Table of Contents Author Guidelines Submit a Manuscript
Mathematical Problems in Engineering
Volume 2017, Article ID 6362505, 9 pages
https://doi.org/10.1155/2017/6362505
Research Article

An Efficient Algorithm with Stabilized Finite Element Method for the Stokes Eigenvalue Problem

School of Mathematics Science, Huaqiao University, Quanzhou 362021, China

Correspondence should be addressed to Zhifeng Weng; moc.361@htamwfz

Received 5 September 2017; Revised 12 November 2017; Accepted 23 November 2017; Published 31 December 2017

Academic Editor: Vassilios C. Loukopoulos

Copyright © 2017 Zhifeng Weng and Yaoxiong Cai. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

This paper provides a two-space stabilized mixed finite element scheme for the Stokes eigenvalue problem based on local Gauss integration. The two-space strategy contains solving one Stokes eigenvalue problem using the finite element pair and then solving an additional Stokes problem using the finite element pair. The postprocessing technique which increases the order of mixed finite element space by using the same mesh can accelerate the convergence rate of the eigenpair approximations. Moreover, our method can save a large amount of computational time and the corresponding convergence analysis is given. Finally, numerical results are presented to confirm the theoretical analysis.

1. Introduction

The Stokes eigenvalue problem is one of the most important eigenvalue problems and plays an important role in the stability analysis of nonlinear partial differential equations [1]. The eigenvalue problems are used in many application areas: structural mechanics and fluid mechanics. Thus, development of the efficient numerical methods for studying the eigenvalue problems has practical meanings and has been noticed by many researchers. At the time of writing, numerous works are devoted to these problems (see [210] and the references cited therein).

Many effective postprocessing strategies that improve the convergence rate for the approximations of the eigenvalue problems by the finite element methods have been well developed. The two-grid method is one of these efficient postprocessing methods. The basic idea of two-grid scheme is first introduced by Xu [11, 12] for the nonsymmetric and nonlinear elliptic problems. Hence, it can be seen as a postprocessing technique and can take less CPU time compared to the one grid methods. To the best of our knowledge, some details of the two-grid scheme can be found in the works of Xu and Zhou [13], Chien and Jeng [14, 15], Chen et al. [7, 16], Hu and Cheng [17], Yang et al. [18, 19], Huang et al. [8], and Weng et al. [20, 21]. The two-space method is actually the iterative Galerkin method, which was first used for solving integral equation eigenvalue problems by Sloan [22] and differential equation eigenvalue problems by Lin and Xie [23]. Particularly, Racheva and Andreev [24] have proposed a postprocessing method for the -order self-adjoint eigenvalue problems by two-grid method or the two-space method. A similar method has been given for the Stokes eigenvalue problem [7, 25], elliptic eigenvalue problem [16], and the biharmonic eigenvalue problem [26] by mixed finite element methods.

In fact, two-space method [2729] can be cast in the framework of Xu’s work regarding the two-grid method. However, the two-space method is different from the two-grid method. This two-space method consists in solving the original Stokes eigenvalue problem in the -order mixed finite element space and one additional Stokes source problem in an augmented mixed finite element space by a -order mixed finite element space on the same mesh. Besides, the two-space method only needs one mesh size while the two-grid method needs two mesh sizes, a coarse mesh, and a fine mesh. In fact, the two-space method can avoid the discussion on the relation of the coarse and fine meshes. For this reason, in the present paper we establish a two-space discretization scheme for the Stokes eigenvalue problem.

Recently, more attention has been paid to the lowest equal order finite element pairs for simulating the incompressible flow. The lowest equal order finite element pairs offer some computational advances; for example, they are simple and have practical uniform data structure and adequate accuracy, because they show an identical degree distribution for both the velocity and pressure. Moreover, they are of practical importance in scientific computation owing to their very convenient computational cost. However, the lowest equal order mixed finite element pairs do not satisfy the inf-sup condition. Numerical tests show that the violation of the inf-sup condition often brings about unphysical pressure oscillations. In order to avoid the instability problem, the stabilized finite element methods are applied to the incompressible flow. Therefore, a lot of work focuses on stabilization (see [3037]) of the lowest equal order pairs. Particularly, based on the work of Bochev et al. [30], Li et al. [31, 32] used the projection of the pressure onto the piecewise constant space to add the stabilized term for element and Zheng et al. [35] used the projection of the pressure-gradient onto the piecewise constant space to add the stabilized term for element.

Influenced by the work mentioned above, the paper focuses on the method, which combines two-space discretization scheme with a stabilized finite element method based on local Gauss integration technique for the Stokes eigenvalue problem. The paper is organized as follows. In Section 2, we introduce the studied Stokes eigenvalue problem and the notations and some well-known results used throughout this paper. Some stabilized finite element strategies based on two local Gauss integrations are recalled in Section 3. In Section 4, a two-space stabilized finite element algorithm is constructed and its error estimates are discussed. In Section 5, numerical experiments are reported for illustrating the theoretical results and the high efficiency of the proposed method. Finally, we will conclude our presentation in Section 6 with a few comments and also possible future research topics.

2. Preliminaries

In this paper, we consider the following Stokes eigenvalue problem:where is a bounded and convex domain with a Lipschitz-continuous boundary , represents the pressure, is the velocity vector, and is the eigenvalue.

We shall introduce the following Hilbert spaces: The spaces , , are equipped with the -scalar product and -norm . The norm and seminorm in are denoted by and , respectively. The space is equipped with the norm or its equivalent norm due to Poincaré inequality. Spaces consisting of vector-valued functions are denoted in boldface. Furthermore, the norm in the space dual to is given byTherefore, we define the following bilinear forms , , and on , , and , respectively, by and a generalized bilinear form on ; that is,

With the above notations, the variational formulation of problem (1) reads as follows: Find with , such thatFrom [1], we know that eigenvalue problem (6) has an eigenvalue sequence and corresponding eigenvectors with the orthogonal property .

Let

Moreover, the bilinear form satisfies the inf-sup condition for all where is a constant depending only on . Therefore, the generalized bilinear form satisfies the continuity property and coercive conditionwhere and are the positive constants depending only on . Throughout the paper we use or to denote a generic positive constant whose value may change from place to place, which remains independent of the mesh parameter.

3. A Stabilized Mixed Finite Element Method

From now on, is a real positive parameter tending to . The finite element subspaces of are characterized by , a partitioning of into triangles with the mesh size , assumed to be uniformly regular in the usual sense [38]. Then we define them as follows: where represents the set of all polynomials on of degree less than and denotes the space of bubble functions. The bubble functions are defined as follows: where are area coordinates on , . The area coordinate is also known as a triangle barycentre coordinate, where the three components are of the ratio between the area of the three triangles and the area of the mother triangle.

It is known that this choice of the approximate spaces or satisfies the inf-sup condition in [38], but this choice of the approximate spaces or does not satisfy the inf-sup condition [30, 32, 35]. As a consequence, we give a stabilized finite element approximation based on local Gauss integration technique (see [32, 35]). The idea is as follows.

Let be the standard -projection: where .

The projection operator has the following properties: The stabilized bilinear terms are used byand the stabilization term is given by

The stabilized term which is defined by local Gaussian quadrature can be rewritten as where denotes a Gaussian quadrature over which is exact for polynomials of degree . In particular, when , the trial function is projected to the piecewise constant space. Besides, the stabilized term can be rewritten as where the trial function must be projected to when for any . Indeed, Becker and Hansbo have found [33] that the stabilized methods of [30, 32] are identical from a numerical point of view for these low-order approximations.

By adding the stabilization term into the generalized bilinear form , we define

Then the corresponding discrete variational formulation for the Stokes eigenvalue problem reads as follows: find with , such thatand find with , such that

Remark 1. For the () pair which satisfy inf-sup condition, there are points of difference between them. The stabilized method in this article only adds the stabilized term with respect to the pressure space. However, the method has the implicit stabilized term in the velocity space.

With and thanks to the positive definiteness of , we deduce that the discrete eigenvalues are positive. Let the eigenvalue of (21) be ordered as follows: and let us consider the corresponding eigenfunctions where denotes the dimension of the finite element space.

Similarly, let be the eigenspace associated with ; that is,

For (22) with pairs, it can be given similarly. The corresponding nature of the eigenvalues is omitted for the sake of simplicity.

The next theorem shows the continuity property and the weak coercivity property of the bilinear form for the finite element space in [35] and for the finite element space in [30, 32].

Theorem 2. For all , there exist positive constants and , independent of , such that Moreover, for all , there exist positive constants and , independent of , such that

The next theorem contains the convergence result of eigenfunctions and eigenvalues for the Stokes eigenvalue problem in [8, 20].

Theorem 3. With belonging to and being the exact solution of (6), one deduces that there exists a discrete eigenpair of (21) which satisfies the following error estimates:

Furthermore, if the exact solution , then of problem (22) satisfies

4. Two-Space Stabilized Finite Element Scheme and Error Estimates

In this section, we shall present a two-space stabilized finite element algorithm to reduce the computational cost. The two-space stabilized finite element approximation consists of three steps.

Step 1. On the mesh size , solve the following Stokes eigenvalue problem by pair and find and , such that, for all ,

Step 2. On the same mesh size , solve the following Stokes problem by pair and find such that for all

Step 3. Compute the eigenvalue by the Rayleigh quotientwhere

Next, we will study the convergence of the two-space stabilized finite element solution. To do this, we define the Galerkin projection operator byBy Theorem 2, is well defined and the following approximation properties are fulfilled in [20].

Lemma 4. For all , one has

The following identity that relates the errors in the eigenvalue and eigenvector can be found in [3].

Lemma 5. Let be an eigenpair of (6); for any and , one has

The next theorem provides the error estimates for our two-space scheme.

Theorem 6. Let be the th discrete eigenpair. Then the th eigenpair of the Stokes operator is such that

Proof. Denoted by , subtracting (6) from (31), we derive from (33)Let in (38), by using Theorem 2, Sobolev embedding theorem, and Theorems 3, we obtain Combining the triangle inequality with Lemma 4, we deduceand finally we obtain (36).
Moreover, using (32) and Lemma 5, we have Taking the norm and using (15) and (36), we conclude the proof that is

Remark 7. From Theorem 3, for the usual stabilized finite element solution which involves solving a Stokes eigenvalue problem with mesh size , we have the following error estimates:

Furthermore, if we use the two-space stabilized finite element method, then we get the convergence rate of the same order as the usual stabilized finite element method from Theorem 6. However, our method is more efficient than the stabilized finite element scheme in the same mesh because our method for solving Stokes eigenvalue problem is to compute an initial approximation based on a lower number of nodes, which takes less CPU time.

Remark 8. For the two-space algorithm with pair which satisfies inf-sup condition, we obtain the same result. The procedure of the two-space method with pair could be described in the following manner: Firstly, we solve the following Stokes eigenvalue problem by () pair on the mesh size ; then, we should solve the new Stokes problem by pair on the same mesh size . Finally we can compute the eigenvalue by the Rayleigh quotient.

5. Numerical Experiments

In this section we present numerical results to check the theoretical analysis contained in Theorem 6. Our goal is to confirm the theoretical results of the new two-space stabilized finite element method for the two-dimensional Stokes eigenvalue approximated by the equal order finite element pairs based on local Gauss integration.

In our numerical experiments, is the unit square domain in . The domain is uniformly divided by the triangulations of mesh size . Here, we just consider the first eigenvalue of the Stokes eigenvalue problem for the sake of simplicity. Following [4], we employ the approximation as the reference solution for the first eigenvalue. Note that in these computations we set .

When solving the Stokes problem with a mesh size , we need the solutions and generated by a lower finite element pair . To do this we interpolate the solutions and onto the grid with the same mesh size , but increasing the order of the mixed finite element space. In conclusion, the solution of the two-space method is obtained by one simple eigenvalue problem by a lower finite element pair and one time interpolation by a higher finite element pair .

Our goal in this test is to validate the merit of the two-space method as compared with the stabilized method and the two-space method with pair. The eigenvalue approximation , the eigenvalue error, the convergence rates, and the CPU time for the stabilized mixed finite element methods for different values of are tabulated in Tables 1, 2, and 3. From Tables 1, 2, and 3, we can see that the three methods work well and keep the convergence rates just as predicted by the theoretical analysis, but our two-space method can take less CPU time. For the two-space method with pair, the two-space method with pair approximates the velocity variable with a lower number of nodes, so our method can save a lot of time.

Table 1: Relative error and convergence rate for pair.
Table 2: Relative error and convergence rate of two-space method with pair.
Table 3: Relative error and convergence rate of two-space method with pair.

Next numerical test is about the second, third, and fourth eigenvalues . The reference values are computed over a fine mesh and the results are . Then, in Figure 1, we exhibit the convergence rate as has been predicted in Theorem 6 with the two-space method.

Figure 1: The convergence rate of the eigenvalue for on the unit square with the two-space method.

Moreover, we give two plots of numerical solutions of two kinds of two-space schemes at the mesh in Figure 2 for the details. Figure 2 shows the stability of two schemes.

Figure 2: Plots of the velocity and pressure at : numerical solution of two-space method with (a–c) and numerical solution of two-space method with (d–f) for .

6. Conclusions

In this paper, we presented the two-space algorithm for the Stokes eigenvalue problem discretized by stabilized mixed finite element scheme, based on local Gauss integration technique. The main feature of our method is to combine two equal order stabilized methods, then use the first-order mixed finite element space to solve the original Stokes eigenvalue problem, and solve the Stokes source problem in the second-order mixed finite element space on the same mesh. Moreover, the related error estimates have been derived. Finally, numerical tests show that the two-space stabilized mixed finite element method is numerically efficient for solving the Stokes eigenvalue problem. The two-space algorithm can achieve the same accuracy as the stabilized finite element solution as the stabilized method by taking less CPU time. Obviously, this method can be extended to the case of three dimensions easily. And there are some open questions including the possible extension of the method to other linear and nonlinear eigenvalue problems.

Conflicts of Interest

The authors declare that they have no conflicts of interest.

Acknowledgments

The work of Z. F. Weng is partially supported by NSF of China (nos. 11701197, 11701196), the Scientific Research Foundation of Huaqiao University (no. 15BS307), and Promotion Program for Young and Middle-Aged Teacher in Science and Technology Research of Huaqiao University (ZQN-YX502).

References

  1. I. Babuska and J. E. Osborn, “Eigenvalue problems,” in Hand-Book of Numerical Analysis, Volume II, Finite Element Method (Part I), P. G. Ciarlet and J. L. Lions, Eds., pp. 641–787, North-Holland Publishing, Amsterdam, The Netherlands, 1991. View at Google Scholar
  2. B. Mercier, J. Osborn, J. Rappaz, and P.-A. Raviart, “Eigenvalue approximation by mixed and hybrid methods,” Mathematics of Computation, vol. 36, no. 154, pp. 427–453, 1981. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  3. I. Babuska and J. E. Osborn, “Finite element-Galerkin approximation of the eigenvalues and eigenvectors of selfadjoint problems,” Mathematics of Computation, vol. 52, no. 186, pp. 275–297, 1989. View at Publisher · View at Google Scholar · View at MathSciNet
  4. W. Chen and Q. Lin, “Approximation of an eigenvalue problem associated with the Stokes problem by the stream function-vorticity-pressure method,” Applications of Mathematics, vol. 51, no. 1, pp. 73–88, 2006. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  5. X. Yin, H. Xie, S. Jia, and S. Gao, “Asymptotic expansions and extrapolations of eigenvalues for the Stokes problem by mixed finite element methods,” Journal of Computational and Applied Mathematics, vol. 215, no. 1, pp. 127–141, 2008. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  6. C. Lovadina, M. Lyly, and R. Stenberg, “A posteriori estimates for the Stokes eigenvalue problem,” Numerical Methods for Partial Differential Equations, vol. 25, no. 1, pp. 244–257, 2009. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  7. H. Chen, S. Jia, and H. Xie, “Postprocessing and higher order convergence for the mixed finite element approximations of the Stokes eigenvalue problems,” Applications of Mathematics, vol. 54, no. 3, pp. 237–250, 2009. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  8. P. Huang, Y. He, and X. Feng, “Two-level stabilized finite element method for Stokes eigenvalue problem,” Applied Mathematics and Mechanics, vol. 33, no. 5, pp. 621–630, 2012. View at Publisher · View at Google Scholar · View at MathSciNet
  9. F. Ahmad, E. S. Al-Aidarous, D. A. Alrehaili, S. Ekström, I. Furci, and S. Serra-Capizzano, “Are the eigenvalues of preconditioned banded symmetric Toeplitz matrices known in almost closed form?” Numerical Algorithms, pp. 1–27, 2017. View at Publisher · View at Google Scholar
  10. S. Ekström, C. Garoni, and S. Serra-Capizzano, “Are the eigenvalues of banded symmetric toeplitz matrices known in almost closed form?” Experimental Mathematics, pp. 1–10, 2017. View at Publisher · View at Google Scholar
  11. J. Xu, “A novel two-grid method for semilinear elliptic equations,” SIAM Journal on Scientific Computing, vol. 15, no. 1, pp. 231–237, 1994. View at Publisher · View at Google Scholar · View at MathSciNet
  12. J. Xu, “Two-grid discretization techniques for linear and nonlinear PDEs,” SIAM Journal on Numerical Analysis, vol. 33, no. 5, pp. 1759–1777, 1996. View at Publisher · View at Google Scholar · View at MathSciNet
  13. J. Xu and A. Zhou, “A two-grid discretization scheme for eigenvalue problems,” Mathematics of Computation, vol. 70, no. 233, pp. 17–25, 2001. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  14. C. S. Chien and B. W. Jeng, “A two-grid finite element discretization scheme for nonlinear eigenvalue problems,” Computational Methods, pp. 1951–1955, 2006. View at Google Scholar
  15. C.-S. Chien and B.-W. Jeng, “A two-grid discretization scheme for semilinear elliptic eigenvalue problems,” SIAM Journal on Scientific Computing, vol. 27, no. 4, pp. 1287–1304, 2006. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  16. H. Chen, S. Jia, and H. Xie, “Postprocessing and higher order convergence for the mixed finite element approximations of the eigenvalue problem,” Applied Numerical Mathematics, vol. 61, no. 4, pp. 615–629, 2011. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  17. X. Hu and X. Cheng, “Acceleration of a two-grid method for eigenvalue problems,” Mathematics of Computation, vol. 80, no. 275, pp. 1287–1301, 2011. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  18. Y. Yang and X. Fan, “Generalized Rayleigh quotient and finite element two-grid discretization schemes,” Science China Mathematics, vol. 52, no. 9, pp. 1955–1972, 2009. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  19. Y. Yang and H. Bi, “Two-grid finite element discretization schemes based on shifted-inverse power method for elliptic eigenvalue problems,” SIAM Journal on Numerical Analysis, vol. 49, no. 4, pp. 1602–1624, 2011. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  20. Z. Weng, J. Z. Yang, and X. Lu, “Two-level quadratic equal-order stabilized method for the Stokes eigenvalue problem,” International Journal of Computer Mathematics, vol. 92, no. 2, pp. 337–348, 2015. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  21. Z. Weng, X. Feng, and S. Zhai, “Investigations on two kinds of two-grid mixed finite element methods for the elliptic eigenvalue problem,” Computers & Mathematics with Applications, vol. 64, no. 8, pp. 2635–2646, 2012. View at Publisher · View at Google Scholar · View at MathSciNet
  22. I. H. Sloan, “Iterated Galerkin method for eigenvalue problems,” SIAM Journal on Numerical Analysis, vol. 13, no. 5, pp. 753–760, 1976. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  23. Q. Lin and G. Q. Xie, “Acceleration of FEA for eigenvalue problems,” Science Bulletin, vol. 26, pp. 449–452, 1981. View at Google Scholar
  24. M. R. Racheva and A. B. Andreev, “Superconvergence postprocessing for eigenvalues,” Computational Methods in Applied Mathematics, vol. 2, no. 2, pp. 171–185, 2002. View at Google Scholar · View at MathSciNet
  25. H. Xie and X. Yin, “Acceleration of stabilized finite element discretizations for the Stokes eigenvalue problem,” Advances in Computational Mathematics, vol. 41, no. 4, pp. 799–812, 2015. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  26. A. B. Andreev, R. D. Lazarov, and M. R. Racheva, “Postprocessing and higher order convergence of the mixed finite element approximations of biharmonic eigenvalue problems,” Journal of Computational and Applied Mathematics, vol. 182, no. 2, pp. 333–349, 2005. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  27. J. Wu, P. Huang, X. Feng, and D. Liu, “An efficient two-step algorithm for steady-state natural convection problem,” International Journal of Heat and Mass Transfer, vol. 101, pp. 387–398, 2016. View at Publisher · View at Google Scholar · View at Scopus
  28. J. Wu, D. Liu, X. Feng, and P. Huang, “An efficient two-step algorithm for the stationary incompressible magnetohydrodynamic equations,” Applied Mathematics and Computation, vol. 302, pp. 21–33, 2017. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  29. P. Huang, X. Feng, and Y. He, “An efficient two-step algorithm for the incompressible flow problem,” Advances in Computational Mathematics, vol. 41, no. 6, pp. 1059–1077, 2015. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  30. P. B. Bochev, C. R. Dohrmann, and M. Gunzburger, “Stabilization of low-order mixed finite elements for the Stokes equations,” SIAM Journal on Numerical Analysis, vol. 44, no. 1, pp. 82–101, 2006. View at Publisher · View at Google Scholar · View at MathSciNet
  31. J. Li, Y. He, and Z. Chen, “A new stabilized finite element method for the transient Navier-Stokes equations,” Computer Methods Applied Mechanics and Engineering, vol. 197, no. 1–4, pp. 22–35, 2007. View at Publisher · View at Google Scholar · View at MathSciNet
  32. J. Li and Y. He, “A stabilized finite element method based on two local Gauss integrations for the Stokes equations,” Journal of Computational and Applied Mathematics, vol. 214, no. 1, pp. 58–65, 2008. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  33. R. Becker and P. Hansbo, “A simple pressure stabilization method for the Stokes equation,” Communications in Numerical Methods in Engineering, vol. 24, no. 11, pp. 1421–1430, 2008. View at Publisher · View at Google Scholar · View at Scopus
  34. H. Zheng, Y. Hou, F. Shi, and L. Song, “A finite element variational multiscale method for incompressible flows based on two local Gauss integrations,” Journal of Computational Physics, vol. 228, no. 16, pp. 5961–5977, 2009. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  35. H. Zheng, L. Shan, and Y. Hou, “A quadratic equal-order stabilized method for stokes problem based on two local gauss integrations,” Numerical Methods for Partial Differential Equations, vol. 26, no. 5, pp. 1180–1190, 2010. View at Publisher · View at Google Scholar · View at Scopus
  36. T. Zhang and H. Liang, “Decoupled stabilized finite element methods for the Boussinesq equations with temperature-dependent coefficients,” International Journal of Heat and Mass Transfer, vol. 110, pp. 151–165, 2017. View at Publisher · View at Google Scholar · View at Scopus
  37. T. Zhang, Y. Qian, and Y. HuangFu, “Two-level finite element variational multiscale method based on bubble functions for the steady incompressible MHD flow,” International Journal of Computer Mathematics, vol. 94, no. 3, pp. 515–535, 2017. View at Publisher · View at Google Scholar · View at MathSciNet
  38. P. G. Ciarlet, The Finite Element Method for Elliptic Problems, North-Holland Publishing, Amsterdam, The Netherlands, 1978. View at MathSciNet