly relevant topic corresponding to the Luenberger compensator program. The deliberate goal is to have this first contribution serve as a basis for further development of investigations in various directions. Among them, we cite:

  1. replacing the heat component with a fluid component, thus accounting for the pressure variable while keeping the Kelvin-Voigt structure (model as in [2] (wave) or as in [3] (plate)). The new technique inspired by the boundary control theory [4], which is required for the extension from the heat component to the fluid component is described in Appendix A, with a focus on the present uncontrolled model, following [2].

  2. replacing the present Kelvin-Voigt damped-wave with a Kelvin-Voigt damped-plate in modeling the structure. This in turn opens up a variety of different, physically relevant, coupling conditions at the interface between the two media (models as in [3] and [5]);

  3. obtain a corresponding “discrete theory” or (rigorous) numerical analysis theory, as done in past genuine PDE models of different types in [4, pp. 495–504], [68] and also [9, 10], to name just a few references. This is a very challenging and technical direction of investigation, and yet highly important in engineering applications);

  4. analysis of the same models as in (i) and (ii), and this time, however, with no Kelvin-Voigt damping, such as in [1114] (wave), where the original stability properties are different. Here, in contrast with the heat-structure case, λ = 0 is a simple eigenvalue of the uncontrolled model. Heat-viscoelastic plates are studied in [15,16];

  5. further extension of both the continuous and the discrete analysis to nonlinear models, with static interface [1724], and even with moving interface [25,26].

1.1 Historical orientation on Luenberger’s compensator theory

The Luenberger theory of “observers” was introduced for lumped (finite-dimensional) linear systems in 1971 [27], and it was met with great success [28, p. 48]. It subsequently stimulated investigations for PDE problems with boundary controls/boundary observations (infinite-dimensional systems with “badly unbounded” control and observation operators) of both parabolic and hyperbolic types [68], [4, pp. 495–504]. It consists, in its first phase, of a continuous theory, followed next by a rigorous numerical implementation, as in the aforementioned references. At the level of numerical implementation, it was in a sense rediscovered with the more recent topic of “data assimilation” that shares the same philosophy as the Luenberger discrete theory. Recent references include [2932]. A more detailed description of these topics is given below.

Step 1. The continuous theory. Here, in a purely informal manner, we shall provide the special setting that we shall select in our application of the continuous Luenberger’s dynamic compensator theory to heat-structure models. For a preliminary conceptual understanding, we may regard the operators below as being all finite-dimensional, in line with Luenberger’s original contribution [27]. Its standard representation is as follows:

y ˙ = A y + B g , g = F z = control , y ( 0 ) = y 0 , (1.4a) z ˙ = ( A + B F K C ) z + K ( C y ) , z ( 0 ) = z 0 . (1.4b)

The basic idea behind is that the full state y is inaccessible, unknown, beyond any measurement, as is often the case in applications. What we have instead at our disposal is the partial observation ( C y ), where C is the known observation operator. Examples abound: (i) the actual state within a furnace or (ii) the true distribution of “noise” within an acoustic chamber are not exactly accessible, and only some information from the boundary may be available in each case. Thus, the (compensator) z -equation (1.4b) is fed, or determined by, only the available partial observation ( C y ) . Subtracting (1.4b) from (1.4a) with B g = B F z , we obtain after a cancellation of the term B F z :

d d t [ y ( t ) z ( t ) ] = ( A K C ) [ y ( t ) z ( t ) ] , (1.5a) [ y ( t ) z ( t ) ] = e ( A K C ) t [ y 0 z 0 ] , t 0 . (1.5b)

One next assumes the detectability condition for the pair { A , C } : there exist K and k such that e ( A K C ) t M e k t , k > 0 . Thus, from (1.5b), we finally obtain

(1.6) [ y ( t ) z ( t ) ] = e ( A K C ) t [ y 0 z 0 ] M e k t y 0 z 0 , t 0

and the dynamic compensator z ( t ) , which is fed only by the known partial observation ( C y ) of the inaccessible state y , asymptotically approximates such state y ( t ) , at an exponential rate. This is the key of Luenberger’s theory in the lumped case where the state of the system is a finite dimensional vector. Nontrivial extensions were subsequently introduced and studied in the case of distributed parameter systems modeled by partial differential equations with boundary control/boundary observation [4, pp. 495–504], [68].

Step 2. The numerical theory. Particularly in the case of PDEs dynamics, it is critically important to provide a (finite element) approximation theory of dynamic compensators of Luenberger’s type for partially observed systems. The aforementioned PDE references include also the discrete/numerical Luenberger theory based on finite element method. The analysis is very technical.

Connections with data assimilation. In recent years, a numerical procedure called “data assimilation” has been introduced, particularly with emphasis on nonlinear dissipative PDE dynamics with finite degrees of freedom, which in spirit is closely related in terms of goals to the discrete Luenberger’s compensator theory. In common with the Luenberger’s theory, in the presence of inadequate knowledge of the original system, a suitable data assimilation algorithm is introduced to force its corresponding solution to approach the original solution at an exponential rate in time. This is done by having access “to data from measurements of the system collected at much coarser spatial grid than the desired resolution of the forecast” [30]. As expected, the efficiency of data assimilation relies also on the finite dimensionality of the proposed algorithm. Inspiration comes from a rigorous result on the 2D Navier-Stokes equations (NSE) given in [33], where it is proved that if a number of Fourier modes of two different solutions of the NSE have the same asymptotic behavior as t goes to infinity, then the remaining infinite number of modes also have the same asymptotic behavior.

It seems unfortunate that data assimilation theory introduced in 2014 has not apparently been aware of the large body of works in Luenberger theory, which was introduced in 1971, to include PDE parabolic and hyperbolic problems as in [4, pp. 495–504], [68]. Luenberger’s theory in these references emphasizes control/observation on the boundary, unlike the literature of data assimilation. The original Luenberger theory was for linear models, but it was later introduced for nonlinear models, which are the core of data assimilation. Moreover, data assimilation is passive in the sense that there is no control action. Mutual awareness and knowledge of the two communities’ research effort may well benefit both. In this spirit, being this an article on the Luenberger theory in systems of coupled PDEs with control/observation at “the boundary,” we are pleased to provide specific references in data assimilation.

The first main data assimilation was done in [29] in 2014. This gives the AOT algorithm. It describes the interpolation operators (nodes, modes, and volume averages) and uses the “nudging” algorithm, which is essentially interior control for the data assimilated problem. The technique utilizes the existence of finitely many determining functional to capture the essential asymptotic dynamics of the system.

The initial article that does data assimilation for 3D NSE without assuming any regularity of the solution is [34]. All previous works critically utilized the regularity of the 2D NSE to show asymptotic convergence of the data assimilated solution to the reference solution. In the absence of global regularity in the 3D case, the previous article [34] achieved the exact same result for the 3D NSE by imposing conditions on the observed (model) data. Next, we quote paper [30], which provides results on the Boussinesq system (and also, therefore, for 3D NSE). While [34] does data assimilation for the 3D NSE with the assumption that the reference solution is obtained via the Galerkin procedure, the article [30] makes no such assumptions and does data assimilation for a general (Leray-Hopf) weak solution, which obeys the energy identity corresponding to the system. Also, the article only uses velocity measurements to perform data assimilation.

Luenberger problems for three interface/boundary feedback controlled models: system (1.1a)–(1.1f) subject to control action g as in (1.1eN) (CASE 1); or (1.1dD) (CASE 2); or (1.1cD) (CASE 3), in the final form y = A y + g as in equation (1.3-controlled). The goal of the present article is to investigate the Luenberger’s dynamic compensator theory (continuous version) as applied to a class of fluid-structure interaction models, in the particular setting where the structure is subject to visco-elastic (Kelvin-Voigt) damping, as in (1.1a)–(1.1f). How to handle the corresponding fluid-wave model is described in Appendix A. This focuses on the new tricky technique that is required in the present homogeneous case (1.1a)–(1.1f) following [2]. In summary, in the present article, we consider a fluid (heat)-structure interaction model with high Kelvin-Voigt damping under three different scenarios: (1) in Part I, with Neumann control g at the interface Γ s as in (1.1eN); (2) in Part II, with Dirichlet control g at the interface Γ s as in (1.1dD); (3) in Part III, with Dirichlet control g at the external boundary Γ f as in (1.1cD).

1.2 Orientation on the contributions of the present article. Conceptual description of the mathematical setting and ultimate results

To ease the reading of this article, we find it most appropriate to provide a focused, synthetic orientation regarding both the mathematical setting of the article and its ultimate, sought-after Luenberger-type results.

1.2.1 Uncontrolled, homogeneous model

The uncontrolled model is a coupled heat-structure interaction, where the structure is modeled by a wave with strong Kelvin-Voigt damping, which interacts with a heat component through the interface between the two media, see the linear, coupled PDE system (1.1a)–(1.1f) and Figure 1. The state of the system is the triple y = { w , w t , u } , displacement, velocity of the elastic structure, and temperature. A comprehensive study of this model was carried out in [1]. Selected results to be used in the present article are reviewed in Section 1.3. The uncontrolled coupled system is described by an operator A , which is the generator of a s.c. contraction semigroup e A t on a natural finite energy functional setting. From the purpose of the Luenberger theory to be here investigated, its main feature is that such semigroup is uniformly (exponentially) stable, Theorem1.3(ii). This is due to the Kelvin-Voigt damping. An additional property of such semigroup is that it is analytic in its natural setting, also Theorem 1.3(ii). This analyticity property – also due to the Kelvin-Voigt damping – adds a positive feature to the uncontrolled dynamics. One then seeks, successfully, to retain it and propagate it to the corresponding Luenberger feedback problem, that is, the dynamics of the observer variable z , expressed in feedback-form with respect to the partial observation ( C y ) of the original unknown state y . But analyticity is not critical for the key Lueberger’s goal to recover asymptotically the originally unknown full state y by using the observer z .

1.2.2 Controlled systems y = A y + g

As already noted, we consider three cases of boundary/interface control g :

  • Case 1 (Section 2): The control g acts on the matching of the stresses condition (1.1f) at the interface between the two media (Neumann control). See the entire system (1.1a)–(1.1f) in Section 2.

  • Case 2 (Section 4): The control g acts this time as a Dirichlet control on the matching of the “velocity” condition (3.1d) also occurring at the interface between the two media. See the entire system (3.1a)–(3.1f) in Section 4.

  • Case 3 (Section 5): The control g acts, still as a Dirichlet control, but this time as a boundary control on the external boundary of the heat domain as in (4.1c). See the entire system (4.1a)–(4.1f) in Section 5.

  • CASE 1. Here, the following analysis selects the feedback form g = F z for the Neumann boundary control g at the interface, by taking the Luenberger operators as follows: F = , C = , K = . This way, the observer equation becomes: z = ( A 2 ) z + ( y ) , with the observation operator of the entire unknown state y . Ultimately, the difference [ y z ] between unknown state y and known observation z satisfies the equation d [ y z ] d t = ( A ) [ y z ] , where the feedback generator A in equation (2.15a) for CASE 1 is deliberately selected to preserve the property of dissipativity of the original free dynamic operator A . The ultimate goal of the analysis is then to establish that such feedback generator is uniformly (exponentially) stable on its natural setting. This is Theorem 3.1, equation (3.14) in CASE 1. This result is established by PDE methods in Section 3.1.3. This way, the known observation variable z , based only on partial knowledge of the unknown state y thorough the unbounded observation/trace operator , see equation (3.8), approaches the unknown full state y asymptotically at exponential speed, the key of the Luenberger theory.

  • CASE 2. Here, the following analysis selects the feedback form g = F z for the Dirichlet boundary control g at the interface, by a different choice from CASE 1: in fact, in CASE 2 one takes: F = , C = , K = , so that the observation equation now becomes: z = A z + ( y ) , with trace operator , see (4.7). This way the difference [ y z ] between unknown state y and known observation z , satisfies the same-looking equation as in CASE 1: d [ y z ] d t = ( A ) [ y z ] , with dissipative feedback generator. Again, the ultimate goal is to establish that such new feedback generator is uniformly (exponentially) stable in its natural setting. This is Theorem 4.1, equation (4.24). This result is again obtained by PDE methods in Section 4.2.4.

  • CASE 3. Here, the following analysis selects the feedback form g = F z for the Dirichlet boundary control g at the external boundary, which notationally is like CASE 2: F = , C = , K = , with a different operator of course, and hence, again observation equation z = A z + ( y ) , and finally the same desired form of [ y z ] : d [ y z ] d t = ( A ) [ y z ] . In this CASE 3, the ultimate goal is again to show that the new feedback operator ( A ) is uniformly stable. This is Theorem 5.1, equation (5.32), whose proof is given in Section 5.2.4. The operator is again a trace operator (cf. equation (5.15)).

  • Insight on the choice F = versus F = in the various cases. For CASE 2, this insight is given in (4.15), and in CASE 3, this insight is given in equation (5.23). In short, this is a purely PDE problem related to the operator for the purpose to achieve the feedback generator still dissipative. Thus, while the setting of the analysis is functional analytic, the key technical parts are based on PDE estimates.

  • Finally, to ease the reading, each case is dealt individually. In other words, one may read CASE 3 without knowledge of Cases 1 or 2.

1.3 Review of homogeneous heat-structure interaction model with Kelvin-Voigt damping: b = 0 , b = 1 [1]

We return to the homogeneous problem (1.1a)–(1.1f), which for easy reading, we reproduce here:

( PDE ) u t Δ u = 0 in ( 0 , T ] × Ω f ; (1.1a) w t t Δ w Δ w t + b w = 0 in ( 0 , T ] × Ω s ; (1.1b)

( BC ) u Γ f = 0 on ( 0 , T ] × Γ f ; (1.1c) u = w t on ( 0 , T ] × Γ s ; (1.1d) ( w + w t ) ν = u ν on ( 0 , T ] × Γ s ; (1.1e)

(1.1f) (IC) [ w ( 0 , ) , w t ( 0 , ) , u ( 0 , ) ] = [ w 0 , w 1 , u 0 ] H b .

The constant b in (1.1b) will take up either the value b = 0 , or else the value b = 1 , as explained later. Accordingly, the space of well-posedness is taken to be the finite energy space

H b = H 1 ( Ω s ) \ R × L 2 ( Ω s ) × L 2 ( Ω f ) , b = 0 ; (1.2a) H 1 ( Ω s ) × L 2 ( Ω s ) × L 2 ( Ω f ) , b = 1 , (1.2b)

for the variable [ w , w t , u ] . (We are using the common notation H s [ H s ] n .) H b is a Hilbert space with the following norm inducing inner product, where ( f , g ) Ω Ω f g ¯ d Ω :

(1.2c) v 1 v 2 f , v ˜ 1 v ˜ 2 f ˜ H b = ( v 1 , v ˜ 1 ) Ω s + ( v 2 , v ˜ 2 ) Ω s + ( f , f ˜ ) Ω f , b = 0 ; ( v 1 , v ˜ 1 ) Ω s + ( v 1 , v ˜ 1 ) Ω s + ( v 2 , v ˜ 2 ) Ω s + ( f , f ˜ ) Ω f , b = 1 .

In (1.2a), the space H 1 ( Ω s ) \ R = H 1 ( Ω s ) \ const is endowed with the gradient norm.

Abstract model of the homogeneous PDE problem (1.1a)–(1.1f). The operator A b and its adjoint A b , b = 0 , 1 . Basic results [1]. The abstract version of the homogeneous PDE model (1.1a)–(1.1f) is given as a first-order equation by

(1.7) d d t w w t u = A b w w t u ,

where the operator A b : H b D ( A b ) H b is given by

(1.8) A b v 1 v 2 h = 0 I 0 Δ b I Δ 0 0 0 Δ v 1 v 2 h = v 2 Δ ( v 1 + v 2 ) b v 1 Δ h .

A description of { v 1 , v 2 , h } D ( A b ) is as follows:

  1. (1.9a) v 1 , v 2 H 1 ( Ω s ) \ R for b = 0 ; v 1 , v 2 H 1 ( Ω s ) for b = 1 ; so that v 2 Γ s = h Γ s H 1 2 ( Γ s ) in both cases; Δ ( v 1 + v 2 ) L 2 ( Ω s ) ;

  2. (1.9b) h H 1 ( Ω f ) , Δ h L 2 ( Ω f ) , h Γ f 0 , h Γ s = v 2 Γ s H 1 2 ( Γ s ) ; h ν Γ s = ( v 1 + v 2 ) ν Γ s H 1 2 ( Γ s ) .

Remark 1.1

The aforementioned description of D ( A b ) in (1.9a)–(1.9b) shows that the point { v 1 , v 2 , h } D ( A b ) enjoys a smoothing of regularity by one Sobolev unit – from L 2 ( ) to H 1 ( ) – but only of the coordinates v 2 and h , with respect to the original finite energy state space H b in (1.2a). In contrast, the first coordinate v 1 experiences no smoothing: it is in H 1 ( Ω s ) , the first coordinate component of the space H b . This amounts to the fact that A has noncompact resolvent R ( λ , A ) on H b . Consistently, it was shown in [1, Proposition 2.4] that the point λ = 1 belongs to the continuous spectrum of A b : 1 σ c ( A b ) .

Theorem 1.1

(The adjoint A b of A b , b = 0 or 1 [1, Appendix A]). The H b -adjoint of the operator A b defined in (1.7)–(1.9) is given by

(1.10a) A b v ˜ 1 v 2 ˜ h ˜ = 0 I 0 Δ + b I Δ 0 0 0 Δ v ˜ 1 v 2 ˜ h ˜ = v ˜ 2 Δ ( v ˜ 2 v ˜ 1 ) + b v ˜ 1 Δ h ˜ .

The PDE version of

(1.10b) d d t v w u = A b v w u

is given by

( PDE ) u t Δ u = 0 i n ( 0 , T ] × Ω f ; (1.10c) w t t Δ w Δ w t + b w = 0 i n ( 0 , T ] × Ω s ; (1.10d)

( BC ) u Γ f = 0 o n ( 0 , T ] × Γ f ; (1.10e) u = w t o n ( 0 , T ] × Γ s ; (1.10f) ( w + w t ) ν = u ν on ( 0 , T ] × Γ s ; (1.10g)

compared with equation (1.1a)–(1.1e). The domain D ( A b ) of the operator A b in (1.10a) is described as follows (compared with D ( A b ) in (1.9a)–(1.9b)): { v ˜ 1 , v ˜ 2 , v ˜ 3 } D ( A b ) means:

  1. (1.11a) v ˜ 1 , v ˜ 2 H 1 ( Ω s ) \ R f o r b = 0 ; v ˜ 1 , v ˜ 2 H 1 ( Ω s ) f o r b = 1 ; s o t h a t v ˜ 2 Γ s = h ˜ Γ s H 1 2 ( Γ s ) i n b o t h c a s e s ; Δ ( v ˜ 2 v ˜ 1 ) L 2 ( Ω s ) ;

  2. (1.11b) h ˜ H 1 ( Ω f ) , Δ h ˜ L 2 ( Ω f ) , h ˜ Γ f 0 , h ˜ Γ s = v ˜ 2 Γ s H 1 2 ( Γ s ) ; h ˜ ν Γ s = ( v ˜ 2 v ˜ 1 ) ν Γ s H 1 2 ( Γ s ) .

Actually, [1, Appendix A] gives the detailed proof for b = 0 . For b = 1 , in [1, equation (A.4)] one adds the term ( w 2 , v 1 ) for the full H 1 -inner product but also the term b ( w 1 , w 2 ) with b = 1 .

Theorem 1.2

(Generation by A b and A b , b = 0 , b = 1 [1, Theorem 1.2])

  1. The operator A b defined by (1.8), (1.9) and its adjoint A b given by (1.10a) and (1.11) are dissipative: For [ v 1 , v 2 , h ] D ( A b ) , and [ v 1 , v 2 , h ] D ( A b ) , we have

    (1.12) Re A b v 1 v 2 h , v 1 v 2 h H b = v 2 Ω s 2 h Ω f 2 0 ,

    (1.13) Re A b v 1 v 2 h , v 1 v 2 h H b = v 2 Ω s 2 h Ω f 2 0 ,

    in the L 2 ( ) norms of Ω s and Ω f .

  2. Thus, A b and A b are maximal dissipative on H b . Then [35] gives that A b generates a s.c. ( C 0 ) -contraction semigroup e A b t on H b , which gives the unique solution of problems (1.1a)–(1.1f):

    (1.14) w 0 w 1 u 0 H b w ( t ) w t ( t ) u ( t ) e A b t w 0 w 1 u 0 C ( [ 0 , T ] ; H b ) .

    The same generation results hold also for A b on H b , with e A b t solving system (1.10c)–(1.10g).

Again, [1, Section 2] gives the proof only for b = 0 . For b = 1 , in [1, equation (2.2a)], one adds the terms ( v 2 , v 1 ) for the full H 1 -norm and the term b ( v 1 , v 2 ) , b = 1 , leading now to a new version of such equation (2.2a) in [1] given by v 2 2 h 2 + 2 Im ( v 2 , v 1 ) + 2 Im ( v 2 , v 1 ) . Thus, taking the real part of the aforementioned expression, one obtains (1.12) for b = 0 and b = 1 .

Theorem 1.3

With reference to the operator A b in (1.7)–(1.9) and its adjoint A b in (1.10a) and (1.11), both defined on H b , b = 0 , 1 , we have

  1. (1.15) 0 ρ ( A b ) , 0 ρ ( A b ) , ρ ( ) = r e s o l v e n t s e t

    with explicit expression of A b 1 given in [1, Lemma 2.2].

  2. [1, Theorem 1.4] The contraction semigroups e A b t and e A b t generated by Theorem 1.2are analytic and uniformly stable on H b , b = 0 , 1 ; there exist constants M 1 , δ > 0 , such that

    (1.16) e A b t ( H b ) + e A b t ( H b ) M e δ t , t 0 , b = 0 , 1 .

Remark 1.2

  1. Section 1.3 (a subset of [1]) shows that the natural functional setting for problems (1.1a)–(1.1f) is: the energy space H b = 0 in (1.2a) for b = 0 ; and the energy space H b = 1 in (1.2b) for b = 1 . In each such case, b = 0 and b = 1 , the free dynamic operator is maximal dissipative, it defines a corresponding expression for the adjoint A b and the resulting contraction semigroups e A b t and e A b t are analytic and uniformly stable. Analyticity in Theorem 1.3(ii) above is consistent with abstract results [3638], in view of the Kelvin-Voigt damping.

  2. If, however, one insists in considering problems (1.1a)–(1.1f) with b = 0 in the energy space H b = 1 with full H 1 -norm for the position variable, then stability is lost: more precisely, one can readily prove or verify that:

    (1.17) λ = 0 is a simple eigenvalue of the free dynamics operator A b = 0 ( with  b = 0 ) with corresponding eigenvector e = [ 1 , 0 , 0 ] D ( A b = 0 ) H b = 1 .

    In fact, setting equal to zero equation (1.8) with b = 0 implies v 2 0 ; hence, h 0 from Δ h = 0 , h Γ f = 0 , h Γ s = v 2 Γ s = 0 . This yields Δ v 1 = 0 in Ω s , v 1 ν Γ s = h ν Γ s = 0 by (1.9b), whose normalized solution is v 1 = 1 in H 1 ( Ω s ) (while it would be v 1 = 0 in H 1 ( Ω s ) \ R ).

  3. In this case, we may view the problem with b = 1 on H b = 1 as having “stabilized” (and regularized) the same problem with b = 0 on H b = 1 : A b = 1 = A b = 0 + S , with stabilizing operator S v 1 v 2 h = 0 v 1 0 .

2 CASE 1. Heat-structure interaction with Kelvin-Voigt damping: Neumann control g at the interface Γ s

The present article begins with this section. In the present CASE 1, we consider problem (1.1a)–(1.1f) subject this time to control g acting in the Neumann interface condition (1.1e); that is,

( PDE ) u t Δ u = 0 in ( 0 , T ] × Ω f ; (2.1a) w t t Δ w Δ w t + b w = 0 in ( 0 , T ] × Ω s ; (2.1b)

( BC ) u Γ f = 0 on ( 0 , T ] × Γ f ; (2.1c) u = w t on ( 0 , T ] × Γ s ; (2.1d) ( w + w t ) ν = u ν + g on ( 0 , T ] × Γ s ; (2.1e)

(2.1f) (IC) [ w ( 0 , ) , w t ( 0 , ) , u ( 0 , ) ] = [ w 0 , w 1 , u 0 ] H b ,

with Neumann boundary control g acting at the interface Γ s . The constant b in (1.1b) will take up either the value b = 0 , or else the value b = 1 , as explained earlier. Accordingly, the space of well-posedness is taken to be the finite energy space defined in (1.2a), or (1.2b).

2.1 Abstract model on H b , b = 0 , 1 , of the nonhomogeneous PDE model (2.1a)–(2.1f) with Neumann control g acting at the interface Γ s

This topic was duly treated in [39], at least for b = 0 . This will be reviewed below and complemented by the case b = 1 . In either case, the abstract version of the nonhomogeneous PDE model (2.1a)–(2.1f) is given by

(2.2) d d t w w t u = A b w w t u + N ( b ) g ,

with the full dynamic operator A b given by (1.8) and (1.9). The definition of the Neumann control N ( b ) depends on the two cases b = 0 and b = 1 , on the respective space H b . We shall provide a unified treatment covering the two cases b = 0 and b = 1 , which will recover the case b = 0 in [39].

We first define the positive self-adjoint operator A N , s ( b ) by

(2.3) A N , s ( b ) φ = ( Δ b I ) φ , D ( A N , s ( b ) ) = φ H 2 ( Ω s ) \ R for b = 0 φ H 2 ( Ω s ) for b = 1 : φ ν Γ s = 0 ,

so that A N , s ( b ) is defined on the first component space of the state space H b , b = 0 , or b = 1 . Next, following an established procedure [40], [4, Chapter 3], we define the following Neumann map N s ( b ) on the structure domain Ω s by

(2.4) ψ = N s ( b ) μ ( Δ b I ) ψ = 0 in Ω s , ψ ν Γ s = μ ,

with regularity [4043]

(2.5) N s ( b ) : H r ( Γ s ) H 3 2 + r ( Ω s ) \ R for b = 0 , H 3 2 + r ( Ω s ) for b = 1 , r real , N s ( b ) : L 2 ( Γ s ) H 3 2 ( Ω s ) \ R H 3 2 2 ε ( Ω s ) \ R = D ( A N , s ( b ) ) 3 4 ε for b = 0 , H 3 2 ( Ω s ) H 3 2 2 ε ( Ω s ) = D ( A N , s ( b ) ) 3 4 ε for b = 1 , continuously or ( A N , s ( b ) ) 3 4 ε N s ( b ) ( L 2 ( Γ s ) ; L 2 ( Ω s ) ) , b = 0 or b = 1 . [30, p. 195]

Next we return to (2.1b) and rewrite it via (2.4) and (2.1e) [40], [4, Chapter 3]

(2.6) w t t = Δ ( w + w t ) b w = ( Δ b I ) ( w + w t ) N s ( b ) u ν Γ s + g Γ s + b ( w + w t ) b w .

By (2.1), (2.1e), the term ( w + w t ) N s ( b ) u ν Γ s + g Γ s satisfies the zero Neumann B.C. of A N , s ( b ) in (2.3), so we can rewrite (2.6) as follows:

(2.7) w t t = A N , s ( b ) ( w + w t ) N s ( b ) u ν Γ s + g + b ( w + w t ) b w L 2 ( Ω s ) ,

(2.8) w t t = A ˜ N , s ( b ) ( w + w t ) + A ˜ N , s ( b ) N s ( b ) u ν Γ s + b ( w + w t ) b w + A ˜ N , s ( b ) N s ( b ) g [ D ( A N , s ( b ) ) ] ,

where now A ˜ N , s ( b ) is the isomorphic extension L 2 ( Ω s ) [ D ( A N , s ( b ) ) ] = dual of D ( A N , s ( b ) ) with respect to L 2 ( Ω s ) , of the original operator A N , s ( b ) in (2.3). Next, we pass to the “fluid” domain Ω f . We let A D , f be the negative, self-adjoint operator on L 2 ( Ω f ) by

(2.9) A D , f φ = Δ φ , D ( A D , f ) H 2 ( Ω f ) H 0 1 ( Ω f ) ,

while D f , s is the Dirichlet map from Γ s to Ω f defined by

φ = D f , s χ Δ φ = 0 in Ω f ; (2.10a) φ Γ f = 0 , φ Γ s = χ . (2.10b)

The following regularity holds true for D f , s : for any r ,

(2.11a) D f , s : H r ( Ω f ) H r + 1 2 ( Ω f ) , [33]

(2.11b) D f , s : L 2 ( Ω f ) H 1 2 ( Ω f ) H 1 2 2 ε ( Ω f ) D ( A D , f 1 4 ε ) ; [31, p. 181], [48]

continuously. Then, as usual [1], we rewrite the u -problem in (2.1a) via (2.10) as follows:

u t = Δ ( u D f , s ( w t Γ s ) ) in ( 0 , T ] × Ω f ; (2.12a) [ u D f , s ( w t Γ s ) ] Γ s = 0 in ( 0 , T ] × Γ s ; (2.12b) [ u D f , s ( w t Γ s ) ] Γ f = 0 in ( 0 , T ] × Γ f ; (2.12c)

(2.13) u t = A D , f ( u D f , s ( w t Γ s ) ) L 2 ( Ω f ) ; u t = A ˜ D , f ( u + A ˜ D , f D f , s ( w t Γ s ) ) [ D ( A D , f ) ] ,

where A ˜ D , f is its isomorphic extension L 2 ( Ω f ) [ D ( A D , f ) ] = dual of D ( A D , f ) with respect to L 2 ( Ω f ) as a pivot space. Henceforth, as in [4, Chapter 3] we drop the ˜ for the extensions A ˜ N , s ( b ) and A ˜ D , f for notational easiness, as no misunderstanding is likely to arise.

By combining (2.8) and (2.13), we obtain

(2.14) d d t w w t u = 0 I 0 A N , s ( b ) A N , s ( b ) + b I A N , s ( b ) N s ( b ) ν 0 A D , f D f , s ( Γ s ) A D , f w w t u + 0 A N , s ( b ) N s ( b ) g 0 .

The operator in (2.14) acting on [ w , w t , u ] ( g 0 ) is the same operator A b in (1.8)–(1.9), except that in (2.14) the relevant BCs (1.9) are included in the operator entries.

Thus, in conclusion, the abstract model for the nonhomogeneous PDE model (2.1a)–(2.1f) is given again by (2.2), where now

(2.15) N ( b ) g = 0 A N , s ( b ) N s ( b ) g 0 ,

(2.16) N ( b ) : continuous L 2 ( Γ s ) D ( A N , s ( b ) ) 1 4 + ε ,

recalling (2.5). The adjoint operator N ( b ) (which, in fact, does not depend on b ) is given by

(2.17) N ( b ) x 1 x 2 x 3 = x 2 Γ s , x 2 H 1 2 + 2 ε ( Ω s ) ; N ( b ) : continuous D A N , s ( b ) 1 4 + ε L 2 ( Γ s ) ,

with x 2 D ( A N , s ( b ) ) 1 4 + ε = H 1 2 + 2 ε ( Ω s ) , in the following sense. For g L 2 ( Γ s ) and { x 1 , x 2 , x 3 } , D ( A N , s ( b ) ) 1 4 + ε , , we compute as a duality pairing via (1.2a)–(1.2b) and (2.15):

(2.18) N ( b ) g , x 1 x 2 x 3 H b = 0 A N , s ( b ) N s ( b ) g 0 , x 1 x 2 x 3 H b = ( A N , s ( b ) N s ( b ) g , x 2 ) L 2 ( Ω s ) = ( g , N s ( b ) A N , s ( b ) x 2 ) L 2 ( Γ s )

(2.19) = ( g , x 2 Γ s ) L 2 ( Γ s ) = g , N ( b ) x 1 x 2 x 3 L 2 ( Γ s ) ,

where we shall establish that

(2.20) N s ( b ) A N , s ( b ) x 2 = x 2 Γ s for x 2 H 1 2 + 2 ε ( Ω s )

for the operator A N , s ( b ) as in (2.3) and N s ( b ) as in (2.4).

Proof of (2.20)

[see also [4, pp. 195–196].] Take initially f D ( A N , s ( b ) ) in (2.3), so that f ν Γ s = 0 , and g L 2 ( Γ s ) . We compute by means of the second Green’s theorem, where we recall that on Γ s , the normal ν is inward. We obtain by (2.3) and (2.4): In the aforementioned computation we have used (2.4) and f ν Γ s = 0 , thus accounting for the two vanishing terms. For the last equality, we have invoked the BC in (2.4). In conclusion from (2.23)

(2.24) ( N s ( b ) A N , s ( b ) f , g ) L 2 ( Γ s ) = ( f Γ s , g ) L 2 ( Γ s ) , for any g L 2 ( Γ s )

initially for f D ( A N , s ( b ) ) . Thus, (2.20) follows for x 2 D ( A N , s ( b ) ) . Moreover, (2.24) can be extended to all f H 1 2 + 2 ε ( Ω s ) [40], [4, Chapter 3].□

In conclusion:

Theorem 2.1

Let b = 0 or 1. Then the abstract model of the nonhomogeneous PDE problem (2.1a)–(2.1f) on the respective energy space H b in (1.2a)–(1.2b) is given by

(2.25) d d t w w t u = A b w w t u + N ( b ) g , N ( b ) = 0 A N , s ( b ) N s ( b ) 0

with A b is given by (1.8), (1.9), or alternatively by (2.14), N ( b ) is given by (2.15), and for x 2 H 1 2 + 2 ε ( Ω s ) = D ( A N , s ( b ) ) 1 4 + ε

(2.26) N ( b ) x 1 x 2 x 3 = x 2 Γ s ; N s ( b ) A N , s ( b ) x 2 = x 2 Γ s ,

in both case, b = 0 and b = 1 , see (2.17), (2.20).

Henceforth, in light of Theorem 2.1, we shall omit the qualifying parameter “ b ” for CASE 1, as the model (2.25) and (2.26) applies to both cases b = 0 and b = 1 .

3 The Luenberger’s compensator model for the heat-structure interaction model (1.1a)–(1.1f) with Neumann boundary control g at the interface Γ s , b = 0 , and b = 1

3.1 Special selection of the data

With reference to the representation (1.4) in Step 1 of the Orientation in Section 1.1, we take in our present case

A “exponentially stable”: e A t c e δ t , δ > 0 , t 0 , (3.1a) F = B ; C = B , K = B . (3.1b)

Thus, the special setting becomes, in this case,

(3.2) [ partial observation of the state y ] = C y = B y , control g = F z = B z ,

leading to the Luenberger’s dynamics

y ˙ = A y B B z , (3.3a) z ˙ = ( A 2 B B ) z + B ( B y ) , (3.3b)

and hence,

(3.4) d d t [ y z ] = ( A B B ) [ y z ] ; [ y ( t ) z ( t ) ] = e ( A B B ) t [ y 0 z 0 ] .

As noted in Section 1, Step 2, it is the PDE argument to be carried out in Section 3.1.3 in the present case of Neumann control on the interface Γ s that will determine that the infinite dimensional version of e ( A B B ) t is exponentially stable, as desired, as well as analytic.

Insight. How did we decide that F = B in (3.1b); that is, that the preassigned control g = F z is given by g = B z or F = B ? We first notice that, regardless of the choice of F , the Luenberger scheme in (1.4)–(1.6), yields that ( A K C ) is the resulting sought-after operator in characterizing the quantity [ y z ] of interest. As A is, in our case, dissipative and we surely seek to retain dissipativity, then we choose K C = B B , or K = B , C = B . Then, the (dissipative) operator ( A B B ) is the key operator to analyze for the purpose of concluding that the semigroup e ( A B B ) t is (analytic as well as) uniformly stable. Thus, at this stage, with F yet not committed and K = B , C = B committed, the z -equation becomes z ˙ = [ A B F B B ] z + B ( B y ) . It is natural to test either F = B or else F = B , whichever choice may yield the desired properties for the feedback operator A F = A B B . What is the right sign? Thus, passing from the historical scheme (1.4a)–(1.4b) to our present PDE problem (2.1a)–(2.1f), the corresponding operator N is critical in imposing the boundary conditions for the feedback operator A F , N ( b ) = A b N N in our CASE 1. In our present CASE 1, if we choose F N and so g = N z 1 z 2 z 3 = z 2 Γ s by (2.26), this then implies the boundary condition ( v 1 + v 2 ) ν Γ s = h ν Γ s + v 2 Γ s for { v 1 , v 2 , h } in D ( A F , N ) as in (3.17b) below, via (1.9b). With this B.C., the argument in (3.34) and (3.35) leads to the trace term i ω h Γ s 2 in (3.37a), and hence to the critical term i ω [ v 2 2 + h 2 + h Γ s 2 ] in (3.38), with the correct “minus” sign “−” for the argument of Theorem 3.4, in particular estimate (3.32), to succeed. Therefore, in view of the interface condition ( w + w t ) ν Γ s = u ν Γ s + g in (2.1e), the B.C. ( v 1 + v 2 ) ν Γ s = h ν Γ s + v 2 Γ s confirms that g = v 2 Γ s , or g = F w w t u = N w w t u = w t Γ s . Hence, the choice F N , as in (2.1b) is the correct one in our present CASE 1.

3.1.1 The counterpart of y ˙ = A y B B z in (2.3a) for the heat structure interaction (2.1a)–(2.1f), with F N

Accordingly, for z = [ z 1 , z 2 , z 3 ] , the Luenberger’s compensator variable, in line with (3.2), we select the Neumann control g in (2.1e) in the form

(3.5) g = N z = z 2 Γ s , z 2 D A N , s 1 4 + ε H 1 2 + 2 ε ( Ω s ) ,

where we have critically invoked the trace result (2.26) of Theorem 2.1 in both cases b = 0 and b = 1 (we are omitting the superscript “ b ”). With y = [ w , w t , u ] , the PDE version of (3.3a) corresponding to the abstract feedback problem ( b = 0 , b = 1 ):

(3.6) y ˙ = A b y N N z , or d d t w w t u = w t Δ ( w + w t ) b w Δ u + N ( z 2 Γ s )

from problem (2.25), A b as in (1.8), (1.9), alternatively in (2.14), with g as in (3.5), is

u t Δ u = 0 in ( 0 , T ] × Ω f ; (3.7a) w t t Δ w Δ w t + b w = 0 in ( 0 , T ] × Ω s ; (3.7b) u Γ f = 0 on ( 0 , T ] × Γ f ; u = w t on ( 0 , T ] × Γ s ; (3.7c) ( w + w t ) ν = u ν + z 2 on ( 0 , T ] × Γ s . (3.7d)

3.1.2 The counterpart of the dynamic compensator equation z ˙ = ( A 2 B B ) z + B ( B y ) in (2.3b) for the heat-structure interaction (2.1a)–(2.1f)

With partial observation as in (3.2) according to (2.26)

(3.8) N ( b ) y = N ( b ) w w t u = w t Γ s = partial observation of state y = w w t u ,

the compensator equation (3.3b) in z = [ z 1 , z 2 , z 3 ] becomes

(3.9) z ˙ = d d t z 1 z 2 z 3 = ( A b 2 N ( b ) N ( b ) ) z + N ( b ) ( w t Γ s )

or, via (1.8) for A b , (2.15) = (2.25) for N ( b ) , (2.26) for N ( b )

(3.10) z 1 t z 2 t z 3 t = z 2 Δ ( z 1 + z 2 ) b z 1 Δ z 3 2 0 A N , s ( b ) N s ( b ) ( z 2 Γ s ) 0 + 0 A N , s ( b ) N s ( b ) ( w t Γ s ) 0 ,

whereby z 1 t = z 2 , z 1 t t = z 2 t , and thus (3.10) is re-written as follows:

(3.11) d d t z 1 z 1 t z 3 = z 1 t Δ ( z 1 + z 1 t ) b z 1 Δ z 3 + 0 A N , s ( b ) N s ( b ) [ 2 z 2 Γ s w t Γ s ] 0 .

The PDE version of the abstract z -model (3.11) with partial observation w t Γ s in the Neumann condition at the interface Γ s (see (3.8)) is

z 3 t Δ z 3 = 0 in ( 0 , T ] × Ω f ; (3.12a) z 1 t t Δ z 1 Δ z 1 t + b z 1 = 0 in ( 0 , T ] × Ω s ; (3.12b) z 3 Γ f = 0 on ( 0 , T ] × Γ f ; z 3 = z 1 t on ( 0 , T ] × Γ s ; (3.12c) ( z 1 + z 1 t ) ν = z 3 ν + [ 2 z 1 t w t ] on ( 0 , T ] × Γ s . (3.12d)

3.1.3 The dynamics d ˙ = ( A b N ( b ) N ( b ) ) d , d ( t ) = y ( t ) z ( t ) , corresponding to (3.4): analyticity, b = 0 and b = 1

We are omitting the superscript “ b ” on N , N . The main result of the present section is as follows:

Theorem 3.1

Let b = 0 , 1 . The (feedback) operator

(3.13) A F , N ( b ) = A b N N , H b D ( A F , N ( b ) ) = { x H b : ( I A b 1 N N ) x D ( A b ) }

is the infinitesimal generator of a s.c. contraction semigroup e A F , N ( b ) t on H b , which moreover is analytic and exponentially stable on H b : there exist constants C 1 , ρ > 0 , possibly depending on b ,” such that

(3.14) e A F , N ( b ) t ( H b ) = e ( A b N N ) t ( H b ) C e ρ t , t 0 .

A more detailed description of D ( A F , N ( b ) ) is given in (3.17a)–(3.17b).

The proof of Theorem 3.1 is by PDE methods, which consist of analyzing the corresponding PDE system (3.16). We proceed through a series of steps.

Step 1. Identification of the [ y z ] -abstract equation.

Lemma 3.2

Let b = 0 , 1 . With d = [ d 1 , d 2 , d 3 ] H b ( d = difference = y z ), the abstract equation

(3.15a) d ˙ = A F , N ( b ) d = ( A b N N ) d , N ( N d ) = 0 A N , s ( b ) N s ( b ) ( d 2 Γ s ) 0

recalling (1.8) for A b , (2.25) and (2.26) for N and N

(3.15b) d 1 t d 2 t d 3 t = d 2 Δ ( d 1 + d 2 ) b d 1 Δ d 3 0 A N , s ( b ) N s ( b ) ( d 2 Γ s ) 0

so that d 2 = d 1 t , d 2 t = d 1 t t corresponds to the following PDE system, where we relabel the variable [ d 1 , d 2 = d 1 t , d 3 ] = [ w ^ , w ^ t , u ^ ] for convenience

u ^ t Δ u ^ = 0 in ( 0 , T ] × Ω f ; (3.16a) w ^ t t Δ w ^ Δ w ^ t + b w ^ = 0 in ( 0 , T ] × Ω s ; (3.16b) u ^ Γ f = 0 on ( 0 , T ] × Γ f ; u ^ = w ^ t on ( 0 , T ] × Γ s ; (3.16c) ( w ^ + w ^ t ) ν = u ^ ν + w ^ t on ( 0 , T ] × Γ s . (3.16d)

Remark 3.1

We note that the term w ^ ν and w ^ t have the same sign across the equality sign in (3.16d). This is due to the normal vector ν being inward with respect to Ω s as shown in Figure 1 and as noted in (2.20). This is consistent with the fluid-structure interaction model in [12], [39, equation 2.16.1e, p. 128], also with ν being inward with respect to Ω s . This model without the Kelvin-Voigt term is known in these references to be uniformly stable by PDE-techniques.

Description of D ( A F , N ( b ) ) . We have { v 1 , v 2 , h } D ( A F , N ( b ) ) = D ( A b N N ) if any only if

  1. (3.17a) v 1 , v 2 H 1 ( Ω s ) \ R for b = 0 ; v 1 , v 2 H 1 ( Ω s ) for b = 1 ; so that v 2 Γ s = h Γ s H 1 2 ( Γ s ) in both cases ; Δ ( v 1 + v 2 ) L 2 ( Ω s ) ;

  2. (3.17b) h H 1 ( Ω f ) , Δ h L 2 ( Ω f ) , h Γ f 0 , h Γ s = v 2 Γ s H 1 2 ( Γ s ) ; h ν Γ s = ( v 1 + v 2 ) ν Γ s v 2 Γ s H 1 2 ( Γ s ) .

The adjoint operator A F , N ( b ) = A b N N . For [ v 1 , v 2 , h ] D ( A F , N ( b ) ) (to be characterized below), we have recalled A b in (1.10a) and N v 1 v 2 h = v 2 Γ s in (2.26):

(3.18) A F , N ( b ) v 1 v 2 h = ( A b N N ) v 1 v 2 h = v 2 Δ ( v 2 v 1 ) + b v 1 Δ h N ( v 2 Γ s )

(3.19) = v 2 Δ ( v 2 v 1 ) + b v 1 Δ h 0 A N , s ( b ) N s ( b ) ( v 2 Γ s ) 0

recalling also N in (2.25) = (2.15).

Description of D ( A F , N ( b ) ) . We have { v 1 , v 2 , h } D ( A F , N ( b ) ) = D ( A b N N ) if and only if the same conditions for D ( A F , N ( b ) ) in (3.17a)–(3.17b) apply, except that new (3.17a) is replaced by Δ ( v 2 v 1 ) L 2 ( Ω s ) and (3.17b) is replaced by

(3.20) h ν Γ s = ( v 2 v 1 ) ν Γ s + v 2 Γ s H 1 2 ( Γ s ) .

The PDE corresponding to A F , N ( b ) is given by d d t w 1 w 2 u = A F , N ( b ) w 1 w 2 u or

h t Δ h = 0 in ( 0 , T ] × Ω f ; (3.21a) w 1 t t Δ w 1 Δ w 1 t b w 1 = 0 in ( 0 , T ] × Ω s ; (3.21b) h Γ f = 0 on ( 0 , T ] × Γ f ; (3.21c) h Γ s = w 1 t Γ s ; ( w 1 + w 1 t ) ν Γ s = h ν Γ s + w 1 t Γ s on ( 0 , T ] × Γ s , (3.21d)

(where w 1 t = w 2 by (3.18), top line) on H b , with { w 10 , w 11 , h 0 } H b , recalling (1.10b)–(1.10g) for A b .

Step 2. (Analysis of the PDE problem (3.16): the operator A F , N ( b ) = A b N N in (3.13))

Proposition 3.3

Let b = 0 , 1 . The operator A F , N ( b ) = A b N N in (3.13) and its H b -adjoint A F , N ( b ) = A b N N are dissipative

(3.22) Re ( A b N N ) v 1 v 2 h , v 1 v 2 h H b = v 2 Ω s 2 h Ω f 2 v 2 Γ s Γ s 2 , { v 1 , v 2 , h } D ( A F , N ( b ) ) ,

(3.23) Re ( A b N N ) v 1 v 2 h , v 1 v 2 h H b = v 2 Ω s 2 h Ω f 2 v 2 Γ s Γ s 2 , { v 1 , v 2 , h } D ( A F , N ( b ) )

in the L 2 ( ) -norms of Ω s and Ω f , and the L 2 ( Γ s ) -norm on Γ s . Hence, both A F , N ( b ) and A F , N ( b ) are maximal dissipative and thus generate s.c. contraction semigroups e A F , N ( b ) t and e A F , N ( b ) t on H b [35]. Explicitly in terms of the corresponding PDE systems, we have:

(3.24) w ^ ( t ) w ^ t ( t ) u ^ ( t ) = e A F , N ( b ) t w ^ 0 w ^ 1 u ^ 0 = e ( A b N N ) t w ^ 0 w ^ 1 u ^ 0

for the { w ^ , w ^ t , u ^ } -fluid-structure interaction model given by (3.16a)–(3.16d) on H b : with I.C. { w ^ 0 , w ^ 1 , u ^ 0 } H b . Similarly,

(3.25) w 1 ( t ) w 1 t ( t ) h ( t ) = e A F , N ( b ) t w 10 w 11 h 0 = e ( A b N N ) t w 10 w 11 h 0

for the { w 1 , w 1 t = w 2 , u } -fluid-structure interaction model given by (3.21a)–(3.21d).

Proof of (3.22)

For { v 1 , v 2 , h } D ( A F , N ( b ) ) , we compute via (3.15a)–(3.15b)

(3.26) Re ( A b N N ) v 1 v 2 h , v 1 v 2 h H b = Re v 2 Δ ( v 1 + v 2 ) b v 1 Δ h , v 1 v 2 h H b + ,

where

(3.27) = 0 A N , s ( b ) N s ( b ) ( v 2 Γ s ) 0 , v 1 v 2 h H b

(3.28) = ( v 2 Γ s , N s ( b ) A N , s ( b ) v 2 ) Γ s = v 2 Γ s 2

recalling (2.26). On the other hand, the first term on the right-hand side (RHS) of (3.26) is equal to [ v 2 2 h 2 ] by (1.12), which added to (3.28) yields (3.22). Similarly for (3.23), starting from (3.19).□

Step 3. This step provides the key PDE-energy estimate, b = 0 and b = 1 , of the entire present section. The case b = 1 is more challenging.

Remark 3.2

Given { v 1 , v 2 , h } H b , and ω R \ { 0 } , we seek to solve the equation

( i ω I A F , N ( b ) ) v 1 v 2 h = i ω I 0 I 0 Δ b I Δ 0 0 0 Δ v 1 v 2 h = v 1 v 2 h

in terms of { v 1 , v 2 , h } D ( A F , N ( b ) ) uniquely. We have

v 1 v 2 h = R ( i ω , A F , N ( b ) ) v 1 v 2 h , A F , N ( b ) R ( i ω , A F , N ( b ) ) v 1 v 2 h = 0 I 0 Δ b I Δ 0 0 0 Δ v 1 v 2 h = v 2 Δ ( v 1 + v 2 ) b v 1 Δ h .

Theorem 3.4

Let b = 0 , 1 . Let ω R and

(3.29) ( i ω I A F , N ( b ) ) v 1 v 2 h = v 1 v 2 h H b

for { v 1 , v 2 , h } D ( A F , N ( b ) ) identified in (3.17a)–(3.17b). Then:

  1. the following estimate holds true: given ε > 0 sufficiently small, there exists a constant C ε > 0 such that:

    (3.30) Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 + v 2 2 + b v 2 2 + h Γ s 2 + h 2 C ε { v 1 2 + b v 1 2 + v 2 2 + h 2 } , ω ε > 0 .

  2. (Analyticity) In view of Remark 3.2, Estimate (3.30) without the term [ h Γ s 2 + h 2 ] is equivalent to

    (3.31) A F , N ( b ) R ( i ω , A F , N ( b ) ) ( H b ) C ε , ω ε > 0 ,

    R ( i ω , A F , N ( b ) ) = ( i ω I A F , N ( b ) ) 1 , which in turn is equivalent to

    (3.32) R ( i ω , A F , N ( b ) ) ( H b ) c ε ω , ω ε > 0 .

    Thus, the s.c. contraction semigroup e A F , N ( b ) t asserted by Proposition 3.3is analytic on H b by [4, Theorem 3E.3 p. 334] and similarly for e A F , N ( b ) t on H b . Their explicit PDE version is given by (3.24) for system { w ^ ( t ) , w ^ t ( t ) , u ^ ( t ) } in (3.16a)–(3.16d) and, respectively, by (3.25) for system { w 1 ( t ) , w 1 t ( t ) , h ( t ) } in (3.21a)–(3.21d).

Proof

(i) The proof of estimate (3.30) follows closely the technical proof of [1, Section 3] for the operator A except that now the argument uses the B.C. of A F , N rather than of A , b = 0 and b = 1 . We indicate the relevant changes.

Step 1. Return to (3.29) re-written for { v 1 , v 2 , h } D ( A F , N ( b ) ) and { v 1 , v 2 , h } H b .

i ω v 1 v 2 = v 1 , (3.33a) i ω v 2 [ Δ ( v 1 + v 2 ) b v 1 ] = v 2 , (3.33b) i ω h Δ h = h , (3.33c)

Step 2. Take the L 2 ( Ω f ) -inner product of equation (3.33c) against Δ h , use Green’s First theorem, recall the B.C. h Γ f = 0 in (3.17b) for D ( A F , N ) and obtain the counterpart of [1, equation (3.10)]

(3.34) i ω Γ s h h ¯ ν d Γ s i ω h 2 Δ h 2 = ( h , Δ h ) .

Similarly, we take the L 2 ( Ω s ) -inner product of (3.33b) against [ Δ ( v 1 + v 2 ) b v 1 ] , use Green’s first theorem to evaluate Ω s v 2 Δ ( v ¯ 1 + v ¯ 2 ) d Ω s , recalling that the normal vector ν is inward with respect to Ω s , and obtain (see [1, equation (3.11)])

(3.35) i ω Γ s v 2 ( v ¯ 1 + v ¯ 2 ) ν d Γ s i ω ( v 2 , ( v 1 + v 2 ) ) i ω ( v 2 , b v 1 ) Δ ( v 1 + v 2 ) b v 1 2 = ( v 2 , [ Δ ( v 1 + v 2 ) b v 1 ] ) .

We now invoke the B.C. h Γ s = v 2 Γ s and ( v 1 + v 2 ) ν Γ s = h ν Γ s + v 2 Γ s for { v 1 , v 2 , h } in D ( A F , N ( b ) ) (see (3.17)), and we rewrite (3.35) as follows:

(3.36) i ω Γ s h h ¯ ν + h ¯ d Γ s i ω v 2 2 i ω ( v 2 , v 1 ) i ω ( v 2 , b v 1 ) Δ ( v 1 + v 2 ) b v 1 2 = ( v 2 , [ Δ ( v 1 + v 2 ) b v 1 ] ) .

Summing up (3.34) and (3.36) yields after a cancellation of the boundary term i ω Γ s h h ¯ ν d Γ s :

(3.37a) i ω h Γ s 2 i ω [ v 2 2 + h 2 ] = Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 + i ω ( v 2 , b v 1 ) + i ω ( v 2 , v 1 ) + ( v 2 , [ Δ ( v 1 + v 2 ) b v 1 ] ) + ( h , Δ h ) .

Using, via (3.33a), the identities

(3.37b) i ω ( v 2 , v 1 ) = ( v 2 , ( i ω v 1 ) ) = v 2 2 + ( v 2 , v 1 ) ,

(3.37c) i ω ( v 2 , v 1 ) = ( v 2 , i ω v 1 ) = v 2 2 + ( v 2 , v 1 ) ,

we obtain from (3.37a) the final identity

(3.38) Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 + i ω [ v 2 2 + h 2 + h Γ s 2 ] = v 2 2 + b v 2 2 + ( v 2 , v 1 ) + b ( v 2 , v 1 ) ( v 2 , [ Δ ( v 1 + v 2 ) b v 1 ] ) ( h , Δ h ) .

Step 3. We take the real part of identity (3.38), thus obtaining the new identity:

(3.39) Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 = v 2 2 + b v 2 2 + Re ( v 2 , v 1 ) + b Re ( v 2 , v 1 ) Re ( v 2 , [ Δ ( v 1 + v 2 ) b v 1 ] ) Re ( h , Δ h )

or

(3.40) ( 1 ε ) [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] ( 1 + ε ) [ v 2 2 + b v 2 2 ] + C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] .

Step 4. We now take the imaginary part of identity (3.38), thus obtaining the new identity

(3.41) ω { v 2 2 + h 2 + h Γ s 2 } = Im ( v 2 , v 1 ) + b Im ( v 2 , v 1 ) Im { ( v 2 , [ Δ ( v 1 + v 2 ) b v 1 ] ) + ( h , Δ h ) } ,

(3.42) ω { v 2 2 + h 2 + h Γ s 2 } ε 2 2 v 2 2 + b ε 3 3 v 2 2 + ε 3 [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + C ε 2 [ v 1 2 + b v 1 2 ] + C ε 3 [ v 2 2 + h 2 ]

or

(3.43) ω ε 2 2 v 2 2 + ω [ h 2 + h Γ s 2 ] b ε 3 2 v 2 2 + ε 3 [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] .

Take now

(3.44) ε 2 2 ω ε 2 2 ε 2 ω ,

so that (3.43) yields for ω as in (3.44):

(3.45) ε 2 2 v 2 2 + ε 2 [ h 2 + h Γ s 2 ] b ε 3 2 v 2 2 + ε 3 [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] ;

hence, for ω as in (3.44)

(3.46) v 2 2 + h 2 + h Γ s 2 b ε v 2 2 + 2 ε [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] .

Remark 3.3

In the case b = 0 , the proof on the space H b = 0 , hence with the first component in H 1 ( Ω s ) \ R topologized by the gradient norm, proceeds as follows. Estimate (3.46) is “too good for our purposes”: we drop the terms h 2 + h Γ s 2 and substitute the new estimate on v 2 2 on the RHS of (3.40) with b = 0 . We obtain

(3.47) ( 1 ε ) [ Δ ( v 1 + v 2 ) 2 + Δ h 2 ] ( 1 + ε ) 2 ε [ Δ ( v 1 + v 2 ) 2 + Δ h 2 ] + ( 1 + ε ) C ε [ v 1 2 + v 2 2 + h 2 ]

or

(3.48) [ ( 1 ε ) ( 1 + ε ) 2 ε ] [ Δ ( v 1 + v 2 ) 2 + Δ h 2 ] C ˜ ε [ v 1 2 + v 2 2 + h 2 ]

or

(3.49) Δ ( v 1 + v 2 ) 2 + Δ h 2 C ^ ε [ v 1 2 + v 2 2 + h 2 ] .

Estimate (3.48) coincides with estimate [1, equation (3.26)], case b = 0 .

Next, substitute estimate (3.49) into the RHS of estimate (3.46) with b = 0 , to obtain for ω as in (3.44):

(3.50) v 2 2 + h 2 + h Γ s 2 C ˜ ε [ v 1 2 + v 2 2 + h 2 ] .

Summing up estimate (3.49) with estimate (3.50) finally yields for ω as in (3.44):

(3.51) Δ ( v 1 + v 2 ) 2 + Δ h 2 + v 2 2 + h 2 + h Γ s 2 C ε [ v 1 2 + v 2 2 + h 2 ]

for all points i ω , with ω ε 2 2 as in (3.44). Then estimate (3.51) coincides with (3.30) with b = 0 as desired. In view of Remark 3.2, such estimate is equivalent to

(3.52) A F , N ( 0 ) R ( i ω , A F , N ( 0 ) ) ( H 0 ) C ε , ω ε > 0 .

The analyticity of the s.c. contraction semigroup e A F , N ( 0 ) t = e ( A 0 N N ) t on H b = 0 is established and similarly for e A F , N ( 0 ) t .

Step 5. We proceed now with the proof of analyticity in the case b = 1 on H b = 1 . This case is more challenging and requires the following additional result (in substitution of the Poincare inequality, which does not hold true for v 2 on Ω s ).

Lemma 3.5

[44, p. 260] On a sufficiently smooth bounded domain Ω in R n , let Ψ H 1 ( Ω ) . Then:

  1. Ω Ψ 2 d Ω c 1 Ω Ψ 2 d Ω + Γ ˜ Ψ 2 d Γ ,

  2. Γ Ψ 2 d Γ c 2 Ω [ Ψ 2 + Ψ 2 ] d Ω ,

  3. hence for positive constant 0 < k 1 < k 2 < ,

    k 1 Ω [ Ψ 2 + Ψ 2 ] d Ω Ω Ψ 2 d Ω + Γ ˜ Ψ 2 d Γ k 2 Ω [ Ψ 2 + Ψ 2 ] d Ω ,

where Γ ˜ is any fixed portion of the boundary Γ = Ω of Ω of positive measure.

We now return to inequality (3.46) with b = 1 , where we use h Γ s = v 2 Γ s from (3.17a) on its left-hand side (LHS) and then invoke Lemma 3.5(a) for Φ = v 2 to obtain

(3.53a) 1 c 1 v 2 2 + h 2 v 2 2 + h 2 + v 2 Γ s 2 b ε v 2 2 + OK ε ,

(3.53b) OK ε = 2 ε [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] ,

hence,

(3.54) 1 2 c 1 v 2 2 + h 2 1 c 1 b ε v 2 2 + h 2 OK ε .

by taking 1 c 1 b ε 1 2 c 1 (where b = 1 ), which yields b v 2 2 2 c 1 OK ε , which along with the RHS inequality in (3.53a) gives the desired estimate for b = 1 :

(3.55) v 2 2 + b v 2 2 + h 2 + v 2 Γ s 2 b ε v 2 2 + OK ε + b v 2 2 = ( 1 + ε ) b v 2 2 + OK ε [ ( 1 + ε ) 2 c 1 + 1 ] OK ε 2 ε [ ( 1 + ε ) 2 c 1 + 1 ] { [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] } .

The LHS in (3.55) estimates the two terms [ v 2 2 + b v 2 2 ] on the RHS in (3.40). We may now proceed as in going from (3.47) to (3.51) in Remark 3.3 for b = 0 .

Step 6. We substitute the new estimate (3.55) on the RHS of (3.40). We obtain

(3.56) ( 1 ε ) [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] ( 1 + ε ) 2 ε [ ( 1 + ε ) 2 c 1 + 1 ] [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ]

or

(3.57) { ( 1 ε ) ( 1 + ε ) 2 ε [ ( 1 + ε ) 2 c 1 + 1 ] } [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] C ˜ ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ]

or

(3.58) Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 C ^ ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] .

Estimate (3.58) for b = 1 is the counterpart of estimate (3.49) for b = 0 .

Step 7. Summing up estimates (3.58) with estimate (3.55) finally yields for ω as in (3.44):

(3.59) Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 + v 2 2 + b v 2 2 + h 2 + v 2 Γ s 2 C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ]

for all points ω , with ω ε 2 2 as in (3.44). Then estimate (3.58) coincides with estimate (3.30) as desired. The analyticity of the s.c. contraction semigroup e A F , N ( 1 ) t = e ( A 1 N N ) t on H b = 1 is established. Similarly for e A F , N ( 1 ) t , b = 1 .□

3.1.4 Exponential stability of e A F , N ( b ) t and e A F , N ( b ) t on H b , b = 0 , 1

In Proposition 3.6, we shall prove, in both cases b = 0 and b = 1 , that we have

(3.60) 0 ρ ( A F , N ( b ) ) , 0 ρ ( A F , N ( b ) ) , ( A F , N ( b ) ) 1 ( H b ) , ( A F , N ( b ) ) 1 ( H b ) ,

ρ ( ) denoting the resolvent set, so that there exists a disk S r 0 centered at the origin and of suitable radius r 0 > 0 such that S r 0 ρ ( A F , N ( b ) ) . Then, the resolvent bound (3.32) combined with ( A F , N ( b ) ) 1 ( H b ) in (3.60) allows one to conclude that the resolvent operator is uniformly bounded on the imaginary axis i R :

(3.61) R ( i ω , A F , N ( b ) ) ( H b ) const , ω R .

Hence, [45] the s.c. analytic semigroup e A F , N ( b ) t is, moreover. (uniformly) exponentially bounded: There exist constants M 1 , δ > 0 , possibly depending on b , such that

(3.62) e A F , N ( b ) t ( H b ) M e δ t , t 0 , b = 0 , b = 1 .

Similarly for the adjoint A F , N ( b ) .

Proposition 3.6

Statement (3.60) holds true. Hence, the exponential stability for e A F , N ( b ) t in (3.62) holds true. More precisely, with reference to A F , N ( b ) , we have: given { v 1 , v 2 , h } H b , the unique solution { v 1 , v 2 , h } D ( A F , N ( b ) ) of

(3.63) A F , N ( b ) v 1 v 2 h = v 2 Δ ( v 1 + v 2 ) b v 1 Δ h = v 1 v 2 h

is given explicitly by

(3.64) v 1 = ( A N , s ( b ) ) 1 [ Δ v 1 + v 2 ] + N s ( b ) ν [ A D , f 1 h + D f , s ( v 1 Γ s ) ] v 1 ν Γ s + v 1 Γ s H 1 ( Ω s ) \ R f o r b = 0 ; H 1 ( Ω s ) f o r b = 1 .

(3.65) v 2 = v 1 H 1 ( Ω s ) \ R f o r b = 0 , H 1 ( Ω s ) f o r b = 1 , h = A D , f 1 h + D f , s ( v 1 Γ s ) H 1 ( Ω f ) .

where the positive self-adjoint operator A D , f and the Dirichlet map D f , s from Γ s into Ω f were defined in (2.9) and (2.10), and are repeated as follows:

(3.66) A D , f φ = Δ φ i n Ω f ; φ D ( A D , f ) = H 2 ( Ω f ) H 0 1 ( Ω f ) ;

D f , s : H s ( Ω f ) H s + 1 2 ( Ω f ) , s R : D f , s μ = ψ Δ ψ = 0 i n Ω f ; (3.67a) ψ Γ f = 0 , ψ Γ s = μ . (3.67b)

While A N , s ( b ) and N s ( b ) , b = 0 , 1 , are defined in (2.3) and (2.4), respectively. In the operator form, we have

(3.68) v 1 v 2 h = ( A F , N ( b ) ) 1 v 1 v 2 h = ( A N , s ( b ) ) 1 [ Δ v 1 + v 2 ] + N s ( b ) ν [ A D , f 1 h + D f , s ( v 1 Γ s ) ] v 1 ν Γ s + v 1 Γ s v 1 A D , f 1 h + D f , s ( v 1 Γ s )

(3.69) = ( A N , s ( b ) ) 1 Δ + N s ( b ) ν D f , s ( Γ s ) + ( Γ s ) ν Γ s ( A N , s ( b ) ) 1 N s ( b ) ν A D , f 1 I 0 0 D f , s ( Γ s ) 0 A D , f 1 v 1 v 2 h ,

(3.70) [ v 1 , v 2 , h ] H b c [ v 1 , v 2 , h ] H b .

Proof

Identity (3.63) yields by (3.17b)

v 2 = v 1 Δ h = h L ( Ω f ) ; (3.71a) h Γ f = 0 , h Γ s = v 2 Γ s = v 1 Γ s H 1 2 ( Γ s ) , (3.71b)

and the h -problem in (3.71) yields the solution h in (3.65).

Moreover, (3.63), v 2 = v 1 in (3.65) and (3.17b) yield

Δ ( v 1 + v 1 ) b v 1 = v 2 , or Δ v 1 b v 1 = Δ v 1 + v 2 ; (3.72a) ( v 1 + v 1 ) ν Γ s = h ν Γ s + v 1 Γ s , or v 1 ν Γ s = v 1 ν + h ν + v 1 Γ s (3.72b)

from which the expression for v 1 in (3.64) followed by invoking the operator A N , s ( b ) and N s ( b ) in (2.3) and (2.4). The proof is complete.□

4 CASE 2. Heat-structure interaction with Kelvin-Voigt damping: Dirichlet control g at the interface Γ s : abstract model

We return to the homogeneous heat-structure interaction model (1.1a)–(1.1f) with Kelvin-Voigt damping. In this CASE 2, we insert a control g in the Dirichlet interface condition (1.1d). Thus, with the same geometry (Figure 1) and notation { w , w t , u } as in Section 1.3 for the uncontrolled problem, in the present CASE 2, we consider the following controlled problem:

( PDE ) u t Δ u = 0 in ( 0 , T ] × Ω f ; (4.1a) w t t Δ w Δ w t + b w = 0 in ( 0 , T ] × Ω s ; (4.1b)

( BC ) u Γ f = 0 on ( 0 , T ] × Γ f ; (4.1c) u = w t + g on ( 0 , T ] × Γ s ; (4.1d) ( w + w t ) ν = u ν on ( 0 , T ] × Γ s ; (4.1e)

(4.1f) (IC) [ w ( 0 , ) , w t ( 0 , ) , u ( 0 , ) ] = [ w 0 , w 1 , u 0 ] H b ,

this time with Dirichlet control g acting at the interface Γ s . Compare against model (2.1a)–(2.1f) of CASE 1 with Neumann control g at the interface Γ s , as in (2.1e). We shall likewise consider two cases: b = 0 and b = 1 . H b is the same finite energy space as in (1.2a)–(1.2b).

4.1 Abstract model on H b , b = 0 , 1 of the nonhomogeneous PDE model (4.1a)–(4.1f) with Dirichlet control g acting at the interface Γ s

This topic was duly treated in [39, Section 6]. Here, it was shown that the abstract version of the nonhomogeneous PDE model (4.1a)–(4.1f) is given by

(4.2) d d t w w t u = A b w w t u + D g ,

where the operator A b : H b D ( A b ) H b is of course the same as given by (1.8) and (1.9). Instead, the (boundary) control operator D is given by

(4.3) D g = 0 0 A D , f D f , s g , D : continuous L 2 ( Γ s ) D A D , f 3 4 + ε .

Here, A D , f is the negative, self-adjoint operator on L 2 ( Ω f ) defined by (2.9) = (3.66), i.e., by

(4.4) A D , f φ = Δ φ , D ( A D , f ) H 2 ( Ω f ) H 0 1 ( Ω f ) ,

while D f , s is the Dirichlet map from Γ s to Ω f defined by (2.10) = (3.66), i.e., by

φ = D f , s χ Δ φ = 0 in Ω f ; (4.5a) φ Γ f = 0 , φ Γ s = χ . (4.5b)

The following regularity holds true for D f , s [40,41], [4, Chapter 3]:

D f , s : L 2 ( Γ s ) D ( A D , f 1 4 ε ) H 1 2 2 ε ( Ω f ) , (4.6a) or A D , f 1 4 ε D f , s ( L 2 ( Γ s ) ; L 2 ( Ω f ) ) . (4.6b)

Thus, with x 3 D A D , f 3 4 + ε H 3 2 + 2 ε ( Ω f ) , we have:

(4.7) D x 1 x 2 x 3 = x 3 ν Γ s ; D : continuous D A D , f 3 4 + ε L 2 ( Γ s )

in the following sense. For g L 2 ( Γ s ) and { x 1 , x 2 , x 3 } , , D A D , f 3 4 + ε , we compute as a duality pairing via (1.2a)–(1.2b) and (4.3):

(4.8) D g , x 1 x 2 x 3 H b = 0 0 A D , f D f , s g , x 1 x 2 x 3 H b

(4.9) = ( A D , f D f , s g , x 3 ) L 2 ( Ω f ) = ( g , D f , s A D , f x 3 ) L 2 ( Γ s )

(4.10) = g , x 3 ν Γ s L 2 ( Γ s ) = g , D x 1 x 2 x 3 L 2 ( Γ s ) ,

where we have recalled (the normal ν is outward with respect to Ω f )

(4.11) D f , s A D , f x 3 = x 3 ν Γ s

from [1, p. 181], [4, Chapter 3]. Thus (4.7) is established. The proof initially takes x 3 D ( A D , f ) , and thus, x 3 Ω f = 0 , and proceeds analogously to the path (2.21)–(2.23) via Green’s second theorem to obtain (4.11) in this case. Next, we extends (4.11) to x 3 H 3 2 + 2 ε ( Ω f ) D ( A D , f 1 4 + ε ) .

As noted in Theorem 0.3 from [1], the operator A b in (1.8) and (1.9) is boundedly invertible on H b : A b 1 ( H b ) is explicitly given by [1], from which it then follows that A b 1 D ( H 1 2 ( Γ s ) ; H b ) .

4.2 The Luenberger’s compensator model for the heat-structure interaction model (4.1a)–(4.1f) with Dirichlet control g at the interface Γ s , b = 0 , 1

4.2.1 Special selection of the data

For the present heat-structure interaction problem with Dirichlet control at the interface Γ s , we shall modify the special selection made in CASE 1 of Neumann control at the interface Γ s , on the basis of the representation (1.4) in Step 1 of the orientation in Section 1.1. In fact, in the present case, we now take, in the notation of (1.4)–(1.6):

A “exponentially stable”: e A t c e δ t , δ > 0 , t 0 , (4.12a) F = B ; C = B , K = B . (4.12b)

Thus, the special setting becomes, in this case,

(4.13) [ partial observation of the state y ] = C y = B y , control g = F z = B z ,

leading to the Luenberger’s dynamics

y ˙ = A y + B B z , (4.14a) z ˙ = A z + B ( B y ) (4.14b)

(as B F K C = 0 in the present case) and hence to

(4.15) d d t [ y z ] = ( A B B ) [ y z ] ; [ y ( t ) z ( t ) ] = e ( A B B ) t [ y 0 z 0 ] .

This is the setting that will be selected in the study of the Luenberger’s theory below, as applied to heat (fluid)-structure interaction models with Dirichlet control g at the interface Γ s , as in (4.1d).

Insight. How did we decide that F = B in (4.12b); that is, that the preassigned control g = F z is given by g = B z or F = B ? We first notice that, regardless of the choice of F , the Luenberger scheme in (1.4)–(1.6), yields that ( A K C ) is the resulting sought-after operator in characterizing the quantity [ y z ] of interest. As A is, in our case, dissipative and we surely seek to retain dissipativity, then we choose K C = B B , or K = B , C = B . Then, the (dissipative) operator ( A B B ) is the key operator to analyze for the purpose of concluding that the semigroup e ( A B B ) t is (analytic as well as) uniformly stable. Thus, at this stage, with F not yet committed and K = B , C = B committed, the z -equation becomes z ˙ = [ A B F B B ] z + B ( B y ) . It is natural to test either F = B or else F = B , whichever choice may yield the desired properties for the feedback operator A F = A B B . What is the right sign? Thus, passing from the historical scheme (1.4a)–(1.4b) to our present PDE problem (4.1a)–(4.1f), the corresponding operator B D is critical in imposing the boundary conditions for the feedback operator A F , D ( b ) = A b D D in our CASE 2. But in our present CASE 2, if we choose F D and so g = D z 1 z 2 z 3 = z 3 ν Γ s by (4.7) . This then implies the boundary condition h Γ s = v 2 Γ s h ν Γ s for { v 1 , v 2 , h } in D ( A F , D ( b ) ) as in (4.28b). With this B.C., the argument in (4.46), (4.47) leads to the trace term i ω h ν Γ s 2 in (4.48), and hence to the critical term i ω v 2 2 + h 2 + h ν Γ s 2 in (4.50) with the correct “minus” sign “-” for the argument of Theorem 4.4, in particular estimate (4.52), to succeed. Therefore, in view of the interface condition u = w t + g in (4.1d), the B.C. u = w t u ν Γ s confirms that g = u ν Γ s , or g = F w w t u = D w w t u = u ν Γ s . Hence, the choice F D as in (4.12b) is the correct one in our CASE 2.

4.2.2 The counterpart of y ˙ = A y + B B z in (4.14a) for the HSI model (4.1a)–(4.1f) with F D

Accordingly, for z = [ z 1 , z 2 , z 3 ] , the Luenberger’s compensator variable, we select the Dirichlet control g in (4.1d) in the form

(4.16) g = D z = z 3 ν Γ s , z 3 D A D , f 3 4 + ε H 3 2 + 2 ε ( Ω f ) ,

i.e., with F D , where we have critically invoked the trace result (4.7). With y = [ w , w t , u ] , the PDE version of (3.3a) corresponding to the abstract feedback problem

(4.17) y ˙ = A b y + D D z , or d d t w w t u = w t Δ ( w + w t ) b w Δ u + D z 3 ν Γ s

from problem (4.2) with operator A b as in (1.8) and with g as in (4.16), is

u t Δ u = 0 in ( 0 , T ] × Ω f ; (4.18a) w t t Δ w Δ w t + b w = 0 in ( 0 , T ] × Ω s ; (4.18b) u Γ f = 0 on ( 0 , T ] × Γ f ; u = w t z 3 ν on ( 0 , T ] × Γ s ; (4.18c) ( w + w t ) ν = u ν on ( 0 , T ] × Γ s . (4.18d)

4.2.3 The counterpart of the dynamic compensator equation z ˙ = A z + B ( B y ) in the special setting (4.14b) for the HSI model (4.1a)–(4.1f)

With partial observation as in (4.13), C D , according to (4.7)

(4.19) C y D y = D w w t u = u ν Γ s = partial observation of state y = w w t u ,

the compensator equation (4.14b) in z = [ z 1 , z 2 , z 3 ] becomes in our present HSI case

(4.20) z ˙ = d d t z 1 z 2 z 3 = A b z + D u ν Γ s

or via (1.8) on A b , (4.3), (4.7) on D , D

(4.21) z 1 t z 2 t z 3 t = z 2 Δ ( z 1 + z 2 ) b z 1 Δ z 3 + 0 0 A D , f D f , s u ν Γ s ,

whereby z 1 t = z 2 , z 1 t t = z 2 t . The PDE version of the abstract z -model (4.21) with partial observation u ν Γ s in the Dirichlet condition at the interface Γ s as in (4.19) is given as follows:

z 3 t Δ z 3 = 0 in ( 0 , T ] × Ω f ; (4.22a) z 1 t t Δ z 1 Δ z 1 t + b z 1 = 0 in ( 0 , T ] × Ω s ; (4.22b) z 3 Γ f = 0 on ( 0 , T ] × Γ f ; z 3 = z 1 t Γ s u ν Γ s on ( 0 , T ] × Γ s ; (4.22c) ( z 1 + z 1 t ) ν = z 3 ν on ( 0 , T ] × Γ s . (4.22d)

4.2.4 The dynamics d ˙ = ( A b D D ) d , d ( t ) = y ( t ) z ( t ) corresponding to (4.15): analyticity and exponential decay, b = 0 , 1

The main result of the present section is as follows:

Theorem 4.1

Let b = 0 , 1 . The (feedback) operator

(4.23) A F , D ( b ) = A b D D , H b D ( A F , D ) = { x H b : ( I A b 1 D D ) x D ( A b ) }

in the infinitesimal generator of a s.c. contraction semigroup e A F , D ( b ) t on H b , which moreover is analytic and exponential stable on H b : there exist constants C 1 , ρ > 0 possibly depending on b such that

(4.24) e A F , D ( b ) t ( H b ) = e ( A b D D ) t ( H b ) C e ρ t , t 0 .

A more detailed description of D ( A F , D ( b ) ) is given in (4.28a)–(4.28b). The proof of Theorem 4.1 is by PDE methods, which consists of analyzing the corresponding PDE system (4.27).

Step 1.

Lemma 4.2

Let b = 0 , 1 . With d = [ d 1 , d 2 , d 3 ] H b , the abstract equation

(4.25) d ˙ = A F , D ( b ) d = ( A b D D ) d

or via (1.8) on A b , (4.3) and (4.7) on D , D

(4.26) d 1 t d 2 t d 3 t = d 2 Δ ( d 1 + d 2 ) b d 1 Δ d 3 0 0 A D , f D f , s d 3 ν Γ s

so that d 2 = d 1 t , d 2 t = d 1 t t , corresponds to the following PDE system, where we relabel the variable [ d 1 , d 2 = d 1 t , d 3 ] = [ w ^ , w ^ t , u ^ ] for convenience

u ^ t Δ u ^ = 0 in ( 0 , T ] × Ω f ; (4.27a) w ^ t t Δ w ^ Δ w ^ t + b w ^ = 0 in ( 0 , T ] × Ω s ; (4.27b) u ^ Γ f = 0 on ( 0 , T ] × Γ f ; u ^ = w ^ t u ^ ν on ( 0 , T ] × Γ s ; (4.27c) ( w ^ + w ^ t ) ν = u ^ ν on ( 0 , T ] × Γ s . (4.27d)

Description of D ( A F , D ( b ) ) . We have { v 1 , v 2 , h } D ( A F , D ( b ) ) = D ( A b D D ) if any only if

  1. (4.28a) v 1 , v 2 H 1 ( Ω s ) \ R for b = 0 , H 1 ( Ω s ) for b = 1 , so that v 2 Γ s = h Γ s H 1 2 ( Γ s ) in both cases ; Δ ( v 1 + v 2 ) L 2 ( Ω s ) ;

  2. (4.28b) h H 1 ( Ω f ) , Δ h L 2 ( Ω f ) , h Γ f 0 , h Γ s = v 2 Γ s h ν Γ s H 1 2 ( Γ s ) ; h ν Γ s = ( v 1 + v 2 ) ν Γ s H 1 2 ( Γ s ) .

The adjoint operator A F , D ( b ) = A b D D . For { v 1 , v 2 , h } D ( A F , D ( b ) ) (to be characterized later), we have after recalling A b in (1.10a) and D v 1 v 2 h = h ν Γ s from (4.7)

(4.29) A F , D ( b ) v 1 v 2 h = ( A b D D ) v 1 v 2 h = v 2 Δ ( v 2 v 1 ) + b v 1 Δ h D h ν Γ s

(4.30) = v 2 Δ ( v 2 v 1 ) + b v 1 Δ h 0 0 A D , f D f , s h ν Γ s

recalling D in (4.3).

Description of D ( A F , D ( b ) ) . We have { v 1 , v 2 , h } D ( A F , D ( b ) ) = D ( A b D D ) if and only if the same conditions for D ( A F , D ( b ) ) in (4.28a)–(4.28b) apply, except that now (4.28a) is replaced by Δ ( v 2 v 1 ) L 2 ( Ω s ) and (4.28b) is replaced by

(4.31) v 2 v 1 ν Γ s = h ν Γ s H 1 2 ( Γ s ) .

The PDE version corresponding to the adjoint operator A F , D ( b ) is given by d d t w 1 w 2 u = A F , D ( b ) w 1 w 2 u or (counterpart of (3.21a)–(3.21d))

h t Δ h = 0 in ( 0 , T ] × Ω f ; (4.32a) w 1 t t Δ w 1 Δ w 1 t + b w 1 = 0 in ( 0 , T ] × Ω s ; (4.32b) h Γ f = 0 on ( 0 , T ] × Γ f ; (4.32c) h Γ s = w 1 t Γ s h ν Γ s ; ( w 1 + w 1 t ) ν Γ s = h ν Γ s on ( 0 , T ] × Γ s , (4.32d)

(where w 1 t = w 2 by (4.29), top line) with I.C. in H b .

Step 2. (Analysis of the PDE problem (4.27): the operator A F , D ( b ) = A b D D in (4.23))

Proposition 4.3

The operator A F , D ( b ) = A b D D in (4.23) and its H b -adjoint A F , D ( b ) = A b D D are dissipative

(4.33) Re ( A b D D ) v 1 v 2 h , v 1 v 2 h H b = v 2 Ω s 2 h Ω f 2 h ν Γ s Γ s 2 , { v 1 , v 2 , h } D ( A F , D ( b ) ) ,

(4.34) Re ( A b D D ) v 1 v 2 h , v 1 v 2 h H b = v 2 Ω s 2 h Ω f 2 h ν Γ s Γ s 2 , { v 1 , v 2 , h } D ( A F , D ( b ) )

in the L 2 ( ) norms of Ω s and Ω f , and the L 2 ( Γ s ) norm on Γ s . Hence, both A F , D ( b ) and A F , D ( b ) are maximal dissipative and thus generate s.c. contraction semigroups e A F , D ( b ) t and e A F , D ( b ) t , respectively, on H b . Explicitly in terms of the corresponding PDE systems (4.27) we have:

(4.35) w ^ ( t ) w ^ t ( t ) u ^ ( t ) = e A F , D ( b ) t w ^ 0 w ^ 1 u ^ 0 = e ( A b D D ) t w ^ 0 w ^ 1 u ^ 0

for the { w ^ , w ^ t , u ^ } -fluid-structure interaction model given by (4.27a)–(4.27d) on H b : with I.C. { w ^ 0 , w ^ 1 , u ^ 0 } H b , similarly, for the adjoint A F , D ( b ) , corresponding to model (4.32a)–(4.32d).

Proof of (4.33)

For { v 1 , v 2 , h } D ( A F , D ) in (4.28a) and (4.28b), we compute via (1.8), (4.26):

(4.36) Re ( A b D D ) v 1 v 2 h , v 1 v 2 h H b = Re v 2 Δ ( v 1 + v 2 ) b v 1 Δ h , v 1 v 2 h H b 1 D ,

where recalling (4.3) for D and (4.7) for D

(4.37) 1 D = 0 0 A D , f D f , s h ν Γ s , v 1 v 2 h H b = h ν Γ s , D f , s A D , f h L 2 ( Γ s )

(4.38) = h ν Γ s L 2 ( Γ s ) 2 ,

recalling D f , s A D , f h = h ν Γ s from (4.11). On the other hand, recalling (1.12)

(4.39) Re A b v 1 v 2 h , v 1 v 2 h H b = v 2 2 h 2 .

Thus, (4.38) and (4.39), used in (4.36), yield (4.33), as desired. Similarly for (4.34) starting from (1.10a) for A b .□

4.2.5 Analyticity of e A F , D ( b ) t and e A F , D ( b ) t on H b , b = 0 , 1

Remark 4.1

Given { v 1 , v 2 , h } H b , and ω R \ { 0 } , we seek to solve the equation

( i ω I A F , D ( b ) ) v 1 v 2 h = i ω I 0 I 0 Δ b I Δ 0 0 0 Δ v 1 v 2 h = v 1 v 2 h

in terms of { v 1 , v 2 , h } D ( A F , D ( b ) ) uniquely. We have (counterpart of Remark 3.2)

v 1 v 2 h = R ( i ω , A F , D ( b ) ) v 1 v 2 h , A F , D ( b ) R ( i ω , A F , D ( b ) ) v 1 v 2 h = 0 I 0 Δ b I Δ 0 0 0 Δ v 1 v 2 h = v 2 Δ ( v 1 + v 2 ) b v 1 Δ h .

Theorem 4.4

Let b = 0 , 1 . Let ω R and

(4.40) ( i ω I A F , D ( b ) ) v 1 v 2 h = v 1 v 2 h H b

for { v 1 , v 2 , h } D ( A F , D ( b ) ) identified in (4.28a)–(4.28b). Then:

  1. the following estimate holds true: there exists a constant C ε > 0 such that:

    (4.41) b v 2 2 + v 2 2 + Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 + h 2 + h ν Γ s Γ s 2 C ε { b v 1 2 + v 1 2 + v 2 2 + h 2 } , ω ε > 0 ; ;

  2. In view of Remark 4.1, estimate (4.41) is equivalent to

    (4.42) A F , D ( b ) R ( i ω , A F , D ( b ) ) ( H b ) C ε , ω ε > 0 ,

    R ( i ω , A F , D ( b ) ) = ( i ω I A F , D ( b ) ) 1 , in turn equivalent to estimate

    (4.43) R ( i ω , A F , D ( b ) ) ( H b ) c ε ω , ω ε > 0 .

    Thus, the s.c. contraction semigroup e A F , D ( b ) t asserted by Proposition 4.3is analytic on H b by [4, Theorem 3E.3 p. 334], similarly for e A F , D ( b ) t on H b . The explicit PDE version of (4.35) is system { w ^ ( t ) , w ^ t ( t ) , u ^ ( t ) } in (4.27a)–(4.27d).

Proof

(i) The proof of estimate (4.41) follows closely the technical proof of estimate (3.30) for the operator A F , N ( b ) except that now the argument uses the B.C. of A F , D ( b ) rather than of A F , N ( b ) . This, in particular, requires a different argument to handle the more challenging case b = 1 , with full H 1 ( Ω ) -norm on the first component space. We indicate the relevant changes.

Step 1. Return to (4.40), re-written for { v 1 , v 2 , h } D ( A F , D ( b ) ) and { v 1 , v 2 , h } H b :

i ω v 1 v 2 = v 1 , (4.44a) i ω v 2 [ Δ ( v 1 + v 2 ) b v 1 ] = v 2 , (4.44b) i ω h Δ h = h . (4.44c)

Step 2. Take the L 2 ( Ω f ) -inner product of equation (4.44c) against Δ h , use Green’s First theorem, recall the B.C. h Γ f = 0 in (4.28b) for D ( A F , D ( b ) ) and obtain the counterpart of [1, equation (3.10)] or (3.34)

(4.45) i ω Γ s h h ¯ ν d Γ s i ω h 2 Δ h 2 = ( h , Δ h ) .

Similarly, we take the L 2 ( Ω s ) -inner product of (4.44b) against [ Δ ( v 1 + v 2 ) b v 1 ] , use Green’s First Theorem to evaluate Ω s v 2 Δ ( v ¯ 1 + v ¯ 2 ) d Ω s , recalling that the normal vector ν is inward with respect to Ω s , and obtain as in (3.35)

(4.46) i ω Γ s v 2 ( v ¯ 1 + v ¯ 2 ) ν d Γ s i ω ( v 2 , ( v 1 + v 2 ) ) i ω ( v 2 , b v 1 ) Δ ( v 1 + v 2 ) b v 1 2 = ( v 2 , [ Δ ( v 1 + v 2 ) b v 1 ] ) .

We now invoke the B.C. h Γ s = v 2 Γ s h ν Γ s and ( v 1 + v 2 ) ν Γ s = h ν Γ s for { v 1 , v 2 , h } in D ( A F , D ( b ) ) (see (4.28b)), we rewrite (4.46) as follows:

(4.47) i ω Γ s h + h ν h ¯ ν d Γ s i ω v 2 2 i ω ( v 2 , v 1 ) i ω ( v 2 , b v 1 ) Δ ( v 1 + v 2 ) b v 1 2 = ( v 2 , [ Δ ( v 1 + v 2 ) b v 1 ] ) .

Summing up (4.45) and (4.47) yields after a cancellation of the boundary terms i ω Γ s h h ¯ ν d Γ s :

(4.48) i ω h ν Γ s 2 i ω [ v 2 2 + h 2 ] = Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 + i ω ( v 2 , b v 1 ) + i ω ( v 2 , v 1 ) + ( v 2 , [ Δ ( v 1 + v 2 ) b v 1 ] ) + ( h , Δ h ) .

By using via (4.44a), the identities

(4.49a) i ω ( v 2 , v 1 ) = ( v 2 , ( i ω v 1 ) ) = v 2 2 + ( v 2 , v 1 ) ,

(4.49b) i ω ( v 2 , v 1 ) = ( v 2 , i ω v 1 ) = v 2 2 + ( v 2 , v 1 ) ,

and we obtain the final identity

(4.50) Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 + i ω v 2 2 + h 2 + h ν Γ s 2 = v 2 2 + b v 2 2 + ( v 2 , v 1 ) + b ( v 2 , v 1 ) ( v 2 , [ Δ ( v 1 + v 2 ) b v 1 ] ) ( h , Δ h ) .

Step 3. We take the real part of identity (4.50), thus obtaining the new identity:

(4.51) Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 = v 2 2 + b v 2 2 + Re ( v 2 , v 1 ) + b Re ( v 2 , v 1 ) Re ( v 2 , [ Δ ( v 1 + v 2 ) b v 1 ] ) Re ( h , Δ h )

or

(4.52) ( 1 ε ) [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] ( 1 + ε ) [ v 2 2 + b v 2 2 ] + C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] .

Step 4. We now take the imaginary part of identity (4.50), thus obtaining the new identity

(4.53) ω v 2 2 + h 2 + h ν Γ s 2 = Im { ( v 2 , v 1 ) } + b Im { ( v 2 , v 1 ) } Im { ( v 2 , Δ ( v 1 + v 2 ) b v 1 ) + ( h , Δ h ) } ,

(4.54) ω v 2 2 + h 2 + h ν Γ s 2 ε 2 2 v 2 2 + b ε 3 2 v 2 2 + ε 3 [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + C ε 2 [ v 1 2 + b v 1 2 ] + C ε 3 [ v 2 2 + h 2 ]

or

(4.55) ω ε 2 2 v 2 2 + ω h 2 + h ν Γ s 2 b ε 3 2 v 2 2 + ε 3 [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] .

Take now

(4.56) ε 2 2 ω ε 2 2 ε 2 ω

so that (4.55) yields for ω as in (4.56):

(4.57) ε 2 2 v 2 2 + ε 2 h 2 + h ν Γ s 2 b ε 3 2 v 2 2 + ε 3 [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] ;

hence for ω as in (4.56)

(4.58) v 2 2 + h 2 + h ν Γ s 2 b ε v 2 2 + 2 ε [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] .

Remark 4.2

In the case b = 0 , the proof on the space H b = 0 , hence via (1.2a) with the first component in H 1 ( Ω s ) \ R topologized by the gradient norm, proceeds as follows. Estimate (4.58) is “too good for our purposes at this stage”: on its LHS, we drop the terms h 2 + h ν Γ s 2 and substitute the new estimate on the remaining v 2 2 on the RHS of (4.52) with b = 0 . We obtain

( 1 ε ) [ Δ ( v 1 + v 2 ) 2 + Δ h 2 ] ( 1 + ε ) 2 ε [ Δ ( v 1 + v 2 ) 2 + Δ h 2 ] + ( 1 + ε ) C ε [ v 1 2 + v 2 2 + h 2 ] ( 4.5 9 0 )

or

[ ( 1 ε ) ( 1 + ε ) 2 ε ] [ Δ ( v 1 + v 2 ) 2 + Δ h 2 ] C ˜ ε [ v 1 2 + v 2 2 + h 2 ] ( 4.6 0 0 )

or

Δ ( v 1 + v 2 ) 2 + Δ h 2 C ^ ε [ v 1 2 + v 2 2 + h 2 ] . ( 4.6 1 0 )

Estimate ( 4.6 1 0 ) coincides with estimate [1, equation (3.26)], case b = 0 .

Substitute estimate ( 4.6 1 0 ) into the RHS of estimate (4.58) with b = 0 to obtain for ω as in (4.56):

v 2 2 + h 2 + h ν Γ s 2 C ˜ ε [ v 1 2 + v 2 2 + h 2 ] . ( 4.6 2 0 )

Summing up estimate ( 4.6 1 0 ) with estimate ( 4.6 2 0 ) finally yields for ω as in (4.56):

Δ ( v 1 + v 2 ) 2 + Δ h 2 + v 2 2 + h 2 + h ν Γ s 2 C ε [ v 1 2 + v 2 2 + h 2 ] ( 4.6 3 0 )

for all points i ω , with ω ε 2 2 as in (4.56). Then estimate ( 4.6 3 0 ) coincides with estimate (4.41) with b = 0 as desired.

Step 5. We proceed now with the proof of the case b = 1 on the space H b = 1 via (1.2b) with H 1 ( Ω s ) first component. As in the Neumann control on Γ s of CASE 1, this case is more challenging and requires the same Lemma 3.5 of CASE 1. However, its use will be tuned to the present case of Dirichlet control on Γ s , with trace h ν Γ s characterized by (4.28b) in D ( A F , D ) :

(4.64) h ν Γ s = v 2 Γ s h Γ s .

Thus, with b = 1 , we return to estimate (4.58) where on its LHS we use (4.64):

v 2 2 + h 2 + v 2 Γ s h Γ s 2 b ε v 2 2 + OK ε , ( 4.5 9 1 )

where

OK ε = 2 ε [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] . ( 4.6 0 1 )

For 0 ε 1 1 to be chosen below, we return to ( 4.5 9 1 ) and readily obtain, also after adding ε 1 h Γ s 2 to both sides:

v 2 2 + h 2 + ε 1 v 2 Γ s h Γ s 2 + ε 1 h Γ s 2 b ε v 2 2 + ε 1 h Γ s 2 + OK ε . ( 4.6 1 1 )

Next from

v 2 Γ s 2 = ( v 2 Γ s h Γ s ) + h Γ s 2 [ v 2 Γ s h Γ s + h Γ s ] 2 2 v 2 Γ s h Γ s 2 + 2 h Γ s 2 , ( 4.6 2 1 )

and hence,

ε 1 2 v 2 Γ s 2 ε 1 v 2 Γ s h Γ s 2 + ε 1 h Γ s 2 . ( 4.6 3 1 )

Using ( 4.6 3 1 ) on the LHS of ( 4.6 1 1 ) yields ( ε 1 1 )

ε 1 2 [ v 2 Γ s 2 + v 2 2 ] + h 2 b ε v 2 2 + ε 1 h Γ s 2 + OK ε . ( 4.6 4 1 ) .

Next we invoke Lemma 3.5(a) on the term [ ] on the LHS of ( 3.6 4 1 ), to obtain

ε 1 2 c 1 b ε v 2 2 + h 2 ε 1 h Γ s 2 + OK ε ( 4.6 5 1 )

ε 1 C p h 2 + OK ε , ( 4.6 6 1 )

where going from ( 4.6 5 1 ) to ( 4.6 5 1 ) we have invoked the Poincare inequality

h Γ s 2 C p h 2 , ( 4.6 7 1 )

which is legal since h Γ f = 0 by (4.28) on D ( A F , D ( b ) ) . Then ( 4.6 6 1 ) is rewritten as follows:

ε 1 2 c 1 b ε v 2 2 + ( 1 ε 1 C p ) h 2 OK ε . ( 4.6 8 1 )

Now we select 1 ε 1 > 0 ( b = 1 ) so that

ε < ε 1 2 c 1 b ε or 2 c 1 ( ε + ε ) < ε 1 1 , ( 4.6 9 1 )

and

1 2 < 1 ε 1 C p or ε 1 < 1 2 C p 1 ( 4.7 0 1 )

(we can always take C p 1 2 ). Using ( 4.6 9 1 ) and ( 4.7 0 1 ) in ( 4.6 8 1 ) yields

ε v 2 2 + 1 2 h 2 2 ε [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] ( 4.7 1 1 )

recalling OK ε in ( 4.6 0 1 ), for ω ε 2 as in (4.56).

We finally obtain the desired estimate also for b = 1 to include also v 2 2 :

v 2 2 + h 2 4 ε [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] ( 4.7 2 1 )

for ω ε 2 as in (4.56). Substituting the estimate for v 2 2 from (4.2.5) into the RHS of (4.58) with b ε b yields

v 2 2 + h 2 + h ν Γ s 2 4 b ε ε [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] . ( 4.7 3 1 )

Now, add the estimate for v 2 2 in ( 4.7 2 1 ) to ( 4.7 3 1 ) and obtain

b v 2 2 + v 2 2 + h 2 + h ν Γ s 2 4 ε ( 1 + b ε ) [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] . ( 4.7 3 1 )

Estimate ( 4.7 3 1 ) is the counterpart of estimate (4.58) for b = 0 .

Step 6. The rest of the proof for b = 1 now proceeds as in the case b = 0 . In ( 4.7 3 1 ) we drop the terms h 2 + h ν Γ s 2 and substitute the resulting estimate for [ b v 2 2 + v 2 2 ] into the RHS of (4.52) and obtain

( 1 ε ) [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] ( 1 + ε ) 4 ε ( 1 + b ε ) [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + ( 1 + ε ) C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] ( 4.7 4 1 )

or, since [ ( 1 ε ) ( 1 + ε ) 4 ε ( 1 + b ε ) ] k > 0 ,

Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 C ε ^ [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] , ( 4.7 5 1 )

which is a counterpart of ( 4.6 3 0 ) for b = 0 . By substituting ( 4.7 5 1 ) into the RHS of ( 4.7 3 1 ), we finally obtain

b v 2 2 + v 2 2 + h 2 + h ν Γ s 2 C ε ˜ [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] . ( 4.7 6 1 )

Finally, summing up ( 4.7 5 1 ) and ( 4.7 6 1 ) yields

b v 2 2 + v 2 2 + Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 + h 2 + h ν Γ s 2 C ε ¯ [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] , ( 4.7 7 1 )

for all ω > ε 2 , as in (4.56).

Estimate ( 4.7 7 1 ) is the sought-after estimate (4.41) for b = 0 , b = 1 .□

4.2.6 Exponential stability of e A F , D ( b ) t and e A F , D ( b ) t on H b , b = 0 , 1

In Proposition 4.5, we shall prove that, in both cases b = 0 and b = 1 , we have

(4.78) 0 ρ ( A F , D ( b ) ) , 0 ρ ( A F , D ( b ) ) , A F , D ( b ) 1 ( H b ) , A F , D ( b ) 1 ( H b ) ,

so that there exists a disk S r 0 centered at the origin and of suitable radius r 0 > 0 such that S r 0 ρ ( A F , D ( b ) ) . Then, the resolvent bound (4.43) combined with A F , D ( b ) 1 ( H b ) in (4.78) allows one to conclude that the resolvent is uniformly bounded on the imaginary axis i R :

(4.79) R ( i ω , A F , D ( b ) ) ( H b ) const , ω R .

Hence, [45] the s.c. analytic semigroup e A F , D ( b ) t is, moreover, (uniformly) exponentially stable: There exist constants M 1 , δ > 0 , possibly depending on “ b ” such that

(4.80) e A F , D ( b ) t ( H b ) M e δ t , t 0 .

It was similar for the adjoint A F , D ( b ) .

Proposition 4.5

Statement (4.78) holds true. Hence, the exponential stability for e A F , D ( b ) t in (4.80) holds true. More precisely, with reference to A F , D ( b ) , we have: given { v 1 , v 2 , h } H b , the unique solution { v 1 , v 2 , h } D ( A F , D ( b ) ) of

(4.81) A F , D ( b ) v 1 v 2 h = v 2 Δ ( v 1 + v 2 ) b v 1 Δ h = v 1 v 2 h

is given explicitly by

(4.82) v 1 = ( A N , s ( b ) ) 1 [ v 2 Δ v 1 ] + N s ( b ) ν [ A R , f 1 h + R f , s ( v 1 Γ s ) v 1 Γ s ] H 1 ( Ω s ) \ R ,

(4.83) v 2 = v 1 H 1 ( Ω s ) \ R f o r b = 0 , H 1 ( Ω s ) f o r b = 1 , h = A R , f 1 h + R f , s ( v 1 Γ s ) H 3 2 ( Ω f ) .

In the operator form, we have

(4.84a) v 1 v 2 h = A F , D ( b ) 1 v 1 v 2 h = ( A N , s ( b ) ) 1 [ Δ v 1 v 2 ] + N s ( b ) ν [ A R , f 1 h + R f , s ( v 1 Γ s ) v 1 Γ s ] v 1 A D , f 1 h + R f , s ( v 1 Γ s ) ,

(4.84b) = ( A N , s ( b ) ) 1 Δ + N s ( b ) ν [ R f , s ( Γ s ) ] A N , s 1 N s ( b ) ν A D , f 1 I 0 0 R f , s ( Γ s ) 0 A D , f 1 v 1 v 2 h

(4.84c) [ v 1 , v 2 , h ] H b c [ v 1 , v 2 , h ] H b ,

where the operators A N , s ( b ) , N s ( b ) are defined in (2.3), (2.4). Moreover, A R , f is the Robin Laplacian on Ω f :

(4.85) A R , f φ = Δ φ i n Ω f ; φ D ( A R , f ) = ϕ H 2 ( Ω f ) : ϕ Γ f = 0 , ϕ ν + ϕ Γ s = 0 ,

and R f , s is the Robin map from Γ s to Ω f :

R f , s : H s ( Ω f ) H s + 3 2 ( Ω f ) , s R : R f , s μ = ψ Δ ψ = 0 in Ω f ; (4.86a) ψ Γ f = 0 , ψ ν + ψ Γ s = μ . (4.86b)

Proof

Identity (4.81) and the characterization of D ( A F , D ( b ) ) in (4.28a)–(4.28b) yield

v 2 = v 1 H 1 ( Ω s ) \ R for b = 0 , H 1 ( Ω s ) for b = 1 , Δ h = h L 2 ( Ω f ) ; (4.87a) h Γ f = 0 , h ν Γ s + h Γ s = v 1 Γ s , (4.87b)

and the h -problem in (4.87) yields the solution h in (4.83). Moreover, (4.81), v 2 = v 1 in (4.87a) and (4.87b) yield

Δ ( v 1 + v 2 ) b v 1 = v 2 , or Δ v 1 b v 1 = Δ v 1 + v 2 , (4.88a) ( v 1 + v 2 ) ν Γ s = h ν Γ s , or v 1 ν Γ s = v 1 ν Γ s + h ν Γ s H 1 2 ( Γ s ) . (4.88b)

Then, the solution of problem (4.88) is given by (4.82) via (4.83), where the operators A N , s ( b ) and N s ( b ) were defined in (2.3), (2.4), and recalled as follows:

(4.89) A N , s ( b ) φ = ( Δ b I ) φ , D ( A N , s ( b ) ) = φ H 2 ( Ω s ) \ R for b = 0 , H 2 ( Ω s ) for b = 1 , : φ ν Γ s = 0 ,

where N s ( b ) is the Neumann map

(4.90) ψ = N s ( b ) μ ( Δ b I ) ψ = 0 in Ω s ; ψ ν Γ s = μ .

Proposition 4.5 regarding (4.79) and (4.80) is proved.□

5 CASE 3: Heat-structure interaction with Kelvin-Voigt damping: Dirichlet control g at the external boundary Γ f

We return to the homogeneous heat-structure interaction model (1.1a)–(1.1f) with Kelvin-Voigt damping. In the present CASE 3, we apply a Dirichlet control g on the external boundary Γ f . Thus, with the same geometry (Figure 1) and notation { w , w t , u } as in Section 1.3 (see (1.1a)–(1.1f)), in the present CASE 3, we consider the problem

( PDE ) u t Δ u = 0 in ( 0 , T ] × Ω f ; (5.1a) w t t Δ w Δ w t + b w = 0 in ( 0 , T ] × Ω s ; (5.1b)

( BC ) u Γ f = g on ( 0 , T ] × Γ f ; (5.1c) u = w t on ( 0 , T ] × Γ s ; (5.1d) ( w + w t ) ν = u ν on ( 0 , T ] × Γ s ; (5.1e)

(5.1f) (IC) [ w ( 0 , ) , w t ( 0 , ) , u ( 0 , ) ] = [ w 0 , w 1 , u 0 ] H ,

with Dirichlet boundary control g acting at the external boundary Γ f . Compare against model (2.1a)–(2.1f) of CASE 1 with Neumann control g at the interface Γ s , as in (2.1e), and against model (4.1a)–(4.1f) of CASE 2 with Dirichlet control g at the interface Γ s , as in (4.1d). We shall likewise consider two cases: b = 0 and b = 1 . H b is the same finite energy space defined throughout in (1.2a)–(1.2b).

5.1 Abstract model on H b , b = 0 , 1 , of the nonhomogeneous PDE model (5.1a)–(5.1f) with Dirichlet control g acting at the external boundary Γ f

This is the counterpart of the treatment in [39], where an interior Neumann or Dirichlet control acts at the interface Γ s . To this end, we define two boundary interior maps, with interior Ω f : the map D f , s (introduced in (2.10) = (3.67) = (4.5)) acting from Γ s , and the map D f , f acting from Γ f :

(5.2a) φ = D f , s χ Δ φ = 0 in Ω f ; φ Γ f = 0 , φ Γ s = χ , ψ = D f , f μ Δ ψ = 0 in Ω f ; ψ Γ f = μ , ψ Γ s = 0 ,

(5.2b) D f , s , D f , f : H r ( Ω f ) H r + 1 2 ( Ω f ) , for any r R ,

(5.2c) D f , s , D f , f : L 2 ( Ω f ) H 1 2 ( Ω f ) H 1 2 2 ε ( Ω f ) D A D , f 1 4 ε ,

[40,41], [4, p. 181], [39], where A D , f is the negative, self-adjoint operator on L 2 ( Ω f ) defined in (2.9) = (3.66) = (4.4) and repeated here

(5.3) A D , f φ = Δ φ , D ( A D , f ) H 2 ( Ω f ) H 0 1 ( Ω f ) .

Similarly, we recall from (2.4) the Neumann map N s ( b ) on Ω s :

(5.4) ψ = N s ( b ) μ ( Δ b I ) ψ = 0 in Ω s ; ψ ν Γ s = μ ,

with regularity given by (2.5), as well as the Neumann-Laplacian in Ω s ; A N , s ( b ) , the negative self-adjoint operator on L 2 ( Ω s ) , introduced in (2.3) and repeated here

(5.5) A N , s ( b ) φ = ( Δ b I ) φ = 0 , D ( A N , s ( b ) ) = φ H 2 ( Ω s ) \ R for b = 0 , H 2 ( Ω s ) for b = 1 , : φ ν Γ s = 0 .

To obtain the abstract model of problem (5.1), we proceed as usual [4,39]. We re-write the u -problem in (5.1) as follows:

u t = Δ ( u D f , f g D f , s ( w t Γ s ) ) in ( 0 , T ] × Ω f ; (5.6a) [ u D f , f g D f , s w t ] Ω f = 0 = g g 0 on Γ f ; w t 0 w t on Γ s ; in ( 0 , T ] × Ω f . (5.6b)

u t = A D , f u D f , f g D f , s ( w t Γ s ) L 2 ( Ω f ) ; (5.7a) u t = A ˜ D , f u + A ˜ D , f D f , f g + A ˜ D , f D f , s ( w t Γ s ) [ D ( A D , f ) ] , (5.7b)

where A ˜ D , f is the isomorphic extension of the operator A D , f in (5.3): L 2 ( Ω f ) [ D ( A D , f ) ] = dual of D ( A D , f ) with respect to L 2 ( Ω f ) as a pivot space. Similarly, we re-write the w -problem in (5.1b) via the RHS of (5.1e) as (compare with (2.6) of CASE 1):

(5.8) w t t = Δ ( w + w t ) b w = Δ b I ( w + w t ) N s ( b ) u ν Γ s + b ( w + w t ) b w in ( 0 , T ] × Ω s .

The term ( w + w t ) N s ( b ) u ν Γ s satisfies the zero Neumann B.C. of the operator A N , s ( b ) in (5.5), so we can re-write (5.8) as (compare with (2.7) and (2.8) of CASE 1):

(5.9a) w t t = A N , s ( b ) ( w + w t ) N s ( b ) u ν Γ s + b ( w + w t ) b w L 2 ( Ω s ) ,

(5.9b) w t t = A ˜ N , s ( b ) ( w + w t ) + A ˜ N , s ( b ) N s ( b ) u ν Γ s + b ( w + w t ) b w [ D ( A N , s ( b ) ) ] ,

where A ˜ N , s ( b ) is the isomorphic extension of the operator A N , s ( b ) in (5.5): L 2 ( Ω s ) [ D ( A N , s ( b ) ) ] = dual of D ( A N , s ( b ) ) with respect to L 2 ( Ω s ) . Henceforth, as in CASE 1, we write simply A N , s ( b ) to denote also the extension A ˜ N , s ( b ) , and likewise A D , f to denote the extension A ˜ D , f ( b ) . The action of A N , s ( b ) on the terms ( w + w t ) in (5.9a) is Δ b I ( w + w t ) and (5.9a) has an extra term b ( w + w t ) and their combination produces a cancellation of the term b ( w + w t ) for b = 1 . Thus, (5.9b) yields via (5.5)

(5.10) w t t = A N , s ( 0 ) ( w + w t ) b w + A ˜ N , s ( b ) N s ( b ) u ν Γ s ,

ultimately via (5.5) with b = 0

(5.11) w t t = Δ ( w + w t ) b w + A ˜ N , s ( b ) N s ( b ) u ν Γ s .

Combining (5.10) for the w -problem with (5.7) for the u -problem, we obtain the corresponding first order system

(5.12) d d t w w t u = 0 I 0 A N , s ( 0 ) b I A N , s ( 0 ) A N , s ( b ) N s ( b ) ν Γ s 0 A D , f D f , s ( Γ s ) A D , f w w t u + 0 0 A D , f D f , f g .

The operator in (5.12) on [ w , w t , u ] ( g 0 ) is of course the same operator A b in (1.8)–(1.9), except that in (5.12) the relevant BCs in (1.9a)–(1.9b) are included in the operator entries. Equation (5.12) can be rewritten as follows:

(5.13) d d t w w t u = A b w w t u + B D g ,

where the operator A b : H b D ( A b ) H b is of course the same as given by (1.8) and (1.9), while the (boundary) control operator B D is given by

(5.14) B D g = 0 0 A D , f D f , f g , B D : continuous L 2 ( Γ s ) D A D , f 3 4 + ε .

The adjoint operator B D of B D in (5.14) is given by

(5.15) B D x 1 x 2 x 3 = x 3 ν Γ f ; B D : continuous D A D , f 3 4 + ε L 2 ( Γ s )

in the following sense. For g L 2 ( Γ s ) and { x 1 , x 2 , x 3 } , , D A D , f 3 4 + ε , we compute as a duality pairing via (5.14).

(5.16) B D g , x 1 x 2 x 3 H = 0 0 A D , f D f , f g , x 1 x 2 x 3 H b

(5.17) = ( A D , f D f , f g , x 3 ) L 2 ( Ω f ) = ( g , D f , f A D , f x 3 ) L 2 ( Γ s )

(5.18) = g , x 3 ν Γ s L 2 ( Γ s ) = g , B D x 1 x 2 x 3 L 2 ( Γ s ) ,

where we have recalled from [4, Chapter 3] (same technique as in obtaining (2.24), CASE 1 or (4.11), CASE 2)

(5.19) D f , f A D , f x 3 = x 3 ν Γ f .

Thus, (5.15) is established. As shown in (4.11), the proof initially takes x 3 D ( A D , f ) , i.e. x 3 Ω f = 0 , and uses Green’s second theorem to establish (5.19) in this case. Then (5.19) is extended to D ( A D , f 3 4 + ε ) H 3 2 + 2 ε ( Ω f ) .

5.2 The Luenberger’s compensator for the heat-structure interaction model (5.1a)–(5.1f) with Dirichlet control g at the external boundary Γ f , b = 0 , 1

5.2.1 Special selection of the data

For the present heat-structure interaction problem with Dirichlet control at the external boundary Γ f in the notation of (1.4)–(1.6), we take

A “exponentially stable”: e A t c e δ t , δ > 0 , t 0 , (5.20a) F = B ; C = B , K = B , (5.20b)

as in CASE 2, and unlike CASE 1. Thus, the special setting becomes, in this case,

(5.21) [ partial observation of the state y ] = C y = B y , control g = F z = B z

leading to the Luenberger’s dynamics

y ˙ = A y + B B z , (5.22a) z ˙ = A z + B ( B y ) (5.22b)

(as B F K C = 0 as in CASE 2, unlike CASE 1) and hence to

(5.23) d d t [ y z ] = ( A B B ) [ y z ] ; [ y ( t ) z ( t ) ] = e ( A B B ) t [ y 0 z 0 ] .

Insight. How did we decide that F = B in (5.20b); that is, that the preassigned control g = F z is given by g = B z or F = B ? We first notice that, regardless of the choice of F , the Luenberger scheme in (1.4)–(1.6), yields that ( A K C ) is the resulting sought-after operator in characterizing the quantity [ y z ] of interest. As A is, in our case, dissipative and we surely seek to retain dissipativity, then we choose K C = B B , or K = B , C = B . Then, the (dissipative) operator ( A B B ) is the key operator to analyze for the purpose of concluding that the semigroup e ( A B B ) t is (analytic as well as) uniformly stable. Thus, at this stage, with F not yet committed and K = B , C = B committed, the z -equation becomes z ˙ = [ A B F B B ] z + B ( B y ) . It is natural to test either F = B or else F = B , whichever choice may yield the desired properties for the feedback operator A F = A B B . What is the right sign? Thus, passing from the historical scheme (1.4a)–(1.4f) to our present PDE problem (5.1a)–(5.1f), the corresponding operator B D is critical in imposing the boundary conditions for the feedback operator A F , D ( b ) = A b B D B D in our CASE 3. But in our present CASE 3, if we choose F B D and so g = B D z 1 z 2 z 3 = z 3 ν Γ f by (5.15) , this then implies the boundary condition h Γ f = h ν Γ f for { v 1 , v 2 , h } in D ( A F , D ( b ) ) as in (5.34) below. With this B.C., the argument in (5.51)–(5.53) below leads to the trace term i ω h Γ f 2 in (5.54), and hence to the critical term i ω [ v 2 2 + h 2 + h Γ f 2 ] in (5.57), with the correct “minus” sign “ ” for the argument of Theorem 5.4, in particular estimate (5.47), to succeed. Therefore, in view of the interface condition u Γ f = g in (5.1c) at the external boundary Γ f , the B.C. h Γ f = h ν Γ f confirms that g = h ν Γ f , or that g = F w w t u = B D w w t u = h ν Γ f . Hence, the choice F B D as in (5.20b) is the correct one in our CASE 3.

5.2.2 The counterpart of y ˙ = A y + B B z in (5.22b) for the HSI model (5.1a)–(5.1f) with F B D

Accordingly, for z = [ z 1 , z 2 , z 3 ] , the Luenberger’s compensator variable, in line with (5.21) we select the Dirichlet control g in the form

(5.24) g = B D z = z 3 ν Γ f , z 3 D A D , f 3 4 + ε H 3 2 + 2 ε ( Ω f )

by (5.15). With y = [ w , w t , u ] , the PDE version of the abstract feedback problem corresponding to (5.22a)

(5.25) y ˙ = A b y + B D B D z , or d d t w w t u = w t Δ ( w + w t ) b w Δ u + B D z 3 ν Γ f

recalling A b from (1.8) and B D from (5.15), with g as in (5.24), is

u t Δ u = 0 in ( 0 , T ] × Ω f ; (5.26a) w t t Δ w Δ w t + b w = 0 in ( 0 , T ] × Ω s ; (5.26b) u Γ f = z 3 ν Γ f on ( 0 , T ] × Γ f ; u = w t on ( 0 , T ] × Γ s ; (5.26c) ( w + w t ) ν = u ν on ( 0 , T ] × Γ s . (5.26d)

5.2.3 The counterpart of the dynamic compensator equation z ˙ = A z + B ( B y ) in (5.22b) for the HSI model (5.1a)–(5.1f)

With partial observation as in (5.21) according to (5.24) or (5.15)

(5.27) g = B D y = B D w w t u = u ν Γ f = partial observation of state y = w w t u ,

the compensator equation (5.22b) in z = [ z 1 , z 2 , z 3 ] becomes in our present HSI case

(5.28) z ˙ = d d t z 1 z 2 z 3 = A b z + B D u ν Γ f

or via (1.8) for A b and (5.14) and (5.15) for B D , B D , respectively,

(5.29) z 1 t z 2 t z 3 t = z 2 Δ ( z 1 + z 2 ) b z 1 Δ z 3 + 0 0 A D , f D f , f u ν Γ f ,

whereby z 1 t = z 2 , z 1 t t = z 2 t . The PDE version of the abstract z -model (5.29) with partial observation u ν Γ f in the Dirichlet condition at the external boundary Γ f as in (5.27), is via (5.24)

z 3 t Δ z 3 = 0 in ( 0 , T ] × Ω f ; (5.30a) z 1 t t Δ z 1 Δ z 1 t + b z 1 = 0 in ( 0 , T ] × Ω s ; (5.30b) z 3 Γ f = u ν Γ f on ( 0 , T ] × Γ f ; z 3 = z 1 t on ( 0 , T ] × Γ s ; (5.30c) ( z 1 + z 1 t ) ν = z 3 ν on ( 0 , T ] × Γ s . (5.30d)

5.2.4 The dynamics d ˙ = ( A b B D B D ) d , d ( t ) = y ( t ) z ( t ) , corresponding to (5.23): analyticity and exponential decay, b = 0 , 1

The main result of the present section is as follows:

Theorem 5.1

Let b = 0 , 1 . The (feedback) operator

(5.31) A F , D ( b ) = A b B D B D , H b D ( A F , D ( b ) ) = { x H b : ( I + A b 1 B D B D ) x D ( A b ) }

is the infinitesimal generator of a s.c. contraction semigroup e A F , D ( b ) t on H b , which moreover is analytic and exponentially stable on H b : there exist constants C 1 , ρ > 0 possibly depending on b such that

(5.32) e A F , D ( b ) t ( H b ) = e ( A b B D B D ) t ( H b ) C e ρ t , t 0 .

A more detailed description of D ( A F , D ( b ) ) is given in (5.35a) and (5.35b).

The proof of Theorem 5.1 is by PDE methods, which consist of analyzing the corresponding PDE system (5.30). We proceed through a series of steps.

Step 1.

Lemma 5.2

Let b = 0 , 1 . With d = [ d 1 , d 2 , d 3 ] H b , the abstract equation

(5.33a) d ˙ = A F , D ( b ) d = ( A b B D B D ) d

or via (1.8) on A b , (5.14) and (5.15) on B D and B D , respectively

(5.33b) d 1 t d 2 t d 3 t = d 2 Δ ( d 1 + d 2 ) b d 1 Δ d 3 0 0 A D , f D f , f d 3 ν Γ f ,

so that d 2 = d 1 t , d 2 t = d 1 t t corresponds to the following PDE system, where we relabel the variable [ d 1 , d 2 = d 1 t , d 3 ] = [ w ^ , w ^ t , u ^ ] for convenience

u ^ t Δ u ^ = 0 in ( 0 , T ] × Ω f ; (5.34a) w ^ t t Δ w ^ Δ w ^ t + b w ^ = 0 in ( 0 , T ] × Ω s ; (5.34b) u ^ Γ f = u ^ ν Γ f on ( 0 , T ] × Γ f ; u ^ = w ^ t on ( 0 , T ] × Γ s ; (5.34c) ( w ^ + w ^ t ) ν = u ^ ν on ( 0 , T ] × Γ s . (5.34d)

Description of D ( A F , D ( b ) ) . We have { v 1 , v 2 , h } D ( A F , D ( b ) ) = D ( A b B D B D ) if any only if

  1. (5.35a) v 1 , v 2 H 1 ( Ω s ) \ R for b = 0 , H 1 ( Ω s ) for b = 1 , so that v 2 Γ s = h Γ s H 1 2 ( Γ s ) in both cases ; Δ ( v 1 + v 2 ) L 2 ( Ω s ) ;

  2. (5.35b) h H 1 ( Ω f ) , Δ h L 2 ( Ω f ) , h Γ f = h ν Γ f H 1 2 ( Γ f ) ; h Γ s v 2 Γ s , h ν Γ s = ( v 1 + v 2 ) ν Γ s H 1 2 ( Γ s ) .

The adjoint operator A F , D ( b ) = A b B D B D . For { v 1 , v 2 , h } D ( A F , D ( b ) ) (to be characterized below), we have after recalling A b in (1.10a) and B D v 1 v 2 h = h ν Γ f from (5.15)

(5.36) A F , D ( b ) v 1 v 2 h = ( A b B D B D ) v 1 v 2 h = v 2 Δ ( v 2 v 1 ) + b v 1 Δ h B D h ν Γ f

(5.37) = v 2 Δ ( v 2 v 1 ) + b v 1 Δ h 0 0 A D , f D f , f h ν Γ f

recalling B D in (5.14).

Description of D ( A F , D ( b ) ) . We have { v 1 , v 2 , h } D ( A F , D ( b ) ) = D ( A b B D B D ) if and only if the same conditions for D ( A F , D ( b ) ) in (5.35a) and (5.35b) apply, except that now (5.35a) is replaced by Δ ( v 2 v 1 ) L 2 ( Ω s ) and (5.35b) is replaced by

(5.38) v 2 v 1 ν Γ s = h ν Γ s H 1 2 ( Γ s ) .

The PDE version corresponding to the adjoint operator A F , D ( b ) is given by

d d t w 1 w 2 u = A F , D ( b ) w 1 w 2 u or h t Δ h = 0 in ( 0 , T ] × Ω f ; (5.39a) w 1 t t Δ w 1 Δ w 1 t + b w 1 = 0 in ( 0 , T ] × Ω s ; (5.39b) h Γ f = h ν Γ f on ( 0 , T ] × Γ f ; (5.39c) h Γ s = w 1 t Γ s ; ( w 1 + w 1 t ) ν Γ s = h ν Γ s on ( 0 , T ] × Γ s , (5.39d)

(where w 1 t = w 2 in (5.36), top line) with I.C. in H b (counterpart of (3.21a)–(3.21d) in CASE 1, and (4.32a)–(4.32d) in CASE 2).

Step 2. (Analysis of the PDE problem (5.34): the operator A F , D ( b ) = A b B D B D in (5.33a))

Proposition 5.3

The operator A F , D ( b ) = A b B D B D in (5.33) and its H b -adjoint A F , D ( b ) = A b B D B D are dissipative

(5.40) Re ( A b B D B D ) v 1 v 2 h , v 1 v 2 h H = v 2 Ω s 2 h Ω f 2 h ν Γ f Γ f 2 , { v 1 , v 2 , h } D ( A F , D ( b ) ) ,

(5.41) Re ( A b B D B D ) v 1 v 2 h , v 1 v 2 h H = v 2 Ω s 2 h Ω f 2 h ν Γ f Γ f 2 , { v 1 , v 2 , h } D ( A F , D ( b ) ) ,

in the L 2 ( ) -norms of Ω s and Ω f , and the L 2 ( Γ s ) -norm on Γ s . Hence, both A F , D ( b ) and A F , D ( b ) are maximal dissipative and thus generate s.c. contraction semigroups e A F , D ( b ) t and e A F , D ( b ) t on H b . Explicitly in terms of the corresponding PDE systems we have:

(5.42) w ^ ( t ) w ^ t ( t ) u ^ ( t ) = e A F , D ( b ) t w ^ 0 w ^ 1 u ^ 0 = e ( A b B D B D ) t w ^ 0 w ^ 1 u ^ 0

for the { w ^ , w ^ t , u ^ } -fluid-structure interaction model given by (5.34a)–(5.34d) on H b : with I.C. { w ^ 0 , w ^ 1 , u ^ 0 } H b . Similarly, for the adjoint A F , D ( b ) , corresponding to the model (5.39a)–(5.39d).

Proof of (5.3)

For { v 1 , v 2 , h } D ( A F , D ( b ) ) in (5.35a) and (5.35b), we compute:

(5.43) Re ( A b B D B D ) v 1 v 2 h , v 1 v 2 h H b = Re A b v 1 v 2 h , v 1 v 2 h H b B D v 1 v 2 h 2

(5.44) = v 2 2 h 2 h ν Γ f L 2 ( Γ f ) 2 ,

recalling (1.12) and (5.24). Thus, (5.40) is established. Similarly for (5.41) recalling (1.13) for A b .□

5.2.5 Analyticity of e A F , D ( b ) t and e A F , D ( b ) t on H b , b = 0 , 1

Remark 5.1

A remark such as Remark 3.2 (CASE 1) or Remark 4.1 (CASE 2) applies in the present case, to justify (5.47) from (5.46).

Theorem 5.4

Let b = 0 , 1 . Let ω R and

(5.45) ( i ω I A F , D ( b ) ) v 1 v 2 h = v 1 v 2 h H b

for { v 1 , v 2 , h } D ( A F , D ( b ) ) identified in (5.35a) and (5.35b). Then:

  1. the following estimate holds true: there exists a constant C ε > 0 such that:

    (5.46) Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 + b v 2 2 + v 2 2 + h 2 + h ν Γ f Γ f 2 C ε { v 1 2 + b v 1 2 + v 2 2 + h 2 } , ω ε > 0 .

  2. In view of Remark 5.1, estimate (5.46) without the term h 2 + h ν Γ f 2 is equivalent to

    (5.47) A F , D ( b ) R ( i ω , A F , D ( b ) ) ( H b ) C ε , ω ε > 0 ,

    R ( i ω , A F , D ( b ) ) = ( i ω I A F , D ( b ) ) 1 , which in turn is equivalent to

    (5.48) R ( i ω , A F , D ( b ) ) ( H b ) c ε ω , ω ε > 0 .

    Thus, the s.c. contraction semigroup e A F , D ( b ) t asserted by Proposition 5.3 is analytic on H b by [4, Theorem 3E.3 p. 334]. Similarly for e A F , D ( b ) t on H b . The explicit PDE version of (5.33) for system { w ^ ( t ) , w ^ t ( t ) , u ^ ( t ) } is given by (5.34a)–(5.34d).

Proof

(i) The proof of estimate (5.46) follows closely the technical proof of estimate (3.30) for the operator A F , N ( b ) in CASE 1, or (4.41) for the operator A F , D ( b ) in CASE 2 except that now the argument uses the B.C. of A F , D ( b ) rather than the B.C. of A F , N ( b ) or the B.C. of A F , D ( b ) . This, in particular, requires an argument different from CASE 1 and CASE 2 to handle the more challenging case b = 1 , with full H 1 ( Ω ) -norm on the first component space. We indicate the relevant changes:

Step 1. Return to (5.45), re-written for { v 1 , v 2 , h } D ( A F , D ( b ) ) and { v 1 , v 2 , h } H b :

i ω v 1 v 2 = v 1 , (5.49a) i ω v 2 [ Δ ( v 1 + v 2 ) b v 1 ] = v 2 , (5.49b) i ω h Δ h = h . (5.49c)

Step 2. Take the L 2 ( Ω f ) -inner product of equation (5.49c) against Δ h , use Green’s First theorem (counterpart (4.45))

(5.50) i ω Γ f h h ¯ ν d Γ f + i ω Γ s h h ¯ ν d Γ s i ω h 2 Δ h 2 = ( h , Δ h ) .

Similarly, we take the L 2 ( Ω s ) -inner product of (5.49b) against [ Δ ( v 1 + v 2 ) b v 1 ] , use Green’s First Theorem to evaluate Ω s v 2 Δ ( v ¯ 1 + v ¯ 2 ) d Ω s , recalling that the normal vector ν is inward with respect to Ω s , and obtain (4.45), repeated here

(5.51) i ω Γ s v 2 ( v ¯ 1 + v ¯ 2 ) ν d Γ s i ω ( v 2 , ( v 1 + v 2 ) ) i ω ( v 2 , b v 1 ) Δ ( v 1 + v 2 ) b v 1 2 = ( v 2 , [ Δ ( v 1 + v 2 ) b v 1 ] ) .

Now we invoke the B.C. in (5.35a) and (5.35b) for D ( A F , D ( b ) ) :

(5.52) h Γ f = h ν Γ f ; h Γ s = v 2 Γ s ; ( v 1 + v 2 ) ν Γ s = h ν Γ s in D ( A F , D ( b ) )

and re-write (5.50) and (5.51) as follows, respectively

(5.53) i ω Γ f h 2 d Γ s + i ω Γ s h h ¯ ν d Γ s i ω h 2 Δ h 2 = ( h , Δ h ) ,

(5.54) i ω Γ s h h ¯ ν d Γ s i ω v 2 2 + ( v 2 , ( i ω v 1 ) ) i ω ( v 2 , b v 1 ) Δ ( v 1 + v 2 ) b v 1 2 = ( v 2 , [ Δ ( v 1 + v 2 ) b v 1 ] ) .

Summing up (5.53) and (5.54) yields after a cancellation of the boundary terms i ω Γ s h h ¯ ν d Γ s :

(5.55) i ω h Γ f 2 i ω [ v 2 2 + h 2 ] = Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 + i ω ( v 2 , b v 1 ) + i ω ( v 2 , v 1 ) + ( v 2 , [ Δ ( v 1 + v 2 ) b v 1 ] ) + ( h , Δ h ) .

We now use the identities from (5.49a)

(5.56a) i ω ( v 2 , v 1 ) = ( v 2 , ( i ω v 1 ) ) = v 2 2 + ( v 2 , v 1 ) ,

(5.56b) i ω ( v 2 , v 1 ) = ( v 2 , i ω v 1 ) = v 2 2 + ( v 2 , v 1 )

to obtain the final identity from (5.55)

(5.57) Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 + i ω [ v 2 2 + h 2 + h Γ f 2 ] = v 2 2 + b v 2 2 + ( v 2 , v 1 ) + b ( v 2 , v 1 ) ( v 2 , [ Δ ( v 1 + v 2 ) b v 1 ] ) ( h , Δ h ) .

Identity (5.57) is the counterpart of identity (4.50) for Dirichlet control g on the interface Γ s , after replacing h ν Γ s 2 in (4.50) with h ν Γ f 2 = h Γ f 2 in (5.57). Identity (5.57) is also the counterpart of identity (3.38) for Neumann control g at the Interface Γ s , after replacing h Γ s 2 in (3.38) with h ν Γ f 2 = h Γ f 2 in (5.57).□

Step 3. We proceed as in Step 3 of CASE 1 or of CASE 2. We take the real part of identity (5.57) and next obtain

(5.58) ( 1 ε ) [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] ( 1 + ε ) [ v 2 2 + b v 2 2 ] + C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] ,

the counterpart of estimate (4.52) CASE 2, or estimate (3.40), CASE 1.

Step 4. We now take the imaginary part of identity (5.57), and proceeding as in Step 4 of CASE 1 or CASE 2, we arrive at the following estimate

(5.59a) v 2 2 + h 2 + h Γ f 2 ε b v 2 2 + OK ε ,

(5.59b) OK ε = 2 ε [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] .

Equation (5.59a) and (5.59b) are the counterpart of (3.46) (CASE 1) or ( 4.5 9 1 ) and ( 4.6 0 1 ) via (4.64) (CASE 2), the difference being that the boundary term h Γ s 2 on the LHS of (3.46) (CASE 1), respectively, and the boundary term h ν Γ s 2 on the LHS of ( 4.5 9 1 ) (CASE 2, where h ν Γ s = v 2 Γ s h Γ s holds as in (4.64)), is now replaced by the boundary term h Γ f 2 on the LHS of (5.59a).

Case b = 0 . In the case b = 0 , the rest of the proof is the same as in Remark 3.1 (CASE 1, with Neumann control on Γ s ) or as in Remark 4.2 (CASE 2 with Dirichlet control on Γ s ) by replacing the term h Γ s 2 in (3.46), respectively, by replacing the term h ν Γ s 2 in (4.58) with the term h Γ f 2 = h ν Γ f 2 in (5.59a). Thus, in the case b = 0 , the sought-after estimate (5.45), and hence, (5.47) are established in the space H b = 0 .

Step 5. Case b = 1 . In the present case, as in CASE 2 of Dirichlet control on Γ s , there are additional challenges in establishing the sought-after estimate (5.46) with b = 1 , i.e., in the space H b = 1 . The required argument is provided in the present step, which is the counterpart of Step 5 in Section 3 (CASE 1) or of Step 5 in Section 4 (CASE 2). In line with these two cases, it will rely on Lemma 3.5.

We return to (5.59a) and add the term ε 1 v 2 Γ s 2 to both sides, 1 ε 1 > 0 to be chosen below, where on the RHS, we use v 2 Γ s = h Γ s by (5.52).

We obtain

v 2 2 + ε 1 v 2 Γ s 2 + h 2 + h Γ f 2 ε b v 2 2 + ε 1 h Γ s 2 + OK ε , ( 5.6 0 1 )

for ω 2 ε > 0 , hence

ε 1 [ v 2 2 + v 2 Γ s 2 ] + 1 2 h 2 + 1 2 h 2 + h Γ f 2 ε b v 2 2 + ε 1 [ h 2 + h 2 ] + OK ε . ( 5.6 1 1 )

In going from ( 5.6 0 1 ) to ( 5.6 1 1 ), we have used: on the LHS, that 0 < ε 1 1 ; on the RHS, that the L 2 ( Γ s ) -norm of [ h Γ s ] is dominated by the H 1 ( Ω f ) -norm of h . Next, on the LHS of ( 5.6 1 1 ), we invoke Lemma 3.5(a) for both v 2 in Ω s and h in Ω f . We obtain

ε 1 c 1 v 2 2 ε 1 [ v 2 2 + v 2 Γ s 2 ] , ( 5.6 2 1 )

1 2 c 1 h 2 1 2 [ h 2 + h Γ f 2 ] . ( 5.6 3 1 )

Invoking ( 5.6 2 1 ), ( 5.6 3 1 ), we re-write ( 5.6 1 1 ) as follows:

ε 1 c 1 ε b v 2 2 + 1 2 ε 1 h 2 + 1 2 c 1 ε 1 h 2 OK ε , ω 2 ε > 0 . ( 5.6 4 1 )

We now select 1 ε > 0 ( b = 1 ) as follows:

ε < ε 1 c 1 ε ; 1 4 < 1 2 ε 1 ; 1 4 c 1 < 1 2 c 1 ε 1 ; ( ε + ε ) < ε 1 min 1 4 , 1 4 c 1 , ( 5.6 5 1 )

(we can always assume c 1 > 1 , see Lemma 3.5(a)). Using ( 5.6 5 1 ) on the LHS of ( 5.6 4 1 ) yields

ε v 2 2 + 1 4 h 2 + 1 4 c 1 h 2 2 ε [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] , ( 5.6 6 1 )

for ω ε 2 , after invoking (5.59b) for OK ε . Finally, ( b = 1 and ε < 1 4 , ε < 1 4 c 1 ), we obtain

b v 2 2 + h 2 + h 2 2 ε [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + C ˜ ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] ( 5.6 7 1 )

for ω ε 2 as in (4.56) (CASE 2). Estimate ( 5.6 7 1 ) is a counterpart of estimate (3.55) (CASE 1) or estimate (4.2.5) (CASE 2). Estimate ( 5.6 7 1 ) corresponds to (5.59) for b = 0 .

Substituting the estimate for b v 2 2 from ( 5.6 7 1 ) into the RHS of (5.59a) yields

v 2 2 + h 2 + h Γ f 2 { ε 2 ε [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + C ˜ ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] } + OK ε . ( 5.6 7 1 )

Next add to ( 5.6 7 1 ) the estimate for b v 2 2 in ( 5.6 7 1 ) to obtain

b v 2 2 + v 2 2 + h 2 + h Γ f 2 [ 2 ε ( 1 + ε ) + 2 ε ] [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] , ( 5.6 8 1 )

recalling OK ε from (5.59b). Estimate ( 5.6 8 1 ) is the counterpart of estimate (5.59) for b = 0 .

Step 6. The rest of the proof for b = 1 now proceeds as in the case b = 0 . In ( 5.6 8 1 ), we drop the terms [ h 2 + h Γ f 2 ] and substitute the resulting estimate for [ b v 2 2 + v 2 2 ] into the RHS of (5.58) and obtain

( 1 ε ) [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] ( 1 + ε ) [ 2 ε ( 1 + ε ) + 2 ε ] [ Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 ] + ( 1 + ε ) C ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] , ( 5.6 9 1 )

or, since { ( 1 ε ) ( 1 + ε ) [ 2 ε ( 1 + ε ) + 2 ε ] } k > 0 , we obtain

Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 C ^ ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] , ( 5.7 0 1 )

which is the counterpart of (3.58) in CASE 1, and ( 4.7 5 1 ) in CASE 2. By substituting ( 5.7 0 1 ) into the RHS of ( 5.6 8 1 ), we finally obtain

b v 2 2 + v 2 2 + h 2 + h Γ f 2 C ˜ ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] . ( 5.7 1 1 )

Summing up ( 5.7 0 1 ) and ( 5.7 1 1 ) yields

b v 2 2 + v 2 2 + Δ ( v 1 + v 2 ) b v 1 2 + Δ h 2 + h 2 + h Γ f 2 C ¯ ε [ v 1 2 + b v 1 2 + v 2 2 + h 2 ] , ( 5.7 2 1 )

which is the sought-after estimate (5.15), b = 0 , b = 1 , h Γ f 2 = h ν Γ f 2 .

5.2.6 Exponential stability of e A F , D ( b ) t and e A F , D ( b ) t on H b , b = 0 , 1

In Proposition 5.5, we shall prove that, in both cases b = 0 and b = 1 , we have

(5.73) 0 ρ ( A F , D ( b ) ) , 0 ρ ( A F , D ( b ) ) , A F , D ( b ) 1 ( H b ) , A F , D ( b ) 1 ( H b ) ,

so that there exists a disk S r 0 centered at the origin and of suitable radius r 0 > 0 such that S r 0 ρ ( A F , D ( b ) ) . Then, the resolvent bound (5.48) combined with A F , D ( b ) 1 ( H b ) in (5.73) allows one to conclude that the resolvent is uniformly bounded on the imaginary axis i R :

(5.74) R ( i ω , A F , D ( b ) ) ( H b ) const .

Hence, [45] the s.c. analytic semigroup e A F , D ( b ) t is, moreover, (uniformly) exponentially bounded: There exist constants M 1 , δ > 0 , possibly depending on “ b ” such that

(5.75) e A F , D ( b ) t ( H b ) M e δ t , t 0 .

It is similar for the adjoint A F , D ( b ) .

Proposition 5.5

Statement (5.73) holds true. Hence, the exponential stability for e A F , D ( b ) t in (5.75) holds true. More precisely, with reference to A F , D ( b ) , we have: given { v 1 , v 2 , h } H b , the unique solution { v 1 , v 2 , h } D ( A F , D ( b ) ) of

(5.76) A F , D ( b ) v 1 v 2 h = v 2 Δ ( v 1 + v 2 ) b v 1 Δ h = v 1 v 2 h

is given explicitly by

(5.77a) v 1 = ( A N , s ( b ) ) 1 ( Δ v 1 + v 2 ) + N s ( b ) ν [ A R , f 1 h + D ˜ f , s ( v 1 Γ s ) ] v 1 ν Γ s H 1 ( Ω s ) ,

(5.77b) v 2 = v 1 H 1 ( Ω s ) \ R f o r b = 0 , H 1 ( Ω s ) f o r b = 1 , h = A R , f 1 h + D ˜ f , s ( v 1 Γ s ) H 3 2 ( Ω f ) .

Here, A N , s ( b ) and N s ( b ) are defined in (5.5) and (5.4). Moreover, the operator A R , f is defined in (4.85). A new operator is the Dirichlet map D ˜ f , s defined by

D ˜ f , s μ = ψ Δ ψ = 0 i n Ω f ; (5.78a) ψ Γ s = μ , ψ ν + ψ Γ f = 0 . (5.78b)

In operator form, we have

(5.79a) v 1 v 2 h = A F , D ( b ) 1 v 1 v 2 h = ( A N , s ( b ) ) 1 ( Δ v 1 + v 2 ) + N s ( b ) ν [ A R , f 1 h + D ˜ f , s ( v 1 Γ s ) ] v 1 ν Γ s v 1 A R , f 1 h + D ˜ f , s ( v 1 Γ s )

(5.79b) = ( A N , s ( b ) ) 1 ( Δ ) + N s ( b ) ν Γ s ( A N , s ( b ) ) 1 N s ( b ) ν [ A R , f 1 ] 0 I 0 D ˜ f , s ( Γ s ) 0 A R , f 1 v 1 v 2 h ,

(5.79c) [ v 1 , v 2 , h ] H c [ v 1 , v 2 , h ] H ,

where the operators A N , s ( b ) , A R , f , D f , f , and N s ( b ) , are defined in the following proof.

Proof

Identity (5.76) and the characterization of D ( A F , D ( b ) ) in (5.35a) and (5.35b) yield

v 2 = v 1 Δ h = h L 2 ( Ω f ) ; (5.80a) h ν + h Γ f = 0 , h Γ s = v 2 Γ s = v 1 Γ s , (5.80b)

and the h -problem in (5.80) yields the solution h in (5.77b), invoking A R , f from (4.85) and D ˜ f , s from (5.78). Moreover, (5.76), v 2 = v 1 in (5.80) and (5.35b) yield

Δ ( v 1 + v 2 ) b v 1 = v 2 , or Δ v 1 b v 1 = Δ v 1 + v 2 ; (5.81a) ( v 1 + v 2 ) ν Γ s = h ν Γ s , or v 1 ν Γ s = v 1 ν Γ s + h ν Γ s . (5.81b)

Then, the solution of problem (5.81) is given by (5.77a) via (5.77b).□

Remark 5.2

A recent contribution of a heat-plate interaction with the plate subject to a (formal) “square root” damping in [46].

Acknowledgments

The authors wish to thank Yongjin Lu, Oakland University, MI, for reading the manuscript and providing useful comments; and Justin Webster, University of Maryland, Baltimore County, for providing insight and references on data assimilation.

  1. Funding information: The research of R.T. was partially supported by the National Science Foundation under Grant NSF DMS-2205508.

  2. Author contributions: All authors have accepted responsibility for the entire content of this manuscript and approved its submission.

  3. Conflict of interest: The authors state no conflict of interest.

Appendix A The genuine fluid-structure interaction model with Kelvin-Voigt (viscoelastic) damping [2,12,14]

In this article, we have deliberately chosen to consider the simpler heat-viscoelastic structure model, as a first step of an entirely new investigation regarding the corresponding Luenberger theory. To be sure, replacing the heat equation with a fluid equation introduces conceptual and technical difficulties. These however have already been dealt with and ultimately resolved in prior work by one co-author, for a structure model originally without viscoelastic damping [12], and next with viscoelastic damping [2]. A first serious obstacle is faced at the very outset: because of the coupled nonhomogeneous boundary conditions involving the linearized Navier-Stokes equations, it is not possible to use the classical, by now standard idea of N-S problems with no-slip boundary conditions to eliminate the pressure by applying the Leray projector on the equation from L 2 ( Ω ) onto the classical space { f ( L 2 ( Ω ) ) d : div f = 0 in Ω f ; f ν = 0 on Ω f } [47, p. 7]. Accordingly, [12] introduces an entirely new idea that is inspired by boundary control theory. This is explained below in the context of the problem under present consideration.

We thus consider the following fluid-plate PDE model in solution variables

u = [ u 1 ( t , x ) , u 2 ( t , x ) , , u d ( t , x ) ] (the velocity field)

and

w = [ w 1 ( t , x ) , w 2 ( t , x ) , , w d ( t , x ) ] (the structural displacement field) ,

while the scalar-valued p denotes the pressure:

( PDE ) u t Δ u + p = 0 in ( 0 , T ] × Ω f Q f ; (A.1a) div u = 0 in Q f ; (A.1b) w t t Δ w Δ w t + b w = 0 in ( 0 , T ] × Ω s Q s ; (A.1c)

( BC ) u Γ f = 0 on ( 0 , T ] × Γ f Σ f ; (A.1d) u = w t on ( 0 , T ] × Γ s Σ s ; (A.1e) u ν ( w + w t ) ν = p ν on Σ s ; (A.1f)

(A.1g) (IC) [ u ( 0 , ) , w ( 0 , ) , w t ( 0 , ) ] = [ u 0 , w 0 , w 1 ] on Ω .

The constant b in (A1.c) will take up either the value b = 0 or else the value b = 1 , as in the article. Accordingly, the space of well-posedness is taken to be the finite energy space:

b ( H 1 ( Ω s ) R ) d × ( L 2 ( Ω s ) ) d × H ˜ f , b = 0 ; (A.2a) ( H 1 ( Ω s ) ) d × ( L 2 ( Ω s ) ) d × H ˜ f , b = 1 , (A.2b)

for the variable [ u , w , w t ] , where

(A.3) H ˜ f = { f ( L 2 ( Ω f ) ) d : div f 0 in Ω f ; f ν 0 on Γ f } .

The norm-including inner product on b is given in (1.2a)–(1.2b).

Abstract model for the free dynamics (A.1a)–(A.1g)

The previous article [12] (as well as paper [48], where the d -dimensional wave equation (A1.c) is replaced by the system of dynamic elasticity) eliminated the pressure by a completely different strategy. Following the idea of [4952] (see also [4]), [12,48] identify a suitable elliptic problem for the pressure p , to be solved for p in terms of u , w and w t .

Elimination of p, by expressing p in terms of u, w, and w t . A key idea of [11,12,48] is that the pressure p ( t , x ) solves the following elliptic problem on Ω f in x , for each t :

Δ p 0 in ( 0 , T ] × Ω f Q f ; (A.4a) p = u ν ν ( w + w t ) ν ν on ( 0 , T ] × Γ s Σ s ; (A.4b) p ν = Δ u ν on ( 0 , T ] × Γ f Σ f . (A.4c)

In fact, (A.4a) is obtained by taking the divergence div across equation (A.1a), and using div u t 0 in Q f by (A.1b), as well as div Δ u = Δ div u 0 in Q f . Next, the B.C. (A.4b) on Γ s is obtained by taking the inner product of equation (A.1e) with ν . Finally, the B.C. (A.4c) on Γ f is obtained by taking the inner product of equation (A.1a) restricted on Γ f , with ν , using u Γ f 0 by (A.1d), so that on Γ f : p ν = p ν Γ f . This then results in (A.4c).

Explicit solution of problem (A) for p. We set

(A.5) p = p 1 + p 2 in Q f ,

where p 1 and p 2 solve the following problems:

Δ p 1 0 in Q f ; p 1 u ν ν ( w + w t ) ν ν on Σ s ; p 1 ν Σ f 0 on Σ f ; Δ p 2 0 in Q f ; (A.6a) p 2 0 on Σ s ; (A.6b) p 2 ν Σ f Δ u ν on Σ f . (A.6c)

Accordingly, define the following “Dirichlet" and “Neumann" maps D s and N f :

h D s g Δ h 0 in Q f ; h = g on Γ s ; h ν 0 on Γ f ; ψ N f μ Δ ψ 0 in Γ s ; (A.7a) ψ 0 on Γ s ; (A.7b) ψ ν μ on Γ f . (A.7c)

Elliptic theory gives that D s and N f are well defined and possess the following regularity [43]:

(A.8a) D s : continuous H r ( Γ s ) H r + 1 2 ( Ω f ) , r R ,

(A.8b) N f : continuous H r ( Γ f ) H r + 3 2 ( Ω f ) , r R .

Accordingly, in view of problems (A.7), we write the solutions p 1 and p 2 in (A.6), finally p in (A.5), as follows:

(A.9) p 1 = D s u ν ν ( w + w t ) ν ν Σ s ; p 2 = N f [ ( Δ u ν ) Σ f ] in Q f ,

(A.10a) p = p 1 + p 2 = Π 1 ( w + w t ) + Π 2 ( u )

(A.10b) = D s u ν ν ( w + w t ) ν ν Σ s + N f [ ( Δ u ν ) Σ f ] in Q f ,

where

(A.11a) Π 1 ( w + w t ) = D s ( w + w t ) ν ν Σ s ,

(A.11b) Π 2 ( u ) = D s u ν ν Σ s + N f [ ( Δ u ν ) Σ f ] in Q f ,

hence via (A.10a) and (A.10b):

(A.12a) p = G 1 ( w + w t ) G 2 ( u ) = Π 1 ( w + w t ) + Π 2 ( u )

(A.12b) = D s u ν ν ( w + w t ) ν ν Σ s + ( N f [ ( Δ u ν ) Σ f ] ) in Q f ;

where

(A.13) G 1 ( w + w t ) = Π 1 ( w + w t ) = D s ( w + w t ) ν ν Σ s in Q f ,

(A.14) G 2 ( w ) = Π 2 u = D s u ν ν Σ s + N f [ ( Δ u ν ) Σ f ] in Q f ,

The linear maps G 1 and G 2 in (A.12)–(A.14) are introduced mostly for notational convenience. Equations (A.10a), (A.10b), and (A.12) have managed to eliminate the pressure p , and, more pertinently, its gradient p , by expressing them in terms of the three key variables: the fluid velocity field u and the wave solution { w , w t } . By using (A.12a), we accordingly rewrite the original model (A.1a)–(A.1g) as follows:

( PDE ) u t = Δ u + G 1 ( w + w t ) + G 2 u in Q f ; (A.15a) div u = 0 in Q f ; (A.15b) w t t = Δ w + Δ w t b w in Q s ; (A.15c)

( BC ) u Γ f = 0 on Σ f ; (A.15d) u = w t on Σ s ; (A.15e)

(A.15f) (IC) [ u ( 0 , ) , w ( 0 , ) , w t ( 0 , ) ] = [ u 0 , w 0 , w 1 ] on Ω ,

only in terms of u , w , and w t , where the pressure p has been eliminated, as desired.

Abstract model of system (A.15). The abstract model of system (A.15) is given by

(A.16a) d d t w w t u = 0 I 0 Δ b I Δ 0 G 1 G 2 Δ + G 2 w w t u = A b w w t u

(A.16b) = w t Δ ( w + w t ) b w Δ u + G 1 ( w + w t ) + G 2 u ,

(A.16c) [ w ( 0 ) , w t ( 0 ) , u ( 0 ) ] = [ w 0 , w 1 , u 0 ] b ,

where the matrix form for A on the L.H.S of (A.16a) is formal and means the action described in (A.16b).

The operator A b . Recalling (A.13) and (A.14) prompts the introduction of the operator

(A.17a) A 0 I 0 Δ 2 ρ Δ 2 0 G 1 G 2 Δ + G 3

(A.17b) = 0 I 0 Δ 2 ρ Δ 2 0 { D s [ ( ( Δ ) ν ) Σ s ] } a 23 a 33 ,

(A.17c) a 33 = Δ D s ν ν Γ s + N f [ ( ( Δ ) ν ) Γ f ] ,

(A.17d) b D ( A b ) b .

The finite energy space b of well-posedness for problems (A.1a)–(A.1g), or its abstract version (A.16)–(A.17) is defined in (A). The domain D ( A b ) of A b will be identified below. To this end, we find it convenient to introduce a function π , whose indicated regularity was ascertained in [12].

The scalar harmonic function π . Henceforth, with reference to (A.17b), for [ v 1 , v 2 , f ] D ( A ) , we introduce the harmonic function π = π ( v 1 , v 2 , f ) :

(A.18) π D s f ν ν Γ s + N f [ ( Δ f ν ) Γ f ] D s ( v 1 + v 2 ) ν ν Γ s L 2 ( Ω f )

(compare with (A.10b) for the dynamic problem). According to the definition of the Dirichlet map D s and Neumann map N f given in (A.7a)–(A.7c), π = π ( v 1 , v 2 , f ) in (A.18) can be equivalently given as the solution of the following elliptic problem (compare with (A.4a)–(A.4c) for the dynamic problem):

Δ π 0 in Ω f ; (A.19a) π = f ν ν ( v 1 + v 2 ) ν ν H 1 2 ( Γ s ) on Γ s ; (A.19b) π ν = Δ f ν H 3 2 ( Γ f ) on Γ f . (A.19c)

It then follows from A b in (A.17b) and (A.18) via the function π defined in (A.18) that

(A.20) A b v 1 v 2 f = v 2 Δ ( v 1 + v 2 ) b v 1 Δ f π v 1 v 2 f b , v 1 v 2 f D ( A b ) .

References

[1] I. Lasiecka and R. Triggiani, Heat-structure interaction with viscoelastic damping: Analyticity with sharp analytic sector, exponential decay, fractional powers, Commun. Pure Appl. Anal. 15 (2015), no. 5, 1515–1543. 10.3934/cpaa.2016001Search in Google Scholar

[2] R. Mahawattege and R. Triggiani, Fluid-structure interaction with Kelvin-Voight damping: analyticity, spectral analysis, exponential decay, Appl. Math. Optim. 84 (2021), 1821–1863. 10.1007/s00245-021-09812-5Search in Google Scholar

[3] R. Mahawattege and R. Triggiani, Fluid-plate interaction with strong damping: analytic, uniformly stable semigroup, International Workshop on Operator Theory and Its Applications, IWOTA, 2021, (to appear). Search in Google Scholar

[4] I. Lasiecka and R. Triggiani, Control Theory for Partial Differential Equations: Continuous and Approximation Theories I, Abstract Parabolic Systems, Encyclopedia of Mathematics and Its Applications, Vol. 1, Cambridge University Press, Cambridge, 2000. 10.1017/CBO9781107340848Search in Google Scholar

[5] R. Mahawattege and R. Triggiani, Fluid-plate interaction with Kelvin-Voigt damping and bending moment at the interface: Well-posedness, spectral analysis, uniform stability, Recent Developments in Operator Theory, Mathematical Physics and Complex Analysis (Cham), D. Alpay, J. Behrndt, F. Colombo, I. Sabadini, and D. C. Struppa, (Eds)., Springer International Publishing, 2023, pp. 217–267. 10.1007/978-3-031-21460-8_6Search in Google Scholar

[6] S. K. Chang, I. Lasiecka, and R. Triggiani, Finite element compensators for thermo-elastic systems with boundary control and point observation, Num. Funct. Anal. Optimiz. 20 (1999), no. 5–6, 419–435. 10.1080/01630569908816903Search in Google Scholar

[7] E. Hendrickson and I. Lasiecka, Finite-dimensional approximations of boundary control problems arising in partially observed hyperbolic systems, Dynam. Contin. Discrete Impuls. Systems 1 (1995), no. 1, 101–142. Search in Google Scholar

[8] G. Ji and I. Lasiecka, Partially observed analytic systems with fully unbounded actuators and sensors-fem algorithms, Comput. Optim. Appl. 11 (1998), no. 2, 111–136. 10.1023/A:1018681526852Search in Google Scholar

[9] R. Curtain, Finite dimensional compensators for parabolic distributed systems with unbounded control and observation, SIAM J. Control Optim. 22 (1984), no. 2, 255–276. 10.1137/0322018Search in Google Scholar

[10] J. M. Schumacher, A direct approach to compensator design for distributed parameter systems, SIAM J. Control Optim. 21 (1983), no. 6, 823–836. 10.1137/0321050Search in Google Scholar

[11] G. Avalos, I. Lasiecka, and R. Triggiani, Higher regularity of a coupled parabolic-hyperbolic fluid-structure interactive system, Georgian Math. J. 15 (2008), no. 3, 403–437.10.1515/GMJ.2008.403Search in Google Scholar

[12] G. Avalos and R. Triggiani, The coupled PDE system arising in fluid-structure interaction. Part I: Explicit semigroup generator and its spectral properties, AMS Contemporary Math. Fluids Waves 440 (2007), 15–55. 10.1090/conm/440/08475Search in Google Scholar

[13] G. Avalos and R. Triggiani, Uniform stabilization of a coupled PDE system arising in fluid-structure interaction with boundary dissipation at the interface, Discrete Contin. Dyn. Syst. 22 (2008), no. 4, 817–833. 10.3934/dcds.2008.22.817Search in Google Scholar

[14] R. Triggiani, Linear parabolic-hyperbolic fluid-structure interaction models. The case of static interface, in: Mathematical Theory of Evolutionary Fluid-Flow Structure Interactions, Oberwolfach Seminars, Vol. 48. Birkhäuser, 2008, pp. 53–171. 10.1007/978-3-319-92783-1_2Search in Google Scholar

[15] R. Triggiani, Heat-viscoelastic plate interaction via bending moments and shear forces operators: analyticity, spectral analysis, exponential decay, Appl. Math. Optim. 82 (2020), no. 2, 755–797. 10.1007/s00245-018-9547-0Search in Google Scholar

[16] R. Triggiani and J. Zhang, Heat-viscoelastic plate interaction: analyticity, spectral analysis, exponential decay, Evol. Equ. Control Theory 1 (2017), no. 1, 153–182. 10.3934/eect.2018008Search in Google Scholar

[17] I. Lasiecka and Y. Lu, Asymptotic stability of finite energy in Navier Stokes-elastic wave interaction, Semigroup Forum 82 (2011), no. 1, 61–82. 10.1007/s00233-010-9281-7Search in Google Scholar

[18] I. Lasiecka and Y. Lu, Interface feedback control stabilization of a nonlinear fluid-structure interaction, Nonlinear Anal. 75 (2012), no. 3, 1449–1460. 10.1016/j.na.2011.04.018Search in Google Scholar

[19] Y. Lu, Uniform stabilization to equilibrium of a non-linear fluid-structure interaction model, Nonlinear Anal. Real World Appl. 25 (2015), 51–63. 10.1016/j.nonrwa.2015.02.006Search in Google Scholar

[20] Y. Lu, Global uniform stabilization to nontrivial equilibrium of a nonlinear fluid viscoelastic-structure interaction, Applicable Anal. 97 (2018), no. 10, 1797–1813. 10.1080/00036811.2017.1341975Search in Google Scholar

[21] I. Kukavica and A. Tuffaha, Regularity of solutions to a free boundary problem of fluid-structure interaction, Indiana Univ. Math. J. 61 (2012), no. 5, 1817–1859. 10.1512/iumj.2012.61.4746Search in Google Scholar

[22] I. Kukavica, A. Tuffaha, and M. Ziane, Strong solutions to a nonlinear fluid structure interaction system, J. Differential Equations 247 (2009), no. 5, 1452–1478. 10.1016/j.jde.2009.06.005Search in Google Scholar

[23] I. Kukavica, A. Tuffaha, and M. Ziane, Strong solutions for a fluid structure interaction system, Adv. Differential Equations 15 (2010), no. 3–4, 231–254. 10.57262/ade/1355854749Search in Google Scholar

[24] I. Kukavica, A. Tuffaha, and M. Ziane, Strong solutions to a Navier-Stokes-Lamé system on a domain with a non-flat boundary, Nonlinearity 24 (2011), no. 1, 159–176. 10.1088/0951-7715/24/1/008Search in Google Scholar

[25] M. Ignatova, I. Kukavica, I. Lasiecka, and A. Tuffaha, On well-posedness for a free boundary fluid-structure model, J. Math. Phys. 53 (2012), no. 11, 115624. 10.1063/1.4766724Search in Google Scholar

[26] M. Ignatova, I. Kukavica, I. Lasiecka, and A. Tuffaha, Small data global existence for a fluid-structure model, Nonlinearity 30 (2017), 848–898. 10.1088/1361-6544/aa4ec4Search in Google Scholar

[27] D. G. Luenberger, An introduction to observers, IEEE Trans. Automatic Control AC-16 (1971), 596–602. 10.1109/TAC.1971.1099826Search in Google Scholar

[28] J. Zabczyk, Mathematical Control Theory: An Introduction, Springer Nature Switzerland AG, 2nd edn., 2020. 10.1007/978-3-030-44778-6Search in Google Scholar

[29] A. Azouani, E. Olson, and E. S. Titi, Continuous data assimilation using general interpolant observables, J. Nonlinear Sci. 24 (2014), no. 2, 277–304. 10.1007/s00332-013-9189-ySearch in Google Scholar

[30] A. Balakrishna and A. Biswas, Determining map, data assimilation and an observable regularity criterion for the three-dimensional Boussinesq system, Appl. Math. Optim. 86 (2022), no. 3, 1–53. 10.1007/s00245-022-09896-7Search in Google Scholar

[31] A. Farhat, E. Lunasin, and E. S. Titi, Continuous data assimilation for a 2D Bénard convection system through horizontal velocity measurements alone, J. Nonlinear Sci. 27 (2017), no. 3, 1065–1087. 10.1007/s00332-017-9360-ySearch in Google Scholar

[32] E. Lunasin and E. S. Titi, Finite determining parameters feedback control for distributed nonlinear dissipative systems – a computational study, Evol. Equ. Control Theory 6 (2017), no. 4, 535–557. 10.3934/eect.2017027Search in Google Scholar

[33] C. Foias and G. Prodi, Sur le comportement global des solutions non-stationnaires des équations de Navier-Stokes en dimension 2, Rend. Semin. Mat. Univ. Padova 39 (1967), 1–34. Search in Google Scholar

[34] A. Biswas and R. Price, Continuous data assimilation for the three dimensional Navier-Stokes equations, SIAM J. Math. Anal. 53 (2021), no. 6, 6697–6723. 10.1137/20M1323229Search in Google Scholar

[35] A. Pazy, Semigroups of Linear Operators and Applications to Partial Differential Equations, Springer, New York, NY, 1983. 10.1007/978-1-4612-5561-1Search in Google Scholar

[36] S. Chen and R. Triggiani, Proof of two conjectures of G. Chen and D. L. Russell on structural damping for elastic systems: The case α=1∕2 (with S. Chen), Springer-Verlag Lect. Notes Math. 1354 (1988), 234–256. 10.1007/BFb0089601Search in Google Scholar

[37] S. Chen and R. Triggiani, Proof of extensions of two conjectures on structural damping for elastic systems: The case 1∕2≤α≤1, Pacific J. Math. 136 (1989), 15–55. 10.2140/pjm.1989.136.15Search in Google Scholar

[38] S. Chen and R. Triggiani, Characterization of domains of fractional powers of certain operators arising in elastic systems, and applications, J. Differential Equations 88 (1990), 279–293. 10.1016/0022-0396(90)90100-4Search in Google Scholar

[39] R. Triggiani, A heat-viscoelastic structure interaction model with Neumann and Dirichlet boundary control at the interface: Optimal regularity, control theoretic implications, Appl. Math. Optim. 73 (2016), no. 3, 571–594. 10.1007/s00245-016-9348-2Search in Google Scholar

[40] I. Lasiecka, Unified theory for abstract parabolic boundary problems – a semigroup approach, Appl. Math. Optim. 6 (1980), 31–62. 10.1007/BF01442882Search in Google Scholar

[41] D. Fujiwawa, Concrete characterization of the domain of fractional powers of some elliptic differential operators of the second order, Proc. Japan Acad. 43 (1967), 82–86. 10.3792/pja/1195521686Search in Google Scholar

[42] P. Grisvard, Caractérisation de quelques espaces d’interpolation, Arch. Rational Mech. Anal. 25 (1967), 40–63. 10.1007/BF00281421Search in Google Scholar

[43] J. L. Lions and E. Magenes, Nonhomogeneous Boundary Value Problems and Applications, Vol. 1, Springer-Verlag, Berlin, Heidelberg, 1972. 10.1007/978-3-642-65217-2_1Search in Google Scholar

[44] I. Lasiecka, R. Triggiani, and X. Zhang, Nonconservative wave equations with unobserved Neumann BC: global uniqueness and observability in one shot, Contemporary Math. 268 (2000), 227–326. 10.1090/conm/268/04315Search in Google Scholar

[45] J. Pruss, On the spectrum of c0 semigroups, Trans. Amer. Math. Soc. 284 (1984), 847–857. 10.1090/S0002-9947-1984-0743749-9Search in Google Scholar

[46] R. Triggiani and X. Wan, A heat-structure interaction model with (formal) ‘square-root’ damping: analyticity and uniform stability, Special issue of Quantum Studies: Mathematics and Foundations (submitted). Search in Google Scholar

[47] P. Constantin and C. Foias, Navier-Stokes Equations, University of Chicago Press, Chicago, London, 1989. 10.7208/chicago/9780226764320.001.0001Search in Google Scholar

[48] G. Avalos and R. Triggiani, Semigroup well-posedness in the energy space of a parabolic-hyperbolic coupled Stokes-Lamé PDE system of fluid-structure interaction, Discrete Contin. Dyn. Syst. Ser. S. 2 (2009), no. 3, 417–448. 10.3934/dcdss.2009.2.417Search in Google Scholar

[49] I. Lasiecka and R. Triggiani, A cosine operator approach to modeling l2 boundary input hyperbolic equations, Appl. Math. Optim. 7 (1981), 35–83. 10.1007/BF01442108Search in Google Scholar

[50] R. Triggiani, A cosine operator approach to modeling boundary input problems for hyperbolic systems, Springer-Verlag Lecture Notes in Control and Information Sciences, Vol. 6, 1978, pp. 380–390. 10.1007/BFb0007256Search in Google Scholar

[51] R. Triggiani, Boundary stabilization of parabolic equations, International Conference on Recent Advances in Differential Equations, Miramare-Trieste (Italy), R. Conti, (Ed)., Springer-Verlag LNCIS, Vol. 14, Academic Press, 1978, (Also, Proceedings of International Conference on Systems Analysis, I.R.I.A., Paris), pp. 305–313. Search in Google Scholar

[52] R. Triggiani, Well-posedness and regularity of boundary feedback parabolic systems, J. Differential Equations 36 (1980), 347–362. 10.1016/0022-0396(80)90054-6Search in Google Scholar

Received: 2023-01-16
Revised: 2023-05-07
Accepted: 2023-05-11
Published Online: 2023-06-22

© 2023 the author(s), published by De Gruyter

This work is licensed under the Creative Commons Attribution 4.0 International License.

Articles in the same Issue

  1. Special Issue on Future Directions of Further Developments in Mathematics
  2. What will the mathematics of tomorrow look like?
  3. On H 2-solutions for a Camassa-Holm type equation
  4. Classical solutions to Cauchy problems for parabolic–elliptic systems of Keller-Segel type
  5. Control of multi-agent systems: Results, open problems, and applications
  6. Logical perspectives on the foundations of probability
  7. Subharmonic solutions for a class of predator-prey models with degenerate weights in periodic environments
  8. A non-smooth Brezis-Oswald uniqueness result
  9. Luenberger compensator theory for heat-Kelvin-Voigt-damped-structure interaction models with interface/boundary feedback controls
  10. Special Issue on Fractional Problems with Variable-Order or Variable Exponents (Part II)
  11. Positive solution for a nonlocal problem with strong singular nonlinearity
  12. Analysis of solutions for the fractional differential equation with Hadamard-type
  13. Hilfer proportional nonlocal fractional integro-multipoint boundary value problems
  14. A comprehensive review on fractional-order optimal control problem and its solution
  15. The θ-derivative as unifying framework of a class of derivatives
  16. Review Articles
  17. On the use of L-functionals in regression models
  18. Minimal-time problems for linear control systems on homogeneous spaces of low-dimensional solvable nonnilpotent Lie groups
  19. Regular Articles
  20. Existence and multiplicity of solutions for a new p(x)-Kirchhoff problem with variable exponents
  21. An extension of the Hermite-Hadamard inequality for a power of a convex function
  22. Existence and multiplicity of solutions for a fourth-order differential system with instantaneous and non-instantaneous impulses
  23. Relay fusion frames in Banach spaces
  24. Refined ratio monotonicity of the coordinator polynomials of the root lattice of type Bn
  25. On the uniqueness of limit cycles for generalized Liénard systems
  26. A derivative-Hilbert operator acting on Dirichlet spaces
  27. Scheduling equal-length jobs with arbitrary sizes on uniform parallel batch machines
  28. Solutions to a modified gauged Schrödinger equation with Choquard type nonlinearity
  29. A symbolic approach to multiple Hurwitz zeta values at non-positive integers
  30. Some results on the value distribution of differential polynomials
  31. Lucas non-Wieferich primes in arithmetic progressions and the abc conjecture
  32. Scattering properties of Sturm-Liouville equations with sign-alternating weight and transmission condition at turning point
  33. Some results for a p(x)-Kirchhoff type variation-inequality problems in non-divergence form
  34. Homotopy cartesian squares in extriangulated categories
  35. A unified perspective on some autocorrelation measures in different fields: A note
  36. Total Roman domination on the digraphs
  37. Well-posedness for bilevel vector equilibrium problems with variable domination structures
  38. Binet's second formula, Hermite's generalization, and two related identities
  39. Non-solid cone b-metric spaces over Banach algebras and fixed point results of contractions with vector-valued coefficients
  40. Multidimensional sampling-Kantorovich operators in BV-spaces
  41. A self-adaptive inertial extragradient method for a class of split pseudomonotone variational inequality problems
  42. Convergence properties for coordinatewise asymptotically negatively associated random vectors in Hilbert space
  43. Relating the super domination and 2-domination numbers in cactus graphs
  44. Compatibility of the method of brackets with classical integration rules
  45. On the inverse Collatz-Sinogowitz irregularity problem
  46. Positive solutions for boundary value problems of a class of second-order differential equation system
  47. Global analysis and control for a vector-borne epidemic model with multi-edge infection on complex networks
  48. Nonexistence of global solutions to Klein-Gordon equations with variable coefficients power-type nonlinearities
  49. On 2r-ideals in commutative rings with zero-divisors
  50. A comparison of some confidence intervals for a binomial proportion based on a shrinkage estimator
  51. The construction of nuclei for normal constituents of Bπ-characters
  52. Weak solution of non-Newtonian polytropic variational inequality in fresh agricultural product supply chain problem
  53. Mean square exponential stability of stochastic function differential equations in the G-framework
  54. Commutators of Hardy-Littlewood operators on p-adic function spaces with variable exponents
  55. Solitons for the coupled matrix nonlinear Schrödinger-type equations and the related Schrödinger flow
  56. The dual index and dual core generalized inverse
  57. Study on Birkhoff orthogonality and symmetry of matrix operators
  58. Uniqueness theorems of the Hahn difference operator of entire function with a Picard exceptional value
  59. Estimates for certain class of rough generalized Marcinkiewicz functions along submanifolds
  60. On semigroups of transformations that preserve a double direction equivalence
  61. Positive solutions for discrete Minkowski curvature systems of the Lane-Emden type
  62. A multigrid discretization scheme based on the shifted inverse iteration for the Steklov eigenvalue problem in inverse scattering
  63. Existence and nonexistence of solutions for elliptic problems with multiple critical exponents
  64. Interpolation inequalities in generalized Orlicz-Sobolev spaces and applications
  65. General Randić indices of a graph and its line graph
  66. On functional reproducing kernels
  67. On the Waring-Goldbach problem for two squares and four cubes
  68. Singular moduli of rth Roots of modular functions
  69. Classification of self-adjoint domains of odd-order differential operators with matrix theory
  70. On the convergence, stability and data dependence results of the JK iteration process in Banach spaces
  71. Hardy spaces associated with some anisotropic mixed-norm Herz spaces and their applications
  72. Remarks on hyponormal Toeplitz operators with nonharmonic symbols
  73. Complete decomposition of the generalized quaternion groups
  74. Injective and coherent endomorphism rings relative to some matrices
  75. Finite spectrum of fourth-order boundary value problems with boundary and transmission conditions dependent on the spectral parameter
  76. Continued fractions related to a group of linear fractional transformations
  77. Multiplicity of solutions for a class of critical Schrödinger-Poisson systems on the Heisenberg group
  78. Approximate controllability for a stochastic elastic system with structural damping and infinite delay
  79. On extremal cacti with respect to the first degree-based entropy
  80. Compression with wildcards: All exact or all minimal hitting sets
  81. Existence and multiplicity of solutions for a class of p-Kirchhoff-type equation RN
  82. Geometric classifications of k-almost Ricci solitons admitting paracontact metrices
  83. Positive periodic solutions for discrete time-delay hematopoiesis model with impulses
  84. On Hermite-Hadamard-type inequalities for systems of partial differential inequalities in the plane
  85. Existence of solutions for semilinear retarded equations with non-instantaneous impulses, non-local conditions, and infinite delay
  86. On the quadratic residues and their distribution properties
  87. On average theta functions of certain quadratic forms as sums of Eisenstein series
  88. Connected component of positive solutions for one-dimensional p-Laplacian problem with a singular weight
  89. Some identities of degenerate harmonic and degenerate hyperharmonic numbers arising from umbral calculus
  90. Mean ergodic theorems for a sequence of nonexpansive mappings in complete CAT(0) spaces and its applications
  91. On some spaces via topological ideals
  92. Linear maps preserving equivalence or asymptotic equivalence on Banach space
  93. Well-posedness and stability analysis for Timoshenko beam system with Coleman-Gurtin's and Gurtin-Pipkin's thermal laws
  94. On a class of stochastic differential equations driven by the generalized stochastic mixed variational inequalities
  95. Entire solutions of two certain Fermat-type ordinary differential equations
  96. Generalized Lie n-derivations on arbitrary triangular algebras
  97. Markov decision processes approximation with coupled dynamics via Markov deterministic control systems
  98. Notes on pseudodifferential operators commutators and Lipschitz functions
  99. On Graham partitions twisted by the Legendre symbol
  100. Strong limit of processes constructed from a renewal process
  101. Construction of analytical solutions to systems of two stochastic differential equations
  102. Two-distance vertex-distinguishing index of sparse graphs
  103. Regularity and abundance on semigroups of partial transformations with invariant set
  104. Liouville theorems for Kirchhoff-type parabolic equations and system on the Heisenberg group
  105. Spin(8,C)-Higgs pairs over a compact Riemann surface
  106. Properties of locally semi-compact Ir-topological groups
  107. Transcendental entire solutions of several complex product-type nonlinear partial differential equations in ℂ2
  108. Ordering stability of Nash equilibria for a class of differential games
  109. A new reverse half-discrete Hilbert-type inequality with one partial sum involving one derivative function of higher order
  110. About a dubious proof of a correct result about closed Newton Cotes error formulas
  111. Ricci ϕ-invariance on almost cosymplectic three-manifolds
  112. Schur-power convexity of integral mean for convex functions on the coordinates
  113. A characterization of a ∼ admissible congruence on a weakly type B semigroup
  114. On Bohr's inequality for special subclasses of stable starlike harmonic mappings
  115. Properties of meromorphic solutions of first-order differential-difference equations
  116. A double-phase eigenvalue problem with large exponents
  117. On the number of perfect matchings in random polygonal chains
  118. Evolutoids and pedaloids of frontals on timelike surfaces
  119. A series expansion of a logarithmic expression and a decreasing property of the ratio of two logarithmic expressions containing cosine
  120. The 𝔪-WG° inverse in the Minkowski space
  121. Stability result for Lord Shulman swelling porous thermo-elastic soils with distributed delay term
  122. Approximate solvability method for nonlocal impulsive evolution equation
  123. Construction of a functional by a given second-order Ito stochastic equation
  124. Global well-posedness of initial-boundary value problem of fifth-order KdV equation posed on finite interval
  125. On pomonoid of partial transformations of a poset
  126. New fractional integral inequalities via Euler's beta function
  127. An efficient Legendre-Galerkin approximation for the fourth-order equation with singular potential and SSP boundary condition
  128. Eigenfunctions in Finsler Gaussian solitons
  129. On a blow-up criterion for solution of 3D fractional Navier-Stokes-Coriolis equations in Lei-Lin-Gevrey spaces
  130. Some estimates for commutators of sharp maximal function on the p-adic Lebesgue spaces
  131. A preconditioned iterative method for coupled fractional partial differential equation in European option pricing
  132. A digital Jordan surface theorem with respect to a graph connectedness
  133. A quasi-boundary value regularization method for the spherically symmetric backward heat conduction problem
  134. The structure fault tolerance of burnt pancake networks
  135. Average value of the divisor class numbers of real cubic function fields
  136. Uniqueness of exponential polynomials
  137. An application of Hayashi's inequality in numerical integration
Downloaded on 11.10.2025 from https://www.degruyterbrill.com/document/doi/10.1515/math-2022-0589/html
Scroll to top button