^{1}

^{*}

^{2}

Methods for the approximation of solution of nonlinear system of equations often fail when the Jacobian s of the systems are singular at iteration points. In this paper, multi-step families of quadrature based iterative methods for approximating the solution of nonlinear system of equations with singular Jacobian are developed using decomposition technique. The methods proposed in this study are of convergence order , and require only the evaluation of first-order Frechet derivative per iteration. The approximate solutions generated by the proposed iterative methods in this paper compared with some existing contemporary methods in literature, show that methods developed herein are efficient and adequate in approximating the solution of nonlinear system of equations whose Jacobians are singular and non-singular at iteration points.

System of equations that is used in describing real life phenomena is often nonlinear in nature. Examples of Mathematical models that are formulated using nonlinear system of equations (NLSE) include mathematical models that describe kinematics, combustion, chemical equilibrium, economic problem and neurophysiology problems, [

Consider the NLSE,

G ( X ) = 0 (1)

where X ∈ ℜ m , 0 is a null vector of dimension m, G : D ⊂ ℜ m → ℜ m is functional define by

G ( X 1 , X 2 , ⋯ , X m ) = [ G i ( X 1 , X 2 , ⋯ , X m ) ] T ,

G i , i = 1 , 2 , ⋯ , m are coordinate functions of G and D is an open domain in ℜ m .

The Newton method in m-dimension is a popular iterative method for approximating the solution of NLSE (1). The sequence of approximations { X k } k ≥ 1 generated using the Newton method, converges to the solution Φ of the NLSE (1) with convergence order ρ = 2 , under the condition that det ( G ′ ( X k ) ) ≠ 0 , [

The development of new iterative methods for approximating the solution of (1) has attracted the attention of researchers in recent years, as evident in the literature such as [

Quadrature formulas are veritable tools for the evaluation of the integrals, [

Let β be the initial approximation close to Φ a solution of the NLSE (1.1.1) and Ω ( X ) a function such that

Ω ( X ) = ψ ( X ) ⊙ G ( X ) = 0 (2)

where ψ ( X ) is a differentiable nonzero scalar function. The notation ⊙ is component-wise element operator such that

ψ ( X ) ⊙ G ( X ) = [ ψ ( X ) G ( X 1 ) , ψ ( X ) G ( X 2 ) , ⋯ , ψ ( X ) G ( X m ) ] T (3)

The solution of Ω ( X ) = 0 and G ( X ) = 0 are same because ψ ( X ) ≠ 0 for all values of X.

With the aid of Taylor series expansion of a multi-dimensional function about β up to the second term and using the generic quadrature formula to approximate G ′ ( β ) , then (2) can be rewritten as

G ( β ) + [ G ( β ) ( ∇ ψ ( X ) T ψ ( X ) ) + ∑ i = 1 q μ i G ′ ( β + θ i ( X − β ) ) ] ( X − β ) + H ( X ) = 0 (4)

where H ( X ) is higher order terms of the Taylor expansion, the division operator in ( ∇ ψ ( X ) T ψ ( X ) ) is element wise, θ i and μ i , i = 1 , 2 , ⋯ , q are knots and weights respectively such that

∑ i = 1 q μ i = 1 , ∑ i = 1 q μ i θ i = 1 2 , (5)

Equation (5) is consistency conditions, [

The (4) is expressed into coupled equation given in (6) and (7).

G ( β ) + [ G ( β ) ( ∇ ψ ( X ) T ψ ( X ) ) + ∑ i = 1 q μ i G ′ ( β + θ i ( X − β ) ) ] ( X − β ) + H ( X ) = 0 (6)

H ( X ) = G ( X ) + G ( β ) − [ G ( β ) ( ∇ p ( X ) T p ( X ) ) + ∑ i = 1 q μ i G ′ ( β + θ i ( X − β ) ) ] ( X − β ) (7)

In compact form, (6) can be expressed as

X = β + ℏ ( X ) (8)

where

ℏ ( X ) = − [ G ( β ) ( ∇ ψ ( X ) T ψ ( X ) ) + ∑ i = 1 q μ i G ′ ( β + θ i ( X − β ) ) ] − 1 ( G ( β ) + H ( X ) ) (9)

is a nonlinear function.

Applying the decomposition technique due to [

ℏ ( X ) = ℏ ( X 0 ) + ∑ i = 1 ∞ [ ℏ ( ∑ j = 0 i X j ) − ℏ ( ∑ j = 0 i − 1 X j ) ] (10)

where X 0 is initial guess.

The idea here is to find the solution vector X of the NLSE (1) in series form through an iterative scheme, such that the solution X is the sum of the initial guess β and the sum of consecutive differences of successive and preceding iterate points approximations of X, that is;

X = ∑ i = 0 ∞ X j = β + ℏ ( X 0 ) + ∑ i = 1 ∞ [ ℏ ( ∑ j = 0 i X j ) − ℏ ( ∑ j = 0 i − 1 X j ) ] (11)

Hence the following scheme from (11) is obtained:

X 0 = β X 1 = ℏ ( X 0 ) ⋮ X s + 1 = ℏ ( ∑ j = 0 i X j ) − ℏ ( ∑ j = 0 i − 1 X j ) , i = 1 , 2 , ⋯ (12)

The sum of the respective sides of (12) is

∑ i = 0 s + 1 X i = β + ℏ ( ∑ i = 0 s X i ) , i = 1 , 2 , ⋯ (13)

From (2) and (12), the solution X of the (1) is approximated as:

X ≈ β + ∑ i = 0 s X i = β + ℏ ( ∑ i = 0 s − 1 X i ) , i = 1 , 2 , ⋯ (14)

As s becomes large, the approximations of the solution X gets closer to the exact solution of (1).

From Equation (12)

X 0 = β (15)

Since X 0 is initial guess, setting X 0 in (7) yields

H ( X 0 ) = 0 (16)

From (9)and (15),

X 1 = ℏ ( X 0 ) = − [ G ( β ) ( ∇ ψ ( X ) T ψ ( X ) ) + ( ∑ i = 1 q μ i ) G ′ ( β ) ] − 1 G ( β ) (17)

For s = 1 in (14), and using (17), the following is obtained.

X ≈ X 0 + X 1 ≈ β + ℏ ( X 0 ) ≈ β − [ G ( β ) ( ∇ ψ ( X ) T ψ ( X ) ) + G ′ ( β ) ] − 1 G ( β ) (18)

Using the formulation in (18), a One-step family of iterative scheme for approximating the solution of (1) is proposed as in Scheme 1.

Scheme 1 Assume X 0 is an initial guess, approximate the solution Φ of (1.1.1) using the iterative scheme:

X k + 1 = X k − [ G ( X k ) ( ∇ ψ ( X k ) T ψ ( X k ) ) + G ′ ( X k ) ] − 1 G ( X k ) , k = 0 , 1 , 2 , ⋯ (19)

Scheme 1 is a family of One-step iterative scheme that can be used to propose iterative methods for solving (1) for some function Φ ( X k ) .

For s = 2 in (14), the solution X of (1) can be approximated as:

X ≈ X 0 + X 1 + X 2 ≈ X 0 + ℏ ( X 0 + X 1 ) ≈ β − [ G ( β ) ( ∇ ψ ( X ) T ψ ( X ) ) + ∑ i = 1 q μ i G ′ ( β + θ i ( X 0 + X 1 − β ) ) ] − 1 × G ( β + H ( X 0 + X 1 ) ) (20)

Set X = X 0 + X 1 in (7) implies

H ( X 0 + X 1 ) = G ( X 0 + X 1 ) − G ( β ) − [ G ( β ) ( ∇ ψ ( X ) T ψ ( X ) ) + ∑ i = 1 q μ i G ′ ( β + θ i ( X 0 + X 1 − β ) ) ] − 1 ( X 0 + X 1 − β ) (21)

From (18),

X 0 + X 1 − β = − [ G ( β ) ( ∇ ψ ( X ) T ψ ( X ) ) + G ′ ( β ) ] − 1 G ( β ) (22)

substituting (22) into (21) yields

H ( X 0 + X 1 ) = G ( X 0 + X 1 ) − G ( β ) − [ G ( β ) ( ∇ ψ ( X ) T ψ ( X ) ) + ∑ i = 1 q μ i G ′ ( β + θ i ( X 0 + X 1 − β ) ) ] − 1 × [ G ( β ) ( ∇ ψ ( X ) T ψ ( X ) ) + G ′ ( β ) ] − 1 G ( β ) (23)

Inserting (23) into (20), gives the equation

X ≈ X 0 + X 1 + X 0 ≈ β − [ G ( β ) ( ∇ ψ ( X ) T ψ ( X ) ) + G ′ ( β ) ] − 1 G ( β ) − [ G ( β ) ( ∇ ψ ( X ) T ψ ( X ) ) + ∑ i = 1 q μ i G ′ ( β + θ i ( X 0 + X 1 − β ) ) ] − 1 G ( X 0 + X 1 ) (24)

Using (24), a Two-step iterative scheme for the approximation of solution Φ of (1) as given in Scheme 2 is proposed.

Scheme 2 Assume X 0 is an initial guess, approximate the solution Φ of (1) using the iterative scheme:

ν k = X k − [ G ( X k ) ( ∇ ψ ( X k ) T ψ ( X k ) ) + G ′ ( X k ) ] − 1 G ( X k ) X k + 1 = ν k − [ G ( X k ) ( ∇ ψ ( X k ) T ψ ( X k ) ) + ∑ i = 1 q μ i G ′ ( X k + θ i ( ν k − X k ) ) ] − 1 G ( ν k ) , k = 0 , 1 , 2 , ⋯ (25)

Scheme 2 is used to propose Two-step iterative methods for approximating the solution Φ of the NLSE (1).

For s = 3 in (14), the solution of (1) can be approximated as follows:

X ≈ X 0 + X 1 + X 2 + X 3 ≈ X 0 + ℏ ( X 0 + X 1 + X 2 ) ≈ β − [ G ( β ) ( ∇ ψ ( X ) T ψ ( X ) ) + ∑ i = 1 q μ i G ′ ( β + θ i ( X 0 + X 1 + X 2 − β ) ) ] − 1 × ( G ( β ) + H ( X 0 + X 1 + X 2 ) ) (26)

Set X = X 0 + X 1 + X 2 in (7) yields

H ( X 0 + X 1 + X 2 ) = G ( X 0 + X 1 + X 2 ) − G ( β ) − [ G ( β ) ( ∇ ψ ( X ) T ψ ( X ) ) + ∑ i = 1 q μ i G ′ ( β + θ i ( X 0 + X 1 + X 2 − β ) ) ] ( X 0 + X 1 + X 2 − β ) (27)

From (26), (28) is obtained.

X 0 + X 1 + X 2 − β = − [ G ( β ) ( ∇ ψ ( X ) T ψ ( X ) ) + ∑ i = 1 q μ i G ′ ( β + θ i ( X 0 + X 1 − β ) ) ] − 1 G ( X 0 + X 1 ) − [ G ( β ) ( ∇ ψ ( X ) T ψ ( X ) ) + G ′ ( β ) ] − 1 G ( β ) (28)

Substituting (28) into (27) yields

H ( X 0 + X 1 + X 2 ) = G ( X 0 + X 1 + X 2 ) − G ( β ) − [ G ( β ) ( ∇ ψ ( X ) T ψ ( X ) ) + ∑ i = 1 q μ i G ′ ( β + θ i ( X 0 + X 1 + X 2 − β ) ) ] × [ ( G ( β ) ( ∇ ψ ( X ) T ψ ( X ) ) + ∑ i = 1 q μ i G ′ ( β + θ i ( X 0 + X 1 + X 2 − β ) ) ] − 1 × G ( X 0 + X 1 + X 2 ) − [ G ( β ) ( ∇ ψ ( X ) T ψ ( X ) ) + G ′ ( β ) ] − 1 G ( β ) ) (29)

Substitute (29) in (26)

X ≈ X 0 + X 1 + X 2 + X 3 ≈ X 0 + ℏ ( X 0 + X 1 + X 2 ) ≈ β − [ G ( β ) ( ∇ ψ ( X ) T ψ ( X ) ) + ∑ i = 1 q μ i G ′ ( β + θ i ( X 0 + X 1 + X 2 − β ) ) ] − 1 × G ( X 0 + X 1 + X 2 ) − [ G ( β ) ( ∇ ψ ( X ) T ψ ( X ) ) + ∑ i = 1 q μ i G ′ ( β + θ i ( X 0 + X 1 − β ) ) ] − 1 × G ( X 0 + X 1 ) − [ G ( β ) ( ∇ ψ ( X ) T ψ ( X ) ) + G ′ ( β ) ] − 1 G ( β ) (30)

The formulation in (30) enable the proposal of the three-step iterative scheme for the solution of (1).

Scheme 3 Assume X 0 is an initial guess, approximate the solution Φ of (1) using the iterative scheme:

ν k = X k − [ G ( X k ) ( ∇ ψ ( X k ) T ψ ( X ) ) + G ′ ( X k ) ] − 1 G ( X k ) W k = ν k − [ G ( X k ) ( ∇ ψ ( X k ) T ψ ( X k ) ) + ∑ i = 1 q μ i G ′ ( X k + θ i ( ν k − X k ) ) ] − 1 G ( ν k ) , X k + 1 = W k − [ G ( X k ) ( ∇ ψ ( X k ) T ψ ( X k ) ) + ∑ i = 1 q μ i G ′ ( X k + θ i ( W k − X k ) ) ] − 1 G ( W k ) , k = 0 , 1 , 2 , ⋯ (31)

Scheme 3 is used to propose Three-step iterative methods for approximating the solution Φ of the NLSE (1).

A suitable choice of the function ψ ( X k ) in the proposed Scheme 1, Scheme 2 and Scheme 3 yields families of quadrature based iterative methods for approximation of the solution Φ of (1). It is worthy of note that for ψ ( X k ) = 1 , Scheme 1 reduces to the classical Newton method, while Scheme 2 and Scheme 3 reduces to the family of approximation methods in [

∇ ψ ( X k ) T ψ ( X k ) = − ∇ ψ ( X k ) T (32)

From (32), a generalization can be made as

∇ ψ ( X k ) T = λ ( X k ) (33)

where λ ( X k ) = [ λ 1 ( X k ) , λ 2 ( X k ) , ⋯ , λ m ( X k ) ] T . Consequently, the following iterative algorithms are obtained from Scheme 1, Scheme 2 and Scheme 3 respectively.

Algorithm 1 Assume X 0 is an initial guess, approximate the solution Φ of (1) using the iterative method:

X k + 1 = X k − [ G ′ ( X k ) − G ( X k ) λ ( X k ) ] − 1 G ( X k ) , k = 0 , 1 , 2 , ⋯ (34)

If the parameter λ ( X k ) = 0 , Algorithm 1 reduces to the m-dimensional classical Newton method (1). The major difference between Algorithm 1 and the Wu method in [

Algorithm 2 Assume X 0 is an initial guess, approximate the solution Φ of (1) using the iterative method:

ν k = X k − [ G ′ ( X k ) − G ( X k ) λ ( X k ) ] − 1 G ( X k ) , X k + 1 = ν k − [ ∑ i = 1 q μ i G ′ ( X k + θ i ( ν k − X k ) ) − G ( X k ) λ ( X k ) ] − 1 G ( ν k ) , k = 0 , 1 , 2 , ⋯ (35)

The Algorithm 2 is a Two-step family of iterative method for approximating the solution of (1).

Algorithm 3 Assume X 0 is an initial guess, approximate the solution Φ of (1) using the iterative method:

ν k = X k − [ G ′ ( X k ) − G ( X k ) λ ( X k ) ] − 1 G ( X k ) , W k = ν k − [ ∑ i = 1 q μ i G ′ ( X k + θ i ( ν k − X k ) ) − G ( X k ) λ ( X k ) ] − 1 G ( ν k ) , X k + 1 = W k − [ ∑ i = 1 q μ i G ′ ( X k + θ i ( W k − X k ) ) − G ( X k ) λ ( X k ) ] − 1 G ( W k ) , k = 0 , 1 , 2 , ⋯ (36)

Remark 1

For numerical implementation, the choice of λ ( X k ) is subjectively chosen, however specific values of λ (for reference purpose, λ ( X k ) is denoted as λ ) are used such that the magnitude of their elements is less one in order to achieve better convergence rate and accuracy. Similarly the choice of θ i is also subjective but must satisfy the consistency condition in (5).

In this section, the convergence of the iterative methods (Algorithm 1, Algorithm 2 and Algorithm 3) are established using the Taylor series approach, [

To establish the convergence of Algorithm 1, the proof of Theorem 1 is considered.

Theorem 1 Suppose the function G : D ⊂ ℜ m → ℜ m is continuous and differentiable in some neighborhood D ⊂ ℜ m of Φ . If X 0 is an initial guess in the neighborhood of Φ , then the sequence of approximations { X k } k ≥ 0 , ( X k ∈ D ) generated by (34) converges to Φ with convergence order ρ = 2 .

Proof. Let E k = ‖ X k − Φ ‖ ∞ be the error in the kth iteration point. Using the Taylor series expansion of G ( X ) and G ′ ( X ) about Φ , the following equations are obtained.

G ( X ) = G ( Φ ) + G ′ ( Φ ) ( X − Φ ) + 1 2 ! G ″ ( Φ ) ( X − Φ ) 2 + 1 2 ! G ‴ ( Φ ) ( X − Φ ) 3 + ⋯ (37)

G ′ ( X ) = G ′ ( Φ ) + G ″ ( Φ ) ( X − Φ ) + 1 2 ! G ‴ ( Φ ) ( X − Φ ) 2 + 1 3 ! G ‴ ' ( Φ ) ( X − Φ ) 3 + ⋯ (38)

Setting X = X k in (37) and (38), implies

G ( X k ) = G ( Φ + E k ) = G ′ ( Φ ) [ E k + ∑ n = 2 4 C n E k n + O ( E k 5 ) ] , k = 0 , 1 , 2 , ⋯ (39)

G ′ ( X k ) = G ′ ( Φ + E k ) = G ′ ( Φ ) [ I + ∑ n = 2 5 C n E k n − 1 + O ( E k 5 ) ] , k = 0 , 1 , 2 , ⋯ (40)

where I is an m × m identity matrix and C n = 1 n ! ‖ ( G ′ ( Φ ) ) − 1 G n ( Φ ) ‖ ∞ , n ≥ 2 .

Using (39) and (40)

[ G ′ ( X k ) − G ( X k ) λ ] − 1 = ( G ′ ( Φ ) ) − 1 [ I + ( − 2 C 2 + λ ) E k + ( 4 C 2 2 − 3 C 3 − 3 C 2 λ + λ 2 ) E k 2 − ( − 8 C 2 3 + 12 C 2 C 3 + 4 C 4 + 8 C 2 2 − 5 C 3 λ + 4 C 2 λ 2 + λ 3 ) E k 3 + ( 16 C 2 4 + 9 C 3 2 − 5 C 5 − 20 C 2 3 − 20 C 2 3 λ − 7 C 4 λ − 7 C 3 λ 2 + λ 4 + C 2 ( 13 λ 2 − 36 C 3 ) + C 2 ( 16 C 4 + 26 C 3 λ − 5 λ 3 ) E k 4 + O ( E k 5 ) ] (41)

multiply (41) and (39), yields

[ G ′ ( X k ) − G ( X k ) λ ] − 1 G ( X k ) = E k + ( λ − C 2 ) E k 2 + ( 2 C 2 2 − 2 C 3 − 2 C 2 λ + λ 2 ) E k 3 − ( − 4 C 2 3 + 7 C 2 C 3 + 4 C 4 − 3 C 4 + 5 C 2 2 − 4 C 3 λ − 3 C 2 λ 2 + λ 3 ) E k 4 + O ( E k 5 ) (42)

substituting (42) in (34), the following equation is obtained.

X k + 1 = Φ + ( λ − C 2 ) E k 2 + ( − 2 C 2 2 + 2 C 3 + 2 C 2 λ − λ 2 ) E k 3 − ( − 4 C 2 3 + 7 C 2 C 3 + 4 C 4 − 3 C 4 + 5 C 2 2 − 4 C 3 λ − 3 C 2 λ 2 + λ 3 ) E k 4 + O ( E k 5 ) (43)

The Equation (45) implies that the sequence of approximations generated by the iterative method (34) converges to the solution Φ of (1) with convergence order ρ = 2 .

Similar to the proof of Theorem 1, the convergence of Algorithm 2 is established in the proof of Theorem 2.

Theorem 2 Suppose the function G : D ⊂ ℜ m → ℜ m is continuously differentiable in some neighborhood D ⊂ ℜ m of Φ . If X 0 is an initial guess in the neighborhood of Φ , then for λ the sequence of approximations { X k } k ≥ 0 , ( X k ∈ D ) generated by (35) converges to Φ with convergence order ρ = 3 .

Proof. From Equation (35), ν k is defined. Setting X = ν k in (37) lead to obtaining the following equation.

G ( ν k ) = G ′ ( Φ ) [ ( C 2 + λ ) E k 2 + ( − 2 C 2 2 + 2 C 3 + 2 C 2 λ − λ 2 ) E k 3 + O ( E k 4 ) ] (44)

Similarly, set X = X k + θ i ( ν k − X k ) in (38), then

∑ i = 1 q μ i G ′ ( X k + θ i ( ν k − X k ) ) = G ′ ( Φ ) [ I + 2 C 2 ( ∑ i = 1 q μ i ( 1 − θ i ) E k + ( ∑ i = 1 q μ i ( 3 C 3 ( θ i − 1 ) 2 ) + 2 C 2 ( C 2 − λ ) θ i ) E k 2 + ( ∑ i = 1 q μ i ( 4 C 3 ( θ i − 1 ) 3 + 2 C 2 ( − 2 C 2 2 + 2 C 3 + 2 C 2 λ − λ 2 ) θ i + 6 C 32 − λ ) ( 1 − θ i ) ) E k 3 + ( ∑ i = 1 q μ i ( 2 C 2 ( 4 C 2 3 − 7 C 2 C 3 + 3 C 4 − 5 C 2 2 λ + 4 C 3 λ + 3 C 2 λ 2 − λ 3 ) θ i ) + 12 C 4 ( C 2 − λ ) ( θ i − 1 ) 2 θ i + 3 C 3 ( 2 ( − 2 C 2 2 + 2 C 3 + 2 C 2 λ − λ 2 ) ( 1 − θ i ) θ i + ( C 2 − λ ) 2 ) θ i E k 4 + O ( E k 5 ) ] (45)

Using (39) and (45);

∑ i = 1 q μ i G ′ ( X k + θ i ( ν k − X k ) ) − G ( X k ) λ = G ′ ( Φ ) [ I + ( ∑ i = 1 q μ i ( − λ − 2 C 2 μ i ( θ i − 1 ) ) ) E k

+ ( − C 2 λ + ∑ i = 1 q μ i ( 3 C 3 ( θ i − 1 ) 2 ) + 2 C 2 ( C 2 − λ ) θ i ) E k 2 + ( − C 3 λ + ∑ i = 1 q μ i ( 4 C 4 ( θ i − 1 ) 3 ) − 2 C 2 ( 2 C 2 2 − 2 C 3 − 2 C 2 λ + λ 2 ) θ i − 6 C 3 ( C 2 − λ ) ( 1 − θ i ) θ i ) E k 3

+ ( − C 4 λ + ∑ i = 1 q μ i ( 2 C 2 ( 4 C 2 3 − 7 C 2 C 3 + 3 C 4 − 5 C 2 2 λ + 4 C 3 λ + 3 C 2 λ 2 − λ 3 ) + 12 C 4 ( C 2 − λ ) ( θ i − 1 ) 2 + 3 C 3 ( 22 C 2 2 − 2 C 3 − 2 C 2 λ + λ 2 ) ( 1 − θ i ) + ( C 2 − λ ) 2 θ i ) E k 4 + O ( E k 4 ) ] (46)

From (46) and (44),

[ ∑ i = 1 q μ i G ′ ( X k + θ i ( ν k − X k ) ) − G ( X k ) λ ] − 1 G ( ν k ) = ( C 2 − λ ) E k 2 + ( ( − 2 C 2 2 − 2 C 2 − 2 C 2 λ + λ 2 ) + ( C 2 − λ ) ( λ + ∑ i = 1 q μ i C 2 ( θ i − 1 ) ) ) E k 3 + ( ( 5 C 2 3 − 7 C 2 C 3 + 3 C 4 − 7 C 2 2 λ + 4 C 3 λ + 4 C 2 λ 2 − λ 3 ) + ( − 2 C 2 2 λ + 2 C 3 + 2 C 2 λ − λ 2 ) ( λ + ∑ i = 1 q μ i C 2 ( θ i − 1 ) ) + ( C 2 − λ ) ( λ + 2 ∑ i = 1 q μ i C 2 ( θ i − 1 ) ) 2 − ( λ C 3 + 3 ∑ i = 1 q μ i C 2 ( θ i − 1 ) 2 + 2 C 2 ( C 2 − λ ) θ i ) E k 4 + O ( E k 4 ) (47)

Using (47) in the second step of (35), with the expansion of ν k as in (43), the following equation is obtained.

X k + 1 = Φ + ( ( C 2 − λ ) ( λ + 2 ∑ i = 1 q μ i C 2 ( θ i − 1 ) ) ) E k 3 + O ( E k 4 ) (48)

Equation (48) implies that the sequence of approximations generated by the iterative method (35) converges to Φ with convergence order ρ = 3 .

The convergence of the propose Algorithm 3 is established by the proof of Theorem 3.

Theorem 3 Suppose the function G : ℜ m → ℜ m is continuously differentiable in some neighborhood D ∈ ℜ m of its solution Φ . If X 0 is an initial guess in the neighborhood of Φ , then the sequence of approximations { X k } k ≥ 0 , ( X k ∈ D ) generated by (36) converges to Φ with convergence order ρ = 4 .

Proof. Set X = W k and X = X k + θ i ( W k − X k ) in (37) and (38) respectively, where W k is the second step of (35) then,

G ( W k ) = G ′ ( Φ ) [ ( ( C 2 − λ ) ( λ + ∑ i = 1 q μ i C 2 ( θ i − 1 ) ) ) E k 3 + ( ( − C 2 3 + 2 C 2 2 − C 2 λ 2 + λ 3 ) + ( − 2 C 2 2 λ + 2 C 3 + 2 C 2 λ − λ 2 ) ( λ + ∑ i = 1 q μ i C 2 ( θ i − 1 ) ) + ( C 2 − λ ) ( λ + 2 ∑ i = 1 q μ i C 2 ( θ i − 1 ) ) 2 − ( λ C 2 + 3 ∑ i = 1 q μ i C 3 ( θ i − 1 ) 2 + 2 C 2 ( C 2 − λ ) θ i ) E k 4 + O ( E k 5 ) ] (49)

and

∑ i = 1 q μ i G ′ ( X k + θ i ( W k − X k ) ) = G ′ ( Φ ) [ I + C 2 E k + ( 3 C 3 ∑ i = 1 q μ i θ i 2 ) E k 2 + ( 4 C 4 ∑ i = 1 q μ i ( 1 − θ i ) 3 − 2 C 2 ( C 2 − λ ) ( λ + 2 C 2 ( ∑ i = 1 q μ i θ i 2 − 1 2 ) ) ) E k 3 + O ( E k 4 ) ] (50)

From (50)

[ ∑ i = 1 q μ i G ′ ( X k + θ i ( W k − X k ) ) + G ( X k ) λ ] − 1 = ( G ′ ( Φ ) ) − 1 [ I − ( λ − C 2 ) E k + ( λ 2 + 4 C 2 2 ( ∑ i = 1 q μ i θ i 2 − 1 2 ) − 3 C 3 ( ∑ i = 1 q μ i θ i 2 − 1 2 ) − C 2 λ ) E k 2 + ( 4 C 4 ( ∑ i = 1 q μ i ( θ i − 1 ) 3 ) + 2 C 2 2 ( 1 2 + 4 ∑ i = 1 q μ i θ i 2 ) ) + 2 C 3 − 2 ( − 1 2 − 5 ∑ i = 1 q μ i θ i 2 + 2 ∑ i = 1 q μ i θ i 3 ) + λ ( λ 2 + C 3 ( 1 − 6 ∑ i = 1 q μ i θ i 2 ) ) E k 3 + O ( E k 4 ) ] (51)

By multiplying (51) by (49), the following equation is obtained.

[ ∑ i = 1 q μ i G ′ ( X k + θ i ( W k − X k ) ) + G ( X k ) λ ] − 1 G ( W k ) = ( C 2 − λ ) 2 E k 3 + ( C 2 3 ( − 2 − 8 ∑ i = 1 q μ i θ i 2 ) + C 2 2 λ ( 7 + 8 ∑ i = 1 q μ i θ i 2 ) + λ ( 3 λ 2 + C 2 ( − 2 − 3 ∑ i = 1 q μ i θ i 2 ) ) + C 2 ( − 8 λ 2 − C 3 ( 2 + 3 ∑ i = 1 q μ i θ i 2 ) ) ) E k 4 + O ( E k 5 ) (52)

Using (48) and (50) in the third step of (36), yields

X k = Φ + ( C 2 − λ ) ( C 2 − λ ) 2 E k 4 + O ( E k 5 ) (53)

Equation (53) implies that the sequence of approximations generated by the iterative method (36) converges to the solution Φ of the (1) with convergence order ρ = 4 .

Here, some particular forms of the iterative methods in Algorithm 2 and Algorithm 3 are developed by assigning arbitrary values to the parameters μ i and θ i , i = 1 , 2 , ⋯ satisfying the conditions given in (5).

For q = 1 , μ 1 = 1 , θ 1 = 1 2 , in Algorithm 2 give rise to the following iterative method for approximating Φ of (1).

Algorithm 4 Assume X 0 is an initial guess, approximate the solution Φ of (1) using the iterative method:

ν k = X k − [ G ′ ( X k ) − G ( X k ) λ ] − 1 G ( X k ) , X k + 1 = ν k − [ G ′ ( X k + ν k 2 ) − G ( X k ) λ ] − 1 G ( ν k ) , k = 0 , 1 , 2 , ⋯ (54)

Algorithm 4 is an iterative method for approximating the solution Φ of (1) with convergence order ρ = 3 and error equation satisfying

E k + 1 = ( C 2 − λ ) 2 E k 3 + O ( E k 4 ) (55)

For q = 2 , μ 1 = 1 4 , μ 2 = 3 4 , θ 1 = 0 , θ 2 = 2 3 , in Algorithm 2 give rise to the following new iterative method for approximating Φ of (1) is obtained.

Algorithm 5 Assume X 0 is an initial guess, approximate the solution Φ of (1) using the iterative method:

ν k = X k − [ G ′ ( X k ) − G ( X k ) λ ] − 1 G ( X k ) , X k + 1 = ν k − 4 [ G ′ ( X k ) + 3 G ′ ( X k + 2 ν k 3 ) − 4 G ( X k ) λ ] − 1 G ( ν k ) , k = 0 , 1 , 2 , ⋯ (56)

Algorithm 5 is of convergence order ρ = 3 iterative method for approximating the solution Φ of (1) error equation satisfying

E k + 1 = ( C 2 − λ ) 2 E k 3 + O ( E k 4 ) (57)

For q = 3 , μ 1 = 1 6 , μ 2 = 2 3 , μ 3 = 1 6 , θ 1 = 0 , θ 2 = 1 2 , and θ 3 = 1 in Algorithm 2 it reduces to the following new iterative method.

Algorithm 6 Assume X 0 is an initial guess, approximate the solution Φ of (1) using the iterative method:

ν k = X k − [ G ′ ( X k ) − G ( X k ) λ ] − 1 G ( X k ) , X k + 1 = ν k − 6 [ G ′ ( X k ) + 4 G ′ ( X k + ν k 2 ) + G ′ ( ν k ) − 6 G ( X k ) λ ] − 1 G ( ν k ) , k = 0 , 1 , 2 , ⋯ (58)

Algorithm 6 is a convergence order ρ = 3 iterative method for approximating the solution Φ of (1) error equation satisfying

E k + 1 = ( C 2 − λ ) 2 E k 3 + O ( E k 4 ) (59)

For q = 3 , μ 1 = 1 4 , μ 2 = 1 2 , μ 3 = 1 4 , θ 1 = 0 , θ 2 = 1 2 , and θ 3 = 1 in Algorithm 2 it reduces to the following new iterative method.

Algorithm 7 Assume X 0 is an initial guess, approximate the solution Φ of (1) using the iterative method:

ν k = X k − [ G ′ ( X k ) − G ( X k ) λ ] − 1 G ( X k ) , X k + 1 = ν k − 4 [ G ′ ( X k ) + 4 G ′ ( X k + ν k 2 ) + G ′ ( X k ) − 4 G ( X k ) λ ] − 1 G ( ν k ) , k = 0 , 1 , 2 , ⋯ (60)

Algorithm 7 is a convergence order ρ = 3 iterative method for approximating Φ of (1) having error equation satisfying

E k + 1 = ( C 2 − λ ) 2 E k 3 + O ( E k 4 ) (61)

Consider some particular forms of Algorithm 3. Set q = 1 , μ 1 = 1 , θ 1 = 1 2 , in Algorithm 3 leads to the iterative method.

Algorithm 8 Assume X 0 is an initial guess, approximate the solution Φ of (1) using the iterative method:

ν k = X k − [ G ′ ( X k ) − G ( X k ) λ ] − 1 G ( X k ) , W k = ν k − [ G ′ ( X k + ν k 2 ) − G ( X k ) λ ] − 1 G ( ν k ) , X k + 1 = W k − [ G ′ ( X k + W k 2 ) − G ( X k ) λ ] − 1 G ( W k ) , k = 0 , 1 , 2 , ⋯ (62)

Algorithm 8 is an iterative method for approximating the solution Φ of (1) with convergence order ρ = 4 and error equation satisfying

E k + 1 = − ( λ − C 2 ) 3 E k 4 + O ( E k 5 ) (63)

For q = 2 , μ 1 = 1 4 , μ 2 = 3 4 , θ 1 = 0 , θ 2 = 2 3 , in Algorithm 3 leads to the following iterative method.

Algorithm 9 Assume X 0 is an initial guess, approximate the solution Φ of (1) using the iterative method:

ν k = X k − [ G ′ ( X k ) − G ( X k ) λ ] − 1 G ( X k ) , W k = ν k − 4 [ G ′ ( X k ) + 3 G ′ ( X k + 2 ν k 3 ) − 4 G ( X k ) λ ] − 1 G ( ν k ) , X k + 1 = W k − 4 [ G ′ ( X k ) + 3 G ′ ( X k + 2 W k 3 ) − 4 G ( X k ) λ ] − 1 G ( W k ) , k = 0 , 1 , 2 , ⋯ (64)

Algorithm 9 is an iterative method for approximating the solution Φ of (1) with convergence order ρ = 4 . The error equation of Algorithm 9 is

E k + 1 = − ( λ − C 2 ) 2 E k 4 + O ( E k 5 ) (65)

For q = 3 , μ 1 = 1 6 , μ 2 = 2 3 , μ 3 = 1 6 , θ 1 = 0 , θ 2 = 1 2 , and θ 3 = 1 in Algorithm 3 the following iterative method is proposed:

Algorithm 10 Assume X 0 is an initial guess, approximate the solution Φ of (1) using the iterative method:

ν k = X k − [ G ′ ( X k ) − G ( X k ) λ ] − 1 G ( X k ) , W k = ν k − 6 [ G ′ ( X k ) + 4 G ′ ( X k + ν k 2 ) + G ′ ( ν k ) − 6 G ( X k ) λ ] − 1 G ( ν k ) , X k + 1 = W k − 6 [ G ′ ( X k ) + 4 G ′ ( X k + W k 2 ) + G ′ ( ϒ k ) − 6 G ( X k ) λ ] − 1 G ( W k ) , k = 0 , 1 , 2 , ⋯ (66)

The Algorithm 10 is of convergence order ρ = 4 for approximating the solution Φ of (1). Its error equation is

E k + 1 = − ( λ − C 2 ) 2 E k 4 + O ( E k 5 ) (67)

For q = 3 , μ 1 = 1 4 , μ 2 = 1 2 , μ 3 = 1 4 , θ 1 = 0 , θ 2 = 1 2 , and θ 3 = 1 in Algorithm 2 it reduces to the following new iterative method.

Algorithm 11 Assume X 0 is an initial guess, approximate the solution Φ of (1) using the iterative method:

ν k = X k − [ G ′ ( X k ) − G ( X k ) λ ] − 1 G ( X k ) , W k = ν k − 4 [ G ′ ( X k ) + 4 G ′ ( X k + ν k 2 ) + G ′ ( ν k ) − 4 G ( X k ) λ ] − 1 G ( ν k ) , X k + 1 = W k − 4 [ G ′ ( X k ) + 4 G ′ ( X k + W k 2 ) + G ′ ( W k ) − 4 G ( X k ) λ ] − 1 G ( W k ) , k = 0 , 1 , 2 , ⋯ (68)

The Algorithm 11 is a convergence order ρ = 3 for approximating the solution Φ of (1). Its error equation is

E k + 1 = ( C 2 − λ ) 2 E k 4 + O ( E k 5 ) (69)

In this section, the efficiency index (EI) of the iterative methods proposed are established. Let A v ρ represents iterative method v with convergence order ρ . For reference purpose, the iterative methods proposed are denoted as indicated in

of the iterative methods, [

2 ( 1 m + m 2 ) , for m ≥ 2 . This is same as the efficiency index (EI) of the classical

Newton method ( N 1 2 ) and the Wu method with convergence order ρ = 2 developed in [

The Wu method is given as:

W 1 2 : X k + 1 = X k − [ G ′ ( X k ) − d i a g ( σ i G i ( X k ) ) ] − 1 G ( X k ) (70)

where the parameter σ i ∈ [ − 1 , 1 ] , i = 1 , 2 , ⋯ , m .

lightaqua Iterative methods | Denotation | Iterative methods | Denotation |
---|---|---|---|

Algorithm 1 | A 1 2 | Algorithm 8 | A 8 3 |

Algorithm 4 | A 4 3 | Algorithm 9 | A 9 3 |

Algorithm 5 | A 5 3 | Algorithm 10 | A 10 3 |

Algorithm 6 | A 6 3 | Algorithm 11 | A 11 3 |

Algorithm 7 | A 7 3 | - | - |

Efficiency Index (EI) | |||
---|---|---|---|

lightaqua Methods | m | m = 10 | m = 20 |

A 1 2 , W 1 2 | 2 ( 1 m + m 2 ) | 1.0063 | 1.0017 |

A 4 3 , A 5 3 | 3 ( 1 2 m + 2 m 2 ) | 1.0050 | 1.0013 |

A 6 3 , A 7 3 | 3 ( 1 2 m + 3 m 2 ) | 1.0034 | 1.0009 |

A 8 4 , A 9 4 | 4 ( 1 3 m + 3 m 2 ) | 1.0042 | 1.0011 |

A 10 4 , A 11 4 | 4 ( 1 3 m + 5 m 2 ) | 1.0026 | 1.0007 |

From

The developed iterative methods are tested on three standard problems in the literature, in order to illustrate their performance and confirm the theoretical convergence order ( ρ ). The computational performance of the iterative methods developed are compared with the performance of the Wu method in [

H 2 3 : X k = X k − [ G ′ ( X k ) − d i a g ( σ i G i ( X k ) ) ] − 1 ( G ( X k ) + G ( η k ) ) (71)

where the parameter σ i ∈ [ − 1 , 1 ] , i = 1 , 2 , ⋯ , m and η k is approximated using W 1 2 .

For the implementation, Intel Celeron(R) CPU 1.6 GHz with 2 GB of RAM processor is used to execute PYTHON 2.7.12 programs. The stopping criterion used for computer programs is ‖ G ( X k + 1 ) ‖ ∞ < ϵ , where ϵ is error tolerance. The Metrics used in comparison are:

Number of iterations (IT), Central Processing Unit Time or Execution time (CPU-Time), Norm function of last iteration ( ‖ G ( X k + 1 ) ‖ ∞ < ϵ ) , and Computational order of convergence ( ρ c o c ) given in [

ρ c o c = ln ( ‖ G ( X k ) ‖ ) ln ( ‖ G ( X k − 1 ) ‖ ) (72)

To test the performance of proposed methods, the following problems are solved.

Problem 1 [

Consider the NLSE

G ( X ) = 0

where

G ( X 1 , X 2 ) = [ X 1 3 + X 1 X 2 X 2 + X 2 2 ]

The solutions of Problem 1 in the domain G : ( − 1.5,1.5 ) × ( − 1.5,1.5 ) are Φ ( 1 ) = ( 0 , 0 ) T and Φ ( 2 ) = ( 1 , − 1 ) T . The initial approximation used is X 0 = ( 0.5 , − 0.5 ) T . The numerical results obtained for each method using different values of the parameters λ i and σ i are presented in Tables 3-7. All computations are carried out with 200 digit precision and ϵ = 10 − 15 .

Problem 2 [

X 1 2 − X 2 + 1 = 0 ,

X 1 − cos ( π X 2 2 ) = 0 .

lightaqua Methods | Φ | IT | CPU time | ‖ G ( X k + 1 ) ‖ ∞ | ρ c o c |
---|---|---|---|---|---|

W 2 1 | Φ ( 1 ) | 32 | 0.535 | 3.1591E−15 | 1.0378 |

A 2 1 | Φ ( 2 ) | 7 | 0.129 | 1.1500E−08 | 2.0757 |

H 3 1 | Fail to converge | ||||

A 4 3 | Φ ( 2 ) | 3 | 0.107 | 1.1659E−05 | 3.3195 |

A 5 3 | Φ ( 2 ) | 5 | 0.177 | 4.7267E−12 | 3.1647 |

A 6 3 | Φ ( 2 ) | 5 | 0.176 | 4.7268E−12 | 3.1647 |

A 7 3 | Φ ( 2 ) | 4 | 0.148 | 1.1503E−10 | 2.9630 |

A 8 4 | Φ ( 2 ) | 3 | 0.148 | 2.3099E−07 | 3.7104 |

A 9 4 | Φ ( 2 ) | 3 | 0.164 | 4.6413E−07 | 4.5660 |

lightaqua Methods | Φ | IT | CPU time | ‖ G ( X k + 1 ) ‖ ∞ | ρ c o c |
---|---|---|---|---|---|

W 2 1 | Φ ( 1 ) | 34 | 0.552 | 2.2560E−15 | 1.0374 |

A 2 1 | Φ ( 2 ) | 8 | 0.147 | 2.4253E−08 | 2.0634 |

H 3 1 | Fail to converge | ||||

A 4 3 | Φ ( 2 ) | 4 | 0.139 | 2.8614E−07 | 2.8677 |

A 5 3 | Φ ( 2 ) | 6 | 0.210 | 7.4986E−15 | 3.1009 |

A 6 3 | Φ ( 2 ) | 6 | 0.211 | 7.4986E−15 | 3.1009 |

A 7 3 | Φ ( 2 ) | 5 | 0.184 | 7.0321E−15 | 2.9704 |

A 8 4 | Φ ( 2 ) | 3 | 0.197 | 2.1598E−09 | 3.7716 |

A 9 4 | Φ ( 2 ) | 4 | 0.217 | 1.4937E−14 | 4.2083 |

lightaqua Methods | Φ | IT | CPU time | ‖ G ( X k + 1 ) ‖ ∞ | ρ c o c |
---|---|---|---|---|---|

W 2 1 | Φ ( 1 ) | 36 | 0.586 | 1.8519E−15 | 1.0372 |

A 2 1 | Φ ( 2 ) | 10 | 0.184 | 2.6789E−13 | 2.0302 |

H 3 1 | Fail to converge | ||||

A 4 3 | Φ ( 2 ) | 5 | 0.165 | 1.9834E−07 | 3.0775 |

A 5 3 | Φ ( 2 ) | 7 | 0.235 | 1.5024E−10 | 3.1184 |

A 6 3 | Φ ( 2 ) | 7 | 0.239 | 1.5024E−10 | 3.1184 |

A 7 3 | Φ ( 2 ) | 5 | 0.188 | 2.0991E−07 | 2.9685 |

A 8 4 | Φ ( 2 ) | 16 | 0.675 | 1.2791E−15 | 1.0770 |

A 9 4 | Φ ( 2 ) | 4 | 0.208 | 5.6012E−06 | 4.6275 |

lightaqua Methods | Φ | IT | CPU time | ‖ G ( X k + 1 ) ‖ ∞ | ρ c o c |
---|---|---|---|---|---|

W 2 1 | Φ ( 1 ) | 37 | 0.632 | 2.0416E−15 | 1.0373 |

A 2 1 | Φ ( 2 ) | 11 | 0.212 | 8.4523E−15 | 2.0242 |

H 3 1 | Fail to converge | ||||

A 4 3 | Φ ( 2 ) | 6 | 0.198 | 4.5660E−09 | 3.0642 |

A 5 3 | Φ ( 2 ) | 8 | 0.276 | 1.9972E−08 | 3.1347 |

A 6 3 | Φ ( 2 ) | 8 | 0.277 | 1.9972E−08 | 3.1347 |

A 7 3 | Φ ( 2 ) | 6 | 0.211 | 1.6872E−13 | 2.9838 |

A 8 4 | Φ ( 2 ) | 4 | 0.191 | 1.0284E−06 | 4.4246 |

A 9 4 | Φ ( 2 ) | 5 | 0.253 | 1.0022E−12 | 4.1769 |

lightaqua Methods | Φ ( 1 ) | IT | CPU time | ‖ G ( X k + 1 ) ‖ ∞ | ρ c o c |
---|---|---|---|---|---|

W 2 1 | Φ ( 1 ) | 38 | 0.621 | 1.5247E−15 | 1.0370 |

A 2 1 | Φ ( 2 ) | 11 | 0.214 | 8.4523E−15 | 2.0242 |

H 3 1 | Fail to converge | ||||

A 4 3 | Φ ( 2 ) | 7 | 0.229 | 1.0329E−14 | 3.0409 |

A 5 3 | Φ ( 2 ) | 9 | 0.311 | 3.0272E−13 | 3.0778 |

A 6 3 | Φ ( 2 ) | 9 | 0.312 | 3.0272E−13 | 3.0778 |

A 7 3 | Φ ( 2 ) | 6 | 0.222 | 4.8002E−10 | 2.9919 |

A 8 4 | Φ ( 2 ) | 4 | 0.188 | 1.0284E−06 | 4.4246 |

A 9 4 | Φ ( 2 ) | 5 | 0.255 | 1.5505E−08 | 4.2787 |

The solutions of Problem 2 within the domain D = ( − 1 , 0 ) × ( 0 , 2 ) are

Φ ( 1 ) = ( 2 2 , 1.5 ) T , Φ ( 2 ) = ( − 1 , 2 ) T and Φ ( 3 ) = ( 0 , 1 ) T . The numerical solutions to Problem 2 are presented in

Problem 3 [

Consider the chemical equilibrium system modeled in NLSE (1) with

X 1 X 2 + X 1 − 3 X 5 = 0

2 X 1 X 2 + X 1 + X 2 X 3 2 + R 8 X 2 − R X 5 + R 10 X 2 2 + R 7 X 1 X 3 + R 9 X 2 X 4 = 0

2 X 2 X 3 2 + 2 R 5 X 3 2 − 8 X 5 + R 6 X 3 + R 7 X 2 X 3 = 0

R 9 X 2 X 4 + 2 X 4 2 − 4 R 5 = 0

X 1 ( X 2 + 1 ) + R 10 X 2 2 + X 2 X 3 2 + R 8 X 2 + R 5 X 3 2 + X 4 2 − 1 + R 6 X 3 + R 7 X 2 X 3 + R 9 X 2 X 4 = 0

where

R = 10 , R 10 = 0.193 , R 6 = 0.002597 40 , R 7 = 0.003448 40 , R 8 = 0.00001799 40 , R 9 = 0.0002155 40 , R 10 = 0.00003846 40

Using X 0 = ( 0.6 , 33.2 , 0.6 , 1.5 , − 0.7 ) T as initial starting point, 200 digits floating point arithmetics and ‖ G ( X k ) ‖ ≤ 10 − 50 as stopping criteria, the solution Φ in D = ( 1 , 1 ) × ( 33.5 , 35.5 ) × ( − 1 , 1 ) × ( − 0.8 , 1.8 ) × ( − 1 , 1 ) approximated to 20 decimal places is

Φ = [ 0.00311410226598496012 34.59792453029012391022 0.06504177869743799154 0.85937805057794058144 0.03695185914804602454 ]

The computational results obtained for different methods are presented in

lightaqua Methods | IT | CPU time | ‖ G ( X k + 1 ) ‖ ∞ | ρ c o c | |
---|---|---|---|---|---|

W 2 1 | Φ ( 2 ) | 8 | 0.119 | 1.7603E−11 | 1.9083 |

A 2 1 | Φ ( 1 ) | 7 | 0.116 | 2.3202E−10 | 2.0382 |

H 3 1 | Fail to converge | ||||

A 4 3 | Φ ( 3 ) | 4 | 0.126 | 1.0466E−08 | 3.2031 |

A 5 3 | Φ ( 3 ) | 10 | 0.316 | 9.2320E−14 | 2.0501 |

A 6 3 | Φ ( 3 ) | 4 | 0.135 | 2.0072E−06 | 5.5442 |

A 8 4 | Φ ( 1 ) | 5 | 0.213 | 3.9010E−09 | 3.2580 |

A 10 4 | Φ ( 2 ) | 6 | 0.286 | 1.3913E−11 | 2.6142 |

lightaqua Methods | λ i = σ i | IT | CPU time | ‖ G ( X k + 1 ) ‖ ∞ | ρ c o c |
---|---|---|---|---|---|

W 2 1 | 1/7 | 10 | 0.69 | 8.3642E−32 | 1.7416 |

A 2 1 | 1/7 | 9 | 0.67 | 1.2446E−33 | 2.6476 |

H 3 1 | 1/7 | 7 | 0.87 | 3.4193E−25 | 3.0227 |

A 4 3 | 1/7 | 6 | 0.81 | 3.7103E−44 | 3.0243 |

A 5 3 | 1/7 | 6 | 0.86 | 3.7103E−44 | 3.0243 |

A 6 3 | 1/7 | 6 | 1.08 | 3.7103E−44 | 3.0243 |

A 7 4 | 1/7 | 6 | 1.10 | 3.7103E−44 | 3.1114 |

A 8 4 | 1/7 | 4 | 0.82 | 2.4234E−16 | 4.4022 |

A 9 4 | 1/7 | 4 | 1.09 | 2.2924E−16 | 4.4599 |

A 10 4 | 1/7 | 4 | 1.19 | 2.2924E−16 | 4.4599 |

A 11 4 | 1/7 | 4 | 1.20 | 2.2924E−16 | 4.4888 |

The numerical results obtained on Tables 3-9, leads to the following observations about the effectiveness of the proposed methods in approximation of the solution of (1).

・ The numerical results obtained in Tables 3-9, clearly implies that the proposed methods are effective in approximation of solution of (1).

・ Most of the computational order of convergence ρ c o c of the proposed methods agrees with theoretical value.

・ It is observed that the proposed convergence order ρ = 2 method ( A 1 2 ) produce better precision compared with Wu method ( W 1 2 ) for small system. The reason is justifiable since G ( X ) λ is a dense matrix, more computation cost is incurred as the system become large.

・ Observe from Tables 3-8, Haijun method ( H 1 3 ) failed in Problem 1 and 2 while the proposed methods converged to solutions in few number of iterations.

・ The choice of λ , its magnitude should be less 1 to get better precision and convergence.

In this paper, multistep quadrature based methods for approximation of the solution of NLSE are proposed. The proposed methods require only first order Frechet derivative to attain convergence order ρ ≤ 4 and effectively approximate solution of NLSE with singular Jacobian. The proposed methods are applied on three standard problems in literature so as to describe their effectiveness. Judging from the computational results obtained and presented in tables, the proposed methods are competent compared to some existing methods.

The authors declare no conflicts of interest regarding the publication of this paper.

Ogbereyivwe, O. and Muka, K.O. (2019) Multistep Quadrature Based Methods for Nonlinear System of Equations with Singular Jacobian. Journal of Applied Mathematics and Physics, 7, 702-725. https://doi.org/10.4236/jamp.2019.73049