Skip to main content

Optimal control for cooperative systems involving fractional Laplace operators

Abstract

In this work, the elliptic \(2\times 2\) cooperative systems involving fractional Laplace operators are studied. Due to the nonlocality of the fractional Laplace operator, we reformulate the problem into a local problem by an extension problem. Then, the existence and uniqueness of the weak solution for these systems are proved. Hence, the existence and optimality conditions are deduced.

1 Introduction

Nonlocal operators have been a useful area of investigation in different branches of mathematics such as operator theory and harmonic analysis. Also, they have gained vital attention because of their strong connection with real-world problems since they form a fundamental part of the modeling and simulation of complex phenomena that span vastly different length scales.

Nonlocal operators appear in several applications such as image processing, boundary-control problems, electromagnetic fluids, materials science, porous-media flow, turbulence, optimization, nonlocal continuum field theories, and others. Consequently, the domain of definition Ω may be in its general form.

In this paper, we discuss the elliptic \(2\times 2\) cooperative system containing one of the nonlocal operators namely the fractional Laplace operator \((- \Delta )^{s}\).

Let \(\Omega \subset \mathbb{R}^{N}\), \(N\geq 2s\), be an open, bounded and connected domain with Lipschitz boundary Ω. Then, we shall study the following system:

$$\begin{aligned} \textstyle\begin{cases} (-\Delta )^{s}y_{1}=a y_{1}+b y_{2}+f_{1} &\text{in}\ \Omega , \\ (-\Delta )^{s}y_{2}=c y_{1}+d y_{2}+f_{2}& \text{in} \ \Omega , \\ y_{1}=y_{2}=0&\text{in} \ \mathbb{R}^{N}\backslash \Omega , \end{cases}\displaystyle \end{aligned}$$
(1.1)

where \(y=\{y_{1},y_{2}\}\) are the states of the system, \(f=\{f_{1},f_{2} \}\) are the external sources. The fractional Laplace operator \((- \Delta )^{s}\) is defined by Fourier transform as follows:

$$ (-\Delta )^{s} y(x)=c_{N, s} \quad \text{P.V.}\quad \int _{\mathbb{R}^{N}} \frac{y(x)-y(t)}{ \vert x-t \vert ^{N+2 s}} \,dt,\quad c_{N, s}>0. $$
(1.2)

From (1.2), it becomes clear that the fractional laplace operator \((- \Delta )^{s}\) is a nonlocal operator.

Definition 1.1

For given numbers a, b, c and d the system

$$\begin{aligned} \textstyle\begin{cases} (-\Delta )^{s}y_{1}=a y_{1}+b y_{2}+f_{1} &\text{in}\ \Omega , \\ (-\Delta )^{s}y_{2}=c y_{1}+d y_{2}+f_{2}& \text{in} \ \Omega , \end{cases}\displaystyle \end{aligned}$$
(1.3)

is called cooperative if \(b,c>0\); otherwise, the system (1.3) is said to be noncooperative.

Optimal control for partial differential equations (PDEs) has been widely studied in many fields such as biology, ecology, economics, engineering, and finance [510, 18, 22, 24, 25, 30, 34, 37]. These results have been expanded in [12, 14, 15, 29, 3133] to cooperative and noncooperative systems. The fractional optimal control problems are the generalization of standard optimal control problems. Hence, it allows treatment of more general applications in physics, chemistry, and engineering [13, 17, 20, 21, 23, 39]. Several papers discuss time-fractional optimal control. In [27, 28], the distributed optimal control problem for a time-fractional diffusion equation is discussed. Moreover, the optimality conditions are derived. In [19], the distributed control for a time-fractional differential system involving a Schrödinger operator is studied, and the optimality conditions are derived. Furthermore, space-fractional optimal control is introduced. In [1, 2, 11], the nonlocal system is reformulated to a local system by an extension problem. Hence, the optimality conditions are achieved.

Henceforth, to treat with the nonlocality of the fractional Laplace operator, in [4], Caffarelli and Silvestre proved that the fractional Laplace operator can be characterized as an operator that maps a Dirichlet boundary condition to a Neumann-type condition via an extension problem as follows:

Let \({D}^{+}\subset \mathbb{R}^{N+1}\) be a semi-infinite cylinder as

$$\begin{aligned} {D}^{+}=\bigl\{ (x,t):x\in \mathbb{R}^{N}, t\in (0,\infty ) \bigr\} \subset \mathbb{R}^{N+1}, \end{aligned}$$
(1.4)

where t is a new extended variable. Therefore, the nonlocal problem (1.1) is reformulated locally as follows:

$$\begin{aligned} \textstyle\begin{cases} \nabla \cdot(t^{1-2s}\nabla Y_{1})=0 & \text{in} \ D^{+}, \\ \nabla \cdot(t^{1-2s}\nabla Y_{2})=0 & \text{in} \ D^{+} , \\ Y_{1}(x,0)=Y_{2}(x,0)=0 & \text{in}\ \mathbb{R}^{N}\backslash \Omega , \\ \frac{1}{k_{s}}\frac{\partial Y_{1}}{\partial \nu }=(f_{1}+a \operatorname{Tr}_{\Omega }Y_{1}+b \operatorname{Tr}_{\Omega }Y_{2})& \text{on}\ \Omega \times \{0\}, \\ \frac{1}{k_{s}}\frac{\partial Y_{2}}{\partial \nu }=(f_{2}+c \operatorname{Tr}_{\Omega }Y_{1}+d \operatorname{Tr}_{\Omega }Y_{2})& \text{on}\ \Omega \times \{0\}, \end{cases}\displaystyle \end{aligned}$$
(1.5)

where \(\frac{\partial Y}{\partial \nu }=-\lim_{t\to 0^{+}}t^{1-2s} \frac{\partial Y(x,t)}{\partial t}\), ν is the unit outer normal to \({D}^{+}\) at \(\Omega \times \{0\}\), \(\lim_{t \to \infty } Y(x, t) = 0\) and \(k_{s}=2^{1-2s} \frac{\Gamma (1-s)}{\Gamma (s)}>0\).

In this paper, we generalize some previous results obtained for the classical cooperative systems. Indeed, we consider the elliptic \(2 \times 2\) cooperative system involving one of the nonlocal operators called the fractional Laplace operator. The nonlocality of the fractional Laplace operator creates some difficulties. To overcome these, we transform the nonlocal system into a local system via an extension problem. Hence, via the Lax–Milgram lemma, we are able to prove the existence and uniqueness of the weak solution for the local system. Moreover, for both local and nonlocal systems, the optimality conditions are derived via the Lions technique. The results obtained tend to the classical results if \(s\rightarrow 1\). This article is organized as follows. In Sect. 2, we introduce some functional spaces to represent the fractional cooperative systems and their extension, and also furnish the existence results. In Sect. 3, the weak solution and the optimality condition are established for the scalar case. In Sect. 4, we can generalize our results to a \(2\times 2\) cooperative system involving the fractional Laplace operator. Section 5 is devoted to a summary and discussion.

2 Preliminaries

In our work, the optimal control of the cooperative system depends on the variational formulation. Hence, we introduce the Sobolev spaces, which are the solution spaces for our problem. This section includes three subsections. Section 2.1 provides a short overview of classical fractional Sobolev spaces. In Sect. 2.2, we recollect the idea of the weighted Sobolev spaces and their embedding properties. The characterization of the principal eigenvalue problem is presented in Sect. 2.3.

2.1 Fractional Sobolev spaces

For \(0< s<1\), define the fractional-order Sobolev space [3, 35]

$$\begin{aligned} H^{s}(\Omega )= \biggl\{ y \in L^{2}(\Omega ): \int _{\Omega } \int _{ \Omega } \frac{ \vert y(x_{1})-y(x_{2}) \vert ^{2}}{ \vert x_{1}-x_{2} \vert ^{N+2s}} \,dx_{1} \,dx_{2}< \infty \biggr\} , \end{aligned}$$
(2.1)

which is a Hilbert space endowed with the norm

$$\begin{aligned} \Vert u \Vert _{H^{s}(\Omega )}:= \biggl( \int _{\Omega } \vert y \vert ^{2} \,dx+ \int _{ \Omega } \int _{\Omega } \frac{ \vert y(x_{1})-y(x_{2}) \vert ^{2}}{ \vert x_{1}-x_{2} \vert ^{N+2 s}} \,dx_{1} \,dx _{2} \biggr)^{\frac{1}{2}}. \end{aligned}$$
(2.2)

Also, define the space \(H_{0}^{s}(\Omega )\) as:

$$\begin{aligned} H_{0}^{s}(\Omega )= \bigl\{ y \in H^{s}( \Omega ): y=0 \text{ on } \partial \Omega \bigr\} , \end{aligned}$$
(2.3)

which can be endowed with norm

$$\begin{aligned} \Vert u \Vert _{H_{0}^{s}(\Omega )}:= \biggl( \int _{\Omega } \int _{\Omega } \frac{ \vert y(x_{1})-y(x_{2}) \vert ^{2}}{ \vert x_{1}-x_{2} \vert ^{N+2 s}} \,dx_{1} \,dx_{2} \biggr)^{\frac{1}{2}}. \end{aligned}$$
(2.4)

Moreover, the Lions–Magenes space is given by [26]

$$\begin{aligned} H_{00}^{\frac{1}{2}}(\Omega )= \biggl\{ u \in H^{\frac{1}{2}}( \Omega ): \int _{\Omega } \frac{u^{2}(x)}{d(x, \partial \Omega )} \,dx< \infty \biggr\} , \end{aligned}$$
(2.5)

where \(d(x, \partial \Omega )\) is the distance from x to Ω. Combining (2.1), (2.3) and (2.5) we have for any \(s\in (0,1)\), the following fractional Sobolev space:

$$ \mathcal{H}^{s}(\Omega )= \textstyle\begin{cases} {H^{s}(\Omega );} & {s \in (0, \frac{1}{2} )}, \\ {H_{00}^{1 / 2}(\Omega );} & {s=\frac{1}{2}}, \\ {H_{0}^{s}(\Omega );} & {s \in (\frac{1}{2}, 1 )}. \end{cases} $$
(2.6)

Moreover, we denote by \(\mathcal{H}^{-s}(\Omega )\) the dual space of \(\mathcal{H}^{s}(\Omega )\) such that

$$ (-\Delta )^{s}:\mathcal{H}^{s}(\Omega )\to \mathcal{H}^{-s}(\Omega ). $$
(2.7)

Also, we have the following embedding

$$ \mathcal{H}^{s}(\Omega )\hookrightarrow L^{2}(\Omega ) \hookrightarrow \mathcal{H}^{-s}(\Omega ). $$
(2.8)

By a Cartesian product, we have the following chain of Sobolev spaces

$$ \bigl(\mathcal{H}^{s}(\Omega )\bigr)^{2}\hookrightarrow \bigl(L^{2}(\Omega )\bigr)^{2} \hookrightarrow \bigl( \mathcal{H}^{-s}(\Omega )\bigr)^{2}. $$
(2.9)

2.2 Weighted Sobolev spaces

To set the weak solution for problem (1.5), it is useful to establish the following weighted space [16]

$$ X^{s} \bigl({D}^{+} \bigr)= \biggl\{ Y \in H_{\mathrm{loc}}^{1} \bigl( \mathcal{D}^{+} \bigr): \int _{{D}^{+}} t^{1-2 s} \bigl\vert \nabla Y(x, t) \bigr\vert ^{2} \,dx \,dt< +\infty \biggr\} , $$
(2.10)

equipped with the norm

$$ \Vert Y \Vert _{X^{s} ({D}^{+} )}:= \biggl( \int _{{D}^{+}} t^{1-2 s} \bigl\vert \nabla Y(x, t) \bigr\vert ^{2} \,dx \,dt \biggr)^{\frac{1}{2}}. $$
(2.11)

Hence, the space of all functions in \(X^{s} ({D}^{+} )\), whose trace over \(\mathbb{R}^{N}\) vanishes outside of Ω, is given by

$$ X_{\Omega }^{s} \bigl({D}^{+} \bigr)= \bigl\{ Y \in X^{s} \bigl({D}^{+} \bigr): Y|_{\mathbb{R}^{N} \times \{0\}}= 0 \text{ in } \mathbb{R}^{N} \backslash \Omega \bigr\} ,$$
(2.12)

which furnishes a precise meaning of the solutions to problem (1.5) in a bounded domain Ω. It is clear that \(\mathcal{H}^{s}(\Omega )= \{ Y | _{\Omega \times \{0\}}: Y \in X_{\Omega }^{s} (D^{+} ) \} \). In addition, we have the following compact embedding.

Lemma 2.1

Let \(1 \leq p<2_{s}^{\#}=\frac{2 N}{N-2s} \). Then, \(\operatorname{Tr}_{ \Omega } (X_{\Omega }^{s} (D^{+} ) )\) is compactly embedded in \(L^{p}(\Omega )\).

Remark 2.1

For a function \(Y\in X_{\Omega }^{s} (D^{+} )\), the operator \(\operatorname{Tr}_{\Omega }: X_{\Omega }^{s} (D^{+} ) \rightarrow \mathcal{H}^{s}(\Omega )\) is called the trace operator and satisfies

$$ \Vert \operatorname{Tr}_{\Omega }Y \Vert _{\mathcal{H}^{s}( \Omega )} \leq \delta \Vert Y \Vert _{X_{\Omega }^{s} (D^{+} )},\quad \delta >0. $$
(2.13)

Furthermore, \({\operatorname{Tr}_{\Omega }}Y=Y(x,0)=y(x)\) is the trace of Y onto \(\Omega \times \{0\}\).

2.3 Eigenvalue problem

In this subsection, we state some results given in [36] concerning the eigenvalue problem for the following fractional elliptic equation

$$ \textstyle\begin{cases} (-\Delta )^{s} y=\lambda y & \text{in } \Omega , \\ y=0 & \text{in } \mathbb{R}^{N} \backslash \Omega . \end{cases} $$
(2.14)

Theorem 2.1

([36])

The first eigenvalue of problem (2.14) is positive and can be characterized as follows:

$$ \lambda =\min_{Y \in X_{\Omega }^{s} (D^{+} )} \int _{D^{+}} t^{1-2 s} \nabla Y \cdot \nabla Y \,dx \,dt,\quad \bigl\Vert Y(x,0) \bigr\Vert _{L^{2}( \Omega )}=1, $$
(2.15)

or equivalently,

$$ \lambda =\min_{Y \in X^{s}_{\Omega } (D^{+} )} \frac{\int _{D^{+}} t^{1-2 s} \nabla Y \cdot \nabla Y \,dx \,dt}{\int _{\Omega } \vert Y(x, 0) \vert ^{2} \,dx}, \quad Y(x,0)\neq 0. $$
(2.16)

3 Scalar case

For \(a>0\), consider the following system:

$$\begin{aligned} \textstyle\begin{cases} (-\Delta )^{s}y(x)=a y(x)+f(x) &\text{in}\ \Omega , \\ y(x)=0 &\text{on}\ \mathbb{R}^{N}\backslash \Omega . \end{cases}\displaystyle \end{aligned}$$
(3.1)

Using the extension problem, the nonlocal problem (3.1) is reformulated in a local way as follows:

$$ \textstyle\begin{cases} \nabla (t^{1-2s}\nabla Y(x,t))=0 & \text{in }{D}^{+}, \\ Y(x, 0)=0 & x \in \mathbb{R}^{N} \backslash \Omega , \\ -\frac{1}{k_{s}} \lim_{t \rightarrow 0^{+}} t^{1-2s} \frac{\partial Y}{\partial t}(x, t)=a \operatorname{Tr}_{\Omega }Y(x,t)+f(x) & x \in \Omega . \end{cases} $$
(3.2)

3.1 Weak solution

Multiplying the first equation in (3.2) by a test function \(\phi (x,t) \in X_{\Omega }^{s} ({D}^{+} )\) and integrating over \({D}^{+}\) we obtain

$$ \int _{{D}^{+}}\nabla \cdot\bigl(t^{1-2s}\nabla Y\bigr)\phi (x,t)\,dx \,dt=0. $$
(3.3)

Applying Green’s formula, we have

$$\begin{aligned} \int _{D^{+}}\bigl(t^{1-2s}\nabla Y\bigr)\nabla \phi (x,t)\,dx \,dt =& \int _{ \Omega \times \{0\}}\frac{\partial Y}{\partial \nu }\operatorname{Tr}_{ \Omega } \phi (x,t) \,dx \end{aligned}$$
(3.4)
$$\begin{aligned} =&k_{s} \int _{\Omega \times \{0\}}\bigl(a \operatorname{Tr}_{\Omega }Y(x,t)+f(x) \bigr) \operatorname{Tr}_{\Omega }\phi (x,t) \,dx. \end{aligned}$$
(3.5)

Take the bilinear form \(a(Y,\phi )\) as follows:

$$ a(Y,\phi )=\frac{1}{k_{s}} \int _{D^{+}}t^{1-2s}\nabla Y\nabla \phi (x,t)\,dx \,dt-a \int _{\Omega \times \{0\}}\bigl(\operatorname{Tr}_{\Omega }Y(x,t) \bigr) \operatorname{Tr}_{\Omega }\phi (x,t) \,dx. $$
(3.6)

Also, take the linear form \(F(\phi )\) as follows:

$$ F(\phi )= \int _{\Omega \times \{0\}}f(x)\operatorname{Tr}_{\Omega } \phi (x,t) \,dx. $$
(3.7)

Lemma 3.1

If \(\lambda >a\ k_{s}\), the bilinear form \(a(Y,\phi )\) defined in (3.6) is coercive.

Proof

Replacing \(\phi (x,t)\) by \(Y(x,t)\) in (3.6) we obtain

$$ a(Y,Y)=\frac{1}{k_{s}} \int _{D^{+}}t^{1-2s} \vert \nabla Y \vert ^{2} \,dx \,dt-a \int _{\Omega \times \{0\}} \bigl\vert \operatorname{Tr}_{\Omega }Y(x,t) \bigr\vert ^{2} \,dx. $$
(3.8)

Hence, using (2.16) we obtain

$$ a(Y,Y)\geq \frac{1}{k_{s}} \biggl(1-\frac{a\ k_{s}}{\lambda } \biggr) \int _{D^{+}}t^{1-2s} \vert \nabla Y \vert ^{2} \,dx \,dt. $$
(3.9)

Then, for \(\lambda >a\ k_{s}\) the coerciveness condition is satisfied, i.e.,

$$ a(Y,Y)\geq C_{1} \Vert Y \Vert ^{2},\quad C_{1}>0. $$
(3.10)

 □

Remark 3.1

([38])

If \(Y(x,t)\) is a solution of the extended problem (3.2), then the trace function \(y(x)=\operatorname{Tr}_{\Omega }Y(x,t)=Y(x,0)\) will be called a weak solution to problem (3.1).

3.2 The optimality condition

Consider \(L^{2}(\Omega )\) as the space of controls. For a control \(u \in L^{2}(\Omega )\), the state \(Y(u)\) solves the systems

$$ \textstyle\begin{cases} \nabla (t^{1-2s}\nabla Y(u))=0 & \text{in }{D}^{+}, \\ Y(u;x, 0)=0 & x \in \mathbb{R}^{N} \backslash \Omega , \\ \frac{1}{k_{s}}\frac{\partial Y}{\partial \nu }(u)=a \operatorname{Tr}_{\Omega }Y(u)+u & x \in \Omega . \end{cases} $$
(3.11)

For a given \(z_{d}\in L^{2}(\Omega )\) and \(v\in L^{2}(\Omega )\), the cost-functional subject to the systems (3.11) is given by

$$\begin{aligned} \text{min} \quad J(v)=\frac{1}{2} \bigl\Vert \operatorname{Tr}_{\Omega }Y(v)-z_{d} \bigr\Vert ^{2}_{L^{2}(\Omega )}+(\operatorname{\mathbf{N}}v,v)_{L^{2}(\Omega )}, \quad \forall v \in \mathcal{U}_{ad}, \end{aligned}$$

where \(\operatorname{\mathbf{N}} \in \mathcal{L}(L^{2}(\Omega )) \)Footnote 1 is a positive-definite Hermitian operator satisfying

$$ (\operatorname{\mathbf{N}}v,v)\geq \alpha \Vert v \Vert ^{2}_{L^{2}(\Omega )}, \quad \alpha >0. $$
(3.12)

Let v belong to a subset \(\mathcal{U}_{ad}\) of \(L^{2}(\Omega )\) (the set of admissable controls); we assume \(\mathcal{U}_{ad}\) is a closed nonempty subset of \(L^{2}(\Omega )\). Then, the optimal control problem is now

$$ \textstyle\begin{cases} \text{ Finding }u \in \mathcal{U}_{ad}, \\ \text{ such that}\ J(u)\leq J(v),\quad \forall v\in \mathcal{U}_{ad}. \end{cases} $$
(3.13)

Theorem 3.1

If the cost functional is given by (3.12) and the condition (3.13) is satisfied, then there exists a unique optimal control \(u\in \mathcal{U}_{ad}\). Moreover, this control is characterized by the following equations:

$$\begin{aligned} \textstyle\begin{cases} \nabla \cdot(t^{1-2s}\nabla P )=0 & \textit{in} \ D^{+}, \\ P(x,0)=0 & x \in \mathbb{R}^{N} \backslash \Omega , \\ (\frac{\partial P}{\partial \nu },\operatorname{Tr}_{\Omega }Y)=( \operatorname{Tr}_{\Omega }P,\frac{\partial Y}{\partial \nu })&\textit{on} \ \Omega \times \{0\}, \end{cases}\displaystyle \end{aligned}$$
(3.14)

together with

$$\begin{aligned} (\operatorname{Tr}_{\Omega }P+\operatorname{\mathbf{N}}u,v-u)\geq 0,\quad \forall v\in \mathcal{U}_{ad}, \end{aligned}$$
(3.15)

where \(P\in X_{\Omega }^{s} (D^{+} )\) is the adjoint state.

Proof

The control \(u\in \mathcal{U}_{ad}\) is optimal, if and only if

$$ J^{\prime }(u)\cdot(v-u)\geq 0,\quad \forall v\in \mathcal{U}_{ad},$$
(3.16)

and hence, via an explicit computation of \(J^{\prime }(u)\), (3.17) is equivalent to [24]

$$ \bigl(\operatorname{Tr}_{\Omega }Y(u)-z_{d}, \operatorname{T}_{\Omega }Y(v)- \operatorname{Tr}_{\Omega }Y(u) \bigr)+(\operatorname{\mathbf{N}} u,v-u) \geq 0,\quad \forall v\in \mathcal{U}_{ad}. $$
(3.17)

In order to transform (3.18) into a more convenient form, we introduce the adjoint state P defined by \((\operatorname{\mathbf{A}}Y,P)=(Y, \operatorname{\mathbf{A}}^{*}P)\), where \(A^{*}\) is the adjoint of A. Now

$$ (\operatorname{\mathbf{A}}Y,P)=\bigl(\nabla \cdot\bigl(t^{1-2s}\nabla Y \bigr),P\bigr)= \int _{D^{+}} \nabla \cdot\bigl(t^{1-2s}\nabla Y \bigr)P(x,t) \,dx \,dt. $$
(3.18)

By applying Green’s formula, (3.19) is transformed to

$$\begin{aligned} (\operatorname{\mathbf{A}}Y,P) =& \int _{D^{+}}\nabla \cdot\bigl(t^{1-2s} \nabla P \bigr)Y(x,t)\,dx \,dt- \int _{\Omega \times \{0\}}\operatorname{Tr}_{ \Omega }Y \frac{\partial P}{\partial \nu }\,dx+ \int _{\Omega \times \{0\}} \operatorname{Tr}_{\Omega }P \frac{\partial Y}{\partial \nu }\,dx \\ =&\bigl(Y,\operatorname{\mathbf{A}}^{*}P\bigr), \end{aligned}$$

and hence (3.15) is satisfied.

Take the adjoint systems as follows:

$$\begin{aligned} \textstyle\begin{cases} \nabla \cdot(t^{1-2s}\nabla P)=0 & \text{in} \ D^{+}, \\ P(x,0)=0 & x \in \mathbb{R}^{N} \backslash \Omega , \\ \frac{1}{k_{s}}\frac{\partial P(u)}{\partial \nu }-a \operatorname{Tr}_{\Omega }P(u)=\operatorname{Tr}_{\Omega }Y(u)-z_{d} &\text{on} \ \Omega \times \{0\}. \end{cases}\displaystyle \end{aligned}$$
(3.19)

Remark 3.2 The variational form of (3.21) is

$$ a^{*}(P,\phi )=\bigl(\operatorname{Tr}_{\Omega }Y(u)-z_{d}, \operatorname{Tr}_{\Omega }\phi \bigr), $$
(3.20)

where we define

$$ a^{*}(\psi ,\phi )=a(\phi ,\psi ). $$
(3.21)

Then, (3.18) is equivalent to

$$\begin{aligned} \biggl(\frac{1}{k_{s}}\frac{\partial P}{\partial \nu }-a \operatorname{Tr}_{\Omega }P ,\operatorname{Tr}_{\Omega }Y(v)- \operatorname{Tr}_{\Omega }Y(u) \biggr)+(\operatorname{\mathbf{N}} u,v-u) \geq 0,\quad \forall v\in \mathcal{U}_{ad}. \end{aligned}$$

Hence, using the last condition in (3.15), we obtain

$$\begin{aligned}& \frac{1}{k_{s}} \biggl(\operatorname{Tr}_{\Omega }P , \frac{\partial Y(v)}{\partial \nu }-\frac{\partial Y(u)}{\partial \nu } \biggr)-a \bigl( \operatorname{Tr}_{\Omega }P,\operatorname{Tr}_{ \Omega }Y(v)- \operatorname{Tr}_{\Omega }Y(u)\bigr) \\& \quad {}+(\operatorname{\mathbf{N}} u,v-u)\geq 0,\quad \forall v\in \mathcal{U}_{ad}. \end{aligned}$$

By using (3.11), we have

$$\begin{aligned}& \bigl(\operatorname{Tr}_{\Omega }P ,a\operatorname{Tr}_{\Omega }Y(v)+v-a \operatorname{Tr}_{\Omega }Y(u)-u \bigr)-a \bigl( \operatorname{Tr}_{ \Omega }P,\operatorname{Tr}_{\Omega }Y(v)- \operatorname{Tr}_{\Omega }Y(u)\bigr) \\& \quad {}+(\operatorname{\mathbf{N}} u,v-u)\geq 0,\quad \forall v\in \mathcal{U}_{ad}, \end{aligned}$$

and hence, the optimality condition becomes

$$\begin{aligned} (\operatorname{Tr}_{\Omega }P,v-u)+(\operatorname{\mathbf{N}}u,v-u) \geq 0,\quad \forall v\in \mathcal{U}_{ad}. \end{aligned}$$
(3.22)

Thereby, the proof is completed. □

4 \(2\times 2\) cooperative system

In this section, we generalize the results obtained in the previous section to a \(2\times 2\) cooperative system. This section is divided into two subsections. In Sect. 4.1, we prove the existence and uniqueness of the weak solution by using the Lax–Miligram Lemma. The optimality condition is obtained in Sect. 4.2.

4.1 The weak solution

To obtain a weak solution of the systems (1.5), we first transform (1.5) into a weak form. Indeed, multiplying the first and second equations in (1.5) by a test function \(\phi (x,t)=\{\phi _{1}(x,t),\phi _{2}(x,t)\} \in (X_{\Omega }^{s}(D^{+}) )^{2}\) and integrating over \(D^{+}\) we obtain

$$\begin{aligned} \int _{D^{+}}\nabla \cdot\bigl(t^{1-2s}\nabla Y_{1}\bigr)\phi _{1}(x,t)\,dx\,dt=0, \end{aligned}$$
(4.1)
$$\begin{aligned} \int _{D^{+}}\nabla \cdot\bigl(t^{1-2s}\nabla Y_{2}\bigr)\phi _{2}(x,t)\,dx\,dt=0. \end{aligned}$$
(4.2)

Applying Green’s formula, we have

$$\begin{aligned} \int _{D^{+}}t^{1-2s}\nabla Y_{1}\nabla \phi _{1}(x,t)\,dx\,dt=- \int _{ \Omega \times \{0\}} \operatorname{Tr}_{\Omega }\phi _{1}(x,t)\lim_{t \to 0^{+}}t^{1-2s} \frac{\partial Y_{1}}{\partial t}\,dx, \end{aligned}$$
(4.3)
$$\begin{aligned} \int _{D^{+}}t^{1-2s}\nabla Y_{2}\nabla \phi _{2}(x,t)\,dx\,dt=- \int _{ \Omega \times \{0\}} \operatorname{Tr}_{\Omega }\phi _{2}(x,t)\lim_{t \to 0^{+}}t^{1-2s} \frac{\partial Y_{2}}{\partial t}\,dx. \end{aligned}$$
(4.4)

Then, we obtain

$$\begin{aligned} \int _{D^{+}}t^{1-2s}\nabla Y_{1}\nabla \phi _{1}(x,t)\,dx\,dt= \int _{\Omega \times \{0\}} \operatorname{Tr}_{\Omega }\phi _{1}(x,t) \frac{\partial Y_{1}}{\partial \nu }\,dx, \end{aligned}$$
(4.5)
$$\begin{aligned} \int _{D^{+}}t^{1-2s}\nabla Y_{2}\nabla \phi _{2}(x,t)\,dx\,dt= \int _{\Omega \times \{0\}} \operatorname{Tr}_{\Omega }\phi _{2}(x,t) \frac{\partial Y_{2}}{\partial \nu }\,dx, \end{aligned}$$
(4.6)

By using the systems (1.5), Eqs. (4.5) and (4.6) are equivalent to

$$\begin{aligned} \frac{1}{k_{s} b} \int _{D^{+}}t^{1-2s}\nabla Y_{1}\nabla \phi _{1}(x,t)\,dx \,dt= \int _{\Omega \times \{0\}} \biggl(f_{1}+\frac{a}{b}y_{1}+y_{2} \biggr) \operatorname{Tr}_{\Omega }\phi _{1}(x,t)\,dx, \end{aligned}$$
(4.7)
$$\begin{aligned} \frac{1}{k_{s} c} \int _{D^{+}}t^{1-2s}\nabla Y_{2}\nabla \phi _{2}(x,t)\,dx \,dt= \int _{\Omega \times \{0\}} \biggl(f_{1}+y_{1}+ \frac{d}{c}y_{2}\biggr) \operatorname{Tr}_{\Omega } \phi _{2}(x,t)\,dx. \end{aligned}$$
(4.8)

To this end, we can define a bilinear form on \((X_{\Omega }^{s}(D^{+}) )^{2}\) as follows:

$$\begin{aligned} a(Y,\phi ) =&\frac{1}{k_{s} b} \int _{D^{+}}t^{1-2s}\nabla Y_{1} \nabla \phi _{1}(x,t)\,dx\,dt+\frac{1}{k_{s} c} \int _{D^{+}}t^{1-2s}\nabla Y_{1}\nabla \phi _{1}(x,t)\,dx \,dt \\ &{}- \int _{\Omega \times \{0\}} \biggl(\frac{a}{b} y_{1}+y_{2} \biggr) \operatorname{Tr}_{\Omega }\phi _{1}(x,t)\,dx- \int _{\Omega \times \{0\}} \biggl(y_{1}+\frac{c}{d} y_{2}\biggr)\operatorname{Tr}_{\Omega }\phi _{2}(x,t)\,dx. \end{aligned}$$

Also, we can define a linear form as follows:

$$ F(\phi )= \int _{\Omega \times \{0\}} \bigl(f_{1}(x)\operatorname{Tr}_{ \Omega } \phi _{1}(x,t)+f_{2}(x)\operatorname{Tr}_{\Omega } \phi _{2}(x,t)\bigr)\,dx, \quad \forall \phi \in \bigl(X_{\Omega }^{s} \bigl(D^{+}\bigr) \bigr)^{2}. $$
(4.9)

Lemma 4.1

The bilinear form (4.9) is coercive and bounded.

Proof

Replacing \(\phi =\{\phi _{1},\phi _{2}\}\) by \(Y=\{Y_{1},Y_{2} \}\) in (3.9) yields

$$\begin{aligned} a(Y,Y) =&\frac{1}{k_{s} b} \int _{D^{+}}t^{1-2s} \vert \nabla Y_{1} \vert ^{2}\,dx\,dt+ \frac{1}{k_{s} c} \int _{D^{+}}t^{1-2s} \vert \nabla Y_{2} \vert ^{2}\,dx\,dt \\ &{}-\frac{a}{b} \int _{\Omega \times \{0\}} \vert \operatorname{Tr}_{\Omega }Y_{1} \vert ^{2}\,dx- \frac{d}{c} \int _{\Omega \times \{0\}} \vert \operatorname{Tr}_{\Omega }Y_{2} \vert ^{2}\,dx \\ &{}-2 \int _{\Omega \times \{0\}}\operatorname{Tr}_{\Omega }Y_{1} \operatorname{Tr}_{\Omega }Y_{2}\,dx. \end{aligned}$$

By using the Cauchy–Schwarz inequality, we have

$$\begin{aligned} a(Y,Y) \geq &\frac{1}{k_{s} b} \int _{D^{+}}t^{1-2s} \vert \nabla Y_{1} \vert ^{2}\,dx \,dt+\frac{1}{k_{s} c} \int _{D^{+}}t^{1-2s} \vert \nabla Y_{2} \vert ^{2}\,dx\,dt \\ &{}-\frac{a}{b} \int _{\Omega \times \{0\}} \vert \operatorname{Tr}_{\Omega }Y_{1} \vert ^{2}\,dx- \frac{d}{c} \int _{\Omega \times \{0\}} \vert \operatorname{Tr}_{\Omega }Y_{2} \vert ^{2}\,dx \\ &{}-2 \biggl( \int _{\Omega \times \{0\}} \vert \operatorname{Tr}_{\Omega }Y_{1} \vert ^{2}\,dx \biggr)^{\frac{1}{2}} \biggl( \int _{\Omega \times \{0\}} \vert \operatorname{Tr}_{\Omega }Y_{1} \vert ^{2}\,dx \biggr)^{\frac{1}{2}}, \end{aligned}$$

from (2.15), we deduce

$$\begin{aligned} a(Y,Y) \geq &\frac{1}{k_{s} b} \biggl(1-\frac{ak_{s}}{\lambda } \biggr) \Vert \nabla Y_{1} \Vert ^{2}+\frac{1}{k_{s} c} \biggl(1- \frac{d k_{s}}{\lambda } \biggr) \Vert \nabla Y_{2} \Vert ^{2}-\frac{2}{\lambda } \Vert Y_{1} \Vert \Vert Y_{2} \Vert \\ =&\frac{1}{\lambda } \biggl[ \biggl(\frac{1}{k_{s} b}(\lambda -ak_{s})- \frac{ \Vert Y_{2} \Vert }{ \Vert Y_{1} \Vert } \biggr) \Vert \nabla Y_{1} \Vert ^{2}+ \biggl( \frac{1}{k_{s} c}( \lambda -d k_{s})-\frac{ \Vert Y_{1} \Vert }{ \Vert Y_{2} \Vert } \biggr) \Vert \nabla Y_{2} \Vert ^{2} \biggr]. \end{aligned}$$

Take

$$ C_{2}= \biggl(\frac{1}{k_{s} b}(\lambda -ak_{s})- \frac{ \Vert Y_{2} \Vert }{ \Vert Y_{1} \Vert } \biggr), $$

and

$$ C_{3}= \biggl(\frac{1}{k_{s} c}(\lambda -d k_{s})- \frac{ \Vert Y_{1} \Vert }{ \Vert Y_{2} \Vert } \biggr). $$

Then, if \(C_{2},C_{3}\geq 0\),we have

$$\begin{aligned} a(Y,Y) \geq &\text{min}\{C_{2},C_{3}\} \bigl[ \Vert Y_{1} \Vert ^{2}_{X_{ \Omega }^{s}(D^{+})}+ \Vert Y_{2} \Vert ^{2}_{X_{\Omega }^{s}(D^{+})} \bigr], \end{aligned}$$
(4.10)

or

$$ a(Y,Y)\geq C \Vert Y \Vert ^{2}_{ (X_{\Omega }^{s}(D^{+}) )^{2}},\quad C=\text{min}\{C_{2},C_{3}\}. $$
(4.11)

Hence, the bilinear form \(a(Y,Y)\) is coercive, if and only if the following conditions are satisfied

$$\begin{aligned} \textstyle\begin{cases} (\lambda >a k_{s},\qquad \lambda > d k_{s}, \\ (\lambda -a k_{s})(\lambda -d k_{s})\geq (k_{s})^{2} bc. \end{cases}\displaystyle \end{aligned}$$
(4.12)

 □

4.2 The optimality conditions

The control-problem formulation is the main target of this work. For the control problem, we construct the adjoint state. Furthermore, we originate the conditions of optimality via the Lions technique [24, 25]. This subsection consists of two parts. Section 4.2.1 contains the derivation of the necessary and sufficient conditions for fractional optimal control. Meanwhile, the equivalence extended optimal control is obtained in Sect. 4.2.2.

4.2.1 Fractional optimal control

Consider \((L^{2}(\Omega ))^{2}\) as the space of controls. For a control \(u= \{u_{1},u_{2}\} \in (L^{2}(\Omega ))^{2}\), the state \(y(u)=\{y_{1}(u),y_{2}(u)\}\) solves the following system:

$$\begin{aligned} \textstyle\begin{cases} (-\Delta )^{s}y_{1}(u)=a y_{1}(u)+b y_{2}(u)+u_{1} &\text{in}\ \Omega , \\ (-\Delta )^{s}y_{2}(u)=c y_{1}(u)+d y_{2}(u)+u_{2}& \text{in} \ \Omega , \\ y_{1}(u)=y_{2}(u)=0&\text{in} \ \mathbb{R}^{N}\backslash \Omega . \end{cases}\displaystyle \end{aligned}$$
(4.13)

The observation equations are given by

$$ \mathbf{y}_{i}(u)=y_{i}(u),\quad i=1,2. $$
(4.14)

For a given \(z_{d}=\{z_{1d},z_{2d}\}\in (L^{2}(\Omega ))^{2}\) and \(v=\{v_{1},v_{2} \}\in (L^{2}(\Omega ))^{2}\), the cost-functional subject to the systems (4.1) is given by

$$ J(v)=\frac{1}{2} \bigl\Vert y_{1}(v)-z_{1d} \bigr\Vert ^{2}_{L^{2}(\Omega )}+\frac{1}{2} \bigl\Vert y_{2}(v)-z_{2d} \bigr\Vert ^{2}_{L^{2}(\Omega )}+ (\operatorname{\mathbf{N}}v,v)_{(L^{2}( \Omega ))^{2}}, $$
(4.15)

where \(\operatorname{\mathbf{N}} \in \mathcal{L}((L^{2}(\Omega ))^{2})\) is a positive-definite Hermitian operator satisfying

$$ (\operatorname{\mathbf{N}}v,v)\geq \alpha \Vert v \Vert ^{2}_{(L^{2}(\Omega ))^{2}}, \quad \alpha >0. $$
(4.16)

Let \(\mathcal{U}_{ad}\) be a closed and convex subset of \(L^{2}(\Omega ) \). Then, the control problem is given as follows:

$$ \textstyle\begin{cases} \text{ Finding }u \in (\mathcal{U}_{ad})^{2}, \\ \text{ such that}\ J(u)\leq J(v),\quad \forall v\in (\mathcal{U}_{ad})^{2}. \end{cases} $$
(4.17)

Theorem 4.1

If the cost functional is given by (4.3), and the condition (4.4) is satisfied, then there exists a unique optimal control \(u\in (\mathcal{U}_{ad})^{2}\). Moreover, this control is characterized by the following equations:

$$\begin{aligned} \textstyle\begin{cases} (-\Delta )^{s}p_{1}(u)-ap_{1}(u)-c p_{2}(u)=y_{1}(u)-z_{1d}& \textit{in}\ \Omega , \\ (-\Delta )^{s}p_{2}(u)-bp_{1}(u)-d p_{2}(u)=y_{2}(u)-z_{2d}& \textit{in}\ \Omega , \\ p_{1}(u)=p_{2}(u)=0 &\textit{on}\ \mathbb{R}^{N}\backslash \Omega . \end{cases}\displaystyle \end{aligned}$$
(4.18)

In addition,

$$ (p_{1},v_{1}-u_{1})+(p_{2},v_{2}-u_{2})+( \operatorname{\mathbf{N}} u,v-u)_{(L^{2}( \Omega )^{2}}\geq 0, \quad \forall v\in ( \mathcal{ U}_{ad})^{2},$$
(4.19)

where \(p=\{p_{1},p_{2}\}\in (\mathcal{H}^{s}(\Omega ))^{2}\) is the adjoint state.

Proof

Since \(\operatorname{\mathbf{N}}>0\), then the cost functional (4.3) is strictly convex. Furthermore, the set \(\mathcal{ U}_{ad}\) is nonempty, closed, bounded and convex in \(L^{2}(\Omega )\). Therefore, the existence and uniqueness of the optimal control is proved.

The control \(u=\{u_{1},u_{2}\}\in (\mathcal{U}_{ad})^{2}\) is optimal, if and only if

$$ J^{\prime }(u)\cdot(v-u)\geq 0,\quad \forall v\in (\mathcal{U}_{ad})^{2},$$
(4.20)

which is equivalent to [24]

$$\begin{aligned}& { \bigl(y_{1}(u)-z_{1d},y_{1}(v)-y_{1}(u) \bigr)+\bigl(y_{2}(u)-z_{2d},y_{2}(v)-y_{2}(u) \bigr)} \\& \quad {}+ (\operatorname{\mathbf{N}} u_{1},v_{1}-u_{1} )+( \operatorname{\mathbf{N}}u_{2},v_{2}-u_{2}) \geq 0,\quad \forall v \in (\mathcal{U}_{ad})^{2}. \end{aligned}$$
(4.21)

Now, since \((\operatorname{\mathbf{A}}y,p)=(y, \operatorname{\mathbf{A}}^{*}p)\), where

$$ \mathbf{A}\bigl(y=\{y_{1},y_{2}\}\bigr)=\bigl\{ (-\Delta )^{s}y_{1}-a y_{1}-b y_{2}, (-\Delta )^{s}y_{2}-c y_{1}-d y_{2}\bigr\} , $$
(4.22)

then

$$\begin{aligned} (\operatorname{\mathbf{A}}y,p) =& \bigl((-\Delta )^{s}y_{1}-a y_{1}-b y_{2},p_{1} \bigr)+ \bigl((- \Delta )^{s}y_{2}-c y_{1}-d y_{2},p_{2} \bigr) \\ =&\bigl((-\Delta )^{s}y_{1},p_{1} \bigr)-a( y_{1},p_{1})-b( y_{2},p_{1})+ \bigl((- \Delta )^{s}y_{2},p_{2}\bigr)-c( y_{1},p_{2}) \\ &{}-b( y_{2},p_{2}) \\ =&\bigl(y_{1},(-\Delta )^{s}p_{1} \bigr)-a( y_{1},p_{1})-b( y_{2},p_{1})+ \bigl(y_{2},(- \Delta )^{s}p_{2}\bigr)-c( y_{1},p_{2}) \\ &{}-b( y_{2},p_{2}) \\ =&(\bigl(y_{1},(-\Delta )^{s}p_{1}-a p_{1}-c p_{2}\bigr)+(\bigl(y_{2},(-\Delta )^{s}p_{2}-b p_{1}-d p_{2}\bigr) \\ =&\bigl(y,\operatorname{\mathbf{A}}^{*}p\bigr), \end{aligned}$$

where A is the adjoint operator of A, p is the adjoint state.

Take the adjoint system as follows:

$$\begin{aligned} \textstyle\begin{cases} (-\Delta )^{s}p_{1}(u)-ap_{1}(u)-c p_{2}(u)=y_{1}(u)-z_{1d}& \text{in}\ \Omega , \\ (-\Delta )^{s}p_{2}(u)-bp_{1}(u)-d p_{2}(u)=y_{2}(u)-z_{2d}& \text{in}\ \Omega , \\ p_{1}(u)=p_{2}(u)=0 &\text{on}\ \mathbb{R}^{N}\backslash \Omega . \end{cases}\displaystyle \end{aligned}$$
(4.23)

By using Eqs. (4.1) and (4.12), we deduce

$$\begin{aligned}& { (p_{1},v_{1}-u_{1} )+(p_{2},v_{2}-u_{2})+( \operatorname{\mathbf{N}}u_{1},v_{1}-u_{1})_{L^{2}(\Omega )}} \\& \quad {}+{ (\operatorname{\mathbf{N}} u_{2},v_{2}-u_{2} )_{L^{2}( \Omega )}\geq 0}, \quad \forall v\in (\mathcal{ U}_{ad})^{2}. \end{aligned}$$
(4.24)

Thus, the proof is completed. □

4.2.2 Extended optimal control

If \(y(v)\in (\mathcal{H}^{s}(\Omega ))^{2}\) is a solution of (4.1) with \(v= \{v_{1},v_{2}\}\in (\mathcal{H}^{-s}(\Omega ))^{2}\) and \(Y(v)\in (X_{\Omega }^{s}(D^{+}) )^{2}\) solves the following systems:

$$\begin{aligned} \textstyle\begin{cases} \nabla \cdot(t^{1-2s}\nabla Y_{1}(u))=0 & \text{in} \ D^{+}, \\ \nabla \cdot(t^{1-2s}\nabla Y_{2}(u))=0 & \text{in} \ D^{+} , \\ Y_{1}(x,0)=Y_{2}(x,0)=0 & \text{in}\ \mathbb{R}^{N}\backslash \Omega , \\ \frac{1}{k_{s}}\frac{\partial Y_{1}(u)}{\partial \nu }=(u_{1}+a y_{1}(u)+b y_{2}(u))& \text{on}\ \Omega \times \{0\}, \\ \frac{1}{k_{s}}\frac{\partial Y_{2}(u)}{\partial \nu }=(u_{2}+c y_{1}(u)+d y_{2}(u))& \text{on}\ \Omega \times \{0\}, \end{cases}\displaystyle \end{aligned}$$
(4.25)

then, we have

$$ \operatorname{Tr}_{\Omega }Y(v)=y(v). $$
(4.26)

Hence, the equivalence extended optimal control problem is given by

$$\begin{aligned} \text{min} \quad \mathbb{J}(v) =&\frac{1}{2} \bigl\Vert \operatorname{Tr}_{\Omega }Y_{1}(v)-z_{1d} \bigr\Vert ^{2}_{L^{2}(\Omega )}+\frac{1}{2} \bigl\Vert \operatorname{Tr}_{\Omega }Y_{2}(v)-z_{2d} \bigr\Vert ^{2}_{L^{2}(\mathcal{B})} \\ &{}+(\operatorname{\mathbf{N}}v,v)_{(L^{2}(\Omega ))^{2}}, \quad \forall v \in ( \mathcal{U}_{ad})^{2}. \end{aligned}$$

Theorem 4.2

If the cost functional is given by (4.16) and the condition (4.4) is satisfied, then there exists a unique optimal control \(u=\{u_{1},u_{2}\}\in ( \mathcal{U}_{ad})^{2}\). Moreover, this control is characterized by the following equations:

$$\begin{aligned} \textstyle\begin{cases} \nabla \cdot(t^{1-2s}\nabla P_{1})=0 & \textit{in} \ D^{+}, \\ \nabla \cdot(t^{1-2s}\nabla P_{2})=0 & \textit{in} \ D^{+}, \\ P_{1}(x,0)=P_{2}(x,0)=0 & \textit{in}\ \mathbb{R}^{N}\backslash \Omega , \\ (\frac{\partial P_{1}}{\partial \nu },\operatorname{Tr}_{\Omega }Y_{1})=( \operatorname{Tr}_{\Omega }P_{1},\frac{\partial Y_{1}}{\partial \nu })&\textit{on} \ \Omega \times \{0\}, \\ (\frac{\partial P_{2}}{\partial \nu },\operatorname{Tr}_{\Omega }Y_{2})=( \operatorname{Tr}_{\Omega }P_{2},\frac{\partial Y_{2}}{\partial \nu })& \textit{on}\ \Omega \times \{0\}. \end{cases}\displaystyle \end{aligned}$$
(4.27)

In addition,

$$\begin{aligned} (\operatorname{Tr}_{\Omega }P_{1},v_{1}-u_{1})+( \operatorname{Tr}_{ \Omega }P_{2},v_{2}-u_{2})+( \operatorname{\mathbf{N}}u,v-u)\geq 0,\quad \forall v\in (\mathcal{U}_{ad})^{2}, \end{aligned}$$
(4.28)

where \(P=\{P_{1},P_{2}\}\in (X_{\Omega }^{s}(D^{+}) )^{2}\) is the adjoint state.

Proof

The control \(u\in (\mathcal{U}_{ad})^{2}\) is optimal, if and only if

$$ \mathbb{J}^{\prime }(u)\cdot(v-u)\geq 0,\quad \forall v\in ( \mathcal{U}_{ad})^{2},$$
(4.29)

which is again equivalent to [24]

$$\begin{aligned}& { \bigl(\operatorname{Tr}_{\Omega }Y_{1}(v)-z_{1d}, \operatorname{Tr}_{ \Omega }Y_{1}(v)-\operatorname{Tr}_{\Omega }Y_{1}(u) \bigr)+\bigl( \operatorname{Tr}_{\Omega }Y_{2}(v)-z_{2d}, \operatorname{Tr}_{\Omega }Y_{2}(v)- \operatorname{Tr}_{\Omega }Y_{2}(u) \bigr)} \\& \quad {}+ (\operatorname{\mathbf{N}} u_{1},v_{1}-u_{1} )+( \operatorname{\mathbf{N}}u_{2},v_{2}-u_{2}) \geq 0,\quad \forall v\in ( \mathcal{U}_{ad})^{2}. \end{aligned}$$
(4.30)

Now, since \((\operatorname{\mathbf{A}}Y,P)=(Y, \operatorname{\mathbf{A}}^{*}P)\), then

$$\begin{aligned} (\operatorname{\mathbf{A}}Y,P) =&\bigl(\nabla \cdot\bigl(t^{1-2s}\nabla Y_{1}\bigr),P_{1}\bigr)+\bigl( \nabla \cdot\bigl(t^{1-2s}\nabla Y_{2}\bigr),P_{2}\bigr) \\ =& \int _{D^{+}}\nabla \cdot\bigl(t^{1-2s}\nabla Y_{1}\bigr)P_{1}\,dx\,dt+ \int _{D^{+}}\nabla \cdot\bigl(t^{1-2s}\nabla Y_{2}\bigr)P_{2}\,dx\,dt \\ =& \int _{D^{+}}\nabla \cdot\bigl(t^{1-2s}\nabla P_{1}\bigr)Y_{1}(x,t)\,dx\,dt- \int _{ \Omega }\operatorname{Tr}_{\Omega }Y_{1} \frac{\partial P_{1}}{\partial \nu }+ \int _{\Omega }\operatorname{Tr}_{ \Omega }P_{1} \frac{\partial Y_{1}}{\partial \nu } \\ &{}+ \int _{D^{+}}\nabla \cdot\bigl(t^{1-2s}\nabla P_{2}\bigr)Y_{2}(x,t)\,dx\,dt- \int _{ \Omega }\operatorname{Tr}_{\Omega }Y_{2} \frac{\partial P_{2}}{\partial \nu }+ \int _{\Omega }\operatorname{Tr}_{ \Omega }P_{2} \frac{\partial Y_{2}}{\partial \nu } \\ =&\bigl(Y,\operatorname{\mathbf{A}}^{*}P\bigr), \end{aligned}$$

and hence, (4.17) is satisfied.

Take the adjoint systems as follows:

$$\begin{aligned} \textstyle\begin{cases} \nabla \cdot(t^{1-2s}\nabla P_{1})=0 & \text{in} \ \mathcal{D}^{+}, \\ \nabla \cdot(t^{1-2s}\nabla P_{2})=0& \text{in} \ \mathcal{D}^{+}, \\ P_{1}(x,0)=P_{2}(x,0)=0 & \text{in}\ \mathbb{R}^{N}\backslash \Omega , \\ \frac{1}{k_{s}}\frac{\partial P_{1}}{\partial \nu }-a\operatorname{Tr}_{ \Omega }P_{1}-c\operatorname{Tr}_{\Omega }P_{2}=\operatorname{Tr}_{ \Omega }Y_{1}(v)-z_{1d} &\text{on}\ \Omega \times \{0\} \\ \frac{1}{k_{s}}\frac{\partial P_{2}}{\partial \nu }-b\operatorname{Tr}_{ \Omega }P_{1}-d\operatorname{Tr}_{\Omega }P_{2}=\operatorname{Tr}_{ \Omega }Y_{2}(v)-z_{2d}& \text{on}\ \Omega \times \{0\}. \end{cases}\displaystyle \end{aligned}$$
(4.31)

Then, (4.20) is equivalent to

$$\begin{aligned} & \biggl(\frac{1}{k_{s}}\frac{\partial P_{1}}{\partial \nu }-a \operatorname{Tr}_{\Omega }P_{1}-c \operatorname{Tr}_{\Omega }P_{2}, \operatorname{Tr}_{\Omega } Y_{1}(v)-\operatorname{Tr}_{\Omega }Y_{1}(u) \biggr) \\ &\quad {}+ \biggl(\frac{1}{k_{s}}\frac{\partial P_{2}}{\partial \nu }-b \operatorname{Tr}_{\Omega }P_{1}-d \operatorname{Tr}_{\Omega }P_{2}, \operatorname{Tr}_{\Omega } Y_{2}(v)-\operatorname{Tr}_{\Omega }Y_{2}(u) \biggr) \\ &\quad {}+(\operatorname{\mathbf{N}} u_{1},v_{1}-u_{1})+( \operatorname{\mathbf{N}}u_{2},v_{2}-u_{2}) \geq 0,\quad \forall v \in (\mathcal{U}_{ad})^{2}. \end{aligned}$$

Hence, using the last two conditions in (4.17), the optimality condition becomes

$$ (\operatorname{Tr}_{\Omega }P_{1},v_{1}-u_{1})+( \operatorname{Tr}_{ \Omega }P_{2},v_{2}-u_{2})+( \operatorname{\mathbf{N}}u,v-u)\geq 0,\quad \forall v\in (\mathcal{U}_{ad})^{2}, $$
(4.32)

which completes the proof. □

5 Summary and conclusion

In the present work, we investigate the optimal control problem for \(2 \times 2\) cooperative systems involving the fractional Laplace operator, wherein these systems are subject to the zero Dirichlet condition. Due to the difficulty arising from the nonlocality of the fractional Laplace operator, we follow the Caffarelli and Silvestre technique to extend our problem to local cooperative systems. With the aid of the Lax–Milgram lemma, the existence and uniqueness of the solution to the extended problem are proved. Moreover, the conditions of optimality are proved by the Lions technique for both fractional and extended optimal control. If \(s\rightarrow 1\), the obtained results are similar to the classical results.

Availability of data and materials

The data that support the findings of this study are available from the authors upon request.

Notes

  1. \(\mathcal{L}(X)\) is the space of all bounded and linear operators from X into itself.

References

  1. Antil, H., Otarola, E.: A FEM for an optimal control problem of fractional powers of elliptic operators. SIAM J. Control Optim. 53, 3432–3456 (2015)

    Article  MathSciNet  Google Scholar 

  2. Antil, H., Otrola, E., Salgado, A.J.: Optimization with respect to order in a fractional diffusion model: analysis, approximation and algorithmic aspects. J. Sci. Comput. 77, 204–224 (2018)

    Article  MathSciNet  Google Scholar 

  3. Bisci, G.M., Servadei, R., Radulescu, V.D.: Variational methods for non-local fractional problems (2016)

  4. Caffarelli, L., Silvestre, L.: An extension problem related to the fractional Laplacian. Commun. Partial Differ. Equ. 32, 1245–1260 (2017)

    Article  MathSciNet  Google Scholar 

  5. Ceng, L.C., Fu, Y.X., Yin, J., et al.: The solvability of generalized systems of time-dependent hemivariational inequalities enjoying symmetric structure in reflexive Banach spaces. Symmetry 13, Article ID 1801 (2021)

    Article  Google Scholar 

  6. Ceng, L.C., Liou, Y.C., Wen, C.F., Lo, C.H.: Convergence and some control conditions of hybrid steepest-descent methods for systems of variational inequalities and hierarchical variational inequalities. J. Nonlinear Sci. Appl. 10, 4574–4596 (2017)

    Article  MathSciNet  Google Scholar 

  7. Ceng, L.C., Liou, Y.C., Yao, J.C., Yao, Y.: Well-posedness for systems of time-dependent hemivariational inequalities in Banach spaces. J. Nonlinear Sci. Appl. 10, 4318–4336 (2017)

    Article  MathSciNet  Google Scholar 

  8. Ceng, L.C., Liou, Y.C., Yao, J.C., Yao, Y., Lo, C.H.: Well-posedness for systems of generalized mixed quasivariational inclusion problems and optimization problems with constraints. J. Nonlinear Sci. Appl. 10, 5373–5392 (2017)

    Article  MathSciNet  Google Scholar 

  9. Ceng, L.C., Liu, Z.H., Yao, J.C., Yao, Y.: Optimal control of feedback control systems governed by systems of evolution hemivariational inequalities. Filomat 32, 5205–5220 (2018)

    Article  MathSciNet  Google Scholar 

  10. Ceng, L.C., Wen, C.F., Yao, J.C., Yao, Y.: A system of evolution problems driven by a system of hemivariational inequalities. J. Nonlinear Sci. Appl. 11, 342–357 (2018)

    Article  MathSciNet  Google Scholar 

  11. D’Elia, M., Glusa, C., Otrola, E.: A priori error estimates for the optimal control of the integral fractional Laplacian. SIAM J. Control Optim. 57, 2775–2798 (2019)

    Article  MathSciNet  Google Scholar 

  12. El-Saify, H.A., Serag, H.M., Shehata, M.A.: Time optimal control problem for cooperative hyperbolic systems involving the Laplace operator. J. Dyn. Control Syst. 15, 405–423 (2009)

    Article  MathSciNet  Google Scholar 

  13. Fallahgoul, H.A., Facardi, S.M.: Fractional Calculus and Fractional Process with Applications to Financial Economics: Theory and Application. Elsevier, Amsterdam (2017)

    Google Scholar 

  14. Fleckinger, J., Serag, H.M.: Semilinear cooperative elliptic systems on \(\mathcal{R}^{n}\). Rend. Mat. Appl. 15, 98–108 (1995)

    MATH  Google Scholar 

  15. Gali, I.M., Serag, H.M.: Optimal control of cooperative elliptic systems defined on \(\mathcal{R}^{n}\). J. Egypt. Math. Soc. 3, 33–39 (1995)

    MATH  Google Scholar 

  16. Göldshtein, V., Ukhlov, A.: Weighted Sobolev spaces and embedding theorems. Trans. Am. Math. Soc. 361, 3829–3850 (2009)

    Article  MathSciNet  Google Scholar 

  17. Guo, B., Pu, X., Huang, F.: Fractional Partial Differential Equations and Their Numerical Solutions. world scientific, Beijing (2015)

    Book  Google Scholar 

  18. Hinze, M., Pinnau, R., Ulbrich, M., Ulbrich, S.: Optimization with PDE Constraints. Mathematical Modelling: Theory and Applications, vol. 23. Springer, New York (2009)

    Book  Google Scholar 

  19. Hyder, A., El-Badawy, M.: Distributed control for time fractional differential system involving Schrödinger operator. J. Funct. Spaces 2019, Article ID 1389787 (2019)

    MATH  Google Scholar 

  20. Hyder, A.: White noise theory and general improved Kudryashov method for stochastic nonlinear evolution equations with conformable derivatives. Adv. Differ. Equ. 2020, Article ID 236 (2020)

    Article  MathSciNet  Google Scholar 

  21. Hyder, A., El-Badawy, M.: Fractional optimal control of Navier-Stokes equations. Comput. Mater. Continua 64, 859–870 (2020)

    Article  Google Scholar 

  22. Hyder, A., Soliman, A.H.: Exact solutions of space-time local fractal nonlinear evolution equations: a generalized conformable derivative approach. Results Phys. 17, Article ID 103135 (2020)

    Article  Google Scholar 

  23. Hyder, A., Soliman, A.H.: An extended Kudryashov technique for solving stochastic nonlinear models with generalized conformable derivatives. Commun. Nonlinear Sci. Numer. Simul. 97, Article ID 105730 (2020)

    Article  Google Scholar 

  24. Lions, J.L.: Optimal Control of Systems Governed by Partial Differential Equations. Springer, Berlin (1971)

    Book  Google Scholar 

  25. Lions, J.L.: Some Methods in the Mathematical Analysis of Systems and Their Control. Science Press, Beijing (1981)

    MATH  Google Scholar 

  26. Lions, J.L., Magenes, E.: Non-homogeneous Boundary Value Problems and Applications Vol.I. Springer, New York (1972)

    Book  Google Scholar 

  27. Mophou, G.M.: Optimal control of fractional diffusion equation with state constraints. Comput. Math. Appl. 62, 1413–1426 (2011)

    Article  MathSciNet  Google Scholar 

  28. Mophou, G.M.: Optimal control of fractional diffusion equation. Comput. Math. Appl. 62, 68–78 (2011)

    Article  MathSciNet  Google Scholar 

  29. Qamlo, A.H., Serag, H.M., El-Zahrany, E.A.: Optimal control for non-cooperative parabolic systems with conjugation conditions. Eur. J. Sci. Res. 131, 215–226 (2015)

    Google Scholar 

  30. Serag, H.M.: Optimal control of systems involving Schrödinger operators. Control Intell. Syst. 32, 154–157 (2004)

    MathSciNet  MATH  Google Scholar 

  31. Serag, H.M.: Distributed control for cooperative systems involving parabolic operators with an infinite number of variables. IMA J. Math. Control Inf. 24, 149–161 (2007)

    Article  MathSciNet  Google Scholar 

  32. Serag, H.M., Khafagy, S.: On nonhomogeneous \(n\times n\) elliptic systems involving p-Laplacian with different weights. J. Adv. Res. Differ. Equ. 33, 1–13 (2009)

    Google Scholar 

  33. Serag, H.M., Qamlo, A.H.: Boundary control for non-cooperative elliptic systems. Adv. Model. Anal. 38, 31–42 (2001)

    Google Scholar 

  34. Soliman, A.H., Hyder, A.: Closed-form solutions of stochastic KdV equation with generalized conformable derivatives. Phys. Scr. 95, Article ID 065219 (2020)

    Article  Google Scholar 

  35. Tartar, L.: An introduction to Sobolev spaces and interpolation spaces. Lecture Notes of the Unione Matematica Italiana 3 (2007)

  36. Tian, Y.: Some results on the eigenvalue problem for a fractional elliptic equation. Bound. Value Probl. (2019)

  37. Tröltzsch, F.: Optimal Control of Partial Differential Equations: Theory, Methods, and Applications, Gruduate Studies in Mathematics. Am. Math. Soc., Providence (2010)

    Google Scholar 

  38. Vázquez, Y., Volzone, J.L.: Symmetrization for fractional elliptic and parabolic equations and an isoperimetric application. Chin. Ann. Math., Ser. B 38(2), 661–686 (2017)

    Article  MathSciNet  Google Scholar 

  39. Zhou, Y.: Basic Theory of Fractional Differential Equations. World Scientific, New Jersy (2014)

    Book  Google Scholar 

Download references

Acknowledgements

The authors extend their appreciation to the Deanship of Scientific Research at King Khalid University for funding this work through Research Groups Program under grant RGP.1/68/42.

Funding

This research was funded by King Khalid University, grant RGP.1/68/42.

Author information

Authors and Affiliations

Authors

Contributions

All authors read and approved the final manuscript.

Corresponding author

Correspondence to M. El-Badawy.

Ethics declarations

Competing interests

The authors declare that they have no competing interests.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Serag, H.M., Hyder, AA. & El-Badawy, M. Optimal control for cooperative systems involving fractional Laplace operators. J Inequal Appl 2021, 196 (2021). https://doi.org/10.1186/s13660-021-02727-3

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s13660-021-02727-3

MSC

Keywords