physics0604223/sr.tex
1: \documentclass{elsart}
2: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
3: \usepackage{natbib,epsfig}
4: \usepackage{amsfonts}
5: %\usepackage{graphicx}
6: %\usepackage{dcolumn}
7: %\usepackage{bm}
8: 
9: %\setcounter{MaxMatrixCols}{10}
10: 
11: \begin{document}
12: \runauthor{Salzmann and Ralchenko}
13: \begin{frontmatter}
14: \title{New parametrization for differences between plasma kinetic codes}
15: \author{David Salzmann\thanksref{aff}} and
16: \author{Yuri Ralchenko\thanksref{corresp_author}}
17: %\email{yuri.ralchenko@nist.gov}
18: \address{Atomic Physics Division, National Institute of Standards and Technology,
19: Gaithersburg, MD 20899-8422}
20: %\date{\today}
21: \thanks[aff]{\emph{Permanent address}: Soreq Nuclear Research Center, Yavne,
22: Israel 81800.}
23: \thanks[corresp_author]{\emph{Corresponding author. Email}:
24: yuri.ralchenko@nist.gov.}
25: 
26: \begin{abstract}
27: Validation and verification of plasma kinetics codes 
28: requires the development of quantitative methods and techniques for code
29: comparisons. We describe two parameters that can be
30: used for characterization of differences between such codes. It is shown
31: that these parameters, which are determined
32: from the most general results of kinetic codes, can provide
33: important information on the differences between the basic rate coefficients
34: employed. Application
35: of this method is illustrated by comparisons of some results from the 3rd
36: NLTE Code Comparison Workshop for carbon, germanium, and gold plasmas.
37: \end{abstract}
38: 
39: %\pacs{52.25.Jm, 52.20.-j}
40: %\maketitle
41: 
42: % PACS, the Physics and Astronomy
43: % Classification Scheme.
44: %\keywords{Suggested keywords}%Use showkeys class option if keyword
45: %display desired
46: 
47: \end{frontmatter}
48: 
49: \section{\label{sec:intro}Introduction}
50: 
51: Spectroscopy of plasmas has a wide range of applications \cite{book,Griem},
52: including the study of astrophysical objects, diagnostics of laser produced
53: plasmas, and analysis of radiation emission from plumes of rockets. In spite of
54: the importance of this subject and the substantial efforts made by numerous
55: groups, there persist significant discrepancies in results of the plasma kinetic
56: codes used to analyze plasma emission spectra under
57: non-local-thermodynamic-equilibrium (NLTE) conditions. For examples of the level
58: of disagreement see, e.g., the reports from the NLTE Code Comparison Workshops
59: \cite{NLTE1,NLTE2,NLTE3}.
60: 
61: Given the plasma particle temperature and density, the first task of kinetic
62: codes is the computation of the charge state and excitation state distributions
63: \cite{book,Griem}. These are typically determined from the set of rate
64: equations:
65: 
66: \begin{eqnarray}
67: \frac{d\,N_{\zeta m}}{dt} & = &\sum_{all ~ populating ~ processes}
68: {n_{e}^{k}N_{\zeta ^{\prime }m^{\prime }}\mathbb{R}_{\zeta
69: ^{\prime }m^{\prime }\rightarrow \zeta m}} - \nonumber \\
70: & & \sum_{all ~ depopulating ~ processes}{n_{e}^{k}N_{\zeta m}\mathbb{R}_{\zeta
71: m\rightarrow \zeta ^{\prime }m^{\prime }}},  \label{Eq1}
72: \end{eqnarray}
73: where $N_{\zeta m}$ is the density of ions having charge $\zeta $ ( $%
74: 0\leq \zeta \leq Z$ with $Z$ being the nuclear charge) 
75: excited to state $m$ (ordered according to their
76: ascending energy, $m=0$ corresponds to the ion's ground state), $n_{e}$ is
77: the electron density, and $\mathbb{R}_{a\rightarrow b}$ denotes 
78: the rate coefficient (r-c) for the transition of an ion from state $a$ to
79: state $b$ due to some atomic process. In Eq. (\ref{Eq1}), $k$ represents the
80: number of
81: electrons taking part in any given process, thus, $k=0$ for spontaneous
82: decay and autoionization, $k=1$
83: for the electron impact processes (e.g., excitation and photorecombination), and
84: $k=2$ for the three-body recombination. In large-size plasmas, where linear
85: dimension is larger than the photon mean free path, photon induced
86: processes must also be included in (\ref{Eq1}). In a steady state plasma,
87: $d\,N_{\zeta m}/dt=0$, and Eq. (\ref{Eq1}) reduces to:
88: 
89: \begin{equation}
90: \sum_{all ~ populating ~ processes}{
91: n_{e}^{k}\,N_{\zeta ^{\prime }m^{\prime }} \mathbb{R}_{\zeta ^{\prime
92: }m^{\prime }\rightarrow \zeta m}} = \sum_{all  ~ 
93: depopulating  ~ processes}{n_{e}^{k}N_{\zeta m} \mathbb{R}_{\zeta
94: m\rightarrow \zeta ^{\prime }m^{\prime }}}.  \label{Eq2}
95: \end{equation}
96: This is a finite set of non-linear coupled equations for $N_{\zeta m}$
97: 's. If one is interested only in the density of the charge states,
98: regardless of the ionic excitations, Eq. (\ref{Eq2}) can be further simplified, and
99: the result is a set of recursive equations:
100: 
101: \begin{equation}
102: \frac{N_{\zeta +1}}{N_{\zeta }}=\frac{I_{\zeta \rightarrow \zeta +1}}{%
103: R_{\zeta +1\rightarrow \zeta }^{(2)}\,+\,\,\,n_{e}\,R_{\zeta +1\rightarrow
104: \zeta }^{(3)}}.  \label{Eq3}
105: \end{equation}
106: Here $N_{\zeta }=\sum_{m}N_{\zeta m}$ is the density of the charge
107: state, and \ $I_{\zeta \rightarrow \zeta +1}$, $R_{\zeta +1\rightarrow \zeta
108: }^{(2)}$ and $R_{\zeta +1\rightarrow \zeta }^{(3)}$ are the \textit{total
109: rate coefficients} for ionization, two-body (radiative+dielectronic)
110: recombination, and three-body recombination, respectively. For quasineutral
111: plasmas, solutions of
112: Eq.(3) have to satisfy two complementary conditions, namely,
113: 
114: \begin{equation}
115: n_{i}=\sum\limits_{\zeta =0}^{Z}\,N_{\zeta }\;;\quad
116: n_{e}=\sum\limits_{\zeta =0}^{Z}\,\zeta \,N_{\zeta }=\overline{Z}\,n_{i}.
117: \end{equation}
118: Here $n_i$ is the total ion density, and $\overline{Z}$ is the
119: average charge of the ions. In the following we assume that $N_{\zeta}$
120: is the fractional abundance of charge state $\zeta$, which is equivalent to
121: the assumption that $n_i$ = 1.
122: 
123: The disagreement between the results of the kinetic codes developed by various
124: researchers may originate from several factors. One of the most important is the
125: approximate character of the r-c's. In the literature one can find several
126: recommended formulas for the relevant r-c's. Some may be more accurate than
127: others, but none was shown to have high accuracy. For higher accuracy, the r-c's
128: can be directly calculated for each transition between atomic states, using
129: advanced quantum-mechanical methods. The necessity to generate a large number of
130: r-c's for kinetic calculations, however, impedes the application of these
131: techniques, thereby forcing a compromise between computational speed and
132: accuracy. The solutions of (\ref{Eq3}) obviously depend on the methods chosen
133: for the determination of the r-c's in the right-hand side of (\ref{Eq3}).
134: 
135: Another source of disagreement is the criterion used for the continuum or ionization
136: potential lowering in plasmas. This is particularly important in high-density
137: plasmas where the plasma potential moves the upper ionic bound states into the
138: continuum, leaving the ion only with a finite number of discrete states.
139: Moreover, due to the fluctuations of the local microfield around each ion, even
140: the "tightly bound states" may change into instantaneous quasi-molecular states
141: whose treatment is, as yet, not clear.
142: 
143: Comparison of the results of NLTE kinetic codes was a subject of several
144: workshops \cite{NLTE1,NLTE2,NLTE3}. The participants were asked to submit large
145: sets of various physical quantities to be compared, and in numerous cases very
146: significant differences were found. This situation is well exemplified in Fig.
147: \ref{fig1}, where the various calculations of the relative ion populations for a
148: germanium plasma are presented for a specific case from the NLTE-3 Workshop
149: \cite{NLTE3}. One can clearly notice a significant spread  both for the mean ion
150: charges and for the distribution widths calculated with different kinetic codes.
151: Although a variety of physical parameters investigated at the Workshop was
152: mostly sufficient to draw conclusions about sources of discrepancies,
153: introduction of simple and clearly defined new parameters that would pinpoint
154: some fundamental underlying differences would greatly facilitate such
155: comparisons. 
156: 
157: %\begin{figure}[tbp]
158: %\centerline{\includegraphics[width=0.8\linewidth]{fig1.eps}}
159: %\label{fig1}
160: %\end{figure}
161: 
162: \begin{figure*}[t]
163: \begin{minipage}{15cm}
164: \epsfig{file=fig1.eps,width=14cm}
165: \end{minipage}
166: \vspace*{1cm}
167: \caption{Relative ion populations for a steady-state germanium plasma at electron
168: temperature T$_e$ = 250 eV and electron density N$_e$ = $10^{17}$ cm$^{-3}$
169: \cite{NLTE3,SAHA}.}
170: \label{fig1}
171: \end{figure*}
172: 
173: 
174: In Ref. \cite{PRA} one of the authors developed a method that allows 
175: determination of variations in an ionization state distribution due to small
176: changes in the rates of basic atomic processes. The purpose of the present paper
177: is the opposite, that is, to introduce new parameters that quantify differences
178: in the {\it input} atomic rates in kinetic codes using the {\it results} of
179: calculations. Such approach is similar to solution of the inverse problem in
180: physics. 
181: 
182: Below we address a simple but critical question, namely: ``How different are two
183: population kinetic codes?" It is obvious that the most comprehensive answer can
184: be obtained provided all input and output parameters as well as the complete
185: description of the approximations used are available. Unfortunately, this is not
186: always the case. Here we introduce parameters that are straightforward to
187: calculate and can provide an answer in a clear manner. An important feature of
188: the proposed method is that only the most general kinetic characteristics, such
189: as the mean ion charge and central momenta, are required to determine these new
190: parameters. 
191: 
192: 
193: The paper is organized as follows. In the next section we define these
194: parameters and explain their meanings. In Section \ref{sec3} we apply the method
195: to characterize the differences between code results for three representative
196: cases from the NLTE-3 Workshop \cite{NLTE3}. Finally, in Section \ref{sec4}, a
197: short summary and conclusions are presented.
198: 
199: \section{\label{sec2}Quantitative characterization of kinetic codes}
200: 
201: We introduce the notation $Q_{\zeta }$ for the
202: r.h.s. of Eq.~(\ref{Eq3}):
203: 
204: \begin{equation}
205: Q_{\zeta }=\frac{I_{\zeta \rightarrow \zeta +1}}{R_{\zeta +1 \rightarrow
206: \zeta }^{(2)} + n_e R_{\zeta +1\rightarrow \zeta }^{(3)}}
207: \end{equation}
208: and denote by $\Delta Q_{\zeta } \equiv Q_{ref} - Q$ the overall difference in
209: this quantity for 
210: a given code relative to the reference. 
211: %It is 
212: %$\Delta Q_{\zeta }$ that is the measure of the inaccuracies of the
213: %calculations. 
214: Using this definition, Eq.(3) is rewritten as:
215: 
216: \begin{equation}
217: \frac{N_{\zeta +1}}{N_{\zeta }}=Q_{\zeta }.
218: \end{equation}
219: It was shown in Ref. \cite{PRA}, that the relative difference in the final
220: results, $\Delta N_{\zeta }/N_{\zeta }$, caused by the relative differences
221: used in the input data, $\Delta Q_{\zeta }/Q_{\zeta }$, is given by
222: 
223: \begin{equation}
224: \frac{\Delta N_{\zeta }}{N_{\zeta }}{\ =\alpha }_{\zeta }\,{\ -\,}\overline{
225: \alpha },  \label{dn1}
226: \end{equation}
227: where
228: 
229: \begin{equation}
230: {\alpha }_{\zeta }=\sum\limits_{\zeta ^{\prime }=0}^{\zeta }\frac{\Delta
231: Q_{\zeta ^{\prime }}}{Q_{\zeta ^{\prime }}}\;; ~~ \\
232: \overline{\alpha }=\sum\limits_{\zeta =0}^{Z}\alpha _{\zeta }\frac{%
233: N_{\zeta }}{n_{i}}. \label{al1}
234: \end{equation}
235: In particular, if the relative differences are all equal, then
236: \begin{equation}
237: \Delta Q_{\zeta}/Q_{\zeta } \equiv p=const, \label{dQ}
238: \end{equation}
239: and Eq.(\ref{dn1}) reduces to the simple form \cite{PRA}:
240: 
241: \begin{equation}
242: \frac{\Delta N_{\zeta }}{N_{\zeta }} = p(\zeta -\overline{Z}).  \label{dn0}
243: \end{equation}
244: In Ref. \cite{PRA} a full discussion is presented about the meaning of
245: Eq. (\ref{dn0}), as well as how and when this difference influences the
246: results of kinetic codes. Even if $\Delta Q_{\zeta }\bigskip /Q_{\zeta }$ is
247: not constant, one can define their average value, $p\equiv \left\langle
248: \Delta Q_{\zeta }\bigskip /Q_{\zeta }\right\rangle _{\zeta }$, over the
249: relevant ion charge states.
250: 
251: The above formulas were derived assuming small differences between
252: code results. In a general case, however, the deviations are not small, and one
253: has to symmetrize the relevant parameters with respect to both compared codes, i.e.:
254: 
255: \begin{equation}
256: N_{\zeta} = \frac{1}{2}(N_{1,\zeta}+N_{2,\zeta}), 
257: ~ \overline{Z} = \frac{1}{2} (\overline{Z}_1+\overline{Z}_2),
258: \end{equation}
259: and so on, which is assumed in what follows.
260: 
261: Multiplying both sides of Eq.(\ref{dn0}) by $(\zeta -\overline{Z}) N_{\zeta}$ and summing over
262: all charge states, one arrives at:
263: \begin{equation}
264: p = \frac{\Delta \overline{Z}}{\overline{Z^2} - {\overline{Z}}^2} \equiv
265: \frac{\Delta \overline{Z}}{\sigma_2} \label{newp}
266: \end{equation}
267: where $\sigma_2$ is the variance, or second central moment, which is related to
268: the ionization distribution width. Remarkably, this equation 
269: links the average difference in the ratios of effective 
270: atomic rates to the most general plasma parameters, namely, mean ion charges
271:  and variances.
272: 
273: However, if two compared codes have the same $\overline{Z}$ but different
274: variances, the parameter $p$ defined by Eq. (\ref{newp}) is zero. This simply
275: means that the approximation (\ref{dQ}) is insensitive to differences in
276: ionization distribution widths. To take this dependence into account, one can
277: add the next term in the expansion with respect to $(\zeta - \overline{Z})$:
278: 
279: \begin{equation}
280: \frac{\Delta Q_{\zeta }}{Q_{\zeta }}=p+2k(\zeta -\overline{Z}). \label{dQ1}
281: \end{equation}
282: Substituting (\ref{dQ1}) into Eqs. (\ref{al1}) and (\ref{dn1}), one obtains:
283: 
284: \begin{equation}
285: \frac{\Delta N_{\zeta }}{N_{\zeta }} = p(\zeta - \overline{Z}) + k
286: \left[(\zeta - \overline{Z})^2 + (\zeta - \overline{Z}) - \sigma_2\right].  \label{dnn}
287: \end{equation}
288: To derive equations relating two parameters, $p$ and $k$, we multiply Eq.
289: (\ref{dnn}) by $(\zeta -\overline{Z}) N_{\zeta}$ and $(\zeta -\overline{Z})^2
290: N_{\zeta}$ and then sum both sides over ion charges. The two ensuing equations are
291: sufficient to
292: obtain the following expressions:
293: 
294: \begin{eqnarray}
295: p &=& \frac{\Delta \overline{Z} \left(\sigma_4 + \sigma_3 - \sigma_2^3\right) 
296: - \Delta \sigma_2 \left(\sigma_2 + \sigma_3\right)}{\sigma_4 \sigma_2 - \sigma_3^2 -
297: \sigma_2^3},  \nonumber \\
298: k &=& \frac{\Delta \sigma_2 \cdot \sigma_2 - \Delta \overline{Z} \cdot \sigma_3}
299: {\sigma_4 \sigma_2 - \sigma_3^2 - \sigma_2^3}, \label{newkp}
300: \end{eqnarray}
301: where 
302: $\sigma_i = \sum_{\zeta =0}^{Z_{max}} \left(\zeta - \overline{Z}\right)^i N_{\zeta}
303: $
304: is the $i$th central moment.
305: 
306: Equation (\ref{newkp}) is the main result of the present paper. We propose to
307: use the parameters $p$ and $k$ for characterization of differences between
308: plasma kinetic codes. Using Eq. (\ref{newkp}), which depends only on the most
309: general kinetic parameters, namely, mean ion charges and central momenta, one
310: can directly evaluate the average differences between the effective rates
311: implemented in various kinetic codes.
312: 
313: \section{\label{sec3}Comparison of kinetic codes}
314: 
315: The above described method is applied here to the computational results from the
316: 3rd NLTE Workshop \cite{NLTE3} that can be accessed in the NIST SAHA database
317: \cite{SAHA}. This database contains various parameters, including mean ion
318: charges, central momenta, and ion populations, which may be used to determine
319: the quantities $p$ and $k$ defined in Eqs. (\ref{newp}) and (\ref{newkp}). The
320: SAHA database also provides other valuable parameters, such as effective rates
321: and partition functions, so that a user can obtain a deep insight into
322: differences between kinetic codes. In accordance with the policy accepted by the
323: Workshop participants, the results will be presented without direct attribution,
324: although a list of participating codes will be given for each case. Note also
325: that not all codes provide a complete set of central momenta up to $\sigma_4$.
326: 
327: The first step of the comparison procedure consists in selection of a reference
328: against which the other codes are to be compared. In the following
329: comparisons the reference code is chosen arbitrarily, as generally there are no
330: {\it a priori} physical grounds to prefer a particular code. Obviously, the
331: average values of $p$ and $k$ would change when selecting another code as a
332: reference. However, the standard deviations $\sigma$ of the corresponding
333: distributions of $p$ and $k$ that reflect the average spread within a group of
334: codes should not change and therefore are reported below as well. In what
335: follows, the parameters $p$ and $k$ determined from Eqs. (\ref{newp}) and
336: (\ref{newkp}) are referred to as ``calculated", while those determined by
337: fitting the $\Delta Q/Q$ ratios (Eqs. \ref{dQ} and \ref{dQ1}) are referred to as
338: ``fit". As the fitting procedure has to include only the physically
339: significant cases, ion states with $N_\zeta < 10^{-4}$ were excluded from the
340: comparison. Also, in order to emphasize the contribution from the most populated
341: states, the fitting was performed with the weights $g_w = \sqrt{N_1 N_2}$, where
342: $N_1$ and $N_2$ are the ion populations of the reference and compared codes.
343: 
344: Among numerous cases available in the SAHA database we selected three cases for
345: germanium, carbon, and gold. These elements cover a wide range of ion
346: charges and their ions represent atomic systems with different level of
347: complexity.
348: 
349: 
350: \subsection{Ge}
351: 
352: For germanium, we selected a relatively simple case of T$_e$ = 600 eV and N$_e$
353: = $10^{17}$ cm$^{-3}$, where almost all codes have a mean ion charge $\overline
354: Z \approx 22$ corresponding to a closed-shell Ne-like Ge. This case will be
355: discussed in more detail than the C and Au cases.
356: 
357: Table \ref{TbGe} presents calculated and fit (superscript ``$f$") values of
358: parameters $p$ and $k$. The values of $p$ determined from the single-parameter
359: formula (\ref{newp}) or fit using Eq. (\ref{dQ}) have subscript ``0". One can
360: immediately notice a generally good agreement between the calculated and fit
361: values of $p$ for all but one code. Table  \ref{TbGe} clearly demonstrates that
362: code 7 is an outlier, which is also emphasized by its very different value of
363: the mean ion charge $\overline Z \approx 27$. Moreover, the two-parameter fit
364: was not performed for code 7, as for only one ion stage both code 7 and the
365: reference code have populations larger than $10^{-4}$.
366: 
367: \begin{table}
368: \caption{Calculated and fit parameters $p$ and $k$ for the Ge case of T$_e$ = 600
369: eV and N$_e$ = $10^{17}$ cm$^{-3}$. Superscript $f$ denotes fit values.
370: Subscript 0 denotes $p$'s determined from the one-parameter formulas (\ref{dQ})
371: and (\ref{newp}). Code 7 was excluded in determination of standard deviation
372: $\sigma$.}
373: %Code 4 is the reference code.
374: \label{TbGe}
375: \begin{center}
376: \begin{tabular}{c|cc|cccc}
377: \hline 
378: Code No. & $p_0$ & $p_0^f$ & $p$ & $k$ & $p^f$ &$k^f$ \\
379: \hline 
380: 1 & -0.566 & -0.708 & -0.509 & -0.249 & -0.510 & -0.308 \\
381: 2 & -0.400 & -0.247 & -0.379 & -0.044 & -0.309 &  0.124 \\
382: 3 & -0.507 & -0.417 & -0.571 &  0.130 & -0.594 &  0.288 \\
383: 4 &  0.223 &  0.222 &  0.159 &  0.150 &  0.140 &  0.256 \\
384: 5 & -0.203 & -0.140 & -0.210 &  0.033 & -0.206 &  0.113 \\
385: 6 & -0.764 & -0.942 & -0.760 & -0.130 & -0.760 & -0.260 \\
386: 7 & -4.266 & -1.298 & -3.800 & -1.389 &  ----- &  ----- \\
387: 8 & -0.592 & -0.122 & -0.334 & -0.175 & -0.018 &  0.413 \\
388: \hline 
389: $\sigma$ & 0.301 & 0.362 & 0.272 & 0.142 & 0.298 & 0.255 \\
390: \hline 
391: \end{tabular}
392: \end{center}
393: \end{table}
394: 
395: Another interesting feature is a very small difference between the ``simple" $p$
396: of Eq. (\ref{dQ}) and calculated and fit values of $p$ determined from the
397: two-parameter formulas. While agreement between differently calculated
398: parameters $p$ is generally very good, the calculated and fit values of $k$
399: show worse level of correspondence. This is not surprising since $k$ is
400: a high-order parameter which may be more sensitive to small variations in
401: data.
402: 
403: As already mentioned, a standard deviation $\sigma$ would unambiguously
404: represent the spread of parameters $p$ and $k$ within a particular group of
405: codes. The last row in Table \ref{TbGe} show $\sigma$'s calculated for each
406: column (the outlier code 7 was not included in the determination of $\sigma$).
407: Remarkably, $\sigma$'s for calculated $p_0$ and two-parameter calculated and fit
408: $p$'s agree within only 6 \%, and even $\sigma(p_0^f)$ deviates by less than 25
409: \%. The value of $\sigma \approx $ 0.3 means that in this group of codes the
410: average deviation of effective ionization and recombination rates is about 30 \%.
411: 
412: Finally, consider the dependence of $\sigma$ on electron temperature $T_e$. The
413: SAHA database contains data for $T_e$ = 150 eV, 250 eV, 450 eV, and 600 eV at
414: $n_e$ = $10^{17}$ cm$^{-3}$. The calculated $\sigma$'s for $p_0$, $p$, and $k$
415: are presented in Fig. \ref{fig2} for the four temperatures. At low $T_e$, the
416: mostly populated ions are those with open shells, and since these cases are most
417: difficult to calculate, the difference between codes is the largest. With the
418: increase of electron temperature, the ionization stage approaches the
419: closed-shell Ne-like ion, and therefore agreement improves dramatically. Note
420: also that $\sigma_k$ is smaller than $\sigma_p$, and therefore in many cases a
421: simple one-parameter formula for $p$ would be sufficient for estimates of the
422: difference between codes.
423: 
424: 
425: %\begin{figure}[tbp]
426: %\centerline{\includegraphics[width=0.8\linewidth]{fig2.eps}}
427: %\caption{\\ Standard deviation $\sigma$ for calculated parameters $p_0$ (circles),
428: %$p$ (squares), and
429: %$k$ (triangles) as a function of electron temperature for the germanium case at
430: %electron density N$_e$ = 10$^{17}$ cm$^{-3}$.}
431: %\label{fig2}
432: %\end{figure}
433: 
434: \begin{figure*}[t]
435: %\begin{minipage}{15cm}
436: \epsfig{file=fig2.eps,width=12cm}
437: %\end{minipage}
438: \vspace*{1cm}
439: \caption{Standard deviation $\sigma$ for calculated parameters $p_0$ 
440: (circles), $p$ (squares), and $k$ (triangles) as a function of electron 
441: temperature for the germanium case at electron density 
442: N$_e$ = 10$^{17}$ cm$^{-3}$.}
443: \label{fig2}
444: \end{figure*}
445: 
446: 
447: \subsection{C}
448: 
449: Carbon cases in the SAHA database were calculated at a single electron density
450: of N$_e$ = $10^{22}$ cm$^{-3}$. At T$_e$ = 20 eV, which is the selected case
451: here, the mean ion charge varies between 1.8 to 3.3 for different codes. There
452: are no obvious outliers in this case, and therefore two-parameter fits were
453: successful for all codes (Table \ref{TbC}). Similar to the Ge case, the
454: correlation between calculated and fit values is generally good, although
455: one-parameter $p$'s for codes 8 and 9 show larger discrepancies. The standard
456: deviations for the different parameters $p$ in Table \ref{TbC} agree even better
457: than for the Ge case, however, they are about a factor of 2 larger and reach 60
458: \%. This would seem to be unexpected as carbon simulations could include only
459: seven ionization stages at most and are thus supposed to be simpler. The reason
460: for the larger $\sigma$'s is that the effects of ionization potential lowering
461: are much more important here due to the significantly higher density, and
462: different treatments of the continuum lowering noticeably contribute to the
463: increased spread of the $p$ and, to a lesser extent, $k$ values.
464: 
465: 
466: \begin{table}
467: \caption{Calculated and fit parameters $p$ and $k$ for the C case of T$_e$ = 20
468: eV and N$_e$ = $10^{22}$ cm$^{-3}$.  Superscript $f$ denotes fit values.
469: Subscript 0 denotes $p$'s determined from the one-parameter formulas (\ref{dQ})
470: and (\ref{newp}).}
471: %Code 4 is the reference code.
472: \label{TbC} 
473: \begin{center}
474: \begin{tabular}{c|cc|cccc}
475: \hline 
476: Code No. & $p_0$ & $p_0^f$ & $p$ & $k$ & $p^f$ &$k^f$ \\
477: \hline 
478: 1 & -0.795 & -1.003 & -1.182 &  0.411 & -1.440 &  0.693 \\
479: 2 &  1.043 &  0.537 &  0.755 &  0.515 &  0.635 &  0.605 \\
480: 3 &  0.629 &  0.434 &  0.443 &  0.295 &  0.435 &  0.361 \\
481: 4 & -1.015 & -0.991 & -1.245 &  0.154 & -1.046 &  0.135 \\
482: 5 & -0.770 & -0.899 & -1.064 &  0.291 & -1.001 &  0.305 \\
483: 6 & -0.345 & -0.337 & -0.393 &  0.042 & -0.340 &  0.009 \\
484: 7 & -0.209 & -0.395 & -0.467 &  0.331 & -0.471 &  0.374 \\
485: 8 &  0.020 & -0.158 & -0.221 &  0.322 & -0.218 &  0.403 \\
486: 9 &  0.257 & -0.374 & -0.132 &  0.547 & -0.362 &  0.851 \\
487: \hline 
488: $\sigma$ & 0.651 & 0.535 & 0.660 & 0.151 & 0.638 & 0.250 \\
489: \hline 
490: \end{tabular}
491: \end{center}
492: \end{table}
493: 
494: \subsection{Au}
495: 
496: The gold cases available in the SAHA database show very significant differences,
497: and for the selected case of N$_e$ = $10^{21}$ cm$^{-3}$ and T$_e$ = 750 eV, the
498: mean ion charge varies between 31 (code 7) and 44 (code 4). Moreover, ionization
499: distributions for codes 1 and 2 show double peak structure unlike other,
500: smoother bell-like distributions. It is therefore not surprising that the
501: absolute values of $p$, which are the main indicators of code disagreements, are
502: much larger than for the Ge and C cases discussed above. As a consequence, the
503: standard deviations for the differently determined $p$ and $k$ values do not
504: show the same level of agreement as previously. This situation simply reflects 
505: very significant differences between code results submitted to the 3rd NLTE Code
506: Comparison Workshop.
507: 
508: 
509: \begin{table}
510: \caption{Calculated and fit parameters $p$ and $k$ for the Au case of T$_e$ = 750
511: eV and N$_e$ = $10^{21}$ cm$^{-3}$. Superscript $f$ denotes fit values.
512: Subscript 0 denotes $p$'s determined from the one-parameter formulas (\ref{dQ})
513: and (\ref{newp}).}
514: \label{Tb Au}
515: \begin{center}
516: \begin{tabular}{c|cc|cccc}
517: \hline 
518: Code No. & $p_0$ & $p_0^f$ & $p$ & $k$ & $p^f$ &$k^f$ \\
519: \hline 
520: 1 &  2.293 &  1.376 &  1.118 & -0.454 &  1.297 &  0.042 \\
521: 2 & -0.534 & -0.410 &  0.418 & -0.085 & -0.434 & -0.025 \\
522: 3 & -1.151 & -1.061 & -0.393 &  0.050 & -1.067 &  0.050 \\
523: 4 & -1.338 & -1.178 & -0.497 &  0.061 & -1.189 &  0.052 \\
524: 5 &  1.638 &  1.387 &  0.384 & -0.101 &  1.419 & -0.039 \\
525: 6 & -1.086 & -1.042 & -0.422 &  0.070 & -1.052 &  0.069 \\
526: 7 &  3.393 & -0.970 &  0.678 & -0.129 & -0.979 & -0.056 \\
527: 8 & -1.162 & -0.302 & -0.302 & -0.028 & -0.218 &  0.403 \\
528: \hline 
529: $\sigma$ & 1.763 & 1.001 & 0.568 & 0.160 & 0.996 & 0.136 \\
530: \hline 
531: \end{tabular}
532: \end{center}
533: \end{table}
534: 
535: \section{\label{sec4}Conclusions}
536: 
537: As the complexity of plasma kinetic codes rapidly increases, their verification
538: and validation is becoming mandatory for establishing credibility of
539: computational results. To this end, a development of new techniques for code
540: comparisons is an urgent and important task. In the present paper we introduced
541: two new parameters for the characterization of discrepancies between plasma
542: kinetic codes. These parameters describe differences between effective
543: ionization and recombination rates used in the codes. Importantly, the only
544: physical quantities required for their calculation are the mean ion charges and
545: central momenta that are the most widely reported characteristics of plasma
546: kinetic calculations. Since the final formulas include only the simplest
547: algebra, this method provides very fast estimates of code differences in the
548: input atomic rates. The new parametrization was applied to the data from the 3rd
549: NLTE Code Comparison Workshop and the presented results clearly prove simplicity
550: and reliability of the method used. We plan to implement this method to the
551: analysis of the data from future NLTE workshops.
552: 
553: \section*{Acknowledgments}
554: 
555: This work was carried out while one of us (D.S.) spent a two-month working visit
556: at the Atomic Physics Division of the National Institute of Standards and
557: Technology. He would like to express his thanks for the cooperative atmosphere
558: and generous hospitality. Authors would also like to thank H. R. Griem and R. W.
559: Lee for reading the manuscript and valuable comments. This research was
560: supported in part by the Office of Fusion Energy Sciences of the US Department
561: of Energy.
562: 
563: 
564: \begin{thebibliography}{99}
565: 
566: \bibitem{book} D. Salzmann, \textit{Atomic Physics in Hot Plasmas}, Oxford
567: University Press, 1998.
568: 
569: \bibitem{Griem} H. R. Griem, \textit{Principles of Plasma Spectroscopy},
570: Cambridge University Press, 1997.
571: 
572: \bibitem{NLTE1} R. W. Lee, J. K. Nash, and Yu. Ralchenko, J. Quant. Spectrosc.
573: Radiat. Transfer \textbf{58}, 737 (1997).
574: 
575: \bibitem{NLTE2} C. Bowen, A. Decoster, C. J. Fontes, K. B. Fournier, O. Peyrusse
576: and Yu. V. Ralchenko, J. Quant. Spectr. Rad. Transfer 81, 71 (2003).
577: 
578: \bibitem{NLTE3} Yu. Ralchenko, R. W. Lee, and C. Bowen, in: 14th APS Topical
579: Conference on Atomic Processes in Plasmas, AIP Conference Proceedings
580: \textbf{730}, 151 (2004); C. Bowen, R. W. Lee, and Yu. Ralchenko, J. Quant.
581: Spectr. Rad. Transfer \textbf{99}, 102 (2006).
582: 
583: \bibitem{SAHA} Yu. Ralchenko (2006). NIST SAHA Plasma Kinetics Database
584: (version 1.0 beta), [Online]. Available: http://nlte.nist.gov/SAHA [2006, March
585: 12]. National Institute of Standards and Technology, Gaithersburg, MD.
586: 
587: \bibitem{PRA} D. Salzmann, Phys. Rev. A\textbf{22}, 2245 (1980).
588: 
589: \end{thebibliography}
590: 
591: \end{document}
592: