Skip to main content
Log in

An enriched finite element method to fractional advection–diffusion equation

  • Original Paper
  • Published:
Computational Mechanics Aims and scope Submit manuscript

Abstract

In this paper, an enriched finite element method with fractional basis \(\left[ 1,x^{\alpha }\right] \) for spatial fractional partial differential equations is proposed to obtain more stable and accurate numerical solutions. For pure fractional diffusion equation without advection, the enriched Galerkin finite element method formulation is demonstrated to simulate the exact solution successfully without any numerical oscillation, which is advantageous compared to the traditional Galerkin finite element method with integer basis \(\left[ 1,x\right] \). For fractional advection–diffusion equation, the oscillatory behavior becomes complex due to the introduction of the advection term which can be characterized by a fractional element Peclet number. For the purpose of addressing the more complex numerical oscillation, an enriched Petrov–Galerkin finite element method is developed by using a dimensionless fractional stabilization parameter, which is formulated through a minimization of the residual of the nodal solution. The effectiveness and accuracy of the enriched finite element method are demonstrated by a series of numerical examples of fractional diffusion equation and fractional advection–diffusion equation, including both one-dimensional and two-dimensional, steady-state and time-dependent cases.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8
Fig. 9
Fig. 10
Fig. 11

Similar content being viewed by others

References

  1. Em Karniadakis G, Hesthaven JS, Podlubny I, Podlubny I (2015) Special issue on “ Fractional PDEs: theory, numerics, and applications”. J Comput Phys 293:1–3

    Article  MathSciNet  Google Scholar 

  2. Chen W, Sun HG, Zhang XD, Korosak D (2010) Anomalous diffusion modeling by fractal and fractional derivatives. Comput Math Appl 59:1754–1758

    Article  MathSciNet  MATH  Google Scholar 

  3. Craiem D, Magin RL (2010) Fractional order models of viscoelasticity as an alternative in the analysis of red blood cell (RBC) membrane mechanics. Phys Biol 7:013001

    Article  Google Scholar 

  4. Tang S, Greene MS, Liu WK (2012) Two-scale mechanism-based theory of nonlinear viscoelasticity. J Mech Phys Solids 60:199–226

    Article  MathSciNet  MATH  Google Scholar 

  5. West BJ (2014) Colloquium: Fractional calculus view of complexity: a turorial. Rev Mod Phys 86:1169–1184

    Article  Google Scholar 

  6. Zhang Y, Meerschaert MM, Packman AI (2012) Linking fluvial bed sediment transport across scales. Geophys Res Lett 39:L20404

    Google Scholar 

  7. Ervian VJ, Roop JP (2005) Variational formulation for the stationary fractional advection dispersion equation. Numer Methods PDE 22:558–576

    Article  MathSciNet  Google Scholar 

  8. Ervin VJ, Roop JP (2007) Variational solution of fractional advectoin dispersion equations on bounded domains in \(\mathbb{R}^d\). Numer Methods PDE 23:256–281

    Article  MATH  Google Scholar 

  9. Fix GJ, Roop JP (2004) Least squares finite-element solution of a fractional order two-point boundary value problem. Comput Math Appl 48:1017–1033

    Article  MathSciNet  MATH  Google Scholar 

  10. Lian YP, Ying YP, Tang SQ, Lin S, Wagner GJ, Liu WK (2016) A Petrov-Galerkin finite element method for the fractional advection–diffusion equation. Comput Methods Appl Mech Eng 309:388–410

    Article  MathSciNet  Google Scholar 

  11. Brooks AN, Hughes TJR (1982) Streamline upwind/Petrov–Galerkin formulations for convection dominated flows with particular emphasis on the incompressible navier-stokes equations. Comput Methods Appl Mech Eng 32(1):199–259

    Article  MathSciNet  MATH  Google Scholar 

  12. Hughes TJR (1978) A simple scheme for developing ’upwind’ finite elements. Int J Numer Methods Eng 12(3):1359–1365

    Article  MATH  Google Scholar 

  13. Roop JP (2006) Computational aspects of FEM approximation of fractional advection dispersion equations on bounded domains in \(\mathbb{R}^d\). J Comput Appl Math 193:243–268

    Article  MathSciNet  MATH  Google Scholar 

  14. Idczak D, Stanislaw W (2013) Fractional Sobolev spaces via Riemann–Liouville derivatives. J Funct Spaces Appl 2013:1–15

    Article  MathSciNet  MATH  Google Scholar 

  15. Lian YP, Luan SZ, Ying YP, Tang SQ, Liu WK (2017) A general Petrov–Galerkin finite element method to two dimentional fractional advection–diffusion equation. In: Preparation (2017)

Download references

Acknowledgements

The support of ARO Grant W911NF-15-1-0569 is gratefully acknowledged. Yuping is grateful for the support from China Scholarship Council. Shaoqiang Tang is partially supported by the National Nature Science Foundation of China under Contract Numbers 11272009 and 11521202.

Author information

Authors and Affiliations

Authors

Corresponding authors

Correspondence to Yanping Lian or Wing Kam Liu.

Appendices

Appendix 1

In this appendix, the following formula is demonstrated.

$$\begin{aligned}&\sum _{J=1}^{k,k\le I-2}\check{K}_{IJ}\left( \alpha ,M\right) \phi _{J}^{\text {ex}} =\frac{\alpha \varGamma \left( 1+\alpha \right) }{\varGamma \left( 1-\alpha \right) \varGamma ^{2}\left( \alpha \right) }\frac{k^{\alpha }\left( k+1\right) ^{\alpha }}{\left( k+1\right) ^{\alpha }-k^{\alpha }}\nonumber \\&\quad \times \frac{1}{I^{\alpha }-\left( I-1\right) ^{\alpha }}\left( M_{G}^{\alpha }\mid _{\frac{k}{I-1}}^{\frac{k}{I}}-M_{G}^{\alpha }\mid _{\frac{k+1}{I-1}}^{\frac{k+1}{I}}\right) \nonumber \\&\qquad - \frac{\alpha \varGamma \left( 1+\alpha \right) }{\varGamma \left( 1-\alpha \right) \varGamma ^{2}\left( \alpha \right) }\frac{k^{\alpha }\left( k+1\right) ^{\alpha }}{\left( k+1\right) ^{\alpha }-k^{\alpha }}\nonumber \\&\quad \times \frac{1}{\left( I+1\right) ^{\alpha }-I^{\alpha }}\left( M_{G}^{\alpha }\mid _{\frac{k}{I}}^{\frac{k}{I+1}}-M_{G}^{\alpha }\mid _{\frac{k+1}{I}}^{\frac{k+1}{I+1}}\right) , \end{aligned}$$
(72)

where \(\check{K}_{IJ}\left( \alpha ,M\right) \) is part of stiffness matrix following the formula in Eq. (17), and \(\phi _{J}^{\text {ex}}\left( x\right) =\left( J\triangle x\right) ^{\alpha }\) is the nodal exact solution. Hence, Eq. (72)can be rewritten as

$$\begin{aligned} \sum _{J=1}^{k,k\le I-2}\check{K}_{IJ}\left( \alpha ,M\right) \phi _{J}^{\text {ex}}=\frac{\alpha \varGamma \left( 1+\alpha \right) }{\varGamma \left( 1-\alpha \right) \varGamma ^{2}\left( \alpha \right) }\sum _{J=1}^{k,k\le I-2}A_{J}B_{J}. \end{aligned}$$
(73)

The mathematical expression of \(A_{J}\) and \(B_{J}\) in Eq. (73) are defined as

$$\begin{aligned} A_{J}= & {} \frac{J^{\alpha }}{\left[ I^{\alpha }-\left( I-1\right) ^{\alpha }\right] \left[ J^{\alpha }-\left( J-1\right) ^{\alpha }\right] }M_{G}^{\alpha }\mid _{\frac{J}{I-1}}^{\frac{J}{I}}\nonumber \\&-\,\frac{\left( J-1\right) ^{\alpha }}{\left[ I^{\alpha }-\left( I-1\right) ^{\alpha }\right] \left[ J^{\alpha }-\left( J-1\right) ^{\alpha }\right] }M_{G}^{\alpha }\mid _{\frac{J-1}{I-1}}^{\frac{J-1}{I}}\nonumber \\&+\,\frac{J^{\alpha }}{\left[ I^{\alpha }-\left( I-1\right) ^{\alpha }\right] \left[ \left( J+1\right) ^{\alpha }-J^{\alpha }\right] }M_{G}^{\alpha }\mid _{\frac{J}{I-1}}^{\frac{J}{I}}\nonumber \\&-\,\frac{\left( J+1\right) ^{\alpha }}{\left[ I^{\alpha }-\left( I-1\right) ^{\alpha }\right] \left[ \left( J+1\right) ^{\alpha }-J^{\alpha }\right] }M_{G}^{\alpha }\mid _{\frac{J+1}{I-1}}^{\frac{J+1}{I}}\nonumber \\&-\,\frac{J^{\alpha }}{\left[ \left( I+1\right) ^{\alpha }-I^{\alpha }\right] \left[ J^{\alpha }-\left( J-1\right) ^{\alpha }\right] }M_{G}^{\alpha }\mid _{\frac{J}{I}}^{\frac{J}{I+1}}\nonumber \\&+\,\frac{\left( J-1\right) ^{\alpha }}{\left[ \left( I+1\right) ^{\alpha }-I^{\alpha }\right] \left[ J^{\alpha }-\left( J-1\right) ^{\alpha }\right] }M_{G}^{\alpha }\mid _{\frac{J-1}{I}}^{\frac{J-1}{I+1}}\nonumber \\&-\,\frac{J^{\alpha }}{\left[ \left( I+1\right) ^{\alpha }-I^{\alpha }\right] \left[ \left( J+1\right) ^{\alpha }-J^{\alpha }\right] }M_{G}^{\alpha }\mid _{\frac{J}{I}}^{\frac{J}{I+1}}\nonumber \\&+\,\frac{\left( J+1\right) ^{\alpha }}{\left[ \left( I+1\right) ^{\alpha }-I^{\alpha }\right] \left[ \left( J+1\right) ^{\alpha }-J^{\alpha }\right] }M_{G}^{\alpha }\mid _{\frac{J+1}{I}}^{\frac{J+1}{I+1}}. \end{aligned}$$
(74)
$$\begin{aligned} B_{J}= & {} J^{\alpha }. \end{aligned}$$
(75)

By mathematical induction, Eq. (72) can be demonstrated as following steps.

$$\begin{aligned} S_{1}= & {} A_{1}B_{1}=\frac{1^{\alpha }2^{\alpha }}{2^{\alpha }-1^{\alpha }}\left[ \frac{1}{I^{\alpha }-\left( I-1\right) ^{\alpha }}M_{G}^{\alpha }\mid _{\frac{1}{I-1}}^{\frac{1}{I}}\right. \nonumber \\&-\,\frac{1}{\left( I+1\right) ^{\alpha }-I^{\alpha }}M_{G}^{\alpha }\mid _{\frac{1}{I}}^{\frac{1}{I+1}}+\frac{1}{\left( I+1\right) ^{\alpha }-I^{\alpha }}M_{G}^{\alpha }\mid _{\frac{2}{I}}^{\frac{2}{I+1}}\nonumber \\&\left. -\,\frac{1}{I^{\alpha }-\left( I-1\right) ^{\alpha }}M_{G}^{\alpha }\mid _{\frac{2}{I-1}}^{\frac{2}{I}}\right] . \end{aligned}$$
(76)

Assuming

$$\begin{aligned} S_{n}=&\sum _{J=1}^{n}A_{J}B_{J}=\frac{n^{\alpha }\left( n+1\right) ^{\alpha }}{\left( n+1\right) ^{\alpha }-n^{\alpha }}\left[ \frac{1}{I^{\alpha }-\left( I-1\right) ^{\alpha }}M_{G}^{\alpha }\mid _{\frac{n}{I-1}}^{\frac{n}{I}}\right. \nonumber \\&\left. -\,\frac{1}{\left( I+1\right) ^{\alpha }-I^{\alpha }}M_{G}^{\alpha }\mid _{\frac{n}{I}}^{\frac{n}{I+1}}\right] \nonumber \\&+\, \frac{n^{\alpha }\left( n+1\right) ^{\alpha }}{\left( n+1\right) ^{\alpha }-n^{\alpha }}\left[ \frac{1}{\left( I+1\right) ^{\alpha }-I^{\alpha }}M_{G}^{\alpha }\mid _{\frac{n+1}{I}}^{\frac{n+1}{I+1}}\right. \nonumber \\&\left. -\,\frac{1}{I^{\alpha }-\left( I-1\right) ^{\alpha }}M_{G}^{\alpha }\mid _{\frac{n+1}{I-1}}^{\frac{n+1}{I}}\right] . \end{aligned}$$
(77)

Then, by some simple algerbaric calculations

$$\begin{aligned} S_{n+1}&=\sum _{J=1}^{n+1}A_{J}B_{J}=\frac{\left( n+1\right) ^{\alpha }\left( n+2\right) ^{\alpha }}{\left( n+2\right) ^{\alpha }-\left( n+1\right) ^{\alpha }}\nonumber \\&\times \left[ \frac{1}{I^{\alpha }{-}\left( I{-}1\right) ^{\alpha }}M_{G}^{\alpha }\mid _{\frac{n+1}{I-1}}^{\frac{n+1}{I}}{-}\frac{1}{\left( I{+}1\right) ^{\alpha }{-}I^{\alpha }}M_{G}^{\alpha }\mid _{\frac{n+1}{I}}^{\frac{n+1}{I+1}}\right] \nonumber \\&\quad +\, \frac{\left( n+1\right) ^{\alpha }\left( n+2\right) ^{\alpha }}{\left( n+2\right) ^{\alpha }-\left( n+1\right) ^{\alpha }}\left[ \frac{1}{\left( I+1\right) ^{\alpha }-I^{\alpha }}M_{G}^{\alpha }\mid _{\frac{n+2}{I}}^{\frac{n+2}{I+1}}\right. \nonumber \\&\quad \left. -\,\frac{1}{I^{\alpha }-\left( I-1\right) ^{\alpha }}M_{G}^{\alpha }\mid _{\frac{n+2}{I-1}}^{\frac{n+2}{I}}\right] . \end{aligned}$$
(78)

Hence, Eq. (72) is proved.

$$\begin{aligned}&\sum _{J=1}^{k,k\le I-2}\check{K}_{IJ}\left( \alpha ,M\right) \phi _{J}^{\text {ex}} =\frac{\alpha \varGamma \left( 1+\alpha \right) }{\varGamma \left( 1-\alpha \right) \varGamma ^{2}\left( \alpha \right) }S_{k}\nonumber \\&\quad = \frac{\alpha \varGamma \left( 1+\alpha \right) }{\varGamma \left( 1-\alpha \right) \varGamma ^{2}\left( \alpha \right) }\frac{k^{\alpha }\left( k+1\right) ^{\alpha }}{\left( k+1\right) ^{\alpha }-k^{\alpha }}\nonumber \\&\qquad \times \frac{1}{I^{\alpha }-\left( I-1\right) ^{\alpha }}\left( M_{G}^{\alpha }\mid _{\frac{k}{I-1}}^{\frac{k}{I}}-M_{G}^{\alpha }\mid _{\frac{k+1}{I-1}}^{\frac{k+1}{I}}\right) \nonumber \\&\qquad - \frac{\alpha \varGamma \left( 1+\alpha \right) }{\varGamma \left( 1-\alpha \right) \varGamma ^{2}\left( \alpha \right) }\frac{k^{\alpha }\left( k+1\right) ^{\alpha }}{\left( k+1\right) ^{\alpha }-k^{\alpha }}\nonumber \\&\qquad \times \frac{1}{\left( I+1\right) ^{\alpha }-I^{\alpha }}\left( M_{G}^{\alpha }\mid _{\frac{k}{I}}^{\frac{k}{I+1}}-M_{G}^{\alpha }\mid _{\frac{k+1}{I}}^{\frac{k+1}{I+1}}\right) . \end{aligned}$$
(79)

Appendix 2

In this appendix, the appropriate fractional element Peclet number \(\hat{P}e\) is formulated, and the matrix form of Galekrin discretization is rewritten with fractional element Peclet number.

Based on the formula of fractional diffusion stiffness matrix in Eq. (15), \(\varvec{K}\left( \alpha ,M\right) \) can be rewritten as

$$\begin{aligned}&\varvec{K}\left( \alpha ,M\right) =\left[ \begin{array}{ccccccc} \lambda _{1} &{} \lambda _{2} &{} \lambda _{3} &{} \cdots &{} \cdots &{} \cdots &{} \lambda _{M-1}\\ \lambda _{1} &{} \lambda _{2} &{} \lambda _{3} &{} \cdots &{} \cdots &{} \cdots &{} \lambda _{M-1}\\ \lambda _{1} &{} \lambda _{2} &{} \lambda _{3} &{} \cdots &{} \cdots &{} \cdots &{} \lambda _{M-1}\\ \vdots &{} \vdots &{} \vdots &{} \vdots &{} \vdots &{} \vdots &{} \vdots \\ \vdots &{} \vdots &{} \vdots &{} \vdots &{} \vdots &{} \vdots &{} \vdots \\ \vdots &{} \vdots &{} \vdots &{} \vdots &{} \vdots &{} \vdots &{} \vdots \\ \lambda _{1} &{} \lambda _{2} &{} \lambda _{3} &{} \cdots &{} \cdots &{} \cdots &{} \lambda _{M-1} \end{array}\right] \nonumber \\&\quad \times \left[ \begin{array}{ccccccc} \omega _{11} &{} -1\\ \omega _{22} &{} \omega _{21} &{} -1 &{} &{} &{} 0\\ \omega _{33} &{} \omega _{32} &{} \omega _{31} &{} -1\\ \vdots &{} \vdots &{} \vdots &{} \vdots &{} \ddots \\ \vdots &{} \vdots &{} \vdots &{} \vdots &{} &{} \ddots \\ \vdots &{} \vdots &{} \vdots &{} \vdots &{} &{} &{} -1\\ \omega _{M-1,1} &{} \omega _{M-1,2} &{} \omega _{M-1,3} &{} \cdots &{} \cdots &{} \cdots &{} \omega _{M-1,M-1} \end{array}\right] .\nonumber \\ \end{aligned}$$
(80)

The stiffness matrix \(\varvec{K}\left( 0,M\right) \) is calculated as

$$\begin{aligned} K_{IJ}\left( 0,M\right) =\left\{ \begin{array}{lll} -0.5, &{} &{} J=I+1\\ 0, &{} &{} J=I\\ 0.5, &{} &{} J=I-1. \end{array}\right. \end{aligned}$$
(81)

Therefore, the matrix form of fractional advection–diffusion equation can be rewritten as

$$\begin{aligned} -\tilde{\varvec{K}}\left( 0,M\right) +\tilde{\varvec{K}}\left( \alpha ,M\right) =\tilde{\varvec{F}}, \end{aligned}$$
(82)

where,

$$\begin{aligned}&\tilde{\varvec{K}}\left( 0,M\right) =\left[ \begin{array}{lllllll} 0 &{} -1\\ 1 &{} 0 &{} -1\\ &{} 1 &{} 0 &{} -1\\ &{} &{} \ddots &{} \ddots &{} \ddots \\ &{} &{} &{} \ddots &{} \ddots &{} \ddots \\ &{} &{} &{} &{} \ddots &{} \ddots &{} -1\\ &{} &{} &{} &{} &{} 1 &{} 0 \end{array}\right] ,\end{aligned}$$
(83)
$$\begin{aligned}&\tilde{\varvec{K}}\left( \alpha ,M\right) =\left[ \begin{array}{ccccccc} \bar{\lambda }_{1} &{} \bar{\lambda }_{2} &{} \bar{\lambda }_{3} &{} \cdots &{} \cdots &{} \cdots &{} \bar{\lambda }_{M-1}\\ \bar{\lambda }_{1} &{} \bar{\lambda }_{2} &{} \bar{\lambda }_{3} &{} \cdots &{} \cdots &{} \cdots &{} \bar{\lambda }_{M-1}\\ \bar{\lambda }_{1} &{} \bar{\lambda }_{2} &{} \bar{\lambda }_{3} &{} \cdots &{} \cdots &{} \cdots &{} \bar{\lambda }_{M-1}\\ \vdots &{} \vdots &{} \vdots &{} \vdots &{} \vdots &{} \vdots &{} \vdots \\ \vdots &{} \vdots &{} \vdots &{} \vdots &{} \vdots &{} \vdots &{} \vdots \\ \vdots &{} \vdots &{} \vdots &{} \vdots &{} \vdots &{} \vdots &{} \vdots \\ \bar{\lambda }_{1} &{} \bar{\lambda }_{2} &{} \bar{\lambda }_{3} &{} \cdots &{} \cdots &{} \cdots &{} \bar{\lambda }_{M-1} \end{array}\right] \nonumber \\&\quad \times \left[ \begin{array}{ccccccc} \omega _{11} &{} -1\\ \omega _{22} &{} \omega _{21} &{} -1 &{} &{} &{} 0\\ \omega _{33} &{} \omega _{32} &{} \omega _{31} &{} -1\\ \vdots &{} \vdots &{} \vdots &{} \vdots &{} \ddots \\ \vdots &{} \vdots &{} \vdots &{} \vdots &{} &{} \ddots \\ \vdots &{} \vdots &{} \vdots &{} \vdots &{} &{} &{} -1\\ \omega _{M-1,1} &{} \omega _{M-1,2} &{} \omega _{M-1,3} &{} \cdots &{} \cdots &{} \cdots &{} \omega _{M-1,M-1} \end{array}\right] ,\nonumber \\ \end{aligned}$$
(84)
$$\begin{aligned}&\tilde{\varvec{F}}=\left[ \begin{array}{c} 0\\ 0\\ 0\\ \vdots \\ \vdots \\ 0\\ \bar{\lambda }_{M}-1 \end{array}\right] , \end{aligned}$$
(85)

where

$$\begin{aligned} \bar{\lambda }_{I}= & {} \frac{2v}{u\triangle x^{\alpha }}\left[ \frac{\varGamma \left( 1+\alpha \right) }{\left( I+1\right) ^{\alpha }-I^{\alpha }}-\frac{\alpha \varGamma \left( 1+\alpha \right) }{\varGamma \left( 1-\alpha \right) \varGamma ^{2}\left( \alpha \right) }\right. \nonumber \\&\quad \times \left. \frac{I^{\alpha }}{\left[ \left( I+1\right) ^{\alpha }-I^{\alpha }\right] ^{2}}M_{G}^{\alpha }\mid _{1}^{\frac{I}{I+1}}\right] =\frac{2v}{u\triangle x^{\alpha }}\vartheta _{I}. \end{aligned}$$
(86)

Here, \(\vartheta _{I}\) always has almost the same value of \(\frac{1}{\varGamma \left( 3-\alpha \right) }\), which can be numerically demonstrated as shown in Fig. 12.

Fig. 12
figure 12

Comparison between \(\vartheta _{I}\) and \(\frac{1}{\varGamma \left( 3-\alpha \right) }\) for different fractional order a \(\alpha =0.3\), b \(\alpha =0.5\) and c \(\alpha =0.7\)

Hence, we assume all the element value of \(\vartheta _{I}\) has the same value of \(\frac{1}{\varGamma \left( 3-\alpha \right) }\) and then all the element value of \(\bar{\lambda }_{I}\) becomes \(\frac{2v}{u\triangle x^{\alpha }\varGamma \left( 3-\alpha \right) }\).

Therefore, the matrix form in Eq. (82) can be rewritten as

$$\begin{aligned}&-\hat{P}e\left[ \begin{array}{lllllll} 0 &{} -1\\ 1 &{} 0 &{} -1\\ &{} 1 &{} 0 &{} -1\\ &{} &{} \ddots &{} \ddots &{} \ddots \\ &{} &{} &{} \ddots &{} \ddots &{} \ddots \\ &{} &{} &{} &{} \ddots &{} \ddots &{} -1\\ &{} &{} &{} &{} &{} 1 &{} 0 \end{array}\right] \nonumber \\&\qquad +\left[ \begin{array}{ccccccc} \omega _{11} &{} -1\\ \omega _{22} &{} \omega _{21} &{} -1 &{} &{} &{} 0\\ \omega _{33} &{} \omega _{32} &{} \omega _{31} &{} -1\\ \vdots &{} \vdots &{} \vdots &{} \vdots &{} \ddots \\ \vdots &{} \vdots &{} \vdots &{} \vdots &{} &{} \ddots \\ \vdots &{} \vdots &{} \vdots &{} \vdots &{} &{} &{} -1\\ \omega _{M-1,1} &{} \omega _{M-1,2} &{} \omega _{M-1,3} &{} \cdots &{} \cdots &{} \cdots &{} \omega _{M-1,M-1} \end{array}\right] \nonumber \\&\quad =\left[ \begin{array}{c} 0\\ 0\\ 0\\ \vdots \\ \vdots \\ 0\\ 1-\hat{P}e \end{array}\right] . \end{aligned}$$
(87)

The matrix form of Eq. (87) is denoted as

$$\begin{aligned} -\hat{P}e\bar{\varvec{K}}\left( 0,M\right) +\bar{\varvec{K}}\left( \alpha ,M\right) =\bar{F}. \end{aligned}$$
(88)

Appendix 3

In this appendix, the detailed formula of stiffness matrices \(\varvec{K}\left( 1,M\right) \) and \(\varvec{K}\left( 1+\alpha ,M\right) \) are derived as follows.

\(\varvec{K}\left( 1,M\right) \) is a tri-diagonal matrix, formulated as

$$\begin{aligned}&K_{IJ}\left( 1,M\right) =\frac{\alpha ^{2}}{2\alpha -1}\frac{1}{\triangle x}\nonumber \\&\quad \times \left\{ \begin{array}{lll} -\frac{J^{2\alpha -1}-I^{2\alpha -1}}{\left( J^{\alpha }-I^{\alpha }\right) ^{2}}, &{} &{} J=I+1\\ \frac{\left( I+1\right) ^{2\alpha -1}-I^{2\alpha -1}}{\left[ \left( I+1\right) ^{\alpha }-I^{\alpha }\right] ^{2}}+\frac{I^{2\alpha -1}-\left( I-1\right) ^{2\alpha -1}}{\left[ I^{\alpha }-\left( I-1\right) ^{\alpha }\right] ^{2}}, &{} &{} J=I\\ -\frac{I^{2\alpha -1}-J^{2\alpha -1}}{\left( I^{\alpha }-J^{\alpha }\right) ^{2}}, &{} &{} J=I-1. \end{array}\right. \end{aligned}$$
(89)

\(\varvec{K}\left( 1+\alpha ,M\right) \) is a lower triangular matrix, formulated as

$$\begin{aligned}&K_{IJ}\left( 1+\alpha ,M\right) =\frac{1}{\varGamma \left( 2-\alpha \right) }\frac{\alpha ^{2}}{\triangle x^{1+\alpha }}\nonumber \\&\quad \times \left\{ \begin{array}{lll} 0, &{} &{} J\ge I+2\\ -\frac{1}{\left( J^{\alpha }-I^{\alpha }\right) ^{2}}\frac{1}{\left( IJ\right) ^{1-\alpha }}, &{} &{} J=I+1\\ \frac{1}{\left[ \left( I+1\right) I\right] ^{1-\alpha }}\left\{ \frac{1}{\left[ \left( I+1\right) ^{\alpha }-I^{\alpha }\right] \left[ I^{\alpha }-\left( I-1\right) ^{\alpha }\right] }+\frac{1}{\left[ \left( I+1\right) ^{\alpha }-I^{\alpha }\right] ^{2}}\right\} , &{} &{} J=I\\ \sum _{m=I-1}^{I}\sum _{n=J-1}^{J}\frac{1}{\left( m+1\right) ^{\alpha }-m^{\alpha }}\frac{1}{\left( n+1\right) ^{\alpha }-n^{\alpha }}\\ \left[ \sum _{p=m}^{m+1}\sum _{q=n}^{n+1}\left( -1\right) ^{p+q-I-J+1}\left( \frac{p-q}{pq}\right) ^{1-\alpha }\right] -\frac{1}{\left[ I^{\alpha }-\left( I-1\right) ^{\alpha }\right] ^{2}}\frac{1}{I^{1-\alpha }\left( I-1\right) ^{1-\alpha }}, &{} &{} J=I-1\\ \sum _{m=I-1}^{I}\sum _{n=J-1}^{J}\frac{1}{\left( m+1\right) ^{\alpha }-m^{\alpha }}\frac{1}{\left( n+1\right) ^{\alpha }-n^{\alpha }}\\ \left[ \sum _{p=m}^{m+1}\sum _{q=n}^{n+1}\left( -1\right) ^{p+q-I-J+1}\left( \frac{p-q}{pq}\right) ^{1-\alpha }\right] , &{} &{} J\le I-2. \end{array}\right. \end{aligned}$$
(90)

From the formula of stiffness \(\varvec{K}\left( 1,M\right) \) , one special case is that \(K_{11}\left( 1,M\right) \) may not exist when the fractional order \(\alpha <0.5\), which reads

$$\begin{aligned} K_{11}\left( 1,M\right) =\frac{\alpha ^{2}}{\triangle x}\left[ \int _{0}^{\triangle x}x^{2\alpha -2}\text {d}x+\int _{\triangle x}^{2\triangle x}x^{2\alpha -2}\text {d}x\right] . \end{aligned}$$
(91)

Therefore, we rewritte the formula of \(K_{11}\left( 1,M\right) \) as

$$\begin{aligned}&K_{11}\left( 1,M\right) \!=\!\frac{\alpha ^{2}}{\triangle x\left( 2\alpha -1\right) \left( 2^{\alpha }-1\right) ^{2}}\nonumber \\&\quad \times \left\{ \begin{array}{lll} \left( 2^{2\alpha -1}\!-\!1\right) +2^{1-2\alpha }\left( 2\alpha -1\right) \left( 2^{\alpha }\!-\!1\right) ^{2}\varepsilon , &{} &{} \alpha \le 0.5\\ \left( 2^{2\alpha -1}-1\right) +\left( 2^{\alpha }-1\right) ^{2}, &{} &{} \alpha >0.5, \end{array}\right. \nonumber \\ \end{aligned}$$
(92)

where \(\varepsilon \) represents the approximation of integration \(\left( \frac{2}{\triangle x}\right) ^{2\alpha -1}\int _{0}^{\triangle x}x^{2\alpha -2}\text {d}x\). In this article, it is recommended by the author to use Gaussian quadrature approximation to calculate \(\varepsilon \) in Eq. (92), and takes three Gaussian points as an example, \(\varepsilon \) is calculated as

$$\begin{aligned}&\varepsilon =\left( \frac{2}{\triangle x}\right) ^{2\alpha -1}\int _{0}^{\triangle x}x^{2\alpha -2}\text {d}x=\int _{-1}^{1}\left( 1+\xi \right) ^{2\alpha -2}\text {d}\xi \nonumber \\&\quad \thickapprox \frac{8}{9}+\frac{5}{9}\left[ \left( 1-\sqrt{0.6}\right) ^{2\alpha -2}+\left( 1+\sqrt{0.6}\right) ^{2\alpha -2}\right] . \end{aligned}$$
(93)

Additionally, the load vector \(\varvec{F}\) is formulated as

$$\begin{aligned} F_{I}=\left\{ \begin{array}{ll@{\quad }l} 0, &{} &{} I<M-1\\ -\frac{1}{2}u+v\left\{ \frac{\varGamma \left( 1+\alpha \right) }{\triangle x^{\alpha }}\frac{1}{M^{\alpha }-\left( M-1\right) ^{\alpha }}-\frac{\alpha \varGamma \left( 1+\alpha \right) }{\triangle x^{\alpha }\varGamma \left( 1-\alpha \right) \varGamma ^{2}\left( \alpha \right) }\frac{\left( M-1\right) ^{\alpha }}{\left[ M^{\alpha }-\left( M-1\right) ^{\alpha }\right] ^{2}}M_{G}^{\alpha }\mid _{1}^{\frac{M-1}{M}}\right\} \\ +u\gamma \frac{\alpha ^{2}}{2\alpha -1}\frac{1}{\triangle x}\frac{M^{2\alpha -1}-\left( M-1\right) ^{2\alpha -1}}{\left[ M^{\alpha }-\left( M-1\right) ^{\alpha }\right] ^{2}}-v\gamma \frac{1}{\varGamma \left( 2-\alpha \right) }\frac{\alpha ^{2}}{\triangle x^{1+\alpha }}\frac{1}{\left[ M^{\alpha }-\left( M-1\right) ^{\alpha }\right] ^{2}}\frac{1}{\left[ M\left( M-1\right) \right] ^{1-\alpha }}, &{} &{} I=M-1. \end{array}\right. \end{aligned}$$
(94)

Appendix 4

In this appendix, the detailed derivations of the artificial viscosity coefficient of the enriched Petrov–Galerkin finite element method to fractional advection–diffusion equation is proposed.

The load vector residual is defined as

$$\begin{aligned} \varvec{R}=\left\| \varvec{F}^{*}-\varvec{F}\right\| , \end{aligned}$$
(95)

Considering the first term in Eq. (95), the load vector at first node reads

$$\begin{aligned} F_{1}= & {} 0,\end{aligned}$$
(96)
$$\begin{aligned} F_{1}^{*}= & {} \left[ -uK_{11}\left( 0,M\right) +vK_{11}\left( \alpha ,M\right) \right. \nonumber \\&\left. +u\gamma K_{11}\left( 1,M\right) -v\gamma K_{11}\left( 1+\alpha ,M\right) \right] \phi _{1}^{\text {ex}}\nonumber \\&+ \left[ -uK_{12}\left( 0,M\right) +vK_{12}\left( \alpha ,M\right) +u\gamma K_{12}\left( 1,M\right) \right. \nonumber \\&\left. -v\gamma K_{12}\left( 1+\alpha ,M\right) \right] \phi _{2}^{\text {ex}} \end{aligned}$$
(97)

Based on the formulas of stiffness matrix in Eq. (15), (81), (89), (90), and exact nodal solution in Eq. (37), the details in Eq. (97) are formulated as

$$\begin{aligned}&K_{11}\left( 0,M\right) =0,\end{aligned}$$
(98)
$$\begin{aligned}&K_{12}\left( 0,M\right) =-\frac{1}{2},\end{aligned}$$
(99)
$$\begin{aligned}&K_{11}\left( \alpha ,M\right) =\frac{\varGamma \left( 1+\alpha \right) }{\triangle x^{\alpha }}\frac{2^{\alpha }}{2^{\alpha }-1}\nonumber \\&\quad -\,\frac{\alpha \varGamma \left( 1+\alpha \right) }{\triangle x^{\alpha }\varGamma \left( 1-\alpha \right) \varGamma ^{2}\left( \alpha \right) }\frac{2^{\alpha }}{\left( 2^{\alpha }-1\right) ^{2}}M_{G}^{\alpha }\mid _{1}^{0.5},\end{aligned}$$
(100)
$$\begin{aligned}&K_{12}\left( \alpha ,M\right) =-\frac{\varGamma \left( 1+\alpha \right) }{\triangle x^{\alpha }}\frac{1}{2^{\alpha }-1}\nonumber \\&\quad +\,\frac{\alpha \varGamma \left( 1+\alpha \right) }{\triangle x^{\alpha }\varGamma \left( 1-\alpha \right) \varGamma ^{2}\left( \alpha \right) }\frac{1}{\left( 2^{\alpha }-1\right) ^{2}}M_{G}^{\alpha }\mid _{1}^{0.5}, \end{aligned}$$
(101)
$$\begin{aligned}&K_{11}\left( 1,M\right) =\frac{\alpha ^{2}}{\triangle x\left( 2\alpha -1\right) \left( 2^{\alpha }-1\right) ^{2}}\nonumber \\&\quad \times \left\{ \begin{array}{lll} \left( 2^{2\alpha {-}1}{-}1\right) +2^{1-2\alpha }\left( 2\alpha -1\right) \left( 2^{\alpha }{-}1\right) ^{2}\varepsilon , &{} &{} \alpha \le 0.5\\ \left( 2^{2\alpha {-}1}{-}1\right) +\left( 2^{\alpha }{-}1\right) ^{2}, &{} &{} \alpha >0.5, \end{array}\right. \nonumber \\ \end{aligned}$$
(102)
$$\begin{aligned}&K_{12}\left( 1,M\right) =-\frac{2^{2\alpha -1}-1}{\left( 2^{\alpha }-1\right) ^{2}}\frac{\alpha ^{2}}{2\alpha -1}\frac{1}{\triangle x}, \end{aligned}$$
(103)
$$\begin{aligned}&K_{11}\left( 1+\alpha ,M\right) =\frac{1}{\varGamma \left( 2-\alpha \right) }\frac{2^{2\alpha -1}}{\left( 2^{\alpha }-1\right) ^{2}}\frac{\alpha ^{2}}{\triangle x^{1+\alpha }}, \end{aligned}$$
(104)
$$\begin{aligned}&K_{12}\left( 1+\alpha ,M\right) =-\frac{1}{\varGamma \left( 2-\alpha \right) }\frac{2^{\alpha -1}}{\left( 2^{\alpha }-1\right) ^{2}}\frac{\alpha ^{2}}{\triangle x^{1+\alpha }}, \end{aligned}$$
(105)
$$\begin{aligned}&\phi _{1}^{\text {ex}}=\frac{E_{\alpha ,1}\left( \frac{u}{v}\triangle x^{\alpha }\right) -1}{E_{\alpha ,1}\left( \frac{u}{v}\right) -1}, \end{aligned}$$
(106)
$$\begin{aligned}&\phi _{2}^{\text {ex}}=\frac{E_{\alpha ,1}\left( \frac{u}{v}\left( 2\triangle x\right) ^{\alpha }\right) -1}{E_{\alpha ,1}\left( \frac{u}{v}\right) -1}. \end{aligned}$$
(107)

Substituting all the formulas above into Eq. (97), by expanding and minimizing \(R_{1}=\left| F_{1}^{*}-F_{1}\right| =0\), the artificial viscosity coefficient is formulated as

$$\begin{aligned} \gamma =\frac{\triangle x}{2}\frac{\hat{P}e\left( 2^{\alpha }-1\right) ^{2}\varTheta +\alpha \varGamma \left( 3-\alpha \right) \left( 2^{\alpha }-\varTheta \right) \left[ \left( 2^{\alpha }-1\right) \varGamma \left( \alpha \right) -\frac{\alpha sin\left( \pi \alpha \right) }{\pi }M_{G}^{\alpha }\mid _{1}^{0.5}\right] }{\left[ 2^{\alpha -2}\left( 2^{\alpha }-\varTheta \right) \left( 2-\alpha \right) -\hat{P}e\chi \right] \alpha ^{2}}, \end{aligned}$$
(108)

where \(\varTheta \) is a function of fractional Peclet number defined as

$$\begin{aligned} \varTheta =\frac{E_{\alpha ,1}\left( 2^{1+\alpha }\hat{P}e/\varGamma \left( 3-\alpha \right) \right) -1}{E_{\alpha ,1}\left( 2\hat{P}e/\varGamma \left( 3-\alpha \right) \right) -1}. \end{aligned}$$
(109)

\(\chi \) is a function of parameter \(\varepsilon \) introduced by the special case of \(K_{11}\left( 1,M\right) \) as shown in “Appendix 3”, and defined as

$$\begin{aligned} \chi =\frac{1}{2\alpha -1}\left\{ \begin{array}{lll} \left( 2^{2\alpha -1}-1\right) \left( 1-\varTheta \right) \\ \quad +2^{1-2\alpha }\left( 2\alpha -1\right) \left( 2^{\alpha }-1\right) ^{2}\varepsilon , &{} &{} \alpha \le 0.5\\ \left( 2^{2\alpha -1}-1\right) \left( 1-\varTheta \right) \\ \quad +\left( 2^{\alpha }-1\right) ^{2}, &{} &{} \alpha >0.5. \end{array}\right. \end{aligned}$$
(110)

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Luan, S., Lian, Y., Ying, Y. et al. An enriched finite element method to fractional advection–diffusion equation. Comput Mech 60, 181–201 (2017). https://doi.org/10.1007/s00466-017-1400-9

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s00466-017-1400-9

Keywords

Navigation