aa r X i v : . [ a s t r o - ph ] M a r Atomic Processes in Planetary Nebulae and H ii Regions
Manuel A. Bautista
Centro de F´ısica, IVIC, P.O. Box 21827 Caracas 1010A, Venezuela E-mail: [email protected]
Abstract.
Spectroscopic studies of Planetary Nebulae (PNe) and H ii regions have drivenmuch development in atomic physics. In the last few years the combination of a generation ofpowerful observatories, the development of ever more sophisticated spectral modeling codes, andlarge efforts on mass production of high quality atomic data have led to important progress inour understanding of the atomic spectra of such astronomical objects. In this paper I review suchprogress, including evaluations of atomic data by comparisons with nebular spectra, detectionof spectral lines from most iron-peak elements and n-capture elements, observations of hyperfineemission lines and analysis of isotopic abundances, fluorescent processes, and new techniques fordiagnosing physical conditions based on recombination spectra. The review is directed towardatomic physicists and spectroscopists trying to establish the current status of the atomic dataand models and to know the main standing issues.
1. Introduction
Nearly all modern observational research in gaseous nebulae (PNe, H ii regions, and circumstellarnebulae) involves some kind of spectra, whose interpretation requires some understanding of theatomic processes involved and the handling of atomic data. The level of understanding ofsuch processes and the data required for various applications ranges from phenomenologicaland statistical studies of commonly prominent lines, to compilation of raw atomic data fordirect inspection (such as line finding lists), and to modeling of synthetic spectra attemptingto fit the observed spectra. On the other hand, the accuracy and quantity of atomic modelsdescribing various features in observed spectra have evolved with time, driven by great advancesin ground- and space-based observatories, computer technology and experimental techniques. Atpresent there are models and data accurate enough for satisfactory analysis of at least the mostprominent spectral features. Also the advent of on-line databases and spectroscopic tools hasrevolutionized the dissemination of results of atomic physics research. However, the demandson the quantity and quality of the atomic data will grow as new instruments delivering greatersensitivity and spectral resolution become available.A review of atomic processes spans an audience that includes specialists in the study ofatomic systems with tools capable of providing models and atomic data, nebular astrophysicists,astronomers seeking to understand the reliability and accuracy of their modeled spectra, andthose who use the models and data to compare synthetic spectra with observations, diagnosethe physical conditions and compute chemical abundances of nebulae. The present review is Currently at Department of Physics, Virginia Polytechnic Institute and State University, VA 24061, USA imed at researchers in atomic physics and spectroscopy who wish to understand the reach ofcurrently available atomic models and identify present needs.The physical conditions in gaseous nebulae of interest here are those of photoionized plasmasas observed in the ultraviolet (UV), optical, and infrared (IR) bands. Roughly speaking theseconditions are: temperatures of the order of 10 K and electron densities between 10 and10 cm − . This review will not treat the X-ray band which in gaseous nebulae results frommechanically heated coronal gas. A good review on this subject can be found in [1].The present review is organized according the nuclear charge of atomic species, starting withhydrogen and helium, and following with the second and third row elements, the iron-peak ions,and then onto the heavier elements. I then finish the paper with a discussion of various generalissues of interest.
2. Hydrogen
The spectrum of hydrogen in H ii regions has been studied extensively for quite some time.In principle the spectrum is determined by the recombination rates into each level and thesubsequent radiative cascades. Such a process is accurately modeled in two idealistic situations,the so called “case A” in which all lines are optically thin and “case B” which assumes that theoptical depth of Ly α goes to infinity. Two complications arise beyond these approximations:(1) a detailed solution to the radiative transfer of Ly α photons including the effects of self-absorption, removal and collisionally induced transitions between the 2 s and 2 p states, and (2)when the electron temperature and density of the plasma are high enough for effective collisionalexcitations on to n > s / − p / and 2 s / − p / lines at 1.1 GHz and 9.9 GHz respectively. Dennison et al. explain that removal of Ly α photonsby dust would limit the pumping of H atoms into the 2 p states, and under these conditions the2 s / − p / transition will appear in stimulated emission while the 2 s / − p / line will appearin absorption. In general, the relative strengths of these two lines should serve as diagnostics ofthe populations of the 2 s and 2 p states and the responsible processes. Further, Dennison et al.suggest that removal by dust is the dominant mechanism acting on Ly α photons in H ii regions,which if correct would circumvent the need to solve the radiative transfer problem.In regards to collisional excitation of hydrogen, P´equignot & Tsamis [3] study the evolutionof calculated collision strengths for 1 s → n = 3 , ,
5. They point out large variations in thetheoretical values up to the last calculation of Anderson et al. [4]. P´equignot & Tsamis comparetheir observations for the PN G135.9+55.9 with predictions of models using various collisionaldata sets. They find that the collision strengths of Anderson et al. yield the most consistentresults, yet these may have not reached ultimate accuracy.Modern high-signal-to-noise optical spectra allow for accurate measurements of lines andcontinuum around the Balmer and Paschen recombination series. By fitting the results ofdetailed spectral models to these measurements it is possible to determine the electron densityand temperature in the nebula [5]. Surprisingly, temperatures obtained from the Balmer seriesoften disagree with the temperatures derived from line ratios of collisionally excited lines, suchas the ratio of nebular to auroral lines of [O iii ]. Moreover, for a large sample of PNe studied byZhang et al., the hydrogen Balmer temperatures are typically lower than the collisional oxygentemperature by several thousand degrees. These temperature differences are used in supportof the so called ’temperature fluctuations’ or ’temperature variations’ in nebulae. However, aflag of warning on the use of hydrogen temperatures comes from the fact that in four PNewhere temperatures from the Paschen series were determined, these disagree with the Balmertemperatures at the three sigma level or more. . Helium
Modeling the He i recombination spectrum has become increasingly reliable, with an accuracy online emissivities of the order of a few percent (see [6, 7, 8]). Yet, this is not at the 1% accuracylevel needed to place useful constrains on the primordial helium abundance in cosmologicalstudies [9, 10]. Recently, Bauman et al. [6] worked out the recombination problem in finestructure in the low density limit. They found no significant effects on the recombinationspectrum from spin-orbit coupling. However, they found that some of lines may be uncertaindue to use of poor quality atomic data, specifically the photoionization cross sections for stateswith 9 < n <
20 and
L < i lines came from Porteret al. [11] who compared model predictions with measured fluxes for 100 lines with n ≤
20 fromthe Orion nebula. They found an average difference of 6.5% between all observed and predictedlines and 3.8% difference for the 22 most accurately measured lines.Zhang et al. [12, 13] present a method to diagnose temperatures from He i line ratios.They write parametric forms for He i line ratios vs. temperature and conclude that the bestdiagnostic is obtained from the λ /λ ∼ i λ s S → p P o ) line. When comparing the results from the λ /λ λ /λ ∼
500 K) arises, which may be related tothe uncertainty on the optical depth of the λ β decrement by an average of 4000 K.
4. Second and third row elements (C-Ar)
Density and temperature diagnostics in PNe are commonly based on line ratios amongcollisionally excited lines. The best known density diagnostics are [S ii ] λ λ ii ] λ /λ cm − , [Cl iii ] λ /λ N e ∼ cm − , and [Ar iv ] λ /λ N e ∼ − cm − .The [O ii ] ratio that arises from transitions among the S / ground level and the D / and D / levels was a subject of controversy for the last few years. It was commonly assumedthat LS -coupling was a valid approximation for terms of the ground configuration of the O + system; thus the ratio of collision strengths from the ground levels to the D / and D / levelswas given by the statistical weights, i.e. 1.5. However, McLaughlin & Bell [14] claimed thatrelativistic effects enhanced the ratio of collision strengths to 1.93, with profound effects on N e determinations for low surface brightness H ii regions. Such claims were contested by evidencefrom the extensive literature survey of Copetti & Writzl [15] and the observational campaignof Wang et al. [16]. The issue seems settled now with a new calculation by Pradhan et al.[17], that accounts for all dominant relativistic effects and confirms the earlier predictions of the LS -coupling approximation. The spectroscopic survey by Wang et al. also served to test the A -values for dipole forbidden transitions of [O ii ]. They found that the transition probabilitiesof Zeippen [18] best fit the observations, the differences being of only a few percent, while theresults of later calculations by the same author ([19]) and by W. Wenaker [20] look problematic.This latter set is what is currently available through the NIST database ([21]).Wang et al. also compared the electron densities obtained from various line ratios andobtained very good agreement between N e ([O ii ]), N e ([S ii ]), and N e ([Cl iii ]). By contrast N e ([Ar iv ]) values yield densities systematically higher than those from the other diagnostics,which sheds some doubts on the accuracy of the N e ([Ar iv ]) A -values. . The iron-peak elements Fe and iron-peak elements are important constituents of gaseous nebulae. Depending on theexcitation of the nebula, iron is frequently seen in stages from Fe i to Fe vii . The observed ionsof nickel span about the same range. Other less abundant elements of the group are occasionallyidentified as well. An extreme example of rich spectra is the η Carinae ejecta (see article byGull in this volume).On this subject the contributions of Sveneric Johansson could hardly be overstated. He hasbeen the driver of extensive and detailed research of energy levels and transition rates for the lowionization stages of iron-peak species. Data often are taken for granted as they becomes easilyavailable through various databases, yet their determination from laboratory work requires veryskillful people and substantial funding. Johansson’s publications on the subject are too numerousto list here, but some of the most important contributions are: on the spectra of Sc ii [22], Fe i [23], Fe ii [24], Ti ii [25], Co ii [26], and on the determination of f -values, lifetimes, and radiativerates for forbidden transitions (e.g. for Fe ii [27, 28, 29, 30], Ti ii [31], Ni i [32]). Johansson hasalso done seminal work the understanding of the Ly α excitation mechanism of Fe ii and otheriron-peak species in astronomical spectra (e.g. [33, 34, 35, 36]).Spectra of singly ionized iron-peak species are particularly complex as they respond to avariety of excitation mechanisms. These are H Ly α fluorescence, continuum fluorescence, self-fluorescence by overlapping Fe ii transitions, and collisional transitions among low and highlevels. In moderately dense plasmas ( N e ∼ cm − ) the populations of high levels involved influorescence are redistributed by collisionally induced transitions through highly excited pseudo-metastable levels [47]. Furthermore, proper modeling of Fe ii spectra requires large systems,approaching one thousand levels and complete and accurate atomic data. This has led to adiverse set of models with little consensus among them and limited understanding about theiraccuracy. For ions of atoms other than Fe and Ni the first detailed spectral models are nowbeing created, e.g. [38, 39]. A good part of this work has been done under the auspices of theIRON Project [40]. A summary of the data is presented in Fig. 1.More work is also in progress by various groups like the IRON Project, the FERRUM Projectat Lund University, at NIST, and the group at Queen’s University Belfast.Another modeling challenge in dealing with dense spectra of iron-peak species is the treatmentof the transfer of line radiation from the point of emission to the boundary of the atmosphere. Foriron-peak species, in addition to continuum opacity and resonant scattering, one must considerfluorescence by line overlaps, such as the well known H Ly α effect in Fe ii [33]. While writingthe present review, I have estimated the number of overlaps between absorption lines arisingfrom low excitation levels, typically populated under nebular conditions, of Fe ii-iv and linesfrom H i , He ii , O ii and O iii , dominant species in typical nebular spectra. Line wavelengthsawere taken from the NIST database and do not represent the entire spectra. For line widthsof 20 km/s the number of overlaps are 592, 243, and 49 for Fe ii , iii , and iv respectively, whilefor widths or 60 km/s the overlaps are 1722, 762, 524 for the same ions. This is the range ofline widths in nebular spectra. The huge number of line overlaps demonstrates the need for adetailed treatment of radiative transfer of Fe spectra. The case also applies to other iron-peakspecies.In modeling spectra one also needs to solve for ionization equilibrium of each element. For thatreason, detailed level specific photoionization cross sections, that account for complex resonancestructures, are being computed for both ground and excited states using the R-matrix method.Cross sections are available for Fe i - iv [50, 51, 52, 53] and Ni ii [54]. Recently, much progress hasbeen made at the Instituto Venezolano de Investigaciones Cientificas (IVIC) on the calculationof cross sections for neutral Sc, Ti, and Cr. igure 1.
Spectral models for low ionization iron-peak species
6. Hyperfine-induced spectra
Observations of hyperfine-induced transitions in PNe have been of great interest in recent years.This is because these spectra allow us to determine relative isotopic compositions to be comparedwith predictions of nucleosynthesis. This, in turn, enables us to disentangle enrichment by stellarnucleosynthesis from the primordial composition in observed present total abundances. Forexample, the abundance of He in our Galaxy could be used to test big bang nucleosynthesisand constrain the baryonic density of the Universe. However, the evolution of He in theGalaxy is a longstanding open problem due to the observed discrepancy (by two orders ofmagnitude) between theoretical yields of low-mass stars and the measured abundances in H ii regions [55, 56]. A proposed solution of this problem could be in suppressing, by non-standardmixing mechanisms, the production of He during the red giant branch and/or asymptotic giantbranch phases of stars of mass up to ∼ M ⊙ [57, 58]. An observable consequence of such anscenario is that the ratio C/ C in the ejecta of PNe should be much lower than in the standardcase. For a 1 M ⊙ star, the predicted ratio is ∼
5, in contrast with the standard ratio of 25-30[59].Isotopic abundances can be derived from beryllium-like ions through observations of thehyperfine-induced transition 2s2p P − S within the UV0.01 multiplet. This transitioncan only occur through coupling of the electronic total momentum with nuclei of nonzero spin.This happens in the spectra of C (with nuclear spin 1/2) but not in C. Thus, C/ Cabundance ratios can be derived from spectra showing the hyperfine transitions together withthe stronger 2s2p P − S (M2) and 2s2p P − S (IC) transitions. Palla et al.[60] observed these lines in NGC 3242 using STIS on board the Hubble Space Telescope . Fromthis, they derived C/ C =38. The same lines from the isoelectronic ion N iv were measuredby Brage et al. [61] from a STIS spectrum of NGC 3918. In this case both stable isotopesf nitrogen ( N and N) have non-zero nuclear spin. Although no abundances were derived,this work served to confirm the theoretical A -value for the hyperfine transitions. More recently,Rubin et al. [62] analyzed spectra of PNe in the IUE archive and determined C/ C = 4 . ± . iv , was reportedby Casassus et al. [63]. The two stable isotopes, Al and Al, have nuclear spins 5/2 and 5respectively. Thus, hyperfine structure causes the P - P transition at 3.66 µ m to split into 9components, 5 of which could be resolved spectroscopically. From the observations, Casassus etal. calculated an upper limit to the Al/ Al abundance ratio of 1/33 in NGC 6302, which isin good agreement with the predicted value of 1/37 from stellar evolution models.
7. Neutron capture elements
One of the frontiers of spectroscopy of PNe is the study of n -capture elements presumablysynthesized by the AGB progenitor of the nebula. The progress on this area since the seminalwork of P´equignot & Baluteau [64] has been slow owing to the need for high spectral resolution( >
10 000) and the absence of atomic data and models. Despite the difficulties, Dinerstein [65]was able to identify lines of [Kr iii ] and [Se iv ] in the K-band of the near infrared spectra ofNGC 7027 and IC 5117. Soon after, Dinerstein & Geballe [66] also detected [Zn IV] lines in thenear infrared spectrum of NGC 7027 . To date these lines have been identified in almost onehundred PNe [67, 68].On the theoretical side, the calculation of collisional data for heavy species represents animportant challenge because intermediate coupling representations are inappropriate in manycases, at the same time as relativistic effects become too large to treat in the Breit-Paulirepresentation. Thus, researchers must turn to fully relativistic codes in the Dirac formalism.Recently, Badnell et al. [69] put the Dirac Coulomb R-matrix package of Berrington et al. [70]on the same footing as the traditional Breit-Pauli R-matrix codes developed by the IRON/RmaXteam. Nonetheless, these sort of calculations are still far from routine work, owing to the sizeof the atomic representations in J K -coupling.
8. Additional considerations
This review would not be complete without pointing out some important general issues ofrelevance to the present subject as well as recently developed tools for atomic spectroscopy.The accuracy of recombination rate coefficients was in recent years an important source ofconcern for modelers for their effects on calculations of ionic fractions. The issue, however,seems to have reached a point of stability as different theoretical methods, e.g. the unified R-matrix approach of Nahar and Pradhan and the Thomas-Fermi-Dirac approach in the version ofBadnell and collaborators [71], begin to yield consistent results and to release large amounts ofdata. Such theoretical results also seem to agree with recent experimental determinations (e.g.[72]). A warning must be raised though regarding low temperature dielectronic recombination,that is dominated by near threshold autoionizing channels, whose positions are difficult to getaccurately from theory. More experimental work on this area is needed to benchmark currentand forthcoming data (e.g. [73]).On its origins, one of the explicit objectives of the IRON Project was to provide a complete andreliable set of data for positive ions of interest in nebular infrared spectroscopy. A comprehensivereview of these data is presented by Badnell et al. [74]. The entire dataset produced by theIRON Project will soon be available through TIPbase [75].A very useful tool for nebular spectroscopy, the EMILI package for emission line identification,has been developed by [76]. This tool should expedite the analysis of high spectral resolutionand signal-to-noise spectra, and enable the identification of faint poorly known features. . Conclusions
The question of what fractions of Ly α photons are self-absorbed and removed by dust inphotoionized regions is rather important, as the treatment of hydrogen spectra in photoionizationmodels is one of the main limitations to their accuracy. There are also various questions in regardto heating and photo-evaporation of grains in PNe. Further research on this is fundamental forentering a new level of detailed understanding of PNe and H ii regions.Temperature diagnostics with H and He are very important as they yield new informationon the long lasting problem of temperature fluctuations in gaseous nebulae. It is important,though, to sort out the apparent inconsistencies between temperature diagnostics from differentrecombination series of lines or line ratios. Moreover, it seems plausible that temperaturefluctuations in nebulae could be accompanied by density variations. Thus, it is importantthat both quantities be derived simultaneously from the same diagnostics.As atomic models have become available for essentially all of the most prominent species inspectra of gaseous nebulae it is important to check that diagnostics of temperature and densityout of various species all yield consistent results. This would allow us to determine whetherthe atomic data have reached ultimate accuracy. A flag of warning must be raised, however,in regard to neutral species and ions with ionization potentials below that of hydrogen becausetheir spectra could be affected by photo-excitation by continuum radiation. Such is the case ofN i [77] that Copetti & Writzl [15] tried to analyze on the same grounds as ionized species.Iron and iron-peak ions are as important as ever, particularly Fe ii , which is prominentthroughout astronomy, beyond PNe research. These ions, however still pose great challenges.The open 3d structure of these species is difficult to describe owing to various factors: (1) largenumbers of strongly correlated LS terms shortly spaced in energy from the ground state; (2)strong angular correlations among configurations such as 4s , 4p , 4d , 4p4f that lead to verydemanding computations; (3) strong core-valence interactions in the vicinity of excitations of3p electrons onto the unoccupied 3d orbital, leading to the so-called giant 3p →
3d resonance;(4) electronic correlations among 3d electrons such as the radial distribution of electrons in a3d N +1 configuration differs from that of electron in a 3d N nl. More experimental work is needed,particularly in the measurement of cross sections that guide theoretical work (e.g. [86]).The importance to modern astronomy of studying isotopic abundances and n -captureelements can hardly be overstressed. Yet, such work imposes extreme demands on spectroscopicinstruments and astronomers. Carefully coordinated efforts between atomic physicists,astronomers, and instrument designers would be desirable on this subject. \ack I wish to thanks the referee of this paper, Prof. D. Morton, important discussions and correctionsto the original manuscript.