0% found this document useful (0 votes)
12 views24 pages

Approximate controllability for linearized compressible barotropic Navier–Stokes system in one and two dimensions

Uploaded by

ghoshaditya949
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
12 views24 pages

Approximate controllability for linearized compressible barotropic Navier–Stokes system in one and two dimensions

Uploaded by

ghoshaditya949
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 24

J. Math. Anal. Appl.

422 (2015) 1034–1057

Contents lists available at ScienceDirect

Journal of Mathematical Analysis and Applications


www.elsevier.com/locate/jmaa

Approximate controllability for linearized compressible barotropic


Navier–Stokes system in one and two dimensions
Shirshendu Chowdhury
T.I.F.R. Centre for Applicable Mathematics, Post Bag No. 6503, GKVK Post Office, Bangalore-560065,
India

a r t i c l e i n f o a b s t r a c t

Article history: In this paper we consider compressible barotropic Navier–Stokes equations in


Received 31 May 2013 one and two dimensions linearized around a constant steady state with Dirichlet
Available online 16 September 2014 boundary conditions. We explore the controllability of this linearized system using
Submitted by P.G. Lemarie-Rieusset
a control only for the velocity equation. We prove that the system with homogeneous
Keywords: Dirichlet boundary conditions, is approximately controllable by a localized interior
Linearized compressible barotropic control when time is sufficiently large.
Navier–Stokes equations © 2014 Elsevier Inc. All rights reserved.
Localized interior control
Approximate controllability

1. Introduction and main results

The Navier–Stokes equations for a viscous compressible barotropic fluid in Ω ⊂ RN is


∂ρ   ⎪
(x, t) + div ρ(x, t)v(x, t) = 0, ⎪



 ∂t 
∂v     (1.1)
ρ(x, t) (x, t) + v(x, t).∇ v(x, t) = −∇p(x, t) + μΔv(x, t) + (λ + μ)∇ div v(x, t) , ⎪

∂t ⎪


p(x, t) = aρ (x, t), t > 0, x ∈ Ω,
γ

where ρ(x, t) is the density of the fluid, v(x, t) = (v1 (x, t), . . . , vN (x, t)) denotes the velocity vector in RN
and p(x, t) denotes the pressure. Note that the second equation of (1.1) component wise is

∂vi ∂p ∂
ρ + v.∇vi =− + μΔvi + (λ + μ) [div v], i = 1, 2, . . . , N.
∂t ∂xi ∂xi

E-mail address: [email protected].

http://dx.doi.org/10.1016/j.jmaa.2014.09.011
0022-247X/© 2014 Elsevier Inc. All rights reserved.
S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057 1035

Throughout this paper, we follow this same notational convention and use bold script to denote vectors
and product spaces. The viscosity coefficients μ, λ are assumed to be constant satisfying the following
thermodynamic restrictions: μ > 0, λ + μ  0 and the constants a > 0, γ ≥ 1.
The Navier–Stokes equations for a viscous compressible barotropic fluid in an interval of R become

ρt (x, t) + (ρv)x (x, t) = 0,


   
ρ(x, t) vt (x, t) + v(x, t)vx (x, t) + p(ρ) x (x, t) − νvxx (x, t) = 0, (1.2)
p(ρ) = aργ for a > 0, γ ≥ 1, (1.3)

where ρ, v, p, ν > 0 denote the density, velocity, pressure and viscosity of the fluid as earlier.
To linearize around (Q0 , V0 ) with Q0 > 0, V0 ≥ 0, we search for solution (ρ(x, t), v(x, t)) of (1.2)–(1.3) in
the following way:

ρ(x, t) = σ(x, t) + Q0 , v(x, t) = u(x, t) + V0 (1.4)

and collect the first order terms in σ and u. In this way we obtain the following coupled linearized system
in (σ, u) around (Q0 , V0 )

σt (x, t) + V0 σx (x, t) + Q0 ux (x, t) = 0,


ν
ut (x, t) − uxx (x, t) + V0 ux (x, t) + aγQγ−20 σx (x, t) = 0. (1.5)
Q0

In this paper we consider first the following compressible Navier–Stokes system in IL := (0, L), linearized
around the constant steady state (Q0 , V0 ), with Q0 > 0, V0 > 0,

σt (x, t) + V0 σx (x, t) + Q0 ux (x, t) = 0,


ν
ut (x, t) − uxx (x, t) + V0 ux (x, t) + aγQγ−2
0 σx (x, t) = f χO , (1.6)
Q0

where χO is the characteristic function of an open subset O ⊂ IL .


Initial conditions are:

σ(x, 0) = σ0 (x), u(x, 0) = u0 (x) ∀x ∈ IL . (1.7)

Dirichlet boundary conditions are:

σ(0, t) = u(0, t) = u(L, t) = 0 ∀t ∈ (0, T ). (1.8)

Let Z = L2 (IL ) × L2 (IL ) be endowed with the inner product

  L L
σ ρ
, := aγQγ−2
0 σ(x)ρ(x) dx + Q0 u(x)v(x)dx.
u v Z
0 0

It is known that system (1.6)–(1.8) is well-posed in Z and details are given in Section 2.1.

Definition 1.1. The system (1.6)–(1.8) is approximately controllable in Z at time T > 0, if, for any U0 =
(σ0 , u0 )T ∈ Z and any other UT = (σT , uT )T ∈ Z and any ε > 0, there exists a control f ∈ L2 (0, T ; L2 (O))
such that the solution to system (1.6)–(1.8) satisfies
 
U(T ) − UT  ≤ ε.
Z
1036 S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057

Our first main result concerning approximate controllability of (1.6)–(1.8) is the following:

Theorem 1.2. Let O = (0, l), 0 < l < L. Our control system (1.6)–(1.8) is approximately controllable in the
space Z = L2 (IL ) × L2 (IL ) at time T , by a localized interior control f ∈ L2 (0, T ; L2 (O)) for the velocity, if
T > (L−l)
V0 .

Next we study the linearized system in two dimensions around the constant steady state solution (Q, V) =
(Q, V1 , V2 ), Q > 0, V1 > 0, V2 ≥ 0 of (1.1) in Ω × (0, T )

∂σ  
(x, t) + V.∇σ(x, t) + Q div u(x, t) = 0, (1.9)
∂t
∂u μ (λ + μ)  
(x, t) − Δu(x, t) − ∇ div u(x, t) + (V.∇)u(x, t) + aγQγ−2 ∇σ(x, t) = f χO (1.10)
∂t Q Q

where χO is the characteristic function of an open subset O ⊂ Ω. We consider the system (1.9)–(1.10) in
 
Ω = x = (x1 , x2 ) ∈ R2 : 0 < x1 < L, 0 < x2 < H

with boundary ∂Ω, having three disjoint portions


     
Γin = x ∈ ∂Ω : V.n(x) < 0 , Γ0 = x ∈ ∂Ω : V.n(x) = 0 , Γout = x ∈ ∂Ω : V.n(x) > 0

where n(x) is the unit outward normal to ∂Ω where it is defined.


Note that for V1 > 0, V2 = 0

Γin = {0} × (0, H), Γ0 = (0, L) × {0, H}, Γout = {L} × (0, H)

and for V1 > 0, V2 > 0


       
Γin = {0} × (0, H) ∪ (0, L) × {0} , Γout = {L} × (0, H) ∪ (0, L) × {H}

and Γ0 is an empty set. Let us denote

ΩT = Ω × (0, T ); ΣT = ∂Ω × (0, T ).

The initial and boundary conditions are

σ(x, 0) = σ0 (x), u(x, 0) = u0 (x) in Ω, (1.11)


σ(x, t) = 0 on Γin × (0, T ), u(x, t) = 0 on ΣT , (1.12)

where

σ0 ∈ L2 (Ω), u0 ∈ L2 (Ω). (1.13)

Let Z = L2 (Ω) × L2 (Ω) be endowed with the inner product


   
σ ρ γ−2
, := aγQ σ(x)ρ(x) dx + Q u(x).v(x)dx
u v Z
Ω Ω

where u = (u1 , u2 ), v = (v1 , v2 ) and u.v = u1 v1 + u2 v2 .


S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057 1037

We obtain the following approximate controllability results for the system (1.9)–(1.12).

Theorem 1.3 (V1 > 0, V2 > 0). Let O = (0, l) × (0, h), 0 < l < L, 0 < h < H. Our control system
(1.9)–(1.12) is approximately controllable in the space Z = L2 (Ω) × L2 (Ω) at time T , by a localized interior
control f ∈ L2 (0, T ; L2 (O)) for the velocity, if T > max{ (L−l)
V1 ,
(H−h)
V2 }.

Theorem 1.4 (V1 > 0, V2 = 0). Let O = (0, l) × (0, H), 0 < l < L. Our control system (1.9)–(1.12)
is approximately controllable in the space Z = L2 (Ω) × L2 (Ω) at time T, by a localized interior control
f ∈ L2 (0, T ; L2 (O)) for the velocity, if T > (L−l)
V1 .

The proof of these approximate controllability results rely on a unique continuation property for the
solutions of the adjoint system. This property will be established in Sections 2 and 3 in one and two
dimensions respectively as a consequence of the classical Holmgren’s uniqueness theorem. The proofs of
Theorems 1.2, 1.3, 1.4 will be in the same spirit as the proof of approximate controllability for wave
equation in the book of Lions [48, Chapter 1, Section 8] and the proof of approximate controllability for
linear system of thermoelasticity by Zuazua [62].
In our system, the interesting feature is the coupling between first order transport equation of hyperbolic
type and second order linearized momentum equation of parabolic type. This leads to some difficulties
particularly in regularity questions. In fact it is very exciting to find out after interaction of these two
different types of equations in the coupled system, which type is dominating the other.
We now mention some controllability results for other models involving both transport and parabolic
effects. Thermoelasticity system, viscoelastic fluid model are some examples of this nature. Zuazua stud-
ied in [62] controllability of a system of linear thermoelasticity. This system consists of a wave equation
coupled with a heat equation. He proved in [62] exact-approximate controllability of the linear system of
thermoelasticity (exact controllability of the displacement and the approximate controllability of the tem-
perature) when the control time is large enough and control acts in the equations of displacement and it
is supported in a neighborhood of the boundary of the thermoelastic body. Teresa and Zuazua in [16] also
proved exact-approximate controllability for linear system of thermoelastic plates in a similar way. Lebeau
and Zuazua in [47] proved that, if the control time and the support of the control satisfy the geometric
control condition for the wave equation and control acts either only hyperbolic equation or only parabolic
equation, then system of thermo elasticity is null-controllable. Fernández-Cara and Teresa in [25] established
the null controllability of a cascade linear system formed by a heat and a wave equation where the localized
interior control acts only on the heat equation. Albano and Tataru in [1] studied boundary observability
for a coupled system of parabolic–hyperbolic type(example:Thermoelasticity). They proved some Carleman
estimates with singular weights for the heat and for the wave equations and combined them to obtain an
observability result for the system. Crépeau and Prieur proved in [15] approximate controllability for a
system coupling a reaction–diffusion system and an ordinary differential equation constructing a control
explicitly. They used a flatness-like property, indeed, the solution is expressed in terms of an infinite series
depending on a flat output, its derivatives and its integrals.
Viscoelastic flows [56] gives an interesting coupled fluid model. Controllability of linear viscoelastic flows
in one dimension using interior control is studied by Renardy [54,55]. He obtained exact controllability result
for Maxwell (hyperbolic nature) fluid using the Hilbert uniqueness method and approximate controllability
for Jeffreys model (parabolic + hyperbolic mixed nature); proving a unique continuation result. Note that
this unique continuation property cannot be obtained from Holmgren’s theorem, since the lines x = const.
are characteristic. Controllability of incompressible linear viscoelastic fluids of the Maxwell and Jeffreys
kinds is studied by Boldrini, Doubova, Fernández-Cara and González-Burgos in two and three dimensions
in [5,18,17]. Recently they established the large time exact controllability for incompressible linear Maxwell
fluids in [5]. For incompressible linear Jeffreys fluids in two or three dimensions only partial approximate
1038 S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057

controllability results are known and it is recently proved by Doubova and Fernández-Cara in [17]. Approx-
imate controllability and null controllability for this system are very interesting open questions.
Controllability problems for incompressible fluids have been extensively studied in the recent years.
There are several results concerning the controllability properties of the constant density incompressible
Navier–Stokes equations by Coron, Fursikov, Imanuilov, Fernández-Cara, Guerrero, Puel, Gonzales-Burgos,
Fabre in [12–14,31,32,45,26,6,39,37,20] or for the Boussinesq system in [36,27,28]. They concern approximate
controllability, null controllability, exact controllability to the trajectories using boundary or distributed
control and also controlling the system with a reduced number of controls.
Fursikov and Imanuvilov proved in [31, Chapter III] (see also [32]) the local exact controllability of the
2-D Navier–Stokes system in a bounded domain around regular trajectories for boundary conditions on the
normal velocity and on the curl. This local result was extended by Imanuvilov in [44,45] to the case of
Dirichlet boundary conditions for velocity in two and three dimensions. Later on in [26] Fernández-Cara,
Guerrero, Imanuvilov and Puel improved it by proving the result with less regularity on the target trajectory.
González-Burgos, Guerrero and Puel in [36] proved local exact controllability to the trajectories of the
Boussinesq system by adding a fictitious control on the divergence equation and then using a lifting argument
to absorb it.
Coron and Guerrero established in [14] the local null controllability of two-dimensional Navier–Stokes
system in a torus with internal controls having one vanishing component. Carreño and Guerrero proved in
[6] local null controllability of the N -dimensional Navier–Stokes system in bounded domain with internal
controls having one vanishing component. Fernández-Cara, Guerrero, Imanuvilov and Puel in [27] established
the local exact controllability to the trajectories for the Navier–Stokes and Boussinesq systems with N − 1
scalar controls.
According to our knowledge, there is no general global controllability result for incompressible Navier–
Stokes equations, partial global results are only known for some particular systems. Coron proved in [12]
global approximate controllability for the 2-D Navier–Stokes system with Navier-slip boundary conditions.
This result was generalized by Coron and Fursikov in [13] and they obtained the global exact controllability
for the Navier–Stokes system on a 2-D manifold without boundary. A similar result for the Boussinesq
system on torus was given by Fursikov and Imanuilov in [28]. Fabre in [21,20] proved approximate control-
lability of a variant of the Navier–Stokes equations with truncated quadratic nonlinearity using distributed
controls. Guerrero, Imanuvilov and Puel proved a result concerning the global approximate controllability
for the 2-D and 3-D Navier–Stokes system in [38] and [39] by means of boundary controls. Note that, it is not
a classical result of global approximate controllability as they need to introduce a sequence of approximate
right-hand sides in [39] and [38].
However for variable density incompressible Navier–Stokes equations, the control results are more difficult
to obtain and only a few partial results are known. Fernández-Cara studies the control of non-homogeneous
(or variable density) incompressible Navier–Stokes equations of a fluid in [24]. He consider in [24] first
optimal control of variable density Navier–Stokes fluids with several different cost functionals. He proves
existence, uniqueness of optimal control and characterizes it through optimality system. He also proposes
some iterative schemes to compute the solutions of optimal control problems. Then he discuss some control-
lability questions. He proves in [24] local null-controllability of the velocity only, using interior control. He
also studies null controllability problem for the variable density Boussinesq system. Very recently, Badra,
Ervedoza and Guerrero in [3] prove local exact controllability to trajectories (controllability in both σ
and u) for non-homogeneous (variable density) incompressible 2-D Navier–Stokes equations using boundary
controls for both σ and u. Their strategy is close to the one in by Ervedoza et al. [19].
Also relatively less control results are known in the case of compressible Navier–Stokes equation. Amosova
in [2] considers compressible viscous fluid in one dimension in Lagrangian coordinates, with zero boundary
condition for the velocity on the boundaries of the interval (0, 1) and an interior control on the velocity
S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057 1039

equation. She proves in [2] local exact controllability to trajectories for the velocity, provided that the initial
density is already on the “targeted trajectory” (with initial condition in H 1 (0, 1) × H01 (0, 1)).
Ervedoza, Glass, Guerrero and Puel consider the compressible Navier–Stokes equation in one space
dimension in a bounded domain (0, L) in [19]. They prove in [19] local exact controllability to constant
states (ρ̄, v̄) with ρ̄ > 0, v̄ = 0 using two boundary controls (both for density and velocity) when initial
conditions are regular (namely both initial density and velocity lies in H 3 (0, L)).
We consider in [11] linearized (around (Q0 , 0), i.e. V0 = 0) compressible Navier–Stokes system in one
dimension and study the controllability of this system in the interval (0, π). We establish in [11] that for any
T > 0, the linearized system is null controllable in time T by an interior control f ∈ L2 (0, T ; L2 (0, π)) acting
everywhere in the velocity equation if and only if initial data (σ0 , u0 ) ∈ (H 1 (0, π) ∩ L̇2 (0, π)) × L2 (0, π)
where L̇2 (0, π) denotes the space of functions in L2 (0, π) with mean value 0. We show in [11] that linearized
coupled system in (σ, u) is not null controllable in L2 (0, π) × L2 (0, π) using any L2 boundary control(or
localized interior control) in the velocity equation. In fact we prove, for any T > 0, no nontrivial finite linear
combination of eigenvectors of the linearized operator can be driven to rest using L2 boundary control for
the velocity. Thus our result is sharp in the sense that the null controllability cannot be achieved by a
localized interior control or by a boundary control. On the other hand, we obtain in [11] that the linearized
system is approximately controllable in L̇2 (0, π) × L2 (0, π) for any time T > 0 using boundary control for
velocity.
t
Guerrero and Imanuvilov prove in [40] that heat equation with the memory term 0 uxx (x, s)ds [or
t
0
u(x, s)ds] is not null controllable in L2 (0, 1) using L2 boundary control(or localized interior control).
Then in [40] they generalize this negative result in any dimensions (initial condition lies in L2 (Ω) and
boundary control lies in the class L2 ((0, T ) × ∂Ω)). Lack of null controllability of the heat equation with
memory in one dimension using boundary control is also noticed by Halanay and Pandolfi in [41, Theorem 2,
Remark 1]. Ivanov and Pandolfi prove in [46] that the one-dimensional heat equation with memory cannot
be ‘controlled to rest’ (means we can find a time T > 0 and a control f (t) such that the control f (t) and
corresponding solution is zero for all t > T ) for large classes of memory kernels and controls.
Note that our system can be viewed as an integro-differential equation. Namely using initial and boundary
conditions for σ:

σ(x, 0) = σ0 (x) ∀x ∈ IL , σ(0, t) = 0 ∀t ∈ (0, T ), (1.14)

we get from the first equation of (1.6)

t
 
σ(x, t) = σ0 (x − V0 t) − Q0 ux V0 (s − t) + x, s ds, for V0 t < x, (1.15)
0

t
 
σ(x, t) = −Q0 ux V0 (s − t) + x, s ds, for V0 t > x, (1.16)
t− Vx
0

t
   
σ(x, t) = −Q0 ux V0 (s − t) + x, s ds for V0 t = x taking σ0 (0) = 0 . (1.17)
0

Then calculating σx (x, t) and plugging it into the 2nd equation of (1.6), we will obtain a single integro-
differential equation for u with homogeneous Dirichlet boundary conditions. Observe that (σ0 )x (x − V0 t)
will be also present as source term when V0 t < x.
1040 S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057

Work of Guerrero–Imanuvilov in one dimension is close to our linearized Navier–Stokes system when

V0 = 0.

In fact if we take σ0 = 0, then linearized Navier–Stokes system around (Q0 , 0) can be written (from first
equation solving σ and then replacing σ in velocity equation) as heat equation with the memory term
t
aγQγ−1
0 u (x, s)ds. Notice that for V0 = 0, we already studied the controllability of the linearized
0 xx
system in our previous work [11]. For any T > 0, approximate controllability of heat equation with memory
in L2 (Ω) is proved by Barbu and Innelli in [4] using L2 boundary or interior control in higher dimensions.
Tao, Gao and Zhang also show in [59], approximate controllability of heat equation with memory (taking
kernel equal to 1) for any T > 0 using boundary control in one dimension. So in this way(reducing and
viewing as the heat equation in velocity with memory term) for V0 = 0 case, we can get system (1.5) is
approximately controllable in only u (not in σ) using boundary control in velocity for any T > 0. But note
that in [11] for V0 = 0 case, we already showed that whole coupled system (1.5) is approximately controllable
(both in σ, u) using boundary control in velocity for any T > 0.
Doubova and Fernández-Cara in [17] study the control of incompressible viscoelastic linear Jeffreys fluids
in two or three dimensions. They establish partial approximate-finite dimensional controllability in velocity
y only (not for stress τ ) in any time T > 0, with distributed or boundary controls. Since stress τ in [17] is
governed by a differential equation where only its partial derivatives with respect to t is present (a family
of ODEs), this linear Jeffreys fluid model is also close to our linearized compressible Navier–Stokes system
around (Q0 , 0) or (Q, 0, 0) i.e. when

V0 = 0 or V = (0, 0).

Using the expression of stress τ , they deduce an integro-differential equation for velocity y with pressure.
Then they consider adjoint problem corresponding to this reduced integro-differential equation not for the
whole system and they got unique continuation only for this reduced integro-differential equation using
Laplace transform approach. Thus they got approximate controllability in velocity y. Unique continuation
for adjoint of the whole system is an open question in [17]. Their Laplace transform approach reducing in to
a integro-differential equation of y will not work due to the presence of nonzero source term corresponding
to stress (see Section 3.1 and Lemma 2.3 in [17]).
If for V0 > 0, we reduce our linearized Navier–Stokes system in to a single equation of u with integral
memory term, then approach of Doubova and Fernández-Cara in [17] (if it works) can at most give only
partial approximate controllability means approximate controllability only for velocity u (not for density σ)
but at any time T > 0. This approach is not that much satisfactory since it will not provide approximate
controllability in σ. However our approach in this paper gives the approximate controllability in both
components (σ, u) for large time. So it indicates need of large time (unless control acts everywhere) for
controlling the transport part and controllability in both components (σ, u) of our coupled system are
connected.
Martin, Rosier and Rouchon in [51] consider the wave equation with structural damping in one dimension
with periodic boundary conditions. Using spectral analysis and the method of moments, they obtain that this
system is null controllable with a moving distributed control for regular initial conditions (in H s+2 × H s ,
s > 15 2 ) in sufficiently large time. The one dimensional compressible Navier–Stokes equation linearized
around a constant steady state (Q0 , V0 ), Q0 > 0, V0 > 0 with periodic boundary conditions is closely related
to the structurally damped wave equation studied by Martin et al. in [51]. Chowdhury and Mitra consider the
controllability of this linearized system around (Q0 , V0 ), Q0 > 0, V0 > 0 with periodic boundary conditions,
using a control only for the velocity equation. Then, following the approach of Martin et al. in [51] using
the method of moments, they establish in [8] that the linearized system with periodic boundary conditions
S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057 1041

is null controllable by a localized interior control when time is large enough, and for regular initial data
s+1
(in Ḣper (0, 2π) × Hper
s
(0, 2π) with s ≥ 7 where Ḣper
s
(0, 2π) denotes the Sobolev space of periodic function
with mean zero). Is the system null controllable in large time for less regular initial data? Very recently
Chowdhury, Mitra, Ramaswamy and Renardy in [9] proved null controllability of this linearized system for
1
initial conditions in Ḣper (0, 2π) × L2 (0, 2π) using only one interior control for velocity when T > 2πV0 . They
also established that result is sharp by showing that the null controllability cannot be achieved in subspace
s
Ḣper (0, 2π) × L2 (0, 2π), 0 ≤ s < 1 by a L2 localized interior control for velocity. Moreover they showed null
controllability in L2 (0, 2π) × L2 (0, 2π) can be obtained if we use two localized L2 interior controls both for
density and velocity when T > 2π V0 . The proofs of these results in [9] rely on an observability inequality for
the solutions of the adjoint system, the spectral analysis of the linearized operator and use of two different
types of Ingham inequality for complex frequencies.
Chaves-Silva, Rosier and Zuazua in [7] consider the wave equation with both viscous Kelvin–Voigt and
frictional damping as a model of viscoelasticity in RN . The structure of the system they consider is also,
in some sense, similar to the linearized compressible Navier–Stokes equations in higher dimensions. They
established null controllability of the system with a moving internal control using the observability of the
adjoint system and some Carleman estimates.
For V0 = 0, with Dirichlet boundary condition the resolvent of the corresponding linearized operator is a
compact operator in Z (see Proposition IV.13 in Chapter 4, [35]). This is no longer true if V0 = 0. Moreover,
if V0 = 0, it is a sectorial operator in Z (see Lemma 2.5 in [11] for some details). Thus the properties of the
two semigroups (the one when V0 = 0 and the one when V0 = 0) corresponding to the linearized operators
are completely different. When V0 = 0, it is an analytic semigroup and the coupled system behaves like
a parabolic equation. When V0 > 0, it is just a strongly continuous semigroup and the coupled system
behaves like a hyperbolic equation. Recall that for V0 = 0 case in [11], spectrum of the linearized operator
has an accumulation point which is responsible for bad control properties. This kind of similar behavior
also observed by Rosier in [57] for boundary controllability of wave equation with structural damping,
Micu in [52] for boundary controllability of the linearized Benjamin–Bona–Mahony equation, Guerrero and
Imanuvilov in [40], Tao et al. in [59] for control of heat equation with the memory term, Renardy in [55],
Doubova and Fernández-Cara in [17], for the control of viscoelastic flows. They have also used Fourier series,
Laplace transform, moment method to study the controllability like us in [11]. For V0 > 0, deduction in to
a single equation for u will involve V0 ux term as well as some more complicated integral terms (see Eqs.
(1.15)–(1.17)) compared to V0 = 0. It is interesting to note that for V0 > 0, there is no accumulation point
in the spectrum of linearized operator with periodic boundary conditions and so in periodic setting better
control properties are obtained by Chowdhury et al. in [9].
To the best of author’s knowledge, any interior null controllability result for the system (1.6)–(1.8) with
homogeneous Dirichlet boundary conditions is not known. In this case since we are unable to use a Fourier
basis, method of moments does not work. Also method in [9] cannot be used. So in this context we stress
the importance of our Theorem 1.2, which answers at least approximately controllability in Z using only
one interior control on velocity for the system (1.6)–(1.8) with homogeneous Dirichlet boundary conditions.
Also we have generalized approximately controllability results (Theorems 1.3, 1.4) in two dimensions. To
the author’s knowledge, these results are totally new for compressible Navier–Stokes system.
Our previous paper [11] appears to be the first work concerning the controllability for linearized (around
(Q0 , 0)) compressible Navier–Stokes system in one dimension. The current paper gives more controllability
results when compressible Navier–Stokes system in one dimension is linearized around (Q0 , V0 ) with V0 > 0.
It also gives similar controllability results in two dimensions for the system linearized around (Q, V1, V2 )
with Q > 0, V1 > 0, V2 ≥ 0. The interesting point in our work is that only one interior control is needed
whereas two boundary controls are used by Ervedoza et al., in [19]. An optimal boundary control problem
for the two dimensional unsteady linearized compressible Navier–Stokes equations in a rectangle has been
studied by Chowdhury and Ramaswamy in [10] and optimality conditions are derived.
1042 S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057

This paper is organized as follows. In Section 2, we study the existence and uniqueness of the linearized
system (1.6)–(1.8) and prove the wellposedness of adjoint system of (1.6)–(1.8). Then we establish that the
adjoint system satisfies the unique continuation property and so approximate controllability (Theorem 1.2)
is proved. In Section 3, we study the linearized system in two dimensions and show adjoint system satisfies
unique continuation property and hence Theorems 1.3 and 1.4 follows. In Section 4, we make some additional
remarks and comments regarding the generalization of the results, our methods, some difficulties, role of
large time and simple geometry. We also mention some open questions and future directions of work.

2. Linearized system

2.1. Existence of unique solution for linearized system

Let us recall that Z = L2 (IL ) × L2 (IL ) and introduce the positive constants

ν
b := aγQγ−2
0 , ν0 := . (2.1)
Q0

Let Z be endowed with the inner product

  L L
σ ρ
, := b σ(x)ρ(x) dx + Q0 u(x)v(x)dx.
u v Z
0 0

We define
 
1
H{0} (IL ) = ϕ ∈ H 1 (IL ), ϕ(0) = 0 .

Similarly we define
 
1
H{L} (IL ) = ϕ ∈ H 1 (IL ), ϕ(L) = 0 .

We now define the unbounded operator (A, D(A)) in Z by


 
D(A) = H{0}
1
(IL ) × H 2 (IL ) ∩ H01 (IL )

and
   
−V0 dx
d
−Q0 dx
d
−V0 dx
d
−Q0 dxd
A= γ−2 d ν d2 = 2 . (2.2)
−aγQ0 dx Q0 dx2 − V0 dx
d
−b dx
d d
ν0 dx 2 − V0 dx
d

Setting U(t) = (σ(·, t), u(·, t))T the system (1.6)–(1.8) with f ∈ L2 (0, T ; L2 (IL )) can be written as

U (t) = AU(t) + Bf, U(0) = U0 ∈ Z (2.3)

where the control operator B ∈ L(L2 (IL ), Z) and defined by

0
B(f ) = . (2.4)
χO f

We now recall the following existence result. Details can be found in [35, Chapter IV, Section 5.2, Proposi-
tion IV.12].
S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057 1043

Lemma 2.1. The operator (A, D(A)) is maximal dissipative in Z. Thus, (A, D(A)) is the infinitesimal gen-
erator of a strongly continuous semigroup of contractions on Z, denoted by (S(t))t≥0 . For every U0 ∈ Z,
with f = 0 there is a unique solution U of (2.3) in C([0, ∞); Z) and
 
U(t) ≤ U0 Z for all t ≥ 0.
Z

Remark 2.2. The adjoint (A∗ , D(A∗ )) of the operator (A, D(A)) in Z is defined by
   
D A∗ = H{L}
1
(IL ) × H 2 (IL ) ∩ H01 (IL )

and
d d
ψ V0 dx Q0 dx ψ
A∗ = d d 2
d ,
φ b dx ν0 dx 2 + V0 dx φ

for (ψ, φ)T ∈ D(A∗ ). Moreover (A∗ , D(A∗ )) is the infinitesimal generator of a strongly continuous semigroup
of contractions (S ∗ (t))t≥0 on Z.

Using Lemma 2.1 and parabolic regularity of the 2nd equation in (1.6) we get the following existence
and uniqueness theorem. Details can be found, for example in [35] and [33].

Theorem 2.3. Given U0 = (σ0 , u0 )T ∈ Z, f ∈ L2 (0, T ; L2 (IL )) the system (1.6)–(1.8) has a unique weak so-
lution U(t) = (σ, u)T ∈ C([0, T ]; L2 (IL ))×[L2 (0, T ; H01 (IL ))∩C([0, T ]; L2 (IL ))] with ut ∈ L2 (0, T ; H −1 (IL )).

Remark 2.4. Since B ∈ L(L2 (IL ), Z), adjoint of B i.e. B ∗ ∈ L(Z, L2 (IL )). Using (2.4) we have

L
 ∗

Bf, ηZ = f, B η L2 (IL )
= Q0 f χO η2 dx, η∈Z
0

and hence

η1
B ∗ η = Q0 χO η2 , η= ∈ Z. (2.5)
η2

2.2. Adjoint problem

In order to study approximate controllability of the system (1.6)–(1.8), we first consider the adjoint
system of (1.6)–(1.8) in IL × (0, T )

−ψt (x, t) − V0 ψx (x, t) − Q0 φx (x, t) = 0,


−φt (x, t) − ν0 φxx (x, t) − V0 φx (x, t) − bψx (x, t) = 0, (2.6)

with following terminal and Dirichlet boundary conditions

ψ(x, T ) = ψT (x), φ(x, T ) = φT (x) ∀x ∈ IL , (2.7)


ψ(L, t) = φ(0, t) = φ(L, t) = 0 ∀t ∈ (0, T ). (2.8)

Reversing the time variable we can write the above system as the following initial and boundary value
problem:
1044 S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057

ψ̃t (x, t) − V0 ψ̃x (x, t) − Q0 φ̃x (x, t) = 0,


φ̃t (x, t) − ν0 φ̃xx (x, t) − V0 φ̃x (x, t) − bψ̃x (x, t) = 0, (2.9)

with initial and Dirichlet boundary conditions:

ψ̃(x, 0) = ψT (x), φ̃(x, 0) = φT (x) ∀x ∈ IL , (2.10)


ψ̃(L, t) = φ̃(0, t) = φ̃(L, t) = 0 ∀t ∈ (0, T ), (2.11)

where ψ̃(x, t) = ψ(x, T − t) and φ̃(x, t) = φ(x, T − t).


In view of Remark 2.2 using semigroup theory and parabolic regularity of the 2nd equation in (2.9) i.e.
as in Theorem 2.3, we get the following existence and uniqueness theorem for the adjoint system.

Theorem 2.5. Given VT = (ψT , φT )T ∈ Z the system (2.9)–(2.11) has a unique weak solution V(t) =
(ψ̃, φ̃)T = S ∗ (t)VT ∈ C([0, T ]; L2 (IL )) × [L2 (0, T ; H01 (IL )) ∩ C([0, T ]; L2 (IL ))] with φ̃t ∈ L2 (0, T ; H −1 (IL )).

Thus (2.6)–(2.8) has a unique weak solution (ψ, φ)T ∈ C([0, T ]; L2 (IL )) × [L2 (0, T ; H01 (IL )) ∩ C([0, T ];
L (IL ))] with φt ∈ L2 (0, T ; H −1 (IL )).
2

2.3. Unique continuation and approximate controllability

In this section we study the approximate controllability of the system (1.6)–(1.8). We recall the following
well known result on approximate controllability. Proof can be found for example in Zabczyk [61, Part IV,
Chapter 2, Section 2.3, Theorem 2.5].

Theorem 2.6. Let us fix T > 0. The following conditions are equivalent.

(i) The system (1.6)–(1.8) is approximately controllable in Z at time T .


(ii) If B ∗ S ∗ (t)VT = 0 for almost all t ∈ [0, T ], then VT = 0.

Remark 2.7. Using (2.5) and Theorem 2.5, we have

B ∗ S ∗ (t)VT = B ∗ V(t) = Q0 χO φ̃.

This helps us to rewrite (ii) in our case as follows.

Corollary 2.8. Thus, the system (1.6)–(1.8) is approximately controllable in Z at time T if and only if the
adjoint problem satisfies the following unique continuation property (UCP): “If (ψ̃, φ̃)T ∈ C([0, T ]; L2 (IL )) ×
[L2 (0, T ; H01 (IL )) ∩ C([0, T ]; L2 (IL ))] with φ̃t ∈ L2 (0, T ; H −1 (IL )) is the solution of (2.9)–(2.11) associated
to (ψT , φT )T ∈ Z and φ̃ = 0 in O × (0, T ), then (ψT , φT )T = (0, 0)T [i.e. (ψ̃, φ̃)T = (0, 0)T in IL × (0, T )].”

The main Theorem of this section is the following.

Theorem 2.9. Assume that T > L−l V0 . Let (ψ̃, φ̃) ∈ C([0, T ]; L (IL )) × [L (0, T ; H0 (IL )) ∩ C([0, T ]; L (IL ))]
T 2 2 1 2
−1
with φ̃t ∈ L (0, T ; H (IL )) be the solution of (2.9)–(2.11) such that φ̃ = 0 in O × (0, T ). Then (ψ̃, φ̃)T =
2

(0, 0)T in IL × (0, T ).

Hence Theorem 1.2 follows from Corollary 2.8 and Theorem 2.9.
To prove Theorem 2.9, let us introduce first a few notations and definitions. Let x1 , . . . , xN denote
independent variables in RN , D = (D1 , D2 , . . . , Dj , . . . , DN ), Dj = −i ∂x

j

(hence ∂x j
= iDj ) for j =
S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057 1045


1, 2, . . . , N and P (x, D) = |α|≤m aα (x)Dα be a differential operator with coefficients aα , where Dα =
αN
D1α1 . . . DN .

Definition 2.10. The principal symbol of P (x, D) = |α|≤m aα (x)Dα is the function

Pm (x, ξ) = aα (x)ξ α .
|α|=m

Definition 2.11. A C 1 surface S ⊂ RN with normal ξ at x is said to be characteristic at x for P (x, D) if


Pm (x, ξ) = 0. A surface which is characteristic at all points is called a characteristic surface.

Proof of unique continuation is based on the following consequence of Holmgren’s uniqueness Theorem
given by Hörmander in [43, Chapter VIII, Section 8.6, Theorem 8.6.8].

Theorem 2.12 (Hörmander). Let P = P (D) = |α|≤m aα Dα be a differential operator with constant coeffi-
cients aα , and let Ω1 and Ω2 be two convex open sets in RN with Ω1 ⊂ Ω2 . Then the following properties
are equivalent:

(i) (UCP) Every distribution Θ ∈ D (Ω2 ) satisfying P (D)Θ = 0 in Ω2 and vanishing in Ω1 must also
vanish on Ω2 .
(ii) Every characteristic hyperplane for P (D) which intersects Ω2 has to intersect Ω1 .

To apply the theorem in our set up for our system, we reduce our two equations in (2.9) into a single
equation for φ̃.

Lemma 2.13. Let (ψ̃, φ̃)T ∈ C([0, T ]; L2 (IL )) × [L2 (0, T ; H01 (IL )) ∩ C([0, T ]; L2 (IL ))] be the weak solution of
the system (2.9)–(2.11) for (ψT , φT )T ∈ Z, then φ̃ will satisfy
 
V0 ν0 φ̃xxx − ν0 φ̃xxt + φ̃tt − 2V0 φ̃tx + V02 − bQ0 φ̃xx = 0, (2.12)

in the sense of distribution.

Proof. Taking the distributional derivative of (2.9) w.r.to x we get

ψ̃tx − V0 ψ̃xx − Q0 φ̃xx = 0, (2.13)


φ̃tx − ν0 φ̃xxx − V0 φ̃xx − bψ̃xx = 0. (2.14)

Taking the distributional derivative of 2nd equation in (2.9) w.r.to t we get

φ̃tt − ν0 φ̃xxt − V0 φ̃xt − bψ̃xt = 0. (2.15)

Multiplying (2.14) by V0 and then subtracting from (2.15) we get

φ̃tt − ν0 φ̃xxt − V0 φ̃xt − b(ψ̃xt − V0 ψ̃xx ) − V0 φ̃tx + V0 ν0 φ̃xxx + V02 φ̃xx = 0.

Hence using (2.13), from above we get (2.12). 2

Lemma 2.14. Let


 
P (D)φ̃ = V0 ν0 φ̃xxx − ν0 φ̃xxt + φ̃tt − 2V0 φ̃tx + V02 − bQ0 φ̃xx
1046 S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057

Fig. 1. Picture of the set A.


where D = (−i ∂x , −i ∂t

). Then the characteristic hyperplanes for this P(D) are given by

x
t = C1 (constant), t+ = C2 (constant).
V0

Proof. Let Π = {(x, t) : ξx + ηt = C} be a hyperplane of R2 with normal vector (ξ, η) = (0, 0) where C is
a constant. Then by Definition 2.11, Π is a characteristic hyperplane for P (D) if

P3 (ξ, η) = V0 ν0 (iξ)3 − ν0 (iξ)2 (iη) = 0.

So we get either ξ = 0 or η = V0 ξ. Thus normal vector of the characteristic hyperplane Π are of the form

either (ξ, η) = constant(0, 1) or (ξ, η) = constant(1, V0 ).

Hence characteristic hyperplanes for P (D) are given by

x
t = C1 (constant), t+ = C2 (constant). 2
V0

Now we are going to prove Theorem 2.9 using Theorem 2.12 and Lemmas 2.13 and 2.14.

Proof of Theorem 2.9. Let (ψ̃, φ̃)T ∈ C([0, T ]; L2 (IL )) ×[L2 (0, T ; H01 (IL )) ∩C([0, T ]; L2 (IL ))] be the solution
of (2.9)–(2.11) such that φ̃ = 0 in O × (0, T ). Then using Lemma 2.13 we have in the distribution sense
 
P (D)φ̃ = V0 ν0 φ̃xxx − ν0 φ̃xxt + φ̃tt − 2V0 φ̃tx + V02 − bQ0 φ̃xx = 0.

By assumption we already have φ̃ = 0 in O × (0, T ), where O = (0, l).


Let T > L−l
V0 . Define (see Fig. 1)


A= (0, l + V0 s) × (0, T − s).
0≤s<T

To apply Theorem 2.12 of Hörmander, we define


 
Ω1 = (0, l) × (0, T ), Ω2 = A ∩ (0, L) × (0, T ) .

Then Ω1 ⊂ Ω2 . Note that if the characteristic line t = C1 or t + x


V0 = C2 hits Ω2 , then it hits Ω1 . Since φ̃
satisfies
S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057 1047

Fig. 2. Picture of the set A in this case.

P (D)φ̃ = 0 in D (Ω2 ) and φ̃ = 0 in Ω1 ,

using Theorem 2.12 we get

φ̃ = 0 in Ω2 . (2.16)

Using (2.16) and (2.9) we obtain

ψ̃x = ψ̃t = 0 in D (Ω2 ).

Thus ψ̃(x, t) = K(constant) in Ω2 . From the boundary conditions (2.11) we already have ψ̃(L, t) = 0 ∀t ∈
(0, T ) and hence

ψ̃ = 0 in Ω2 . (2.17)
(L−l)
As (0, L) × (0, T − V0 ) ⊂ Ω2 , from (2.16) and (2.17) we have in particular

(L − l)
ψ̃ = φ̃ = 0 in (0, L) × 0, T − .
V0

Hence by forward uniqueness we obtain that

ψ̃ = φ̃ = 0 in (0, L) × (0, T ).

This completes the proof of Theorem 2.9. 2

Remark 2.15. We may expect a similar following unique continuation result holds when O is any sub
interval of (0, L).
Let O = (l1 , l2 ), 0 ≤ l1 < l2 ≤ L, T > max{ Vl10 , L−l
V0 } and (ψ̃, φ̃) ∈ C([0, T ]; L (IL )) ×[L (0, T ; H0 (IL )) ∩
2 T 2 2 1
−1
C([0, T ]; L (IL ))] with φ̃t ∈ L (0, T ; H (IL )) be the solution of (2.9)–(2.11) such that φ̃ = 0 in O × (0, T ).
2 2

Then (ψ̃, φ̃)T = (0, 0)T in IL × (0, T ).


To prove this we try to proceed in the similar way as in the proof of Theorem 2.9. Let T > max{ Vl10 , L−l V0 }.
2

Define
 
A1 = (l1 , l2 + V0 s) × (0, T − s), A2 = (l1 − V0 s, l2 ) × (s, T ),
0≤s<T 0<s<T

and A = A1 ∪ A2 .
1048 S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057

To apply Theorem 2.12 of Hörmander, we define


 
Ω1 = (l1 , l2 ) × (0, T ), Ω2 = A ∩ (0, L) × (0, T ) .

Exactly like the proof of Theorem 2.9, using Fig. 2 of A in this case, we get

l1 (L − l2 )
ψ̃ = φ̃ = 0 in (0, L) × ,T − .
V0 V0

Then by forward uniqueness we obtain that

l1
ψ̃ = φ̃ = 0 in (0, L) × ,T .
V0

But to conclude ψ̃ = φ̃ = 0 in (0, L) × (0, T ) we need to prove



l1
ψ̃ = φ̃ = 0 in (0, L) × 0, . (2.18)
V0

This is true if we have a backward uniqueness result for our system. But we do not have any such result
yet and so approximate controllability for our system with a localized control in the set O = (l1 , l2 ),
0 < l1 < l2 ≤ L is an open question. See the first comment in Section 4 for more details.

3. Approximate controllability in two dimensions

In this section we study the approximate controllability of the linearized system (1.9)–(1.13) in the
rectangle.

3.1. Wellposedness of linearized and adjoint system

Let us recall that Z = L2 (Ω) × L2 (Ω) be endowed with the inner product
   
σ ρ
, := b σ(x)ρ(x) dx + Q u(x).v(x)dx
u v Z
Ω Ω

where b := aγQγ−2 , u = (u1 , u2 ), v = (v1 , v2 ) and u.v = u1 v1 + u2 v2 .


We also recall the following existence and uniqueness result for (1.9)–(1.12). Details can be found, for
example in [35] and [33].

Theorem 3.1. Given (σ0 , u0 ) ∈ Z, f ∈ L2 (0, T ; L2 (Ω)) the system (1.9)–(1.12) has a unique solution
(σ, u) ∈ C([0, T ]; L2 (Ω)) × [L2 (0, T ; H10 (Ω)) ∩ C([0, T ]; L2 (Ω))] with ut ∈ L2 (0, T ; H−1 (Ω)).

In order to study the approximate controllability of the linearized system (1.9)–(1.12) we consider the
following adjoint system in ΩT with terminal and boundary conditions

∂ψ  
(x, t) − V.∇ψ(x, t) − Q div φ(x, t) = 0,
− (3.1)
∂t
∂φ μ (λ + μ)   
− (x, t) − Δφ(x, t) − ∇ div φ(x, t) − (V.∇)φ(x, t) − b∇ψ(x, t) = 0 (3.2)
∂t Q Q
ψ(x, T ) = ψT (x), φ(x, T ) = φT (x) in Ω, (3.3)
ψ(x, t) = 0 on Γout × (0, T ), φ(x, t) = 0 on ΣT . (3.4)
S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057 1049

Reversing the time variable we can write the above system as the following initial and boundary value
problem:

∂ ψ̃  
(x, t) − V.∇ψ̃(x, t) − Q div φ̃(x, t) = 0 (3.5)
∂t
∂ φ̃ μ (λ + μ)   
(x, t) − Δφ̃(x, t) − ∇ div φ̃(x, t) − (V.∇)φ̃(x, t) − b∇ψ̃(x, t) = 0 (3.6)
∂t Q Q
ψ̃(x, 0) = ψT (x), φ̃(x, 0) = φT (x) in Ω, (3.7)
ψ̃(x, t) = 0 on Γout × (0, T ), φ̃(x, t) = 0 on ΣT , (3.8)

where ψ̃(x, t) = ψ(x, T − t), φ̃(x, t) = φ(x, T − t).


We get the following existence and uniqueness theorem for the adjoint system.

Theorem 3.2. Given (ψT , φT ) ∈ Z the system (3.5)–(3.8) has a unique solution (ψ̃, φ̃) ∈ C([0, T ]; L2 (Ω)) ×
[L2 (0, T ; H10 (Ω)) ∩ C([0, T ]; L2 (Ω))] with φ̃t ∈ L2 (0, T ; H−1 (Ω)).

3.2. Unique continuation

Now we are going to prove the following unique continuation result from which Theorem 1.3 follows using
Theorem 2.6 as before.

Theorem 3.3 (V1 > 0, V2 > 0). Let O = (0, l) × (0, h), 0 < l < L, 0 < h < H. Assume
that T > max{ (L−l) V1 ,
(H−h)
V2 } and (ψ̃, φ̃) ∈ C([0, T ]; L (Ω)) × [L (0, T ; H0 (Ω)) ∩ C([0, T ]; L (Ω))] with
2 2 1 2
−1
φ̃t ∈ L (0, T ; H (Ω)) be the solution of (3.5)–(3.8) such that φ̃ = 0 in O × (0, T ). Then (ψ̃, φ̃) = (0, 0) in
2

Ω × (0, T ).

Proof. Note that

∂ φ̃2 ∂ φ̃1
curl φ̃ = − .
∂x1 ∂x2

Then from (3.6) we get

∂ μ
(curl φ̃) − Δ(curl φ̃) − V.∇(curl φ̃) = 0 in Ω × (0, T ) (3.9)
∂t Q

in the sense of distribution.


Since φ̃ = 0 in O × (0, T ), curl φ̃ = 0 in O × (0, T ). Notice that the characteristic lines for the PDE (3.9)
are t = C. So taking

Ω1 = O × (0, T ), Ω2 = Ω × (0, T )

and applying Theorem 2.12 of Hörmander, we get

curl φ̃ = 0 in Ω × (0, T ).

Since Ω is a bounded simply-connected domain with Lipschitz boundary and φ̃ ∈ L2 (0, T ; H10 (Ω)), there
exists a scalar function g = g(x, t) ∈ H 2 (Ω) such that

φ̃ = ∇g in Ω × (0, T ). (3.10)
1050 S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057

Details of this existence of g can be found for example in Girault and Raviart [34, Chapter 1, Section 2.3,
Theorem 2.9].
From (3.5)–(3.6)

∂ φ̃1 ∂ φ̃2
W = div φ̃ = +
∂x1 ∂x2

satisfy

∂W μ (λ + μ)
− ΔW − ΔW − V.∇W − bΔψ̃ = 0 (3.11)
∂t Q Q
∂ ψ̃
− V.∇ψ̃ − QW = 0 (3.12)
∂t
in the sense of distribution, i.e.

∂W (λ + 2μ)
− ΔW − V.∇W = bΔψ̃, (3.13)
∂t Q
∂(Δψ̃)
− V.∇(Δψ̃) = QΔW . (3.14)
∂t
Using (3.13), from (3.14) we get W satisfies the following

∂2W (λ + 2μ) ∂W ∂W
P (D)W = 2
− Δ − V.∇
∂t Q ∂t ∂t
∂W (λ + 2μ)
− V.∇ − ΔW − V.∇W − bQΔW = 0. (3.15)
∂t Q

In the same way like Lemma 2.14, we find the characteristic hyperplanes in R3 for the above PDE and they
are given by

t = C1 , ξ1 x1 + ξ2 x2 + ηt = C2 where V1 ξ1 + V2 ξ2 = η. (3.16)

Let (ψ̃, φ̃) ∈ C([0, T ]; L2 (Ω)) × [L2 (0, T ; H10 (Ω)) ∩ C([0, T ]; L2 (Ω))] be the solution of (3.5)–(3.6) such
that φ̃ = 0 in O × (0, T ). Then from (3.15) we have in the distribution sense

∂2W (λ + 2μ) ∂W ∂W
P (D)W = 2
− Δ − V.∇
∂t Q ∂t ∂t
∂W (λ + 2μ)
− V.∇ − ΔW − V.∇W − bQΔW = 0
∂t Q

where W = div φ̃. By assumption we already have φ̃ = 0 in O × (0, T ), and so W = 0 in O × (0, T ) where
O = (0, l) × (0, h).
Let T > M = max{ (L−l)
V1 ,
(H−h)
V2 }. Define


A= (0, l + V1 s) × (0, h + V2 s) × (0, T − s).
0≤s<T

To apply Theorem 2.12 of Hörmander, we define


 
Ω1 = (0, l) × (0, h) × (0, T ), Ω2 = A ∩ Ω × (0, T ) .
S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057 1051

Note that if the characteristic line t = C1 or ξ1 x1 + ξ2 x2 + ηt = C2 with V1 ξ1 + V2 ξ2 = η hits Ω2 , then it


hits Ω1 . Since W satisfies

P (D)W = 0 in D (Ω2 ) and W = 0 in Ω1 ,

using Theorem 2.12 we get

W = 0 in Ω2 , (3.17)

i.e. div φ̃ = 0 in Ω2 . So from (3.10) we get

Δg = 0 in Ω2 . (3.18)

Thus g is a harmonic function and hence analytic. Since φ̃ = ∇g in Ω × (0, T ) and φ̃ = 0 in O × (0, T ), we
have

g(x, t) = e(t) in O × (0, T ). (3.19)

From (3.18)–(3.19) we get

g(x, t) = e(t) in Ω2 . (3.20)

Therefore from (3.10) and (3.20), we get

φ̃ = ∇g = ∇e(t) = 0 in Ω2 . (3.21)

Using (3.21) and (3.5)–(3.6) we obtain

∇ψ̃ = 0, ψ̃t = 0 in D (Ω2 ),

where Ω2 is a connected set. Thus ψ̃(x, t) = K(constant) in Ω2 . From the boundary conditions (3.8) we
already have ψ̃(x, t) = 0 on Γout × (0, T ) and hence

ψ̃ = 0 in Ω2 . (3.22)

As Ω × (0, T − M ) ⊂ Ω2 , from (3.21) and (3.22) we have in particular

ψ̃ = 0, φ̃ = 0 in Ω × (0, T − M ).

Hence by forward uniqueness we obtain that

ψ̃ = 0, φ̃ = 0 in Ω × (0, T ).

This completes the proof of Theorem 3.3. 2

Similarly for V1 > 0, V2 = 0 case, taking

 (L − l)
A= (0, l + V1 s) × (0, H) × (0, T − s) for T > ,
V1
0≤s<T

we will get the following unique continuation result and hence Theorem 1.4 follows.
1052 S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057

Theorem 3.4 (V1 > 0, V2 = 0). Let O = (0, l) × (0, H), 0 < l < L. Assume that T > (L−l) V1 and
(ψ̃, φ̃) ∈ C([0, T ]; L2 (Ω)) × [L2 (0, T ; H10 (Ω)) ∩ C([0, T ]; L2 (Ω))] with φ̃t ∈ L2 (0, T ; H−1 (Ω)) be the solution
of (3.5)–(3.8) such that φ̃ = 0 in O × (0, T ). Then (ψ̃, φ̃) = (0, 0) in Ω × (0, T ).

4. Further remarks and open questions

(i) Control supported in more general interval and backward uniqueness


Proof of Theorem 1.2 relies on the fact that control acts on the set (0, l) touching the left boundary of the
domain (0, L). If control acts on the set O = (l1 , l2 ) not touching the left boundary i.e. 0 < l1 < l2 ≤ L then
to prove approximate controllability we need an additional backward uniqueness result for our system. In this
context, note that the transport equation in a bounded interval with Dirichlet boundary condition does not
have backward uniqueness whereas heat equation has backward uniqueness. In our coupled system both the
types of equations are combined and we do not know yet backward uniqueness is true or not for our system.
If we have the backward uniqueness result then Theorem 1.2 can be generalized (in the way we indicated in
Remark 2.15) where control set O can be any (l1 , l2 ), 0 ≤ l1 < l2 ≤ L. Then approximate controllability holds
for the system (1.6)–(1.8) in Z for T > max{ Vl10 , L−l V0 }. Similarly Theorem 1.3 can also be generalized for
2

the subset O = (l1 , l2 ) ×(h1 , h2 ), 0 ≤ l1 < l2 ≤ L, 0 ≤ h1 < h2 ≤ H if we have a backward uniqueness result.
In that case, we will get approximate controllability in Z for T > max{ Vl11 , (L−l 2 ) h1 (H−h2 )
V1 , V2 , V2 }. However,
proof of approximate controllability for any control set may need totally different approach without using
backward uniqueness.
For the linearized system with Dirichlet boundary conditions, proving backward uniqueness seems to be
a challenging open problem. We are not able to use Fourier’s series technique. Odd order spacial derivative
present in the system and Dirichlet boundary condition for density are only a part of boundary (not in the
total boundary) are some of the difficulties. We need other ideas to get any backward uniqueness result if
it is true.
(ii) Need of a large time to have approximate controllability in our system
Large time is natural for the control of hyperbolic type equation and time T plays the role of a critical
parameter. For 1-D wave equation in Ω = (0, L) when interior control acts in O = (l1 , l2 ), 0 ≤ l1 < l2 ≤ L,
Haraux in [42] shows approximate controllability is true if T > 2 max{l1 , L −l2 } and for T < 2 max{l1 , L −l2 }
it fails. Because of the hyperbolic nature of equation of the density in our coupled system, time is expected
to be large and critical for the control properties of the equation of the density in (1.6). Note that Ervedoza
et al. in [19] also need large time for controllability of nonlinear system around (Q0 , V0 ), Q0 > 0, V0 > 0,
namely T > VL0 using boundary controls.
To get a better view, let us consider the following transport equation with interior control:

σt (x, t) + V0 σx (x, t) = f χ(0,l) ,


σ(x, 0) = σ0 (x) ∀x ∈ IL , σ(0, t) = 0 ∀t ∈ (0, T ), (4.1)

where χ(0,l) is the characteristic function of (0, l), 0 < l ≤ L.


Reversing the time variable we can write the adjoint system as

ψ̃t (x, t) − V0 ψ̃x (x, t) = 0,


ψ̃(x, 0) = ψT (x) ∀x ∈ IL , ψ̃(L, t) = 0 ∀t ∈ (0, T ). (4.2)

Let

L−l
T < .
V0
S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057 1053

We take ψT ∈ L2 (0, L) as

ψT (x) = 0 ∀x ∈ (0, l + V0 T ), (4.3)


ψT (x) = 0 ∀x ∈ (l + V0 T, L). (4.4)

Then we obtain

ψ̃(x, t) = 0 in (0, l) × (0, T )

but

ψ̃(x, t) = 0 in (0, L) × (0, T ).

L−l
So unique continuation fails for T < V0 and hence approximate controllability fails for transport equation
when T < L−l
V0 . Thus we know for transport equation optimal time for approximate controllability is (L−l) V0 .
In this paper, we need large time to prove approximate controllability for linearized coupled system
(1.6)–(1.8). But to show for T < (L−l) V0 , system is not approximate controllable, is an interesting open
problem. We expect here this will happen as transport case since in our coupled system, density equation
is present. But finding counter example(like transport equation) to show that unique continuation will fail
for T < (L − l)/V0 in our coupled system is not easy. It becomes difficult since heat equation has infinite
speed of propagation and its effect on our coupled system. Very recently D. Maity in [50] proves that the one
dimensional compressible Navier–Stokes system for non-barotropic fluid linearized around a constant steady
state (ρ̄, v̄, θ̄) with ρ̄ > 0, v̄ > 0, θ̄ > 0 is not null controllable by localized interior control or boundary control
for small time T . He constructs highly localized solutions known as Gaussian Beam to prove this negative
result. Constructing particular terminal condition for the adjoint problem, he makes in [50] observation
small and contradicts the observability inequality. But to get non-approximate controllability in small time
for our coupled system, we need to construct terminal condition such that observation is exactly zero (not
only small), which becomes very difficult due to the heat part (infinite speed of propagation).
(iii) Our method and its limitation
The method we used is natural, since we have differential operator with constant coefficients and x = c
are not characteristic. So Hörmander’s theorem helps in this case (see for examples, Lions [48] in the context
of Wave equation). For heat equation characteristic hyperplanes are horizontal, parallel to the hyperplane
t = 0. So this method is also used in the case of heat equation to prove approximate controllability for any
T > 0 (see for examples, Zuazua Notes [53]). The other popular techniques to establish unique continuation
based on the derivation of a Carleman estimate. (See for example Saut [58].) For simplicity we give the results
in a rectangle in two dimensions. This method will also give approximate controllability in a rectangular
parallelepiped in three dimensions for the linearized system around constant states (Q, V1 , V2 , V3 ). Note that
in three dimensions

∂ φ̃3 ∂ φ̃2 ∂ φ̃1 ∂ φ̃3 ∂ φ̃2 ∂ φ̃1


curl φ̃ = − , − , −
∂x2 ∂x3 ∂x3 ∂x1 ∂x1 ∂x2

and so Eq. (3.9) will be satisfied component wise in that case.


We have mentioned already Fourier analysis will not help to get approximate controllability for the whole
system in this case since Dirichlet boundary condition for density are only a part of boundary(not in the
total boundary). Fourier approach reducing in to a integro-differential equation for velocity can at most
give some partial approximate controllability for velocity only (not density) for any T > 0 i.e. results like
Doubova [17].
1054 S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057

In one dimension when V0 = 0, i.e. for the linearized system around (Q0 , 0) we have proved null and
approximate controllability results in [11] using Fourier analysis. Note that for this V0 = 0 case, unique
continuation property cannot be obtained directly from Holmgren’s theorem, since the lines x = const. are
characteristic. Similar difficulties arise to get controllability when V1 = 0, V2 = 0 in two dimensions also
and we need a different approach. For V1 = 0, V2 = 0, in two dimensions with special boundary conditions
introduced by Vaigant and Kazhikhov in [60], we can capture the main structures of the spectrum of the
linearized operator. Using this knowledge of the spectrum of linearized models and Fourier analysis like in
one dimension [11], we can get similar approximate and null controllability results in two dimensions. But
for V1 = 0, V2 = 0, in two dimensions with Dirichlet boundary conditions we do not know the behavior
of the spectrum of the linearized operator and are unable to use Fourier’s series technique. Controllability
questions are open in this case.
(iv) Role of simple geometry
We choose simple geometry (example (0, l) ⊆ (0, L)) to get an idea of time needed for approximate
controllability for whole system using interior control on velocity. We find it is the same time required
for approximate controllability of transport equation using interior control. Although we are not able to
prove that yet it is optimal time for approximate controllability like transport. We now explain that simple
geometry gives better value of time for approximate controllability. Let V = (1, 1, 1, . . . , 1) ∈ RN and Ω
be an open, bounded, connected domain of RN and ω an open subset of Ω contained in the nbd of the
boundary where density is prescribed. (For example in our case (0, l) or (0, l) × (0, h)). For any x ∈ Ω, we
define δ(x, ω) = inf{δ(x, y), y ∈ ω}, where δ(x, y) is the infimum of the lengths of all polygonal lines joining
x and y and contained in Ω. We define,
 
δ(Ω, ω) := sup δ(x, ω), x ∈ Ω .

Then it is natural here (as transport part is present) to expect approximate controllability for

T > δ(Ω, ω).

[However in wave equation case we need T > 2δ(Ω, ω) (see [42,48]).]


Note that for V0 = 1, ω = (0, l), Ω = (0, L) we have

δ(Ω, ω) = L − l

but for V = (1, 1), ω = (0, l) × (0, h), Ω = (0, L) × (0, H) we get

δ(Ω, ω) > max{L − l, H − h}.

So simple geometry like our case can give better lower bound for time.
(v) Exact-approximate controllability
Here we prove approximate controllability for V0 > 0 in one dimension and V1 > 0, V2 ≥ 0 in two
dimensions. Exact-approximate controllability in one dimension(as a starting point) for the linearized system
around (Q0 , V0 ), Q0 > 0, V0 > 0 is an interesting problem and our work in this direction is in progress.
(vi) Controllability problems for nonlinear system
Since we have the approximate controllability for the linearized system around constant steady states,
it is natural to ask can we get some local controllability properties for the nonlinear system from the
linearized one. This has been done for approximate controllability of semilinear heat equation in many
papers, for instance by Fabre, Zuazua, Puel and Fernández in [22,23,63] (the references therein). They
uses a variant of a classical fixed point method or the penalization of an optimal control problem. For
approximate controllability of semilinear heat equation in [22,23,63], the assumption that the nonlinear
S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057 1055

term satisfies the globally Lipschitz condition is essential. The situation changes when this condition is not
satisfied. Moreover, in our compressible case, nonlinear term becomes substantially more difficult compared
to incompressible case due to density.
Fursikov and Imanuvilov proved in [30, Theorem 2.1] and also in [29, Theorem 7.1] that the viscous
Burgers equation in the interval (0, L) itself is not globally approximately controllable in L2 (0, L) with
respect to a distributed control in L2 ((0, L) × (0, T )) concentrated in subinterval. They also proved that
it is not globally approximately controllable as well as with respect to a boundary control in L2 (0, T ) at
the end x = L. This kind of result is obtained by means of a new estimate on a solution of the viscous
Burgers equation. It can be also found in the book [31] (see Chapter 1, Theorems 6.3 and 6.4). So in general
we should look for (one can hope to get in our case using localized interior control) a positive local result
instead of global result.
At present, according to our knowledge, as long as the control is acting in an interior domain or on
part of the boundary, there is no global approximate controllability result for incompressible Navier–Stokes
equations itself. Thus approximate controllability is an open question yet for incompressible Navier–Stokes
system. (Conjecture on approximate controllability in [49] formulated by J.L. Lions is still not solved.) Only
results in this direction have been obtained by Coron in [12] for different boundary conditions (Navier slip
boundary conditions), by Coron and Fursikov in [13] (on a manifold without boundary), by Fabre in [20]
for truncated nonlinearty and by Guerrero, Imanuvilov, Puel in [39,38] (‘concerning’ global approximate
controllability but not classical one). Study of compressible case becomes more involved since it is coupled
with density equation. Compressible case, only local exact controllability to ‘constant trajectories’ is known
now due to Ervedoza et al. [19]. So as a starting point, study of local approximate controllability for the
nonlinear system near constant steady states using interior control for velocity only will be our aim in future.
To get approximate controllability for compressible Navier–Stokes system, fixed point argument (Schaud-
er’s fixed point theorem) seems to be a natural strategy. For that, we have to prove first approximate
controllability of linearized systems with variable L∞ (Ω × (0, T )) coefficients. In order to have compactness,
we have to study the approximate controllability problem with this irregular variable coefficients. Thus it
reduces to prove unique continuation property, for the adjoint system with L∞ (Ω × (0, T )) coefficients.
We do not know and also able to prove this kind of result yet. In this context we mention that to study
approximate controllability of incompressible Navier–Stokes equation, unique continuation of Stokes kind
system and then fixed-point argument is used by Fabre in [20]. Although note that this approach does not
provide approximate controllability for incompressible Navier–Stokes, but only for an approximation(for a
different nonlinearity) of it. In conclusion, approximate controllability is an open question for compressible
Navier–Stokes system also. In fact compressible case, it becomes more difficult and involved.

Acknowledgments

The author would like to thank Prof. Jean-Pierre Raymond and Prof. Mythily Ramaswamy for useful
discussions. The author would like to express his sincere thanks to the referee for helpful comments, valuable
questions and suggestions to improve the first version of the paper. The author acknowledge the financial
support from the Indo French Centre for Applied Mathematics (IFCAM) under the project “PDE control”.

References

[1] P. Albano, D. Tataru, Carleman estimates and boundary observability for a coupled parabolic–hyperbolic system, Electron.
J. Differential Equations 22 (2000), 15 pp.
[2] E.V. Amosova, Exact local controllability for the equations of viscous gas dynamics, Differential Equations 47 (12) (2011)
1776–1795.
[3] M. Badra, S. Ervedoza, S. Guerrero, Local controllability to trajectories for non-homogeneous 2-D incompressible Navier–
Stokes equations, preprint, 2013.
1056 S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057

[4] V. Barbu, M. Iannelli, Controllability of the heat equation with memory, Differential Integral Equations 13 (10–12) (2000)
1393–1412.
[5] J.L. Boldrini, A. Doubova, E. Fernández-Cara, M. González-Burgos, Some controllability results for linear viscoelastic
fluids, SIAM J. Control Optim. 50 (2) (2012) 900–924.
[6] N. Carreño, S. Guerrero, Local null controllability of the N-dimensional Navier–Stokes system with N − 1 scalar controls
in an arbitrary control domain, J. Math. Fluid Mech. 15 (1) (2013) 139–153.
[7] F.W. Chaves-Silva, L. Rosier, E. Zuazua, Null controllability of a system of viscoelasticity with a moving control, J. Math.
Pures Appl. (2014), in press.
[8] S. Chowdhury, D. Mitra, Null controllability of the linearized compressible Navier–Stokes equations using moment method,
submitted for publication.
[9] S. Chowdhury, D. Mitra, M. Ramaswamy, M. Renardy, Null controllability of the linearized compressible Navier–Stokes
system in one dimension, J. Differential Equations 257 (10) (2014) 3813–3849, http://dx.doi.org/10.1016/j.jde.2014.07.010.
[10] S. Chowdhury, M. Ramaswamy, Optimal control of linearized compressible Navier–Stokes equations, ESAIM Control
Optim. Calc. Var. 19 (2) (2013) 587–615.
[11] S. Chowdhury, M. Ramaswamy, J.-P. Raymond, Controllability and stabilizability of the linearized compressible Navier–
Stokes system in one dimension, SIAM J. Control Optim. 50 (2012) 2959–2987.
[12] J.-M. Coron, On the controllability of the 2-D incompressible Navier–Stokes equations with the Navier slip boundary
conditions, ESAIM Control Optim. Calc. Var. 1 (1995/96) 35–75.
[13] J.-M. Coron, A.V. Fursikov, Global exact controllability of the 2-D Navier–Stokes equations on a manifold without bound-
ary, Russian J. Math. Phys. 4 (1996) 429–448.
[14] J.-M. Coron, S. Guerrero, Local null controllability of the two-dimensional Navier–Stokes system in the torus with a control
force having a vanishing component, J. Math. Pures Appl. 9 (92) (2009) 528–545.
[15] E. Crépeau, C. Prieur, Approximate controllability of a reaction–diffusion system, Systems Control Lett. 57 (2008)
1048–1057.
[16] L. de Teresa, E. Zuazua, Controllability of the linear system of thermoelastic plates, Adv. Differential Equations 1 (3)
(1996) 369–402.
[17] A. Doubova, E. Fernández-Cara, On the control of viscoelastic Jeffreys fluids, Systems Control Lett. 61 (4) (2012) 573–579.
[18] A. Doubova, E. Fernández-Cara, M. González-Burgos, Controllability results for linear viscoelastic fluids of the Maxwell
and Jeffreys kinds, C. R. Acad. Sci. Paris Sér. I Math. 331 (7) (2000) 537–542.
[19] S. Ervedoza, O. Glass, S. Guerrero, J.-P. Puel, Local exact controllability for the one-dimensional compressible Navier–
Stokes equation, Arch. Ration. Mech. Anal. 206 (1) (2012) 189–238.
[20] C. Fabre, Uniqueness results for Stokes equations and their consequences in linear and nonlinear control problems, ESAIM
Control Optim. Calc. Var. 1 (1995/96) 267–302.
[21] C. Fabre, Résultats d’unicité pour les équations de Stokes et applications au contrôle (Uniqueness results for Stokes
equations and their consequences in linear and nonlinear control problems), C. R. Acad. Sci. Paris Sér. I Math. 322 (12)
(1996) 1191–1196.
[22] C. Fabre, J.-P. Puel, E. Zuazua, Approximate controllability of the semilinear heat equation, Proc. Roy. Soc. Edinburgh
Sect. A 125 (1) (1995) 31–61.
[23] L.A. Fernández, E. Zuazua, Approximate controllability for the semilinear heat equation involving gradient terms, J. Op-
tim. Theory Appl. 101 (2) (1999) 307–328.
[24] E. Fernández-Cara, Motivation, analysis and control of the variable density Navier–Stokes equations, Discrete Contin.
Dyn. Syst. Ser. S 5 (6) (2012) 1021–1090.
[25] E. Fernández-Cara, L. de Teresa, Null controllability of a cascade system of parabolic–hyperbolic equations, Discrete
Contin. Dyn. Syst. 11 (2–3) (2004) 699–714.
[26] E. Fernández-Cara, S. Guerrero, O.Yu. Imanuvilov, J.-P. Puel, Local exact controllability of the Navier–Stokes system,
J. Math. Pures Appl. (9) 83 (12) (2004) 1501–1542.
[27] E. Fernández-Cara, S. Guerrero, O.Yu. Imanuvilov, J.-P. Puel, Some controllability results for the N-dimensional Navier–
Stokes and Boussinesq systems with N − 1 scalar controls, SIAM J. Control Optim. 45 (1) (2006) 146–173.
[28] A.V. Fursikov, O.Yu. Imanuilov, Exact controllability of the Navier–Stokes and Boussinesq equations (Russian), Uspekhi
Mat. Nauk 54 (3(327)) (1999) 93–146, translation in Russian Math. Surveys 54 (3) (1999) 565–618.
[29] A.V. Fursikov, O.Yu. Imanuvilov, On approximate controllability of the Stokes system, Annales de la Faculte des sciences
de Toulous 11 (1993) N2.
[30] A.V. Fursikov, O.Yu. Imanuvilov, On controllability of certain systems simulating a fluid flow, in: Flow Control, Min-
neapolis, MN, 1992, in: IMA Vol. Math. Appl., vol. 68, Springer, New York, 1995, pp. 149–184.
[31] A.V. Fursikov, O.Yu. Imanuvilov, Controllability of Evolution Equations, Lecture Notes Series, vol. 34, Seoul National
University, Research Institute of Mathematics, Global Analysis Research Center, Seoul, 1996.
[32] A.V. Fursikov, O.Yu. Imanuvilov, Local exact controllability of the two-dimensional Navier–Stokes equations (Russian),
Mat. Sb. 187 (9) (1996) 103–138, translation in Sb. Math. 187 (9) (1996) 1355–1390.
[33] G. Geymonat, P. Leyland, Transport and propagation of a perturbation of a flow of a compressible fluid in a bounded
region, Arch. Rational Mech. Anal. 100 (1) (1987) 53–81.
[34] V. Girault, P.-A. Raviart, Finite element methods for Navier–Stokes equations, in: Theory and Algorithms, in: Springer
Series in Computational Mathematics, vol. 5, Springer-Verlag, Berlin, 1986.
[35] V. Girinon, Quelques problemes aux limites pour les equations de Navier–Stokes compressibles, PhD Thesis, Université
de Toulouse, 2008.
[36] M. González-Burgos, S. Guerrero, J.-P. Puel, Local exact controllability to the trajectories of the Boussinesq system via
a fictitious control on the divergence equation, Commun. Pure Appl. Anal. 8 (1) (2009) 311–333.
[37] S. Guerrero, Local exact controllability to the trajectories of the Navier–Stokes system with nonlinear Navier-slip boundary
conditions, ESAIM Control Optim. Calc. Var. 12 (3) (2006) 484–544.
S. Chowdhury / J. Math. Anal. Appl. 422 (2015) 1034–1057 1057

[38] S. Guerrero, O.Yu. Imanuvilov, J.-P. Puel, Remarks on global approximate controllability for the 2-D Navier–Stokes system
with Dirichlet boundary conditions, C. R. Math. Acad. Sci. Paris 343 (9) (2006) 573–577.
[39] S. Guerrero, O.Yu. Imanuvilov, J.-P. Puel, A result concerning the global approximate controllability of the Navier–Stokes
system in dimension 3, J. Math. Pures Appl. (9) 98 (6) (2012) 689–709.
[40] S. Guerrero, O.Yu. Imanuvilov, Remarks on non controllability of the heat equation with memory, ESAIM Control Optim.
Calc. Var. 19 (1) (2013) 288–300.
[41] A. Halanay, L. Pandolfi, Lack of controllability of the heat equation with memory, Systems Control Lett. 61 (2012)
999–1002.
[42] A. Haraux, A generalized internal control for the wave equation in a rectangle, J. Math. Anal. Appl. 153 (1) (1990)
190–216.
[43] L. Hörmander, The Analysis of Linear Partial Differential Operators I, Distribution Theory and Fourier Analysis, Second
edition, Springer study edition, Springer-Verlag, Berlin, 1990.
[44] O.Yu. Imanuvilov, On exact controllability for the Navier–Stokes equations, ESAIM Control Optim. Calc. Var. 3 (1998)
97–131.
[45] O.Yu. Imanuvilov, Remarks on exact controllability for the Navier–Stokes equations, ESAIM Control Optim. Calc. Var. 6
(2001) 39–72.
[46] S. Ivanov, L. Pandolfi, Heat equation with memory: lack of controllability to rest, J. Math. Anal. Appl. 355 (2009) 1–11.
[47] G. Lebeau, E. Zuazua, Null-controllability of a system of linear thermoelasticity, Arch. Rational Mech. Anal. 141 (4)
(1998) 297–329.
[48] J.L. Lions, Contrôlabilité exacte perturbations et stabilisation de systèmes distribués. Tome 1. Contrôlabilite exacte,
Masson, RMA 8 (1988).
[49] J.-L. Lions, Are there connections between turbulence and controllability?, in: A. Bensoussan, J.-L. Lions (Eds.), Lecture
Notes in Control and Inform. Sci., vol. 144, Springer-Verlag, New York, 1990, Remarques sur la contrôlabilité approchée,
in Jornadas Hispano-Francesas sobre Control de Sistemas Distribuidos, Málaga, Spain, 1990.
[50] D. Maity, Some controllability results for linearized conpressible Navier–Stokes system, submitted for publication, 2014.
[51] P. Martin, L. Rosier, P. Rouchon, Null controllability of the structurally damped wave equation with moving control,
SIAM J. Control Optim. 51 (2013) 660–684.
[52] S. Micu, On the controllability of the linearized Benjamin–Bona–Mahony equation, SIAM J. Control Optim. 39 (6) (2001)
1677–1696.
[53] S. Micu, E. Zuazua, An introduction to the controllability of partial differential equations, http://www.uam.es/
personal_pdi/ciencias/ezuazua/informweb/argel.pdf.
[54] M. Renardy, Are viscoelastic flows under control or out of control?, Systems Control Lett. 54 (12) (2005) 1183–1193.
[55] M. Renardy, A note on a class of observability problems for PDEs, Systems Control Lett. 58 (3) (2009) 183–187.
[56] M. Renardy, Mathematical analysis of viscoelastic flows, CBMS-NSF Regional Conference Series in Applied Mathematics,
vol. 73, Society for Industrial and Applied Mathematics (SIAM), Philadelphia.
[57] L. Rosier, P. Rouchon, On the controllability of a wave equation with structural damping, Int. J. Tomogr. Stat. 5 (W07)
(2007) 79–84.
[58] J.-C. Saut, B. Scheurer, Unique continuation for some evolution equations, J. Differential Equations 66 (1) (1987) 118–139.
[59] Q. Tao, H. Gao, B. Zhang, Approximate controllability of a parabolic equation with memory, Nonlinear Analysis: Hybrid
Systems 6 (2012) 839–845.
[60] A.V. Val̆gant, A.V. Kazhikhov, On the existence of global solutions of two-dimensional Navier–Stokes equations of a
compressible viscous fluid, (Russian) Sibirsk. Mat. Zh. 36 (1995) 1283–1316, ii, translation in Siberian Math. J. 36 (6)
(1995) 1108–1141.
[61] J. Zabczyk, Mathematical Control Theory, An Introduction, Reprint of the 1995 edition Modern Birkhäuser Classics,
Birkhäuser Boston, Inc., Boston, MA, 2008.
[62] E. Zuazua, Controllability of the linear system of thermoelasticity, J. Math. Pures Appl. (9) 74 (4) (1995) 291–315.
[63] E. Zuazua, Approximate controllability for semilinear heat equations with globally Lipschitz nonlinearities, Recent ad-
vances in control of PDEs, Control Cybernet. 28 (3) (1999) 665–683.

You might also like