Abstract
In this paper, we analyse Mickens’ type non-standard finite difference schemes (NSFD) and establish their convergence. We then apply these schemes on Lane Emden equations. The numerical results thus obtained are compared with existing analytical solutions or with solutions computed with standard finite difference (FD) schemes. NSFD and FD solutions and their errors have also been compared graphically and observed that the errors in NSFD tends to zero as step size tends to zero. The result shows that the NSFD behave qualitatively in the same way as the original equations. NSFD approximate solution near singular point efficiently where FD fails to do so (Buckmire in Numer Methods Partial Differ Equ 19:380–398, 2003).
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Chambre [6] discusses that in the theory of thermal explosion critical condition for inflammability is achieved when we have
amount of heat developed by the chemical reaction = amount lost to the surroundings.
The mathematical formulation for the thermal balance between the heat generated by the chemical reaction and that conducted away is described as
where T is temperature of the gas, Q the heat of reaction, \(\lambda \) the thermal conductivity, W the reaction velocity, and \(\nabla ^2\) is Laplacian. Author assumes that the reaction is monomolecular and that its velocity is governed by Arrhenius law, so that
where c is the concentration of the reactant, a the frequency factor, and E the energy of activation of the reaction. Hence
After some approximations and circular symmetry Chambre arrived at the following Poisson Boltzmann equation
where \(\delta \) is appropriate constant, k is the geometry of the vessel:
-
(i)
\(k=0\) for infinite plane-parallel vessel,
-
(ii)
\(k=1\) for cylindrical vessel where length is much greater than radius,
-
(iii)
\(k=2\) for spherical vessel.
Chambre imposed boundary conditions as
In this article, motivated by Chambre’s [6] work we propose to use Micken’s type NSFD schemes to solve the following second order non-linear singular boundary value problems of the form:
where c, m, n and \( \beta \) are real constants and \(c\ge 1\). We assume that g(t, w) is continuous and Lipschitz in \(D = \left\{ (t,w) \in [0,1] \times \mathbb {R}\right\} \). Various real life problems are governed by non-linear second order singular boundary value problems (SBVP) [3, 6, 7, 10, 11, 13, 14]. For both exact [9, 15, 21,22,23] and numerical results [1, 2, 8, 12, 19, 25, 26, 28] a huge literature is available in these papers as well as in their references.
Numerical instabilities are found due to the use of FD methods which have been shown in [16]. Therefore NSFD were proposed to remove these instabilities. One way to characterize these instabilities is to say that they are extra solutions that occur in FD and that do not correspond to any solution of ordinary differential equations (ODE). Also in FD, different step sizes exhibit different behavioral patterns which creates a problem in their selection. It is observed that the central difference scheme allows for the existence of chaotic orbits for all positive time steps for the logistic differential equation [27, 29]. Also for decay equation, Euler central scheme has numerical instabilities for all step sizes and therefore it is unstable.
The idea of solving these type of problems through NSFD schemes originates from the works of Mickens [16] and Ron Buckmire [4, 5]. Very recently in [18] authors have further studied importance of NSFD schemes. They also pointed out that the use of the central difference scheme forces all the fixed-points to become unstable.
SBVPs are solved more accurately and efficiently through NSFD than FD schemes. The NSFD scheme helps in approximating the solution easily near the singularity where the FD schemes generally fail (see [4]).
The paper is organised as follows. In second section, details of the NSFD methods have been given which are used to solve the singular, non-linear ODE problems. We have mentioned three methods here. First is from Mickens [16], second and third are from Buckmire [4, 5].
In third section we have shown the convergence of the three schemes proposed in [4, 5, 16].
In fourth section, we have considered 9 test problems and are discretized by using the NSFD schemes defined in Sect. 2. Then the system of non-linear equations thus obtained are solved by using Newton Raphson method. Numerical solutions and their absolute errors produced by NSFD schemes are compared with the analytical solutions and also with the standard FD schemes for step size \(h=0.0125\). Three kinds of graphs for each problem have been plotted. First kind of graphs show that as we decrease the step size, the approximate solution tends towards the analytical solution. In the second kind of graphs, the error between the analytical solution and the numerical solution obtained by the NSFD scheme and the FD scheme are drawn, which shows that the accuracy of NSFD scheme is better than the standard FD scheme. In the third kind of graphs, the absolute errors between the exact and NSFD solutions are given which illustrate that as we decrease our step size absolute errors decrease.
The article ends with the conclusions.
2 Non-standard finite difference methods (NSFD)
Here we will construct qualitatively stable methods for some nonlinear, singular second order ode problems. We consider ODE of the form defined by Eq. (5) where g(t, w) is non-zero real valued functions of t and w.
2.1 Method 1
By applying the Micken’s rules [16, 17] of NSFD schemes,
where \( \varphi (h)\rightarrow h^2+O(h^4)\) as \(h\rightarrow 0\), and
where \(\chi (h)\) and \(\varphi (h)\) are, respectively called the numerator and denominator functions; they satisfy the requirements \( \chi (h)\rightarrow 1+o(h^2)\), \(\varphi (h)\rightarrow h+o(h^2)\) as \(h \rightarrow 0\).
Here, we have \(h = \delta t,~ t \rightarrow t_{i} = ih,~ w(t)\rightarrow w_{i}\), where i is an integer. It should be noted that, in general, the numerator function can always be selected to be \(\chi (h) = 1\).
The linear and non-linear terms of w(t) should be in general be modelled non-locally on the computational grid in many different ways. For example
2.2 Method 2
We apply Mickens-type NSFD schemes for nonlinear singular boundary value problems in cylindrical coordinates [4, 5]. We write
and approximate \(tw'\) from [4] as
So, by using Eq. (11) in (10), we get
2.3 Method 3
We apply Mickens-type NSFD schemes [4] for non-linear singular boundary value problems in spherical coordinates. We write
and from [4] we approximate \(t^2 \frac{dw}{dt}\) as follows:
So, using Eq. (14) in (13) , we get
3 Convergence of difference scheme
Here we establish the convergence of proposed three NSFD schemes for the nonlinear SBVP defined by Eq. (5) subject to boundary condition \(w'(0)=0,~w(1)=\beta \).
3.1 Convergence of NSFD for Method 1
Consider the NSFD scheme of (5) as defined in Sect. 2.1,
The boundary conditions become
Here we take \(\varphi =4 \sin ^{2}\big (\frac{h}{2}\big )\) and \( \chi =\sin {(h)}\). So, our difference scheme takes the following form:
The exact solution w(t) of (18) satisfies
where \( T_{i}\) is the truncation error given below:
Subtracting (19) from (18) and applying Mean Value Theorem, and substituting \(\varepsilon _{i}=w_{i}-w(t_{i})\), we get the error equation
For \(i=1\), using boundary condition \(w'_0=0=w_1-w_0\), we get
For \(i=N-1\) and using the boundary condition \(w_N=\beta \), we get
where \(\varepsilon _{N}=w_N-w(t_N)=\beta -\beta =0\).
In the matrix notation, we write (21), (22) and (23) as
where \(M=D+Q\), \(E=[\varepsilon _{1},\varepsilon _{2},\ldots ,\varepsilon _{N-1}]^T\), \( T=[T_{1},T_{1},\ldots , T_{N-1}]^T\),
and
We now show that the matrix \(M=D+Q\) is an irreducible, monotone matrix such that \( M\ge D \) and \(Q\ge 0 \).
Since D is tridiagonal matrix where \( a_{i,j}=0\) for \( |i-j|>1\) and \(a_{i,i-1}\ne 0\) for \(i=2(1)N-1\) and \( a_{i,i+1}\ne 0\) for \(i=1(1)N-2\). So, D is irreducible. Also D is diagonally dominant, i.e.,
So, D is irreducibly diagonally dominant matrix and it has also non-positive off diagonal elements. So, D is monotone. Since \( g_{w_{i}}>0, ~ i=1(1)N-1\), so, we have \( Q \ge 0\)
Also, D is monotone. Now, Q is diagonal matrix with positive entries. Hence \(M=D+Q\) is also irreducibly diagonal dominant and have non-positive off-diagonal elements. Therefore M is also monotone and \(M \ge D\). So, we have \( 0<M^{-1}\le D^{-1}. \)
From (24), we have
and
Let \(W=(1,1,\ldots , 1)^{T}\) and \(P=(P_{1},P_{2},\ldots , P_{N-1})^T = DW\) denote the vector of row sums of D.
Also, let \( Z=(Z_{1},Z_{2},\ldots , Z_{N-1})^T\) where \(Z_{j}=2-\frac{(t_{j}+1)^2}{2}\), and \(S=(S_{1},S_{2},\ldots , S_{N-1})^T = DZ\). We next obtain bounds for \(D^{-1}=(d_{ij}^{-1}).\) Now
hence
Now \(S_{k}=D_{k-1}Z_{k-1}+D_{k}Z_{k}+D_{k+1}Z_{k+1},~k=2(1)N-2\),
hence
Since \( D^{-1} S =Z \) and \( Z_{i}< \frac{3}{2} \) for \(i= 1(1)N-2\), then \( d^{-1}_{i,1} S_{1} \le Z_{i} <(3/2), ~ i=1(1)N-2\),
where \(\max _k\left[ \frac{c}{k} \frac{\tan (h/2)}{h/2}\right] =\frac{c}{2}\frac{\tan (h/2)}{h/2}\).
Now since \(P_{N-1}=\left( 1+ \frac{2 c \tan (h/2)}{(N-1)h} \right) \) then with the help of \(D^{-1} P = W\), we obtain
and thus from Eqs. (20), (25), (29)–(31), we get
For limiting case when h is sufficiently small then \( \tan \frac{h}{2}= \frac{h}{2}\) and since \(i \ge 1, N-1\ge 1\) so, \( \frac{1}{i}\le 1\) and \(\frac{1}{N-1} \le 1\),
where \( \lambda = \frac{3(3c+2)}{4(c+1)}\displaystyle \sup _{[0,1]}w''(t)\).
Now in view of \(\Vert E \Vert _{\infty } \leqslant \Vert D^{-1} T \Vert _{\infty }\), we have
Similarly we establish the convergence for NSFD schemes described in Sects. 2.2 and 2.3.
3.2 Convergence of NSFD for Method 2
Consider the NSFD for (5) subject to boundary condition \(w'(0)=0,~w(1)=\beta \) given in Sect. (2.2) with \(c=1\). Here we will we get, the error equation as given below
where
For \(i=1\), using boundary condition \(w'(0)=0\), we get \(w_1-w_0=0\) and hence
For \(i=N-1\), using the boundary condition \(w(1)=\beta \) and that \(\varepsilon _N=w_N-w(t_n)=\beta -\beta =0\), we get
In the matrix notation, we write (33), (34), (35) and (36) as
where \(M=D+Q\), \(E=[\varepsilon _{1},\varepsilon _{2},\ldots ,\varepsilon _{N-1}]^T\), \( T=[T_{1},T_{1},\ldots , T_{N-1}]^T,\)
Similarly to convergence analysis of method 2.1 we can show that M is an irreducible and monotone matrix. Also \(\Vert E\Vert \le \Vert M^{-1}\Vert \Vert T \Vert \le \Vert D^{-1}\Vert \Vert T \Vert \). Here we get matrix P as
Also, let \(Z=(Z_{1},Z_{2},\ldots , Z_{N-1})^T\), where \(Z_{j}=2-\frac{(t_{j}+1)^2}{2}\). And \(S=(S_{1},S_{2},\ldots , S_{N-1})^T = DZ\). In the same way, here we get
We next obtain bounds for \( D^{-1}=(d_{ij}^{-1}) \) in the same way as done above and put them in the equation given below,
where \(\max \left[ \frac{1}{\ln \left( (k+1)/k\right) }-\frac{1}{\ln \left( k/k-1\right) }\right] <2\) and \( \lambda =\left[ \frac{\ln 2+2}{8}\right] \displaystyle \sup _{[0,1]}\Big [t^2 w^{(iv)}(t)+2tw'''(t)-w''(t)+\frac{w'}{t}\Big ]\). Now in view of \(\Vert E \Vert _{\infty } \leqslant \Vert D^{-1} T \Vert _{\infty }\), we have
3.3 Convergence of NSFD for Method 3
Consider the NSFD scheme described in Sect. 2.2 for (5) subject to \(w'(0)=0\) and \(w(1)=\beta \) with \(c=2\). Here we will we get, the error equation
where
For \(i=1\), using boundary condition \(w'(0)=0\), we get
For \(i=N-1\), using the boundary condition \(w(1)=\beta \), we get
In the matrix notation, we write (41), (42), (43) and (44) as
where \(M=D+Q\), \(E=[\varepsilon _{1},\varepsilon _{2},\ldots ,\varepsilon _{N-1}]^T\), \( T=[T_{1},T_{1},\ldots , T_{N-1}]^T\),
Similarly we can show that M is an irreducible and monotone matrix. Also \(\Vert E\Vert \le \Vert M^{-1}\Vert \Vert T \Vert \le \Vert D^{-1}\Vert \Vert T \Vert \). Here we get matrix P as
Also, let \( Z=(Z_{1},Z_{2},\ldots , Z_{N-1})^T\) where \(Z_{j}=2-\frac{(t_{j}+1)^2}{2}\). And \(S=(S_{1},S_{2},\ldots , S_{N-1})^T = DZ\). Here we get
We next obtain bounds for \( D^{-1}=(d_{ij}^{-1}) \) and put them in the equation given below,
where \(\lambda =\frac{3}{16}\displaystyle \sup _{[0,1]}\left[ t^2w^{iv}(t)+4tw'''(t)\right] .\) Now in view of \(\Vert E \Vert _{\infty } \leqslant \Vert D^{-1} T \Vert _{\infty }\), we have
4 Non-linear singular BVPs: Lane Emden type equations
4.1 Problem 1
Polytropes in hydrostatic equilibrium as simple models of a star: Polytropes can be described by the equation of state in which the pressure P is given as a power-law in density \(\rho \)
where \(\kappa \) and \(\gamma \) are constants.
The equation of hydrostatic equilibrium is given by
By solving it for M(r) and comparing it to what we obtain from considering the spherical shell in hydrostatic equilibrium \(dM = 4\pi r^2\rho dr\) we get
Since \(P=\kappa \rho ^\gamma \) we get
Substitute \(\rho \equiv \lambda \theta ^n\), \(\gamma \equiv \frac{n+1}{n}\) and then \(\xi \equiv \frac{r}{\alpha }\), \(\alpha \equiv \sqrt{\frac{n+1}{4\pi G}\kappa \lambda ^{(1-n)/n}}\) we obtain the so called Lane Emden for polytropes in hydrostatic equilibrium:
The above SBVP was derived by Chandrasekhar [7].
In (5) we take \(c=2\) and \(g(t,w) = w^\beta \) and \(\beta =5\) (or \(n=5\) in the above equation) and write the following SBVP:
with boundary condition
Analytical solution of the above problem is:
4.1.1 Mathematical illustrations
Applying NSFD scheme as described above in Sect. 2.1 to the above problem, we get
where \(\varphi =4 \sin ^{2}\big (\frac{h}{2}\big )\) and \( \chi =\sin {(h)}\).
The nonlinear term \(w_{i}^5\) is approximated non-locally by the term \(w_{i-1}^5 \). Here \(h=t_i-t_{i-1}\) is the step size. Now, we discretize the inner boundary condition \(w'(0)=0\) by using backward Euler method,
The outer boundary condition can be dicretized as \(w_N=w(t_N)=\frac{\sqrt{3}}{2}\).
4.1.2 Numerical illustrations
Table 1 Analytic and NSFD solutions of Problem 1
t | NSFD (W1) | NSFD (W2) | NSFD (W3) | NSFD (W4) | Analytic | FD |
---|---|---|---|---|---|---|
0 | 0.994312521 | 0.997301048 | 0.999051 | 0.999923 | 1 | 1.000163654 |
0.1 | 0.994312521 | 0.996682835 | 0.998901 | 0.9983425 | 0.998337488 | 0.998525527 |
0.2 | 0.991072907 | 0.992740998 | 0.993692 | 0.99339242 | 0.993399268 | 0.993582356 |
0.3 | 0.984593678 | 0.985489555 | 0.985382 | 0.98532142 | 0.985329278 | 0.985504418 |
0.4 | 0.974969408 | 0.975154355 | 0.974381 | 0.9743613 | 0.974354704 | 0.974519309 |
0.5 | 0.962384523 | 0.961498659 | 0.960665 | 0.9607711 | 0.960768923 | 0.960920933 |
0.6 | 0.947102766 | 0.945639849 | 0.9451061 | 0.94492053 | 0.944911183 | 0.945049084 |
0.7 | 0.929449782 | 0.927935575 | 0.9271935 | 0.92715193 | 0.927145541 | 0.927268353 |
0.8 | 0.909792535 | 0.908464262 | 0.9079459 | 0.90785121 | 0.907841299 | 0.9079486 |
0.9 | 0.888517674 | 0.888005817 | 0.887505486 | 0.88736231 | 0.887356 | 0.887448355 |
Fig. 1 Comparison of NSFD solutions as \(h\rightarrow 0\) with analytical solution for Problem 1
Table 2 Absolute Errors in solution of Problem 1
t | E1 | E2 | E3 | E4 | E in FD |
---|---|---|---|---|---|
0 | 0.005687479 | 0.002698952 | 0.000949 | 7.7E\(-\)05 | 0.000163654 |
0.1 | 0.004024968 | 0.001654654 | 0.000563512 | 5.01154E\(-\)06 | 0.000188039 |
0.2 | 0.002326361 | 0.00065827 | 0.000292732 | 6.84781E\(-\)06 | 0.000183088 |
0.3 | 0.0007356 | 0.000160277 | 5.27218E\(-\)05 | 7.85818E\(-\)06 | 0.000175139 |
0.4 | 0.000614704 | 0.000799652 | 2.62963E\(-\)05 | 6.59628E\(-\)06 | 0.000164605 |
0.5 | 0.0016156 | 0.000729737 | 0.000103923 | 2.17713E\(-\)06 | 0.00015201 |
0.6 | 0.002191583 | 0.000728666 | 0.000194917 | 9.34743E\(-\)06 | 0.000137901 |
0.7 | 0.002304242 | 0.000790034 | 4.79591E\(-\)05 | 6.38911E\(-\)06 | 0.000122812 |
0.8 | 0.001951236 | 0.000622963 | 0.000104601 | 9.91092E\(-\)06 | 0.000107301 |
0.9 | 0.001161165 | 0.000649307 | 0.000148976 | 5.80049E\(-\)06 | 9.18457E\(-\)05 |
Fig. 2 Numerical errors in NSFD and FD for \(h=0.0125\) for Problem 1
Fig. 3 Error in NSFD solutions as \(h\rightarrow 0\) for Problem 1
4.2 Problem 2
Thermal explosion in cylindrical vessel: In the thermal explosion, the following two point nonlinear SBVP \((c = 1)\) and \(g(t,w) = e^{w(t)})\) arises, which has been derived by Chamber [6],
with boundary condition
Analytical solution of the above problem is:
where \( B=3-2\sqrt{2}.\)
4.2.1 Mathematical illustrations
Above problem can be written as
After discretizing it, we get
Applying NSFD scheme 2.2 to the above problem, we get
that is
Here \(h=t_i-t_{i-1}\) is the step size.
Now using the non-standard finite difference method, we discretize the inner boundary condition since \(w'(0)=0\), so \(tw'\) must be zero at its first grid point, i.e.,
The outer boundary condition can be dicretized as \(w_N=w(t_N)=0\).
4.2.2 Numerical illustrations
Table 3 Analytic and NSFD solutions of Problem 2
t | NSFD (W1) | NSFD (W2) | NSFD (W3) | NSFD (W4) | Analytic | FD |
---|---|---|---|---|---|---|
0 | 0.310600377 | 0.315837 | 0.316213 | 0.316562 | 0.316694368 | 0.316264073 |
0.1 | 0.310600377 | 0.312594201 | 0.313097553 | 0.313223749 | 0.313266063 | 0.313054366 |
0.2 | 0.301144158 | 0.302545108 | 0.30289766 | 0.302985969 | 0.303015636 | 0.302861647 |
0.3 | 0.284653686 | 0.285697177 | 0.285959617 | 0.286025344 | 0.286047478 | 0.285928733 |
0.4 | 0.261481022 | 0.26226734 | 0.262465086 | 0.26251461 | 0.26253134 | 0.262438741 |
0.5 | 0.231913687 | 0.232500053 | 0.232647529 | 0.232684465 | 0.232696997 | 0.232625621 |
0.6 | 0.196259942 | 0.196684376 | 0.196791145 | 0.196817886 | 0.196827019 | 0.196773669 |
0.7 | 0.154860498 | 0.1551507 | 0.155223717 | 0.155242006 | 0.15524832 | 0.155210861 |
0.8 | 0.108085861 | 0.108263218 | 0.108307853 | 0.108319034 | 0.108322977 | 0.108299491 |
0.9 | 0.056329533 | 0.056411183 | 0.056431736 | 0.056436885 | 0.056438816 | 0.056427608 |
Fig. 4 Comparison of NSFD solutions as \(h\rightarrow 0\) with analytical solution for Problem 2
Table 4 Absolute Errors in solution of Problem 2
t | E1 | E2 | E3 | E4 | E in FD |
---|---|---|---|---|---|
0 | 0.006093991 | 0.000857368 | 0.000481368 | 0.000132368 | 0.000430295 |
0.1 | 0.002665687 | 0.000671862 | 0.000168511 | 4.23142E\(-\)05 | 0.000211697 |
0.2 | 0.001871478 | 0.000470528 | 0.000117976 | 2.9667E\(-\)05 | 0.000153989 |
0.3 | 0.001393793 | 0.000350302 | 8.78617E\(-\)05 | 2.21345E\(-\)05 | 0.000118746 |
0.4 | 0.001050319 | 0.000264 | 6.62547E\(-\)05 | 1.67304E\(-\)05 | 9.25997E\(-\)05 |
0.5 | 0.00078331 | 0.000196944 | 4.9468E\(-\)05 | 1.25321E\(-\)05 | 7.1376E\(-\)05 |
0.6 | 0.000567077 | 0.000142642 | 3.58743E\(-\)05 | 9.13241E\(-\)06 | 5.33494E\(-\)05 |
0.7 | 0.000387822 | 9.76201E\(-\)05 | 2.4603E\(-\)05 | 6.31347E\(-\)06 | 3.74591E\(-\)05 |
0.8 | 0.000237115 | 5.97584E\(-\)05 | 1.51237E\(-\)05 | 3.94263E\(-\)06 | 2.34854E\(-\)05 |
0.9 | 0.000109283 | 2.76332E\(-\)05 | 7.07977E\(-\)06 | 1.93078E\(-\)06 | 1.12079E\(-\)05 |
Fig. 5 Numerical errors in NSFD and FD for \(h=0.0125\) for Problem 2
Fig. 6 Error in NSFD solutions as \(h\rightarrow 0\) for Problem 2
4.3 Problem 3
We consider the nonlinear differential equation by taking \(b_{0}={0.5}\) in example 1 of [20]
with boundary condition
Analytical solution of the above problem is:
4.3.1 Mathematical illustrations
Above problem can be written as
Dividing by \( \sqrt{t} \) on both sides, we get
After discretizing it, we get
Applying Nonstandard finite difference scheme to the above problem, we get
By modelling non-linear terms nonlocally on the computational grid, we get
Here \(h=t_i-t_{i-1}\) is the step size.
Now using the NSFD method, we discretize the inner boundary condition since \(w'(0)=0\), so \(tw'\) must be zero at its first grid point i.e.,
The outer boundary condition can be dicretized as \( w(1)= w_N = \ln (1/5)\).
4.3.2 Numerical illustrations
Table 5 Analytic and NSFD solutions of Problem 3
t | NSFD (W1) | NSFD (W2) | NSFD (W3) | NSFD (W4) | Analytic | FD |
---|---|---|---|---|---|---|
0 | \(-\) 1.387531417 | \(-\) 1.386085024 | \(-\) 1.386109839 | \(-\) 1.386297508 | \(-\) 1.386296861 | \(-\) 1.386299777 |
0.1 | \(-\) 1.387531417 | \(-\) 1.386085024 | \(-\) 1.386116345 | \(-\) 1.386295508 | \(-\) 1.386296861 | \(-\) 1.3863022 |
0.2 | \(-\) 1.387609369 | \(-\) 1.386165898 | \(-\) 1.386219228 | \(-\) 1.386375308 | \(-\) 1.386374358 | \(-\) 1.38637915 |
0.3 | \(-\) 1.388156389 | \(-\) 1.386714916 | \(-\) 1.386591871 | \(-\) 1.386903172 | \(-\) 1.386901677 | \(-\) 1.386904987 |
0.4 | \(-\) 1.390189617 | \(-\) 1.388731807 | \(-\) 1.38898212 | \(-\) 1.388854361 | \(-\) 1.38885109 | \(-\) 1.388859532 |
0.5 | \(-\) 1.395623342 | \(-\) 1.394105507 | \(-\) 1.393893568 | \(-\) 1.394074361 | \(-\) 1.394076502 | \(-\) 1.394072317 |
0.6 | \(-\) 1.407472669 | \(-\) 1.405833473 | \(-\) 1.405561245 | \(-\) 1.405548351 | \(-\) 1.405547818 | \(-\) 1.40554287 |
0.7 | \(-\) 1.42988375 | \(-\) 1.428088553 | \(-\) 1.427617241 | \(-\) 1.427452891 | \(-\) 1.427453099 | \(-\) 1.427450767 |
0.8 | \(-\) 1.467836303 | \(-\) 1.465988428 | \(-\) 1.465392298 | \(-\) 1.465032998 | \(-\) 1.465031602 | \(-\) 1.465030002 |
0.9 | \(-\) 1.526380196 | \(-\) 1.524923482 | \(-\) 1.524386807 | \(-\) 1.523985998 | \(-\) 1.523986772 | \(-\) 1.523985495 |
Fig. 7 Comparison of NSFD solutions as \(h\rightarrow 0\) with analytical solution for Problem 3
Table 6 Absolute Errors in solution of Problem 3
t | E1 | E2 | E3 | E4 | E in FD |
---|---|---|---|---|---|
0 | 0.001234556 | 0.000211837 | 0.000187022 | 6.46422E\(-\)07 | 2.91592E\(-\)06 |
0.1 | 0.001234556 | 0.000211837 | 0.000180516 | 1.35358E\(-\)06 | 5.33841E\(-\)06 |
0.2 | 0.001235011 | 0.00020846 | 0.00015513 | 9.49846E\(-\)07 | 4.79216E\(-\)06 |
0.3 | 0.001254712 | 0.000186761 | 0.000309806 | 1.49532E\(-\)06 | 3.31003E\(-\)06 |
0.4 | 0.001338527 | 0.000119282 | 0.00013103 | 3.2711E\(-\)06 | 8.44202E\(-\)06 |
0.5 | 0.001546841 | 2.90059E\(-\)05 | 0.000182933 | 2.14013E\(-\)06 | 4.1844E\(-\)06 |
0.6 | 0.001924851 | 0.000285655 | 1.34271E\(-\)05 | 5.32648E\(-\)07 | 4.94811E\(-\)06 |
0.7 | 0.002430651 | 0.000635454 | 0.000164142 | 2.07632E\(-\)07 | 2.33182E\(-\)06 |
0.8 | 0.002804702 | 0.000956826 | 0.000360696 | 1.39636E\(-\)06 | 1.59996E\(-\)06 |
0.9 | 0.002393424 | 0.00093671 | 0.000400035 | 7.74396E\(-\)07 | 1.27675E\(-\)06 |
Fig. 8 Numerical errors in NSFD and FD for \(h=0.0125\) for Problem 3
Fig. 9 Error in NSFD solutions as \(h\rightarrow 0\) for Problem 3
4.4 Problem 4
We consider the following differential equation by taking \(b_{0}={0.75}\) in example 1 of [20]
with boundary condition
Analytical solution of the above problem is:
4.4.1 Mathematical illustrations
Above problem can be written as
Dividing by \( {t^{0.75}} \) on both sides, we get
After discretizing it, we get
Applying NSFD scheme to the above problem, we get
By modelling non-linear terms nonlocally on the computational grid, we get
Here \(h=t_{i}-t_{i-1}\) is the step size. Now using the non-standard finite difference method, we discretize the inner boundary condition since \(w'(0)=0\), so \(tw'\) must be zero at its first grid point, i.e.,
The outer boundary condition can be dicretized as \( w(1)= w_N =\ln (1/5)\).
4.4.2 Numerical illustrations
Table 7 Analytic and NSFD solutions of Problem 4
t | NSFD (W1) | NSFD (W2) | NSFD (W3) | NSFD (W4) | Analytic | FD |
---|---|---|---|---|---|---|
0 | \(-\) 1.387871361 | \(-\) 1.386471195 | \(-\) 1.386223505 | \(-\) 1.386270664 | \(-\) 1.386294361 | \(-\) 1.386347681 |
0.1 | \(-\) 1.387871361 | \(-\) 1.38647291 | \(-\) 1.386225859 | \(-\) 1.38627806 | \(-\) 1.386296861 | \(-\) 1.386350099 |
0.2 | \(-\) 1.387926235 | \(-\) 1.386546512 | \(-\) 1.386303354 | \(-\) 1.386362606 | \(-\) 1.386374358 | \(-\) 1.386427017 |
0.3 | \(-\) 1.388402406 | \(-\) 1.387069527 | \(-\) 1.386833822 | \(-\) 1.38687643 | \(-\) 1.386901677 | \(-\) 1.386952763 |
0.4 | \(-\) 1.390277732 | \(-\) 1.38902302 | \(-\) 1.38879553 | \(-\) 1.388863997 | \(-\) 1.38885109 | \(-\) 1.388899134 |
0.5 | \(-\) 1.395423103 | \(-\) 1.394274737 | \(-\) 1.394050644 | \(-\) 1.394039798 | \(-\) 1.394076502 | \(-\) 1.394119615 |
0.6 | \(-\) 1.40683325 | \(-\) 1.405809966 | \(-\) 1.405577086 | \(-\) 1.405535558 | \(-\) 1.405547818 | \(-\) 1.405583899 |
0.7 | \(-\) 1.428712182 | \(-\) 1.427819856 | \(-\) 1.427563932 | \(-\) 1.427455567 | \(-\) 1.427453099 | \(-\) 1.427480254 |
0.8 | \(-\) 1.466259759 | \(-\) 1.465508457 | \(-\) 1.465229328 | \(-\) 1.465029171 | \(-\) 1.465031602 | \(-\) 1.465048605 |
0.9 | \(-\) 1.52497968 | \(-\) 1.524448449 | \(-\) 1.524204375 | \(-\) 1.523981537 | \(-\) 1.523986772 | \(-\) 1.523994056 |
Fig. 10 Comparison of NSFD solutions as \(h\rightarrow 0\) with analytical solution for Problem 4
Table 8 Absolute Errors in solution of Problem 4
t | E1 | E2 | E3 | E4 | E in FD |
---|---|---|---|---|---|
0 | 0.001577 | 0.000176834 | 7.0856E\(-\)05 | 2.36971E\(-\)05 | 5.33203E\(-\)05 |
0.1 | 0.0015745 | 0.000176048 | 7.10024E\(-\)05 | 1.88012E\(-\)05 | 5.3238E\(-\)05 |
0.2 | 0.001551877 | 0.000172154 | 7.10044E\(-\)05 | 1.17524E\(-\)05 | 5.26587E\(-\)05 |
0.3 | 0.00150073 | 0.000167851 | 6.78545E\(-\)05 | 2.52463E\(-\)05 | 5.10868E\(-\)05 |
0.4 | 0.001426642 | 0.00017193 | 5.55599E\(-\)05 | 1.29074E\(-\)05 | 4.80438E\(-\)05 |
0.5 | 0.001346601 | 0.000198235 | 2.58579E\(-\)05 | 3.67037E\(-\)05 | 4.31139E\(-\)05 |
0.6 | 0.001285432 | 0.000262148 | 2.92676E\(-\)05 | 1.22596E\(-\)05 | 3.60806E\(-\)05 |
0.7 | 0.001259083 | 0.000366758 | 0.000110833 | 2.46779E\(-\)06 | 2.71548E\(-\)05 |
0.8 | 0.001228157 | 0.000476855 | 0.000197726 | 2.43089E\(-\)06 | 1.70028E\(-\)05 |
0.9 | 0.000992908 | 0.000461677 | 0.000217603 | 5.23519E\(-\)06 | 7.2841E\(-\)06 |
Fig. 11 Numerical errors in NSFD and FD for \(h=0.0125\) for Problem 4
Fig. 12 Error in NSFD solutions as \(h\rightarrow 0\) for Problem 4
4.5 Problem 5
Polytropes in hydrostatic equilibrium as simple models of a star: The following two point nonlinear SBVP \( (c = 2\) and \(g(t,w) = w^\beta )\), where \(\beta \) is a physical constant has been derived by Chandrasekhar [7]. Here we are taking the case when \(\beta = 1\).
with boundary condition
Analytical solution of the above problem is:
4.5.1 Mathematical illustrations
Applying NSFD scheme as described above in (2.3) to the above problem, we get
where \(h=t_{i}-t_{i-1}\) is the step size.
4.5.2 Numerical illustrations
Table 9 Analytic and NSFD solutions of Problem 5
t | NSFD (W1) | NSFD (W2) | NSFD (W3) | NSFD (W4) | Analytic | FD |
---|---|---|---|---|---|---|
0 | 0.998488547 | 0.999620798 | 0.999905215 | 0.999976866 | 1 | 1.000022634 |
0.1 | 0.998488547 | 0.998371272 | 0.998343447 | 0.998337049 | 0.998334166 | 0.998424728 |
0.2 | 0.993496105 | 0.993382532 | 0.993355631 | 0.993349458 | 0.993346654 | 0.993391908 |
0.3 | 0.98520865 | 0.985101206 | 0.985075828 | 0.985070032 | 0.985067356 | 0.985104589 |
0.4 | 0.973678043 | 0.973576917 | 0.973553637 | 0.97354835 | 0.973545856 | 0.973606775 |
0.5 | 0.958970255 | 0.958878585 | 0.958858067 | 0.958853347 | 0.958851077 | 0.95889549 |
0.6 | 0.941173644 | 0.941094038 | 0.941076834 | 0.941072777 | 0.941070789 | 0.941067611 |
0.7 | 0.92039222 | 0.920329286 | 0.920316063 | 0.920312601 | 0.920310982 | 0.920228369 |
0.8 | 0.896750362 | 0.896707841 | 0.896698854 | 0.896696322 | 0.896695114 | 0.89660596 |
0.9 | 0.870391135 | 0.870369842 | 0.870365247 | 0.870364026 | 0.870363233 | 0.8703153 |
Fig. 13 Comparison of NSFD solutions as \(h\rightarrow 0\) with analytical solution for Problem 5
Table 10 Absolute Errors in solution of Problem 5
t | E1 | E2 | E3 | E4 | E in FD |
---|---|---|---|---|---|
0 | 0.001511453 | 0.000379202 | 9.47851E\(-\)05 | 2.31338E\(-\)05 | 2.26343E\(-\)05 |
0.1 | 0.000154381 | 3.71054E\(-\)05 | 9.28045E\(-\)06 | 2.88218E\(-\)06 | 9.05617E\(-\)05 |
0.2 | 0.000149451 | 3.58783E\(-\)05 | 8.9771E\(-\)06 | 2.80437E\(-\)06 | 4.52539E\(-\)05 |
0.3 | 0.000141294 | 3.38504E\(-\)05 | 8.47228E\(-\)06 | 2.67639E\(-\)06 | 3.72335E\(-\)05 |
0.4 | 0.000132187 | 3.10616E\(-\)05 | 7.7817E\(-\)06 | 2.49382E\(-\)06 | 6.09195E\(-\)05 |
0.5 | 0.000119178 | 2.75077E\(-\)05 | 6.99014E\(-\)06 | 2.26967E\(-\)06 | 4.44131E\(-\)05 |
0.6 | 0.000102855 | 2.32495E\(-\)05 | 6.04526E\(-\)06 | 1.9877E\(-\)06 | 3.17842E\(-\)06 |
0.7 | 8.12387E\(-\)05 | 1.83045E\(-\)05 | 5.08137E\(-\)06 | 1.61939E\(-\)06 | 8.26132E\(-\)05 |
0.8 | 5.52479E\(-\)05 | 1.27275E\(-\)05 | 3.74032E\(-\)06 | 1.20881E\(-\)06 | 8.91539E\(-\)05 |
0.9 | 2.79018E\(-\)05 | 6.6092E\(-\)06 | 2.01397E\(-\)06 | 7.93069E\(-\)07 | 4.79333E\(-\)05 |
Fig. 14 Numerical errors in NSFD and FD for \(h=0.0125\) for Problem 5
Fig. 15 Error in NSFD solutions as \(h\rightarrow 0\) for Problem 5
4.6 Problem 6
Polytropes in hydrostatic equilibrium as simple models of a star: The following two point nonlinear SBVP \( (c = 2\) and \(g(t,w) = w^\beta )\), where \(\beta \) is a physical constant has been derived by Chandrasekhar [7]. Here we are taking the case when \(\beta = 0\).
with boundary condition
Analytical solution of the above problem is:
4.6.1 Mathematical illustrations
Applying NSFD scheme as described above in (2.3) to the above problem, we get
where \(h=t_{i+1}-t_{i}=t_{i}-t_{i-1}\) is the step size.
4.6.2 Numerical illustrations
Table 11 Analytic and NSFD solutions of Problem 6
t | NSFD (W1) | NSFD (W2) | NSFD (W3) | NSFD (W4) | Analytic | FD |
---|---|---|---|---|---|---|
0 | 0.331667827 | 0.332916864 | 0.333229227 | 0.333307441 | 0.3333333 | 0.333290917 |
0.1 | 0.331667827 | 0.331666864 | 0.331666725 | 0.331666817 | 0.331666667 | 0.331682038 |
0.2 | 0.326667827 | 0.326666861 | 0.326666728 | 0.326666818 | 0.326666667 | 0.326673276 |
0.3 | 0.318334493 | 0.318333524 | 0.318333396 | 0.318333488 | 0.318333333 | 0.318340846 |
0.4 | 0.306668512 | 0.306666872 | 0.30666674 | 0.306666821 | 0.306666667 | 0.306684483 |
0.5 | 0.291668923 | 0.291666869 | 0.291666837 | 0.291666827 | 0.291666667 | 0.291681745 |
0.6 | 0.273335863 | 0.273333536 | 0.273333627 | 0.273333488 | 0.273333333 | 0.273335831 |
0.7 | 0.251668757 | 0.251666843 | 0.251667238 | 0.25166677 | 0.251666667 | 0.251647285 |
0.8 | 0.226667826 | 0.226666788 | 0.226667282 | 0.226666716 | 0.226666667 | 0.226645107 |
0.9 | 0.198333769 | 0.198333402 | 0.198333731 | 0.198333359 | 0.198333333 | 0.19832179 |
Fig. 16 Comparison of NSFD solutions as \(h\rightarrow 0\) with analytical solution for Problem 6
Table 12 Absolute Errors in solution of Problem 6
t | E1 | E2 | E3 | E4 | E in FD |
---|---|---|---|---|---|
0 | 0.001665473 | 0.000416436 | 0.000104073 | 2.58586E\(-\)05 | 4.23825E\(-\)05 |
0.1 | 1.16001E\(-\)06 | 1.97084E\(-\)07 | 5.85398E\(-\)08 | 1.50308E\(-\)07 | 1.53714E\(-\)05 |
0.2 | 1.16001E\(-\)06 | 1.93959E\(-\)07 | 6.1144E\(-\)08 | 1.51783E\(-\)07 | 6.6093E\(-\)06 |
0.3 | 1.16001E\(-\)06 | 1.90834E\(-\)07 | 6.2446E\(-\)08 | 1.55014E\(-\)07 | 7.51309E\(-\)06 |
0.4 | 1.84501E\(-\)06 | 2.05417E\(-\)07 | 7.28625E\(-\)08 | 1.54112E\(-\)07 | 1.78162E\(-\)05 |
0.5 | 2.25602E\(-\)06 | 2.02084E\(-\)07 | 1.70518E\(-\)07 | 1.60297E\(-\)07 | 1.5078E\(-\)05 |
0.6 | 2.53002E\(-\)06 | 2.02639E\(-\)07 | 2.9389E\(-\)07 | 1.54315E\(-\)07 | 2.49728E\(-\)06 |
0.7 | 2.09002E\(-\)06 | 1.75953E\(-\)07 | 5.70999E\(-\)07 | 1.03296E\(-\)07 | 1.93814E\(-\)05 |
0.8 | 1.15918E\(-\)06 | 1.21563E\(-\)07 | 6.14931E\(-\)07 | 4.88593E\(-\)08 | 2.15595E\(-\)05 |
0.9 | 4.35189E\(-\)07 | 6.90742E\(-\)08 | 3.97581E\(-\)07 | 2.55243E\(-\)08 | 1.15411E\(-\)05 |
Fig. 17 Numerical errors in NSFD and FD for \(h=0.0125\) for Problem 6
Fig. 18 Error in NSFD solutions as \(h\rightarrow 0\) for Problem 6
4.7 Problem 7
We consider the following differential equation
with boundary condition
Analytical solution of the above problem is:
4.7.1 Mathematical illustrations
Above problem can be written as
After discretizing it, we get
Applying NSFD scheme (2.2) to the above problem, we get
that is
Here \(h=t_i-t_{i-1}\) is the step size. Now using the non-standard finite difference method, we discretize the inner boundary condition since \(w'(0)=0\), so \(tw'\) must be zero at its first grid point, i.e.,
The outer boundary condition can be dicretized as \(w(1)=w_N=0\).
4.7.2 Numerical illustrations
Table 13 Analytic and NSFD solutions of Problem 7
t | NSFD (W1) | NSFD (W2) | NSFD (W3) | NSFD (W4) | Analytic | FD |
---|---|---|---|---|---|---|
0 | 0.009559563 | 0.003003089 | 0.00090062 | 0.000261825 | 0 | 0.000337719 |
0.1 | 0.009559563 | 0.009153682 | 0.009040324 | 0.009010155 | 0.009 | 0.009037187 |
0.2 | 0.031043248 | 0.031765354 | 0.031942588 | 0.031985678 | 0.032 | 0.031983681 |
0.3 | 0.061458723 | 0.062616703 | 0.062905244 | 0.062976331 | 0.063 | 0.062982141 |
0.4 | 0.094271592 | 0.095569041 | 0.095893158 | 0.095973304 | 0.096 | 0.095975332 |
0.5 | 0.123308409 | 0.124577932 | 0.124895239 | 0.12497382 | 0.125 | 0.124973675 |
0.6 | 0.142491324 | 0.143623645 | 0.14390646 | 0.143976623 | 0.144 | 0.143975374 |
0.7 | 0.145776084 | 0.14669501 | 0.146924048 | 0.146981017 | 0.147 | 0.146979379 |
0.8 | 0.127134778 | 0.127784701 | 0.127946257 | 0.127986568 | 0.128 | 0.1279851 |
0.9 | 0.080547656 | 0.080887364 | 0.080971876 | 0.080992971 | 0.081 | 0.080992024 |
Fig. 19 Comparison of NSFD solutions as \(h\rightarrow 0\) with analytical solution for Problem 7
Table 14 Absolute Errors in solution of Problem 7
t | E1 | E2 | E3 | E4 | E in FD |
---|---|---|---|---|---|
0 | 0.009559563 | 0.003003089 | 0.00090062 | 0.000261825 | 0.000337719 |
0.1 | 0.000559563 | 0.000153682 | 4.03239E\(-\)05 | 1.01547E\(-\)05 | 3.71874E\(-\)05 |
0.2 | 0.000956752 | 0.000234646 | 5.74123E\(-\)05 | 1.4322E\(-\)05 | 1.63194E\(-\)05 |
0.3 | 0.001541277 | 0.000383297 | 9.47557E\(-\)05 | 2.3669E\(-\)05 | 1.7859E\(-\)05 |
0.4 | 0.001728408 | 0.000430959 | 0.000106842 | 2.66963E\(-\)05 | 2.46679E\(-\)05 |
0.5 | 0.001691591 | 0.000422068 | 0.000104761 | 2.61797E\(-\)05 | 2.63253E\(-\)05 |
0.6 | 0.001508676 | 0.000376355 | 9.35402E\(-\)05 | 2.33773E\(-\)05 | 2.46258E\(-\)05 |
0.7 | 0.001223916 | 0.00030499 | 7.59523E\(-\)05 | 1.89826E\(-\)05 | 2.06208E\(-\)05 |
0.8 | 0.000865222 | 0.000215299 | 5.37432E\(-\)05 | 1.34324E\(-\)05 | 1.48999E\(-\)05 |
0.9 | 0.000452344 | 0.000112636 | 2.81237E\(-\)05 | 7.02929E\(-\)06 | 7.97627E\(-\)06 |
Fig. 20 Numerical errors in NSFD and FD for \(h=0.0125\) for Problem 7
Fig. 21 Error in NSFD solutions as \(h\rightarrow 0\) for Problem 7
4.8 Problem 8
Equilibrium of isothermal gas sphere: We consider the following nonlinear SBVP [7]:
with boundary conditions
This equation arises in modelling the equilibrium of isothermal gas sphere. Analytical solution of the above problem is:
4.8.1 Mathematical illustrations
Above problem can be written as
After discretizing it, we get
Applying NSFD scheme 2.2 to the above problem, we get
Here \(h=t_i-t_{i-1}\) is the step size.
Now using the NSFD method, we discretize the inner boundary condition since \(w'(0)=0\), so \(tw'\) must be zero at its first grid point, i.e.,
The outer boundary condition can be dicretized as \(w(1)=w_N=\frac{1}{\sqrt{2}}\).
4.8.2 Numerical illustrations
Table 15 Analytic and NSFD solutions of Problem 8
t | NSFD (W1) | NSFD (W2) | NSFD (W3) | NSFD (W4) | Analytic | FD |
---|---|---|---|---|---|---|
0 | 0.996332253 | 0.998871634 | 0.999677271 | 0.99997214 | 1 | 0.999885538 |
0.1 | 0.996332253 | 0.995429323 | 0.995169214 | 0.99504235 | 0.99503719 | 0.995079661 |
0.2 | 0.98277194 | 0.981194759 | 0.980783562 | 0.980589015 | 0.980580676 | 0.980644482 |
0.3 | 0.960235379 | 0.958487455 | 0.958003395 | 0.95783625 | 0.957826285 | 0.957894974 |
0.4 | 0.930749026 | 0.929098487 | 0.928580805 | 0.92847424 | 0.928476691 | 0.928541545 |
0.5 | 0.896366768 | 0.894959087 | 0.894489422 | 0.894428333 | 0.894427191 | 0.894483312 |
0.6 | 0.859019184 | 0.857913896 | 0.857520941 | 0.8574998 | 0.857492926 | 0.857537843 |
0.7 | 0.820329393 | 0.819538202 | 0.819239468 | 0.819232346 | 0.819231921 | 0.819264993 |
0.8 | 0.781560027 | 0.781064889 | 0.780858965 | 0.780862401 | 0.780868809 | 0.780890166 |
0.9 | 0.743618689 | 0.743386195 | 0.743284162 | 0.74329 | 0.743294146 | 0.743304417 |
Fig. 22 Comparison of NSFD solutions as \(h\rightarrow 0\) with analytical solution for Problem 8
Table 16 Absolute Errors in solution of Problem 8
t | E1 | E2 | E3 | E4 | E in FD |
---|---|---|---|---|---|
0 | 0.003667747 | 0.001128366 | 0.000322729 | 2.786E\(-\)05 | 0.000114462 |
0.1 | 0.001295063 | 0.000392133 | 0.000132023 | 5.15979E\(-\)06 | 4.24711E\(-\)05 |
0.2 | 0.002191264 | 0.000614083 | 0.000202887 | 8.33931E\(-\)06 | 6.38066E\(-\)05 |
0.3 | 0.002409093 | 0.000661169 | 0.000177109 | 9.96478E\(-\)06 | 6.86887E\(-\)05 |
0.4 | 0.002272335 | 0.000621797 | 0.000104114 | 2.45089E\(-\)06 | 6.48545E\(-\)05 |
0.5 | 0.001939577 | 0.000531896 | 6.22315E\(-\)05 | 1.142E\(-\)06 | 5.61206E\(-\)05 |
0.6 | 0.001526258 | 0.00042097 | 2.80156E\(-\)05 | 6.87429E\(-\)06 | 4.49171E\(-\)05 |
0.7 | 0.001097473 | 0.000306281 | 7.547E\(-\)06 | 4.25481E\(-\)07 | 3.30725E\(-\)05 |
0.8 | 0.000691218 | 0.000196079 | 9.84399E\(-\)06 | 6.40844E\(-\)06 | 2.13562E\(-\)05 |
0.9 | 0.000324542 | 9.20484E\(-\)05 | 9.98383E\(-\)06 | 4.14625E\(-\)06 | 1.02708E\(-\)05 |
Fig. 23 Numerical errors in NSFD and FD for \(h=0.0125\) for Problem 8
Fig. 24 Error in NSFD solutions as \(h\rightarrow 0\) for Problem 8
4.9 Problem 9
Thermal distribution in the human head: The following two point nonlinear SBVP \( (c = 2\) and \(g(t,w) = e^{-w(t)})\) has been derived by Duggan and Goodman [11] which describes the thermal distribution profile in the human head
with boundary condition
Analytical solution of the above problem is:
where \( B=3-2\sqrt{2}\).
4.9.1 Mathematical illustrations
Above problem can be written as
After discretizing it, we get
Applying NSFD scheme as described above in (2.3) to the above problem, we get
where \(h=t_{i}-t_{i-1}\) is the step size.
4.9.2 Numerical illustrations
Table 17 Analytic, FD and NSFD solutions of Problem 9
t | NSFD (W1) | NSFD (W2) | NSFD (W3) | NSFD (W4) | Analytic | FD |
---|---|---|---|---|---|---|
0 | 0.266263465 | 0.268451865 | 0.269316716 | 0.269692931 | – | 0.269195827 |
0.1 | 0.266263465 | 0.267495248 | 0.268122301 | 0.268439455 | 0.268756903 | 0.267966449 |
0.2 | 0.262417516 | 0.263660205 | 0.264292812 | 0.264612699 | 0.26493282 | 0.264131907 |
0.3 | 0.255974454 | 0.257242592 | 0.257887677 | 0.25821368 | 0.258539792 | 0.257734361 |
0.4 | 0.246894301 | 0.248207425 | 0.24887461 | 0.249211433 | 0.249548183 | 0.248743821 |
0.5 | 0.235122416 | 0.236506816 | 0.237208719 | 0.237562447 | 0.237915891 | 0.237099948 |
0.6 | 0.220590786 | 0.222079284 | 0.222831647 | 0.223209958 | 0.22358771 | 0.22275 |
0.7 | 0.203215373 | 0.20484866 | 0.205670897 | 0.206083078 | 0.206494486 | 0.205625033 |
0.8 | 0.182895848 | 0.184722875 | 0.185638095 | 0.186095721 | 0.186552018 | 0.185664293 |
0.9 | 0.159513476 | 0.161592358 | 0.162628197 | 0.163145114 | 0.163659686 | 0.162760884 |
1 | 0.132927843 | 0.135328001 | 0.13651715 | 0.137110278 | 0.137698751 | 0.13678735 |
Here the analytic solution has been taken from Table 7 of [24].
Fig. 25 Comparison of NSFD solutions as \(h\rightarrow 0\) with analytical solution for Problem 9
Table 18 Absolute Errors in solution of Problem 9
t | E1 | E2 | E3 | E4 | E in FD |
---|---|---|---|---|---|
0.1 | 0.002493438 | 0.001261655 | 0.000634602 | 0.000317448 | 0.000790454 |
0.2 | 0.002515304 | 0.001272615 | 0.000640008 | 0.000320121 | 0.000800913 |
0.3 | 0.002565338 | 0.0012972 | 0.000652115 | 0.000326112 | 0.000805431 |
0.4 | 0.002653882 | 0.001340758 | 0.000673573 | 0.00033675 | 0.000804362 |
0.5 | 0.002793475 | 0.001409075 | 0.000707172 | 0.000353444 | 0.000815943 |
0.6 | 0.002996924 | 0.001508426 | 0.000756063 | 0.000377752 | 0.00083771 |
0.7 | 0.003279113 | 0.001645826 | 0.000823589 | 0.000411408 | 0.000869453 |
0.8 | 0.00365617 | 0.001829143 | 0.000913923 | 0.000456297 | 0.000887725 |
0.9 | 0.00414621 | 0.002067328 | 0.001031489 | 0.000514572 | 0.000898802 |
1 | 0.004770908 | 0.00237075 | 0.001181601 | 0.000588473 | 0.0009114 |
Fig. 26 Numerical errors in NSFD and FD for \(h=0.0125\) for Problem 9
Fig. 27 Error in NSFD solutions as \(h\rightarrow 0\) for Problem 9
5 Remark
Here W1, W2, W3, W4 represent NSFD solutions and E1, E2, E3, E4 represent error in solutions for \(h=0.1\), \(h=0.05\), \(h=0.025\), \(h=0.0125\) respectively. In Figs. 1, 4, 7, 10, 13, 16, 19, 22 and 25, we have shown how NSFD solutions are behaving qualitatively in the same way as \(h\rightarrow 0\).
The numerical errors generated by NSFD and FD for \(h=0.0125\) are shown in Figs. 2, 5, 8, 11, 14, 17, 20, 23 and 26. We can see that errors in NSFD is very less as compared to FD even in the neighbourhood of singular point \(t=0\). And in Figs. 3, 6, 9, 12, 15, 18, 21, 24 and 27 we show that \(E \rightarrow 0\) as \(h\rightarrow 0\).
In Tables 1, 3, 5, 7, 9, 11, 13, 15 and 17, we have computed the values of NSFD solutions for various h and compared them with solutions of FD and Exact. In Tables 2, 4, 6, 8, 10, 12, 14, 16 and 18, we have compared the errors in NSFD for various h and also the errors in FD and NSFD for \(h=0.0125\).
6 Conclusions
The tables of approximate solution obtained from non-standard schemes show that the NSFD schemes give better results than standard schemes for non-linear SBVPs. They tackle singular BVP more accurately and efficiently than the FD schemes. These schemes are also quite stable as they rarely contain any instabilities [16]. The solution is easily approximated by NSFD near the singularity at the origin where the FD methods usually fail. The non-linear terms which are approximated non-locally on the computational grid also increases the accuracy and efficiency of the solution.
Change history
28 February 2019
In the original article, the analytical solution of Problem 9 is published incorrectly.
28 February 2019
In the original article, the analytical solution of Problem 9 is published incorrectly.
References
G. Adomian, R. Rach, Inversion of nonlinear stochastic operators. J. Math. Anal. Appl. 91(1), 39–46 (1983)
G. Adomian, R. Rach, Modified decomposition solution of linear and nonlinear boundary-value problems. Nonlinear Anal. TMA 23(5), 615–619 (1994)
J.V. Baxley, Y. Gu, Nonlinear boundary value problems for shallow membrane caps. Commun. Appl. Anal. 3, 327–344 (1999)
R. Buckmire, Investigations of nonstandard, Mickens-type, finite-difference schemes for singular boundary value problems in cylindrical or spherical coordinates. Numer. Methods Partial Differ. Equ. 19, 380–398 (2003)
R. Buckmire, Application of a Mickens finite-difference scheme to the cylindrical bratu-gelfand problem. Numer. Methods Partial Differ. Equ. 20, 327–337 (2004)
P.L. Chamber, On the solution of the Possion-Boltzmann equation with the application to the theory of thermal explosions. J. Chem. Phys. 20, 1795–1797 (1952)
S. Chandrasekhar, Introduction to the Study of Stellar Structure (Dover, New York, 1967)
M.M. Chawla, C.P. Katti, A uniform mesh finite difference method for a class of singular two-point boundary value problems. SIAM J. Numer. Anal. 22(3), 561–565 (1985)
M.M. Chawla, P.N. Shivkumar, On the existence of solutions of a class of singular nonlinear two-point boundary value problems. J. Comput. Appl. Math. 19, 379–388 (1987)
R.W. Dickey, Rotationally symmetric solutions for shallow membrane caps. Q. Appl. Math. XLVII, 571–581 (1989)
R. Duggan, A. Goodman, Pointwise bounds for a nonlinear heat conduction model of the human head. Bull. Math. Biol. 48(2), 229–236 (1986)
A. Ebaid, A new analytical and numerical treatment for singular two-point boundary value problems via the adomian decomposition method. J. Comput. Appl. Math. 235(8), 1914–1924 (2011)
J.B. Keller, Electrohydrodynamics I. The equilibrium of a charged gas in a container. J. Ration. Mech. Anal. 5, 715–724 (1956)
S.A. Khuri, A. Sayfy, A novel approach for the solution of a class of singular boundary value problems arising in physiology. Math. Comput. Model. 52, 626–636 (2010)
S.J. Liao, A new analytic algorithm of Lane Emden type equations. Appl. Math. Comput. 142, 1–16 (2003)
R.E. Mickens, Nonstandard Finite Difference Models of Differential Equations (World Scientific, Singapore, 1994), p. 65
R.E. Mickens, Calculation of denominator functions for nonstandard finite difference schemes for differential equations satisfying a positivity condition. Numer. Methods Partial Differ. Equ. 23, 672–691 (2007)
A.A. Obayomi, M.O. Oke, A non-standard numerical approach to the solution of some second-order ordinary differential equations. Asian-Eur. J. Math. 08(04), 1550076 (2015)
R.K. Pandey, A finite difference method for a class of singular two-point boundary value problems arising in physiology. Int. J. Comput. Math. 65, 131–140 (1997)
R.K. Pandey, A.K. Singh, On the convergence of fourth order finite difference method for weakly regular singular boundary value problems. Int. J. Comput. Math. 81, 227–238 (2004)
R.K. Pandey, A.K. Verma, Existence-uniqueness results for a class of singular boundary value problems arising in physiology. Nonlinear Anal. RWA 9, 40–52 (2008)
R.K. Pandey, A.K. Verma, Existence-uniqueness results for a class of singular boundary value problems-ii. J. Math. Anal. Appl. 338, 1387–1396 (2008)
R.D. Russell, L.F. Shampine, Numerical methods for singular boundary value problems. SlAM J. Numer. Anal. 12, 13–36 (1975)
M. Singh, A.K. Verma, An effective computational technique for a class of Lane Emden equations. J. Math. Chem. 54, 231–251 (2016)
R. Singh, J. Kumar, An efficient numerical technique for the solution of nonlinear singular boundary value problems. Comput. Phys. Commun. 185, 1282–1289 (2014)
M. Turkyilmazoglu, Effective computation of exact and analytic approximate solutions to singular nonlinear equations of Lane Emden Fowler type. Appl. Math. Model. 37, 7539–7548 (2013)
S. Ushiki, Central difference scheme and chaos. Int. J. Non-Linear Mech. D4, 407–424 (1982)
A.M. Wazwaz, Adomian decomposition method for a reliable treatment of the Emden Fowler equation. Appl. Math. Comput. 161, 543–560 (2005)
M. Yamaguti, S. Ushiki, Chaos in numerical analysis of ordinary differential equations. Phys. D Nonlinear Phenom. 3, 618–626 (1981)
Acknowledgements
We are thankful to Prof. R. E. Mickens and Ron Buckmire whose ideas are applied while constructing the schemes for all of the above problems. We are also thankful to anonymous reviewers for their valuable comments.
Author information
Authors and Affiliations
Corresponding author
Additional information
Amit Kumar Verma: In the memory of my loving mother Late Shrimati Mithlesh Verma, a teacher by profession.
Rights and permissions
About this article
Cite this article
Verma, A.K., Kayenat, S. On the convergence of Mickens’ type nonstandard finite difference schemes on Lane-Emden type equations. J Math Chem 56, 1667–1706 (2018). https://doi.org/10.1007/s10910-018-0880-y
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10910-018-0880-y