Upload
jiwei-li
View
212
Download
0
Embed Size (px)
Citation preview
ARTICLE IN PRESS
Contents lists available at ScienceDirect
Journal of Quantitative Spectroscopy &Radiative Transfer
Journal of Quantitative Spectroscopy & Radiative Transfer 110 (2009) 293–299
0022-40
doi:10.1
� Cor
E-m
journal homepage: www.elsevier.com/locate/jqsrt
Angular anisotropy of group averaged absorption coefficient and itseffect on the behavior of diffusion approach in radiative transfer
Jiwei Li �, Jinghong Li
Institute of Applied Physics and Computational Mathematics, P.O. Box 8009-14, Beijing 100088, China
a r t i c l e i n f o
Article history:
Received 17 April 2008
Received in revised form
15 October 2008
Accepted 5 November 2008
Keywords:
Radiative transfer
Multi-group method
Diffusion approximation
Anisotropy of group averaged absorption
coefficient
73/$ - see front matter Crown Copyright & 2
016/j.jqsrt.2008.11.001
responding author.
ail address: [email protected] (J. Li).
a b s t r a c t
Multi-group method, a generally accepted procedure for handling the radiative transfer
equation, is accompanied by the group averaged absorption coefficient, and actually the
coefficient is angularly anisotropic. In the paper, we present a brief discussion how the
anisotropy of the coefficient makes the material absorb photons at different rate in each
direction, also study its effect on the diffusion approach by comparing the results
calculated using multi-group diffusion and multi-group discrete ordinate SN for the
isotropic and anisotropic group averaged absorption coefficients respectively, and find
that the anisotropy deteriorates the behavior of diffusion approach.
Crown Copyright & 2008 Published by Elsevier Ltd. All rights reserved.
1. Introduction
Time-dependent non-equilibrium radiative transfer can be important in astrophysics, inertial confinement fusion, andhigh energy density physics applications. It is always of great cost to solve the equation of radiation transport accuratelydue to its complexity. To deal with the frequency-dependent variables in the equation, multi-group method is introduced[1,2], which really amounts to a discretization of the frequency variable. That is to say, the photons within one frequencygroup are treated the same, assigning them the same frequency-independent average properties, such as emission andabsorption coefficients. The method is least accurate for coarsen frequency groups and it gives increasingly better resultsas more and finer groups are used; of course, it also becomes very costly. To effectively reduce the computational time,multi-group diffusion approximation is preferable. The approach is a method of approximately describing the angulardistribution of radiation field in which the radiation flux in one group is assumed to be proportional to the gradient of theradiative energy density in the group and actually it solves the equations of the energy density and the flux for each groupregardless of the anisotropy of the radiation intensity in angular direction. Considerable work [3–7] has been done to studythe diffusion approximation and compare its results with the exact solutions for the gray case in which the absorptioncoefficient is presumed to be independent on the frequency. Su and Olson [8] also compared them for non-gray cases wherethey assumed the mean absorption coefficients for each group are constants, respectively. However their work does notconsider the angular anisotropy of the group averaged absorption coefficient in the multi-group radiative transfer equation.Actually the group averaged absorption coefficient strictly introduced in the multi-group method is angularly anisotropic[1] though it is independent on frequency, and the anisotropy may affect the behavior of diffusion approach. In this paper,first we will give a discussion on the anisotropy of the group averaged absorption coefficient in multi-group radiative
008 Published by Elsevier Ltd. All rights reserved.
ARTICLE IN PRESS
J. Li, J. Li / Journal of Quantitative Spectroscopy & Radiative Transfer 110 (2009) 293–299294
transfer equation, then compare the multi-group diffusion approximation to the multi-group discrete ordinate SN for boththe isotropic and anisotropic group averaged absorption coefficients and study the effect of the anisotropy in one group onthe accuracy of diffusion approximation for the group.
2. Multi-group method
Under the assumption of local thermodynamic equilibrium (LTE), the time-dependent radiative transfer equationwithout considering the source and scattering terms in a one-dimensional planar geometry can be given by
1
c
qInðx;m; tÞqt
þ mqInðx;m; tÞqx
¼ s0aðn; TmÞ½BnðTmÞ � Inðx;m; tÞ�, (1)
where x, m and t are the space, angular direction cosine, and time variables, respectively; Iv is the spectral radiationintensity; BnðTmÞ ¼ ð2hn3=c2Þð1=ehn=kTm � 1Þ the spectral intensity for equilibrium radiation; c the speed of the light; Tm thematerial temperature and sa
0 the effective absorption coefficient.Taking integration of both sides of Eq. (1) over the photon frequency interval of each group (frequency interval Dvg), we
obtain the multi-group radiative transfer equation
1
c
qIgðx;m; tÞqt
þ m qIgðx;m; tÞqx
¼ sbgBgðTmÞ �
ZDng
s0aInðx;m; tÞdn, (2)
where Ig ¼RDng
In dn, Bg ¼RDng
Bn dn and sbg ¼
RDng
s0aBnðTmÞdn=Bg is the Planck’s mean value in each group for sa0. In order
to deal with the absorption term, we expand the radiative intensity in Legendre polynomials according to
Inðx;m; tÞ ¼X1n¼0
2nþ 1
4pInðx; n; tÞPnðmÞ. (3)
where due to the orthogonality of the PnðmÞ, the expansion coefficients are given by
Inðx; n; tÞ ¼ 2pZ 1
�1PnðmÞIðx; n;m; tÞdm.
Inserting Eq. (3) into the frequency-integrated absorption term in Eq. (2), the multi-group radiative transfer equation iswritten as
1
c
qIgðx;m; tÞqt
þ m qIgðx;m; tÞqx
¼ sbgBgðTmÞ �
X1n¼0
2nþ 1
4p PnðmÞsngIg
n, (4)
where sng ¼
RDng
s0aInðx; n; tÞdn=Ign, the nth order group averaged absorption coefficient and Ig
n ¼RDng
Inðx; n; tÞdn. For n ¼ 0,I0ðnÞ ¼ 2p
R 1�1 InðmÞdm ¼ cEn, here Ev is the spectral radiative energy, then we have s0
g ¼RDng
s0aEn dn=Eg and name it theisotropic group averaged absorption coefficient, where Eg ¼
RDng
En dn. Because Ev here is unknown, the zeroth order meancoefficient is always used with an approximation s0
g ¼ spg ¼
RDng
s0aBnðTrÞdn=BgðTrÞ, the Planck mean for each group, wherethe Planck function having a local radiation temperature Tr is to get a better spectrum [9–11] where the radiative field is outof equilibrium with the material.
Then Eq. (4) can be written as
1
c
qIg
qtþ m
qIg
qx¼ sb
gBg � spgIg þ
X1n¼1
2nþ 1
4p PnðmÞðspg � s
ng ÞI
gn. (5)
2.1. Group averaged absorption coefficient is angularly isotropic
If we assume that the effective absorption coefficient sa0 happens to be constant in each group, the high order terms
(nX1) in Eq. (5) will get disappeared. Then Eq. (5) degenerates into be
1
c
qIg
qtþ m qIg
qx¼ sb
gBg � spgIg . (6)
clearly the group averaged absorption coefficient spg is angularly isotropic.
Now we will derive the multi-group P1 diffusion equation by virtue of the angular moments of Eq. (6). Actually here ourprocedure is to take integration of radiative transfer Eq. (1) first over frequency then over angle rather than the traditionalone to take integration first over angle then over frequency [12]. Multiplying Eq. (6) successively by 1 and m, and integratingover angle, we can obtain the equations of spectral radiative energy and radiative flux in each group
qEg
qtþqFg
qx¼ 4psb
gBgðTmÞ � cspgEg , (7)
ARTICLE IN PRESS
J. Li, J. Li / Journal of Quantitative Spectroscopy & Radiative Transfer 110 (2009) 293–299 295
1
c
qFg
qtþ
Z 1
�1mm qIg
qxdm ¼ �sp
gFg , (8)
here Fg ¼RDng
Fn dn> and Fn ¼ 2pRDng
mIn dm, the spectral radiative flux. To obtain the P1 diffusion approximation, we makethe further assumption that the time derivative of the radiation flux is negligible and substitute Ig ¼ ðc=4pÞEg þ ð3=4pÞmFg
into Eq. (8), then the radiative diffusion flux in one group is given by [6]
Fg ¼ �c
3spg
qEg
qx. (9)
Eqs. (7) and (9) are the corresponding multi-group diffusion equations of Eq. (6). Since we neglect the angle-dependentabsorption terms in Eq. (5), the group diffusion coefficient in Eq. (9) is determined by the Planck mean free path rather thanthe traditional Rosseland mean free path.
For this case that the group averaged absorption coefficient is isotropic, Su and Olson [8] compared the results of two-group diffusion and two-group radiative transfer and concluded that the accuracy and validity of two-group diffusionapproximation will be deteriorated as the material’s absorption becomes weaker, which also means that the diffusionapproach behaves poorly for optically thin materials.
We can also know that it seems invalid to only keep the zeroth order term in Eq. (5) unless the effective absorptioncoefficient sa
0 is constant in each group because the corresponding diffusion coefficient is determined by the Planck meanfree path rather than the Rosseland mean free path. However it does not matter because our motivation here is to study theaccuracy and validity of diffusion approach.
2.2. Group averaged absorption coefficient is angularly anisotropic
Because the absorption coefficients encountered in practice are generally complex and widely varying functions offrequency, the higher order terms in Eq. (5) have to be kept. Considering them, we can introduce an effective groupaveraged absorption coefficient sg
eff which satisfies
seffg Ig ¼ sp
gIg �X1n¼1
2nþ 1
4pPnðmÞðsp
g � sng ÞI
gn. (10)
It is obvious that the sgeff is angularly anisotropic because sp
gasng for nX1. As we know, the validity of diffusion
approximation is dependent on the angular distribution of the radiative intensity, so the anisotropy may affect the accuracyof diffusion approximation for it will redistribute the radiative intensity in angular direction.
From the angular moments of Eq. (5), we could obtain its corresponding multi-group diffusion equations
qEg
qtþqFg
qx¼ 4psb
gBgðTeÞ � cspgEg , (11)
Fg ¼ �c
3s1g
qEg
qx. (12)
Now we can see that the diffusion coefficient here is determined by the first order group averaged absorption coefficient(it can be approximated by Rosseland mean free path as discussed below), which agrees with the traditional multi-groupdiffusion approximation. The two different procedures to derive the multi-group diffusion reach the same result. Physicallyspeaking, the multi-group Eq. (5) is much more reasonable and accurate than Eq. (6).
Allowing for the convenience to achieve the numerical simulation, we truncate the high order terms at n ¼ 1, thenEq. (5) becomes
1
c
qIg
qtþ mqIg
qx¼ sb
gBg � spgIg þ
3
4pmðspg � s
1g ÞFg . (13)
For n ¼ 1, I1ðnÞ ¼ Fn, the radiation flux and s1g Fg ¼
RDng
s0aFn dn. Actually here we handle the group absorption term in
Eq. (5) with P1 approximation, so we can assume s0aFn ¼ �ðc=3ÞqEn=qx, then s1g
RDngð1=s0aÞqEn=qx dn ¼
RDngðqEn=qxÞdn.
Further we have s1g ¼ sR
g ¼RDng
qBnðTrÞ=qTr dn=RDngð1=s0aÞqBnðTrÞ=qTr dn, the Rosseland mean for each group, then Eq. (13)
becomes the new multi-group method [1] and its effective absorption coefficient for each group
seffg ðmÞ ¼ sp
g �3
4pmðsp
g � sRg ÞFg
Ig. (14)
From this equation, we can see that the absorption rate of forward moving photons becomes lower than that of backwardmoving photons, and the anisotropy plays a role where the Rosseland and Planck means vary widely.
Comparing Eq. (11) to Eq. (7), we would find that the energies transported from radiation to materials have the samedependence on the radiation energy density, though the group averaged absorption coefficient is isotropic in Eq. (6) andanisotropic in Eq. (5). That is to say, though the anisotropy makes the material become absorbing the photons within onegroup at different rate for each direction, the total absorption integrated over all the direction and frequency remains
ARTICLE IN PRESS
J. Li, J. Li / Journal of Quantitative Spectroscopy & Radiative Transfer 110 (2009) 293–299296
literally sameP
g
R4p s
pgIg dO ¼
Pg
R4p s
effg Ig dO ¼ c
Pgs
pgEg . In the following numerical simulation, we will compare the
multi-group diffusion to the multi-group discrete ordinate SN for the two cases respectively to study the effect of theanisotropy of the group averaged absorption coefficient on the accuracy and validity of diffusion approximation.
3. Numerical simulation
The simulation is achieved by virtue of the code RDMG [13] (1D radiation diffusion multi-group) with an average atomopacity model. In this simulation, the plastic (CH) planar slab is exposed to a constant radiation field with T ¼ 149.3 eV onone side and the other side is the vacuum boundary, and the radiation field is assumed to be Planckian in frequency andisotropic in angle. The hydrodynamics is treated by solving the standard mass, momentum and energy conservationequations. For the material CH in the simulation, the 20-group opacities and tabulated equation of state are used [10].Results are compared between the calculations using multi-group discrete ordinate S8 and multi-group diffusion for theisotropic and anisotropic group averaged absorption coefficient, respectively. To better compare the 20-group diffusionapproach and 20-group S8, first we get the time-dependent net radiation flux of each group at the exposed boundary fromS8 results, then set those net flux of each group as the drive source (net diffusive flux) of the multi-group diffusion. Bydoing that, we can guarantee that the drive spectrum of the diffusion approach is closely matched to that of full transfer[14]. Compared to the isotropic case, the anisotropic group averaged absorption coefficient will result in the net radiationflux at the exposed side increase as time goes by though it contributes nothing at the initial stage, correspondingly for somegroups, such as group No. 11 (348.1�425.7 eV), their net radiation flux at the exposed side also increase, as shown in Fig. 1,and the heat wave also propagates faster [15]. To clearly illustrate the effect of the anisotropy on the diffusion approach, wenormalized the space by the position of heat wave front xF and the radiation flux in one group by its net radiation flux at theexposed side Fnet
g_b, respectively when comparing the multi-group SN and multi-group diffusion for the two cases.Fig. 2 is the spatial distribution of the radiation temperature and Fig. 3 is the spectrum of the radiation flux at the
position x=xF ¼ 0:8 at time t ¼ 1.5 ns for two cases, respectively. For the isotropic case, the diffusive heat wave propagatesonly a little ahead of the transport theory and the corresponding spectrum also agrees well with that of the S8 results. Bythe comparison between the anisotropic case and isotropic one, it is clear that the anisotropy of the group averagedabsorption coefficient makes the diffusion approach predict the worse frequency spectrum compared to the S8 results,though the worse of radiation temperature is not too significant. The radiation flux of those photons whose energy are300 eV or so is overestimated by the diffusion approach, and that will also be shown in Fig. 6.
The angular anisotropy caused by the high order term (n ¼ 1) in Eq. (13) retards the decrease of the radiative intensityfor those photons moving forward and the increase of that for those photons moving backward, therefore it willredistribute the radiative field in angular direction. Fig. 4 is the angular distribution of the group’s mean absorptioncoefficient and radiative intensity of group No.11 at time t ¼ 1.5 ns at the position x=xF ¼ 0:6 where the Planck andRosseland means for the group vary widely. Near the position, the distribution of the absorption coefficient becomesanisotropic seriously, correspondingly the radiative intensity also becomes much more anisotropic. That is to say, the roleof the anisotropy of group mean absorption coefficient is to prevent the radiative field from tending to being isotropic. Thespectral radiation flux is determined by the angular distribution of radiative intensity, so it is sensitive to the anisotropy ofgroup averaged absorption coefficient and easy to be affected by it, as shown in Fig. 1. As the radiative field becomesangularly anisotropic, the behavior of the diffusion approach will become deteriorated, such as the diffusion flux. From thecomparison of the radiation flux for group No.11 between the diffusion approach and discrete ordinate S8 shown in Fig. 5,
Fig. 1. The net radiation flux for group No. 11 at the exposed side for the isotropic and anisotropic group averaged absorption coefficients respectively.
ARTICLE IN PRESS
Fig. 2. Comparison of the radiation temperature between 20-group S8 and 20-group diffusion at time t ¼ 1.5 ns for the isotropic and anisotropic group
averaged absorption coefficients respectively.
Fig. 3. Comparison of the spectral radiation flux between 20-group S8 and 20-group diffusion at the position x=xF ¼ 0:8 at time t ¼ 1.5 ns for the isotropic
and anisotropic group averaged absorption coefficients, respectively.
Fig. 4. The angular distribution of the mean absorption coefficient (normalized by sgp) and radiative intensity (normalized by I0max the maximum for all
the direction each case) for group No.11 at the position x=xF ¼ 0:6 and time t ¼ 1.5 ns.
J. Li, J. Li / Journal of Quantitative Spectroscopy & Radiative Transfer 110 (2009) 293–299 297
ARTICLE IN PRESS
Fig. 5. Comparison of the radiation flux in group No.11 at time t ¼ 1.5 ns between 20-group S8 and 20-group diffusion for the isotropic and anisotropic
mean absorption coefficients respectively.
Fig. 6. Comparison of the radiation flux in group No.10 at time t ¼ 1.5 ns between 20-group S8 and 20-group diffusion for the isotropic and anisotropic
mean absorption coefficients respectively.
J. Li, J. Li / Journal of Quantitative Spectroscopy & Radiative Transfer 110 (2009) 293–299298
we can see that the diffusion description becomes less accurate behind the heat wave for the case that the group averagedabsorption coefficient is anisotropic. For the radiation flux of group No.10 (284.7�348.1 eV) shown in Fig. 6, it is much moreevident that the diffusive flux is impracticably overestimated near the heat wave front where the anisotropy disables thediffusion approach.
4. Conclusion and discussion
In the paper, we analyzed the angular anisotropy of the group averaged absorption coefficient and compared the multi-group diffusion and multi-group discrete ordinate SN for the coefficients are isotropic and anisotropic, respectively. Theanisotropy plays a significant role where the Rosseland and Planck means for each group vary widely—for some groupsnear the heat wave front, while others behind the heat wave, and it deteriorates the results predicted by diffusionapproximation, especially for the spectrum and the radiation flux for each group. For the fine enough groups, theanisotropy will become less important because two means are closed to each other.
Here we only consider the anisotropy caused by the high order term with n ¼ 1. If other higher order terms are included,the anisotropy will become more complicated and may result in a more significant effect on the accuracy of diffusionapproximation.
The scattering is not considered in all the work here, otherwise, the group averaged absorption coefficient should beredefined by sn
g ¼R
nngðs0a þ ssÞInðx; n; tÞdn=Ig
n, where the ss is the scattering coefficient.
Acknowledgements
We are grateful to the radiation transport crew at IAPCM and Prof. Wenbing Pei for fruitful and valuable discussion. Wewould like also to thank Prof. Xujun Meng for his Atomic Parameter Databases.
This work was jointly supported by the National Natural Science Foundation of China under Grant No. 10775019 and bythe Laboratory of Computational Physics, China.
ARTICLE IN PRESS
J. Li, J. Li / Journal of Quantitative Spectroscopy & Radiative Transfer 110 (2009) 293–299 299
References
[1] Pomraning GC. The equations of radiation hydrodynamics. New York: Pergamon Press; 1973.[2] Mihalas D, Mihalas BW. Foundations of radiation hydrodynamics. New York Oxford: Oxford University Press; 1984.[3] Zel’dovich YB. Physics of shock waves and high-temperature hydrodynamic phenomena. New York and London: Academic Press; 1966.[4] Castor JI. Radiation hydrodynamics. New York: Cambridge University Press; 2004.[5] Olson GL, Auer LH, Hall ML. Diffusion, P1, and other approximate forms of radiative transport. JQSRT 2000;64:619–34.[6] Brantley PS. Angularly adaptive P1-double P0 flux-limited diffusion solutions of non-equilibrium grey radiative transfer problems. JQSRT
2007;104:116–32.[7] Su B, Olson GL. An analytical benchmark for non-equilibrium radiative transfer in an isotropically scattering medium. Annals of Nuclear Energy
1997;24:1033–55.[8] Su B, Olson GL. Non-grey benchmark results for two temperature non-equilibrium radiative transfer. JQSRT 1999;62:279–302.[9] Xu Y, Lai DX, Feng TG. Establishment of multi-groups atomic parametric database. High Power Laser and Particle Beams 1997;9:533–7 [in Chinese].
[10] Feng TG, Lan K, Lai DX. A comparison between two averaging methods of multi-group parameters in ICF radiation transfer calculation. ChineseJournal of Computational Physics 2001;18:206–10 [in Chinese].
[11] Williams RB. Adaptive multi-group radiation diffusion. PhD thesis, Massachusetts Institute of Technology; 2005.[12] Brunner TA, Mehlhorn T, McClarren R, Kurecka CJ. Advances in radiation modeling in ALEGRA: a final report for LDRD-67120, Efficient implicit
multigroup radiation calculations. Sandia Report, Sandial National Laboratories; 2005.[13] Feng TG, Lai DX, Xu Y. An artificial-scattering iteration method for calculating multigroup radiation transfer problems. Chinese Journal of
Computational Physics 1999;16:199–205 [in Chinese].[14] Haan SW, Herrmann MC, Dittrich TR, Fetterman AJ, Marinak MM, Munro DH, et al. Increasing robustness of indirect drive capsule designs against
short wavelength hydrodynamic instabilities. Physics of Plasmas 2005;12:056316.[15] Li JW, Li H. Study on the multi-group method of radiative transfer equation. Annual Report, Institute of Applied Physics and Computational
Mathematics 2007 [in Chinese].