- Research
- Open access
- Published:
Optimal control for a chemotaxis–haptotaxis model in two space dimensions
Boundary Value Problems volume 2022, Article number: 79 (2022)
Abstract
This paper deals with a chemotaxis–haptotaxis model which described the process of cancer invasion on the macroscopic scale. We first explore the global-in-time existence and uniqueness of a strong solution. For a class of cost functionals, we prove first-order necessary optimality conditions for the corresponding optimal control problem and establish the existence of Lagrange multipliers. Finally, we derive some extra regularity for the Lagrange multiplier.
1 Introduction
In this paper, we investigate the chemotaxis–haptotaxis model with the initial-boundary conditions
where \(\Omega \subset \mathbb{R}^{n}(n = 2)\) is a bounded domain with smooth boundary ∂Ω; ν is the outward normal vector to ∂Ω, and χ, μ, ξ are positive constants. The scalar functions \(u=u(x, t)\), \(v=v(x, t)\), and \(w=w(x, t)\) represent the density of cancer cells, the concentration of enzyme, and the density of healthy tissue, respectively. Notice that in the region of Ω where \(f\geq 0\) the control acts as a proliferation source of the chemical substance, and inversely, in the region of Ω where \(f \leq 0\) the control acts as a degradation source of the chemical substance [23]. In this work, the function \(f\geq 0\) lies in a closed convex set \(\mathcal{F}\).
Anderson et al. [1] presented the following mathematical model which described the invasion of host tissue by tumour cells:
Marciniak-Czochra and Ptashnyk [24] considered the haptotaxis model
They proved the existence of global solutions of the haptotaxis model of cancer invasion for arbitrary non-negative initial conditions. Niño-Celis, Rueda-Gómez and Villamizar-Roa [27] developed two fully discrete schemes for approximating the solutions based on a semi-implicit Euler discretization in time and Finite Element (FE) discretization on space (restricted to triangularization made up of right-angled simplices) of two equivalent systems for the above haptotaxis model.
Chaplain and Lolas [3] first described the process of the cancer invasion on the macroscopic scale by the chemotaxis–haptotaxis system. Tao and Winkler [30] studied the problem
They discussed the global solvability of classical solutions in a bounded domain \(\Omega \subset R^{n} (n\leq 3)\). Cao [2] proved that for nonnegative and suitably smooth initial data, if \(\chi /\mu \) is sufficiently small, the problem possesses a global classical solution, which is bounded in \(\Omega \times (0,\infty )\). The relevant equations have also been studied in [14, 17, 19, 31, 32].
Jin [15] considered the following system:
Under zero-flux boundary conditions, they showed that, for any \(m > 0\), the problem admits a global bounded weak solution for any large initial datum if \(\chi /\mu \) is appropriately small.
Mizukami [25] studied the chemotaxis–haptotaxis system with signal-dependent sensitivity
They established the global existence and boundedness for the above system. The relevant system has also been studied in [33].
During the past years, many authors have been very interested in the optimal control problems governed by the coupled partial differential equations. Colli, Gilardi, Marinoschi and Rocca [8] studied the distributed optimal control problems for a diffuse interface model of tumor growth. Liu and Zhang [21] discussed the optimal distributed control for a new mechanochemical model in biological patterns. Dai and Liu [10] obtained an optimal control problem for a haptotaxis model of solid tumor invasion by considering the multiple treatments of cancer. Recently, Guillén-González, Mallea-Zepeda and Villamizar-Roa [13] studied the following parabolic chemo-repulsion with nonlinear production model in 2D domains:
They proved the existence and uniqueness of global-in-time strong state solution for each control, and the existence of global optimum solution. Guillén-González, Mallea-Zepeda and Rodriguez-Bellido [12] considered a bilinear optimal control problem associated to the above 3D chemo-repulsion model. Guillén-González et al. [11] studied a bilinear optimal control problem for the chemo-repulsion model with linear production term. The existence, uniqueness and regularity of strong solutions of this model were deduced. They also derived the first-order optimality conditions by using a Lagrange multipliers theorem. López-Ríos and Villamizar-Roa [23] studied an optimal control problem associated to a 3D-chemotaxis-Navier–Stokes model. Some other results can be found in [4–7, 16, 20, 22, 29, 35, 36].
In this paper, we are interested in the optimal control problem for the system (1.1)–(1.5). The main difficulties for treating the problem (1.1)–(1.5) are caused by the nonlinearity of \(-\xi \nabla \cdot (u \cdot \nabla w)\) and \(\mu u(1-u-w)\). Our method is based on a Lagrange multiplier theorem.
This paper is organized as follows. In Sects. 2 and 3, we show the well-posedness of the state system (1.1)–(1.5). In Sect. 4, the existence of optimal controls is established. Finally, we derive the first-order necessary optimality conditions in Sect. 5.
Notations: \(L^{p} = L^{p}(\Omega )\,(1\leq p \leq \infty )\) denotes the usual Lebesgue space with the usual norm \(\|\cdot \|_{L^{p}}\). The Sobolev space in Ω of order \(k, k = 0, 1, 2, \dots \), is denoted by \(H^{k}(\Omega )\) with norm \(\|\cdot \|_{H^{k}}\), and the space \(H^{-k}(\Omega )\) is the dual space of \(H^{k}(\Omega )\). The Sobolev space of fractional order \(s > 0\) is denoted by \(H^{s}(\Omega )\) with norm \(\|\cdot \|_{H^{s}}\). \(H^{s}_{N}(\Omega )\) denotes a closed subspace of \(H^{s}(\Omega )\) such that
2 Local solutions
We first review the existence theorem for local solutions to an abstract equation in a Banach space (see Chap. 4 in [34]). Let Z and \(\mathcal{B}\) be two separable Hilbert spaces with dense and compact embedding \(Z\subset \mathcal{B}\). Let \(\|\cdot \|_{Z}\) and \(\|\cdot \|_{\mathcal{B}}\) be the norms of Z and \(\mathcal{B}\), respectively. Let \(Z\subset \mathcal{B}\subset Z^{*}\) be a triplet of spaces. Let \(\|\cdot \|_{Z^{*}}\) be the norm of \(Z^{*}\). We consider the following Cauchy problem for a semilinear abstract evolution equation:
Here, A is a sectorial operator of \(Z^{*}\). The nonlinear operator F is a mapping from Z to \(Z^{*}\), and for any positive number \(\eta > 0\), there exist continuous increasing functions \(\varphi (\cdot )\geq 0\) and \(\psi (\cdot )\geq 0\) such that the following estimates hold:
Then, we have the existence theorem of the local solutions to (2.1).
Proposition 2.1
([34, Theorem 4.6])
Let (2.2) and (2.3) be satisfied. Then, for \(G \in L^{2}(0,T ;Z^{*})\) and any \(U_{0}\in \mathcal{B}\), there exists a unique local solution U to (2.1) in the function space
where \(T_{U_{0},G} > 0\) is determined by the norms \(\|U_{0}\|_{\mathcal{B}}\) and \(\|G\|_{L^{2}(0,T ;Z^{*})}\). In addition, U satisfies the estimate
where \(C_{G, U_{0}}>0\) is a constant depending on the norm \(\|U_{0}\|_{\mathcal{B}}\) and \(\|G\|_{L^{2}(0,T ;Z^{*})}\).
Applying Proposition 2.1, we can show the existence of the local-in-time solutions to (1.1)–(1.5).
Theorem 2.1
For all initial functions \((u_{0}, v_{0}, w_{0})\in H^{1} (\Omega )\times H^{2}_{N}(\Omega ) \times H_{N}^{3} (\Omega ) \), \(u_{0}\geq 0\), \(v_{0}\geq 0\), \(w_{0}\geq 0\) and \(0\leq f\in L^{2}(0,T;H^{1}(\Omega ))\), the problem (1.1)–(1.5) admits a unique local-in-time nonnegative solution \((u, v, w)\) in the function space
with the estimate
where T and C are positive constants depending only on the norms \(\Vert u_{0} \Vert _{H^{1}}+ \Vert v_{0} \Vert _{H^{2}} + \Vert w_{0} \Vert _{H^{2}}\) and \(\|h\|_{L^{2}(0,T;H^{1}(\Omega ))}\).
Proof
Let \(A_{1}=- \Delta +1\), \(A_{2}=- \Delta +1\), and \(A_{3}=1\). Then, \(A_{i}\) are three positive definite self-adjoint operators. We define the linear operator A by
Problem (1.1)–(1.5) is, then, formulated as an abstract equation,
in a product Banach space \(\mathcal{B}=H^{1}(\Omega ) \times H^{2}_{N}(\Omega )\times H^{3}_{N}( \Omega )\). The nonlinear operator F is defined by
The initial value \(U_{0}= (u_{0}, v_{0}, w_{0} )\) is taken in the function space \(H^{1}(\Omega ) \times H_{N}^{2}(\Omega ) \times H_{N}^{3}(\Omega )\). In this setting, we only need to verify conditions (2.2) and (2.3). Let \(U=(u, v, w)\) and \(\tilde{U}=(\tilde{u}, \tilde{v}, \tilde{w})\in Z\). Then, using the interpolation of Sobolev spaces \((\|u\|_{H^{3/2}}\leq C\|u\|_{H^{2}}^{1/2}\|u\|_{H^{1}}^{1/2} )\) and the Young inequality, for any positive number \(\eta > 0\), we have
where C is a positive constant depending only on the known quantities.
On the other hand, we derive
For the first term of the right-hand side, we see that
Similarly, we deduce that
For the third term of the right-hand side,
Similarly, we deduce that
Hence, we can obtain
Thus, we have verified (2.2) and (2.3). Similarly as in the proof of Proposition 2 in [26], we obtain \(u\geq 0\). On the other hand, by the comparison principle, we can be sure that v and w are nonnegative. The proof is complete. □
3 Global existence
In this section, we construct several a priori estimates. At first, we introduce the following lemma.
Lemma 3.1
([28, Lemma 4.3])
For any nonnegative \(u \in H^{1}(\Omega )\), the estimate
holds for any number \(\delta >0\) and some increasing function \(p(\cdot )\).
Lemma 3.2
Let \((u, v, w)\) be a local solution to (1.1)–(1.5). Then, it holds that
Proof
Using the property \(u(t)\geq 0\), \(v(t)\geq 0\) and \(w(t)\geq 0\) for all \(t>0\), and integrating equation (1.1) over Ω, we have
By the comparison argument of ODE, we derive
Integrating (3.4) over \((0,t)\), it follows from (3.5) that
Multiplying equation (1.3) by \(w^{p-1}\) and integrating over Ω, we have
For all \(t\in [0,T]\), due to the fact that v, w are nonnegative, we obtain
which yields
Consequently, (3.3) follows by taking the limit \(p \rightarrow \infty \). Therefore, we complete the proof. □
Lemma 3.3
Let \((u, v, w)\) be a local solution to (1.1)–(1.5). Then, it holds that
Proof
Step 1. v is bounded in \(L^{\infty}(0,T;H^{1}(\Omega ))\cap L^{2}(0,T;H^{2}(\Omega ))\).
Multiplying equation (1.2) by v and integrating over Ω, we have
Integrating (3.9) over \((0,t)\), we derive
Multiplying equation (1.2) by \(-\Delta v\) and integrating over Ω, we have
Integrating (3.11) over \((0,t)\), we obtain
Moreover, using (3.2) and combining (3.10) and (3.12), we have
Step 2. w is bounded in \(L^{\infty}(0,T;H^{2}(\Omega ))\).
Multiplying equation (1.3) by \(-\Delta w\) and integrating over Ω, from (3.3) and the negative of v, we have
Thanks to (3.3), from (3.13), (3.14) and Gronwall lemma, we have
Applying Δ to equation (1.3), multiplying by Δw, and integrating the product over Ω, we have
For the term on the right-hand side, using the interpolation of Sobolev spaces,
Combining (3.16) and (3.17), we can get
From (3.3), (3.13), (3.15), and Gronwall lemma, we have
Step 3. u is bounded in \(L^{\infty}(0,T;H^{2}(\Omega ))\).
We observe that, thanks to the positivity of u, we have \(0 \leq \ln (u+1) \leq u\). Then
We also note that
Taking into account that \(u \in L^{2} (0, T ; H^{1}(\Omega ) )\), from (3.20) and (3.21), we deduce that \(\ln (u+1) \in L^{2} (0, T ; H^{1}(\Omega ) )\). Note that
Testing equation (1.1) with \(\ln (u+1) \in L^{2} (0, T ; H^{1}(\Omega ) )\), and integrating by parts, we have
Applying the Young inequality, we obtain
Similarly, we have
Then combining (3.22)–(3.24) and (3.1), we have
Then, from (3.2), (3.13) and (3.19), as well as applying Gronwall lemma to (3.25), we deduce
Multiplying equation (1.1) by u and integrating over Ω, we see that
Here, we note that
for some positive constant C. Applying Young’s inequality, we further deduce that
So, in the same way, we can derive
Combining (3.27)–(3.30), (3.13) and (3.15), we then deduce
Next, applying ∇ to the equation of (1.2), multiplying by \(\nabla \Delta v\), and integrating the product over Ω, we have
Applying operator ∇Δ to equation (1.3), multiplying by \(\nabla \Delta w\), and integrating over Ω, we have
Thanks to (3.3), (3.19), we further deduce that
Replacing (3.34) in (3.33), we have
Then, choosing δ small enough to absorb \(\|v\|_{H^{3}}\), from (3.7) with \(p=2\), (3.9), (3.11), (3.14), (3.18), (3.31), (3.32), and (3.35), we have
Then, applying the Gronwall lemma to (3.36), we deduce
Integrating (3.36) over \((0,t)\), we obtain
Multiplying equation (1.1) by \(-\Delta u\) and integrating over Ω, we have
The first two terms in the right-hand side can be estimated as follows:
Through a similar calculation as in obtaining the above inequality, it is easy to get
For the third term of the right-hand side, thanks to the nonnegativity of u and w, applying the Gagliardo–Nirenberg inequality and (3.19), we have
Therefore, we have
Taking \(\delta >0\) small enough, and using (3.39), we can get
The proof is complete. □
Theorem 3.1
For all initial functions \((u_{0}, v_{0}, w_{0})\in H^{1} (\Omega )\times H^{2}_{N}(\Omega ) \times H^{3} (\Omega ) \) and \(f\in L^{2}(0,T;H^{1}(\Omega ))\), the problem (1.1)–(1.5) admits a unique global-in-time nonnegative solution \((u, v, w)\) in the function space
In addition, the solution satisfies the uniform estimate involving the norms of initial functions such that
Proof
From Theorem 2.1 and Proposition 2.1, for each triplet of nonnegative initial functions \((u_{0}, v_{0}, w_{0} )\), there exists a unique nonnegative local solution \((u,v,w)\) on an interval \([0,T]\), where the existence time \(T>0\) depends only on the norms of those functions, \(\Vert u_{0} \Vert _{H^{1}}+ \Vert v_{0} \Vert _{H^{2}}+ \Vert w_{0} \Vert _{H^{3}}\). In addition, from Lemma 3.3, the norm \(\|u(t)\|_{H^{1}}+\|v(t)\|_{H^{2}}+\|w(t)\|_{H^{3}}\), \(0 \leq t \leq T\), is estimated from above by a uniform constant C, depending only on the norm \(\Vert u_{0} \Vert _{H^{1}}+ \Vert v_{0} \Vert _{H^{2}}+ \Vert w_{0} \Vert _{H^{3}}\). Then, we consider the problem in \([T, 2T]\). Hence, the interval can be extended to \([0, 2T]\), and the norm \(\|u(t)\|_{H^{1}}+\|v(t)\|_{H^{2}}+\|w(t)\|_{H^{3}}\), \(0 \leq t \leq 2T\), is estimated again by the same constant C from (3.8). Then, the existence time can be extended to 3T. Iterating this procedure proves the global-in-time existence of solutions with the estimate (3.8). □
4 Existence of an optimal control
In this section, we will prove the existence of the optimal solution of the control problem. The method we use for treating this problem was inspired by some ideas of Guillén-González et al. [11]. Assume that \(\mathcal{F} \subset L^{2} (0, T ; H^{1} (\Omega _{c} ) )\) is a nonempty, closed and convex set, where \(\Omega _{c} \subset \Omega \) is the control domain, and \(\Omega _{d} \subset \Omega \) is the observability domain. We consider data \((u_{0}, v_{0}, w_{0})\in H^{1} (\Omega )\times H^{2}(\Omega ) \times H^{3} (\Omega ) \) with \(u_{0}\geq 0\), \(v_{0}\geq 0\) and \(w_{0}\geq 0\) in Ω, and the function \(f \in \mathcal{F}\) that describes the control acting on the v-equation.
Now, we consider the optimal control problem for system (1.1)–(1.5) as follows:
where
Here \((u_{d}, v_{d}, w_{d})\in L^{2}(Q_{d})\times L^{2}(Q_{d}) \times L^{2}(Q_{d})\) represents the desired states and the \(\beta _{i} (i=1,2,3,4)>0\). We will use
which denotes the set of admissible solutions of (4.1).
First, we will consider the existence of a global optimal solution of problem (4.1). To this end, we start with the definition of optimal solution.
Definition 4.1
An element \((\tilde{u}, \tilde{v}, \tilde{w}, \tilde{f} )\in \mathcal{S}_{\mathrm{ad}}\) will be called a global optimal solution of problem (4.1) if
Here we state the following result.
Theorem 4.1
Let \(u_{0}\in H^{1}(\Omega )\), \(v_{0}\in H^{2}(\Omega )\) and \(w_{0}\in H^{3}(\Omega )\) with \(u_{0}\geq 0\), \(v_{0}\geq 0\) and \(w_{0}\geq 0\) in Ω. Then the optimal control problem (4.1) has at least one global optimal solution \((\tilde{u}, \tilde{v}, \tilde{w}, \tilde{f}) \in \mathcal{S}_{\mathrm{ad}}\).
Proof
From Theorem 3.1, recalling that \(\mathcal{S}_{\mathrm{ad}}\) is nonempty, there exists a minimizing sequence \(\{s_{m}\}_{m\in \mathbb{N}}\subset \mathcal{S}_{\mathrm{ad}}\) such that \(\lim_{m\to +\infty} J(s_{m})=\inf_{s\in \mathcal{S}_{\mathrm{ad}}}J(s)\). Then, by the definition of \(\mathcal{S}_{\mathrm{ad}}\), we know that for each \(m\in \mathbb{N}\), \(s_{m}\) satisfies
Hence, it follows that
By (3.37), (3.38), (3.44), and (3.45), we see that there exists \(C>0\) such that
Therefore, by (4.6), (4.7) and since \(\mathcal{F}\) is a closed convex subset of \(L^{2}(Q_{c})\), we deduce that there exist \(\tilde{s}=(\tilde{u},\tilde{v},\tilde{w},\tilde{f})\in \mathcal{M}\) and a subsequence of \(\{s_{m}\}_{m\in \mathbf{N}}\), not relabeled, such that, as \(m\rightarrow +\infty \),
and
From (4.8)–(4.13) and the Aubin–Lions lemma, we have
In particular, since \(\nabla \cdot (u_{m} \nabla v_{m} )=\nabla u_{m} \cdot \nabla v_{m}+u_{m} \Delta v_{m}\) and \(\nabla \cdot (u_{m} \nabla w_{m} )=\nabla u_{m} \cdot \nabla w_{m}+u_{m} \Delta w_{m}\) is bounded in \(L^{2}(0,T;L^{2}(\Omega ))\), one has the weak convergences:
On the other hand, from (4.8)–(4.17), one has
Therefore, we can identify \(\psi _{1}=\nabla \cdot (\tilde{u} \nabla \tilde{v})\) and \(\psi _{2}=\nabla \cdot (\tilde{u} \nabla \tilde{w})\) a.e. in Q, and thus
Moreover, by (4.15)–(4.17), we see \((u_{m}(0), v_{m}(0), w_{m}(0) )\to (\tilde{u}(0), \tilde{v}(0), \tilde{w}(0))\), in \(L^{2}(\Omega )\times H^{1}(\Omega )\times H^{2}(\Omega )\). Since \(u_{m}(0)=u_{0}\), \(v_{m}(0)=v_{0}\), \(w_{m}(0)=w_{0}\), we conclude that \(\tilde{u}(0)=u_{0}\), \(\tilde{v}(0)=v_{0}\) and \(\tilde{w}(0)=w_{0}\), thus s̃ satisfies the initial conditions given in (1.1)–(1.5). Therefore, considering the convergences (4.8)–(4.19), we can pass to the limit in (4.5) as \(m \to +\infty \), and conclude that \(\tilde{s}=(\tilde{u},\tilde{v},\tilde{w},\tilde{f})\) is a solution of the system (1.1)–(1.5), that is, \(\tilde{s} \in \mathcal{S}_{\mathrm{ad}}\). Hence,
On the other hand, since J is lower semicontinuous on \(\mathcal{S}_{\mathrm{ad}}\), we have \(J(\tilde{s}) \leq \liminf_{m \rightarrow +\infty} J (s_{m} )\), which, jointly with (4.20), implies (4.4). □
5 First-order necessary optimality condition
Now, we will study the first-order necessary optimality conditions for a local optimal solution \((\tilde{u},\tilde{v},\tilde{w},\tilde{f})\) of problem (4.1). To this end, we will use a result on existence of Lagrange multipliers in Banach spaces [37]. First, we discuss the following problem:
where \(J: X \rightarrow \mathbb{R}\) is a functional, \(G: X \rightarrow Y\) is an operator, X and Y are Banach spaces, \(\mathcal{M}\) is a nonempty closed convex subset of X, and \(\mathcal{N}\) is a nonempty closed convex cone in Y with vertex at the origin.
For a subset A of X (or Y), \(A^{+}\) denotes its polar cone, that is,
Definition 5.1
A point \(\tilde{s}\in \mathcal{S}\) is said to be a local optimal solution of problem (5.1), if there exits \(\varepsilon >0\) such that for all \(s\in \mathcal{S}\) satisfying \(\|s-\tilde{s}\|_{X} \leq \varepsilon \) one has \(J(\tilde{s}) \leq J(s)\).
Definition 5.2
Let \(\tilde{s}\in \mathcal{S}\) be a local optimal solution for problem (5.1) with respect to the X-norm. Suppose that J and G are Fréchet differentiable in s̃, with derivatives \(J^{\prime}(\tilde{s})\) and \(G^{\prime}(\tilde{s})\), respectively. Then, any \(\lambda \in Y^{\prime}\) is called a Lagrange multiplier for (5.1) at the point s̃ if
where \(\mathcal{C}(\tilde{s})=\{\theta (s-\tilde{s}): s \in \mathcal{M}, \theta \geq 0\}\) is the conical hull of s̃ in \(\mathcal{M}\).
Definition 5.3
Let \(\tilde{s}\in \mathcal{S}\) be a local optimal solution for problem (5.1). We say that s̃ is a regular point if
where \(\mathcal{N}(G(\tilde{s}))=\{(\theta (n-G(\tilde{s})): n \in \mathcal{N}, \theta \geq 0\}\) is the conical hull of \(G(\tilde{s})\) in \(\mathcal{N}\).
Theorem 5.1
([37, Theorem 3.1])
Let \(\tilde{s}\in \mathcal{S}\) be a local optimal solution for problem (5.1). Suppose that J is a Fréchet differentiable function and G is continuously Fréchet-differentiable. If s̃ is a regular point, then the set of Lagrange multipliers for (5.1) at s̃ is nonempty.
Remark 5.1
To obtain the existence of first-order necessary optimality conditions, because of the nonlinearity of \(-\xi \nabla \cdot (u \cdot \nabla w)\) and \(\mu u(1-u-w)\), the method used in [10] seems not applicable to the present situation. Our method is based on the Lagrange multiplier theorem.
Now, we will reformulate the optimal control problem (4.1) in the abstract setting (5.1). We consider the following Banach spaces:
where
By Theorem 3.1, we know that the operator \(G= (G_{1}, G_{2}, G_{3}, G_{4},G_{5},G_{6} ): X \rightarrow Y\), where
are defined at each point \(s=(u,v,w,f)\in X\) by
By taking \(\mathcal{M}=\mathcal{W}_{u} \times \mathcal{W}_{v} \times \mathcal{W}_{w} \times \mathcal{F}\) a closed convex subset of X and \(\mathcal{N}=\{0\}\), the optimal control problem (4.1) is reformulated as follows:
Similar to [21], by the definition of the Fréchet derivative, using a direct calculation, we have the following results.
Lemma 5.1
The operator \(G:X\rightarrow Y\) is continuously Fréchet differentiable and the Fréchet derivative of G in \(\tilde{s}=(\tilde{u},\tilde{v},\tilde{w},\tilde{f})\in X\), in the direction \(r=(U, V, W, F) \in X\), is the linear operator
defined by
Lemma 5.2
The functional \(J:X\rightarrow \mathbb{R}\) is Fréchet differentiable and the Fréchet derivative of J in \(\tilde{s}=(\tilde{u},\tilde{v},\tilde{w},\tilde{f})\in X\) in the direction \(r=(U, V, W, F) \in X\) is given by
We wish to prove the existence of Lagrange multipliers, which is guaranteed if a local optimal solution of problem (5.9) is a regular point of operator G.
Lemma 5.3
If \(\tilde{s}=(\tilde{u},\tilde{v},\tilde{w},\tilde{f})\in \mathcal{S}_{\mathrm{ad}}\), then s̃ is a regular point.
Proof
Fix \((\tilde{u},\tilde{v},\tilde{w},\tilde{f})\in \mathcal{S}_{\mathrm{ad}}\) and let \((g_{u}, g_{v}, g_{w},U_{0}, V_{0}, W_{0} ) \in Y\). Since \(0 \in \mathcal{C}(\tilde{f})\), it suffices to show the existence of \((U,V,W)\) such that
Step 1. Local existence of a solution,
In order to prove the existence of a solution of (5.12), we use Proposition 2.1 to solve the problem
endowed with the corresponding initial and boundary conditions. On the product Banach space \(\mathcal{B}=H^{1}(\Omega ) \times H^{2}_{N}(\Omega )\times H^{3}_{N}( \Omega )\), we define the linear operator A by
The nonlinear operator F is defined by
The remaining part of the proof can be done in the same way as that in the proof of Theorem 2.1, so we omit the details.
Now, we prove the global-in-time solutions in the following part.
Step 2. \((U,V,W)\in L^{\infty}(0,T;L^{2}(\Omega ))\cap L^{2}(0,T; H^{1}( \Omega ))\times L^{\infty}(0,T;H^{1}(\Omega ))\cap L^{2}(0,T; H^{2}( \Omega )) \times L^{\infty}(0,T;H^{2}(\Omega ))\).
By testing the first equation of (5.13) with U, one has
Applying the Hölder and Young inequalities as well as (3.8) to the terms on the right-hand side of (5.14),
and
In the same way, we can get
and
For the other terms on the right,
Replacing (5.15)–(5.19) in (5.14), we have
By testing the second equation of (5.13) with V, we conclude
By testing the second equation of (5.13) with ΔV, we get
By testing the third equation of (5.13) with W, we obtain
Next, applying ∇ to the third equation of (5.13), multiplying by ∇W, and integrating the product over Ω, we have
From (3.3) and (3.8), we derive that
and
So, combining (5.23)–(5.27), we can get
Applying Δ to the third equation of (5.13), multiplying by ΔW, and integrating the product over Ω, we have
Applying the Hölder and Young inequalities to the terms on the right-hand side of (5.29), we have
and
So, we can get
Therefore, we can obtain
By choosing δ small enough, and utilizing the Gronwall inequality, we have
Step 3. \((U,V,W)\in L^{\infty}(0,T;H^{1}(\Omega ))\cap L^{2}(0,T; H^{2}( \Omega ))\times L^{\infty}(0,T;H^{2}(\Omega ))\cap L^{2}(0,T; H^{3}( \Omega )) \times L^{\infty}(0,T;H^{2}(\Omega ))\).
By testing the first equation of (5.13) with \(-\Delta U\), one has
By applying the boundedness of \(\|\tilde{v}\|^{2}_{H^{2}}\), (5.35), and the Gagliardo–Nirenberg interpolation inequality, we have
and
By utilizing the same ideas, we can get
and
With (5.35) and the boundedness of \(\|\tilde{w}\|^{2}_{H^{1}}\), \(\|\tilde{u}\|^{2}_{H^{1}}\) in hand, we derive
Replacing (5.37)–(5.41) in (5.36), and using the fact that \(\|\tilde{w}\|^{2}_{H^{3}}\leq C\) and (5.35), we have
Next, applying Δ to the second equation of (5.13), multiplying by ΔV, and integrating the product over Ω, we see
Applying the Hölder and Young inequalities, we obtain
Therefore, from (5.43) and (5.44), we get
Applying ∇Δ to the third equation of (5.13), multiplying by \(\nabla \Delta W\), and integrating the product over Ω yields
For the first term on the right,
Thanks to the boundedness of \(\|\nabla \Delta \tilde{w}\|_{L^{2}}\) and (5.35), we can get
and
Then collecting (5.35) and \(\|\tilde{v}\|_{H^{2}}\leq C\), we arrive at
Applying the Hölder and Young inequalities, we obtain
Thus, in light of (5.46)–(5.51), we arrive at
We invoke (5.34), (5.42), (5.45), and (5.52) to obtain
By utilizing the Gronwall inequality and \(\tilde{v}\in L^{2}(0,T;H^{3}(\Omega ))\), \(\tilde{w}\in L^{\infty}(0,T;H^{3}(\Omega )) \) \(\tilde{u}\in L^{2}(0,T;H^{2}(\Omega ))\), we have
Thus, we conclude the proof. □
Now we show the existence of Lagrange multipliers.
Theorem 5.2
Let \(\tilde{s}=(\tilde{u}, \tilde{v}, \tilde{w}, \tilde{f}) \in \mathcal{S}_{\mathrm{ad}}\) be a local optimal solution for the control problem (5.9). Then, there exist Lagrange multipliers \((\lambda , \eta , \rho , \varphi ,\phi , \psi )\in L^{2}(Q)\times (L^{2}(0,T;H^{1}( \Omega )))^{\prime}\times (L^{\infty}(0,T;H^{3}(\Omega )))^{\prime} \times (H^{1}(\Omega ))^{\prime}\times (H^{2}(\Omega ))^{\prime} \times (H^{3}(\Omega ))^{\prime} \) such that for all \((U,V,W,F)\in \mathcal{W}_{u}\times \mathcal{W}_{v}\times \mathcal{W}_{w} \times \mathcal{C}(\tilde{f})\) one has
Proof
From Lemma 5.3, \(\tilde{s} \in \mathcal{S}_{\mathrm{ad}}\) is a regular point, so by Theorem 5.1 there exist Lagrange multipliers
such that
for all \(r=(U,V,W,F)\in \mathcal{W}_{u}\times \mathcal{W}_{v}\times \mathcal{W}_{w}\times \mathcal{C}(\tilde{f})\). Thus, the proof follows from (5.11) and (5.10). □
From Theorem 5.2, we derive an optimality system for which we consider the following spaces:
Corollary 5.1
Let \(\tilde{s}=(\tilde{u}, \tilde{v}, \tilde{w}, \tilde{f})\) be a local optimal solution for the optimal control problem (5.9). Then the Lagrange multiplier \((\lambda , \eta , \rho )\in L^{2}(Q)\times (L^{2}(0,T;H^{1}(\Omega )))^{ \prime}\times (L^{\infty}(0,T;H^{3}(\Omega )))^{\prime}\), provided by Theorem 5.2, satisfies the system
which corresponds to the concept of very weak solution of the linear system
and the optimality condition
Proof
From (5.55), taking \((V,W,F)=(0,0,0)\), and taking into account that \(\mathcal{W}_{u_{0}}\) is a vector space, we have (5.58). Similarly, taking \((U,W,F)=(0,0,0)\) in (5.55), and considering that \(\mathcal{W}_{v_{0}}\) is a vector space, we obtain (5.59). Taking \((U,V,F)=(0,0,0)\) in (5.55), and considering that \(\mathcal{W}_{w_{0}}\) is a vector space, we obtain (5.60). Finally, taking \((U, V ,W) = (0, 0, 0)\) in (5.55), we have
Therefore, choosing \(F=f-\bar{f} \in \mathcal{C}(\bar{f})\) for all \(f \in \mathcal{F}\) in the last inequality, we derive (5.62). □
In the following result we show that the Lagrange multiplier \((\lambda , \eta , \rho )\), provided by Theorem 5.2, has some extra regularity.
Theorem 5.3
Under conditions of Theorem 5.2, the system (5.61) has a unique strong solution \((\lambda , \eta , \psi )\) such that
Proof
Let \(s=T-t\), with \(t \in (0, T)\) and \(\tilde{\lambda}(s)=\lambda (t)\), \(\tilde{\eta}(s)=\eta (t)\), \(\tilde{\eta}(s)=\psi (t)\). Then system (5.61) is equivalent to
The proof employs a Galerkin approximation. By testing (5.67)1 with \(-\Delta \tilde{\lambda}\), we have
Applying the Hölder, Young, and Nirenberg inequalities, as well as the boundedness of \(\|\tilde{v}\|^{2}_{H^{2}}\), we obtain
Then, utilizing the same procedure gives
In light of the boundedness of ũ, w̃, we see that
In the light of (5.68)–(5.71), we have
Similarly, testing (5.67)1 with λ̃ yields
By testing (5.67)2 with η̃, we conclude
Using the Hölder, Young, and Nirenberg inequalities, we obtain
Applying the Hölder and Young inequalities, as well as the boundedness of w̃, we get
Then collecting (5.74)–(5.77), we see
Testing (5.67)3 with ρ̃, we have
First, we see that
Similarly, we know that
and
Then collecting (5.79)–(5.83), we get
Accordingly, we invoke (5.72), (5.73), (5.78), and (5.84) to obtain
By taking δ small enough, a Gronwall argument leads to
Testing the first equation of (5.67) with \(\tilde{\lambda}_{s}\), integrating over \(\Omega \times (0, T)\), and using the above inequality, we have
Similarly, we obtain
The proof is complete. □
Corollary 5.2
(Optimality System)
Let \(\tilde{s}=(\tilde{u}, \tilde{v}, \tilde{w}, \tilde{f}) \in \mathcal{S}_{\mathrm{ad}}\) be a local optimal solution for the control problem (5.9). Then, the Lagrange multiplier \((\lambda , \eta , \rho )\) has the regularity as in (5.67) and satisfies the following optimality system:
Remark 5.2
The first-order necessary optimality conditions for chemotaxis models have been intensively studied [10, 12, 23]. Recently, Colli, Signori and Sprekels [9] established both first-order necessary and second-order sufficient conditions for a tumor growth model of Cahn–Hilliard type, including chemotaxis with possibly singular potentials.
Remark 5.3
For the numerical analysis of the optimal control problem, the ringlike diffusion and aggregation patterns and the dynamics of tumor invasion, as well as the optimal control strategies, are presented numerically in Dai and Liu [10] for a haptotaxis model. Khajanchi and Ghosh [18] studied the numerical aspect of the optimal control problem for the immunogenic tumors model. They demonstrated the numerical illustrations that the optimal regimens reduce the tumor burden under different scenarios.
Availability of data and materials
Not applicable.
References
Anderson, A.R.A., Chaplain, M.A.J., Newman, E.L., Steele, R.J.C., Thompson, A.M.: Mathematical modelling of tumour invasion and metastasis. J. Theor. Med. 2, 129–154 (2000)
Cao, X.: Boundedness in a three-dimensional chemotaxis-haptotaxis model. Z. Angew. Math. Phys. 67, 11–13 (2016)
Chaplain, M.A.J., Lolas, G.: Mathematical modelling of cancer invasion of tissue: dynamic heterogeneity. Netw. Heterog. Media 1, 399–439 (2006)
Chen, B., Li, H., Liu, C.: Optimal distributed control for a coupled phase-field system. Discrete Contin. Dyn. Syst., Ser. B 27(3), 1789–1825 (2022)
Chen, B., Liu, C.: Optimal distributed control of a Allen–Cahn/Cahn–Hilliard system with temperature. Appl. Math. Optim. 84, S1639–S1684 (2021)
Colli, P., Gilardi, G., Marinoschi, G., Rocca, E.: Optimal control for a phase field system with a possibly singular potential. Math. Control Relat. Fields 6, 95–112 (2016)
Colli, P., Gilardi, G., Marinoschi, G., Rocca, E.: Optimal control for a conserved phase field system with a possibly singular potential. Evol. Equ. Control Theory 7(1), 95–116 (2018)
Colli, P., Gilardi, G., Rocca, E., Sprekels, J.: Optimal distributed control of a diffuse interface model of tumor growth. Nonlinearity 30(6), 2518–2546 (2017)
Colli, P., Signori, A., Sprekels, J.: Second-order analysis of an optimal control problem in a phase field tumor growth model with singular potentials and chemotaxis. ESAIM Control Optim. Calc. Var. 27, 73 (2021)
Dai, F., Liu, B.: Optimal control and pattern formation for a haptotaxis model of solid tumor invasion. J. Franklin Inst. 356(16), 9364–9406 (2019)
Guillén-González, F., Mallea-Zepeda, E., Rodríguez-Bellido, M.: Optimal bilinear control problem related to a chemo-repulsion system in 2D domains. ESAIM Control Optim. Calc. Var. 26, 29 (2020)
Guillén-González, F., Mallea-Zepeda, E., Rodriguez-Bellido, M.A.: A regularity criterion for a 3D chemo-repulsion system and its application to a bilinear optimal control problem. SIAM J. Control Optim. 58(3), 1457–1490 (2020)
Guillén-González, F., Mallea-Zepeda, E., Villamizar-Roa, É.J.: On a bi-dimensional chemo-repulsion model with nonlinear production and a related optimal control problem. Acta Appl. Math. 170, 963–979 (2020)
Hu, X., Wang, L., Mu, C., Li, L.: Boundedness in a three-dimensional chemotaxis–haptotaxis model with nonlinear diffusion. C. R. Math. Acad. Sci. Paris 355(2), 181–186 (2017)
Jin, C.: Boundedness and global solvability to a chemotaxis–haptotaxis model with slow and fast diffusion. Discrete Contin. Dyn. Syst., Ser. B 23, 1675–1688 (2018)
Kahle, C., Lam, K.F.: Parameter identification via optimal control for a Cahn–Hilliard–chemotaxis system with a variable mobility. Appl. Math. Optim. 82(1), 63–104 (2020)
Ke, Y., Zheng, J.: A note for global existence of a two-dimensional chemotaxis–haptotaxis model with remodelling of non-diffusible attractant. Nonlinearity 31(10), 4602–4620 (2018)
Khajanchi, S., Ghosh, D.: The combined effects of optimal control in cancer remission. Appl. Math. Comput. 271, 375–388 (2015)
Li, D., Mu, C., Yi, H.: Global boundedness in a three-dimensional chemotaxis–haptotaxis model. Comput. Math. Appl. 77(9), 2447–2462 (2019)
Liu, C., Yuan, Y.: Optimal control of a fully parabolic attraction–repulsion chemotaxis model with logistic source in 2D. Appl. Math. Optim. 85(1), 1–38 (2022)
Liu, C., Zhang, X.: Optimal distributed control for a new mechanochemical model in biological patterns. J. Math. Anal. Appl. 478(2), 825–863 (2019)
Liu, C., Zhang, X.: Optimal control of a new mechanochemical model with state constraint. Math. Methods Appl. Sci. 44(11), 9237–9263 (2021)
López-Ríos, J., Villamizar-Roa, É.J.: An optimal control problem related to a 3D-chemotaxis–Navier–Stokes model. ESAIM Control Optim. Calc. Var. 27, 58 (2021)
Marciniak-Czochra, A., Ptashnyk, M.: Boundedness of solutions of a haptotaxis model. Math. Models Methods Appl. Sci. 20, 449–476 (2010)
Mizukami, M., Otsuka, H., Yokota, T.: Global existence and boundedness in a chemotaxis-haptotaxis system with signal-dependent sensitivity. J. Math. Anal. Appl. 464(1), 354–369 (2018)
Nakaguchi, E., Noda, K., Osaki, K., Uemichi, K.: Global attractor for a two-dimensional chemotaxis system with linear degradation and indirect signal production. Jpn. J. Ind. Appl. Math. 37(1), 49–80 (2020)
Niño-Celis, V., Rueda-Gómez, D.A., Villamizar-Roa, É.J.: Convergence and positivity of finite element methods for a haptotaxis model of tumoral invasion. Comput. Math. Appl. 89, 20–33 (2021)
Osaki, K., Tsujikawa, T., Yagi, A., Mimura, M.: Exponential attractor for a chemotaxis-growth system of equations. Nonlinear Anal. 51(1), 119–144 (2002)
Ryu, S.U., Yagi, A.: Optimal control of Keller–Segel equations. J. Math. Anal. Appl. 256(1), 45–66 (2001)
Tao, Y., Winkler, M.: Large time behavior in a multidimensional chemotaxis–haptotaxis model with slow signal diffusion. SIAM J. Math. Appl. 47, 4229–4250 (2015)
Tao, Y., Winkler, M.: Critical mass for infinite-time aggregation in a chemotaxis model with indirect signal production. J. Eur. Math. Soc. 19(12), 3641–3678 (2017)
Wang, Y., Ke, Y.: Large time behavior of solution to a fully parabolic chemotaxis–haptotaxis model in higher dimensions. J. Differ. Equ. 260, 6960–6988 (2016)
Xu, T., Ji, S., Mei, M., Yin, J.: Global existence of solutions to a chemotaxis–haptotaxis model with density-dependent jump probability and quorum-sensing mechanisms. Math. Methods Appl. Sci. 41(11), 4208–4226 (2018)
Yagi, A.: Abstract Parabolic Evolution Equations and Their Applications. Springer Monographs in Mathematics. Springer, Berlin (2010)
Yuan, Y., Liu, C.: Optimal control for the coupled chemotaxis-fluid models in two space dimensions. Electron. Res. Arch. 29(6), 4269–4296 (2021)
Zhang, X., Li, H., Liu, C.: Optimal control problem for the Cahn–Hilliard/Allen–Cahn equation with state constraint. Appl. Math. Optim. 82(2), 721–754 (2020)
Zowe, J., Kurcyusz, S.: Regularity and stability for the mathematical programming problem in Banach spaces. Appl. Math. Optim. 5, 49–62 (1979)
Acknowledgements
Not applicable.
Funding
This research is supported by the National Natural Science Foundation of China (No. 11701042), and the Science and Technology program of the Education Department of Jilin province(No. JJKH20190501KJ).
Author information
Authors and Affiliations
Contributions
All authors contributed equally to the manuscript and read and approved the final manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare 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/.
About this article
Cite this article
Tang, H., Yuan, Y. Optimal control for a chemotaxis–haptotaxis model in two space dimensions. Bound Value Probl 2022, 79 (2022). https://doi.org/10.1186/s13661-022-01661-7
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s13661-022-01661-7