Microsoft Word - 22 513 | Mathematics 2014) عام 3العدد ( 27مجلة إبن الھيثم للعلوم الصرفة و التطبيقية المجلد Ibn Al-Haitham Jour. for Pure & Appl. Sci. Vol. 27 (3) 2014 Efficient Semi-Analytic Technique for Solving Nonlinear Singular Initial Value Problems Luma N. M. Tawfiq Reem W. Hussein Dept .of Mathematics \ College of Education for pure Scince (Ibn Al-Haitham)/ University of Bagdad Received in :23 September 2012 , Accepted in: 9 December 2012 Abstract The aim of this paper is to present a semi - analytic technique for solving singular initial value problems of ordinary differential equations with a singularity of different kinds to construct polynomial solution using two point osculatory interpolation. The efficiency and accuracy of suggested method is assessed by comparisons with exact and other approximate solutions for a wide classes of non–homogeneous, non–linear singular initial value problems. A new, efficient estimate of the global error is used for adaptive mesh selection. Also, analyze some of the numerical aspects relevant for the implementation, describe measures to increase the efficiency of the code and compare its performance with the performance of established standard codes for singular initial value problems. Many examples are presented to demonstrate the applicability and efficiency of the suggested method on one hand and to confirm the convergence order on the other hand. Kay ward : ODE, Singular initial value problems . 514 | Mathematics 2014) عام 3العدد ( 27مجلة إبن الھيثم للعلوم الصرفة و التطبيقية المجلد Ibn Al-Haitham Jour. for Pure & Appl. Sci. Vol. 27 (3) 2014 1. Introduction The modeliy of nonlinear phenomena in physics, engineering and other sciences, leads to singular initial value problems (SIVP) associated with nonlinear second order ordinary differential equations (ODE) [1]. The general form of the SIVP is : xm y" + f(x, y, y') = 0 , 0 < x < 1 , (1) with some initial conditions. The object is to find a function y that satisfies the differential equation together with IC. We view f as a generally nonlinear function of y and y', but here it wins, we as look f = f(x) only. For such a problem to have a solution, it is generally necessary either that f(x) ≠ 0 hold. When f(x) ≡ 0, and/or A = 0 , B = 0 the IVP is said to be homogeneous and will in general have only the trivial solution, y(x) ≡ 0 [2]. 2. Solution of Second Order SIVP In this section, we suggest a semi analytic technique which is based on osculatory interpolating polynomials P2n+1 and Taylor series expansion to solve second order SIVP, we consider the problem (1) , where f are in general nonlinear functions of their arguments . Now, to solve the problem by suggested method we do the following steps: Step one :- Evaluate Taylor series of y(x) abut x = 0 : y = a 0 + a 1 x +   2i a i x i (2) where y(0) = a0 , y'(0) = a1 ,y"(0) / 2! = a2 ,…, y(i)(0) / i! = ai , i= 3, 4,… And evaluate Taylor series of y(x) about x = 1 : y = b 0 + b 1 (x-1) +   2i b i (x-1) i (3) where y(1) = b0 , y'(1) = b1 , y"(1) / 2! = b2 ,… , y(i)(1) / i! = bi ,i = 3,4,… Step two :- Insert the series form (3) into equation (1) and put x = 1 ,then equate the coefficients of powers of (x-1) to obtain b2 . Step three :- Derive equation(1) with respect to x, to get new form of equation say (4) xm y''' + m xm-1 y'' + df(x, y, y' ) / dx = 0 (4) then, insert the series form (2) into equation (4) and put x = 0 and equate the coefficients of powers of x to obtain a2 ,again insert the series form (3) into equation(4) and put x = 1, then equate the coefficients of powers of (x-1) to obtain b3 . Step four :- Iterate the above process many times to obtain a3 ,b4 ,then a4 , b5 and so on, that is , to get ai and bi for all i ≥ 2, the resulting equations can be solved using MATLAB version 7.10, to obtain ai and bi for all i  2 . Step five :- The notation implies that the coefficients depend only on the indicated unknowns a0, a1, b0, b1, and use the IC to get a0 , a1, therefore we have only two unknown coefficients b0 and b1. Now, we can construct two point osculatory interpolating polynomials P2n+1(x) from these coefficients ( aisۥ and bisۥ ) by the following : 515 | Mathematics 2014) عام 3العدد ( 27مجلة إبن الھيثم للعلوم الصرفة و التطبيقية المجلد Ibn Al-Haitham Jour. for Pure & Appl. Sci. Vol. 27 (3) 2014 P2n+1 =   n i 0 { ai Qi(x) + (–1)i bi Qi(1–x) } (5) where Q j (x) / j! = ( x j / j!) (1–x) 1n    jn s 0        s sn xs Step six :- To find the unknowns coefficients b0 and b1, integrate equation (1) on [0, x] to obtain : xm y'(x) – mxm-1y(x) + m(m–1)  x 0 xm-2 y(x) dx +  x 0 f(x, y, y') dx = 0 (6a) and again integrate equation (6a) on [0, x] to obtain : xmy(x) –2m x 0 xm-1y(x) dx +m(m–1) x 0 (1–x)xm-2y(x)dx+ x 0 (1–x)f(x,y,y') = 0 (6b) Step seven :- Putting x = 1 in equations (6) to get : b1 – mb0 + m(m-1)  1 0 xm-2 y(x) dx +  1 0 f(x, y, y') dx = 0 (7a) and b0–2m 1 0 xm-1y(x)dx +m(m–1) 1 0 (1–x)xm-2 y(x)dx +  1 0 (1–x)f(x,y, y')dx = 0 (7b) Step eight :- Use P2n+1(x) which defined in equation (5) as a replacement of y(x) , we see that equations (7) have only two unknown coefficients b0 and b1. So, we can find b0 and b1 by solving the system of algebraic equations (7) using MATLAB, so insert the value of b0 and b1 into equation (5) , thus equation (5) represents the solution of the problem . Note Extensive computations have shown that this generally provides a more accurate polynomial representation for a given n . 3. Examples In this section, many examples will be given to illustrate the efficiency, accuracy , implementation and utility of the suggested method. The bvp4c solver of MATLAB has been modified accordingly so that it can solve some class of SIVP as effectively as it previously solved nonsingular IVP. 516 | Mathematics 2014) عام 3العدد ( 27مجلة إبن الھيثم للعلوم الصرفة و التطبيقية المجلد Ibn Al-Haitham Jour. for Pure & Appl. Sci. Vol. 27 (3) 2014 Example 1 order nonlinear SIVP : ndConsider the following 2 +18x + 4 3+ 4x 6= x 2y" + (4/x) y' + y with IC : y(0) = 2 , y'(0) = 0 It is clear that x = 0 , is singular point and it is singularity of first kind and the exact solution is y(x) = x3 + 2. Now, this example is solved by using semi-analytic technique, from equation (5) we have : P3 = x3 +2 Example 2 order nonlinear SIVP : ndConsider the following 2 y" + (6/x) y' + (6/x2) y + y2 = 20 + x4 with IC : y(0) = 0 , y'(0) = 0 It is clear that x = 0 , is singular point and the exact solution is y(x) = x2. Now, this example is solved by using semi–analytic technique, from equation(5) we have :P3 = x2 Now,the suggested method is used to investigate the solution of Emden–Fowler equations as a class of nonlinear SIVP, which plays very important role because of the singularity behaviour at the origin. We study the Emden–Fowler equations of the form : , b ϵ R (8) 0 ; 0 ≤ x ≤ 1 , a = g(x) n y 1-my" + ( a/x ) y' + b x under the following initial conditions : y(0) = A , y'(0) = 0 where A, a, b, m are constants, n ≠ 0, n ≠ 1. When m = 1, a = 2 and b = 1 equation (8) reduces to the Lane – Emden equation [3]. Example 3 Consider the following Emden – Fowler equation as a SIVP : ; 0 ≤ x ≤ 1 4+ x 5= x 2y" + ( 8/x ) y' + x y with IC : y(0) = 1 , y'(0) = 0 It is clear that x= 0, is nonlinear singular point of the first kind, we solve this example by using semi-analytic technique , from equation (5), we have : P9 = – 0.00078826882 x9 + 0.0012952791 x8 + 0.00879760172 x7 + 0.014414828109 x6– 0.000150192x5 – 0.0333333333337578 x3 + 1.0 higher accuracy can be obtained by evaluating higher n, now, take n= 5, i.e., + 7+ 0.0014538812 x +0.009581210401 x90.00189361335x -10+ 0.00086427359x110.00021822695x – 11=P + 1.0 30.033333333333758 x -6 x 0.0137816833 Now, increase n, to get higher accuracy , let n = 6 , i.e. , 517 | Mathematics 2014) عام 3العدد ( 27مجلة إبن الھيثم للعلوم الصرفة و التطبيقية المجلد Ibn Al-Haitham Jour. for Pure & Appl. Sci. Vol. 27 (3) 2014 P13= - 0.0000996316x13 + 0.0004488478x12 - 0.000849659317x11 + 0.0007848859x10 - 0.0007538654 x9 + 0.0001852901x8 + 0.01017813169x7 + 0.01367521368x6 - 0.03333333333375776x3 + 1.0 Demir[3] solved this example by DTM and gave the following solution: Also, this example is solved by numerical method by using the Maple11 as given in [3] and the results are compared with the suggested method and its results are given in table (1a) and (1b). Figure (1) give a comparison between the numerical solution in[3] and P13 ; table (2) gave the results obtained with DTM for different values of n which are given [3] . 4. Error / Defect Weights Every known IVP software package reports an estimate of either the relative error or the maximum relative defect. The weights used to scale either the error or the maximum defect differ among IVP software. Therefore, the IVP component of pythODE [4] allows users to select the weights they wish to use. The default weights depend on whether an estimate of the error or maximum defect is being used. If the error is being estimated, then the IVP component of pythODE uses, in this thesis we modify this package to consist SIVP and named " pythSODE ", defined as : ║y(x) − P(x)║∞ / ( 1 + ║P(x)║∞ ) ; 0 ≤ x ≤ 1 , (9) where y(x) is exact solution and P(x) is suggested solution of SIVP (1). If the maximum defect is being estimated, then the SIVP component of pythSODE uses : ║P(x)−f(x, P(x), P(x))║∞ /(1 +║f(x, P(x), P(x) )║∞) ; 0 ≤ x ≤ 1 , (10) The relative estimate of both the error and the maximum defect are slightly modified from the one used in SIVP SOLVER . Again applied equation (10) for example 3 is as following : 4+ x 5x= 2Consider the nonlinear SIVP : y" + (8/x) y' + x y with IC : y(0) =1 , y'(0) = 0 , apply equation (10) as the following : Then equation (10) became : ( for more detail see table 3 ) ║P13(x) - f(x, P13(x), P13(x))║∞ / ( 1 + ║f(x, P13(x), P13(x) )║∞ ) = e-007 / 1.999966681360863 = 8.419941391601780e-0081.683960224221478 5. Accuracy of The Computed Values The convergence orders of the approximations appear to be very stable, indicating that our results are reliable and accurate. In this section, we try to quantify the accuracy we can obtain. To examine the error of the approximation, let E(n) = max | y(xi) − yi | (11) 518 | Mathematics 2014) عام 3العدد ( 27مجلة إبن الھيثم للعلوم الصرفة و التطبيقية المجلد Ibn Al-Haitham Jour. for Pure & Appl. Sci. Vol. 27 (3) 2014 where y(xi) is the exact solution at the xi and yi suggested solution at xi . It is known that for a large N, if M is the order of convergence of the method, then : E(n)  K |n|M (12) where K is the error constant and E(n) is the max error . Following the generalized algorithm for the order verification of methods given in [5], we can find M and K from the line : y = M x + log K (13) that best fits the equation : log E(n) = log K + M log n (14) where y represent log E(n) and x represent log n . Applying this algorithm for the following example : y" + (2/x) y' − 6y = − 4ylny , with the IC: y(0) =1 , y'(0) = 0 the exact solution is y(x) = exp( x− 2), we have from table (4) the best fit line whose graph is shown in figure (2) ,i.e., y = − 6.6214 x − 1.3330 From equation (13), we conclude that the order of convergence |M|  6 with error constant : K =101.3330. 6. Conclusions Our decision to use semi-analytic technique for the solution of the underlying initial value problem was motivated by its advantageous convergence properties , whereas in presence of a singularity other direct higher order methods (finite difference) show order reductions and become inefficient. Moreover, the standard acceleration techniques like Iterated . Our reason for choosing the global error estimate instead of monitoring the local error is the non-smoothness of the latter near the singular point and the order reductions it suffers from. The results show that the convergence order estimates are increased, if we increase n . This is not surprising. References 1- Karakostas (2010) C1-Approximate Solutions of second-order Singular Ordinary Differential Equations, Electronic Journal of Differential Equations., 125: 1- 47 . 2- Robert and Courtney (1996) " Differential Equations a Modeling Perspective " , United States of America . 3- Demir and Süngü (2009) Numerical Solution of a Class of Nonlinear Emden-Fowler Equations by Using Differential Transform Method , Journal of Arts and Sciences,. 75-82 . 4- Shampine ( 2002 ) " Singular Boundary Value Problems for ODEs " , Southern Methodist University J., . 1 . 5- Rasheed (2011) "Efficient Semi-Analytic Technique for Solving Second Order Singular Ordinary Boundary Value Problems" , MSc. Thesis, University of Baghdad , College of Education – Ibn – Al- Haitham . 519 | Mathematics 2014) عام 3العدد ( 27مجلة إبن الھيثم للعلوم الصرفة و التطبيقية المجلد Ibn Al-Haitham Jour. for Pure & Appl. Sci. Vol. 27 (3) 2014 Table No.(1a):The result of suggested method for n =2, 3, 4& method in[3] for example 3 0.990235879509409 0.9902358748522030.990235914910346 0b 0.050589305215627 0.0505893414767570.0505890348752984 1b 13P 11P 9P Numerical solution (y) ix 1 1 1 1 0 .0.999966681360863 ..999966681419201 0.999966680471501 0.99996668000000 0.1 0.999734338980186 0.999734340836996 0.999734323342549 0.999734330000000 0.2 0.999112195858900 0.999112205091858 0.999112136946651 0.999112190000000 0.3 0.997939333529043 0.997939352185325 0.997939228062444 0.997939330000000 0.4 0.996126225634366 0.996126243561995 0.996126122885822 0.996126220000000 0.5 0.993720978536793 0.993720981359821 0.993720947623996 0.993720970000000 0.6 0.991004633063186 0.991004619559273 0.991004695132483 0.9910046300000000 0.7 0.988619280474356 0.988619263799973 0.988619382084590 0.988619280000000 0.8 0.987731920033536 0.987731909996439 0.987731993184585 0.987723192000000 0.9 0.990235879509409 0.990235874852203 0.990235914910346 0.990235880000000 1 Table No.(1b):Comparison between suggested method with method in[3] for example 3 |13P-Error |y |11P-Error |y |9P-Error |y ix 0 0 0 0 1.360863421950850e-009 1.419200645891300e-009 4.715006163279999e-010 0.1 8.980186372298251e-009 1.083699618931180e-008 6.65745103487000e-009 0.2 .5.858900142818871e-009 1.509185754144940e-008 5.305334893623799e-008 0.3 3.529043057781680e-009 2.218532491049530e-008 1.019375558408700e-007 0.4 5.634365973783420e-009 2.356199502173690e-008 9.711417825464299e-008 0.5 8.536792939217721e-009 1.135982052868910e-008 2.237600371657300e-008 0.6 3.063186371221600e-009 1.044072739375680e-008 6.513248329564900e-008 0.7 4.743559989250900e-010 1.620002665347900e-008 1.020845900034930e-007 0.8 3.353606281564000e-011 1.000356053548760e-008 7.318458528082800e-008 0.9 4.905914563480000e-010 5.147797188520500e-009 3.491034628044800e-008 1 2.437493949383256e-016 2.021641245529962e-015 4.442053594825927-014 SSE 520 | Mathematics 2014) عام 3العدد ( 27مجلة إبن الھيثم للعلوم الصرفة و التطبيقية المجلد Ibn Al-Haitham Jour. for Pure & Appl. Sci. Vol. 27 (3) 2014 Table No.(2): Comparison of numerical results with the Maple11&DTM solution for different values of n n = 10 n = 20 n = 30 n = 35 Maple11 Solution x 0.99996668 0.99973433 0.99911219 0.99793933 0.99612622 0.99372096 0.99100452 0.98861874 0.98772971 0.99022826 0.99996668 0.99973433 0.99911219 0.99793933 0.99612622 0.99372097 0.99100463 0.98861927 0.98773191 0.99023588 0.99996668 0.99973433 0.99911219 0.99793933 0.99612622 0.99372097 0.99100463 0.98861927 0.98773191 0.99023587 0.99996668 0.99973433 0.99911219 0.99793933 0.99612622 0.99372097 0.99100463 0.98861927 0.98773191 0.99023587 0.99996668 0.99973433 0.99911219 0.99793933 0.99612622 0.99372097 0.99100463 0.98861928 0.98773192 0.99023588 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Table No. (3) : Max. error of example 3 0.990235879509409 0b 0.050589305215627 1b '(x) |13(x), P 13f(x, P-''13Error | P ''13P ix 9.826181241272956e-009 0.019954693976617- 0.1 9.851733395474671e-008 0.039206687602793- 0.2 1.564345449579437e-007 0.055639331969225- 0.3 6.350270132559042e-009 0.065139098987678 - 0.4 e-0071.683960224221478 0.061118434331738- 0.5 1.159391418267258e-007 0.034078685565394- 0.6 4.083249415301991e-008 0.028767057993287 0.7 6.01923548237699e-008 0.143788913036560 0.8 7.582132859063062e-009 0.331125863858524 0.9 4.761522773288895e-012 0.614718461206487 1 e-0071.683960224221478 Max. Error Table No. ( 4) : The log of max. error of the suggested solution of example Log of max error Max error Logn 2n+1P -3.408460003282497 3.904271375693158e-004 0.301029995663981 5P -4.294089562455153 5.080546580960021e-005 0.477121254719662 7P -5.435358725358707 3.669790521687588e-006 0.602059991327962 9P 521 | Mathematics 2014) عام 3العدد ( 27مجلة إبن الھيثم للعلوم الصرفة و التطبيقية المجلد Ibn Al-Haitham Jour. for Pure & Appl. Sci. Vol. 27 (3) 2014 for example 3 13 numerical solution in[3] and P gure No.( 1): Comparison between theFi Figure No.( 2) : The best fit line for the order of convergence 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.985 0.99 0.995 1 x y the soulution at n=6 Numerical solution p13 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 -5.5 -5 -4.5 -4 -3.5 -3 The best fit line for the order of convergent log n lo g m a x e rr e r 522 | Mathematics 2014) عام 3العدد ( 27مجلة إبن الھيثم للعلوم الصرفة و التطبيقية المجلد Ibn Al-Haitham Jour. for Pure & Appl. Sci. Vol. 27 (3) 2014 خطيةالر كفاءة التقـنيـة شبـه التحـليليـة في حـل مسائل القيم االبتدائية غي الشاذة لمى ناجي محمد توفيق حسين ريم وليد جامعة بغداد /)ابن الھيثم للعلوم الصرفة ( كلية التربية /قسم الرياضيات 2012كانون االول 9، قبل في : 2012ايلول 23استلم في : الخالصة للمعــادالت التفاضــلية االعتياديــة اقتــراح تقنيــة شــبه التحليليــة لحــل مســائل القــيم االبتدائيــة الشــاذة الهــدف مــن هــذا البحــث النقطتــين أوجــدنا كفــاءة الطريقــة يباســتخدام االنــدراج التماســي ذمتعــددة حــدود بوصــفها وبأنواع مختلفة للحصول على الحل من خالل المقارنــات مــع الحــل المضــبوط و حلــول تقريبيــة أخــرى ألصــناف واســعة متجانســة وغيــر متجانســة ودقتها المقترحة خطية و غير خطية لمسائل القيم االبتدائية الشاذة . العالقـــة ذييـــار الشـــبكة المناســـبة أيضـــا ناقشـــنا بعـــض الصـــفات العدديـــة اســـتخدم تخمـــين جديـــد و كفـــوء للخطـــأ الـــرئيس الخت لمســائل القــيم االبتدائيــة الشــاذة . كــذلك عــدد مــن األمثلــة نوقشــت )codeو لزيــادة كفــاءة الشــفرة (ومناقشــتهالتطبيــق المقــاييس لتقارب من جهة أخرى.لتوضيح الدقة و الكفاءة وسهولة أداء الطريقة المقترحة من جهة ولتعزيز و تأكيد رتبة ا معادالت تفاضلية اعتيادية , مسائل القيم االبتدائية الشاذةالكلمات المفتاحية :