The ADER high-order approach

for solving evolutionary PDEs

Eleuterio TORO
Laboratory of Applied Mathematics
University of Trento, Italy
www.ing.unitn.it/toro
[email protected]
1
 1

Introduction

2
I am indebted to many collaborators:

Richard Millington
Mauricio Caceres
Tomas Schwarzkopff
Claus-Dieter Munz
Vladimir Titarev
Yoko Takakura
Michael Dumbser
Martin Kaeser
Cedric Enaux
Cristobal Castro
Giovanni Russo
Carlos Pares
Manuel Castro
Arturo Hidalgo
Gianluca Vignoli
Giovanna Grosso
Matteo Antuono
Alberto Canestrelli
Annunziato Siviglia
Gino Montecinos
Lucas Mueller
Junbo Cheng
.......
.......
3
 We are interested in developing numerical methods
for approximating
time-dependent partial differential equations of the form

∂ t Q + A(Q) = S(Q) + D(Q)

4
 AIM:
solve to high (arbitrary) accuracy
in space and time

non-linear systems of hyperbolic balance laws

with stiff/non-stiff source terms
in multiple space dimensions on
structured/unstructured meshes
in the frameworks of
Finite Volume and
Discontinuous Galerkin Finite Element
Methods
5
Two basic design constraints

Ø Methods must be conservative

(because of the Lax-Wendroff theorem, 1960)

Ø Methods must be non-linear

(because of the Godunov theorem, 1959)

6
 The ADER approach
First results for linear equations in:

E. F. Toro, R. C. Millington and L. A. M. Nejad.

Towards Very High–Order Godunov Schemes.
In Godunov Methods: Theory and Applications. Edited Review.
E. F. Toro (Editor), pages 905–937. Kluwer Academic/Plenum Publishers, 2001

7
 Key features of ADER:

High-order non-linear spatial reconstruction

+
high-order Riemann problem
(also called the Generalized Riemann problem)
This generalized Riemann problem has initial conditions
with a high-order (spatial) representation, such as polynomials,
and source terms are included

8
High accuracy.

But why ?

9
 Test for linear acoustics ADER

Collaborators:
M. Dumbser, T. Schwartzkopff, and C.-D. Munz. Arbitrary high order finite volume schemes for linear wave
propagation. Book Series Notes on Numerical Fluid mechanics and Multidisciplinary Design. Springer Berlin / Heidelberg
ISSN 1612-2909, Volume 91/2006

10
11
 2

ADER in 1D
Finite volume formulation

12
Exact relation between integral averages
∂ t Q + ∂ x F(Q) = S(Q)
Integration in space and time
on control volume
[ xi −1/ 2 , xi +1/ 2 ] × [0, Δt ]
n +1 nΔt
Q i =Q −i [Fi+1/ 2 − Fi−1/ 2 ] + ΔtSi Exact relation
Δx
xi +1 / 2
n
⎫
Q =
i
1
Δx ∫ Q( x,0)dx ⎪
xi −1 / 2 ⎪ Integral averages
Δt ⎪
⎪
Fi +1/ 2 = Δ1t ∫ F (Qi +1/ 2 (τ ))dτ ⎬
0 ⎪
Δt xi +1 / 2 ⎪
Si = Δ1t Δ1x ∫ ∫ S (Qi ( x, t ))dxdt ⎪
0 xi −1 / 2
⎪
⎭

13
 Data reconstruction:

M. Dumbser, M. K¨aser, V. A. Titarev, and E. F. Toro. Quadrature-Free Non-Oscillatory Finite Volume Schemes on Unstructured
Meshes for Nonlinear Hy- perbolic Systems. J. Comput. Phys., 226(8):204–243, 2007.

14
Data variation across interface

15
 In 3D

The numerical flux requires the calculation of an integral in space along

the volume/element interface and in time.
16
A key ingredient:

the high-order
(or generalized)
Riemann problem
17
The high-order (or generalized) Riemann problem:

∂ t Q + ∂ x F (Q) = S (Q) ⎫
⎪
⎪
⎬ GRPK
Q
⎧ L ( x ) if x < 0
Q( x,0) = ⎨ ⎪
⎩QR ( x) if x > 0⎪⎭
Initial conditions: two smooth functions

QL (x), QR (x)
For example, two polynomials of degree K

The generalization is twofold:

(1) the intial conditions are two polynomials of arbitrary degree

(2) the equations include source terms

18
19
 15

12

9
Time

0
-10 0 10
x

20
 3

A solver for the

generalized Rieman problem

21
 K
(k ) τk
QLR (τ ) = Q(0,0+ ) + ∑ ∂ t Q(0,0+ )
k =1 k!
Q(0,0 + ) = lim t −0+ Q(0, t )

Extension of work of Ben-Artzi and Falcovitz, 1984, see also Raviart and LeFloch 1989

See also the related work of Harten et al, 1987.

The leading term

and
higher-order terms

22
Available information at time t=0

23
Computing the leading term in
K
(k) τk
QLR ( τ) = Q(0,0+ ) + ∑ ∂ t Q(0,0+ )
k =1 k!

Solve the classical RP

( 0)
Solution: D ( x / t )
Take Godunov state at x/t=0

Leading term: Q(0,0+ ) = D( 0) (0)

24
Computing the higher-order terms:
K
(k) τk
QLR ( τ) = Q(0,0+ ) + ∑ ∂ t Q(0,0+ )
k =1 k!

First use the Cauchy-Kowalewski (*) procedure

(k) (k) ( 0) (k)
∂ t Q(x, t ) = G (∂ x Q,...., ∂ x Q)
Example:
⎧ ∂ t q = − λ∂ x q
⎪
∂ t q + λ∂ x q = 0 ⇒ ⎨ ∂ (t 2 )q = ( −λ )2 ∂ (x2 )q
⎪∂ ( m )q = ( −λ) m ∂ ( m )q
⎩ t x

Must define spatial derivatives at x=0 for t>0

(*) Cauchy-Kowalewski theorem. One of the most fundamental results in the theory of PDEs.
Applies to problems in which all functions involved are analytic.

25
 Computing the higher-order terms…cont..
K
(k) τk
QLR ( τ) = Q(0,0+ ) + ∑ ∂ t Q(0,0+ )
k =1 k!

Then construct evolution equations for the variables:

(k)
∂ Q( x, t )
x
Note:

∂ t q + λ∂ x q = 0 ⇒ ∂ t (∂ x q) + λ∂ x (∂ x q) = 0
For the general case it can be shown that:
∂ t (∂ (xk )Q) + A(Q)∂ x (∂ (xk )Q) = H ( k ) (∂ (x0 )Q, ∂ (x1)Q,..., ∂ (xk )Q)

Neglecting source terms and linearizing we have

∂ t (∂ (xk )Q) + A(Q(0,0+ )) ∂ x (∂ (xk )Q) = 0

26
 Computing the higher-order terms…cont..
K
(k) τk
QLR ( τ) = Q(0,0+ ) + ∑ ∂ t Q(0,0+ )
k =1 k!

For each k solve classical Riemann problem:

∂ t (∂ (xk )Q) + A(Q(0,0 + ))∂ x (∂ (xk )Q) = 0⎫

⎪ (k )
(k ) ⎧ ∂ (k )
x QL ( 0 ) if x < 0 ⎬ sol : D (x / t)
∂ x Q( x,0) = ⎨ ( k ) ⎪
∂
⎩ x R Q ( 0 ) if x > 0 ⎭

Evaluate solution at x/t=0

All spatial derivatives at x=0 are now defined

(k) (k)
∂ Q(0,0+ ) = D (0)
x

27
 Computing the higher-order terms…cont..
K
(k) τk
QLR ( τ) = Q(0,0+ ) + ∑ ∂ t Q(0,0+ )
k =1 k!

All time derivatives at x=0 are then defined

(k) ( 0) (k)
∂ t Q(0,0+ ) = G( k ) (∂ x Q(0,0+ ),...., ∂ x Q(0,0+ ))
Solution of DRP is:
K k
(k) τ
QLR ( τ) = Q(0,0+ ) + ∑ ∂ t Q(0,0+ )
k =1 k!
GRP-K = 1( non-linear RP) + K (linear RPs)
Options: state expansion and flux expansion

28
 4

Complete ADER scheme

29
ADER finite volume method for

∂ t Q + ∂ x F(Q) = S(Q)
One-step scheme
n +1 nΔt
Q i =Q −i [Fi+1/ 2 − Fi−1/ 2 ] + ΔtSi
Δx
Numerical flux:
Δt
Fi +1/ 2 = 1
Δt ∫ F (Q i +1/ 2 (τ ))dτ
0

Numerical source:
Δt xi+1 / 2
Si = 1 1
Δt Δx ∫ ∫ S (Q ( x, t ))dxdt
i
0 xi −1 / 2

30
 More solvers for the
generalized Riemann problem:

C E Castro and E F Toro. Solvers for the high-order Riemann problem for hyperbolic balance
laws. Journal Computational Physics Vol. 227, pp 2482-2513,2008

M Dumbser, C Enaux and E F Toro. Finite volume schemes of very high order of accuracy for
stiff hyperbolic balance laws . Journal of Computational Physics, Vol 227, pp 3971-4001, 2008.

31
 Summary of ADER schemes
one-step
fully discrete schemes for
∂ t Q + ∂ x F (Q) + ∂ y G (Q) + ∂ z H (Q) = S (Q)

Accuracy in space and time is arbitrary

General meshes
Unified framework
Finite volume, DG finite elements

32
 The Cauchy-Kowalewski procedure:

A Fortran Example Code for the Cauchy-Kowalewski Procedure

for the 3D Euler Equations

M. Dumbser, M. Käser, V.A. Titarev, E.F. Toro. Quadrature-free non-oscillatory finite

volume schemes on unstructured meshes for nonlinear hyperbolic systems.
Journal of Computational Physics. Vol. 226, Issue 1, Pages 204-243,
10 September 2007.

33
 How to avoid the Cauchy-Kowalewski procedure:

Numerical evolution of data (related to Harten s method, 1987).

See

M Dumbser, C Enaux and E F Toro. Finite volume schemes of very high order of
accuracy for stiff hyperbolic balance laws . Journal of Computational Physics,
Vol 227, pp 3971-4001, 2008.

With this solver we can deal with stiff source terms

Work in progress to simplify solvers for the generalized Riemann

problem with stiff source terms

34
 Work in progress to simplify solvers for the generalized Riemann
problem with stiff source terms

Evaluation/comparison of four currently available

generalized Riemann problem solvers:

G I Montecinos, C E Castro, M Dumbser and E F Toro. Comparison of solvers for the

generalized Riemann problem for hyperbolic systems with source terms.
Journal of Computational Physics, 2012.
Available on line. DOI: http://dx.doi.org/10.1016/j.jcp.2012.06.011

35
 Main applications so far
1, 2, 3D Euler equations on unstructured meshes
3D compressible Navier-Stokes equations
Reaction-diffusion (parabolic equations)
Dispersive systems
Sediment transport in water flows (single phase)
Two-phase sediment transport (Pitman and Le model)
Two-layer shallow water equations
Aeroacoustics in 2 and 3D
Seismic wave propagation in 3D
Tsunami wave propagation
Magnetohydrodynamics
3D Maxwell equations
3D compressible two-phase flow, etc.
36
 5

Sample numerical results

37
Linear advection

38
39
40
WENO-5

ADER-3

41
2D and 3D Euler equations

42
2D Euler equations: reflection from triangular object

43
3D Euler equations: reflection from cone

44
2D and 3D Baer-Nunziato equations

45
3D Baer-Nunziato equations for compressible
two-phase flow

11 nonlinear hyperbolic PDES

Stiff source terms: relaxation terms

Michael Dumbser, Arturo Hidalgo, Manuel Castro, Carlos Parés and Eleuterio F. Toro.
FORCE schemes on unstructured meshes II: Non-conservative hyperbolic systems . Computer methods
in Applied Science and Engineering, Vol. 199, Issues 9-12, pp 625-647, January 2010.

46
47
3D Baer-Nunziato equations: spherical explosion

48
49
Double Mach reflection for the 2D Baer-Nunziato equations

50
 6

Convergence rates

51
52
 2D non-linear Euler equations on unstructured meshes

M. Dumbser, M. K¨aser, V. A. Titarev, and E. F. Toro. Quadrature-Free Non-Oscillatory Finite

Volume Schemes on Unstructured Meshes for Nonlinear Hy- perbolic Systems. J. Comput. Phys.,
226(8):204–243, 2007.

53
54
 Non-linear Euler equations with very stiff source terms

6.

Convergence rates

M. Dumbser, C. Enaux, and E. F. Toro. Finite Volume Schemes of Very High Order of Accuracy for Stiff Hyperbolic Balance
Laws. J. Comput. Phys., 227(8):3971–4001, 2008.
55
 Diffusion-reaction equations

E. F. Toro and A. Hidalgo. ADER Finite Volume Schemes for Diffusion–Reaction Equations.
Applied Numerical Mathematics, 59:73–100, 2009.
56
Convergence rates for the Baer-Nunziato equations in 2D unstructured meshes

Michael Dumbsera, , , Arturo Hidalgob, , Manuel Castroc, , Carlos Parésc, and Eleuterio F. Toroa, FORCE schemes on unstructured meshes II:
Non-conservative hyperbolic systems . Computer methods in Applied Science and Engineering, Vol. 199, Issues 9-12, pp 625-647, January 2010.

57
58
 7

Summary and
Concluding remarks

59
Ø Schemes of arbitrary accuracy in space and time for solving
time-dependent PDEs (eg hyperbolic balance laws with stiff source
terms) on unstructured meshes have been presented

Ø Non-linear reconstruction + generalized Riemann problem

Ø One-step, fully discrete, conservative and non-linear

Ø Unified frame, all orders in single scheme

60
Schemes are well established in two important scientific communities:

Acoustics

Seismology

Important advances in:

tsunami wave propagation

and
astrophysics

Current work: further simplification of algorithms

61
 Introduction to ADER approach in chapter 19 and 20:

Eleuterio Toro.
Riemann solvers and numerical methods for fluid dynamics.
A practical introduction.
Third edition. Springer-Verlag, Berlin Heidelberg, 2009.
Book (724 pages). ISBN 978-3-540-25202- 3, 2009.

62
Thank you

63