Upload
adam-robertson
View
224
Download
0
Tags:
Embed Size (px)
Citation preview
A comparison of radiation transport and diffusionusing PDT and the CRASH code
Fall 2011 Review
Eric S. MyraWm. Daryl Hawkins
Our goal is to quantify error associated with using flux-limited diffusion in CRASH
Key goals:
• Using PDT and CRASH, perform “method verification,” with the aim of improving the implementation of radiation diffusion and better understanding its shortcomings
• As necessary, perform code-to-code comparison and verification in the diffusion regime
• To the extent possible, set up the full CRASH problem in both codes and quantify the uncertainty of using diffusion vs. full transport
2
• This study was recommended by the 2010 Review Committee
• PDT/CRASH coupling not presently an option
An objective comparison oftransport and diffusion is challenging
• Differences in
— discretization and solution methods
— phase space coverage (full vs. a subset)
— treatment of multiphysics coupling (e.g., matter-radiation energy exchange)
• Characterizing the effects of ad hoc features of a model
— flux limiters in diffusion
— use of microphysics (e.g., opacities)
• Procedural differences
— e.g., the code may be used for a test problem in a different mode than for “real” problems (timestep selection, use of converged temperatures, etc.)
• A problem that’s easy for one code can be difficult for the other
3
Flux-limited diffusion approximates transport
The full transport equation (used by PDT).
The radiation energy equation (used by CRASH) is the zeroth angular moment of the transport equation with diffusive closure attained by Fick’s law.
with and
4
Flux-limited diffusion approximates transport
The full transport equation (used by PDT).
The radiation energy equation (used by CRASH) is the zeroth angular moment of the transport equation with diffusive closure attained by Fick’s law.
with and
5
Target problems determine how we use each code
PDT: a deterministic radiation transport code• Rad energy: gray and multigroup (both used)
• Rad angle: discrete ordinates (256 angles used)
• Spatial: discontinuous finite element method
• Time: fully implicit
CRASH: an Eulerian rad-hydro, flux-limited-diffusion code• Rad energy: gray and multigroup (both used)
• Rad angle: angle-averaged—0th angular moment equation, with 1st angular moment equation replaced by flux-limited diffusion
• Spatial: finite volume method
• Time: fully implicit
6
The starting point for comparison isdiffusion-limit test problems
• Gray transport
• Simple opacities, but which may vary sharply across an interface
• Examples:
— Infinite medium problems to test rates
— Front problems to test wave propagation
— Marshak waves to test propagation and rates
— Added heat sources as a proxy for shock heating
• Concerns:
— Choosing physically relevant timescales
— Computationally tractable in a reasonable time by both codes
— Defining “diffusive” for purposes of code comparison
If done with care, the codes should agree closely
7
Both codes advance a diffusive front similarly
• Gray transport
• Uniform density of 1 g cm-3
• Opacity = 105 cm2 g-1 in strip
• Opacity = 104 cm2 g-1 outside, but no emission-absorption
Te
Trad = 1 eV
Initial conditions
Results for radiationAt t = 3.0 ps…
• Results for each code are virtually identical for Trad
(PDT in maroon; CRASH in blue dashes)
• Te unchanged for both
• tdiff ~ 10 ns, tfs ~ 3.0 ps,
therefore diffusive8
A Marshak wave with a heat source also agrees well
• Gray transport
• Uniform density of 1 g cm-3
• Opacity = 105 cm2 g-1 in strip
• Opacity = 103 cm2 g-1 outside
• Emission-absorption active everywhere
• dQ/dt = 4.25 x 1033 eV cm-3 s-1 in central strip
Te
Trad = 1 eV
Initial conditions
At t = 100 ps, agreement is good
Material energy transport matches
Volume vs. surface effect?
9
PDTCRASH
Q added
A more realistic test problem has been formulated
0.10 cm 0.08 cm 0.20 cm
0.02 cm0.05 cm
Plastic
Plastic
Au
Au
Be:higher opacity
Post-shock
Xe
Pre-shockXe:
lower opacity
Sh
oc
ke
d X
e
0.0025 cm
0.0025 cm
0.0575 cm
• Hydrostatic
• 2D Cartesian
• No heat conduction
• Realistic opacities, using the CRASH tables
• A heat source acts as a proxy for shock heating
• Te = Trad = 1.0 eV, initially
• Cv (Xe, Au) = 9.9 x 1017 eV g-1 K-1
• Cv (Be, Pl) = 1.1 x 1019 eV g-1 K-1
• dQ/dt = 4.25 x 1033 eV cm-3 s-1
opacitycliff
The heat source is active within this region
10
A 1D gray version of the problem provides a first look
t = 2.0 ps t = 5.0 ps
t = 20.0 ps t = 50.0 ps
____ CRASH FLD on
_ _ _ CRASH FLD off
____ PDT Transport
Trad shows only qualitative agreement on this problem
t = 50.0 ps
Material energy transport differs
significantly
11
Agreement starts to improve in1D multigroup comparisons
____ CRASH FLD on
____ PDT Transport
Trad shows good agreement at early
times, then starts to diverge
• Material energy transport still
differs significantly.• However, in multigroup, PDT now
moves more energy, esp. upstream
“Upstream” radiative pre-
heating
t = 2.0 ps t = 5.0 ps
t = 20.0 ps t = 50.0 ps
t = 50.0 ps
12
10 groups, geometrically spaced, 1.0 eV–20 keV
These results suggest some next steps
• 1D Xe-on-polyimide problem—relevant to wall ablation
• Complete the suite of runs using the 2D version of the CRASH setup
• Implement a second problem using snapshots from full-system CRASH rad-hydro runs as initial conditions.
— Provides more realistic initial conditions (e.g., temperatures)
— Mitigates initial transients and uncertainties in the appropriate timescale over which to make comparisons
— Allows direct comparison between successive rad-hydro CRASH timesteps and PDT
A preliminary 2D result using CRASH showing Trad
13
Conclusions
• We have constructed a test environment that allows comparison of radiation transport and diffusion for problems relative to the CRASH.
• PDT and CRASH show good agreement on a set of problems where they should agree.
• PDT and CRASH show a mixture of agreement and discrepancy for more realistic CRASH-relevant problems.
• Further study is warranted to determine if these discrepancies are significant for predictive simulations of the CRASH experiment.
14