Application of perturbation–iteration method to Lotka–Volterra equations

==Abstract==

Perturbation–iteration method is generalized for systems of first order differential equations. Approximate solutions of Lotka–Volterra systems are obtained using the method. Comparisons of our results with each other and with numerical solutions are given. The method is implemented in ''Mathematica'', a major computer algebra system. The package ''PerturbationIteration.m'' automatically carries out the tedious calculations of the method.

==Keywords==

Perturbation method; Perturbation–iteration method; Symbolic computation; Lotka–Volterra equations; Systems of first order differential equations

==1. Introduction==

The study of methods for approximate solutions of nonlinear models in real life has always been a growing branch of applied mathematical sciences. Many methods with different capabilities and limitations have been developed. The well-known approximate analytical method, the perturbation technique [[#b0005|[1]]] can deal with weakly nonlinear systems due to the small parameter assumption. To overcome this limitation, methods such as the linearized perturbation method [[#b0010|[2]]], the Lindstedt–Poincaré method with modified frequency expansion [[#b0015|[3]]], the multiple-scale Lindstedt–Poincaré method [[#b0020|[4]]] and the parameter expanding method [[#b0025|[5]]] were developed. Also methods such as the Adomian decomposition method [[#b0030|[6]]], the variational iteration method [[#b0035|[7]]], and the homotopy analysis method [[#b0040|[8]]] were among the non-perturbative methods that were applied to many interesting mathematical problems.

Other attempts to treat both weakly and strongly nonlinear problems were through iteration procedures which used pre-formed alternative equations to obtain approximate solutions. Just to list a few, He [[#b0045|[9]]] linearized the nonlinear terms by substitution of iterative solution functions from previous iteration results, Mickens’ iteration procedure [[#b0050|[10]]] was for specific problems, and variational iteration method [[#b0055|[11]]] was used to solve boundary value problems.

Recently, the perturbation–iteration method, which gives valuable solutions for strongly nonlinear problems [[#b0060|[12]]] and [[#b0065|[13]]] has been developed. In [[#b0070|[14]]], approximate solutions of some nonlinear heat transfer problems were obtained, and the comparison of the results showed that perturbation–iteration method fits better than the variational iteration method as the parameter measuring the nonlinearity takes larger values.

The aim of this study was to develop perturbation–iteration algorithms for systems of first order differential equations and to obtain accurate solutions of Lotka–Volterra differential equations. Many analytical methods such as Adomian decomposition [[#b0075|[15]]], variational iteration method [[#b0080|[16]]], homotopy analysis method [[#b0085|[17]]], and optimal parametric iteration method [[#b0090|[18]]] were successfully applied to this type of problems.

The paper is organized as follows. In Section [[#s0010|2]], the perturbation–iteration algorithm for systems is explained. The method is applied to Lotka–Volterra equations in Section [[#s0015|3]]. In Section [[#s0050|4]], comparisons of our results with other methods are given. The package ''PerturbationIteration.m'' is described in Section  [[#s0055|5]]. Concluding remarks are given in Section [[#s0060|6]].

<span id='s0010'></span>
==2. Perturbation–iteration method for systems of first order differential equations==

In this section, the perturbation–iteration method, described in [[#b0060|[12]]] and [[#b0065|[13]]], is generalized toward systems of first order differential equations. The generalization is developed by taking arbitrary number of terms in Taylor series expansion and one correction term in perturbation expansion PIA(1,M) i.e. the first number expresses the correction terms in the perturbation expansion and the second number expresses the derivative orders in the Taylor expansion. Consider the following system of first order differential equations:

<span id='e0005'></span>
{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>F_k({\overset{\cdot}{u}}_k\mbox{,}u_j\mbox{,}\epsilon \mbox{,}t)=</math><math>0\mbox{,}\quad j=1\mbox{,}2\mbox{,}\ldots \mbox{,}K\mbox{,}\quad k=</math><math>1\mbox{,}2\mbox{,}\ldots \mbox{,}K\mbox{,}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (1)
|}

where ''u<sub>j</sub>'' are the dependent variables, ''t'' is the independent variable'', K  '' is the number of dependent variables, <math display="inline">\epsilon </math> is the artificially introduced small parameter and the dot stands for the derivative. Reconsider Eq. [[#e0005|(1)]] in the following iterative form:

{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>F_k({\overset{\cdot}{u}}_{k\mbox{,}n+1}\mbox{,}u_{j\mbox{,}n+1}\mbox{,}\epsilon \mbox{,}t)=</math><math>0\mbox{,}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (2)
|}

where the subscript ''n'' expresses the number of iterations completed in the procedure. More clearly, the system of equations is given below

<span id='e0015'></span>
{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>F_1=F_1({\overset{\cdot}{u}}_1\mbox{,}u_1\mbox{,}u_2\mbox{,}\ldots \mbox{,}u_K\mbox{,}\epsilon \mbox{,}t)</math>
|-
|<math>F_2=F_2({\overset{\cdot}{u}}_2\mbox{,}u_1\mbox{,}u_2\mbox{,}\ldots \mbox{,}u_K\mbox{,}\epsilon \mbox{,}t)</math>
|-
|<math>\vdots </math>
|-
|<math>F_K=F_K({\overset{\cdot}{u}}_K\mbox{,}u_1\mbox{,}u_2\mbox{,}\ldots \mbox{,}u_K\mbox{,}\epsilon \mbox{,}t)\mbox{.}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (3)
|}

Next, define the following iterative perturbation series:

<span id='e0020'></span>
{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>u_{j\mbox{,}n+1}=u_{j\mbox{,}n}+\epsilon u_{j\mbox{,}n}^c\mbox{,}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (4)
|}

Taylor series expansion of equation [[#e0005|(1)]] is given as follows:

<span id='e0025'></span>
{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>F_k=\sum_{m=0}^M\frac{1}{m!}{\left[{\left(\frac{d}{d\epsilon }\right)}^mF_k\right]}_{\epsilon =0}{\epsilon }^m\mbox{,}\quad k=</math><math>1\mbox{,}2\ldots \ldots .K</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (5)
|}

with the derivative operator defined as follows:

<span id='e0030'></span>
{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>\frac{d}{d\epsilon }=\frac{\partial {\overset{\cdot}{u}}_{k\mbox{,}n+1}}{\partial \epsilon }\frac{\partial }{\partial {\overset{\cdot}{u}}_{k\mbox{,}n+1}}+</math><math>\sum_{j=1}^K\left(\frac{\partial u_{j\mbox{,}n+1}}{\partial \epsilon }\frac{\partial }{\partial u_{j\mbox{,}n+1}}\right)+</math><math>\frac{\partial }{\partial \epsilon }\mbox{.}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (6)
|}

Calculating derivatives at ''ε'' = 0 and substituting Eq. [[#e0030|(6)]] into Eq. [[#e0025|(5)]] yields

<span id='e0035'></span>
{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>F_k=\sum_{m=0}^M\frac{1}{m!}{\left\{{\left({\overset{\cdot}{u}}_{k\mbox{,}n}^c\frac{\partial }{\partial {\overset{\cdot}{u}}_{k\mbox{,}n+1}}+\sum_{j=1}^Ku_{j\mbox{,}n}^c\frac{\partial }{\partial u_{j\mbox{,}n+1}}+\frac{\partial }{\partial \epsilon }\right)}^mF_k\right\}}_{\epsilon =0}{\epsilon }^m=</math><math>0\mbox{,}\quad k=1\mbox{,}2\ldots \ldots .K</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (7)
|}

Note that, by taking larger values of ''M,'' i.e., increasing the number of terms in the Taylor series expansion, one can develop different algorithms for the specific problem with the help of Eq.  [[#e0035|(7)]]. Since one correction term in the perturbation expansion and m’th order derivatives in the Taylor expansion are taken, the algorithm is PIA(1,m). A more general algorithm with PIA(n,m) can be constructed but would cause too much complexity in the applications.

<span id='s0015'></span>
==3. Application to Lotka–Volterra problems==

The Lotka–Volterra equations, also known as the ''predator–prey equations'', are frequently used by mathematicians to describe the time evolution of the species in the dynamics of biological systems  [[#b0075|[15]]], [[#b0080|[16]]] and [[#b0085|[17]]]. In the more general framework one can study the following multidimensional Lotka–Volterra system:

<span id='e0040'></span>
{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>\frac{{du}_k}{dt}=u_k\left(b_k+\sum_{j=1}^Ka_{kj}u_j\right)\mbox{,}\quad \quad u_k(0)=</math><math>c_k\mbox{,}\quad k=1\mbox{,}2\mbox{,}\ldots \mbox{,}K\mbox{,}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (8)
|}

where ''c<sub>k</sub>'' represent the populations of the species at the beginning of the evolution. In the following subsections, solutions of one dimensional and multidimensional systems such as ''K'' ''='' 1, 2, 3,… are investigated by the perturbation–iteration method.

<span id='s0020'></span>
===3.1. One dimensional system (''K'' = 1)===

One dimensional Lotka–Volterra equation, known as the Verhulst equation [[#b0095|[19]]], describes the behavior of population in time of one species competing for a given finite source of food:

<span id='e0045'></span>
{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>\frac{du}{dt}=u(b+au)\mbox{,}\quad b>0\mbox{,}\quad a<0\mbox{,}\quad u(0)>0\mbox{.}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (9)
|}

The exact solution of [[#e0045|(9)]] is

{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>u(t)=\begin{array}{ll}
\frac{{be}^{bt}}{\frac{b+au(0)}{u(0)}-{ae}^{bt}}\mbox{,} & b\quad \not =\quad 0\\
\frac{u(0)}{1-au(0)t}\mbox{,} & b=0\mbox{.}
\end{array}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (10)
|}

Approximate solutions of Eq. [[#e0045|(9)]] will be obtained using two different perturbation–iteration algorithms by taking ''M'' = 1 and ''M'' = 2.

====3.1.1. Perturbation iteration method with first order derivatives in the Taylor series expansion PIA(1,1)====

First rewrite Eqs. [[#e0045|(9)]] and [[#e0015|(3)]] in the following form:

{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>F_1={\overset{\cdot}{u}}_{1\mbox{,}n+1}-{bu}_{1\mbox{,}n+1}-</math><math>\epsilon {au}_{1\mbox{,}n+1}^2=0\mbox{,}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (11)
|}

<span id='e0060'></span>
{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>u_{1\mbox{,}n+1}=u_{1\mbox{,}n}+\epsilon u_{1\mbox{,}n}^c\mbox{,}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (12)
|}

where ''ε'' is artificially introduced as a small parameter. Reorganizing Eq.  [[#e0035|(7)]] for Eq. [[#e0045|(9)]] yields

<span id='e0065'></span>
{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>\frac{{du}_{1\mbox{,}n}^c}{dt}+\frac{\partial F_1/\partial u_{n+1}}{\partial F_1/\partial {\overset{\cdot}{u}}_{n+1}}u_{1\mbox{,}n}^c=</math><math>-\frac{\epsilon F_1}{\partial F_1/\partial {\overset{\cdot}{u}}_{n+1}}-</math><math>\frac{\partial F_1/\partial \epsilon }{\partial F_1/\partial {\overset{\cdot}{u}}_{n+1}}\mbox{,}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (13)
|}

where the derivatives are

<span id='e0070'></span>
{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>{F_1}_{\epsilon =0}={\overset{\cdot}{u}}_n-{bu}_n\mbox{,}\quad {\frac{\partial F_1}{\partial u_{n+1}}}_{\epsilon =0}=</math><math>-b\mbox{,}\quad {\frac{\partial F_1}{\partial {\overset{\cdot}{u}}_{n+1}}}_{\epsilon =0}=</math><math>1\mbox{,}\quad {\frac{\partial F_1}{\partial \epsilon }}_{\epsilon =0}=</math><math>-a\quad u_n^2\mbox{.}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (14)
|}

Finally, the iteration equation is obtained by substituting [[#e0070|(14)]] into [[#e0065|(13)]] and setting ''ε'' = 1:

<span id='e0075'></span>
{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>\frac{{du}_{1\mbox{,}n}^c}{dt}-{bu}_{1\mbox{,}n}^c=</math><math>-{\overset{\cdot}{u}}_{1\mbox{,}n}+{bu}_{1\mbox{,}n}+</math><math>{au}_{1\mbox{,}n}^2\mbox{.}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (15)
|}

One can take the initially assumed function as follows:

{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>u_{1\mbox{,}0}=\alpha e^{bt}\mbox{,}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (16)
|}

which satisfy the initial condition exactly.

Starting with the above initial function, first <math display="inline">u_{1\mbox{,}0}^c</math> is calculated from Eq. [[#e0075|(15)]] and then substituted into Eq. [[#e0060|(12)]] to obtain ''u''<sub>1,1</sub> as the solution of the first iteration. The iteration process is repeated using the previous solution as an initial guess until satisfactory result is obtained. This way, one can obtain the first and second iteration solutions as follows:

{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>u_{1\mbox{,}1}=\alpha \mbox{e}^{bt}+\frac{{\alpha }^2a}{b}\mbox{e}^{bt}(e^{bt}-</math><math>1)\mbox{,}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (17)
|}

{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>u_{1\mbox{,}2}=\alpha \mbox{e}^{bt}+\frac{{\alpha }^2a\mbox{e}^{bt}(\mbox{e}^{bt}-1)}{b}+</math><math>\frac{{\alpha }^3a^2\mbox{e}^{bt}{\left(\mbox{e}^{bt}-1\right)}^2}{b^2}+</math><math>\frac{{\alpha }^4a^3\mbox{e}^{bt}{\left(\mbox{e}^{bt}-1\right)}^3}{3b^3}\mbox{.}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (18)
|}

====3.1.2. Perturbation iteration method with second order derivatives in the Taylor series expansion PIA(1,2)====

Taking up to second order derivative terms in the Taylor series expansion leads to the iteration formula:

<span id='e0095'></span>
{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>\frac{{du}_{1\mbox{,}n}^c}{dt}-(b+2{au}_{1\mbox{,}n})u_{1\mbox{,}n}^c=</math><math>-{\overset{\cdot}{u}}_{1\mbox{,}n}+{bu}_{1\mbox{,}n}+</math><math>{au}_{1\mbox{,}n}^2\mbox{.}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (19)
|}

Using the same initial guess function <math display="inline">u_{1\mbox{,}0}=\alpha e^{bt}\mbox{,}</math> one can obtain the first iteration solution:

{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>u_{1\mbox{,}1}=\frac{\alpha }{2}e^{bt}\left(1+e^{\frac{2\alpha a}{b}(e^{bt}-1)}\right)\mbox{,}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (20)
|}

The second and third iteration solutions are as follows:

{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>u_{1\mbox{,}2}=\frac{\alpha \mbox{e}^{bt-\frac{2a\alpha }{b}}}{8}\left(3\mbox{e}^{\frac{2a\alpha }{b}}+\right. </math><math>\left. 4\mbox{e}^{\frac{2a\alpha \mbox{e}^{bt}}{b}}+\right. </math><math>\left. \mbox{e}^{\frac{2\alpha a(2\mbox{e}^{bt}-1)}{b}}-\right. </math><math>\left. \frac{4\alpha a}{b}\mbox{e}^{\frac{2\alpha a\mbox{e}^{bt}}{b}}(\mbox{e}^{bt}-\right. </math><math>\left. 1)\right)\mbox{,}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (21)
|}

{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>u_{1\mbox{,}3}=\frac{1}{384}\mbox{e}^{bt-\frac{6a\alpha }{b}}\alpha 117\mbox{e}^{\frac{6a\alpha }{b}}+</math><math>15\mbox{e}^{\frac{6a\mbox{e}^{bt}\alpha }{b}}+161\mbox{e}^{\frac{2a(2+\mbox{e}^{bt})\alpha }{b}}+</math><math>90\mbox{e}^{\frac{2a(1+2\mbox{e}^{bt})\alpha }{b}}+</math><math>\mbox{e}^{\frac{2a(-1+4\mbox{e}^{bt})\alpha }{b}}</math>
|-
|<math>-\frac{12a}{b}\mbox{e}^{\frac{2a\mbox{e}^{bt}\alpha }{b}}(-</math><math>1+\mbox{e}^{bt})(20\mbox{e}^{\frac{4a\alpha }{b}}+</math><math>\mbox{e}^{\frac{4a\mbox{e}^{bt}\alpha }{b}}+12\mbox{e}^{\frac{2a(1+\mbox{e}^{bt})\alpha }{b}})\alpha </math>
|-
|<math>+\frac{24{\alpha }^2a^2}{b^2}\mbox{e}^{\frac{2a(1+\mbox{e}^{bt})\alpha }{b}}{\left(-1+\mbox{e}^{bt}\right)}^2\left(5\mbox{e}^{\frac{2a\alpha }{b}}+\right. </math><math>\left. 2\mbox{e}^{\frac{2a\mbox{e}^{bt}\alpha }{b}}\right)\mbox{.}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (22)
|}

In second and third iteration solutions, as the calculations get more involved, the second term in [[#e0095|(19)]], ''b''+2''au''<sub>1,</sub>''<sub>n</sub>'' is approximated as ''b''+2''au''<sub>1,0</sub> for simplicity and help of a computer algebra system such as ''Mathematica'' is essential.

<span id='s0035'></span>
===3.2. Two dimensional system (system with two species, ''K'' = 2)===

Two dimensional Lotka–Volterra system was first proposed by Lotka [[#b0100|[20]]] and Volterra [[#b0105|[21]]] without knowing each other’s work. The following two dimensional competitive Lotka–Volterra equation models a pair of species competing for a common resource:

<span id='e0115'></span>
{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>\frac{{du}_1}{dt}=u_1(b_1+a_{11}u_1+a_{12}u_2)\mbox{,}u_1(0)=</math><math>\alpha \mbox{,}</math>
|-
|<math>\frac{{du}_2}{dt}=u_2(b_2+a_{21}u_1+a_{22}u_2)\mbox{,}u_2(0)=</math><math>\beta \mbox{,}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (23)
|}

where ''a''<sub>11</sub>'', a''<sub>12</sub>, ''a''<sub>21</sub>, ''a''<sub>22</sub> are parameters representing the interaction of the species and the self-interaction, and ''α'', ''β'' are the initial size of populations. Perturbation–iteration method with first order derivatives in the Taylor series PIA(1,1) will be applied to Eq.  [[#e0115|(23)]]. First, re-write Eq. [[#e0115|(23)]] as follows:

{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>F_1={\overset{\cdot}{u}}_{1\mbox{,}n+1}-u_{1\mbox{,}n+1}(b_1+</math><math>\epsilon a_{11}u_{1\mbox{,}n+1}+\epsilon a_{12}u_{2\mbox{,}n+1})=</math><math>0</math>
|-
|<math>F_2={\overset{\cdot}{u}}_{2\mbox{,}n+1}-u_{2\mbox{,}n+1}(b_2+</math><math>\epsilon a_{21}u_{1\mbox{,}n+1}+\epsilon a_{22}u_{2\mbox{,}n+1})=</math><math>0\mbox{.}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (24)
|}

Setting ''K'' = 2 and ''M'' = 1, transforms Eq. [[#e0035|(7)]] into the following:

<span id='e0125'></span>
{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>F_1\cong {F_1}_{\epsilon =0}+{\left\{\left[{\overset{\cdot}{u}}_{1\mbox{,}n}^c\frac{\partial }{\partial {\overset{\cdot}{u}}_{1\mbox{,}n+1}}+u_{1\mbox{,}n}^c\frac{\partial }{\partial u_{1\mbox{,}n+1}}+u_{2\mbox{,}n}^c\frac{\partial }{\partial u_{2\mbox{,}n+1}}+\frac{\partial }{\partial \epsilon }\right]F_1\right\}}_{\epsilon =0}\epsilon =</math><math>0</math>
|-
|<math>F_2\cong {F_2}_{\epsilon =0}+{\left\{\left[{\overset{\cdot}{u}}_{2\mbox{,}n}^c\frac{\partial }{\partial {\overset{\cdot}{u}}_{2\mbox{,}n+1}}+u_{2\mbox{,}n}^c\frac{\partial }{\partial u_{2\mbox{,}n+1}}+u_{1\mbox{,}n}^c\frac{\partial }{\partial u_{1\mbox{,}n+1}}+\frac{\partial }{\partial \epsilon }\right]F_2\right\}}_{\epsilon =0}\epsilon =</math><math>0</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (25)
|}

with

<span id='e0130'></span>
{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>{F_1}_{\epsilon =0}={\overset{\cdot}{u}}_{1\mbox{,}n}-</math><math>b_1u_{1\mbox{,}n}\mbox{,}\quad {F_2}_{\epsilon =0}=</math><math>{\overset{\cdot}{u}}_{2\mbox{,}n}-b_2u_{2\mbox{,}n}\mbox{,}\quad {\frac{\partial F_1}{\partial {\overset{\cdot}{u}}_{1\mbox{,}n+1}}}_{\epsilon =0}=</math><math>1\mbox{,}\quad {\frac{\partial F_2}{\partial {\overset{\cdot}{u}}_{2\mbox{,}n+1}}}_{\epsilon =0}=</math><math>1\mbox{,}</math>
|-
|<math>{\frac{\partial F_1}{\partial u_{1\mbox{,}n+1}}}_{\epsilon =0}=</math><math>-b_1\mbox{,}\quad {\frac{\partial F_1}{\partial u_{2\mbox{,}n+1}}}_{\epsilon =0}=</math><math>0\mbox{,}\quad {\frac{\partial F_1}{\partial \epsilon }}_{\epsilon =0}=</math><math>-a_{11}u_{1\mbox{,}n}^2-a_{12}u_{1\mbox{,}n}u_{2\mbox{,}n}\mbox{,}</math>
|-
|<math>{\frac{\partial F_2}{\partial \epsilon }}_{\epsilon =0}=</math><math>-a_{22}u_{2\mbox{,}n}^2-a_{21}u_{1\mbox{,}n}u_{2\mbox{,}n}\mbox{,}\quad {\frac{\partial F_2}{\partial u_{2\mbox{,}n+1}}}_{\epsilon =0}=</math><math>-b_2\mbox{,}\quad {\frac{\partial F_2}{\partial u_{1\mbox{,}n+1}}}_{\epsilon =0}=</math><math>0\mbox{.}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (26)
|}

Substituting [[#e0130|(26)]] into [[#e0125|(25)]] leads to the final form of the iteration equation:

{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>{\overset{\cdot}{u}}_{1\mbox{,}n}^c-b_1u_{1\mbox{,}n}^c=</math><math>a_{11}u_{1\mbox{,}n}^2+a_{12}u_{1\mbox{,}n}u_{2\mbox{,}n}-</math><math>{\overset{\cdot}{u}}_{1\mbox{,}n}+b_1u_{1\mbox{,}n}\mbox{,}</math>
|-
|<math>{\overset{\cdot}{u}}_{2\mbox{,}n}^c-b_2u_{2\mbox{,}n}^c=</math><math>a_{22}u_{2\mbox{,}n}^2+a_{21}u_{1\mbox{,}n}u_{2\mbox{,}n}-</math><math>{\overset{\cdot}{u}}_{2\mbox{,}n}+b_2u_{2\mbox{,}n}\mbox{.}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (27)
|}

As Eq. [[#e0020|(4)]] become:

<span id='e0140'></span>
{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>u_{1\mbox{,}n+1}=u_{1\mbox{,}n}+u_{1\mbox{,}n}^c\mbox{,}</math>
|-
|<math>u_{2\mbox{,}n+1}=u_{2\mbox{,}n}+u_{2\mbox{,}n}^c\mbox{,}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (28)
|}

taking

<span id='e0145'></span>
{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>u_{1\mbox{,}0}=\alpha e^{b_1t}\mbox{,}\quad u_{2\mbox{,}0}=</math><math>\beta e^{b_2t}\mbox{,}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (29)
|}

as an initial approximation satisfying the initial conditions leads to the following:

<span id='e0150'></span>
{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>u_{1\mbox{,}0}^c=\frac{{\alpha }^2a_{11}\mbox{e}^{{tb}_1}(-1+\mbox{e}^{{tb}_1})}{b_1}+</math><math>\frac{\alpha \beta a_{12}\mbox{e}^{{tb}_1}(-1+\mbox{e}^{{tb}_2})}{b_2}\mbox{,}</math>
|-
|<math>u_{2\mbox{,}0}^c=\frac{\alpha \beta a_{21}\mbox{e}^{{tb}_2}(-1+\mbox{e}^{{tb}_1})}{b_1}+</math><math>\frac{{\beta }^2a_{22}\mbox{e}^{{tb}_2}(-1+\mbox{e}^{{tb}_2})}{b_2}\mbox{,}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (30)
|}

as correction terms. Combining [[#e0145|(29)]] and [[#e0150|(30)]] in [[#e0140|(28)]] yields the first iteration solutions:

<span id='e0155'></span>
{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>u_{1\mbox{,}1}=\alpha \mbox{e}^{{tb}_1}+\frac{{\alpha }^2a_{11}\mbox{e}^{{tb}_1}(-1+\mbox{e}^{{tb}_1})}{b_1}+</math><math>\frac{\alpha \beta a_{12}\mbox{e}^{{tb}_1}(-1+\mbox{e}^{{tb}_2})}{b_2}\mbox{,}</math>
|-
|<math>u_{2\mbox{,}1}=\beta \mbox{e}^{{tb}_2}+\frac{\alpha \beta a_{21}\mbox{e}^{{tb}_2}(-1+\mbox{e}^{{tb}_1})}{b_1}+</math><math>\frac{{\beta }^2a_{22}\mbox{e}^{{tb}_2}(-1+\mbox{e}^{{tb}_2})}{b_2}\mbox{.}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (31)
|}

For briefness, higher iteration results are not given here, but numerical results of those are given in numerical comparison tables.

<span id='s0040'></span>
===3.3. Three dimensional system (system with three species, ''K'' = 3)===

The three dimensional Lotka–Volterra system models population dynamics of three competitive species in an ecosystem [[#b0110|[22]]]:

<span id='e0160'></span>
{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>\frac{{du}_1}{dt}=u_1(b_1+a_{11}u_1+a_{12}u_2+a_{13}u_3)\mbox{,}\quad u_1(0)=</math><math>\alpha \mbox{,}</math>
|-
|<math>\frac{{du}_2}{dt}=u_2(b_2+a_{21}u_1+a_{22}u_2+a_{23}u_3)\mbox{,}\quad u_2(0)=</math><math>\beta \mbox{,}</math>
|-
|<math>\frac{{du}_3}{dt}=u_3(b_3+a_{31}u_1+a_{32}u_2+a_{33}u_3)\mbox{,}\quad u_3(0)=</math><math>\gamma \mbox{.}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (32)
|}

Using <math display="inline">u_{1\mbox{,}0}=\alpha \mbox{e}^{b_1t}\mbox{,}</math><math display="inline">u_{2\mbox{,}0}=\beta \mbox{e}^{b_2t}\mbox{,}</math><math display="inline">u_{3\mbox{,}0}=\gamma \mbox{e}^{b_3t}</math> as initial functions, approximate solutions after the first iteration are as follows:

{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>u_{1\mbox{,}1}=\alpha \mbox{e}^{{tb}_1}+\frac{{\alpha }^2a_{11}\mbox{e}^{{tb}_1}(-1+\mbox{e}^{{tb}_1})}{b_1}+</math><math>\frac{\alpha \beta a_{12}\mbox{e}^{{tb}_1}(-1+\mbox{e}^{{tb}_2})}{b_2}+</math><math>\frac{\alpha \gamma a_{13}\mbox{e}^{{tb}_1}(-1+\mbox{e}^{{tb}_3})}{b_3}\mbox{,}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (33)
|}

{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>u_{2\mbox{,}1}=\beta \mbox{e}^{{tb}_2}+\frac{\alpha \beta a_{21}\mbox{e}^{{tb}_2}(-1+\mbox{e}^{{tb}_1})}{b_1}+</math><math>\frac{{\beta }^2a_{22}\mbox{e}^{{tb}_2}(-1+\mbox{e}^{{tb}_2})}{b_2}+</math><math>\frac{\beta \gamma a_{23}\mbox{e}^{{tb}_2}(-1+\mbox{e}^{{tb}_3})}{b_3}\mbox{,}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (34)
|}

{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>u_{3\mbox{,}1}=\gamma \mbox{e}^{{tb}_3}+\frac{\alpha \gamma a_{31}\mbox{e}^{{tb}_3}(-1+\mbox{e}^{{tb}_1})}{b_1}+</math><math>\frac{\beta \gamma a_{32}\mbox{e}^{{tb}_3}(-1+\mbox{e}^{{tb}_2})}{b_2}+</math><math>\frac{{\gamma }^2a_{33}\mbox{e}^{{tb}_3}(-1+\mbox{e}^{{tb}_3})}{b_3}\mbox{.}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (35)
|}

For second and higher iteration solutions, the parameters have to be set to values.

===3.4. Multi-dimensional Lotka–Volterra Equations===

In this subsection, we will obtain perturbation–iteration solutions of multi-dimensional Lotka–Volterra equations having arbitrary number of competitive species. We seek solutions with first order derivatives in the Taylor series expansion. First, rewrite Eq. [[#e0040|(8)]] as follows:

{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>F_k={\overset{\cdot}{u}}_{k\mbox{,}n+1}-u_{k\mbox{,}n+1}\left(b_k+\right. </math><math>\left. \epsilon \sum_{j=1}^Ka_{kj}u_{j\mbox{,}n+1}\right)=</math><math>0\mbox{.}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (36)
|}

For ''m'' = 1, the terms included in Eq. [[#e0035|(7)]] are as follows:

{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>{F_k}_{\epsilon =0}={\overset{\cdot}{u}}_{k\mbox{,}n}-</math><math>b_ku_{k\mbox{,}n}\mbox{,}{\frac{\partial F_k}{\partial {\overset{\cdot}{u}}_{k\mbox{,}n+1}}}_{\epsilon =0}=</math><math>1\mbox{,}\quad {\frac{\partial F_k}{\partial u_{k\mbox{,}n+1}}}_{\epsilon =0}=</math><math>-b_k\mbox{,}\quad {\frac{\partial F_k}{\partial \epsilon }}_{\epsilon =0}=</math><math>-\sum_{j=1}^Ka_{kj}u_{k\mbox{,}n}u_{j\mbox{,}n}\mbox{.}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (37)
|}

Hence substituting these terms in Eq. [[#e0035|(7)]] yields

<span id='e0190'></span>
{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>{\overset{\cdot}{u}}_{k\mbox{,}n}^c-b_ku_{k\mbox{,}n}^c=</math><math>-{\overset{\cdot}{u}}_{k\mbox{,}n}+b_ku_{k\mbox{,}n}+</math><math>\sum_{j=1}^Ka_{kj}u_{k\mbox{,}n}u_{j\mbox{,}n}\mbox{.}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (38)
|}

Integrating Eq. [[#e0190|(38)]] gives the following:

{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>u_{k\mbox{,}n}^c=e^{b_kt}\left(-\int e^{-b_kt}{\overset{\cdot}{u}}_{k\mbox{,}n}dt+\right. </math><math>\left. \int e^{-b_kt}b_ku_{k\mbox{,}n}+\int e^{-b_kt}\sum_{j=1}^Ka_{kj}u_{k\mbox{,}n}u_{j\mbox{,}n}dt+\right. </math><math>\left. c_n\right)\mbox{,}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (39)
|}

where ''c<sub>n</sub>'' is integration constant to be determined by the initial conditions. Then ''n’'' th iteration solution for multidimensional Lotka–Volterra equations becomes the following:

{| class="formulaSCP" style="width: 100%; text-align: center;" 
|-
| 
{| style="text-align: center; margin:auto;" 
|-
| <math>u_{k\mbox{,}n+1}=u_{k\mbox{,}n}+e^{b_kt}\left(-\int e^{-b_kt}{\overset{\cdot}{u}}_{k\mbox{,}n}dt+\right. </math><math>\left. \int e^{-b_kt}b_ku_{k\mbox{,}n}+\int e^{-b_kt}\sum_{j=1}^Ka_{kj}u_{k\mbox{,}n}u_{j\mbox{,}n}dt+\right. </math><math>\left. c_n\right)\mbox{.}</math>
|}
| style="width: 5px;text-align: right;white-space: nowrap;" | (40)
|}

<span id='s0050'></span>
==4. Numerical results and comparisons==

Comparisons of approximate solutions of Eq. [[#e0045|(9)]] with the exact solutions are presented in [[#t0005|Table 1]]. Solutions up to five iterations are computed using PIA(1,1). Solutions resulted from PIA(1,2) are given in the last three columns of [[#t0005|Table 1]]. It is easy to see that both iteration algorithms rapidly converge to the exact solution. For the PIA(1,2) case, third iteration solutions acquire almost the same accuracy with the fifth iteration solutions for the PIA(1,1) case.

<span id='t0005'></span>

{| class="wikitable" style="min-width: 60%;margin-left: auto; margin-right: auto;"
|+

Table 1.

Comparisons of perturbation–iteration solutions of [[#e0045|(9)]] with exact solution [[#b0080|[16]]] (''a'' = −3, ''b'' = 1 and ''α'' = 0.1).

|-

! colspan="2" | Exact solution
! colspan="5" | PIA(1,1)
! colspan="3" | PIA(1,2)
|-

! ''t''
! ''u''<sub>1</sub>
! ''u''<sub>1,1</sub>
! ''u''<sub>1,2</sub>
! ''u''<sub>1,3</sub>
! ''u''<sub>1,4</sub>
! ''u''<sub>1,5</sub>
! ''u''<sub>1,1</sub>
! ''u''<sub>1,2</sub>
! ''u''<sub>1,3</sub>
|-

| 0.0
| 0.100000
| 0.100000
| 0.100000
| 0.100000
| 0.100000
| 0.100000
| 0.100000
| 0.100000
| 0.100000
|-

| 0.1
| 0.107137
| 0.107030
| 0.107139
| 0.107137
| 0.107137
| 0.107137
| 0.107138
| 0.107137
| 0.107137
|-

| 0.2
| 0.114533
| 0.114028
| 0.114555
| 0.114532
| 0.114533
| 0.114533
| 0.114543
| 0.114533
| 0.114533
|-

| 0.3
| 0.122164
| 0.120818
| 0.122253
| 0.122159
| 0.122164
| 0.122164
| 0.122206
| 0.122164
| 0.122164
|-

| 0.4
| 0.130001
| 0.127171
| 0.130259
| 0.129983
| 0.130002
| 0.130001
| 0.130121
| 0.130002
| 0.130001
|-

| 0.5
| 0.138013
| 0.132785
| 0.138625
| 0.137958
| 0.138017
| 0.138012
| 0.138293
| 0.138016
| 0.138013
|-

| 0.6
| 0.146163
| 0.137272
| 0.147445
| 0.146021
| 0.146175
| 0.146162
| 0.146738
| 0.146174
| 0.146163
|-

| 0.7
| 0.154414
| 0.140132
| 0.156869
| 0.154084
| 0.154449
| 0.154411
| 0.155492
| 0.154445
| 0.154415
|-

| 0.8
| 0.162726
| 0.140729
| 0.167126
| 0.162023
| 0.162814
| 0.162717
| 0.164618
| 0.162802
| 0.162728
|-

| 0.9
| 0.171057
| 0.138259
| 0.178536
| 0.169658
| 0.171261
| 0.171033
| 0.174207
| 0.171228
| 0.171064
|-

| 1.0
| 0.179367
| 0.131705
| 0.191525
| 0.176725
| 0.179808
| 0.179306
| 0.184389
| 0.179721
| 0.179386
|}

Representative values of known numerical and perturbation–iteration solutions of [[#e0115|(23)]] are given in [[#t0010|Table 2]]. Only in two iterations, very high accuracy has been achieved as expected.

<span id='t0010'></span>

{| class="wikitable" style="min-width: 60%;margin-left: auto; margin-right: auto;"
|+

Table 2.

Comparisons of perturbation–iteration solutions of [[#e0115|(23)]] with numerical solutions [[#b0080|[16]]] (''b''<sub>1</sub> = 0.1, ''a''<sub>11</sub> = −0.0014, ''a''<sub>12</sub> = −0.0012, ''b''<sub>2</sub> = 0.08, ''a''<sub>21</sub> = −0.0009, ''a''<sub>22</sub> = −0.001, ''α'' = 4 and ''β'' = 10).

|-

! colspan="3" | Numerical solution
! colspan="4" | PIA(1,1)
|-

! ''t''
! ''u''<sub>1</sub>
! ''u''<sub>2</sub>
! ''u''<sub>1,1</sub>
! ''u''<sub>2,1</sub>
! ''u''<sub>1,2</sub>
! ''u''<sub>2,2</sub>
|-

| 0.0
| 4.000000
| 10.000000
| 4.000000
| 10.000000
| 4.000000
| 10.000000
|-

| 0.1
| 4.033071
| 10.066572
| 4.033059
| 10.066553
| 4.033071
| 10.066573
|-

| 0.2
| 4.066363
| 10.133490
| 4.066316
| 10.133411
| 4.066364
| 10.133490
|-

| 0.3
| 4.099879
| 10.200753
| 4.099771
| 10.200573
| 4.099879
| 10.200754
|-

| 0.4
| 4.133617
| 10.268363
| 4.133422
| 10.268037
| 4.133618
| 10.268365
|-

| 0.5
| 4.167580
| 10.336320
| 4.167269
| 10.335802
| 4.167582
| 10.336322
|-

| 0.6
| 4.201767
| 10.404623
| 4.201312
| 10.403865
| 4.201770
| 10.404627
|-

| 0.7
| 4.236180
| 10.473273
| 4.235549
| 10.472226
| 4.236185
| 10.473281
|-

| 0.8
| 4.270819
| 10.542271
| 4.269980
| 10.540883
| 4.270826
| 10.542282
|-

| 0.9
| 4.305684
| 10.611618
| 4.304604
| 10.609834
| 4.305694
| 10.611634
|-

| 1.0
| 4.340776
| 10.681312
| 4.339420
| 10.679076
| 4.340796
| 10.681334
|}

Comparisons for approximate solutions of three dimensional Lotka–Volterra system [[#e0160|(32)]] are given in [[#t0015|Table 3]]. Satisfactory results on the whole domain are obtained in four or five iterations. The results indicate that solutions obtained with the perturbation–iteration method rapidly converge to the known numerical solutions.

<span id='t0015'></span>

{| class="wikitable" style="min-width: 60%;margin-left: auto; margin-right: auto;"
|+

Table 3.

Comparisons of perturbation–iteration solutions of [[#e0160|(32)]] with numerical solutions [[#b0080|[16]]] (''b''<sub>1</sub> = 1, ''a''<sub>11</sub> = −1, ''a''<sub>12</sub> = −0.1, ''a''<sub>13</sub> = −0.1, ''b''<sub>2</sub> = 1, ''a''<sub>21</sub> = −0.1, ''a''<sub>22</sub> = −1, ''a''<sub>23</sub> = −0.1, ''b''<sub>3</sub> = 1, ''a''<sub>31</sub> = −0.1, ''a''<sub>32</sub> = −0.1, ''a''<sub>33</sub> = −1, ''α'' = 0.2, ''β'' = 0.3, ''γ'' = 0.5).

|-

! colspan="4" | Numerical solutions
! colspan="6" | PIA(1,1)
|-

! ''t''
! ''u''<sub>1</sub>
! ''u''<sub>2</sub>
! ''u''<sub>3</sub>
! ''u''<sub>1,4</sub>
! ''u''<sub>2,4</sub>
! ''u''<sub>3,4</sub>
! ''u''<sub>1,5</sub>
! ''u''<sub>2,5</sub>
! ''u''<sub>3,5</sub>
|-

| 0.0
| 0.20000
| 0.30000
| 0.50000
| 0.20000
| 0.30000
| 0.50000
| 0.20000
| 0.30000
| 0.50000
|-

| 0.1
| 0.21473
| 0.31914
| 0.52234
| 0.21473
| 0.31915
| 0.52234
| 0.21473
| 0.31915
| 0.52234
|-

| 0.2
| 0.23010
| 0.33873
| 0.54428
| 0.23010
| 0.33873
| 0.54429
| 0.23010
| 0.33873
| 0.54428
|-

| 0.3
| 0.24609
| 0.35867
| 0.56572
| 0.24609
| 0.35867
| 0.56573
| 0.24609
| 0.35867
| 0.56572
|-

| 0.4
| 0.26264
| 0.37888
| 0.58655
| 0.26265
| 0.37889
| 0.58662
| 0.26264
| 0.37888
| 0.58654
|-

| 0.5
| 0.27973
| 0.39927
| 0.60667
| 0.27975
| 0.39931
| 0.60693
| 0.27973
| 0.39927
| 0.60664
|-

| 0.6
| 0.29729
| 0.41974
| 0.62601
| 0.29734
| 0.41987
| 0.62679
| 0.29729
| 0.41973
| 0.62591
|-

| 0.7
| 0.31527
| 0.44020
| 0.64449
| 0.31540
| 0.44054
| 0.64654
| 0.31526
| 0.44015
| 0.64420
|-

| 0.8
| 0.33361
| 0.46054
| 0.66208
| 0.33393
| 0.46138
| 0.66683
| 0.33356
| 0.46042
| 0.66128
|-

| 0.9
| 0.35222
| 0.48068
| 0.67871
| 0.35294
| 0.48254
| 0.68887
| 0.35210
| 0.48037
| 0.67675
|-

| 1.0
| 0.37105
| 0.50053
| 0.69438
| 0.37256
| 0.50438
| 0.71455
| 0.37075
| 0.49979
| 0.68991
|}

<span id='s0055'></span>
==5. The mathematica package==

To use the code, first load the ''Mathematica'' package PerturbationIteration.m  [[#b0115|[23]]] using the command

{| class="wikitable" style="min-width: 60%;margin-left: auto; margin-right: auto;"

|-

| In[2]:=Get[“PerturbationIteration.m”];
|}

Proceeding with the two dimensional Lotka–Volterra system [[#e0115|(23)]] as an example, call the function PerturbationIterationAlgorithm (which is part of the package):

{| class="wikitable" style="min-width: 60%;margin-left: auto; margin-right: auto;"

|-

| In[3]: = PerturbationIterationAlgorithm[
|-

| {u′[t] == u[t] (b1 + a11 u[t] + a12 v[t]),
|-

| v′[t] == v[t] (b2 + a21 u[t] + a22 v[t])}, {u[0] == α,
|-

| v[0] == β}, {u, v}, t, 1, 1, 2]
|}

which will produce the solutions in [[#e0155|(31)]] and go one more iteration. Similarly, results given in subsections  [[#s0020|3.1]], [[#s0035|3.2]] and [[#s0040|3.3]] of this paper can be reproduced by using our package.

The usage of the main function PerturbationIterationAlgorithm is as follows:

{| class="wikitable" style="min-width: 60%;margin-left: auto; margin-right: auto;"

|-

| PerturbationIterationAlgorithm[
|-

|  <equations>,
|-

|  <list of initial conditions>,
|-

|  <functions>,
|-

|  <independent variable>,
|-

|  <''n'': number of correction terms (integer) >,
|-

|  <''m'': number of terms included in the Taylor series expansion (integer) >,
|-

|  <number of iterations to be carried out (integer)>,
|-

|  <options>]
|}

where <options> can be Perturbation Parameter → <parameter>, IntroduceEpsilon → <True | False >, or GivenUZero → <list of <math display="inline">u_{j\mbox{,}0}</math> initial guess values>. One should keep in mind that ''n'' has to be smaller than or equal to ''m''.

The default value of PerturbationParameter is ε. If set to another parameter, then this new parameter is used as the perturbation parameter. IntroduceEpsilon option is used for controlling the introduction of ε. If the system has the perturbation parameter already in place, then setting IntroduceEpsilon to False will skip introduction of the perturbation parameter. The default value of GivenUZero is set to Automatic, in which case the function automatically computes <math display="inline">u_{j\mbox{,}0}</math>. If GivenUZero is set to a user defined expression as GivenUZero ->{<math display="inline">u_{1\mbox{,}0}</math>,<math display="inline">u_{2\mbox{,}0}\mbox{,}</math>…} then computations continue using the initial guess values <math display="inline">u_{1\mbox{,}0}</math>,<math display="inline">u_{2\mbox{,}0}\mbox{,}</math>….

Further documentation and more examples about the package can be found in [[#b0115|[23]]].

<span id='s0060'></span>
==6. Concluding remarks==

In this study, the perturbation–iteration method [[#b0060|[12]]] and [[#b0065|[13]]] is extended to the systems of first order differential equations and applied to the Lotka–Volterra equations. Contrary to the classical perturbation methods, small parameter assumption is not needed for the new perturbation–iteration method. In the study of one dimensional Lotka–Volterra equations, solutions arising from PIA(1,2) have better accuracy in comparison with those of PIA(1,1). Hence increasing the number of terms in Taylor series expansion results in a higher accuracy as in [[#b0120|[24]]].

Also solutions of general multi-dimensional Lotka–Volterra equations without any small parameter are obtained using PIA(1,1). It is easy to see that these results cover the one-two and three dimensional cases.

==References==

<ol style='list-style-type: none;margin-left: 0px;'><li><span id='b0005'></span>
[[#b0005|[1]]] A.H. Nayfeh; Introduction to Perturbation Techniques; Wiley and Sons, New York (1981)</li>
<li><span id='b0010'></span>
[[#b0010|[2]]] J.H. He; Linearized perturbation technique and its applications to strongly nonlinear oscillators; Comput. Math. Appl., 45 (2003), pp. 1–8</li>
<li><span id='b0015'></span>
[[#b0015|[3]]] H. Hu; A classical perturbation technique which is valid for large parameters; J. Sound Vib., 269 (2004), pp. 409–412</li>
<li><span id='b0020'></span>
[[#b0020|[4]]] M. Pakdemirli, M.M.F. Karahan, H. Boyacı; A new perturbation algorithm with better convergence properties: multiple scales Lindstedt Poincaré method; Math. Comput. Appl., 14 (2009), pp. 31–44</li>
<li><span id='b0025'></span>
[[#b0025|[5]]] L. Xu; Determination of limit cycle by He’ s parameter expanding method for strongly nonlinear oscillators; J. Sound Vib., 302 (2007), pp. 178–184</li>
<li><span id='b0030'></span>
[[#b0030|[6]]] A.M. Wazwaz; Adomian decomposition method for a reliable treatment of the Bratu-type equations; Appl. Math. Comput., 166 (2005), pp. 652–663</li>
<li><span id='b0035'></span>
[[#b0035|[7]]] J.H. He; Variational iteration method — a kind of non-linear analytical technique: some examples; Int. J. Non-Linear Mech., 34 (1999), pp. 699–708</li>
<li><span id='b0040'></span>
[[#b0040|[8]]] S.J. Liao; On the homotopy analysis method for nonlinear problems; Appl. Math. Comput., 147 (2004), pp. 499–513</li>
<li><span id='b0045'></span>
[[#b0045|[9]]] J.H. He; Some asymptotic methods for strongly nonlinear equations; Int. J. Mod. Phys. B, 20 (2006), pp. 1141–1199</li>
<li><span id='b0050'></span>
[[#b0050|[10]]] R.E. Mickens; A generalized iteration procedure for calculating approximations to periodic solutions of ‘‘truly nonlinear oscillators’’; J. Sound Vib., 287 (2005), pp. 1045–1051</li>
<li><span id='b0055'></span>
[[#b0055|[11]]] H. Tari; Modified variational iteration method; Phys. Lett. A, 369 (2007), pp. 290–293</li>
<li><span id='b0060'></span>
[[#b0060|[12]]] Y. Aksoy, M. Pakdemirli; New perturbation–iteration solutions for Bratu-type equations; Comput. Math. Appl., 59 (2010), pp. 2802–2808</li>
<li><span id='b0065'></span>
[[#b0065|[13]]] M. Pakdemirli, Y. Aksoy, H. Boyacı; A new perturbation iteration approach for first order differential equations; Math. Comput. Appl., 16 (2011), pp. 890–899</li>
<li><span id='b0070'></span>
[[#b0070|[14]]] Y. Aksoy, M. Pakdemirli, S. Abbasbandy, H. Boyacı; New perturbation–iteration solutions for nonlinear heat transfer equations; Int. J. Numer. Meth. Heat Fluid Flow, 22 (2012), pp. 814–828</li>
<li><span id='b0075'></span>
[[#b0075|[15]]] S. Olek; An accurate solution to the multispecies Lotka–Volterra equations; Soc. Ind. Appl. Math., 36 (3) (1994), pp. 480–488</li>
<li><span id='b0080'></span>
[[#b0080|[16]]] B. Batiha, M.S.M. Noorani, I. Hashim; Variational iteration method for solving multispecies Lotka-Volterra equations; Comput. Math. Appl., 54 (2007), pp. 903–909</li>
<li><span id='b0085'></span>
[[#b0085|[17]]] A.S. Bataineh, M.S.M. Noorani, I. Hashim; Series solution of the multispecies Lotka–Volterra equations by means of the homotopy analysis method; Diff. Equat. Nonlinear Mech., 2008 (2008), Article 816787 [http://dx.doi.org/10.1155/2008/816787 http://dx.doi.org/10.1155/2008/816787]</li>
<li><span id='b0090'></span>
[[#b0090|[18]]] Vasile Marinca, Nicolae Herişanu; Optimal parametric iteration method for solving multispecies Lotka–Volterra Equations; Discrete Dyn. Nat. Soc., 2012 (2012), Article 842121 [http://dx.doi.org/10.1155/2012/842121 http://dx.doi.org/10.1155/2012/842121]</li>
<li><span id='b0095'></span>
[[#b0095|[19]]] P.H. Verhulst; Notice sur la loi que la population suit dans son accroissement; Corresp. Mathématique et Physique, 10 (1838), pp. 113–121</li>
<li><span id='b0100'></span>
[[#b0100|[20]]] V. Volterra; Variazioni e fluttuazioni del numero d’individui in specie animali conviventi; Memorie della Reale Accademia dei Lincei, 2 (1926), pp. 31–113</li>
<li><span id='b0105'></span>
[[#b0105|[21]]] A.J. Lotka; Elements of Physical Biology; Williams & Wilkins, Baltimore (1925)</li>
<li><span id='b0110'></span>
[[#b0110|[22]]] R.M. May, W.J. Leonard; Nonlinear aspects of competition between three species; SIAM J. Appl. Math., 29 (1975), pp. 243–253</li>
<li><span id='b0115'></span>
[[#b0115|[23]]] Ü. Göktaş, 2011. Software is available on <http://www.turgutozal.edu.tr/ugoktas/>.</li>
<li><span id='b0120'></span>
[[#b0120|[24]]] Ü. Göktaş, M. Pakdemirli and Y. Aksoy, Symbolic Computation of Perturbation–Iteration Solutions for Differential Equations, in: Proc. 2nd International Symposium on Computing in Science and Engineering (ISCSE 2011), June 1–4, 2011, Kuşadası, Aydın, Turkey, 60–66, ed. M. Güneş, İzmir (Turkey), Gediz University Publications, 2011.</li>
</ol>