Abstract
In this paper, by introducing a reconstruction operator based on the Legendre moments, we construct a reduced discontinuous Galerkin (RDG) space that could achieve the same approximation accuracy but using fewer degrees of freedom (DoFs) than the standard discontinuous Galerkin (DG) space. The design of the “narrow-stencil-based” reconstruction operator can preserve the local data structure property of the high-order DG methods. With the RDG space, we apply the local discontinuous Galerkin (LDG) method with the implicit-explicit time marching for the nonlinear unsteady convection–diffusion–reaction equation, where the reduction of the number of DoFs allows us to achieve higher efficiency. In terms of theoretical analysis, we give the well-posedness and approximation properties for the reconstruction operator and the \(L^2\) error estimate for the semi-discrete LDG scheme. Several representative numerical tests demonstrate the accuracy and the performance of the proposed method in capturing the layers.
Similar content being viewed by others
Data Availability
The datasets generated during the current study are available from the corresponding author upon reasonable request. They support our published claims and comply with field standards.
Notes
The kth order Legendre polynomials are defined as: \(\widehat{L}^k(\widehat{x}) = \frac{(-1)^k}{2^kk!}(\frac{d}{dx})^k[(1-x^2)^k]\) [1].
References
Abramowitz, M., Stegun, I.A.: Handbook of mathematical functions with formulas, graphs, and mathematical tables. US Government printing office (1964)
Ayuso, B., Marini, L.D.: Discontinuous Galerkin methods for advection-diffusion-reaction problems. SIAM J. Numer. Anal. 47(2), 1391–1420 (2009)
Barth, T., Frederickson, P.: Higher order solution of the Euler equations on unstructured grids using quadratic reconstruction. In 28th Aerospace Sciences Meeting, p. 13 (1990)
Baysal, O.: Stabilized finite element methods for time dependent convection–diffusion equations. Izmir Institute of Technology (Turkey) (2012)
Brenner, S.C., Scott, L.R., Scott, L.R.: The Mathematical Theory of Finite Element Methods. Springer, Berlin (2008)
Brooks, A.N., Hughes, T.J.: Streamline upwind/Petrov–Galerkin formulations for convection dominated flows with particular emphasis on the incompressible Navier–Stokes equations. Comput. Methods Appl. Mech. Eng. 32(1–3), 199–259 (1982)
Calvo, M., De Frutos, J., Novo, J.: Linearly implicit Runge–Kutta methods for advection–reaction–diffusion equations. Appl. Numer. Math. 37(4), 535–549 (2001)
Ciarlet, P.G.: The Finite Element Method for Elliptic Problems. SIAM (2002)
Cockburn, B., Dong, B., Guzmán, J., Restelli, M., Sacco, R.: A hybridizable discontinuous Galerkin method for steady-state convection–diffusion–reaction problems. SIAM J. Sci. Comput. 31(5), 3827–3846 (2009)
Cockburn, B., Karniadakis, G.E., Shu, C.-W.: The development of discontinuous Galerkin methods. In: Discontinuous Galerkin Methods, pp. 3–50. Springer (2000)
Cockburn, B., Shu, C.-W.: TVB Runge–Kutta local projection discontinuous Galerkin finite element method for conservation laws. II. General framework. Math. Comput. 52(186), 411–435 (1989)
Cockburn, B., Shu, C.-W.: The local discontinuous Galerkin method for time-dependent convection–diffusion systems. SIAM J. Numer. Anal. 35(6), 2440–2463 (1998)
Cockburn, B., Shu, C.-W.: Runge-Kutta discontinuous Galerkin methods for convection-dominated problems. J. Sci. Comput. 16(3), 173–261 (2001)
Ding, M., Cai, X., Guo, W., Qiu, J.-M.: A semi-Lagrangian discontinuous Galerkin (DG)-local DG method for solving convection-diffusion equations. J. Comput. Phys. 409, 109295 (2020)
Henderson, H.V., Pukelsheim, F., Searle, S.R.: On the history of the Kronecker product. Linear Multilinear Algebra 14(2), 113–120 (1983)
Houston, P., Schwab, C., Süli, E.: Discontinuous hp-finite element methods for advection–diffusion–reaction problems. SIAM J. Numer. Anal. 39(6), 2133–2163 (2002)
Hughes, T.J., Engel, G., Mazzei, G., Larson, M.G.: A comparison of discontinuous and continuous Galerkin methods based on error estimates, conservation, robustness and efficiency. In Discontinuous Galerkin Methods, pp. 135–146. Springer, Berlin (2000)
LeVeque, R.J.: Finite Difference Methods for Ordinary and Partial Differential Equations: Steady-State and Time-Dependent Problems. SIAM (2007)
Li, J., Zhai, S., Weng, Z., Feng, X.: H-adaptive RBF-FD method for the high-dimensional convection–diffusion equation. Int. Commun. Heat Mass Transf. 89, 139–146 (2017)
Li, R., Yang, F.: A least squares method for linear elasticity using a patch reconstructed space. Comput. Methods Appl. Mech. Eng. 363, 112902 (2020)
Nguyen, N.C., Peraire, J., Cockburn, B.: An implicit high-order hybridizable discontinuous Galerkin method for nonlinear convection–diffusion equations. J. Comput. Phys. 228(23), 8841–8855 (2009)
Ollivier-Gooch, C., Van Altena, M.: A high-order-accurate unstructured mesh finite-volume scheme for the advection–diffusion equation. J. Comput. Phys. 181(2), 729–752 (2002)
O’Sullivan, D.: Exploring spatial process dynamics using irregular cellular automaton models. Geogr. Anal. 33, 1–18 (2010)
Rashidinia, J., Khasi, M., Fasshauer, G.E.: A stable Gaussian radial basis function method for solving nonlinear unsteady convection–diffusion–reaction equations. Comput. Math. Appl. 75(5), 1831–1850 (2018)
Roos, H.-G., Stynes, M., Tobiska, L.: Robust Numerical Methods for Singularly Perturbed Differential Equations: Convection–Diffusion–Reaction and Flow Problems. Springer Science & Business Media, Berlin (2008)
Safdari-Vaighani, A., Heryudono, A., Larsson, E.: A radial basis function partition of unity collocation method for convection–diffusion equations arising in financial applications. J. Sci. Comput. 64(2), 341–367 (2015)
Sun, Z., Liu, J., Wang, P.: A discontinuous Galerkin method by patch reconstruction for convection–diffusion problems. Adv. Appl. Math. Mech. 12(3), 729–747 (2020)
Wang, H., Shu, C.-W., Zhang, Q.: Stability and error estimates of local discontinuous Galerkin methods with implicit-explicit time-marching for advection-diffusion problems. SIAM J. Numer. Anal. 53(1), 206–227 (2015)
Wang, H., Wang, S., Zhang, Q., Shu, C.-W.: Local discontinuous Galerkin methods with implicit-explicit time-marching for multi-dimensional convection–diffusion problems. ESAIM Math. Modell. Numer. Anal. 50(4), 1083–1105 (2016)
Wang, H., Zhang, Q., Shu, C.-W.: Third order implicit-explicit Runge-Kutta local discontinuous Galerkin methods with suitable boundary treatment for convection-diffusion problems with Dirichlet boundary conditions. J. Comput. Appl. Math. 342, 164–179 (2018)
Xu, Y., Shu, C.-W.: Error estimates of the semi-discrete local discontinuous Galerkin method for nonlinear convection–diffusion and KdV equations. Comput. Methods Appl. Mech. Eng. 196(37–40), 3805–3822 (2007)
Yap, P.-T., Paramesran, R.: An efficient method for the computation of Legendre moments. IEEE Trans. Pattern Anal. Mach. Intell. 27(12), 1996–2002 (2005)
Zhang, H., Ding, F.: On the Kronecker products and their applications. J. Appl. Math. 2013 (2013)
Zhang, Q., Shu, C.-W.: Error estimates to smooth solutions of Runge-Kutta discontinuous Galerkin methods for scalar conservation laws. SIAM J. Numer. Anal. 42(2), 641–666 (2004)
Funding
Research of Yinhua Xia was partially supported by National Key R &D Program of China No. 2022YFA1005202/2022YFA1005200 and NSFC Grant No. 12271498.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they 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.
Appendices
The Numerical Analysis of Assumption 1
Given one-dimensional mesh \({\mathcal {T}}_h = \{K_j\}_{j=1}^N\), we select the element itself and the direct Moore neighbors as the element stencil. Periodic boundary value problems can be denoted by
where \(K_{0}, K_{N+1}\) are obtained by the periodic extension. For other non-periodic boundary conditions, we denote the following bias stencil
For two dimensions, the stencil can be written as the tensor product of the one-dimensional stencil defined above. Consider the mesh \({\mathcal {T}}_h = \{K_\theta |\ \theta \in \Theta ^2\}\), where \(K_\theta =I^1_{\theta _1}\times I^2_{\theta _2}\) as denoted in the previous section. Here, we define the one-dimensional stencil for the first direction by \(\{S^1(I^1_{\theta _1})\}_{\theta _1=1}^{N_1}\), similarly, \(\{S^1(I^2_{\theta _2})\}_{\theta _2=1}^{N_2}\) for the second direction. The two-dimensional stencil is denoted by
The width of the stencil is \(d(S) = 3^d\). In what follows, we denote the stencil by \(S(K_\theta ) = \{K_{\theta ^1},K_{\theta ^2},\cdots ,K_{\theta ^{s}}\}\) without distinguishing one or two dimension. For clear description, we define the trace of a multiple indicator \(\alpha = (\alpha _1,\cdots , \alpha _n)\) by \(tr(\alpha ) = \sum _{i = 1}^n\alpha _i\) and sort the set \(A^k\) based on the ascendant sequence of the trace, i.e.,
where \(tr(\alpha ^1)\le tr(\alpha ^2)\le ,\cdots ,\le tr(\alpha ^{(k+1)^d})\).
As a result, we can express the approximation \(R_{K_\theta }^ku\) as the following Legendre expansion
where \(\pmb {u} = (u^{\alpha ^1},u^{\alpha ^2},\cdots ,u^{\alpha ^{(k+1)^d}})^T\) is the unknown coefficient vector to be determined.
For any element \(K_{\theta ^i}\in S^d(K_\theta )\), let \(M_{K_{\theta ^i}}^k\) denote an \((m+1)^d\times (k+1)^d\) matrix whose jth row is the Legendre moments vector obtained by applying the operator \(I_{K_{\theta ^i}}^{\alpha ^j}\) on the Legendre basis vector \((L^{\alpha ^1}_{K_\theta }(\pmb x),L^{\alpha ^2}_{K_\theta }(\pmb x),\cdots ,L^{\alpha ^{(k+1)^d}}_{K_\theta }(\pmb x))\), i.e.,
Repeating the same process for every element of the stencil \(S^d(K_\theta )\), we can define the coefficient matrix of the linear system (3.4) by
Ultimately, we obtain the matrix version of the system (3.4) as follows:
where \(\pmb 0\) is the zeros vector.
For this system to have a unique zero solution, there are two conditions to be simultaneously satisfied:
-
1.
Matrix \(M_{K_\theta }^k\) is a square matrix, i.e., \(3(m+1) = k+1\).
-
2.
The determinant of matrix \(M_{K_\theta }^k\) must be non-zero.
Under the first condition, let us discuss the second condition. To do that, we need to figure out each element of \(M_{K_\theta }^k\). Take the element \(I_{K_{\theta ^i}}^{\alpha ^r}L^{\alpha ^l}_{K_\theta }(\pmb x)\) for example, we have
where \(a_{\theta ^i}^j = \frac{h_{K_{\theta ^i}}^j}{h_{K_\theta }^j}\) and \(b_{\theta ^i}^j = \frac{2}{h_{K_\theta }^j}(x_{K_{\theta ^i}}^j-x_{K_\theta }^j)\).
Next, we will give numerical proof for different cases. Let us start with the one-dimensional case, i.e., \(d = 1\). Where \(\theta ,\alpha \) are all just single indices. The element \(K_\theta \) can be briefly written as \(K_j\). Besides, We have \(\alpha ^i = i-1\) for \(i \ge 1\). For \(k = 2\), we have \(m = 0\) according to the first condition. When taking the center stencil \(S_c^1(K_j) = \{K_{j-1},K_j,K_{j+1}\}\), we have
Naturally, the determinant of \(M_{K_j}^2\) can be computed as
To determine the value of \(\det (M_{K_j}^2)\), we need more information. First, the regularity of mesh can induce the following condition
Moreover, by simple calculation, there are \(b_{j-1} = -(1+a_{j-1})\), \(b_{j+1}=1+a_{j+1}\). With these informations, the following result can be directly obtained
For the backward stencil \(S_b^1(K_j) = \{K_{j-2},K_{j-1},K_{j}\}\), the similar determinant property results can be derived from the conditions \(b_{j-2} = -(1+2a_{j-1}+a_{j-2}),\ b_{j-1}=-(1+a_{j-1})\) and A.3, which is presented as follows,
In the same way, with the conditions \(b_{j+1} = 1+a_{j+1},\ b_{j+2}=1+2a_{j+1}+a_{j+2}\) and A.3, we can obtain the determinant property for the forward stencil \(S_f^1(K_j) = \{K_j,K_{j+1},K_{j+2}\}\) as follows,
The determinant property for the case of \(k = 5\) can be deduced following the proof process for \(k = 2\). The results are presented in Table 1.
Next, we turn our attention to the two-dimensional case, i.e., \(d = 2\). Here, \(\theta ,\alpha \) are all double indexes with the form (i, j). The element \(K_\theta \) can be briefly defined as \(K_{i,j} = I_i^1\times I_j^2\). For \(k = 2\), we have \(m = 0\). We recall the definition of the two-dimensional stencil, \(S^2(K_{i,j}) = S^1(I^1_i)\times S^1(I^2_j)\). Consequently, three different options for one-dimensional stencil produce nine different stencil strategies for two dimensions, which are shown as follows,
Take the stencil \(S_{cc}^2(K_{i,j})\) as an example. Let us illustrate the proof process. We know that the basis functions of \(Q^k(S^2(K_{i,j}))\) are the tensor product of the basis functions of \(Q^k(S^1(I_i))\) and \(Q^k(S^1(I_j))\). Considering the same product form of the stencils, the coefficient matrix can be expressed by the Kronecker product form
Given square matrices A and B with degrees m and n, there is a well-known determinant conclusion for the Kronecker product [15, 33],
Therefore, we have
Based on the determinant property in one dimension, we can conclude that the same property also holds for every stencil strategy for two dimensions. The same result can be obtained for \(k = 5\) according to the fact that
Proof of Theorem 1
Proof
The local reconstruction operator \(R_{K}^k\) can be regarded as one interpolation operator. Thus, proving the k-exactness property (3.7) is equivalent to proving the uniqueness of the polynomial interpolation problem, which has been given by Assumption 1.
With the k-exactness property (3.7), the operator \(R_{K}^k\) can be regarded as a projection operator which projects the Sobolev space \(H^{k+1}(S(K))\) on the polynomial space \(Q^k(S(K))\). Considering the \(l_\infty \) norm \(\Vert \cdot \Vert _{l_\infty (S(K))}\), as defined in Sect. 3.2, we have the following property
Following [5, Chapter 4.6], there exists an averaged Taylor polynomial \(T^{k+1}u\in Q^{k}(S(K))\) such that
With (3.5), (B.5), and (3.6), we have
Therefore,
which gives (3.8).
Following (3.8), the \(L^2\) error estimate (3.9) can be derived directly as follows,
By the approximation property (2.2), we can choose an approximation polynomial \(u_h\in Q^k(S(K)\) such that
With (2.2), (3.9), and the inverse property (2.3), we can derive (3.10) smoothly
Here, the proof is completed. \(\square \)
Proof of Theorem 3
Proof
Given that the exact solution u and \(\pmb q\) also satisfies the global weak formulation (4.3), we obtain the following error equation by a simple subtraction
Here, we denote
Thus, a more neat error equation can be described as follows,
Denote
and take the test function as
We obtain the energy equality
Next, we will estimate each term on the right-hand side of the energy equality.
To estimate the nonlinear convection term \({\mathcal {H}}(u,\xi )-{\mathcal {H}}(u_h,\xi )\), we need to make \(a\ priori\) assumption, for small enough h, we have
With this assumption, we have \(\Vert u-u_h\Vert _{0,\infty }\le Ch^{\frac{1}{2}}\). Moreover, the property (4.6) implies that \(\Vert R^ku-u_h\Vert _{0,\infty }\le Ch^{\frac{1}{2}}\).
Here, we have
According to [31, Lemma 3.4], the second part can be estimated as
Where \(\alpha (\hat{f};u_h)\) is non-negative and bounded, which is defined in [34]. Next, we estimate the first part I. Follow [31, Appendix A.1], I can be divided into six parts as follows,
where \(f_u'''\) and \(\tilde{f}_u'''\) are the factors in the remainder of Taylor expansion of \(f(u_h)\) and \(f(\bar{u}_h)\) separately. By integration by parts, we can estimate each term now:
-
\(I_1\) term
$$\begin{aligned} I_1 = \frac{1}{2}(\nabla \cdot (\pmb bf'(u)),\xi ^2)\le C\Vert \xi \Vert _{0}^2. \end{aligned}$$ -
\(I_2\) term
$$\begin{aligned} I_2 = \frac{1}{6}(\nabla \cdot (\pmb b f''(u))\xi ,\xi ^2)+\frac{1}{24}\langle \pmb bf''(u)[\xi ],[\xi ]^2\rangle .\\ \end{aligned}$$By a Taylor expansion, we have
$$\begin{aligned} \begin{aligned} f''(u)[\xi ]&= f''(u_h)[\xi ] +f''_u(u-u_h)[\xi ]\\&=-f''(u_h)[u_h]+f''(u_h)[\xi ^e]+f''_u(u-u_h)[\xi ]\\&\le 8\alpha (\hat{f};u_h)+8|[u-u_h]|^2+f''(u_h)[\xi ^e]+f''_u(u-u_h)[\xi ]. \end{aligned} \end{aligned}$$Thus,
$$\begin{aligned} \begin{aligned} I_2 \le&\frac{1}{3}\Vert \pmb b\Vert _{1,\infty }\langle \alpha (\hat{f};u_h),[\xi ]^2\rangle \\ {}&+C\left( \Vert \xi \Vert _{0,\infty }+h^{-1}(\Vert u-u_h\Vert _{0,\infty }^2+\Vert \xi ^e\Vert _{0,{\mathcal {E}}_h}^2)\right) \Vert \xi \Vert _{0}^2\\ \le&\frac{1}{3}\Vert \pmb b\Vert _{1,\infty }\langle \alpha (\hat{f};u_h),[\xi ]^2\rangle +C\Vert \xi \Vert _{0}^2. \end{aligned} \end{aligned}$$ -
\(I_3\) term
$$\begin{aligned} \begin{aligned} I_3&= -(\pmb bf'(u)\xi ^e,\nabla \cdot \xi )-\langle \pmb b(f'(u)-f'(\bar{u}_h))\bar{\xi }^e,[\xi ]\rangle -\langle \pmb bf'(\bar{u}_h)\bar{\xi }^e,[\xi ]\rangle \\ {}&\le C \Vert \xi \Vert _{0}^2 +2\langle \pmb b (\alpha (\hat{f};u_h)+C|[u-u_h]|)\bar{\xi }^e,[\xi ]\rangle +Ch^{2k}\\ {}&\le \frac{1}{6}\Vert \pmb b\Vert _{1,\infty }\langle \alpha (\hat{f};u_h),[\xi ]^2\rangle +C\Vert \xi \Vert _{0}^2+Ch^{2k}. \end{aligned} \end{aligned}$$ -
\(I_4\), \(I_5\), and \(I_6\) terms
$$\begin{aligned} \begin{aligned} I_4&\le Ch^{-1}\Vert \xi ^e\Vert _{0,\infty }\Vert \xi \Vert _{0}\le C\Vert \xi \Vert _{0}^2,\\ I_5&\le C\Vert \xi ^e\Vert _{0,\infty }(h^{-1}\Vert \xi ^e\Vert _{0} +h^{-\frac{1}{2}}\Vert \xi ^e\Vert _{0,{\mathcal {E}}_h})\Vert \xi \Vert _{0}\le C\Vert \xi \Vert _{0}^2+Ch^{2k+2},\\ I_6&\le C\Vert u-u_h\Vert _{0,\infty }^2(h^{-1}(\Vert \xi \Vert _{0}+\Vert \xi ^e\Vert _{0}) +h^{-\frac{1}{2}}\Vert \xi ^e\Vert _{0,{\mathcal {E}}_h})\Vert \xi \Vert _{0}\\ {}&\le C\Vert \xi \Vert _{0}^2+Ch^{2k+2}. \end{aligned} \end{aligned}$$
Combing with all the above estimates, we can derive that
Naturally, we have the following estimate
Then we focus on the diffusion part. We have
Finally, we estimate the reaction part as follows,
With the above estimates and Young’s inequality, the energy equation becomes
Thus,
By the Gronwall inequality and triangle inequality, the proof is completed as follows,
\(\square \)
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
Hou, S., Xia, Y. Discontinuous Galerkin Method Based on the Reduced Space for the Nonlinear Convection–Diffusion–Reaction Equation. J Sci Comput 99, 19 (2024). https://doi.org/10.1007/s10915-024-02486-5
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s10915-024-02486-5