Exp. Rosenbrock methods ReLPM Numerical experiments
Implementation of
exponential Rosenbrock-type methods
M. Caliari
∗A. Ostermann
∗3rd Austrian Numerical Analysis Day, TU Wien April 26–27, 2007
∗Universit¨at Innsbruck, Institut f¨ur Mathematik
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Exponential Rosenbrock-type methods: the idea
Let us consider
u
0(t) = f (u(t)) = Au(t) + g (u(t)), u(t
0) = u
0M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Exponential Rosenbrock-type methods: the idea
Let us consider
u
0(t) = f (u(t)) = Au(t) + g (u(t)), u(t
0) = u
0For the numerical approximation u
nof u(t
n) at time t
n, first we linearise at each step
u
0(t) = J
nu(t) + g
n(u(t)), t
n≤ t ≤ t
n+1where
J
n= D
uf (u
n), g
n(u(t)) = f (u(t)) − J
nu(t)
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Exponential Rosenbrock-type methods: the idea
Let us consider
u
0(t) = f (u(t)) = Au(t) + g (u(t)), u(t
0) = u
0For the numerical approximation u
nof u(t
n) at time t
n, first we linearise at each step
u
0(t) = J
nu(t) + g
n(u(t)), t
n≤ t ≤ t
n+1where
J
n= D
uf (u
n), g
n(u(t)) = f (u(t)) − J
nu(t) and then we consider the variation-of-constants formula.
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Exponential Rosenbrock–Euler method
u(t
n+1) = exp(h
nJ
n)u
n+ Z
hn0
exp ((h
n− τ )J
n) g
n(u(t
n+ τ ))dτ
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Exponential Rosenbrock–Euler method
u(t
n+1) = exp(h
nJ
n)u
n+ Z
hn0
exp ((h
n− τ )J
n) g
n(u(t
n+ τ ))dτ and with g
n(u(t
n+ τ )) ≈ g
n(u
n) we get the exponential
Rosenbrock–Euler method
u
n+1= exp(h
nJ
n)u
n+ h
nϕ
1(h
nJ
n)g
n(u
n) where
ϕ
1(h
nJ
n) = 1 h
nZ
hn0
exp ((h
n− τ )J
n) dτ, ϕ
1(z) = e
z− 1 z
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
s -stage exponential Rosenbrock methods
More generally, we consider the s-stage exponential method
U
ni= exp(c
ih
nJ
n)u
n+ h
ni −1
X
j=1
a
ij(h
nJ
n)g
n(U
nj),
u
n+1= exp(h
nJ
n)u
n+ h
ns
X
i=1
b
i(h
nJ
n)g
n(U
ni).
where b
iand a
ijare linear combinations of ϕ
k(h
nJ
n), ϕ
k(h
nJ
n) = 1
h
nkZ
hn0
exp ((h
n− τ )J
n) τ
k−1(k − 1)! d τ
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Methods of order 2, 3 and 4
It is easy to construct a 2-stage third order method (with a second order error estimator embedded) and a 3-stage fourth order
method (with a third order error estimator embedded).
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Methods of order 2, 3 and 4
It is easy to construct a 2-stage third order method (with a second order error estimator embedded) and a 3-stage fourth order
method (with a third order error estimator embedded).
The exponential Rosenbrock–Euler method
u
n+1= exp(h
nJ
n)u
n+ h
nϕ
1(h
nJ
n)g
n(u
n)
is computationally attractive since it achieves second order with one stage only. With the additional ϕ
1(h
nJ
n)g
n(u
n+1) evaluation, it is possible to have a third order error estimator.
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Matrix functions evaluation
In the applications, these methods need the efficient evaluation of the underlying matrix functions ϕ
k(hJ)v (J ∈ R
N×N, v ∈ R
N).
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Matrix functions evaluation
In the applications, these methods need the efficient evaluation of the underlying matrix functions ϕ
k(hJ)v (J ∈ R
N×N, v ∈ R
N).
Most authors regard Krylov-like as the methods of choice.
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Matrix functions evaluation
In the applications, these methods need the efficient evaluation of the underlying matrix functions ϕ
k(hJ)v (J ∈ R
N×N, v ∈ R
N).
Most authors regard Krylov-like as the methods of choice.
An alternative class of polynomial methods is based on direct interpolation or approximation of the ϕ
kfunctions on the spectrum (or the field of values) of the relevant matrix.
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
ReLPM: Real Leja Points Method
Among others, the ReLPM has shown very attractive computational features:
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
ReLPM: Real Leja Points Method
Among others, the ReLPM has shown very attractive computational features:
I
it rests on Newton interpolation of the exponential functions at a sequence of real Leja points;
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
ReLPM: Real Leja Points Method
Among others, the ReLPM has shown very attractive computational features:
I
it rests on Newton interpolation of the exponential functions at a sequence of real Leja points;
I
they guarantee maximal convergence of the interpolant and thus superlinear convergence of the corresponding matrix polynomials;
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
ReLPM: Real Leja Points Method
Among others, the ReLPM has shown very attractive computational features:
I
it rests on Newton interpolation of the exponential functions at a sequence of real Leja points;
I
they guarantee maximal convergence of the interpolant and thus superlinear convergence of the corresponding matrix polynomials;
I
with respect to other sets of interpolation points, (Chebyshev, e.g.), the Leja points allow to increase the degree of
interpolation just by adding new nodes of the same sequence.
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
The first 13 Leja points on [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
The first 13 Leja points on [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
The first 13 Leja points on [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
The first 13 Leja points on [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
The first 13 Leja points on [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
The first 13 Leja points on [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
The first 13 Leja points on [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
The first 13 Leja points on [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
The first 13 Leja points on [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
The first 13 Leja points on [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
The first 13 Leja points on [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
The first 13 Leja points on [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
The first 13 Leja points on [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
The first 13 Leja points on [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
The first 13 Leja points on [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Estimating the spectrum: Gershgorin’s circles
A key step is a cheap estimate of a real focal interval [a, b] such that the minimal ellipse of the confocal family which contains the spectrum has a small capacity (half sum of the semi-axes).
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Estimating the spectrum: Gershgorin’s circles
A key step is a cheap estimate of a real focal interval [a, b] such that the minimal ellipse of the confocal family which contains the spectrum has a small capacity (half sum of the semi-axes).
The numerical experience with matrices arising from stable spatial discretizations of parabolic equations has shown that good results can be obtained simply by intersecting the Gershgorin’s circles of the matrix with the real axis.
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Kernel of ReLPM
The kernel of the ReLPM code is given by the interpolation of ϕ
k(hz) at Leja points of the real focal interval
[a, b] = [c − 2γ, c + 2γ].
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Kernel of ReLPM
The kernel of the ReLPM code is given by the interpolation of ϕ
k(hz) at Leja points of the real focal interval
[a, b] = [c − 2γ, c + 2γ].
In practice, it is numerically convenient to interpolate the function ϕ
k(h(c + γξ)) at Leja points {ξ
s} of the reference interval [−2, 2].
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Kernel of ReLPM
The kernel of the ReLPM code is given by the interpolation of ϕ
k(hz) at Leja points of the real focal interval
[a, b] = [c − 2γ, c + 2γ].
In practice, it is numerically convenient to interpolate the function ϕ
k(h(c + γξ)) at Leja points {ξ
s} of the reference interval [−2, 2].
Then, given the corresponding divided differences {d
i} for such a function, the matrix Newton polynomial of degree m is
ϕ
k(hJ)v ≈ p
m(J)v =
m
X
i=0
d
iΩ
i, Ω
i= ((J − cI )/γ − ξ
i −1I ) Ω
i −1Ω
0= v
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
ReLPM: key features
I
real arithmetic;
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
ReLPM: key features
I
real arithmetic;
I
no Krylov subspace to store;
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
ReLPM: key features
I
real arithmetic;
I
no Krylov subspace to store;
I
linear complexity due to the 2-term recurrence;
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
ReLPM: key features
I
real arithmetic;
I
no Krylov subspace to store;
I
linear complexity due to the 2-term recurrence;
I
no linear system to solve;
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
ReLPM: key features
I
real arithmetic;
I
no Krylov subspace to store;
I
linear complexity due to the 2-term recurrence;
I
no linear system to solve;
I
well structured for a parallel implementation.
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Time sub-steps
In general, it is not feasible to interpolate with the original time step h, which has to be split. This happens, e.g., when the expected degree for convergence is too large.
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Time sub-steps
In general, it is not feasible to interpolate with the original time step h, which has to be split. This happens, e.g., when the expected degree for convergence is too large.
Setting K = hJ, ϕ
k(K )v (k > 0) is the solution at t = 1 of y
0(t) = Ky (t) + t
k−1(k − 1)! v, y(0) = y
0= 0
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Time sub-steps
In general, it is not feasible to interpolate with the original time step h, which has to be split. This happens, e.g., when the expected degree for convergence is too large.
Setting K = hJ, ϕ
k(K )v (k > 0) is the solution at t = 1 of y
0(t) = Ky (t) + t
k−1(k − 1)! v, y(0) = y
0= 0 Given δ = 1/L, then y
L= ϕ
k(K )v , where
y
`+1= exp(δK )y
`+δ
kk−1
X
i=0
`
k−1−i(k − 1 − i)! ϕ
i+1(δK )v , ` = 0, . . . , L−1.
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Time sub-steps
In general, it is not feasible to interpolate with the original time step h, which has to be split. This happens, e.g., when the expected degree for convergence is too large.
Setting K = hJ, ϕ
k(K )v (k > 0) is the solution at t = 1 of y
0(t) = Ky (t) + t
k−1(k − 1)! v, y(0) = y
0= 0 Given δ = 1/L, then y
L= ϕ
k(K )v , where
y
`+1= exp(δK )y
`+δ
kk−1
X
i=0
`
k−1−i(k − 1 − i)! ϕ
i+1(δK )v , ` = 0, . . . , L−1.
It requires k evaluations of a matrix function for ` = 0. When increasing ` by one, only one additional evaluation, exp(δK )y
`, is required. The other terms are scalar multiples of the previous evaluations, which have to be stored.
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Diffusion-advection-reaction equations
We consider
∂
tu = ε(∂
xxu + ∂
yyu ) − α(∂
xu + ∂
yu) + γ u
u − 1
2
(1 − u) on the unit square subject to homogeneous Neumann b. c.
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Diffusion-advection-reaction equations
We consider
∂
tu = ε(∂
xxu + ∂
yyu ) − α(∂
xu + ∂
yu) + γ u
u − 1
2
(1 − u) on the unit square subject to homogeneous Neumann b. c.
Standard finite differences in space with ∆x = ∆y = 0.005 (d.o.f.
40401), final time T = 0.3.
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Time integrators
We consider 5 second order time integrators:
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Time integrators
We consider 5 second order time integrators:
erow2l
exponential Rosenbrock–Euler method with ϕ
kevaluations by ReLPM;
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Time integrators
We consider 5 second order time integrators:
erow2l
exponential Rosenbrock–Euler method with ϕ
kevaluations by ReLPM;
erow2k
exponential Rosenbrock–Euler method with ϕ
kevaluations by Krylov method (phipro routine by Y. Saad);
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Time integrators
We consider 5 second order time integrators:
erow2l
exponential Rosenbrock–Euler method with ϕ
kevaluations by ReLPM;
erow2k
exponential Rosenbrock–Euler method with ϕ
kevaluations by Krylov method (phipro routine by Y. Saad);
rkc
explicit Runge–Kutta–Chebyshev method (code by Sommeijer, Shampine and Verwer);
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Time integrators
We consider 5 second order time integrators:
erow2l
exponential Rosenbrock–Euler method with ϕ
kevaluations by ReLPM;
erow2k
exponential Rosenbrock–Euler method with ϕ
kevaluations by Krylov method (phipro routine by Y. Saad);
rkc
explicit Runge–Kutta–Chebyshev method (code by Sommeijer, Shampine and Verwer);
rock2
explicit Orthogonal Runge–Kutta–Chebyshev (code by Abdulle and Medovikov);
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Time integrators
We consider 5 second order time integrators:
erow2l
exponential Rosenbrock–Euler method with ϕ
kevaluations by ReLPM;
erow2k
exponential Rosenbrock–Euler method with ϕ
kevaluations by Krylov method (phipro routine by Y. Saad);
rkc
explicit Runge–Kutta–Chebyshev method (code by Sommeijer, Shampine and Verwer);
rock2
explicit Orthogonal Runge–Kutta–Chebyshev (code by Abdulle and Medovikov);
cn
classical Crank–Nicolson scheme (inexact Newton nonlinear solver, BiCGStab linear solver, ILU with no fill-in perconditioner).
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Results: small advection, small reaction
ε = 0.1, α = −1, γ = 1 (tolerances: 10
−3–10
−6)
1 2 4 8 16 32
CPU seconds 10-6
10-5 10-4 10-3 10-2
relative error in 2-norm
rkcrock2 erow2k erow2l
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Results: large reaction
ε = 0.1, α = −1, γ = 100 (tolerances: 10
−3–10
−6)
2 4 8 16 32 64
CPU seconds 10-6
10-5 10-4 10-3 10-2
relative error in 2-norm
rkcrock2 erow2k erow2l
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. Rosenbrock methods ReLPM Numerical experiments
Results: large advection
ε = 0.1, α = −10, γ = 1 (tolerances: 10
−3–10
−6)
2 4 8 16 32 64
CPU seconds 10-7
10-6 10-5 10-4 10-3 10-2
relative error in 2-norm
rkcrock2 erow2k erow2l
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
Collocation with cardinal functions
Given the PDE
∂
tu(t, x) = Lu(t, x)
(with L a second order, linear operator), we consider the interpolation
u(t, x) ≈
N
X
j=1
φ
j(x)u(t; x
j) = φ(x) ◦ u(t)
where φ
j(x
i) = δ
ijand u(t; x
j) = u(t, x
j).
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
Collocation with cardinal functions
Given the PDE
∂
tu(t, x) = Lu(t, x)
(with L a second order, linear operator), we consider the interpolation
u(t, x) ≈
N
X
j=1
φ
j(x)u(t; x
j) = φ(x) ◦ u(t)
where φ
j(x
i) = δ
ijand u(t; x
j) = u(t, x
j). Then, φ(x) ◦ u
0(t) = L (φ(x) ◦ u(t))
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
Collocation with cardinal functions
Given the PDE
∂
tu(t, x) = Lu(t, x)
(with L a second order, linear operator), we consider the interpolation
u(t, x) ≈
N
X
j=1
φ
j(x)u(t; x
j) = φ(x) ◦ u(t)
where φ
j(x
i) = δ
ijand u(t; x
j) = u(t, x
j). Then, φ(x) ◦ u
0(t) = L (φ(x) ◦ u(t)) Now, we can collocate at the same {x
i}.
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
Exponential meshless method
∗We get
u
0(t) = L (φ(x) ◦ u(t)) |
(x1,...,xN)T= (L(φ)(x) ◦ u(t)) |
(x1,...,xN)T= Au(t) where A = a
ij= L(φ
j)(x
i).
∗In collaboration with M. Vianello, University of Padua
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
Exponential meshless method
∗We get
u
0(t) = L (φ(x) ◦ u(t)) |
(x1,...,xN)T= (L(φ)(x) ◦ u(t)) |
(x1,...,xN)T= Au(t) where A = a
ij= L(φ
j)(x
i).
An iterative method (such as ReLPM or Krylov method) for the approximation of exp(h
nA)u
nis based on successive products
Au
(m)n, m = 0, 1, . . . , u
(0)n= u
n∗In collaboration with M. Vianello, University of Padua
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
Exponential meshless method
∗We get
u
0(t) = L (φ(x) ◦ u(t)) |
(x1,...,xN)T= (L(φ)(x) ◦ u(t)) |
(x1,...,xN)T= Au(t) where A = a
ij= L(φ
j)(x
i).
An iterative method (such as ReLPM or Krylov method) for the approximation of exp(h
nA)u
nis based on successive products
Au
(m)n, m = 0, 1, . . . , u
(0)n= u
nIf the nodes {x
i} are arbitrary and we have a tool to compute Au
n(m)directly as L
φ(x) ◦ u
(m)n|
(x1,...,xN)T, we get an exponential meshless method.
∗In collaboration with M. Vianello, University of Padua
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
2D examples: CShep2D and QShep2D
CShep2D, by R. J. Renka, is among the most accurate and
efficient scattered data interpolation algorithms. It constructs a C
2interpolant in a moving least-square fashion. The costs range between O(N) and O(N
2) (preprocessing) and O(1) and O(N) (evaluation), depending on the distribution of the nodes. It can compute also the first and the second derivative. The required storage is 9N.
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
2D examples: CShep2D and QShep2D
CShep2D, by R. J. Renka, is among the most accurate and
efficient scattered data interpolation algorithms. It constructs a C
2interpolant in a moving least-square fashion. The costs range between O(N) and O(N
2) (preprocessing) and O(1) and O(N) (evaluation), depending on the distribution of the nodes. It can compute also the first and the second derivative. The required storage is 9N.
QShep2D constructs a C
1interpolant. It can compute also the first and the second derivative (although not continuous). The required storage is 5N.
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
A monitor for the check points
Since the collocation, or center, points are arbitrary, we can add or remove points at each time step, depending on the behaviour of the solution on a set of check points.
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
A monitor for the check points
Since the collocation, or center, points are arbitrary, we can add or remove points at each time step, depending on the behaviour of the solution on a set of check points.
For example, we can consider a curvature monitor
|∆x
2u
xx(t; (ˇ x
i, ˇ y
i))| + |∆y
2u
yy(t; (ˇ x
i, ˇ y
i))|
which has to be evaluated at each check point (ˇ x
i, ˇ y
i).
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
Center and check points
check point center point
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
Center and check points
center point check point
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
Unstable derivatives
Derivatives computed directly by {C,Q}Shep2D give raise to instabilities during time integration (under investigation).
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
Unstable derivatives
Derivatives computed directly by {C,Q}Shep2D give raise to instabilities during time integration (under investigation).
Idea: derivatives in a FD fashion. They are much more stable.
center point check point FD stencil
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The Molenkamp–Crowley test
Let us consider
∂
tu + (au)
x+ (bu)
y= 0 with
a(x, y ) = −2π
y − 1
2
, b(x, y ) = 2π
x − 1
2
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The Molenkamp–Crowley test
Let us consider
∂
tu + (au)
x+ (bu)
y= 0 with
a(x, y ) = −2π
y − 1
2
, b(x, y ) = 2π
x − 1
2
The initial profile is rotated around the center of the domain. At time t = 1 one rotation will be completed. Homogeneous Dirichlet conditions are prescribed at the inflow boundaries.
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The Molenkamp–Crowley test
0 0.2
0.4 0.6
0.8 1 0
0.2 0.4
0.6 0.8
1 0
0.2 0.4 0.6 0.8 1
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The Molenkamp–Crowley test
0 0.2
0.4 0.6
0.8 1 0
0.2 0.4
0.6 0.8
1 0
0.2 0.4 0.6 0.8 1
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The Molenkamp–Crowley test
0 0.2
0.4 0.6
0.8 1 0
0.2 0.4
0.6 0.8
1 0
0.2 0.4 0.6 0.8 1
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The Molenkamp–Crowley test
0 0.2
0.4 0.6
0.8 1 0
0.2 0.4
0.6 0.8
1 0
0.2 0.4 0.6 0.8 1
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The Molenkamp–Crowley test
0 0.2
0.4 0.6
0.8 1 0
0.2 0.4
0.6 0.8
1 0
0.2 0.4 0.6 0.8 1
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The Molenkamp–Crowley test
0 0.2
0.4 0.6
0.8 1 0
0.2 0.4
0.6 0.8
1 0
0.2 0.4 0.6 0.8 1
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The Molenkamp–Crowley test
0 0.2
0.4 0.6
0.8 1 0
0.2 0.4
0.6 0.8
1 0
0.2 0.4 0.6 0.8 1
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The Molenkamp–Crowley test
0 0.2
0.4 0.6
0.8 1 0
0.2 0.4
0.6 0.8
1 0
0.2 0.4 0.6 0.8 1
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The Molenkamp–Crowley test
0 0.2
0.4 0.6
0.8 1 0
0.2 0.4
0.6 0.8
1 0
0.2 0.4 0.6 0.8 1
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The Molenkamp–Crowley test
0 0.2
0.4 0.6
0.8 1 0
0.2 0.4
0.6 0.8
1 0
0.2 0.4 0.6 0.8 1
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The Molenkamp–Crowley test
0 0.2
0.4 0.6
0.8 1 0
0.2 0.4
0.6 0.8
1 0
0.2 0.4 0.6 0.8 1
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The Molenkamp–Crowley test
0 0.2
0.4 0.6
0.8 1 0
0.2 0.4
0.6 0.8
1 0
0.2 0.4 0.6 0.8 1
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The Molenkamp–Crowley test
0 0.2
0.4 0.6
0.8 1 0
0.2 0.4
0.6 0.8
1 0
0.2 0.4 0.6 0.8 1
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The Molenkamp–Crowley test
0 0.2
0.4 0.6
0.8 1 0
0.2 0.4
0.6 0.8
1 0
0.2 0.4 0.6 0.8 1
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The Molenkamp–Crowley test
0 0.2
0.4 0.6
0.8 1 0
0.2 0.4
0.6 0.8
1 0
0.2 0.4 0.6 0.8 1
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The Molenkamp–Crowley test
0 0.2
0.4 0.6
0.8 1 0
0.2 0.4
0.6 0.8
1 0
0.2 0.4 0.6 0.8 1
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The Molenkamp–Crowley test
0 0.2
0.4 0.6
0.8 1 0
0.2 0.4
0.6 0.8
1 0
0.2 0.4 0.6 0.8 1
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The Molenkamp–Crowley test
0 0.2
0.4 0.6
0.8 1 0
0.2 0.4
0.6 0.8
1 0
0.2 0.4 0.6 0.8 1
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The Molenkamp–Crowley test
0 0.2
0.4 0.6
0.8 1 0
0.2 0.4
0.6 0.8
1 0
0.2 0.4 0.6 0.8 1
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The Molenkamp–Crowley test
0 0.2
0.4 0.6
0.8 1 0
0.2 0.4
0.6 0.8
1 0
0.2 0.4 0.6 0.8 1
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The first symmetrized 13 Leja points on i [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The first symmetrized 13 Leja points on i [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The first symmetrized 13 Leja points on i [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The first symmetrized 13 Leja points on i [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The first symmetrized 13 Leja points on i [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The first symmetrized 13 Leja points on i [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The first symmetrized 13 Leja points on i [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The first symmetrized 13 Leja points on i [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods
Exp. meshless Numerical experiment
The first symmetrized 13 Leja points on i [−2 , 2]
-3 -2 -1 0 1 2 3
M. Caliari, A. Ostermann Implementation of exponential Rosenbrock-type methods