 Research
 Open Access
On the Cauchy problem for lower semicontinuous differential inclusions
 Paolo Cubiotti^{1} and
 JenChih Yao^{2, 3}Email author
https://doi.org/10.1186/s1366201609330
© Cubiotti and Yao 2016
 Received: 21 April 2016
 Accepted: 4 August 2016
 Published: 23 August 2016
Abstract
We provide a new proof of a classical result by Bressan on the Cauchy problem for firstorder differential inclusions with null initial condition. Our approach allows us to prove the result directly for kth order differential inclusions, under weaker regularity assumptions on the involved multifunction. Our result is the following: let a, b, M be positive real numbers, with \(M\cdot\max\{a,a^{k}\}\le b\), and let B and X be the closed balls in \(\mathbf{R}^{n}\), centered at the origin with radius b and M, respectively. Let \(F:[0,a]\times B^{k}\to2^{X}\) be a multifunction with nonempty closed values, such that F is \(\mathcal{L}([0,a])\otimes\mathcal{B}(B^{k})\)measurable, and for all \(t\in[0,a]\) the multifunction \(F(t,\cdot)\) is lower semicontinuous. Then there exists \(u\in W^{k,\infty}([0,a],\mathbf{R}^{n})\) such that \(u^{(k)}(t)\in F(t,u(t), u^{\prime}(t),\ldots,u^{(k1)}(t))\) a.e. in \([0,a]\), and \(u^{(i)}(0)=0_{\mathbf{R}^{n}}\) for all \(i=0,\ldots, k1\).
Keywords
 differential inclusions
 Cauchy problem
 generalized solutions
 discontinuous selections
1 Introduction
The importance of the theory of differential inclusions is well documented in the literature. Indeed, they play a crucial role in the study of many dynamical problems coming from economics, social sciences, biology. They also provide a fundamental tool in control theory. Moreover, differential inclusions are very useful when studying differential equations with discontinuous righthand side. For a detailed introduction to differential inclusions and their applications, we refer the reader to [1, 2].
In the paper [3], A. Bressan proved his classical and celebrated result on the Cauchy problem for firstorder differential inclusions, which we now state. In what follows, the space \(\mathbf{R}^{n}\) is considered with its Euclidean norm.
Theorem 1.1
(Theorem 2 of [3])
Let \(a>0\), and let B and X be the closed balls in \(\mathbf{R}^{n}\), centered at the origin with radius b and M, respectively, with \(a\le b/M\). Let \(F: [0,a]\times B\to2^{X}\) be a lower semicontinuous multifunction with nonempty closed values.
Then there exists an absolutely continuous \(u:[0,a]\to\mathbf{R}^{n}\) such that \(u(0)=0_{\mathbf{R}^{n}}\) and \(u^{\prime}(t)\in F(t,u(t))\) for a.e. \(t\in[0,a]\).
The aim of this note is simply to propose an alternative proof of Theorem 1.1, completely independent from the original one. Our approach also allows one to state Theorem 1.1 directly for kthorder differential inclusions, under slightly weaker regularity assumptions on F. That is, the joint lower semicontinuity of the multifunction F is replaced by joint measurability and by the lower semicontinuity of \(F(t,\cdot )\). In this connection, we observe that the original proof of Theorem 1.1 does not work if F is not jointly lower semicontinuous (see the Remark at p.96 of [3]).
Before stating our result, we fix some notations. If \(a>0\), \(k,n\in \mathbf{N} \) and \(p\in[1,+\infty]\), we denote by \(W^{k,p}([0,a],\mathbf{R}^{n})\) the set of all \(u\in C^{k1}([0,a],\mathbf{R}^{n})\) such that \(u^{(k1)}\) is absolutely continuous in \([0,a]\) and \(u^{(k)}\in L^{p}([0,a],\mathbf{R}^{n})\). Moreover, we denote by \(\mathcal{L}([0,a])\) and \(\mathcal{B}(S)\) the family of all measurable subsets of \([0,a]\) and the Borel family of the topological space S, respectively. The following is our result.
Theorem 1.2
The proof of Theorem 1.2, which will be given in Section 3, will follow from a more general result, Theorem 3.1 below. It is mainly based on a recent existence result for discontinuous selections and on an existence result for operator inclusions, stated below as Theorems 2.1 and 2.3, respectively. In Section 2, conversely, we shall give some notations and recall the results that we shall use in our proofs.
2 Preliminaries
Let \(n\in\mathbf{N}\). For all \(i\in\{1,\ldots,n\}\), we shall denote by \(P_{n,i}:\mathbf{R}^{n}\to\mathbf{R}\) the projection over the ith axis. The space \(\mathbf{R}^{n}\) is considered with its Euclidean norm \(\ \cdot\_{n}\). If \(A\subseteq \mathbf{R}^{n}\), we denote by \(\overline{\mathrm{conv}}(A)\) the closed convex hull of A. Moreover, we denote by \(m_{n}\) the ndimensional Lebesgue measure in \(\mathbf{R}^{n}\). If \([a,b]\) is a compact interval, we shall denote by \(AC([a,b],\mathbf{R}^{n})\) the space of all absolutely continuos functions from \([a,b]\) into \(\mathbf{R}^{n}\).
If \((T,\mathcal{A})\) is a measurable space, a multifuncion \(F:T\to 2^{V}\) is said to be measurable (resp., weakly measurable) if for any closed (resp., open) set \(W\subseteq V\) one has \(F^{}(W)\in \mathcal{A}\). For the basic definitions and facts about multifunctions, we refer the reader to [4] and [5]. For what concerns measurable multifunctions, we also refer to [6]. In the following, we shall make the obvious identification \((\mathbf{R}^{n})^{k}=\mathbf{R}^{nk}\) whenever it is convenient for simpler exposition.
For the sake of completeness, we now recall the results that we use in our proofs. We start with the following very recent selection theorem.
Theorem 2.1
(Theorem 2.2 of [7])
Let T and \(X_{1},X_{2},\ldots, X_{k}\) be Polish spaces, with \(k\in\mathbf{N}\), and let \(X:=\prod_{j=1}^{k}X_{j}\) (endowed with the product topology). Let \(\mu, \psi_{1},\ldots,\psi_{k}\) be positive regular Borel measures over \(T,X_{1},X_{2},\ldots, X_{k}\), respectively, with μ finite and \(\psi_{1},\ldots,\psi_{k}\) σfinite.
 (i)
the multifunction F is \(\mathcal{T}_{\mu}\otimes\mathcal{B}(X)\)measurable (\(\mathcal {T}_{\mu}\) denoting the completion of the Borel σalgebra \(\mathcal{B}(T)\) of T with respect to the measure μ);
 (ii)for a.e. \(t\in T\), one has$$\bigl\{ x:=(x_{1},\ldots,x_{k})\in X: F_{t} \textit{ is not lower semicontinuous at } x \bigr\} \subseteq E. $$
 (a)
\(\phi(t,x)\in F(t,x) \) for all \((t,x)\in T\times X\);
 (b)
for all \(x:=(x_{1},x_{2},\ldots,x_{k})\in X\setminus [ (\bigcup_{i=1}^{k}P_{*,i}^{1}(Q_{i}) )\cup E ]\), the function \(\phi( \cdot ,x)\) is \(\mathcal{T}_{\mu}\)measurable over T;
 (c)for a.e. \(t\in T\), one has$$\begin{aligned}& \bigl\{ x:=(x_{1},x_{2},\ldots,x_{k})\in X: \phi( t,\cdot ) \textit{ is discontinuous at } x \bigr\} \\& \quad \subseteq E\cup \Biggl[ \Biggl(\bigcup _{i=1}^{k} P_{*,i}^{1}(Q_{i}) \Biggr) \Biggr]. \end{aligned}$$
We also recall the following proposition.
Proposition 2.2
(Proposition 2.6 of [8])
 (i)
for all \(t\in[a,b]\), the function \(\psi(t,\cdot )\) is bounded;
 (ii)
for all \(x\in D\), the function \(\psi( \cdot ,x)\) is \(\mathcal{L}([a,b])\)measurable.
 (a)
G has nonempty closed convex values;
 (b)
for all \(x\in\mathbf{R}^{n}\), the multifunction \(G(\cdot,x)\) is \(\mathcal{L}([a,b])\)measurable;
 (c)
for all \(t\in[a,b]\), the multifunction \(G(t,\cdot)\) has closed graph;
 (d)if \(t\in[a,b]\), and \(\psi(t,\cdot )_{\mathbf{R} ^{n}\setminus E}\) is continuous at \(x\in\mathbf{R}^{n}\setminus E\), then one has$$G(t,x)=\bigl\{ \psi(t,x)\bigr\} . $$
The following existence result for operator inclusions is due to Naselli Ricceri and Ricceri.
Theorem 2.3
(Theorem 1 of [9])
 (i)
for μalmost every \(t\in T\), the multifunction \(F(t,\cdot)\) has closed graph;
 (ii)the setis dense in X;$$\bigl\{ x\in X: \textit{ the multifunction } F(\cdot,x) \textit{ is weakly measurable}\bigr\} $$
 (iii)there exists a number \(r>0\) such that the functionbelongs to \(L^{s}(T)\) and its norm in \(L^{p}(T)\) is less or equal to r.$$t\to\sup_{\x\_{X}\le\varphi(r)} d\bigl(0_{Y},F(t,x)\bigr) $$Under such hypotheses, there exists \(\tilde{u}\in V\) such that$$\begin{aligned}& \Psi (\tilde{u}) (t)\in F\bigl(t,\Phi (\tilde{u}) (t)\bigr)\quad\mu\textit{a.e.}, \\ & \bigl\Vert \Psi (\tilde{u}) (t)\bigr\Vert _{Y}\le\sup _{\x\_{X}\le\varphi(r)} d\bigl(0_{Y},F(t,x)\bigr)\quad\mu\textit{a.e. in } T. \end{aligned}$$
The two following results, concerning absolutely continuous functions, will be fundamental in the sequel. We recall that a function \(f:[a,b]\to[\alpha ,\beta ]\) is called an Nfunction if it maps null sets into null sets.
Theorem 2.4
(Theorem 2 of [10])
Let \(f:[a,b]\to\mathbf{R}\) be continuous and strictly monotonic. Then \(f^{1}\) is absolutely continuous if and only if \(f^{\prime}\ne0\) almost everywhere on \([a,b]\).
Theorem 2.5
(Theorem 18.25 of [11])
Let f be a continuous function of finite variation with domain \([a,b]\subseteq \mathbf{R}\) and range \([\alpha ,\beta ]\subseteq \mathbf{R}\). Then f is an Nfunction if and only if f is absolutely continuous.
Finally, we recall the following lemma.
Lemma 2.6
(Lemma 3.2 of [12])
 (a)
F is \(\mathcal{D}\otimes\mathcal{B}(X)\)measurable;
 (b)
H is \(\mathcal{D}\otimes\mathcal{B}(Y)\)measurable and has closed values.
Then the multifunction G defined by \(G(t,y)=F(t,H(t,y))\) for all \((t,y)\in T\times Y\) is \(\mathcal{D}\otimes\mathcal{B}(Y)\)measurable.
3 The result
Theorem 1.2 above follows from the following more general result.
Theorem 3.1
Let \(n,k\in\mathbf{N}\), \(a>0\), \(F:[0,a]\times(\mathbf{R}^{n})^{k}\to2^{\mathbf{R}^{n}}\) a bounded multifunction with nonempty closed values. Assume that F is \(\mathcal{L}([0,a])\otimes\mathcal{B}((\mathbf{R} ^{n})^{k})\)measurable, and that for all \(t\in[0,a]\) the multifunction \(F(t,\cdot)\) is lower semicontinuous. Then there exists a solution \(u\in W^{k,\infty}([0,a],\mathbf{R}^{n})\) of problem (1).
Proof
Let \(H:=\overline{\mathrm{conv}}(F([0,a]\times(\mathbf{R}^{n})^{k}))\). We now divide the proof into two steps.
 (a)
\(\psi(t,\xi )\in F(t,\xi )\) for all \((t,\xi )\in [0,a]\times (\mathbf{R}^{n})^{k}\);
 (b)
for all \(\xi \in(\mathbf{R}^{n})^{k}\setminus[\bigcup_{i=1}^{nk} P_{nk,i}^{1}(C)]\), the function \(\psi( \cdot ,\xi )\) is \(\mathcal{L}([0,a])\)measurable;
 (c)for all \(t\in[0,a]\setminus K_{0}\), one has$$\bigl\{ \xi \in\bigl(\mathbf{R}^{n}\bigr)^{k}:\psi(t,\cdot ) \mbox{ is discontinuous at } \xi \bigr\} \subseteq \Biggl[\bigcup _{i=1}^{nk}P_{nk,i}^{1}(C) \Biggr]. $$
Since \(m_{nk}(\bigcup_{i=1}^{nk}P_{nk,i}^{1}( C))=0\), there exists a countable set \(D\subseteq (\mathbf{R}^{n})^{k}\setminus[\bigcup_{i=1}^{nk}P_{nk,i}^{1}(C)]\), which is dense in \(\mathbf{R}^{nk}\).
 (a)′:

G has nonempty closed convex values;
 (b)′:

for all \(\xi \in(\mathbf{R}^{n})^{k}\), the multifunction \(G( \cdot ,\xi )\) is \(\mathcal{L}([0,a])\)measurable;
 (c)′:

for all \(t\in[0,a]\), the multifunction \(G( t, \cdot )\) has closed graph;
 (d)′:

if \(t\in[0,a]\), and the function \(\psi(t,\cdot )\) is continuous at \(\xi \in(\mathbf{R}^{n})^{k}\), then one has \(G(t,\xi )=\{\psi(t,\xi )\}\).
Moreover, we have \(G([0,a]\times(\mathbf{R}^{n})^{k})\subseteq H\subseteq \overline{B}_{n}(0,M)\).
 (a)^{′′} :

For every \(v\in L^{\infty}([0,a],\mathbf{R}^{n})\), and every sequence \(\{v_{m}\}\) in \(L^{\infty}([0,a],\mathbf{R}^{n})\), weakly converging to v in \(L^{1}([0,a],\mathbf{R}^{n})\), the sequence \(\{\Phi (v_{m})\}\) converges strongly to \(\Phi (v)\) in \(L^{1}([0,a],\mathbf{R}^{nk})\). To see this, let \(\{v_{m}\}\) and v in \(L^{\infty}([0,a],\mathbf{R}^{n})\) be fixed, with \(\{v_{m}\}\) weakly convergent to v in \(L^{1}([0,a],\mathbf{R}^{n})\). It is quite easy to check that the sequence \(\{\phi_{1}(v_{m})\}\) converges pointwise in \([0,a]\) to \(\phi_{1}(v)\). That is, for every \(t\in[a,b]\) one has \(\lim_{m\to\infty} \\phi_{1}(v_{m})(t)\phi_{1}(v)(t)\_{n}=0\). Since \(\{v_{m}\}\) is weakly convergent in \(L^{1}([0,a],\mathbf{R}^{n})\), \(\{v_{m}\}\) is bounded in \(L^{1}([0,a],\mathbf{R}^{n})\). Consequently, for all \(m\in\mathbf{N}\) and \(t\in[0,a]\) we getBy the dominated convergence theorem we get$$\begin{aligned} \bigl\Vert \phi_{1}(v_{m}) (t)\phi_{1}(v) (t) \bigr\Vert _{n} \le& \biggl\Vert \int_{0}^{t}v_{m}(s) \,ds\biggr\Vert _{n}+ \bigl\Vert \phi_{1}(v) (t)\bigr\Vert _{n} \\ \le & \bigl\Vert \phi_{1}(v) (t)\bigr\Vert _{n} +\sup _{j\in\mathbf{N}}\v_{j}\_{L^{1}([0,a],\mathbf{R}^{n})}. \end{aligned}$$hence \(\{\phi_{1}(v_{m})\}\) converges strongly to \(\phi_{1}(v)\) in \(L^{1}([0,a],\mathbf{R}^{n})\). Of course, this implies that \(\{\phi_{1}(v_{m})\}\) also converges weakly to \(\phi_{1}(v)\) in \(L^{1}([0,a],\mathbf{R}^{n})\). Consequently, using exactly the same argument as before, it is immediately seen that \(\{\phi_{2}(v_{m})\}\) also converges strongly to \(\phi_{2}(v)\) in \(L^{1}([0,a],\mathbf{R}^{n})\). By using recursively the same argument as before, we get that for each fixed \(j=1,\ldots, k\), the sequence \(\{\phi_{j}(v_{m})\}\) converges strongly to \(\phi_{j}(v)\) in \(L^{1}([0,a],\mathbf{R}^{n})\). Of course, this implies that the whole sequence \(\{\Phi (v_{m})\}\) converges strongly to \(\Phi (v)\) in \(L^{1}([0,a],\mathbf{R}^{nk})\), as desired.$$\lim_{m\to+\infty} \int_{0}^{a}\bigl\Vert \phi_{1}(v_{m}) (t)\phi_{1}(v) (t)\bigr\Vert _{n} \,dt=0, $$
 (b)^{′′} :

Let d denote the usual pointtoset distance in \(\mathbf{R}^{n}\), with respect to the norm \(\\cdot\_{n}\). Then he functionbelongs to \(L^{\infty}([0,a])\) and \(\\omega\_{L^{\infty}([a,b])}\le M\) (as regards the measurability of ω, we refer to [9]).$$\omega:t\in[a,b]\to\sup_{\xi \in\mathbf{R}^{nk}} d\bigl(0_{\mathbf {R}^{n}},G(t,\xi )\bigr) $$
Of course, for every fixed \(t\in[0,a]\) the multifunction \(\hat{F}(t,\cdot )\) is lower semicontinuous in \((\mathbf{R}^{n})^{k}\). Moreover, by Lemma 2.6, F̂ is \(\mathcal{L}([0,a])\otimes\mathcal {B}((\mathbf{R}^{n})^{k})\)measurable.
Proof of Theorem 1.2
Remark 3.2
Declarations
Acknowledgements
The second author was partially supported by the Grant MOST 1032923E039001MY3.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
Authors’ Affiliations
References
 Aubin, JP, Cellina, A: Differential Inclusions: SetValued Maps and Viability Theory. Springer, Berlin (1984) View ArticleMATHGoogle Scholar
 Smirnov, GV: Introduction to the Theory of Differential Inclusions. Graduate Studies in Mathematics, vol. 41. Am. Math. Soc., Providence (2001) Google Scholar
 Bressan, A: On differential relations with lower continuous righthand side. An existence theorem. J. Differ. Equ. 37, 8997 (1980) MathSciNetView ArticleMATHGoogle Scholar
 Aubin, JP, Frankowska, H: SetValued Analysis. Birkhäuser, Boston (1990) MATHGoogle Scholar
 Klein, E, Thompson, AC: Theory of Correspondences. Wiley, New York (1984) MATHGoogle Scholar
 Himmelberg, CJ: Measurable relations. Fundam. Math. 87, 5372 (1975) MathSciNetMATHGoogle Scholar
 Cubiotti, P, Yao, JC: Twopoint problem for vector differential inclusions with discontinuous righthand side. Appl. Anal. 93, 18111823 (2014) MathSciNetView ArticleMATHGoogle Scholar
 Cubiotti, P, Yao, JC: On the twopoint problem for implicit secondorder ordinary differential equations. Bound. Value Probl. 2015, 211 (2015) MathSciNetView ArticleMATHGoogle Scholar
 Naselli Ricceri, O, Ricceri, B: An existence theorem for inclusions of the type \(\Psi(u)(t)\in F(t,\Phi(u)(t))\) and application to a multivalued boundary value problem. Appl. Anal. 38, 259270 (1990) MathSciNetView ArticleMATHGoogle Scholar
 Villani, A: On Lusin’s condition for the inverse function. Rend. Circ. Mat. Palermo 33, 331335 (1984) MathSciNetView ArticleMATHGoogle Scholar
 Hewitt, E, Stromberg, K: Real and Abstract Analysis. Springer, Berlin (1965) View ArticleMATHGoogle Scholar
 Anello, G: An existence theorem for an implicit integral equation with discontinuous righthand side. J. Inequal. Appl. 2006, Article ID 71396 (2006). doi:10.1155/JIA/2006/71396 MathSciNetView ArticleMATHGoogle Scholar