405 Internat. J. Math. & Math. Sci. Vol. 9 No. 2 (1986) 405-408 RESEARCH NOTES AN INTEGRAL EQUATION ASSOCIATED WITH LINEAR HOMOGENEOUS DIFFERENTIAL EQfJATIONS A.K. BOSE Department of Mathematical Sciences Clemson University Clemson, South Carolina 29631 U.S.A. (Received March 7, 1985) ABSTRACT. y Associated with each linear homogeneous differential equation n-I (n) Z ai(x)y i=O equation (i) of order x f(x O) + I x f(x) 0 n on the real line, there is an equivalent integral x u h(u)du + f x [I x 0 G 0 f(x) which is satisfied by each solution n-1 (u v)a 0(v)f(v)dv]du of the differential equation. KEY WORDS AND PHRASES. Linear homogeneous differential equations, Integral equations, Initial value problems, Variation of parameters formula, Uniform convergence. 1980 MATHEMATICS SUBJECT CLASSIFICATION CODE. 34A i. INTRODUCTION. Let be a positive integer, I n I. be the class of all functions continuous on y (n) a n-I (x)y (n-l) R be an interval on the real line + + and C(1) Let aO(x)y x (I I) I be any n-th order (normalized) ordinary linear homogeneous differential equation, where a i(x) (n-l). e C(I), i=0,I,2 The purpose of this article is to derive an equivalent integral equation satisfied by the solutions of the linear homogeneous differential equation (I.I). 2. MAIN RESULTS. be a solution of (l.l) defined on f(x) Let THEOREM. I and x 0 e I. Then f(x) is also a solution of the integral equation x x f(x) where h(x) f(x O) u + f h(u)du + I If x x 0 0 x Gn_l(U,V)ao(v)f(v)dv]du, (2.1) 0 is the unique solution of the (n-l)-th order linear homogeneous differential equation y (n-l) an_ (x)y(n-2) + + a l(x)y x I (2.2) A.K. BOSE 406 satisfying the initial conditions Y(X0) and G f n-I (x,u) (x0), Y (x 0) f (x 0) Y Green’s is the well-known (n-2) (x 0) f (n-l) (x0), Function associated with the homogeneous equation (2.2) PROOF OF THE THEOREM. In order to deduce the integral equation (2. I), we will use the well-known Variation of Parameters formula X h(x) + f y(x) x Gn_ (x,u) (u)du (2.3) 0 solving uniquely the non-homogeneous initial value problem y (n-l) y(x 0) a (x)y (n-2) + n-1 (Xo), f Y (x 0) + f (x a + (x) l(x)y 0) Y (n-2) (Xo) f(n-l) (x 0 for each (x) c C(1). Consider the sequence of functions: fl (x), f2(x) fk(x) I, where defined on X fl(x) + I h(u)du f(x0) f2(x) x fl(x) fk(x) Clearly, for each k, X U + I Ef x fl (x) 0 0 x X U + I [I x fk(x0) 0 x G n-1 (u,v)a0(v)f (v)dv3du 0 (2.4) Gn_ (u,V)ao(V)fk_l (v)dv]du 0 f(Xo) fk(x) is differentiable on I and for k 2 X fk(x) h(x) + I x Gn_l(X,u)a0(u)fk_l(U)du. (2.5) 0 Using (2.3), we conclude that, for each k a 2, fk(x) is the unique solution of the non-homogeneous initial value problem y (n-l) a n-l (x)y (n-2) + + a l(x)y + aO(x)fk_ l(x) (2.6) Hence each (x0), Y(X0) f fk(x) n e C (I). Y (x 0) f (x 0) Y (n-2) (x 0) (n-l) f (Xo). Both the sequences {fk(x)} and {fk(x)} converge uniformly on every compact subset on the interval I. To see this, let B be a compact subset of I. Then there exists a closed and bounded interval [a,b] such that B I and [a,b] INTEGRAL EQUATION ASSOCIATED WITH DIFFERENTIAL EQUATIONS x 0 407 [a,b]. e M Let IGn-l(U’v)a0(v)l max s Ifk+ l(x) Ifk+ l(x) fk(x) -< Mksk(b-a)2k/2k! fk(x) _< Mksk(b-a)k/k! fi+ l(x) using recursively the bounds for the series Ifl(v) max for each u,v [a,b]. One can [a,b], now see very easily that for each x e fi(x) Mksk(b-a)2k/2k!, Mksk(b-a)k/k! 1,2,3, i Since each of converges, we conclude by Weierstrass’ M-test that each of the series of functions + fl(x) + fk(x)), fl(x) (fk+l (x) k=1 converges uniformly on [a,b] and hence on g(x) e C such that B. k=l (fk+l (x) fk(x)) Therefore there is a function I(I) for all fl (x) + fl(x) + . I k=l k-I (fk+l (x) fk (x)) (fk+l(X) fk(x)) In particular x e I. g(x O) lira k+= lira k= fk(x) g(x) fk(x) g (x) f(Xo). Also from (2.5) we get by taking limit as k- x h(x) + I g (x) x G n-i (x U)ao(u)g(u)du x e I (2.7) 0 Hence g(x) f(Xo) x x u + I h(u)du + I [ x x 0 x 0 Gn_l(U,V)ao(v)g(v)dv]du (2.8) 0 Again, relation (2.7) implies by (2.3) that g (x) is the unique solution of the initial value problem Therefore Y(n-l) an_l(x)y(n-2) + Y (x f (x n g(x) e C (I) g (n) In other words, y f(x) O) Y (x O) + l(x)y f aO(x)g(x (n-l) (Xo). and (x)gn-I (x) + + a g(x) is the unique solution of the homogeneous initial value problem a n-i n-I (x)y O) f(Xo), g(x) for all (n-l) Y (x f(x 0) (x) + O) x e I. x f(x) (n-l) a (x) (n) y(x Hence O) + + x 0 + f (x aO(x)g(x) a0(x)y O) Y x e I x e I (n-l) (x O) f (n-l) (Xo). Therefore, by (2.8) x u h(u)du + I [ x x 0 G 0 n-I (u,v)ao(v)f(v)dv]du A.K. BOSE 408 This completes the proof. REMARK. The above proof clearly shows how a solution of a linear homogeneous equation with prescribed initial values can be constructed out Qf a solution and the Green’s Function G n-I (x u) h(x) of a lower order homogeneous linear equation. This is specially significant in case of second order homogeneous equations, as ceA-X’t solutions and the Green’s function Gl(X,U) e A(x)-A(u) x [A(x) a f x l(u)du], of first order homogeneous equation y a l(x)y are readily 0 available. REFERENCES I. CODDINGTON, E.A. 2. WIDDER, D.V. An Introduction to Ordinary Differential Equat.!ons Cliffs, New Jersey: Prentice-Hall, Inc. Inc. Advanced Calculus, Englewood Cliffs, New Jersey: Englewood Prentice-Hall,