A natural boundary element method for the ... - Semantic Scholar

Teng et al. Boundary Value Problems (2017) 2017:179 DOI 10.1186/s13661-017-0910-x

RESEARCH

Open Access

A natural boundary element method for the Sobolev equation in the 2D unbounded domain Fei Teng1 , Zhendong Luo2* and Jing Yang1 *

Correspondence: [email protected] School of Mathematics and Physics, North China Electric Power University, No. 2, Bei Nong Road, Beijing, Changping District 102206, China Full list of author information is available at the end of the article 2

Abstract In this article, we devote ourselves to establishing a natural boundary element (NBE) method for the Sobolev equation in the 2D unbounded domain. To this end, we first constitute the time semi-discretized super-convergence format for the Sobolev equation by means of the Newmark method. Then, using the principle of natural boundary reduction, we establish a fully discretized NBE format based on the natural integral equation and the Poisson integral formula of this problem and analyze the errors between the exact solution and the fully discretized NBE solutions. Finally, we use some numerical experiments to verify that the NBE method is effective and feasible for solving the Sobolev equation in the 2D unbounded domain. MSC: 65N30; 35Q10 Keywords: natural boundary element method; Sobolev equation; error estimate; numerical experiment

1 Introduction Let ⊂ R be a connected and bounded region with smooth boundary := ∂, c := R \, x = (x , x ), |x| = x + x . For given time upper bound T, we consider the following Sobolev equation in the D unbounded domain: Problem I Find that satisfies ⎧ c ⎪ ⎪ ⎨t – εt – γ = f (x, t), (x, t) ∈ × (, T), ∂ = g(x, t), (x, t) ∈ × (, T), ∂n ⎪ ⎪ ⎩ (x, ) = u (x), x ∈ c ,

()

represents the partial derivative for the unknown function (x, t) about where t = ∂ ∂t time, ε and γ are positive constants, f (x, t), g(x, t), and u (x) are three known functions ∂ denotes the external normal derivative operator, satisfying the appropriate conditions, ∂n and n represents a normal vector onto boundary of domain c toward the interior of domain . In addition, we also assume the function (x, t) is bounded at infinity point (see [, ]). © The Author(s) 2017. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Teng et al. Boundary Value Problems (2017) 2017:179

Page 2 of 15

Just like the heat equation (see []) and the reaction diffusion equation (see []), Problem I is an important system of equations with real-life application background. It has been widely applied to many practical engineering fields (see [, ]), for example, it is used to describe the procedure of fluid flow permeating rocks, the soils moisture migration, and the different media heat transfer. It usually includes complex computing domain, initial and boundary value functions, or source term for the Sobolev equation in the D unbounded domain in the real world so that it has no analytic solution. Hence, one has to depend on the numerical methods. The natural boundary element (NBE), which was proposed by Feng and Yu at the end of s (see [–]), is a novel type of boundary element method (BEM) and suitable for solving the unbound regional problem and is an attractive and promising numerical method. It has been effectively applied to complicated boundary and infinity regional problem. It has more specific advantages than the usual BEM, for instance, it has a unique form of boundary integral equation, remains the unchanged energy functional, holds very good numerical stability, adopts the same variation principle as the finite element method (FEM), can couple with FEM and we need not calculate a great many singular integrations in practice. The main idea of the NBE method consists in introducing an appropriate artificial boundary and then restricting the computation to an appropriate large finite spatial domain. It divides the domain into two subregions, a bounded inner region  (bounded annular region by  and R ) and a regular unbounded region  (unbounded domain outside circle R ) (see [, ]). One can obtain the natural integral equation on boundary R and corresponding Poisson integral formula of the subproblem over unbounded domain  by the natural boundary reduction. Only the function itself, not its normal derivative at the common boundary R , appears in the variational. It has been used to solve the second-order elliptic equations, standard parabolic equations, and hyperbolic equations (see [, , , , ]). However, for all we know, up to now, the NBE method has yet not been used to solve the Sobolev equation. Especially, the Sobolev equation not only includes a first-order derivative term of time and two second-order derivative terms of spatial variables but also does two mixed derivative terms of spatial variables (second-order) and time (first-order) so that either the establishment of NBE format or the theoretical analysis needs more skills and is confronted with more difficulties than the second-order elliptic equations, standard parabolic equations, and hyperbolic equations as mentioned above, but it has certain special applications as mentioned above. Therefore, it is worth to study the NBE method for the Sobolev equation in the D unbounded domain. The forthcoming contents are scheduled as follows. In Section , we establish a semidiscretized super-convergence format about time for the Sobolev equation in the D unbounded domain and deduce the super-convergence error estimates of the semidiscretized solutions about time. In the next Section , by the principle of natural boundary reduction, we establish a fully discretized NBE format based on the natural integral equation and the Poisson integral formula of this problem and provide the error estimates between the exact solution and the fully discretized NBE solutions. In Section , we supply some numerical experimentations to validate that the numerical computational consequences are concordant with the theoretical ones. Section  summarizes main conclusions.


Page 3 of 15

2 Semi-discretized formulation by the Newmark method and error estimate about time for the Sobolev equation in the 2D unbounded domain By using Green’s formula, we can obtain the following variational form for the Sobolev equation. Problem II For t ∈ (, T), find (t) ∈ H  (c ) that satisfies (I – ε)(t , υ) + γ (∇ , ∇υ) = (f , υ) + g, υ,

∀υ ∈ H  c ,

()

where (·, ·) represents the inner product in L (c ), but ·, · is the inner product in L (). The existence and uniqueness of the solution for the variational form, Problem II, are well known (see [, , ]). ˙ (x, tk )= ∂ (x, tk ). By the Newmark Set τ is the time-step, tk = k · τ , k = (x, tk ), and ∂t method (see, e.g., []), we can establish the following iterative scheme: ˙ k (x) = γ (I – ε)– k (x) + (I – ε)– f k (x), k = , , . . . , N,

˙ k– (x) + β ˙ k (x) , k = , , . . . , N, k (x) = k– (x) + τ ( – β)

() ()

where β ∈ (, ] and N = [T/τ ] ([T/τ ] denotes the integer part of T/τ ). ˙ k– ]/β to approximate to tk , we obtain the following By using [( k – k– )/τ – ( – β) semi-discretized format about time for the Sobolev equation. Problem III Find k ∈ H  (c ) such that

k – k– , υ + (ε + βτ γ ) ∇ k , ∇υ + τ γ ( – β) – ε ∇ k– , ∇υ = βτ f k + τ ( – β)f k– , υ + (βτ γ + ε)g k + τ γ ( – β) – ε g k– , υ , ∀υ ∈ H  c , k = , , . . . , N,

()

where  = u (x), f k = f (x, tk ), and g k = g(x, tk ). The solutions to Problem III possess the following conclusions. Theorem  If f ∈ L (, T; L (c )), g ∈ L (, T; L ()), and u ∈ H  (c ), then Problem III  c has a unique set of solutions { k }N k= ⊂ H ( ) satisfying

k

≤  u  + (ε + τβγ ) ∇u    

+ 

k

  C βγ + ετ – g i , + β  τ f i ,c

 exp(T/),

()

i=

where C is the nonnegative constant in the trace theorem and k = , , . . . , N . This signifies that the solutions of Problem III are stable and consecutively rely on the source function f ,


Page 4 of 15

boundary value function g, and initial value function u . In addition, when is sufficiently smooth about t, we have the following error estimates:

∇ (tk ) – k = ∇ek ≤ Cτ ,   where C  = τ ε– tk+ .

k

i= ( (I

k = , , . . . , N,

()

– ε)tt (ξi )  + (I – ε)tt (ξi )  ) · exp(T), tk– ≤ ξk , ξk ≤

Proof Because Problem III is a linear equation about unknown function , to prove the existence and uniqueness of solutions of Problem III is equivalent to demonstrating that it has only a zero solution when g = f = u = . Choosing υ = k in Problem III together with utilizing the Hölder and Cauchy-Schwarz inequalities and the trace theorem (see []), we have

k 

+ (ε + βτ γ ) ∇ k   





   ≤  β  τ f k  + C τ – (ε + βτ γ ) g k , + k  + k–  

τ (ε + βτ γ )  

∇ k–  + ∇ k  , + k  + (ε + βτ γ ) ∇ k  +    

()

where C is the nonnegative constant in the trace theorem (see []). By summing () from  to k, we have

k 

+ (ε + βτ γ ) ∇ k    ≤

k

τ

ui  + (ε + βτ γ ) ∇ui  + u  + (ε + βτ γ ) ∇u       i=

+

k



 C τ – (ε + βτ γ ) g i , + β  τ f i  ,

 ≤ k ≤ N.

()

i=

When τ is adequately small such that τ ≤ , by Gronwall’s lemma (see [, ]), we have

k

 ≤  u  + (ε + τβγ ) ∇u   k

 – i   i  C βγ + ετ g , + β τ f ,c +  · exp(T/),

 ≤ k ≤ N. ()

i=

Thus, when f = g = u = , from (), we obtain k  = , implying k =  (k = , , . . . , N ). Therefore, Problem III has a unique set of solutions. With Taylor’s expanding formula, we acquire

  t (tk ) = (tk ) – (tk– ) – ( – β)t (tk– ) β τ +

( – β)τ τ tt ξk – tt ξk , β β

()


Page 5 of 15

where tk– ≤ ξk , ξk ≤ tk+ . From Problem II, we obtain ( (tk ) – (tk– , υ) + (ε + βτ γ ) ∇ (tk ), ∇υ + τ γ ( – β) – ε ∇ (tk– ), ∇υ = βτ f (tk ) + τ ( – β)f (tk– ), υ + τ  ( – β) (I – ε)tt ξk , υ τ + (βτ γ + ε)g(tk ) + τ γ ( – β) – ε g(tk– ), υ – (I – ε)tt ξk , υ ,  c  ∀υ ∈ H .

()

Let ek = (tk ) – k . By subtracting () from () taking t = tk , we obtain

ek – ek– , υ + (ε + βτ ) ∇ek , ∇υ + τ γ ( – β) – ε ∇ek– , ∇υ τ (I – ε)tt ξk , υ . = τ  ( – β) (I – ε)tt ξk , υ – 

()

By taking υ = ek in () and using the Cauchy and Hölder-Schwarz inequalities, we have

k 

e + (ε + βτ γ ) ∇ek    ≤



ek  + ek–  + τ  (I – ε)tt ξ k      

 (ε + βτ γ ) k–  k  τ 

∇e + ∇e . + ek  + τ  β  (I – ε)tt ξk  +    

()

By summing () from  to k, when τ is adequately small such that τ / ≤ /, we have k– k

i 

k 

(I – ε)tt ξ i 

e + τ 

e + (ε + βτ γ ) ∇ek  ≤ τ      i=

i=

 + ( – β) (I – ε)tt ξi  . 

()

By Gronwall’s lemma (see [, ]), we have k

k  

e + (ε + βτ γ ) ∇ek  ≤ τ  ( – β) (I – ε)tt ξi    i=

 + (I – ε)tt ξi  · exp(T).

()

From (), we obtain

∇ (tk ) – k = ∇ek ≤ Cτ ,  

k = , , . . . , N,

()

where C  = τ ε– ki= ( (I – ε)tt (ξi )  + (I – ε)tt (ξi )  ) · exp(T). This finishes the proof of Theorem .

3 Natural boundary reduction on the outside circle area and error estimate for the fully discretized NBE solutions We define μ := ( (ε + τβγ ))– , k := (I – ε) k– + τ (I – ε)( – β) ˙ k– , f := – k – τβf k .


Page 6 of 15

The procedure for solving the above semi-discretized Problem III is given as follows. () Prediction k := (I – ε) k– + τ (I – ε)( – β) ˙ k– .

()

() Solve the problem k – μ k = μ f k,

x ∈ c ,

()

∂ k = g k , x ∈ , ∂n k < +∞, |x| → +∞.

() ()

() Update value ˙k=

 k k . – τβ

()

From the above procedure, it is not difficult to find that our main task is to solve elliptic boundary value problems at each time level. Let In (x) and Kn (x) (n = , , , . . .) be, severally, the first and second type of modified Bessel functions (see []), and μ and μ be, respectively, the natural and Poisson integral operators (see [, ]). From the NBE method (see [, , ]), the Dirichlet boundary ˆ k ∂ k and the Neumann boundary ∂n satisfy the following relationship: ∂ k f k , θ = μ ˆ k , + N μ, r; ∂n

()

and the relationship between the solution k of Problem III with its Dirichlet boundary value ˆ k is as follows: f k , R, θ , ˆ k + F μ, r; k = μ

()

where +∞ +∞ μ

¯ n (μ, r; σ ) fñk,c (σ ) cos nθ + fñk,s (σ ) sin nθ dσ , N μ, r; f˜ k , θ = ξn G  n= r ξ = ; ξn = , n = , , . . . , ¯ n (μ, r; σ ) = – Kn (μσ ) · σ , G Kn (μr) r

n = , , , . . . ,

+∞ +∞ μ

F μ, r; f˜ k , R, θ = ξn σ  Gn (R, σ ) fñk,c (σ ) cos nθ + fñk,s (σ ) sin nθ dσ ,  n= r ⎧ ⎨ φn (σ )ψn (R) , R ≤ σ , En (σ ) n = , , , . . . , σ  Gn (R, σ ) = ⎩ ψn (σ )φn (R) , R ≥ σ , En (σ ) φn (σ ) = Kn (μσ ),

ψn (σ ) = In (μσ )Kn (μr) – Kn (μσ )In (μr),

n = , , , . . . ,


Page 7 of 15

En (σ ) = ψn (σ )φn (σ ) – φn (σ )ψn (σ ), n = , , , . . . ,  π ˜ k f (σ , θ ) cos nθ dθ , n = , , , . . . , fñk,c (σ ) = π   π ˜ k f (σ , θ ) sin nθ dθ , n = , , . . . . fñk,s (σ ) = π  The above () and () are named the natural and the Poisson integral equations, respectively. Thus, () is equivalent to the following variation form:  Find k ∈ H  () ( ≤ k ≤ N ) that satisfy Bˆ k , ν k = g k (r, θ ) + N μ, r; fk , θ , ν k , ˆ k , ν k ) = μ where B( ˆ k , ν k =:

ˆ k )ν k ( μ



∀ν k ∈ H  (), ds and ω, =:

()

ω ds.

3.1 Natural boundary reduction on the external circle area Now, let the domain be a circle with radius r and center at origin. For convenience, we also suppose that the solutions k of Problem III are appropriately smooth. Under the polar coordinates, = {(R, θ ) : R = r, θ ∈ [, π]}, c = {(R, θ ) : R = |x| > r, θ ∈ [, π]}, ∂ ∂ and the external normal derivative operator onto satisfies ∂n = – ∂R . The solution of equations ()-() can be expressed with the following form in the polar coordinates:

 an (R) cos nθ + bn (R) sin nθ , k (R, θ ) = a (R) +  n= +∞

()

where  an (R) = π bn (R) =

 π

π

k (R, θ ) cos nθ dθ ,

n = , , , . . . ;

k (R, θ ) sin nθ dθ ,

n = , , . . . .

 π



By calculating, we get the solution k (R, θ ) of equations ()-() as follows. k (R, θ ) =

π +∞  Kn (μR) cos n θ – θ k r, θ dθ ξn π n= Kn (μr)  + F μ, r; f˜ k , R, θ ,

R > r, π μ ∂ k (r, θ ) + N μ, r; f˜ k , θ = K¯ n μ, r; θ – θ k r, θ dθ , ∂n π  where K¯ n (μ, r; θ – θ ) = –

+∞

n= ξn cos n(θ

()

()

– θ ) · Kn (μr)/Kn (μr).

Remark  The formats () and () are, respectively, the Poisson and the natural integral equations. We can attain the solution k (r, θ ) from the natural integral equation () and then obtain the solution of the original boundary value, i.e., Problem I, by the Poisson integral formula (). But the solution of Problem I can be acquired directly by the Poisson integral formula () for the Cauchy-Dirichlet initial boundary value problem, because the function k (r, θ ) is known.


Page 8 of 15

Remark  In numerical computations, we can limit r ≤ R < +∞ on the r ≤ R ≤ Rmax and use the corresponding numerical integral to calculate the integral calculation of the N (μ, r; f˜ k , θ ) and F (μ, r; f˜ k , R, θ ). Meanwhile, the expressions of an infinite series summation can be substituted with a finite series summation in the practical applications.

3.2 Error analysis of NBE solutions We divide the circumference into some finite elements, which satisfy regular conditions. For the sake of simplicity, we take uniform subdivision. Let Sh () ⊂ H / () be a finite element subspace spanned by appropriate basis functions. Thus, the NBE approximation to Problem II is as follows. k Problem IV Find h ∈ Sh () ( ≤ k ≤ N ) that satisfy

k k k Bˆ h , ν = g (r, θ ) + N μ, r; fk , θ , ν k ,

∀ν k ∈ Sh (),

()

and hk (R, θ ) =

π +∞ k  Kn (μR) r, θ dθ cos n θ – θ h ξn π n= K (μr) n  + F μ, r; f˜ k , R, θ , R > r.

()

The above formula () is the approximate expression of Poisson integral formula (). In order to analyze the errors of NBE solutions of Problem IV, it is necessary to introduce the following L -projection and its property (see []). Definition  An operator Ph : L () → Sh () (where Sh () ⊂ L () is a finite element space) is known as an L -projection if, for any υ ∈ L (), there exists unique Ph υ ∈ Sh () satisfying υ – Ph υ, υh = ,

∀υh ∈ Sh ().

Lemma  If Sh () is a subspace spanned by piecewise linear polynomials and υ ∈ H  (), then the L -projection Ph satisfies υ – Ph υ s ≤ Ch–s υ , ,

s = –, , ,

where C used next represents a generic positive real independent of τ and h. Problem IV possesses the following conclusion. 

k be, respectively, solutions to () together with (), Theorem  Let k ∈ H  () and h  τ = O(h ), and Sh () be the piecewise linear polynomial subspace. Then we have the following error estimates:

k

– k ≤ Ch ,  h ,

k = , , . . . , N.

()


Page 9 of 15

Proof By subtracting () from () taking ν k = νhk , we obtain k , νhk Bˆ k – h π +∞ +∞ k– μ k– ¯ –  – h ξn = Gn (μ, r; σ ) π n= r 

k– k– + ε k– – h – τ ( – β)γ k– – h cos nθˆ dθˆ cos nθ π k– k– + –(I – ε) k– – h – τ γ k– – h 

k– k– k ˆ ˆ + τβγ  – h sin nθ dθ sin nθ dσ , νh ,

∀ν k ∈ Sh ().

()

ˆ ·) being positive definite on H  () × H  () (see []), by using the Hölder inDue to B(·, equality, we have

k  M k – h , k k k ≤ Bˆ  – h , k – h k k k , k – Ph k + Bˆ k – h , Ph k – h ≤ Bˆ k – h +∞ +∞ μ

k

k k k

¯ n (μ, r; σ ) ξn ≤  – h ,  – Ph  , + G π r n= π k– k– k– –  – h + ε k– – h × 

k– cos nθˆ dθˆ cos nθ – τ ( – β)γ k– – h π k– –(I – ε) k– – h + 

k– k– k k ˆ ˆ – τ ( – β)γ  – h sin nθ dθ sin nθ dσ , Ph  – h

k

k

– Ph k + C k– – k– Ph k – k ≤ k – h   ,    , h , ,

+ ε k– – k– + τ ∇ k– – k– Ph k – k 

h

,



h

,

k–

Ph k – k – τβ ∇ k– – h   , ,

k

k–

– k + ( + ε) k– – h  h , ,

k k–

– k . + τ γ ( – β) ∇ k– – h  h , ,



 ,

Further, by the Cauchy-Schwarz inequality and Lemma  as well as the inverse estimating theorem (see []), we acquire

k

– k  ≤ C k– – k–  + h .   h , h ,


Page 10 of 15

Further, we acquire

k

– k ≤ Ch .  h ,

This finishes the proof of Theorem . The solutions for approximate expression () have the following error estimates.

Theorem  Let k and hk be, respectively, the solutions of () and () and τ = O(h ). Then we have the error estimates:

k 

– k h ,∞,c ≤ Ch .

()

Proof By the literature [], we can immediately derive Kn (x) =

π n n ex

n

 + o n– ,

n → +∞.

˜ ≥  that satisfies Therefore, we have Kn (μR)/Kn (μr) ≤  (r < R). Thus, there is real M +∞ K (μR) n ˜ ξn ≤ M. Kn (μr) n=

By using (), we have f˜ k = [–(I – ε) – τ ( – β)γ ] k– – τβf k – τ ( – β)f k– . Thus, from () and (), we have k – k h +∞ +∞ π k Kn (μR)  k ≤ cos n θ – θ ·  – h dθ ξn π Kn (μr)  n=

n=

π +∞ μ –(I – ε) k– – k– cos nθˆ dθˆ ξn σ  Gn (R, σ )  h π n= r  π k– k– ˆ ˆ + –τ ( – β)γ  – h cos nθ dθ cos nθ +∞ 

+



+ 

–(I – ε) k– – k– sin nθˆ dθˆ  h

π

+ 

–τ ( – β)γ k– – k– sin nθˆ dθˆ sin nθ dσ  h

π

+∞  π Kn (μR) ξn cos n θ – θ dθ ≤ π Kn (μr)  n=

 

k · k – h ,

π +∞ +∞

μ k– ξn σ  Gn (R, σ ) · k– – h cos nθˆ dθˆ , π n= r 

k– k– + ε  – h cos nθˆ  , +

 


Page 11 of 15

k– k–

ˆ + τ ( – β)γ ∇  – h , ∇ cos nθ  cos nθ

k– + k– – h ,

+ ε k– – k– 

h

,

π

sin nθˆ dθˆ

 



sin nθˆ 

k– ˆ  sin nθ dσ + τ ( – β)γ ∇ k– – h ∇ sin n θ , ≤ Ch + Ch

+∞ +∞ μ εn σ  Gn (R, σ )(sin nθ + cos nθ ) dσ π n= r

≤ Ch .

()

From (), we immediately gain (). This finishes the proof of Theorem . For the fully discretized NBE format, Problem IV, we have the following conclusion.

Theorem  If f k ∈ L (c ), g ∈ L (), and ϑ ∈ H  (c ), Problem IV has a unique solution k ∈ Sh () satisfying h k

k i

g + τβ f i c · exp ( + ε + βγ )T ,

≤ C u + h , h  , ,

()

i=

where uh = Ph u (x). This signifies that the solutions of Problem IV are steady and consecutively rely on the source function f , boundary value function g, and initial value function ϑ . Furthermore, when τ = O(h ), we have the error estimates.

(tk ) – k c ≤ C τ + h , h ,

k = , , . . . , N.

()

ˆ ·) being symmetrical, continuous, and positive definitive on H  () × Proof Due to B(·,   H () (see [, ]), by the Lax-Milgram theorem (see [, , ]), we know that Problem IV has a unique set of solutions. k in () and using the Hölder inequality, we have Then, by taking νhk = h 

k 

≤ Bˆ k , k = g k (r, θ ) + N μ, r; f˜ k , θ , k M h h h h ,

π +∞ +∞

μ

k

¯ –(I – ε) ≤ g , + ξn Gn (μ, r; σ )

π r  n=

k– – τ ( – β)f k– – τβf k · cos nθˆ dθˆ cos nθ – τ ( – β)γ h π k– + –(I – ε) – τ ( – β)γ h 

k

– τ ( – β)f k– – τβf sin nθˆ dθˆ sin nθ dσ h ,

,

k–

+ τβ f k c + g k k . ≤ C ( + ε + βτ ) h h , , , ,

k

()


Page 12 of 15

Furthermore, we have

k

≤ C ( + ε + βτ ) k– + g k + τβ f k c . h  h  , ,

()

By summing () from  to k and using Gronwall’s lemma (see [, ]), we get k

k

i i

≤ C u , +

g + τβ f c · exp ( + ε + βγ )T . h  , ,

()

i=

By using the triangle inequality,

(tk ) – k c ≤ (tk ) – k c + C k – k h , h ,∞,c , ,

()

and combining () and () with (), we can acquire (). This finishes the proof of Theorem .

4 Some numerical experiments At the moment, we utilize some numerical experimentations to validate that the numerical computational conclusions coincide with the theoretical ones and that the NBE format is effective and feasible for solving the Sobolev equation in the D unbounded domain. Let c be the external region outside the unit circle. The source term is chosen as

f (x, t) =  R– – .π  R– – R– sin(.πR) + .π  R– cos(.πR) cos(t)

+ R– – .π  R– sin(.πR) – .πR– cos(.πR) sin(t), where R = |x| = x + y ≥  and take ε = γ = . The boundary and initial functions are, respectively, chosen as g(x, t) = – sin(t) and u (x) =  as in []. We approximately re M ˜k place +∞ n= with n= and adopt the numerical integral to calculate N (μ, r; f , θ ) and k F (μ, r; f˜ , R, θ ) in our numerical experimentations. We divide the circumference into  arc paragraphes with side length θ = π/, which satisfies usual regular conditions, and take M =  and the time step size τ = k at time t = , , , , .. We find exact solution k and numerical solutions h , , , which are shown in Photo (a)’s and (b)’s of Figs. -, severally. Whereas the errors between the exact solution and numerical solutions at t = , , , , , ,  are exhibited graphically in Photo (c)’s of Figs. -, severally. From each group of photos in

Figure 1 The exact and NBE solutions and the error between them at t = 1. Photos (a) and (b) are severally the exact and NBE solutions and Photo (c) is the error between them.


Page 13 of 15





Figs. -, we can clearly see that the exact solutions are basically the same as the numerical solutions. In particular, the error photos indicated that the numerical computing consequences are consistent with the theoretical ones since both theoretical and numerical


Page 14 of 15



errors are O(– ). Especially, it is super-convergent about time accuracy. Even if t = , the numerical solution still converges and maintains accuracy O(– ). These sufficiently signify that the NBE method is very effective and feasible for solving the Sobolev equation in the D unbounded domain.

5 Conclusions In this article, we have established the semi-discretized format about time for the Sobolev equation in the D unbounded domain by the Newmark method and gained the error estimates of super-convergence of the semi-discretized solutions about time. Especially, we have built the fully discretized NBE format and analyzed the errors between the analytical solution and the fully discretized NBE solutions. We have also provided some numerical experiments to validate that our method is effective and feasible. The most important thing is that the NBE method applied to solve the Sobolev equation in the D unbounded domain is first presented, it is new and original. Moreover, the method can also solve many practical problems. Acknowledgements This research was supported by the Fundamental Research Funds for the Central Universities grant 2017XS067 and National Science Foundation of China grant 11671106. Availability of data and materials The authors declare that all the data and material in the article are available and veritable. Competing interests The authors declare that they have no competing interests. Authors’ contributions All authors contributed equally and significantly in writing this article. All authors wrote, read, and approved the final manuscript.


Page 15 of 15

Author details 1 School of Control and Computer Engineering, North China Electric Power University, No. 2, Bei Nong Road, Beijing, Changping District 102206, China. 2 School of Mathematics and Physics, North China Electric Power University, No. 2, Bei Nong Road, Beijing, Changping District 102206, China.

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Received: 5 October 2017 Accepted: 21 November 2017 References 1. Du, QK, Chen, JR: Development Equation Boundary Element Method and Its Application. Science Press, Beijing (2013) (in Chinese) 2. Zhu, CJ, Du, QK: Natural boundary element method for parabolic equations in an unbounded domain. J. Math. Res. Exposition 22(2), 177-188 (2002) 3. Abdelwahed, M, Chorfi, N, Radulescu, V: Numerical solutions to heat equations via the spectral method. Electron. J. Differ. Equ. 2016, 68 (2016) 4. Zhan, H: Reaction diffusion equations with boundary degeneracy. Electron. J. Differ. Equ. 2016, 81 (2016) 5. Shi, DM: On the initial boundary value problem of nonlinear the equation of the moisture in soil. Acta Math. Appl. Sin. 13(1), 33-40 (1990) 6. Ting, TW: A cooling process according to two-temperature theory of heat conduction. J. Math. Anal. Appl. 45(1), 23-31 (1974) 7. Feng, K: Finite element method and natural boundary reduction. In: Proceedings of the International Congress of Mathematicians. Polish Academy Press, Warszawa (1983) 8. Feng, K, Yu, DH: Canonical integral equations of elliptic boundary value problem and their numerical solutions. In: Proceeding of China-France Symposium on the Finite Element Method, pp. 211-252. Science Press, Beijing (1983) 9. Yu, DH: Mathematical Theory of Natural Boundary Element Method. Science Press, Beijing (1993) 10. Yu, DH: Numerical solutions of harmonic and biharmonic canonical integral equations in interior or exterior circular domains. J. Comput. Math. 1(1), 52-62 (1983) 11. Feng, K: Canonical Boundary Reduction and Finite Element Method. Science Press, Beijing (1982) 12. Barenblett, GI, Zheltov, IP, Kochina, IN: Basic concepts in the theory of homogeneous liquids in fissured rocks. J. Appl. Math. Mech. 24(5), 1286-1303 (1990) 13. Wood, WL: Practical Time-Stepping Schemes. Clarendon, London (1990) 14. Adams, RA: Sobolev Spaces. Academic Press, New York (1975) 15. Girault, V, Raviant, PA: Finite Element Approximations of the Navier-Stokes Equations, Theorem and Algorithms. Springer, New York (1986) 16. Luo, ZD: Mixed Finite Element Methods and Applications. Science Press, Beijing (2006) (in Chinese) 17. Wang, ZX, Guo, DR: Special Functions. World Scientific, Singapore (1989) 18. Courant, R, Hilbert, D: Methods of Mathematical Physics (II). Wiley, New York (1962) 19. Francis, B, Hilde, B: Advance Calculus for Applications, 2nd edn. Prentice Hall, New York (1976)