Internat. J. Math. & Math. Sci. VOL. 13 NO. (1990) 159-164 159 ITERATIVE SOLUTION OF NEGATIVE EXPONENT EMDEN-FOWLER PROBLEMS C.D. LUNING Department of Mathematics Sam Houston State University IIunt sville, TX 77340 W.L. PERRY Department of Mathematics Texas A & M University College Station, TX 77843 (Received June 21, 1988) ABSTRACT. A monotone iterative technique for proving existence of a positive solution pair (,y) of y"(x) +X shown. a(x)y(x) < 0, 0 x < I, y(0+) y(l-) u < -I is 0, The method is computationally effective and an example is given. KEY WORDS AND PHRASES. Monotone iterations, singular Emden-Fowler problems. 34B15. 1980 AMS SUBJECT CLASSIFICATION CODE. INTRODUCTION. In this paper a monotone Iteratlve technique for finding a positive solution pair (k,y) of /’(x) + y(O+) is shown. and a(x) y(1-) y(x) < 0, 0 x < (I.I) O, We require a(x) to be non-negative and continuous on [0,I] with x -I a(x) (l-x) -I a(x) being bounded on (0,1). and y’(0+) and y’(l-) are finite. Under these conditions, a solution exists was This shown by Tallaferro methods in that paper were not constructive in nature. sequence which is paper. proved to converge In other work the authors [2] to a positive [I]; Development solution is have obtained results for tJ o however the an iteratlve this our aim in >- I. Negative C.D. LUNING AND W.L. PERRY 160 problems Emden-Fowler exponent have some significant catalysis and fluid mechanics (Lunlng and Perry in applications permeable [3], [4]). THE MONOTONE ITERATION. 2. The iteratlve sequence is defined as follows: (1-x) (1-)x K(x,) 0 < x x ( 1, Let Uo(X) x, and n u (x) n f 0 K(x,) a()u n-I () d (I-O a() 0 n-- 1,2 .... (2.1) Un_ 1,2,3,..., on (x), n Moreover, since K(x,) is the Green’s The conditions on a(x) guarantee the existence of u < 0 < x and that u’(l-) is finite for all n. u" function for O, u(O) + u"(X)n u(1) O, the sequences " (x) n a(X)Un-I Un(l-) u (0+) n {Un(X)} {n satisfy 0 (2.2) 0 u’ (0+) n The sequences LEMMA. 0 < x < 2, {Un(X)} {n I, For n are actually alternating monotone. U2n_l(X) < U2n+l(X) < U2n(X) < UZn_Z(X) 2n < 2n+2 < 2n+l < 2n-l" one value of x e (0,I) such that Moreover, for any 0 Muir(x) B(x) u < M < there is at most O, where the pair (i,J) can be (2n-2,2n-l), (2n,2n-l), (2n-2,2n), or (2n+l,2n-l). Before proving the lemnm, we note that the result implies that {u ,o (l-)}n= is a bounded sequence. PROOF OF THE LEMMA. (u0 u l)’’(x) so that u 0 > Ul, u )’la(x) 0 < x First we compare u < I. (x) > Let O, 0 < fO, l(X) x < M and u - 0 1, I. We have (Uo-U I)’(0+)’= (Uo-UI)(0) u0(x) Ul(X) where 0 < M < O, I. Now fO,l ’’(x) -Ul"(x) )’I a(x) u((x) > O, 0 < x < 1, fO,l(O) O, fO,l’(O) M-1/-1 < Thus fo ,I (x) can cross the x axis at most once interior to (0,I). Since u0 > Ul, the definition of k leads to < I" O. ITERATIVE SOLUTION OF NEGATIVE EXPONENT EMDEN-FOWLER PROBLEMS In a completely similar way it is shown that u 0 M-l/Bu0(x)-u2(x) has f0,2(x) (0,I) such that 12n+2 < 2n+! O, 0 < x < l, To show that For the induction > () (U2n-U2n+2) -I/u2n-u2n+l (M-I/U2n-U2n+?)(x)_ O. 2n+3 < 12n+l" and > < 0, 0 x < l, O, and there is at (x) (O,l) such that (M there is at most one x most one x (x) and (0,I). at most one zero interior to (U2n-U2n+l) hypothesis we assume 3 < ll’ > ’*2’ 161 This implies that (U2n+2-U2n+l)(X) > 0, 0 < < x I, we proceed as follows (u 2 n+ 2 -u X2n+ 2 a ’2n+ 2n+ -n+la(x)Un(X)U2n+l (x) 12n+2 n+l X2n+2/X2n+l < Since (x)] 2n- ( x)-u n+ (U2n+2-U2n+l) we conclude that (x) (0,1) such that is shown that there is at most one x e > O, 0 X2n+3 > )’2n+2" One then proceeds in the same manner (U2n+3-U2n+l) (x), (M-1/U2n+3-U2n+l) (x), (U2n+2-U2n+3) (x), (U2n+2-U2n+4) the ’s. < 1. Similarly it O. to consider (M-1/U2n+2-U2n+3) (x) along with the definition of (x), Xn to order This completes the induction proof of the lemma. < k2n < 12n+2 < A2n+l < A2n-l’ , llmn Because 0 2n llmn Uo(X), (M-1/laU2n+2-U2n+4) (x), x (M-1/U2n+2-U2n+l) (x) Therefore and < 0 < x < " 2n_l} sequences {u < A < A such that < U2n+l(X) < U2n(X) < U2n_2(x) < Since 0 -2n+l I, the there exists 0 and {U2n} are uniformly bounded monotone sequence s. We can also conclude that the sequences are equicontinuous because By Ascoli’s lemma there exist functions uniformly bounded. 0 < (x) [O,l]. 4 (x), 0 < x < I, and limn and (R)U2n-I 5, llmn + -U2n {lu n ’(x) i} C[O,I] such that -u uniformly on Using the dominated convergence theorem and (2.1) we have (x) f K(x,)a()51() d, 0 (x) (2.3) X f K(x,)a()tl() 0 d or, equivalently, a"(x) + a(x)[tl(x) O, 0 < x < is C.D. LUNING AND W.L. PERRY 162 "(x) . + G(o) a(x)(x) G(o) fi(1) 0 G(1) o. Now we have two possibilities. some subinterval (0,I). of l,(o) G(1) Either x c o , (2.4) 1, - (x) on (0,I) or (x) < fi(x) on the pair (k,y) -(,) is a and Otherwise we proceed as follows. Now let the constant C e (0,1) be chosen so that -(cG") < (x) then If solution to the problem (I.I). < O, a(x)G" < c 3, a(x) CI- . Since we have Then we have ," c-P(x) (c[,)" , c,(o):cG() o. and a subsolutlon In this way we have a supersolut[on a(x) -y"(x) y(O) . yU(x), y(1) (cG)" a(x) 0 < x < C for the problem I, (2.5) 0 p-I Since x a(x) and (l-x)-I a(x) are bounded, the function @ ) Therefore there f(x,u)= ka(x)u has @f/@u bounded, 0 < x < I, (x) < u < x). with exists m > 0 such that f(x,) for all x e [0,I] and f(x,n) , > -m(-l) n e [,@] with > n (where one sided limits are used at 0 and I). The foregoing conditions are sufficient for the iteration schemes (see Amann page 648). [5], ITERATIVE SOLUTION OF NEGATIVE EXPONENT EMDEN-FOWLER PROBLEMS -Vk+ l(x) +Vk+ l(x) Vk+ 1(0) v 0 , Vk+ l(0) or Vk+ 1(l) in ,taxi.at qol,tion Iteratlo scheme (2.6). 0 ,) of the problem positive solution pair (\,y) exists for problem (1.[) (?.I). Moreover the ei,envalue the If )y (that is if iteration scheme (2.[) i teratlve scheme, in (if v [,]. and is obtainable by monotone iteration. the (2.6) 0 @) or a 0 Thus we have shown t-hat by Vk(X) Vo to converge to a minimal (if (2,5), the solution lying a(x) vkU(x) + i 163 functiou y s not is determined determined by the is found by the appllca{ion of the second To implement the method numerically, special care must be taken in all calculations near the endpolnts. 3. EXAL LE. Calculations were performed for the problem y"() + k 3(i_x)3 0. y(1) y(0) y- 2 () =0 The iterative method was applied using the integral formulation, using Simpson’s rule away from the endpoints. "- ’10 and y Xll "(ulO + u II)/2; 7 173, "pinched together". The sequences were alternating monotone and we found 7 2213 that maxlu Because is, we -u < 10 considered that .0003 we used Athe Iterative (0 + kl sequences )/2 had The calculated solution was "- 7.1975 0 0 .2 1.3 .4 .5 where y(x) is symmetric about x I/. I181 I".2002 .222 .2292 Vntqueness of solution follows from a result of "raliaferro [I]. 4. NOTE. Our conditions on a(x) existence res,It. if f 0 t(l-t)a(t)dt are more For a(x) continuous < -Ill. restrictive e,n than those for a nonconstructive (0,I) a solution exists if and only 164 C.D. LUNING AND W.L. PERRY REFERENCES I. TALIAFERRO, S.., A nontlnear singular boundary value problem Nonlinear Aalysi 3 (1979), 897-904. 2. LUNING, C.D. and PERRY, W.L., Positive solutions of negative exponent generalized Emden-Fowler boundary value problems, qIAM J. ,Math. Analysis 12 3. (198[), 874-879. PERRY, W.L., A monotone terative technique for solution of reaction-diffusion problems 4. (1984), 353-357. LUNING, C.D. and PERRY, .L., in permeable tteratlve catalysts. J. p order (p < 0) Comput. Chemistry 5 method for solution of a boundary value problem in non-Newtonlan fluid flow, J. of Non-Newtonlan Fluid Mech. 15 5. (1984), 145-154. AMANN, H., Fixed point equations aud nonlinear elgenvalue problems Banach spaces, SIAM Review. 18 (1976), 620-70q. in ordered