# JOURNAL

OF MATHEMATICAL

ANALYSIS

AND

APPLICATIONS

152, l-5 (1990)

On the Cauchy Differential Discontinuous
NIZAR

Problem for Ordinary Equations with Right-Hand Sides
RZYMOWSKI

HASSAN AND WITOLD

Instytut Matematyki UMCS, 20-031 Lublin, Poland Submitted by V. Lakshmikantham Received December 27, 1988

We deal with the Cauchy problem for a differential equation x’ =f(r, x), where f is a scalar discontinuous function defined in a cone in R*. We are only interested in solutions which have right-hand derivatives satisfying the equation for all t>O. As an application we give an existence theorem for autonomous systems on the plane. These results can be treated as a very particular answer to the question raised in [S, Sect. 16.5, p. 1131. 0 1990 Academic press, IIIC.

1. AN EXISTENCE THEOREM FOR REAL-VALUED Let [a,b]cRand
DEFINITION

RIGHT-HAND

SIDES

K={(t,x)ER’:t>O,

at<x<bt}.

a-continuous,

1. We say that f: K + [a, b] is angularly continuous or if for any (t, X) E K and any E> 0 there exists 6 > 0 such that

if t<s<t+6, (s-t)a<y-x<(s-t)b, and (S - t)(f(t, x) -E) < y -x 6 (S - t)(f( t, x) + E), then Lo& Y)-f(4 XII <&.

(AC)

Clearly, any continuous function f: K + [a, b] is a-continuous. If f is r-continuous in the sense of [S] or d-continuous in the sense of [S] then it is also a-continuous. The inverse statements are not true as the following example shows. Let K= {(t,x)eR2: t>O, Odx<t} andf: K-+ [0, l] be defined by the formula f(t, xl = x/t j-(4 x) = c where c E [0, 11 is a fixed number. 1
0022-247X/90 \$3.00
Copynght 0 1990 by Academic Press, Inc. All rtghts of reproduction in any form reserved.

if if

t>O and x20, t = x = 0,

x) is neither r-continuous nor d-continuous LEMMA 1. Q(t. we can find such a y defined in a certain interval [to. For (t. O<x<t} {w.2 HASSAN AND RZYMOWSKI Remark 1.x)+(l. y(t))EK. x) + (1. x) E K there exists a cone Cc K with vertex at (t.. > t. co) -+ R such that (a) (b) (c) (t. . For any (t.f(t. tl). x0) E K and any E > 0 there exists an absolutely continuous function y: [to.x) 2-“. E int C(t. By (AC).x(t)).x)=(l+t+x)-l c n~Q(t. x) and with the properties: int C # \$3 (C has non-empty interior). y(t))<& for all tat. (t. x)}. where Dy denotes the right-hand derivative of the function y.. and let W= of K with both C(t. b] satisfy the following condition. xl thenf(t. Continuing this procedure by transtinite induction we complete the proof. x) E K we set (t. x(O)=O.O<(y-x)(l+t+x)<s-t).y)~K:s>t. Let K={(t.x)={(s... (t. and O<Dy(t)-f(t.)=x. This function at any point w E W. x)) E C.)~C and 18 from [Z]. Zf f:K--+[a. if (t.x)<b. f is a-continuous.. with t.) -+ (t. is continuous in K and decreases satisfies the condition of Remark will be appropriate. y(tl)) as a new initial condition. b] is a-continuous then for any (to.x(t))EKfor t>O. Observing that the function (1 + t + x) ~’ when t or x increases. tl]. For any (t. x) = (n E N: w. Proof. and with (tl. Suppose that f: K-+ [a. f(t.. we can show that f 1. x) E K the cone C = C( t. w2. THEOREM 1.x))~intC if a<f(t. Dx(t)=f(t. Dx is continuous from the right. . x)..x. Then.x)~R’:t>o. > t.for tB0. 4 +f(t. t2]. co) -+ R such that y(&. rff: K + [a.. The same holds for an interval [tl.. (t. and with y’ being constant in (to..b] is a-continuous then there exists an absolutely continuous function x: [0. x.. We now give an example off based on the idea of Example EXAMPLE 1.} be th e set of all elements from the interior coordinates rational.. f(t. with t.

z(t)) for t>O. x(t)) + E. We are t 3 0. we have D+x( t) <f(t. The function x* defined by the formula x*(s) = x(s) for SE [0. x(t)). x( . x(t)).. (2) DPz(t)>f(t. It follows from Lemma 1 that the inequality D-x(t) > f( t. As a simple consequence we obtain a known result concerning autonomous systems in R. then D -z(s) > f(s. D-x(t) =f(t. co) -+ R be taken from Lemma 1 for t. t > 0. x(t)). The proof is complete. Since z is an arbitrary element of Z and E> 0 is chosen arbitrarily. co) + R satisfying the conditions: (1) z(O)=0 and z(t)+u(s--)<z(s)<z(t)+b(s-t) for Ogt<s. t] and x*(s) = min{x(s). b] for all t > 0. t*) for some t* > t. Theorem 4.)) is continuous from the right. if z E Z then z(s) > y-(s) in (t.11. x(t)).t)(f(t. Theorem 2. we set and Y’(S) =x(t) + (3 . b]. be continuous from . = t. 12]). D-x(t) E [a.2. Thus. which in turn implies that Dx(t) =f(t.g. x(t)=inf{z(t):z~Z). [IS. and. We have just proved that x satisfies condition (b). We shall prove that x satisfies (2). p. COROLLARY 1. x(t)) contradicts the definition of x. By the definition of x. The family 2 is non-empty going to prove that because z(t) = bt. x(t)) . moreover. and E. Then. see [3. o-x(t) >f(t. y(s)} for s > t belongs to Z and satisfies the inequality D+x*(t) <f(t. chosen for (t. Observation 2. The same equality is valid for the right-hand upper Dini’s derivative D +x( t). 00) + [a. let E > 0 be fixed and let y: [to. Let us fix arbitrary t > 0. x(t)) and E according to condition (AC) of Definition 1. if y~~(s)<z(s)6y’(s) for an s~(t. x(t)). Let h: [0. x(t)) 6 b we fix arbitrary E > 0 and take 6 > 0. satisfies conditions (a) and (b). obviously.E. e. x(t)) = a then. belongs to 2.CAUCHY PROBLEMFORDIFFERENTIALEQUATIONS 3 ProoJ Denote by 2 the set of all functions z: [0. If f(t. t + S]. Utilizing the well-known facts on Dini’s derivatives (see. Indeed. So. z(s)) > f( t. we have D-x(t) >f(t. we conclude from this that z(s) 3 y ~ (s) in [t. x(t)) + El.. It now follows from the a-continuity off that the function f( . In the case of a <f(t. x0 = x(t). with a > 0. Clearly. where D-z(t) denotes the right-hand lower Dini’s derivative of the function z at the point t. for s 2 t. t+6]. 1. It is easy to see that x satisfies (a).

(y. The function f is a-continuous. 1.4 HASSAN AND RZYMOWSKI the right. if e # 0. with vertex at the origin. for t>O. Remark 2. x) E K is a-continuous in the sense of Definition 2. x) E K we set Let K be defined as in Definition f(t. x)) for (t. THEOREM 2. Let us decompose g into the pair (gi. Let Kc R2 be a closed convex cone with vertex at the origin. there exists an absolutely differential equation x’ = h(x). x).S(e.. y E R* and (xl= . If a E R2 and A c R* then a + A denotes the set of all sums a + b. . Let K be a closed convex cone in R2 with vertex at the origin and g: K--P K\ { 0 >.p)= {xER~: (x.)eK. x(0)) = c) iff c = 0. The symbol (x. continuous solution x to the x(0) = 0. For e E R2 and . x) = h(x). y2). with Dy continuous from the right. y) denotes the Euclidean scalar product of the vectors x. where b E A. has a solution in the sense of Theorem 1 (with Dx(0) = f(0.. If g: K+ K\(O) is a-continuous. f (t. then g: K + K defined by the formula g(t. then there exists an absolutely continuous function y: [0. the differential equation Dx = f( t. It is easy to see that if f: K+ [a. Obviously. Proof.O) iff c = 0 and. co) + K satisfying the differential equation Dy(t)=g(y(t)). l] we set . where c is a fixed non-negative number.x)=cift=O. also. for y=(y. containing the half-line { te: t > 0} in its interior.) of two real-valued Then.e)>(l---) 1x1 IeI}. x) = (1.&)andIy-xl~r then g(y) E S( g(x). y. g. y(O)=O. 2. we have g(y)=(g. p) is a closed convex cone in R2. EXAMPLE 2. Then. bounded. We say that g is a-continuous if for any x E K andany&>Othereexistsr>OsuchthatifyEx+S(g(x). Let. and such that inf{ Ig(x)1 : x E K} )O./‘m.f(t. then S(e. Proof functions. s) and (g(y) -g(x)\ < LE (for some L independent of y and E). x(O) = 0. DEFINITION 2. For (t. b] is a-continuous in the sense of Definition 1.uL(O.fort. AN EXISTENCE THEOREM FOR AUTONOMOUS SYSTEMS IN R2 Before giving an analogous result for autonomous systems in R2 we introduce some definitions. The function f is a-continuous at (0.x>O.x)=x/2tift>Oandf(t.

au<z<bu}cK. SZARSKI. “Differential Inequalities. “Counterexamples in Analysis. Obviously h is bounded. h(u) 2 a > 0 for u 2 0.b] is a-continuous. Atfi Sem. 1965. 5. u(O) = 0. cc). by Corollary 1. P. and h is continuous from the right. Let us take a = inf{g2b9/gI(y): b = suPMY)/g. there exists a solution z of the differential equation Dz =f(u. Equazioni dilferenziali ordinarie con second0 membro discontinue. University of Lund. GIUNTINI AND G. Differential Equations 31 (1979) 183-199.” Holden-Day. B. The differential equation x’ =fox. PEETRE AND J. Let us consider the next auxiliary differential equation Du = h(u).(Y): f(k z) =g2(4 Z)kI(% z) system we may assume that Y E KIT YE KI> for (u. where h(u) =gl(u. co) --+ [0. with Du continuous from the right.” PWN. K*={(u. yJ). co). San Francisco/London/Amsterdam. z(u)) for u 2 0. satisfying that equation in [0. z(0) = 0. PERSSON. and f:K*-+[a. Warszawa. let us define yr(t) = u(t) and y2(t)=z(u(t)) f or t>O. BINDING. KRASNOSEL~KII AND A.” Nauka.(t)=Du(t)= gI(yIWy y2(t)) and Dy2(t) = WN) DW =g2bIM y2(th which corn- pletes the proof. .z)ER~:u>O. Then.CAUCHY PROBLEM FOR DIFFERENTIAL EQUATIONS 5 g. z). Now. Lund. Mat.(y. Mimeograph. Choosing a suitable coordinate inf{g. A. For all t>O we have Dy. The Peano existence theorem under weaker assumptions. 233-240. 1964. Thus. J. GELBAUM AND J. PIANIGIANI.. J.(y): YEK} >O. with Dz continuous from the right. Fix Univ. z) E K. Moscow. OLMSTED. 3. 1971. Modem 23 (1974). 1983. S. J. 4. 2. Therefore. there exists an absolutely continuous U: [0. REFERENCES 1. 6. V. M. POKROVSKII. “Systems with Hysteresis.