Abstract
In this paper, we propose an efficient B-spline finite element method for a class of fourth order nonlinear differential equations with variable coefficient. For the temporal discretization, we choose the Crank–Nicolson scheme. Boundedness and error estimates are rigorously derived for both semi-discrete and fully discrete schemes. A numerical experiment confirms our theoretical analysis.
Similar content being viewed by others
1 Introduction
The epitaxial growth of nanoscale thin films has attracted a lot of attention in recent years [1–6]. The key reason for this concern is that compositions like YBa2Cu3O\(_{7-\delta }\) (YBCO) are expected to be high temperature superconducting materials that can be used in semiconductor design. King et al. [1] proved the existence, uniqueness, and regularity of solution in an appropriate function space for the initial boundary value problem of the epitaxial thin-film growth. Kohn et al. [3] considered a fourth order parabolic equation, which is a specific example of energy-driven coarsening in two dimension space, and proved that the time-averaged energy per unit area decays no faster than \(t^{-\frac{1}{3}}\).
The finite element method (FEM) plays an important role in solving differential equations [7–11]. There are some papers which have already been published to study the FEM for fourth order nonlinear parabolic equations [12–18]. Liu et al. [12] considered a nonlinear model describing epitaxial thin-film growth with constant coefficient and demonstrated that the Hermite FEM has the convergence rate of \(O(\Delta t+h^{3})\). Choo [13] constructed a finite element scheme for the viscous Cahn–Hilliard equation with a nonconstant gradient energy coefficient and obtained the error estimate using the extended Lax–Richtmyer equivalence theorem. In [18], Qiao et al. presented a mixed FEM for the molecular beam epitaxy mode and showed that the semi-discrete and fully discrete schemes satisfy the nonlinear energy stability property. Moreover, the authors gave the error analysis.
In 1946, the B-spline method was first introduced by Schoenberg [19]. In 1966, Curry and Schoenberg [20] presented one element B-spline functions. In 1976, B-splines were extended to multiple situations [21]. As a class of piecewise polynomials, B-splines are often used in finite element analysis [22–27]. Erfanian et al. [25] used the linear B-spline FEM and cubic B-spline FEM for solving linear Volterra integro-differential equation in the complex plane. Dhawan et al. [26] applied the linear and quadratic B-spline functions to the advection-diffusion equations.
The main advantages of B-splines are the freedom to choose the order and smoothness, the simple data structure with one parameter in \([0,1]\), and the exact representation of boundary conditions. Compared with Lagrange and Hermite type elements, B-spline shape functions involve only one type of basis function. Thus the scale of matrix from B-spline FEM is smaller than that from Lagrange and Hermite elements. Moreover, B-spline shape functions are smoother. It is known that quadratic B-splines, which are in \(C^{1}(-\infty,+\infty )\), satisfy the weak form of fourth order differential equations. However, to deal with boundary conditions, the B-spline basis functions need to be modified. In the present work, we choose the cubic B-spline FEM for a fourth order nonlinear parabolic equation with variable coefficient. It is proved that the convergence order of the Crank–Nicolson scheme is higher than that of the backward Euler scheme in [12].
The following sections are organized as follows. In Sect. 2, we introduce the model and some basic preliminaries. In Sect. 3, we show the boundedness and error estimates for the semi-discrete scheme. In Sect. 4, a fully discrete scheme based on the Crank–Nicolson method is studied. In Sect. 5, a numerical experiment is provided to confirm theoretical results.
In this work, we denote \(L^{2}\), \(L^{k}\), \(L^{\infty }\), \(H^{k}\) norms in I by \(\|\cdot \|\), \(\|\cdot \|_{L^{k}}\), \(|\cdot |_{\infty }\), and \(\|\cdot \|_{k}\), respectively.
2 Initial boundary value problem and some preliminaries
In this paper, we consider the following problem:
where \(I=[0,1]\) and \(u_{t}=\frac{\partial u}{\partial t}\).
For the variable coefficient, the following assumptions hold:
where s, S, \(M_{1}\), and \(M_{2}\) are positive constants.
Considering the boundary value conditions, we need the following space:
The weak formulation associated with problem (1) is: Find \(u=u(\cdot,t)\in H_{0}^{2}(I)\ (0\leq t\leq T) \) such that
where \(Du=\frac{\partial u}{\partial x}\).
According to [12], the solution of problem (1) exists.
Theorem 2.1
Suppose that\(u_{0}\in H_{0}^{2}(I)\cap W^{1,4}(I)\), then there exists a unique global solution\(u(x,t)\)for problem (1) such that
3 Semi-discrete finite element scheme
Let L be a positive integer. Define a uniform partition \(I_{h}:0=x_{0}< x_{1}<\cdots <x_{L}=1\), \(h=x_{i}-x_{i-1}=\frac{1}{L}\), \(I_{i}=[x_{i-1}, x_{i}]\). To deal with boundary value conditions, we define six additional knots \(x_{-3}=-3h\), \(x_{-2}=-2h\), \(x_{-1}=-h\), \(x_{L+1}=1+h\), \(x_{L+2}=1+2h\), \(x_{L+3}=1+3h\).
The cubic B-spline function with integer knots can be defined as follows:
then it is easy to get the cubic B-spline in the interval \([x_{i},x_{i+4}]\) which is
The modified basis functions are defined as follows [28]:
For the sake of convenience, we denote the modified basis functions with respect to \(\{x_{i}\}\) by \(\{\varphi _{i}(x)\}\), which satisfies the following properties:
The modified basis functions can deal with homogeneous as well as non-homogeneous boundary conditions.
Let \(U_{h}\) be the cubic B-spline space. One can see that the cubic B-spline space is in \(C^{2}(-\infty,\infty )\), thus \(U_{h}\subset H_{0}^{2}\). The approximation solution \(u_{h}(x,t)\in U_{h}\) satisfies
where \(\delta _{i}(t)\) are time-dependent quantities.
The semi-discrete finite element scheme based on B-splines for problem (5) is: Find \(u_{h}=u_{h}(\cdot,t)\in U_{h}\ (0< t\leq T)\) such that
The bandwidth of stiffness matrix is 7, and the matrix order is \(L-1\), which is only half of the Lagrange and Hermite finite element scheme.
In order to estimate the errors of the B-spline FEM, we introduce the elliptic projection \(R_{h} u\):
then \(R_{h} u\) is uniquely defined, and
where \(C_{0}\) is a positive constant depending on \(\alpha (x,t)\). Hence, \(a(u,v)\) is a symmetrical positive definite bilinear form and (see [12])
We shall discuss the boundedness of the semi-discrete scheme, which is important for error analysis.
Theorem 3.1
Let\(u_{h}(0)\in H^{2}_{0}(I)\cap W^{1,4}(I)\), then there exists a unique solution\(u_{h}(t)\in U_{h}\)for problem (6) such that
whereCis a positive constant depending on\(\alpha (x,t)\)andT, independent of mesh sizeh.
Proof
According to ordinary differential equation theory, there exists a unique local solution to problem (5) in the interval \([0,t_{n})\). If we have (10), then according to the extension theorem, we can also obtain the existence of unique global solution. So, we only need to prove (10).
Taking \(v_{h}=u_{h}\) in (6), based on (3), we get
By the interpolation inequality, we have
Therefore
According to the method of solving separable equations, we can get the result
Integrating (12) with respect to t, we have
Integrating (11) with respect to t, we get
By (13), we obtain
Choose \(v_{h}=u_{h,t}\) in (6) to get
A direct calculation gives
Define the energy function
It is clear that \(E_{h}(t)\geq 0\). Differentiating \(E_{h}(t)\) with respect to t, we get
By (14), we have
Integrating (17) with respect to t, we have
It is obvious that
In view of (3) and (16), we obtain
It is clear to see from Cauchy’s inequality that
Therefore
It follows that
where C is a positive constant depending on \(\alpha (x,t)\) and \(u_{h}(0)\).
Owing to the interpolation inequality, we obtain
Thus (10) holds. The proof of the theorem is completed. □
Now, we give the error estimates between the solution to problem (5) and the solution in \(L^{2}\) and \(H^{2}\) norms.
Theorem 3.2
Letube the solution to (5), \(u_{h}\)be the solution to (6), \(u(0)\in H^{4} (I)\), \(u, u_{t}\in L^{2}(0,T;H^{4}(I))\), the initial value satisfies
As\(0\leq t\leq T\), we have the following error estimate:
whereCis a positive constant depending on\(\alpha (x,t)\)andT, independent of mesh sizeh.
Proof
Denote \(\theta (t)=R_{h}u-u_{h}\) and \(\rho (t)=u-R_{h}u\). Then
Taking \(v_{h}=\theta \) in (22), we have
By estimating the nonlinear term, we have
Based on Sobolev’s embedding theorem, \(H^{2}(I)\hookrightarrow W^{1,\infty }(I)\), i.e.,
we obtain
In addition, it is easy to get
Using (3), Hölder’s inequality, and Young’s inequality, we can deduce
Hence
By Gronwall’s inequality, we have
Moreover, using the triangle inequality, we know
Hence, when \(0\leq t\leq T\), it follows from (21) and (27)–(28) that formula (20) is derived. This completes the proof. □
Theorem 3.3
Letube the solution to (5), \(u_{h}\)be the solution to (6), \(u(0)\in H^{4}(I)\), \(u,u_{t}\in L^{2}(0,T;H^{4}(I))\), and the initial value satisfies
Then we have the following error estimate:
whereCis a positive constant depending on\(\alpha (x,t)\), independent of mesh sizeh.
Proof
Letting \(v_{h}=\theta _{t}\) in (22), we get
Using the triangle inequality and Sobolev’s embedding theorem, we get
Based on (4) and ε-inequality, we have
Integrating (32) with respect to t, we find
By (3), we have
Then
Note that
Combining (33) and (34), we have
Finally, using (29), we obtain (30). The proof is completed. □
4 Fully discrete finite element scheme
To construct the Crank–Nicolson scheme, we define the following function:
where \(H(Du_{h}^{n})\) is a double well potential function. Obviously, \(H'(Du_{h})=|Du_{h}|^{2} Du_{h}-Du_{h}\).
The fully discrete finite element scheme for problem (1) is: Find \(u_{h}^{n}\in U_{h}\ (n=1,2,\ldots, N)\) such that
where N is a given positive integer, \(\Delta t=T/N\) denotes the time step size, \(t_{n}=n\Delta t\) and
Firstly, we analyze the boundedness of the fully discrete scheme (36). It is a key step for deducing the error estimate.
Theorem 4.1
Let\(u_{h}^{0}\in H^{2}_{0}(I)\cap W^{1,4}(I)\), then there exists a unique solution\(u_{h}^{n}\)for problem (36) such that
whereCis a positive constant depending on\(\alpha (x,t)\)andT, independent ofhand Δt.
Proof
A direct calculation gives
Setting \(v_{h}=u_{h}^{n}+u_{h}^{n-1}\) in (36), we get
Using Cauchy’s inequality, we obtain
Further, we derive
Letting \(\gamma =\frac{1}{2s}\), we have
It is easy to show
If Δt is small enough, we conclude
Choosing \(v_{h}=\partial _{t}u_{h}^{n}\) in (36), we have
Then we get
Define the function
then \(G(u_{h}^{n},t^{n-\frac{1}{2}})\geq 0\). By (44) and (45), we have
With the differential mean value theorem and the boundedness of variable coefficient, we obtain
where \(t^{n-\frac{3}{2}}<\xi <t^{n-\frac{1}{2}}\). Then
Taking the sum over n, we get
It is obvious that
Therefore we know
Based on (44) and \(u_{h}^{0}\in H^{2}_{0}(I)\cap W^{1,4}(I)\), we have
where \(C(u_{h}^{0})\) is a constant depending on \(u_{h}^{0}\). Then
Using discrete Gronwall’s inequality, we derive
Based on (48), it is easy to see
We also know
By (42) and (49), we obtain (37). The proof is completed. □
Next, we give the error estimate in \(L^{2}\) norm.
Theorem 4.2
Let\(u^{n}\)be the solution to problem (5), \(u_{h}^{n}\)be the solution to the fully discrete scheme (36), \(u(0)\in H^{4}(I)\), \(u_{t}\in L^{2}(0,T;H^{4}(I))\cap L^{2}(0,T;W^{1,4}(I)) \), \(u_{ttt}\in L^{2}(0,T;L^{2}(I))\)and\(u_{h}^{0}\in U_{h}\)satisfying
Then we have the following error estimate:
whereCis a positive constant depending on\(\alpha (x,t)\)andT, independent of mesh sizeh.
Proof
Denote \(u_{t}^{n}=u_{t}(x,t^{n})\) and \(u^{n}=u(x,t^{n})\). Setting \(t=t^{n-1}\) and \(t=t^{n}\) in (5), respectively, we obtain
Denote
and
It follows from (52)–(54) and (36) that
Let \(\rho ^{n}=u^{n}-R_{h} u^{n}\) and \(\theta ^{n}=R_{h} u^{n}-u_{h}^{n}\), then \(u^{n}-u_{h}^{n}=\rho ^{n}+\theta ^{n}\). It is clear to get
where
An easy calculation gives
Using Taylor’s theorem, we have
and
With (4), we get
From (7), we have
Setting \(v_{h}=\theta ^{n}+\theta ^{n-1}\) in (58), we get
Based on the Newton–Leibniz formula and Hölder’s inequality, we have
Thus
A direct calculation gives
From (37) and Sobolev’s embedding theorem, \(H^{2}_{0}(I)\hookrightarrow H^{1,\infty }(I)\), we know
Using Hölder’s inequality, we have
Due to (60), we get
By the triangle inequality, we obtain
In view of (62)–(64) and (9), we have
Based on the ε-inequality and Hölder’s inequality, we obtain
Let \(r^{n}=r_{1}^{n}+r_{2}^{n}\), where
It is clear to see that
Using Taylor’s formula, we derive
We easily get
Adding (59), (65), and (66), we have
We know
Then
Taking the sum over n, by \(n\Delta t=t_{n}\leq T\), we have
Hence
If Δt is small enough, we have
By discrete Gronwall’s inequality, it gives
Finally, we obtain (51). The proof is completed. □
In the following theorem, we introduce the error estimate in \(H^{2}\) norm.
Theorem 4.3
Let\(u^{n}\)be the solution to (5), \(u_{h}^{n}\)be the solution to the fully discrete problem (36), \(u(0)\in H^{4}(I)\), \(u_{t}\in L^{2}(0,T;H^{4}(I))\cap L^{2}(0,T;W^{2,4}(I)) \), \(u_{ttt}\in L^{2}(0,T;L^{2}(I))\), and\(u_{h}^{0}\in U_{h}\)satisfying
Then we have the following error estimate:
Proof
Letting \(v_{h}=\partial _{t} \theta ^{n}\) in (58), we get
By Cauchy’s inequality, we have
Using the Newton–Leibniz formula and Hölder’s inequality, we obtain
Based on (70), we have
There exists \(\xi \in (t^{n-\frac{3}{2}},t^{n-\frac{1}{2}})\) such that
Then we have
Taking the sum over n and using (4), we can obtain
It follows from (3) that
Then one has
Introducing some symbols \(I_{1}\), \(I_{2}\), \(I_{3}\), then a direct calculation gives
It is obvious that
First, applying the triangle inequality to \(\|I_{1}\|\), we get
Based on Sobolev’s embedding theorem, we have
Further, Hölder’s inequality yields
Second, we analyze \(\|I_{2}\|\). A direct calculation gives
With the help of Sobolev’s embedding theorem, we can obtain
Then
For \(\|I_{3}\|\), by the triangle inequality and (9), one can have
Substituting (66) and (78) into (74), we obtain
Letting \(n=1\) in (71), based on (66) and (78), we have
Using (51), we have
If Δt is sufficiently small, discrete Gronwall’s inequality yields
This completes the proof. □
5 Numerical approximation
In this section, a numerical example is provided to illustrate the proposed B-spline FEM for solving the nonlinear parabolic equation. The efficiency of the cubic B-spline finite element scheme is tested. We consider
where \(\alpha (x,t)=1+xt\) is selected to satisfy the primary assumptions. We take the analytical solution \(u(x,t)=t^{2}(1-\cos 2\pi x)\). Then the concrete functional form of \(f(x,t)\) is
Figure 1 illustrates the behavior of the exact solution to problem (81), and Fig. 2 demonstrates the profile of the solution to the fully discrete scheme.
In this example, the numerical solution is in good accordance with the exact solution, indicating that the numerical scheme is valid and efficient.
Then choosing \(t=1\), the corresponding errors and convergence rates of the cubic B-spline FEM are shown in Tables 1–3.
In Table 1, to analyze the spatial convergence order, we take the time step \(\Delta t=\frac{1}{8000}\). The values in Table 1 indicate that with the decreasing of the space size, the error is monotone decreasing. We also find that the numerical solution to the scheme is fourth order convergent in \(L^{2}\) norm and is second order convergent in \(H^{2}\) norm.
In Table 2, we consider the error estimates and convergence orders in time direction when the space step is fixed to \(h=\frac{1}{1000}\). It is easy to see that the orders of error estimate both are second order in \(L^{2}\) and \(H^{2}\) norms.
In Table 3, we analyze the convergence rate when the space and time step change at the same time. We choose \((\Delta t,h)=(\frac{1}{100},\frac{1}{10}), (\frac{1}{400}, \frac{1}{20}), (\frac{1}{1600},\frac{1}{40}), (\frac{1}{6400}, \frac{1}{80})\), respectively. One can see that the relation between the space step and the time step is \(\Delta t/h^{2}=1\). This shows that the B-spline finite element scheme is very stable.
The numerical experiment indicates that the cubic B-spline FEM is an efficient approximation tool for solving the fourth order nonlinear parabolic equation.
6 Conclusion
In this paper, we propose the B-spline FEM for a class of fourth order nonlinear parabolic equations. On the one hand, B-splines have better smoothness than the Lagrange and Hermite type elements. On the other hand, B-spline finite element only has one type of basis functions, so the scale of matrix from B-spline FEM is lower.
The coefficient \(\alpha (x,t)\) is variable, which broadens the application fields and also increases the difficulty of analysis. By defining the biharmonic projection operator and the energy function, we prove the boundedness of the semi-discrete and fully discrete schemes based on B-splines. Further, the error estimates in \(L^{2}\) norm and \(H^{2}\) norm are deduced by using the boundedness, Sobolev’s embedding theorem, and so on. The results of numerical example confirm our theoretical analysis.
In general, the B-spline FEM is an efficient method for solving higher order nonlinear parabolic equations. By using central difference, the convergence rate in time direction, which can be improved, is of second order.
Change history
21 July 2020
In the original publication of this article [1] the name of the second author is incorrect. The correct name of the second author is Jiawei Tan rather than John Jiawei Tan. The error in this Correction has been updated in the original article.
References
King, B.B., Stein, O., Winkler, M.: A fourth order parabolic equation modeling epitaxial thin film growth. J. Math. Anal. Appl. 286, 459–490 (2003)
Zangwill, A.: Some causes and a consequence of modeling roughening. J. Cryst. Growth 163, 8–21 (1996)
Kohn, R.V., Yan, X.: Upper bound on the coarsening rate for an epitaxial growth model. Commun. Pure Appl. Math. 56, 1549–1564 (2003)
Liu, C.: Regularity of solutions for a fourth order parabolic equation. Bull. Belg. Math. Soc. Simon Stevin 13(3), 527–535 (2006)
Liu, C.: A fourth order parabolic equation with nonlinear principal part. Nonlinear Anal. 68, 393–401 (2008)
Evans, J.D., Galaktionov, V.A., King, J.R.: Blow-up similarity solutions of the fourth-order unstable thin film equation. Eur. J. Appl. Math. 18, 195–231 (2007)
Ciarlet, P.G.: The Finite Element Method for Ellipitic Problems. North-Holland, Amsterdam (1978)
Douglas, J., Dupont, T.: Galerkin methods for parabolic equations. SIAM J. Numer. Anal. 7, 575–626 (1970)
Brenner, S.C., Scott, L.R.: The Mathematical Theory of Finite Element Methods. Springer, Berlin (2002)
Chen, H., Chen, Y.: A combined mixed finite element and discontinuous Galerkin method for compressible miscible displacement. SIAM J. Numer. Anal. 7, 575–626 (1970)
Brezzi, F., Fortin, M.: Mixed and hybrid finite element methods. Natur. Sci. J. Xiangtan Univ. 26(2), 119–126 (2004)
Liu, F.N., Zhao, X.P., Liu, B.: Finite element analysis of a nonlinear parabolic equation modeling epitaxial thin-film growth. Bound. Value Probl. 2014, 46 (2014)
Choo, S.M., Kim, Y.H.: Finite element scheme for the viscous Cahn–Hilliard equation with a nonconstant gradient energy coefficient. J. Appl. Math. Comput. 19(1–2), 385–395 (2005)
Zhang, T.: Finite element analysis for Cahn–Hilliard equation. Math. Numer. Sin. 28, 281–292 (2006)
Elliott, C.M., French, D.A.: A nonconforming finite element method for the two-dimensional Cahn–Hilliard equation. SIAM J. Numer. Anal. 26(4), 889–903 (1989)
Barrett, J.W., Blowey, J.F., Garcke, H.: Finite element approximation of a fourth order nonlinear degenerate parabolic equation. Numer. Math. 80(4), 525–556 (1998)
Kästner, M., Metsch, P., de Borst, R.: Isogeometric analysis of the Cahn–Hilliard equation-a convergence study. J. Comput. Phys. 305, 360–371 (2016)
Qiao, Z.H., Tang, T., Xie, H.H.: Error analysis of a mixed finite element method for the molecular beam epitaxy model. SIAM J. Numer. Anal. 53(1), 184–205 (2015)
Schoenberg, I.J.: Contributions to the problem of approximation of equidistant data by analytic functions. Q. Appl. Math. 4, 45–99 (1946)
Curry, H.B., Schoenberg, I.J.: On Pólya frequency functions IV: the fundamental spline functions and their limits. J. Anal. Math. 17(1), 71–107 (1966)
de Boor, C.: B-form basis. In: Farin, G. (ed.) Geometric Modelling, pp. 131–148. SIAM, Philadelphia (1978)
Soliman, A.A.: A Galerkin solution for Burgers’ equation using cubic B-spline finite elements. Abstr. Appl. Anal. 2012, Article ID 527467 (2012)
Pourgholi, R., Tabasi, S.H., Zeidabadi, H.: Numerical techniques for solving system of nonlinear inverse problem. Eng. Comput. 34, 487–502 (2018)
Kutluay, S., Esen, A.: A B-spline finite element method for the thermistor problem with the modified electrical conductivity. Appl. Math. Comput. 156(3), 621–632 (2005)
Erfanian, M., Zeidabadi, H.: Approximate solution of linear Volterra integro-differential equation by using cubic B-spline finite element method in the complex plane. Adv. Differ. Equ. 2019(1), 62 (2019)
Dhawan, S., Kapoor, S., Kumar, S.: Numerical method for advection diffusion equation using FEM and B-splines. J. Comput. Sci. 3, 429–437 (2012)
Kolman, R., Okrouhlík, M., Berezovski, A., Gabriel, D., Kopačka, J., Plešek, J.: B-spline based finite element method in one-dimensional discontinuous elastic wave propagation. Appl. Math. Model. 46, 382–395 (2017)
Qin, D.D., Du, Y.W., Liu, B., Huang, W.Z.: A B-spline finite element method for nonlinear differential equations describing crystal surface growth with variable coefficient. Adv. Differ. Equ. 2019(1), 175 (2019)
Acknowledgements
The authors would like to express their deep thanks for the referee’s valuable suggestions about the revision and improvement of the manuscript.
Availability of data and materials
Please contact author for data requests.
Funding
This paper is supported by the Science and Technology Fund Project of Guizhou Health Commission (gzwjkj2019-1-050).
Author information
Authors and Affiliations
Contributions
DQ wrote the first draft. DQ and WH made the figure of numerical solution and errors, BL and JT corrected and improved the final version. All authors read and approved the final draft.
Corresponding author
Ethics declarations
Competing interests
The authors declare that they have no competing interests.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Qin, D., Tan, J., Liu, B. et al. A B-spline finite element method for solving a class of nonlinear parabolic equations modeling epitaxial thin-film growth with variable coefficient. Adv Differ Equ 2020, 172 (2020). https://doi.org/10.1186/s13662-020-02629-6
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s13662-020-02629-6