Abstract
We introduce and analyze a very simple but efficient postprocessing technique for improving the accuracy of the continuous Galerkin (CG) time stepping method for nonlinear first-order initial value problems with smooth and singular solutions. The key idea of the postprocessing technique is to add a higher order Lobatto polynomial of degree \(k+1\) to the CG solution of degree k. We first establish global superconvergent error bounds for the postprocessed CG approximations over arbitrary time partitions, and as a by-product, it is shown that the convergence rates of the \(L^2\)-, \(H^1\)- and \(L^\infty \)-estimates for the CG method with quasi-uniform meshes for smooth solutions are improved by one order. Moreover, for solutions with initial singularities, we prove that the optimal global error estimates and nodal superconvergent estimate can be obtained for the CG method with graded meshes, and after postprocessing, the convergence rates of the \(L^2\)-, \(H^1\)- and \(L^\infty \)-estimates are also improved by one order. As an application, we apply the superconvergent postprocessing technique to the CG time discretization of nonlinear parabolic equations. Numerical examples are presented to verify the theoretical results.
Similar content being viewed by others
Data Availability
Data sharing is not applicable to this paper as no datasets were generated or analysed during the current study.
References
Ahmed, N., Matthies, G.: Higher order continuous Galerkin–Petrov time stepping schemes for transient convection–diffusion–reaction equations. ESAIM Math. Model. Numer. Anal. 49, 1429–1450 (2015)
Akrivis, G., Makridakis, C.: Galerkin time-stepping methods for nonlinear parabolic equations. M2AN Math. Model. Numer. Anal. 38, 26–289 (2004)
Antonietti, P.F., Mazzieri, I., Dal Santo, N., Quarteroni, A.: A high-order discontinuous Galerkin approximation to ordinary differential equations with applications to elastodynamics. IMA J. Numer. Anal. 38, 1709–1734 (2018)
Aziz, A.K., Monk, P.: Continuous finite elements in space and time for the heat equation. Math. Comput. 52, 255–274 (1989)
Baccouch, M.: Analysis of a posteriori error estimates of the discontinuous Galerkin method for nonlinear ordinary differential equations. Appl. Numer. Math. 106, 129–153 (2016)
Baccouch, M.: A posteriori error estimates and adaptivity for the discontinuous Galerkin solutions of nonlinear second-order initial-value problems. Appl. Numer. Math. 121, 18–37 (2017)
Brenner, S.C., Scott, L.R.: The Mathematical Theory of Finite Element Methods, Texts in Applied Mathematics, vol. 15, 3rd edn. Springer, New York (2008)
Cao, W.X., Zhang, Z.M., Zou, Q.S.: Finite volume superconvergence approximation for one-dimensional singularly perturbed problems. J. Comput. Math. 31, 488–508 (2013)
Celiker, F., Zhang, Z.M., Zhu, H.Q.: Nodal superconvergence of SDFEM for singularly perturbed problems. J. Sci. Comput. 50, 405–433 (2012)
Ciarlet, P.G.: The Finite Element Method for Elliptic Problems, Studies in Mathematics and Its Applications, vol. 4. North-Holland Publishing Co., Amsterdam (1978)
Delfour, M., Hager, W., Trochu, F.: Discontinuous Galerkin methods for ordinary differential equations. Math. Comput. 36, 455–473 (1981)
Estep, D., French, D.: Global error control for the continuous Galerkin finite element method for ordinary differential equations. RAIRO Modél. Math. Anal. Numér. 28, 815–852 (1994)
Guo, B.Y., Wang, Z.Q.: Legendre-Gauss collocation methods for ordinary differential equations. Adv. Comput. Math. 30, 249–280 (2009)
Holm, B., Wihler, T.P.: Continuous and discontinuous Galerkin time stepping methods for nonlinear initial value problems with application to finite time blow-up. Numer. Math. 138, 767–799 (2018)
Huang, Q.M., Xu, X.X., Brunner, H.: Continuous Galerkin methods on quasi-geometric meshes for delay differential equations of pantograph type. Discrete Contin. Dyn. Syst. 36, 5423–5443 (2016)
Hulme, B.L.: One-step piecewise polynomial Galerkin methods for initial value problems. Math. Comput. 26, 415–426 (1972)
Hulme, B.L.: Discrete Galerkin and related one-step methods for ordinary differential equations. Math. Comput. 26, 881–891 (1972)
Kyza, I., Metcalfe, S., Wihler, T.P.: \(hp\)-adaptive Galerkin time stepping methods for nonlinear initial value problems. J. Sci. Comput. 75, 111–127 (2018)
McLean, W., Mustapha, K.: A second-order accurate numerical method for a fractional wave equation. Numer. Math. 105, 481–510 (2007)
Meng, T.T., Yi, L.J.: An \(h\)-\(p\) version of the continuous Petrov–Galerkin method for nonlinear delay differential equations. J. Sci. Comput. 74, 1091–1114 (2018)
Schieweck, F.: A-stable discontinuous Galerkin–Petrov time discretization of higher order. J. Numer. Math. 18, 25–57 (2010)
Schötzau, D., Schwab, C.: An \(hp\) a-priori error analysis of the DG time-stepping method for initial value problems. Calcolo 37, 207–232 (2000)
Schötzau, D., Wihler, T.P.: A posteriori error estimation for \(hp\)-version time-stepping methods for parabolic partial differential equations. Numer. Math. 115, 475–509 (2010)
Schwab, C.: \(p\)- and \(hp\)- Finite Element Methods. Oxford University Press, New York (1998)
Wei, Y.C., Yi, L.J.: An \(hp\)-version of the \(C^0\)-continuous Petrov–Galerkin time stepping method for nonlinear second-order initial value problems. Adv. Comput. Math. 46, Paper No. 56 (2020)
Wihler, T.P.: An a priori error analysis of the \(hp\)-version of the continuous Galerkin FEM for nonlinear initial value problems. J. Sci. Comput. 25, 523–549 (2005)
Yi, L.J.: An \(L^\infty \)-error estimate for the \(h\)-\(p\) version continuous Petrov–Galerkin method for nonlinear initial value problems, East Asian. J. Appl. Math. 5, 301–311 (2015)
Yi, L.J., Guo, B.Q.: An \(h\)-\(p\) version of the continuous Petrov–Galerkin finite element method for Volterra integro-differential equations with smooth and nonsmooth kernels. SIAM J. Numer. Anal. 53, 2677–2704 (2015)
Yi, L.J., Guo, B.Q.: The \(h\)-\(p\) version of the continuous Petrov–Galerkin method for nonlinear Volterra functional integro-differential equations with vanishing delays. Int. J. Numer. Anal. Model. 15, 26–47 (2018)
Zhu, H.Q., Celiker, F.: Nodal superconvergence of the local discontinuous Galerkin method for singularly perturbed problems. J. Comput. Appl. Math. 330, 95–116 (2018)
Funding
The authors have not disclosed any funding.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The work of this author is supported in part by the National Natural Science Foundation of China (Nos. 12171322, 11771298 and 12271366), the Natural Science Foundation of Shanghai (Nos. 21ZR1447200 and 22ZR1445500), and the Science and Technology Innovation Plan of Shanghai (No. 20JC1414200).
Appendix A. Some Proofs
Appendix A. Some Proofs
1.1 A.1. Proof of Lemma 2.5
Proof
We generalize the idea in [5] (where nodal superconvergence of the DG method has been studied) to the CG method. For convenience, we set \(e:=u-U\). By (1.1) and (2.2), we have
Assume that there is a positive constant M such that \(|f_{uu}(t, u)|\le M\) on \({\bar{I}}\times {\mathbb {R}}\). Using the classical Taylor’s theorem with Lagrange remainder in the variable u, there exists a function \(\chi \), whose value \(\chi (t)\) at t is between u(t) and U(t), such that
where
Clearly, there holds
Inserting (A.2) into (A.1), yields
We next construct an auxiliary problem: find v such that
Obviously, the exact solution of (A.6) can be expressed as
where the function \(w(t)=\exp \left( \int _{t}^{t_n}\theta (y)dy\right) \) with \(\theta (y)= f_u(y,u(y))\). Assume that \(f_u(t,u(t))\in C^{k-1}({\bar{I}})\). Then we have \(w \in C^{k}([0,t_n))\), which implies that
Using integration by parts, the fact \(e(0)=0\) and (A.6), gives
For any function \(w\in L^2(I_n)\), we denote by \(\pi _{I_n}^{k-1,*}w\in P_{k-1}(I_n)\) the usual \(L^2\)-projection of w onto \(P_{k-1}(I_n)\), namely,
for \(k \ge 1\). Then, for \(w\in H^k(I_n)\) there holds (cf. [24])
Using (A.5), (A.4), the Cauchy–Schwarz inequality and \(L^2\)-stability of the \(L^2\)-projection operator \(\pi _{I_n}^{k-1,*}\), we get
Noticing the well-known embedding inequality, namely, \(\Vert e\Vert _{L^\infty (0,t_n)}\le C \Vert e\Vert _{H^1(0,t_n)}\), we obtain
Combining (A.8), (A.11), (A.10), (A.12), (2.23) and (A.7), yields
which implies that
Using (A.13) and the \(H^1\)-error estimate (2.24) in Lemma 2.4, gives (2.29). As a direct consequence of (2.29), we get (2.30) easily. This ends the proof. \(\square \)
1.2 A.2. Proof of Lemma 3.6
Proof
We start by dividing the error \(e:=u-U\) into two parts that can be analyzed separately:
where U is the CG solution given by (2.1) and \(\Pi ^ku\) is the global projection defined by (2.16). Since Lemma 2.2 can be used to bound \(\eta \), the main task reduces to estimate the term \(\xi \).
We first prove (3.4). Since \(e=\eta +\xi \), using (A.1) we get
In view of (2.15), we have
Selecting \(\varphi =\xi '\) in (A.15), then using (1.2) and the Cauchy–Schwarz inequality, we obtain
which implies that
Summing up (A.16) over all element \(I_n,\ 1\le n\le N\), yields
which together with (2.23) leads to (3.4).
We next prove (3.5). Since
then using the Cauchy–Schwarz inequality, gives
Integrating (A.18) with respect to t on \(I_n\), yields
From (2.15), (2.16) and (A.13), we find that
Then, summing up (A.19) over all elements \(I_n,\ 1\le n\le N\), using (A.20) and (A.16), we obtain
Inserting (2.23) and (2.24) into (A.21), gives (3.5).
It remains to prove (3.6). According to the Sobolev inequality (see (3.9) of [13])
we have
Inserting (A.19) into (A.23), then using (A.20) and (A.16) we get
Noting the fact that (see pages 532–533 of [26])
and using (A.24), we get
Consequently, using (2.17), (2.24) and the fact \(\Vert u\Vert ^2_{H^{k+1}(I_n)} \le C h_n \Vert u\Vert ^2_{W^{k+1,\infty }(I_n)}\), yields
This completes the proof of (3.6).
Finally, using (3.4)–(3.6) and the quasi-uniformity of the time partition, we obtain (3.7)–(3.9). \(\square \)
1.3 A.3. Proof of Theorem 3.2
Proof
We proceed in several steps.
Step 1 We first prove (3.30). Due to (2.20) and (2.16), there holds
By (3.27), we have
for \(\sigma >\frac{1}{2}\). Then, using (2.17) with \(s_1=0\), (A.28) and (3.28), we get
Moreover, using (2.17) with \(s_n=k\), (3.27) and (3.28), yields
Suppose that \(2\sigma -\frac{2k+2}{\beta }>-1\), i.e., \(\beta >\frac{k+1}{\sigma +1/2}\). From (A.29) and (A.30), we obtain
Combining (A.27), (A.31) and (A.32) gives (3.30).
Step 2 We next prove (3.31). Due to (2.21) and (2.16), there holds
Using (2.18) with \(s_1=0\), (A.28) and (3.28), we get
Moreover, using (2.18) with \(s_n=k\), (3.27) and (3.28), gives
Suppose that \(2\sigma -2-\frac{2k}{\beta }>-1\), i.e., \(\beta >\frac{k}{\sigma -1/2}\). From (A.34) and (A.35), we have
Combining (A.33), (A.36) and (A.37) gives (3.31).
Step 3 We further show (3.32). In view of Lemma 2.3, we have
On the one hand, using the Sobolev inequality (A.22), (3.28), (A.29) and (A.34), we get
On the other hand, using (2.19) with \(s_n=k\), (3.27) and (3.28), we obtain
for \(2\le n\le N\). Suppose that \(2\sigma -\frac{2k+2}{\beta }\ge 0\), i.e., \(\beta \ge \frac{k+1}{\sigma }\). From (A.39) and (A.40), we have
Combining (A.38), (A.41) and (A.42) gives (3.32).
Step 4 It remains to show (3.33). Similar to the proof of (2.30) for regular solutions, we consider the same auxiliary problem (A.6) with exact solution
Here, the function \(w(t)=\exp \left( \int _{t}^{t_n}\theta (y)dy\right) \) with \(\theta (y)= {f_u(y,u(y))}\). Assume that \(f_u(t,u(t))\in C({\bar{I}})\). In view of the definition of the function \(\theta \), we may assume that the solution v of the auxiliary problem (A.6) exhibits the similar regularity as the solution of the original problem (1.1), namely,
for some positive constants C.
Combining (A.8), (A.43) and (A.11), gives
which together with (A.12) leads to
Here, \(\pi _{I_n}^{k-1,*}w\) is the \(L^2\)-projection of w onto \(P_{k-1}(I_n)\) as defined in (A.9).
Using the \(L^2\)-stability of the projection operator \(\pi _{I_n}^{k-1,*}\), (A.44) and (3.28), we have
for \(\beta >\frac{k}{\sigma +1/2}\). Similarly to (A.30), using (A.10), (A.44) and (3.28), we have
for \(\beta >\frac{k}{\sigma +1/2}\). In view of (A.46) and (A.44), we get
Inserting (A.46)–(A.48) into (A.45), gives
for \(\beta >\frac{k}{\sigma +1/2}\), which together with (3.31) and (3.30) leads to
for \(\beta > \max \{\frac{k+1}{\sigma +1/2}, \frac{k}{\sigma -1/2}\}\). This proves (3.33). \(\square \)
1.4 A.4. Proof of Lemma 3.7
Proof
We first recall the error splitting (A.14) that \(e=\xi +\eta \), where \(\eta =u-\Pi ^k u\) and \(\xi =\Pi ^k u-U\). In view of (A.17) and (3.30), there holds
This proves (3.34).
We next show (3.35). Noting the facts that \(\xi (t_n)=e(t_n)\) and \(h_n\le C \tau \), then using (A.21), (3.33) and (3.34), we have
for \(\beta > \max \{\frac{k+1}{\sigma +1/2}, \frac{k}{\sigma -1/2}\}\), which implies (3.35).
It remains to prove (3.36). Using again the facts that \(\xi (t_n)=e(t_n)\) and \(h_n\le C \tau \), combining (A.24), (3.33), (A.16), (A.25), (A.27), (A.31) and (A.32), results in
for \(\beta > \max \{\frac{k+1}{\sigma +1/2}, \frac{k}{\sigma -1/2}\}\). Therefore, we have
for \(\beta > \max \{\frac{k+1}{\sigma +1/2}, \frac{k}{\sigma -1/2}\}\).
For \(n=1\), using (2.17) with \(s_1=0\), (A.28) and (3.28), gives
for \(\beta \ge \frac{k+2}{\sigma +1}\). For \(2\le n\le N\), using (2.17) with \(s_n=k\), the fact \(\Vert u\Vert ^2_{H^{k+1}(I_n)} \le C h_n \Vert u\Vert ^2_{W^{k+1,\infty }(I_n)}\), (3.27) and (3.28), results in
provided that \(2\sigma -\frac{2k+4}{\beta }+2\ge 0\), i.e., \(\beta \ge \frac{k+2}{\sigma +1}\).
Combining (A.49), (A.50) and (A.51), yields
provided that \(\beta > \max \{\frac{k+2}{\sigma +1}, \frac{k+1}{\sigma +1/2}, \frac{k}{\sigma -1/2}\}\). This ends the proof of (3.36). \(\square \)
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Zhang, M., Yi, L. Superconvergent Postprocessing of the Continuous Galerkin Time Stepping Method for Nonlinear Initial Value Problems with Application to Parabolic Problems. J Sci Comput 94, 31 (2023). https://doi.org/10.1007/s10915-022-02086-1
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s10915-022-02086-1
Keywords
- Continuous Galerkin method
- Initial value problem
- Parabolic problem
- Postprocessing technique
- Superconvergence