Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks...
Transcript of Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks...
![Page 1: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/1.jpg)
Wave Propagation Software, Computational Science,
and Reproducible Research
Randall J. LeVequeDepartment of Applied Mathematics
University of Washington
Supported in part by NSF and DOE
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 2: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/2.jpg)
Outline
• Hyperbolic PDEs and Riemann problems
• Wave propagation and shock formation
• Applications to many practical problems
• Finite volume methods: Godunov and high-resolution
• General formulation based on Riemann solvers
• Implementation in CLAWPACK software
• Some extensions: adaptive mesh refinement, manifolds
• Example: Shallow water on the sphere
• Shallow water equations over bathymetry with innundation
• f-wave formulation, well-balanced schemes
• TsunamiClaw and tsunami simulations
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 3: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/3.jpg)
Outline
• Hyperbolic PDEs and Riemann problems
• Wave propagation and shock formation
• Applications to many practical problems
• Finite volume methods: Godunov and high-resolution
• General formulation based on Riemann solvers
• Implementation in CLAWPACK software
• Some extensions: adaptive mesh refinement, manifolds
• Example: Shallow water on the sphere
• Shallow water equations over bathymetry with innundation
• f-wave formulation, well-balanced schemes
• TsunamiClaw and tsunami simulations
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 4: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/4.jpg)
Outline
• Hyperbolic PDEs and Riemann problems
• Wave propagation and shock formation
• Applications to many practical problems
• Finite volume methods: Godunov and high-resolution
• General formulation based on Riemann solvers
• Implementation in CLAWPACK software
• Some extensions: adaptive mesh refinement, manifolds
• Example: Shallow water on the sphere
• Shallow water equations over bathymetry with innundation
• f-wave formulation, well-balanced schemes
• TsunamiClaw and tsunami simulations
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 5: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/5.jpg)
Outline
• Hyperbolic PDEs and Riemann problems
• Wave propagation and shock formation
• Applications to many practical problems
• Finite volume methods: Godunov and high-resolution
• General formulation based on Riemann solvers
• Implementation in CLAWPACK software
• Some extensions: adaptive mesh refinement, manifolds
• Example: Shallow water on the sphere
• Shallow water equations over bathymetry with innundation
• f-wave formulation, well-balanced schemes
• TsunamiClaw and tsunami simulations
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 6: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/6.jpg)
Many physical applications of hyperbolic problems...
• Compressible flow:• Aerodynamics — shocks near transonic or supersonic
aircraft• Detonation waves — coupled with combustion• Astrophysics — stellar dynamics, supernova explosions, ...
• Magnetohydrodynamics:• magnetic storms, stellar dynamics• plasma physics, fusion reactors
• Relativistic flows:• black hole accretion• general relativity: gravitational waves
hyperbolic formulations of Einstein equations
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 7: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/7.jpg)
Many physical applications of hyperbolic problems...
• Compressible flow:• Aerodynamics — shocks near transonic or supersonic
aircraft• Detonation waves — coupled with combustion• Astrophysics — stellar dynamics, supernova explosions, ...
• Magnetohydrodynamics:• magnetic storms, stellar dynamics• plasma physics, fusion reactors
• Relativistic flows:• black hole accretion• general relativity: gravitational waves
hyperbolic formulations of Einstein equations
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 8: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/8.jpg)
Many physical applications of hyperbolic problems...
• Compressible flow:• Aerodynamics — shocks near transonic or supersonic
aircraft• Detonation waves — coupled with combustion• Astrophysics — stellar dynamics, supernova explosions, ...
• Magnetohydrodynamics:• magnetic storms, stellar dynamics• plasma physics, fusion reactors
• Relativistic flows:• black hole accretion• general relativity: gravitational waves
hyperbolic formulations of Einstein equations
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 9: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/9.jpg)
Many physical applications of hyperbolic problems...
• Elastic waves• Seismology — earthquakes, imaging• Shock waves in tissue and bone — lithotripsy and shock
wave therapy• Shock induced phase transitions
• Volcanic flows• Dusty gas jets and pyroclastic flows• Lava flows• Debris flows
• Shallow water equations• global atmospheric and ocean modeling• river flows, dam breaks• tsunami propagation and inundation
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 10: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/10.jpg)
Many physical applications of hyperbolic problems...
• Elastic waves• Seismology — earthquakes, imaging• Shock waves in tissue and bone — lithotripsy and shock
wave therapy• Shock induced phase transitions
• Volcanic flows• Dusty gas jets and pyroclastic flows• Lava flows• Debris flows
• Shallow water equations• global atmospheric and ocean modeling• river flows, dam breaks• tsunami propagation and inundation
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 11: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/11.jpg)
Many physical applications of hyperbolic problems...
• Elastic waves• Seismology — earthquakes, imaging• Shock waves in tissue and bone — lithotripsy and shock
wave therapy• Shock induced phase transitions
• Volcanic flows• Dusty gas jets and pyroclastic flows• Lava flows• Debris flows
• Shallow water equations• global atmospheric and ocean modeling• river flows, dam breaks• tsunami propagation and inundation
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 12: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/12.jpg)
First order hyperbolic PDE in 1 space dimension
Linear: qt +Aqx = 0, q(x, t) ∈ lRm, A ∈ lRm×m
Conservation law: qt + f(q)x = 0, f : lRm → lRm (flux)
Quasilinear form: qt + f ′(q)qx = 0
Hyperbolic if A or f ′(q) is diagonalizable with real eigenvalues.
Models wave motion ar advective transport.
Eigenvalues are wave speeds.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 13: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/13.jpg)
Example: Linear acoustics in a 1d tube
q =[pu
]p(x, t) = pressure perturbationu(x, t) = velocity
Equations:
pt + κux = 0 κ = bulk modulusρut + px = 0 ρ = density
or [pu
]t
+[
0 κ1/ρ 0
] [pu
]x
= 0.
Eigenvalues: λ = ±c, where c =√κ/ρ = sound speed
Second order form: Can combine equations to obtain
ptt = c2pxx
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 14: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/14.jpg)
Riemann Problem
Special initial data:
q(x, 0) ={ql if x < 0qr if x > 0
Example: Acoustics with bursting diaphram
Pressure:
Acoustic waves propagate with speeds ±c.R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 15: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/15.jpg)
Riemann Problem
Special initial data:
q(x, 0) ={ql if x < 0qr if x > 0
Example: Acoustics with bursting diaphram
Pressure:
Acoustic waves propagate with speeds ±c.R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 16: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/16.jpg)
Riemann Problem
Special initial data:
q(x, 0) ={ql if x < 0qr if x > 0
Example: Acoustics with bursting diaphram
Pressure:
Acoustic waves propagate with speeds ±c.R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 17: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/17.jpg)
Riemann Problem
Special initial data:
q(x, 0) ={ql if x < 0qr if x > 0
Example: Acoustics with bursting diaphram
Pressure:
Acoustic waves propagate with speeds ±c.R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 18: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/18.jpg)
Riemann Problem
Special initial data:
q(x, 0) ={ql if x < 0qr if x > 0
Example: Acoustics with bursting diaphram
Pressure:
Acoustic waves propagate with speeds ±c.R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 19: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/19.jpg)
Riemann Problem
Special initial data:
q(x, 0) ={ql if x < 0qr if x > 0
Example: Acoustics with bursting diaphram
Pressure:
Acoustic waves propagate with speeds ±c.R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 20: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/20.jpg)
Riemann Problem for acoustics
Waves propagating in x–t space:
Left-going wave W1 = qm − ql andright-going wave W2 = qr − qm are eigenvectors of A.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 21: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/21.jpg)
Nonlinear compressible flow — Euler equations
Conservation of mass, momentum, energy: qt + f(q)x = 0 with
q =
ρρuE
, f(q) =
ρuρu2 + pu(E + p)
where
p = pressure = p(ρ,E) (Equation of state)
The Jacobian f ′(q) has eigenvalues u− c, u, u+ c where
c =
√dp
dρat constant entropy
Eigenvalues vary with q =⇒ shocks, rarefactions.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 22: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/22.jpg)
Riemann Problem for Euler equations
Special initial data:
q(x, 0) ={ql if x < 0qr if x > 0
Example: Shock tube problem in gas dynamics
Pressure:
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 23: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/23.jpg)
Riemann Problem for Euler equations
Special initial data:
q(x, 0) ={ql if x < 0qr if x > 0
Example: Shock tube problem in gas dynamics
Pressure:
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 24: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/24.jpg)
Riemann Problem for Euler equations
Special initial data:
q(x, 0) ={ql if x < 0qr if x > 0
Example: Shock tube problem in gas dynamics
Pressure:
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 25: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/25.jpg)
Riemann Problem for Euler equations
Special initial data:
q(x, 0) ={ql if x < 0qr if x > 0
Example: Shock tube problem in gas dynamics
Pressure:
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 26: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/26.jpg)
Riemann Problem for Euler equations
Special initial data:
q(x, 0) ={ql if x < 0qr if x > 0
Example: Shock tube problem in gas dynamics
Pressure:
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 27: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/27.jpg)
Riemann Problem for gas dynamics
Waves propagating in x–t space:
Similarity solution (function of x/t alone).
Waves can be approximated by discontinuties:Approximate Riemann solvers
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 28: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/28.jpg)
Shock formation
For nonlinear problems wave speed generally depends on q.
Waves can steepen up and form shocks=⇒ even smooth data can lead to discontinuous solutions.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 29: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/29.jpg)
Shock formation
For nonlinear problems wave speed generally depends on q.
Waves can steepen up and form shocks=⇒ even smooth data can lead to discontinuous solutions.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 30: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/30.jpg)
Shock formation
For nonlinear problems wave speed generally depends on q.
Waves can steepen up and form shocks=⇒ even smooth data can lead to discontinuous solutions.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 31: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/31.jpg)
Shock formation
For nonlinear problems wave speed generally depends on q.
Waves can steepen up and form shocks=⇒ even smooth data can lead to discontinuous solutions.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 32: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/32.jpg)
Shock formation
For nonlinear problems wave speed generally depends on q.
Waves can steepen up and form shocks=⇒ even smooth data can lead to discontinuous solutions.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 33: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/33.jpg)
Shock formation
For nonlinear problems wave speed generally depends on q.
Waves can steepen up and form shocks=⇒ even smooth data can lead to discontinuous solutions.
Computational challenges!
Need to capture sharp discontinuities.
PDE breaks down, standard finite difference approximation toqt + f(q)x = 0 can fail badly: nonphysical oscillations,convergence to wrong weak solution.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 34: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/34.jpg)
Integral form of conservation law
d
dt
∫ x2
x1
q(x, t) dx = f(q(x1, t))− f(q(x2, t))
For smooth q: = −∫ x2
x1
∂
∂xf(q(x, t)) dx
=⇒∫ x2
x1
qt + f(q)x dx = 0 ∀x1, x2
Integral form holds for discontinuous q.
Integrating in t gives∫ x2
x1
q(x, t2) dx =∫ x2
x1
q(x, t1) dx−∫ t2
t1
[f(q(x2, t))−f(q(x1, t))] dt
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 35: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/35.jpg)
Integral form of conservation law
d
dt
∫ x2
x1
q(x, t) dx = f(q(x1, t))− f(q(x2, t))
For smooth q: = −∫ x2
x1
∂
∂xf(q(x, t)) dx
=⇒∫ x2
x1
qt + f(q)x dx = 0 ∀x1, x2
Integral form holds for discontinuous q.
Integrating in t gives∫ x2
x1
q(x, t2) dx =∫ x2
x1
q(x, t1) dx−∫ t2
t1
[f(q(x2, t))−f(q(x1, t))] dt
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 36: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/36.jpg)
Integral form of conservation law
d
dt
∫ x2
x1
q(x, t) dx = f(q(x1, t))− f(q(x2, t))
For smooth q: = −∫ x2
x1
∂
∂xf(q(x, t)) dx
=⇒∫ x2
x1
qt + f(q)x dx = 0 ∀x1, x2
Integral form holds for discontinuous q.
Integrating in t gives∫ x2
x1
q(x, t2) dx =∫ x2
x1
q(x, t1) dx−∫ t2
t1
[f(q(x2, t))−f(q(x1, t))] dt
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 37: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/37.jpg)
Finite volume methods
Approximate cell averages: Qni ≈
1∆x
∫ xi+1/2
xi−1/2
q(x, tn) dx
Updating formula (conservation form):
Qn+1i = Qn
i −∆t∆x
[Fni+1/2 − F
ni−1/2]
where the numerical flux is
Fni−1/2 ≈
1∆t
∫ tn+1
tn
f(q(xi−1/2, t)) dt.
Note: can rewrite as
Qn+1i −Qn
i
∆t+Fn
i+1/2 − Fni−1/2
∆x= 0.
Looks like finite difference approximation to qt + fx = 0, butFn
i−1/2 is properly viewed as average flux through cell edge.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 38: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/38.jpg)
Finite volume methods
Approximate cell averages: Qni ≈
1∆x
∫ xi+1/2
xi−1/2
q(x, tn) dx
Updating formula (conservation form):
Qn+1i = Qn
i −∆t∆x
[Fni+1/2 − F
ni−1/2]
where the numerical flux is
Fni−1/2 ≈
1∆t
∫ tn+1
tn
f(q(xi−1/2, t)) dt.
Note: can rewrite as
Qn+1i −Qn
i
∆t+Fn
i+1/2 − Fni−1/2
∆x= 0.
Looks like finite difference approximation to qt + fx = 0, butFn
i−1/2 is properly viewed as average flux through cell edge.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 39: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/39.jpg)
Godunov’s Method for qt + f(q)x = 0
1. Solve Riemann problems at all interfaces, yielding wavesWp
i−1/2 and speeds spi−1/2, for p = 1, 2, . . . , m.
Riemann problem: Original equation with piecewise constantdata.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 40: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/40.jpg)
Godunov’s Method for qt + f(q)x = 0
Then either:
1. Compute new cell averages by integrating over cell at tn+1,
2. Compute fluxes at interfaces and flux-difference:
Qn+1i = Qn
i −∆t∆x
[Fni+1/2 − F
ni−1/2]
3. Update cell averages by contributions from all waves entering cell:
Qn+1i = Qn
i −∆t∆x
[A+∆Qi−1/2 +A−∆Qi+1/2]
where A±∆Qi−1/2 =m∑
i=1
(spi−1/2)
±Wpi−1/2.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 41: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/41.jpg)
Godunov’s Method for qt + f(q)x = 0
Then either:
1. Compute new cell averages by integrating over cell at tn+1,
2. Compute fluxes at interfaces and flux-difference:
Qn+1i = Qn
i −∆t∆x
[Fni+1/2 − F
ni−1/2]
3. Update cell averages by contributions from all waves entering cell:
Qn+1i = Qn
i −∆t∆x
[A+∆Qi−1/2 +A−∆Qi+1/2]
where A±∆Qi−1/2 =m∑
i=1
(spi−1/2)
±Wpi−1/2.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 42: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/42.jpg)
Godunov’s Method for qt + f(q)x = 0
Then either:
1. Compute new cell averages by integrating over cell at tn+1,
2. Compute fluxes at interfaces and flux-difference:
Qn+1i = Qn
i −∆t∆x
[Fni+1/2 − F
ni−1/2]
3. Update cell averages by contributions from all waves entering cell:
Qn+1i = Qn
i −∆t∆x
[A+∆Qi−1/2 +A−∆Qi+1/2]
where A±∆Qi−1/2 =m∑
i=1
(spi−1/2)
±Wpi−1/2.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 43: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/43.jpg)
Upwind wave-propagation algorithm
Qn+1i = Qn
i −∆t∆x
m∑p=1
(spi−1/2)
+Wpi−1/2 +
m∑p=1
(spi+1/2)
−Wpi+1/2
where
s+ = max(s, 0), s− = min(s, 0).
Note: Requires only waves and speeds.
Applicable also to hyperbolic problems not in conservation form.
For qt + f(q)x = 0, conservative if waves chosen properly,e.g. using Roe-average of Jacobians.
Great for general software, but only first-order accurate (upwindmethod for linear systems).
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 44: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/44.jpg)
Wave-propagation form of high-resolution method
Qn+1i = Qn
i −∆t∆x
m∑p=1
(spi−1/2)
+Wpi−1/2 +
m∑p=1
(spi+1/2)
−Wpi+1/2
− ∆t
∆x(F̃i+1/2 − F̃i−1/2)
Correction flux:
F̃i−1/2 =12
Mw∑p=1
|spi−1/2|
(1− ∆t
∆x|sp
i−1/2|)W̃p
i−1/2
where W̃pi−1/2 is a limited version of Wp
i−1/2 to avoid oscillations.
(Unlimited waves W̃p = Wp =⇒ Lax-Wendroff for a linearsystem =⇒ nonphysical oscillations near shocks.)
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 45: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/45.jpg)
Limiter methodsDifferencing Wp
i+1/2 −Wpi−1/2 approximates qxx.
Gives second order terms in Taylor series (Lax-Wendroff)This improves solution only if q is sufficiently smooth.
Limiters: Compare magnitude of Wpi−1/2 to corresponding wave
from adjacent Riemann problem in upwind direction.Apply limiter based on ratio of wave strengths.
Host of high-resolution methods developed since late 70’s: fluxcorrected transport, TVD methods, flux limiters, slope limiters,PPM, ENO, WENO, ...
Developed by: Boris, Book, Harten, Zwas, van Leer, Roe,Osher, Zalesak, Sweby, Colella, Woodward, Engquist,Chakravarthy, Shu, ...
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 46: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/46.jpg)
Limiter methodsDifferencing Wp
i+1/2 −Wpi−1/2 approximates qxx.
Gives second order terms in Taylor series (Lax-Wendroff)This improves solution only if q is sufficiently smooth.
Limiters: Compare magnitude of Wpi−1/2 to corresponding wave
from adjacent Riemann problem in upwind direction.Apply limiter based on ratio of wave strengths.
Host of high-resolution methods developed since late 70’s: fluxcorrected transport, TVD methods, flux limiters, slope limiters,PPM, ENO, WENO, ...
Developed by: Boris, Book, Harten, Zwas, van Leer, Roe,Osher, Zalesak, Sweby, Colella, Woodward, Engquist,Chakravarthy, Shu, ...
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 47: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/47.jpg)
Limiter methodsDifferencing Wp
i+1/2 −Wpi−1/2 approximates qxx.
Gives second order terms in Taylor series (Lax-Wendroff)This improves solution only if q is sufficiently smooth.
Limiters: Compare magnitude of Wpi−1/2 to corresponding wave
from adjacent Riemann problem in upwind direction.Apply limiter based on ratio of wave strengths.
Host of high-resolution methods developed since late 70’s: fluxcorrected transport, TVD methods, flux limiters, slope limiters,PPM, ENO, WENO, ...
Developed by: Boris, Book, Harten, Zwas, van Leer, Roe,Osher, Zalesak, Sweby, Colella, Woodward, Engquist,Chakravarthy, Shu, ...
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 48: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/48.jpg)
Wave-propagation form of high-resolution method
Qn+1i = Qn
i −∆t∆x
m∑p=1
(spi−1/2)
+Wpi−1/2 +
m∑p=1
(spi+1/2)
−Wpi+1/2
− ∆t
∆x(F̃i+1/2 − F̃i−1/2)
Correction flux:
F̃i−1/2 =12
Mw∑p=1
|spi−1/2|
(1− ∆t
∆x|sp
i−1/2|)W̃p
i−1/2
where W̃pi−1/2 is a limited version of Wp
i−1/2 to avoid oscillations.
(Unlimited waves W̃p = Wp =⇒ Lax-Wendroff for a linearsystem =⇒ nonphysical oscillations near shocks.)
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 49: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/49.jpg)
CLAWPACK — Conservation LAWs PACKagehttp://www.amath.washington.edu/~claw/
• Fortran codes with Matlab graphics routines.
• Many examples and applications to run or modify.
• 1d, 2d, and 3d.
• Adaptive mesh refinement, MPI for parallel computing.
User supplies:
• Riemann solver, splitting data into waves and speeds(Need not be in conservation form)
• Boundary condition routine to extend data to ghost cellsStandard bc1.f routine includes many standard BC’s
• Initial conditions — qinit.f
• Source terms — src1.f
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 50: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/50.jpg)
Software goals
• General framework for broad application.
• Adaptive Mesh Refinement (Berger/Oliger/Colella AMR).
Teaching:• Develop students intuition of how solutions behave, make it
easy to test, compare, extend methods.
• Textbook Finite Volume Methods for Hyperbolic Problems(Cambridge) — every figure has CLAWPACK code.
Research:• Application to new scientific/engineering problems.
• Extensions of methods to more challenging problems.
• Compare different methods on standard test problems.
Reproducible research!
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 51: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/51.jpg)
Software goals
• General framework for broad application.
• Adaptive Mesh Refinement (Berger/Oliger/Colella AMR).
Teaching:• Develop students intuition of how solutions behave, make it
easy to test, compare, extend methods.
• Textbook Finite Volume Methods for Hyperbolic Problems(Cambridge) — every figure has CLAWPACK code.
Research:• Application to new scientific/engineering problems.
• Extensions of methods to more challenging problems.
• Compare different methods on standard test problems.
Reproducible research!
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 52: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/52.jpg)
Software goals
• General framework for broad application.
• Adaptive Mesh Refinement (Berger/Oliger/Colella AMR).
Teaching:• Develop students intuition of how solutions behave, make it
easy to test, compare, extend methods.
• Textbook Finite Volume Methods for Hyperbolic Problems(Cambridge) — every figure has CLAWPACK code.
Research:• Application to new scientific/engineering problems.
• Extensions of methods to more challenging problems.
• Compare different methods on standard test problems.
Reproducible research!R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 53: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/53.jpg)
Some extensions and related software
• AMRCLAW — adaptive mesh refinement (Marsha Berger)
• CLAWMAN — on 2d manifolds (James Rossmanith, Derek Bale)
• BEARCLAW — f90 version (Sorin Mitran)
• AstroBEAR — including MHD (Adam Frank, Alexei Poludenko)
• ChomboClaw — interface to C++ (Donna Calhoun, Phil Colellaet.al.)
• AMROC — C++ AMR code (Ralf Deiterding)
• TsunamiClaw — SW over bathymetry (David George)
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 54: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/54.jpg)
Quadrilateral grids on the sphere
Recent work with Donna Calhoun and Christiane Helzel:
Mapping of Cartesian grid in rectangle [−3, 1]× [−1, 1] tosphere.
Ratio of largest to smallest cell is < 2.
Grid is highly non-orthogonal at a fewpoints near equator.
Movie of mapping
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 55: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/55.jpg)
Numerical results on the sphere
Direct application of CLAWPACK — wave-propagation finitevolume method
AMRCLAW can also be used.
Movie — shallow water on the sphere
Movie — depth vs. “latitude” compared to 1d solution
Movie — with adaptive mesh refinement
Movie — in computational plane
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 56: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/56.jpg)
Extensions and an example
• Shallow water equations over bathymetry with dry states
• Tsunami modeling
But first...
• Wave propagation algorithms for spatially varying fluxes
• Source terms and well balanced schemes
• Implementation via change in Riemann solver
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 57: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/57.jpg)
Extensions and an example
• Shallow water equations over bathymetry with dry states
• Tsunami modeling
But first...
• Wave propagation algorithms for spatially varying fluxes
• Source terms and well balanced schemes
• Implementation via change in Riemann solver
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 58: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/58.jpg)
Riemann problem for spatially-varying flux
qt + f(q, x)x = 0
Applications:
• Wave propagation in heterogeneous nonlinear media
• Flow in heterogeneous porous media
• Traffic flow with varying road conditions
• Solving conservation laws on curved manifolds
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 59: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/59.jpg)
Riemann problem for spatially-varying flux
qt + f(q, x)x = 0
Cell-centered discretization: Flux fi(q) defined in ith cell.
Need fi−1(Q∗i−1/2) = fi(Q∗
i−1/2) =⇒ m propagating wavesplus jump in q (= m waves for the m components of q).
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 60: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/60.jpg)
Flux-based wave decomposition (f-waves)
Choose waveforms rp (e.g. eigenvectors of Jacobian on eachside).
Then decompose flux difference:
fi(Qi)− fi−1(Qi−1) =m∑
p=1
βprp ≡m∑
p=1
Zp
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 61: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/61.jpg)
Wave-propagation algorithm using f-waves
Qn+1i = Qn
i −∆t∆x
[A+∆Qi−1/2 +A−∆Qi+1/2]
− ∆t∆x
[F̃i+1/2 − F̃i−1/2]
Standard version: Qi −Qi−1 =∑m
p=1Wpi−1/2
A−∆Qi+1/2 =m∑
p=1
(spi+1/2)
−Wpi+1/2
A+∆Qi−1/2 =m∑
p=1
(spi−1/2)
+Wpi−1/2
F̃i−1/2 =12
m∑p=1
|spi−1/2|
(1− ∆t
∆x|sp
i−1/2|)W̃p
i−1/2.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 62: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/62.jpg)
Wave-propagation algorithm using f-waves
Qn+1i = Qn
i −∆t∆x
[A+∆Qi−1/2 +A−∆Qi+1/2]
− ∆t∆x
[F̃i+1/2 − F̃i−1/2]
Using f -waves: fi(Qi)− fi−1(Qi−1) =∑m
p=1Zpi−1/2
A−∆Qi−1/2 =∑
p:spi−1/2<0
Zpi−1/2,
A+∆Qi−1/2 =∑
p:spi−1/2>0
Zpi−1/2,
F̃i−1/2 =12
m∑p=1
sgn(spi−1/2)
(1− ∆t
∆x|sp
i−1/2|)Z̃p
i−1/2
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 63: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/63.jpg)
Source terms and quasi-steady solutionsqt + f(q)x = ψ(q)
Steady-state solution:
qt = 0 =⇒ f(q)x = ψ(q)
Balance between flux gradient and source.
Quasi-Steady solution:
Small perturbation propagating against steady-state background.
qt � f(q)x ≈ ψ(q)
Want accurate calculation of perturbation.
Examples:• Shallow water equations with bottom topography and flat surface• Stationary atmosphere where pressure gradient balances gravity
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 64: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/64.jpg)
Fractional steps for a quasisteady problem
Alternate between solving homogeneous conservation law
qt + f(q)x = 0 (1)
and source termqt = ψ(q). (2)
Whenqt � f(q)x ≈ ψ(q):• Solving (1) gives large change inq• Solving (2) should essentially cancel this change.
Numerical difficulties:
• (1) and (2) are solved by very different methods. Generally will nothave proper cancellation.
• Nonlinear limiters are applied tof(q)x term, not to small-perturbationwaves. Large variation in steady state hides structure of waves.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 65: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/65.jpg)
Incorporating source term in f-waves
qt + f(q)x = ψ with f(q)x ≈ ψ.
Concentrate source at interfaces: Ψi−1/2 δ(x− xi−1/2)
Split f(Qi)− f(Qi−1)−∆xΨi−1/2 =∑
pZpi−1/2
Use these waves in wave-propagation algorithm.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 66: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/66.jpg)
Incorporating source term in f-waves
qt + f(q)x = ψ with f(q)x ≈ ψ.
Concentrate source at interfaces: Ψi−1/2 δ(x− xi−1/2)
Split f(Qi)− f(Qi−1)−∆xΨi−1/2 =∑
pZpi−1/2
Use these waves in wave-propagation algorithm.
Steady state maintained:
If f(Qi)−f(Qi−1)∆x = Ψi−1/2 then Zp ≡ 0
Near steady state:
Deviation from steady state is split into waves and limited.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 67: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/67.jpg)
Tsunamis
• Small amplitude in ocean (< 1 meter) but can grow to10s of meters at shore.
• Run-up along shore can inundate 100s of meters inland
• Long wavelength (as much as 200 km)
• Propagation speed√gh (bunching up at shore)
• Average depth of Pacific or Indian Ocean is 4km=⇒ average speed 200 m/s ≈ 450 mph
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 68: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/68.jpg)
Cross section of Indian Ocean & tsunami
Surface elevationon scale of 10 meters:
Cross-sectionon scale of kilometers:
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 69: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/69.jpg)
Shallow water equations with topography B(x)
ht + (hu)x = 0
(hu)t +(hu2 +
12gh2
)x
= −ghBx(x)
h(x, t) = depth of wateru(x, t) = horizontal velocity
This has the form of a conservation law with a source term:
qt + f(q)x = ψ(q, x),
where
q =[hhu
], f(q) =
[hu
hu2 + gh2/2
], ψ(q, x) =
[0
−ghB′(x)
].
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 70: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/70.jpg)
Shallow water equations with topography B(x, y)
ht + (hu)x + (hv)y = 0
(hu)t +(hu2 +
12gh2
)x
+ (huv)y = −ghBx(x, y)
(hv)t + (huv)x +(hv2 +
12gh2
)y
= −ghBy(x, y)
Applications:
• Tsunamis
• Estuaries
• River flooding, dam breaks
• Debris flows from volocanic eruptions
Methods to be discussed all extend to 2d (and 3d)
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 71: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/71.jpg)
The Riemann problem
The Riemann problem for qt + f(q)x = 0 has special initial data
q(x, 0) ={ql if x < xi−1/2
qr if x > xi−1/2
Dam break problem for shallow water equations
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 72: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/72.jpg)
The Riemann problem
The Riemann problem for qt + f(q)x = 0 has special initial data
q(x, 0) ={ql if x < xi−1/2
qr if x > xi−1/2
Dam break problem for shallow water equations
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 73: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/73.jpg)
The Riemann problem
The Riemann problem for qt + f(q)x = 0 has special initial data
q(x, 0) ={ql if x < xi−1/2
qr if x > xi−1/2
Dam break problem for shallow water equations
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 74: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/74.jpg)
The Riemann problem
The Riemann problem for qt + f(q)x = 0 has special initial data
q(x, 0) ={ql if x < xi−1/2
qr if x > xi−1/2
Dam break problem for shallow water equations
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 75: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/75.jpg)
The Riemann problem
The Riemann problem for qt + f(q)x = 0 has special initial data
q(x, 0) ={ql if x < xi−1/2
qr if x > xi−1/2
Dam break problem for shallow water equations
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 76: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/76.jpg)
Dry states and inundation
One-dimensional example:
Solved on uniform grid with h = 0 in dry cells.
Cells can change between wet and dry.
Riemann problems may have one side dry or generate drystate in solution.
Augmented Riemann solver of David George: Splits jump in[q1, q2 = f1, f2, B]T into four waves.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 77: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/77.jpg)
Dry states and inundation
One-dimensional example:
Solved on uniform grid with h = 0 in dry cells.
Cells can change between wet and dry.
Riemann problems may have one side dry or generate drystate in solution.
Augmented Riemann solver of David George: Splits jump in[q1, q2 = f1, f2, B]T into four waves.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 78: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/78.jpg)
Dry states and inundation
One-dimensional example:
Solved on uniform grid with h = 0 in dry cells.
Cells can change between wet and dry.
Riemann problems may have one side dry or generate drystate in solution.
Augmented Riemann solver of David George: Splits jump in[q1, q2 = f1, f2, B]T into four waves.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 79: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/79.jpg)
Dry states and inundation
One-dimensional example:
Solved on uniform grid with h = 0 in dry cells.
Cells can change between wet and dry.
Riemann problems may have one side dry or generate drystate in solution.
Augmented Riemann solver of David George: Splits jump in[q1, q2 = f1, f2, B]T into four waves.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 80: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/80.jpg)
Dry states and inundation
One-dimensional example:
Solved on uniform grid with h = 0 in dry cells.
Cells can change between wet and dry.
Riemann problems may have one side dry or generate drystate in solution.
Augmented Riemann solver of David George: Splits jump in[q1, q2 = f1, f2, B]T into four waves.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 81: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/81.jpg)
Dry states and inundation
One-dimensional example:
Solved on uniform grid with h = 0 in dry cells.
Cells can change between wet and dry.
Riemann problems may have one side dry or generate drystate in solution.
Augmented Riemann solver of David George: Splits jump in[q1, q2 = f1, f2, B]T into four waves.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 82: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/82.jpg)
Dry states and inundation
One-dimensional example:
Solved on uniform grid with h = 0 in dry cells.
Cells can change between wet and dry.
Riemann problems may have one side dry or generate drystate in solution.
Augmented Riemann solver of David George: Splits jump in[q1, q2 = f1, f2, B]T into four waves.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 83: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/83.jpg)
Sumatra event of December 26, 2004Magnitude 9.1 quake near Sumatra, where Indian tectonic plateis being subducted under the Burma platelet.
Rupture along subduction zone≈ 1200 km long, 150 km wide
Propagating at ≈ 2 km/sec (for ≈ 10 minutes)
Fault slip up to 15 m, uplift of several meters.(Fault model from Caltech Seismolab.)
www.livescience.com
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 84: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/84.jpg)
Tsunami simulations
• 2D shallow water + bathymetry
• Finite volume method
• Cartesian grid
• Cells can be dry (h = 0)
• Cells become wet/dry as wavemoves on shore
• Mesh refinement on rectangularpatches
• Adaptive — follows wave, morelevels near shore
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 85: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/85.jpg)
Tsunami simulations
Movies:
Fault area
Bay of Bengal
Sri Lanka
Indian Ocean
Zoom on Madras
For movies, see
http://www.amath.washington.edu/ ∼dgeorge/research.html
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 86: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/86.jpg)
Summary
• Wave propagation problems are generally formulated ashyperbolic systems.
• Many practical applications in science and engineering.
• General software for high-resolution methods, AMR:http://www.amath.washington.edu/~claw
• This talk, paper and codes:http://www.amath.washington.edu/~rjl/pubs/icm06
• Other papers and simulations:http://www.amath.washington.edu/~rjl/research.html
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 87: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/87.jpg)
xxx
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 88: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/88.jpg)
Satellite data
Jason-1 Satellite passed over the Indian Ocean during thetsunami event.
Surface height on two passes(one a week before)
Disparity shows tsunami:
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 89: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/89.jpg)
Comparison of simulation with satellite data
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 90: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/90.jpg)
Shock Wave Therapy and Lithotripsy
Setup: Pressure pulse:
• Shock wave lithotripsy is a well-established procedure fornoninvasive destruction of kidney stones.
• Typically several thousand shocks applied at rate of 1 to 4pulses per second.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 91: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/91.jpg)
Lithotripter at the UW Applied Physics Lab
Tom Matula (APL researcher) and Kirsten Fagnan (AMath grad student)
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 92: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/92.jpg)
Stresses in lithotripsy
Kidney stone Idealized cylinder
movie
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 93: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/93.jpg)
Mathematical model
• 2D elastic wave equations (+ axisymmetry)
• Preliminary results in 3D
• Models compression and shear waves
• Heterogeneous material
• Nonlinearity in compression waves
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 94: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/94.jpg)
Equations of linear elasticity
σ11t − (λ+ 2µ)ux − λvy = 0
σ22t − λux − (λ+ 2µ)vy = 0
σ12t − µ(vx + uy) = 0
ρut − σ11x − σ12
y = 0
ρvt − σ12x − σ22
y = 0
whereλ(x, y) andµ(x, y) are Lamé parameters.
This has the formqt +Aqx +Bqy = 0.
The matrix(A cos θ +B sin θ) has eigenvalues−cp, −cs, 0, cs, cp
where the P-wave speed and S-wave speed arecp =√
λ+2µρ , cs =
√µρ
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 95: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/95.jpg)
Extracorporeal Shock Wave Therapy (ESWT)
New uses are currently being tested
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 96: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/96.jpg)
Ankle bone data
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 97: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/97.jpg)
Model parameters
ρ λ µ cp cs
water 1000 2190.4 0.0 1.48 0.0brass 8560 6624.1 4528.2 1.36 0.73bone 1850 9305.5 3126.5 2.9 1.3“air” 100 3.9 0.0 0.2 0.0
Cartesian grid cells cut be interface contain averaged values.
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006
![Page 98: Wave Propagation Software, Computational Science, and ...faculty.washington.edu › rjl › talks › icm06 › icm06.pdfFor nonlinear problems wave speed generally depends on q. Waves](https://reader033.fdocuments.net/reader033/viewer/2022060403/5f0e86837e708231d43faa1e/html5/thumbnails/98.jpg)
Shock wave therapy simulations
Movie: With good acoustic coupling
Movie: With poor acoustic coupling
R. J. LeVeque International Congress of Mathematicians, Madrid, 2006