Abstract
Exponential integrators are special time discretization methods where the traditional linear system solves used by implicit schemes are replaced with computing the action of matrix exponential-like functions on a vector. A very general formulation of exponential integrators is offered by the Exponential Propagation Iterative methods of Runge–Kutta type (EPIRK) family of schemes. The use of Jacobian approximations is an important strategy to drastically reduce the overall computational costs of implicit schemes while maintaining the quality of their solutions. This paper extends the EPIRK class to allow the use of inexact Jacobians as arguments of the matrix exponential-like functions. Specifically, we develop two new families of methods: EPIRK-W integrators that can accommodate any approximation of the Jacobian, and EPIRK-K integrators that rely on a specific Krylov-subspace projection of the exact Jacobian. Classical order conditions theories are constructed for these families. Practical EPIRK-W methods of order three and EPIRK-K methods of order four are developed. Numerical experiments indicate that the methods proposed herein are computationally favorable when compared to a representative state-of-the-art exponential integrator, and a Rosenbrock–Krylov integrator.
Similar content being viewed by others
References
Al-Mohy, A.H., Higham, N.J.: Computing the action of the matrix exponential, with an application to exponential integrators. SIAM J. Sci. Comput. 33(2), 488–511 (2011)
Allen, S.M., Cahn, J.W.: A microscopic theory for antiphase boundary motion and its application to antiphase domain coarsening. Acta Metall. 27(6), 1085–1095 (1979). https://doi.org/10.1016/0001-6160(79)90196-2
Berland, H., Owren, B., Skaflestad, B.: B-series and order conditions for exponential integrators. SIAM J. Numer. Anal. 43(4), 1715–1727 (2005)
Berland, H., Skaflestad, B., Wright, W.M.: Expint—a matlab package for exponential integrators. ACM Trans. Math. Softw. (TOMS) 33(1), 4 (2007)
Butcher, J.: Numerical Methods for Ordinary Differential Equations, 2nd edn. Wiley, New York (2008). https://doi.org/10.1002/9780470753767
Butcher, J.: Trees, B-series and exponential integrators. IMA J. Numer. Anal. 30, 131–140 (2010)
Caliari, M., Ostermann, A.: Implementation of exponential rosenbrock-type integrators. Appl. Numer. Math. 59(3–4), 568–581 (2009)
D’Augustine, A.F.: MATLODE: A MATLAB ODE solver and sensitivity analysis toolbox. Masters Thesis, Virginia Tech (2018)
Hairer, E., Norsett, S., Wanner, G.: Solving Ordinary Differential Equations I: Nonstiff Problems. Springer Series in Computational Mathematics, vol. 8. Springer, Berlin (1993)
Hairer, E., Wanner, G.: Solving Ordinary Differential Equations II: Stiff and Differential-Algebraic Problems. Springer Series in Computational Mathematics, vol. 14, 2nd edn. Springer, Berlin (1996)
Heineken, W., Warnecke, G.: Partitioning methods for reaction–diffusion problems. Appl. Numer. Math. 56(7), 981–1000 (2006)
Hochbruck, M., Lubich, C.: On Krylov subspace approximations to the matrix exponential operator. SIAM J. Numer. Anal. 34(5), 1911–1925 (1997)
Hochbruck, M., Lubich, C., Selhofer, H.: Exponential integrators for large systems of differential equations. SIAM J. Sci. Comput. 19(5), 1552–1574 (1998)
Hochbruck, M., Ostermann, A.: Explicit exponential Runge–Kutta methods for semilinear parabolic problems. SIAM J. Numer. Anal. 43(3), 1069–1090 (2005)
Hochbruck, M., Ostermann, A.: Exponential integrators. Acta Numer. 19, 209–286 (2012)
Loffeld, J., Tokman, M.: Comparative performance of exponential, implicit, and explicit integrators for stiff systems of odes. J. Comput. Appl. Math. 241, 45–67 (2013)
Lorenz, E.N.: Predictability—a problem partly solved. In: Palmer, T., Hagedorn, R. (eds.) Predictability of Weather and Climate, pp. 40–58. Cambridge University Press (CUP), Cambridge (1996). https://doi.org/10.1017/cbo9780511617652.004
Lu, Y.Y.: Computing a matrix function for exponential integrators. J. Comput. Appl. Math. 161(1), 203–216 (2003)
Minchev, B.V., Wright, W.: A review of exponential integrators for first order semi-linear problems. Technical report (2005)
Niesen, J., Wright, W.M.: Algorithm 919: a krylov subspace algorithm for evaluating the \(\varphi \)-functions appearing in exponential integrators. ACM Trans. Math. Softw. 38(3), 22:1–22:19 (2012). https://doi.org/10.1145/2168773.2168781
Rainwater, G., Tokman, M.: A new approach to constructing efficient stiffly accurate Epirk methods. J. Comput. Phys. 323, 283–309 (2016). https://doi.org/10.1016/j.jcp.2016.07.026
Roberts, S., Popov, A., Sandu, A.: ODE Test Problems. https://sibiu.cs.vt.edu/ode-test-problems/index.html. Accessed 16 Dec 2017
Saad, Y.: Iterative Methods for Sparse Linear Systems. SIAM, Philadelphia (2003)
Schulze, J.C., Schmid, P.J., Sesterhenn, J.L.: Exponential time integration using Krylov subspaces. Int. J. Numer. Meth. Fluids 60(6), 561–609 (2008)
Sidje, R.B.: Expokit: a software package for computing matrix exponentials. ACM Trans. Math. Softw. 24(1), 130–156 (1998). https://doi.org/10.1145/285861.285868
Steihaug, T., Wolfbrandt, A.: An attempt to avoid exact Jacobian and nonlinear equations in the numerical solution of stiff differential equations. Math. Comput. 33(146), 521–521 (1979). https://doi.org/10.1090/s0025-5718-1979-0521273-8
Tokman, M.: Efficient integration of large stiff systems of ODEs with exponential propagation iterative (EPI) methods. J. Comput. Phys. 213(2), 748–776 (2006)
Tokman, M.: A new class of exponential propagation iterative methods of Runge–Kutta type (EPIRK). J. Comput. Phys. 230, 8762–8778 (2011)
Tokman, M., Loffeld, J., Tranquilli, P.: New adaptive exponential propagation iterative methods of Runge–Kutta type. SIAM J. Sci. Comput. 34(5), A2650–A2669 (2012)
Tranquilli, P., Glandon, S.R., Sarshar, A., Sandu, A.: Analytical Jacobian-vector products for the matrix-free time integration of partial differential equations. J. Comput. Appl. Math. (2016). https://doi.org/10.1016/j.cam.2016.05.002
Tranquilli, P., Sandu, A.: Exponential–Krylov methods for ordinary differential equations. J. Comput. Phys. 278, 31–46 (2014). https://doi.org/10.1016/j.jcp.2014.08.013
Tranquilli, P., Sandu, A.: Rosenbrock–Krylov methods for large systems of differential equations. SIAM J. Sci. Comput. 36(3), A1313–A1338 (2014). https://doi.org/10.1137/130923336
Trefethen, L.N.: Evaluating matrix functions for exponential integrators via Carathéodory–Fejér approximation and contour integrals. Electron. Trans. Numer. Anal. 29, 1–18 (2007)
Verwer, J.G., Spee, E.J., Blom, J.G., Hundsdorfer, W.: A second-order rosenbrock method applied to photochemical dispersion problems. SIAM J. Sci. Comput. 20(4), 1456–1480 (1999)
van der Vorst, H.A.: Iterative Krylov Methods for Large Linear Systems. Cambridge University Press, Cambridge (2003). https://doi.org/10.1017/cbo9780511615115
Acknowledgements
This work has been supported in part by NSF through Awards NSF DMS-1419003, NSF DMS-1419105, NSF CCF-1613905, by AFOSR through Award AFOSR FA9550-12-1-0293-DEF, and by the Computational Science Laboratory at Virginia Tech.
Author information
Authors and Affiliations
Corresponding author
Appendices
A Derivation of K-Methods
The epirkk method is derived from epirkw method, where a specific Krylov-subspace approximation of the Jacobian is used instead of an arbitrary approximate Jacobian as admitted by the W-method. We start the derivation by first stating the general form of the epirkw method:
In the above equation, the \(\varvec{\psi }\) function is as defined in Eq. (8) and the following simplifying assumption is made about it:
where \(\varvec{\varphi }\) function is as defined in Eqs. (5) and (6). Additionally, the remainder function (\(\mathbf {r}(\mathbf {y})\)) and the forward difference operator (\(\Delta ^{(j)}\mathbf {r}(\mathbf {Y}_i)\)) are defined accordingly below:
The K-method uses a specific Krylov-subspace based approximation of the Jacobian (\(\mathbf {J}_{n}\)). An M-dimensional Krylov-subspace is built as,
whose basis is the orthonormal matrix \(\mathbf {V}\) and \(\mathbf {H}\) is the upper-Hessenberg matrix obtained from Arnoldi iteration defined as
The corresponding Krylov-subspace based approximation of the Jacobian is built as
The use of Krylov-subspace based approximation of the Jacobian reduces the \(\varvec{\varphi }\) and \(\varvec{\psi }\) function in accordance with Lemmas 2 and 3 to the following
where \(\widetilde{p}_j\) is defined as
In order to derive the reduced stage formulation of the epirkk method we need to resolve the vectors in the formulation into components in the Krylov-subspace and orthogonal to it. Repeating the splittings from the main text:
-
Splitting the internal stage vectors noting that \(\mathbf {Y}_0 \equiv \mathbf {y}_n\):
$$\begin{aligned} \mathbf {Y}_i = \mathbf {V}\varvec{\lambda }_i + \mathbf {Y}_i^{\bot } \qquad \text {where} \quad \mathbf {V}^{T} \mathbf {Y}_i = \varvec{\lambda }_i, \quad \left( \mathbf {I} - \mathbf {V}\mathbf {V}^{T}\right) \,\mathbf {Y}_i = \mathbf {Y}_i^{\bot }. \end{aligned}$$(40) -
Splitting the right-hand side function evaluated at internal stage vectors while noting that \(\mathbf {f}_0 \equiv \mathbf {f}(\mathbf {y}_n)\):
$$\begin{aligned} \mathbf {f}_i := \mathbf {f}(\mathbf {Y}_i) = \mathbf {V}\varvec{\eta }_i + \mathbf {f}_i^{\bot } \qquad \text {where} \quad \mathbf {V}^{T} \mathbf {f}_i = \varvec{\eta }_i, \quad \left( \mathbf {I} - \mathbf {V}\mathbf {V}^{T}\right) \mathbf {f}_i = \mathbf {f}_i^{\bot }. \end{aligned}$$(41) -
Splitting the non-linear Taylor remainder of the right-hand side functions:
$$\begin{aligned}&\mathbf {r}(\mathbf {Y}_i) = \mathbf {f}(\mathbf {Y}_i) - \mathbf {f}(\mathbf {y}_{n}) - \mathbf {A}_{n}\, (\mathbf {Y}_i - \mathbf {y}_{n}) = \mathbf {f}_i - \mathbf {f}_0 - \mathbf {V}\, \mathbf {H}\, \mathbf {V}^{T}\, (\mathbf {Y}_i - \mathbf {y}_{n}), \nonumber \\&\quad \text {where} \quad \mathbf {V}^{T}\, \mathbf {r}(\mathbf {Y}_i) = \varvec{\eta }_i - \varvec{\eta }_0 - \mathbf {H}\, (\varvec{\lambda }_i - \varvec{\lambda }_0), \nonumber \\&\quad \left( \mathbf {I} - \mathbf {V}\mathbf {V}^{T}\right) \, \mathbf {r}(\mathbf {Y}_i) = \mathbf {f}_i^{\bot } - \mathbf {f}_0^{\bot }. \end{aligned}$$(42) -
Splitting the forward differences of the non-linear remainder terms:
$$\begin{aligned}&\mathbf {\widetilde{r}}_{(j-1)}:=\Delta ^{(j-1)}\mathbf {r}(\mathbf {y}_{n}) = \mathbf {V}\, \mathbf {d}_{(j-1)} + \mathbf {\widetilde{r}}_{(j-1)}^{\bot }, \nonumber \\&\quad \text {where} \quad \mathbf {V}^{T}\,\mathbf {\widetilde{r}}_{(j-1)} = \mathbf {d}_{(j-1)}, \nonumber \\&\quad \left( \mathbf {I} - \mathbf {V}\mathbf {V}^{T}\right) \, \mathbf {\widetilde{r}}_{(j-1)} = \mathbf {\widetilde{r}}_{(j-1)}^{\bot }. \end{aligned}$$(43)
Using these each internal stage of the epirkk method can be expressed as below:
The reduced stage formulation of the epirkk method is obtained by multiplying the above equation by \(\mathbf {V}^{T}\) from the left
And the full stage vector can be recovered by first computing the reduced stage vector and adding the orthogonal piece \(\mathbf {Y}_i^{\bot }\) obtained when multiplying Eq. (44) by \(\left( \mathbf {I} - \mathbf {V}\mathbf {V}^{T}\right) \)
The final stage can also be written in the above form with multipliers \(b_i\) in place of \(a_{ij}\). Notice that the expensive computations are performed in the reduced space, i.e. the \(\varvec{\psi }\) function is computed in the reduced space instead of the full space offering potential computational savings. In the above equations, the quantities \(\mathbf {d}_{(j-1)}\) and \(\mathbf {\widetilde{r}}_{(j-1)}^{\bot }\) can be shown to be
as is done in the following “Appendix”.
B Proofs
In order to prove Eqs. (23a) and (23b), we start with the definition of the remainder function and forward difference.
Lemma 4
\(\Delta ^{(j)}\mathbf {r}(\mathbf {Y}_i) = \displaystyle \sum _{k=0}^{j} (-1)^k {j \atopwithdelims ()k} \mathbf {r}(\mathbf {Y}_{i+j-k})\).
Proof
In order to prove the lemma, we resort to mathematical induction. Base case \(j = 1\),
The base case is true by definition. We now assume that the proposition holds true for all j up to \(k-1\). We have,
Then for \(j = k\),
We perform a change of variable for the second summation, \(m = l + 1 \; \implies l = (m - 1)\),
The summations run between the same start and end indices, and can be collapsed.
We use the identity,
and arrive at the desired result,
\(\square \)
Lemma 5
Given
we need to prove
Proof
We start with Lemma 4 where we have proven that
We plugin the value \(i = 0\) which corresponds to \(\Delta ^{(j)}\mathbf {r}(\mathbf {Y}_0) \equiv \Delta ^{(j)}\mathbf {r}(\mathbf {y}_n)\) and we get,
WLOG replacing j by \(j-1\) yields
Since the left-hand side of Eq. (57) can be written as
we have the following result
The remainder function \(\mathbf {r}(\mathbf {Y}_i)\) is defined as
Plugging in the definition of remainder function in (58) and observing that \(\mathbf {r}(\mathbf {y}_n) = 0\), we get
Consider the identity involving alternating sum and difference of binomial coefficients,
Additionally, since we are replacing the Jacobian by the approximation in the Krylov-subspace, i.e. \(\mathbf {A}_{n} = \mathbf {V}\,\mathbf {H}\,\mathbf {V}^T\) we have
We get the expression for \(\mathbf {d}_{(j-1)}\) by multiplying Eq. (62) from the left by \(V^{T}\) and that for \(\mathbf {\widetilde{r}}_{(j-1)}^{\bot }\) by multiplying by \((\mathbf {I} - \mathbf {V}\mathbf {V}^{T})\). \(\square \)
Rights and permissions
About this article
Cite this article
Narayanamurthi, M., Tranquilli, P., Sandu, A. et al. EPIRK-W and EPIRK-K Time Discretization Methods. J Sci Comput 78, 167–201 (2019). https://doi.org/10.1007/s10915-018-0761-3
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10915-018-0761-3