«Received ; accepted –2– ABSTRACT We describe the latest release of AtomDB (version 2.0), a database of atomic data and a plasma modeling code ...»
AtomDB 2.0: Updated Atomic Data for X-ray Astrophysics
A. R. Foster, L. Ji, R. K. Smith and N. S. Brickhouse
Smithsonian Astrophysical Observatory, Cambridge, MA 02144, USA
Received ; accepted
We describe the latest release of AtomDB (version 2.0), a database of atomic
data and a plasma modeling code with a focus on X-ray astronomy. This release includes several major updates to the fundamental atomic structure and
transition data held within AtomDB, incorporating new ionization balance data, state-selective recombination data, and updated collisional excitation data for many ions, including the iron L-shell ions from Fe+16 to Fe+23. We also describe some of the eﬀects that these changes have on calculated emission and diagnostic line ratios, such as changes in the temperature-dependent G-ratio for He-like ions of up to 33%. Finally, we compare observations of Capella using the new and the old datasets to show ???FIXME???.
1. Introduction Analysis of astrophysical spectra recorded by instruments in X-ray and other wavelength bands is a powerful tool for understanding the composition and plasma conditions in a wide variety of astrophysical phenomena. Successful analysis of this data requires both a model of the plasma and a large selection of data detailing the various atomic processes occuring in the plasma.
The recognition of the importance of the dielectronic recombination (DR) process for plasmas, even at high temperatures, (Burgess 1964) led to huge strides in modeling collisionally ionized, optically thin astrophysical plasmas.: Cox & Tucker (1969) collated some of the earliest data. Since then there have been a steady stream of reﬁnements to collisional models, usually incorporating improvements to either or both of the physical model (including more processes, for example), and the underlying data (using more –3– accurate or more complete data sets). Diﬀerent evolutions of such models have included those of Cox & Daltabuit (1971), Mewew (1972), Landini & Fossi (1972), Raymond & Smith (1977), and Brickhouse et al. (1995).
The quantity of relevant atomic data continues to grow as the computational power available for performing calculations improves. In modern analysis of X-ray astrophysical spectra there are three widely used atomic databases: XSPEC version 2.0.1 (Arnaud 1996), CHIANTI version 6.0.1 (Dere et al. 2009) and AtomDB version 1.3 (Smith & Brickhouse 2001). Each database has a slightly diﬀerent focus: CHIANTI’s main focuse is on the EUV wavelengths for analyzing solar spectra, while XSPEC and AtomDB focus on the X-ray ranges. These databases continueto undergo periodic review, and are updated as newer data become available; this paper introduces the release of version 2.0 of AtomDB, describing the new data and improvements, how they have been implemented and what is planned for future releases.
Changes have been made to large areas of the AtomDB database. Several ions which were not previously included are now in the database. Data for many other ions have been improved by including more recent calculation data, by increasing the number of energy levels and transitions included for the ion, or a combination of the both. In addition, major changes have been made to the ionization and, particularly, recombination data, with some signiﬁcant changes to the resulting ionization balance and therefore to line emissivities.
Here we outline each change in detail; in section 3 we will summarize the eﬀect of these changes on the spectra observed by users. We note that in this release there are no changes to the format of the data ﬁles which the database uses and produces, which were described in Smith et al. (2001).
2.1. Ionization and Recombination Rates
While calculation of the ionization balance is just one part of calculating an emission spectrum for a plasma, it has a very strong eﬀect on the result. As experimental and theoretical methods improve, there are periodical eﬀorts to collect the rates for ionization and recombination, both radiative (RR) and dielectronic (DR), for easy use by astrophysicists. Previous examples include the widely used data sets of Arnaud & Raymond (1992) and Mazzotta et al. (1998): the latter was included in AtomDB 1.3. Bryans et al. (2006, 2009) have produced a new compilation of ionization balance data for elements from hydrogen (Z0 = 1) to zinc (Z0 =30), and this has been included in AtomDB v2.0.
This dataset provides new rates for DR and RR for all astrophysically relevant ions from fully stripped up to Na-like systems (where Na-like refers to the recombining ion): for the remaining ions, the data of Mazzotta et al. (1998) is retained. In addition, the Bryans et al. (2009) data includes collisional ionization from Dere (2007) for all ions.
In their papers Bryans et al. (2006, 2009) compare the ionization and recombination data with the earlier data sets and with experimental measurements. While the electron impact ionization (EII) and RR rates have not changed dramatically, for many L-shell ions there have been signiﬁcant changes in the DR rates. The DR and RR calculations now generally agree to within 35% of available measurements. For EII there is disagreement for some heavier, lightly ionized ions (Mg-like and above) between the results of Dere (2007) and Mattioli (2007), the former has been included here as incorporated in the Bryans data set: hopefully further study will provide further enlightenment as to the correct values.
The Badnell (2006) data sets which Bryans used for the DR and RR rates are part of a compilation which is undergoes periodic update and review. Since their publication in 2006, the rates for DR into nitrogen-like ions have been revised; we have included this revised data here. For all other ions the DR & RR rates are the same as those of Bryans et –5– al. (2009).
The DR and RR data in the Bryans et al. (2006) collection give total rates for recombination from the ground state of one ion to the next. The Badnell (2006) calculations on which these are based are resolved into state-selective recombination rates, so that the recombination rate into each excited state of the ion is speciﬁed. This allows inclusion of these direct rates when calculating the excited-state populations of the recombined ion, and as a result will aﬀect emission lines, particularly in regions of strong DR (see examples in Section 3.5). For this reason we have included state-selective data for recombination to all H-like, He-like and Li-like ions, and all iron ions from Fe+16 to Fe+22.
The DR & RR source data are split into level-resolved rates for capture into lower n (typically n ≤ 8), with capture rates into higher n provided as totals into each n-shell.
Since the levels identiﬁed in the DR & RR calculations and those in AtomDB are rarely identical, some level matching and cascade calculations have been required.
This has been achieved with a several step process. Using the Autostructure (Badnell 1986) code, transition probabilities have been calculated for all the energy levels which appear in either the AtomDB datasets or the DR/RR calculations. In addition, transition probabilities between n shells have been estimated using the hydrogenic approximation of Burgess & Summers (1976). Using this data, the capture into the high-n levels has been cascaded down, assuming zero electron density and thus no collisional re-ionization/re-distribution eﬀects, to the lower n levels for which resolved data exists.
The cascade contribution is then distributed according to the statistical weight of each level within the n shell.
–6– The Einstein A-values from Autostructure are then used to further cascade the data as required from these levels until levels which are in the AtomDB database for collisional excitation are encoutered. Thus a total eﬀective recombination rate to each level in the AtomDB database is constructed from a direct rate and cascade from higher n levels.
This treatment of the cascade process does cause certain omissions which we hope to address in future: ﬁrstly, the photons emitted during the cascade are not tracked and therefore are not included in any spectral models or radiated power estimates. In addition, the treatment of DR satellite lines in not consistent with this information: those lines are tabulated as a function of temperature obtained from separate calculations: these would be better handled in a self consistent manner during the original calculation of the DR rates.
Future work will aim to address these problems.
The data for hydrogenic ions have been upgraded in two major ways: (1) extended to include higher n-shells (from nupper = 5 to nupper = 10), and, (2) where available, data of a higher quality have been used for ions of astrophysical interest.
Due to the simplicity of the hydrogenic system, it is often used as a training case when learning R-Matrix techniques or other sophisticated collional techniques. However, due to the assumed simplicity of the system, it is rare for the data to be actually recorded.
The ions for which modern H-like R-Matrix collision strengths could be obtained are shown in Table 1. The atomic structure and A-values for the higher-n data were obtained using Autostructure. The electron impact excitation data for these higher n was calculated using the Flexible Atomic Code (FAC, Gu (2003)). R-Matrix calculations exist for collisional excitation for transitions with nupper ≤ 5 for several ions. These calculations –7– have been performed in LS coupling, not LSJ for all such ions except Fe+25 (see Table 1). LS coupling is suﬃcient for most of the lighter ions as the relativistic splitting between J-resolved levels is not generally observable: an energy resolution E/∆E 2000@1 keV is required to observe the 2p2 P[1/2,3/2] → 1s1 S splitting of Neon; for lighter elements it is higher.
The possible advent of higher resolution detectors in the future means that AtomDB is targeted at LSJ resolved data throughout: all other data in the database is already LSJ resolved. We therefore have used a combination of Autostructurefor exnergy levels and A-values, with collisional excitation rates from FAC, to create the data for AtomDB. The exception here is for excitation from ns levels, for which the R-matrix collision strength has been used, split by the statistical weight of the ﬁnal state.
For completeness, for all ions not mentioned in Table 1, data were produced using Autostructure for energy levels and transition probabilities, and using FAC for the collision strengths for all energy levels with nupper ≤ 10.
For the He-like ions, data have been incorporated from new Intermediate Coupling Frame Transformation (ICFT) R-Matrix (Griﬃn et al. 1998) calculations provided by Whiteford (2005) for all elements from C5+ to Kr35+. In this work, structure data is again generated by Autostructure, with the ICFT method being used for the collision strengths. As with the H-like ions we have extended the range of these calculations using FAC to obtain collision strengths for collisions with 5 nupper ≤ 10.
2.5. Iron L-shell Data
For the iron L-shell ions (Fe+16 to Fe+23 ) signiﬁcant advances have been made in collisional excitation calculations since the release of AtomDB 1.3. The collision strength data for these ions have been upgraded to R-matrix calculations. Details of which particular calculation were used for each ion are given in Table 2. These new calculations include, in general, fewer conﬁgurations than the existing Distorted Wave data (Liedahl 1997) in AtomDB. As a result the newer data have been merged with the existing data, updating or adding as appropriate. Much of the older data remains for transitions involving some levels not included in the newer data.
2.6. Non-X-ray focused Data
For many other ions, especially those that are not notably strong emitters in the X-ray region, AtomDB version 1.3 used data from CHIANTI version 2.0 (Landi et al. 1999). We have updated the wavelengths, transition probabilities and collision strengths of all data for all ions excluding the H-like & He-like sequences, and the Ni and Fe L-shell ions to now incorporate the CHIANTI version 6.0.1 data (Dere et al. 2009).
2.7. Minor Changes
In addition to these major updates to the database, several minor corrections have been made which have little or no eﬀect on the resulting spectra from AtomDB but which do aﬀect the database. This includes the correct use of L and S quantum numbers for all levels where LS or LSJ coupling is used; consistent formatting of conﬁguration strings, minor corrections to the H-like two-photon transition rate from the 2s1 → 1s1 level, including adding the rate for hydrogen and reducing it for all other hydrogenic ions.
The eﬀect of the new ionization data is most signiﬁcant for heavier elements. Detailed information can be found in the papers of Bryans et al. (2006, 2009). In Figure 1 we show the change in the ionization balance for iron, including our use of diﬀerent recombination rates for Fe XIX as speciﬁed in Section 2.1. Signiﬁcant changes can be seen around Fe XVII, in some cases leading to changes in fractional abundance of upwards if 30%: this has in turn signiﬁcant eﬀects on emission lines present at these temperatures.
3.2. Total Radiated Power
Figure 2 shows the total power radiated from all elements given a solar elemental abundance (Anders & Grevesse 1989) and an electron density of 1cm−3. The top ﬁgure shows a comparison between the total emission from the old and new versions of the database: as can be seen, there are signiﬁcant ( 10%) increases in radiation in four diﬀerent temperature ranges. The increase at 2 × 104 K is due to the inclusion of two-photon emission from hydrogen for the ﬁrst time (this falls outside the traditional X-ray band and so was previously omitted). The remaining three increases are due to changes in the ionization balance for oxygen, neon and iron respectively: the new data have a higher fraction of less ionized ions present, which radiate more strongly in these hotter regions.