The Composition, Excitation, and Physical State of Atomic Gas in the Debris Disk Surrounding 51 Oph
DDraft version May 12, 2020
Typeset using L A TEX modern style in AASTeX63
The Composition, Excitation, and Physical State of Atomic Gas in theDebris Disk Surrounding 51 Oph ∗ Edward B. Jenkins and C´ecile Gry Princeton University Observatory,Princeton, NJ 08544 Aix Marseille Univ., CNRS, CNES, LAM,Marseille, France
ABSTRACTWe measured 304 absorption features in the ultraviolet and visible spectra of the star51 Oph, which is known to have a debris disk with a high inclination. We analyzed therelative populations of atoms in excited fine-structure and metastable levels that aremaintained by optical pumping and collisional excitation by electrons, and we foundthat most of the gas is situated at about 6 AU from the star, has an electron volumedensity 10 < n ( e ) < × cm − , and a temperature T = 8000 K. Our interpretationsreveal that the gas is partly ionized, has a column density of neutral hydrogen equalto 10 cm − , and has a composition similar to that of a mildly depleted interstellarmedium or that of Jupiter-family comets. Compared to results for disks around someother stars, such as β Pic and 49 Cet, we find surprisingly little neutral carbon. Nomolecular features were detected, which indicates that our line of sight misses themolecule-rich central plane of the disk. The tilt of the disk is also validated by ourbeing able to detect resonant scattering of the starlight by oxygen atoms.
Keywords:
Circumstellar gas (238), Circumstellar matter (241), Circumstellar disks(235), Abundance ratios (11), Debris disks (363) INTRODUCTIONThe emergence and growth in our understanding of gas and dust in orbit aroundstars that are approaching or are on the main sequence has been an important newendeavor for both observers and theoreticians (as reviewed by Hughes et al. 2018).This topic has also provided a critical link to our knowledge on the formation of
Corresponding author: E. B. [email protected]@lam.fr ∗ Based on observations with the NASA/ESA Hubble Space Telescope obtained from the Data Archiveat the Space Telescope Science Institute, which is operated by the Associations of Universities forResearch in Astronomy, Incorporated, under NASA contract NAS5-26555. c (cid:13) a r X i v : . [ a s t r o - ph . S R ] M a y Jenkins & Gry extrasolar planets (Vidal-Madjar et al. 1998 ; Wyatt 2018) and alterations of el-ement abundances in some stellar atmospheres, such as those of white dwarf stars(Jura & Young 2014). Discoveries from orbiting observatories, such as the
InfraredSpace Observatory (ISO), Spitzer Space Telescope, Wide Field Infrared Survey Ex-plorer (WISE), and
Herschel Space Observatory have revealed that approximately20% of both A-type and the F, G, and K-type stars near the main sequence exhibitmeasurable excesses of infrared emission that are well above an extrapolation of theRayleigh-Jeans tails for the radiation from the stellar photospheres (Wyatt 2018).Coronagraphic, polarimetric, and interferometric observing facilities working at vis-ible wavelengths (e.g., STIS on HST, SPHERE/ZIMPOL on the VLT, and the Kecknulling interferometer), as well as radio interferometers (ALMA, SMA, NOEMA,CARMA), the far infrared PACS instrument on the Herschel Space Observatory , andmillimeter observations taken by SCUBA-2 on the JCMT have allowed us to obtaindetailed information on the strengths and morphologies of the emissions by atoms,molecules, and dust surrounding the stars. There is an immense volume of literaturecovering these observations. There are also many publications that have highlightedthe absorption features produced by Ca II , Na I and sometimes Fe I that appear atvisible wavelengths in the spectra of the central stars. These species have even beenimaged in emission for the iconic example of an edge-on disk around the star β Pic(Olofsson et al. 2001 ; Brandeker et al. 2004).If a central star is hot enough and close enough to provide a spectrum in the ultra-violet, an opportunity to detect with great sensitivity various atomic and molecularconstituents in the circumstellar medium is presented. Intensive investigations ofthe matter surrounding β Pic got underway with observations using the
Interna-tional Ultraviolet Explorer (IUE) (Kondo & Bruhweiler 1985 ; Lagrange et al. 1987; Lagrange-Henri et al. 1988 ; Lagrange-Henri et al. 1989). These early investiga-tions highlighted the existence of atoms in a stable component at the velocity of thestar, but in addition, there were occasional appearances of absorption componentsat significantly large positive velocities, which were interpreted as falling evaporatingbodies (FEB) (Vidal-Madjar et al. 1998). The high frequency of detecting the FEBphenomenon has been interpreted to arise from perturbations by planets on the or-bits of comet-sized objects (Karmann et al. 2001 ; Th´ebault & Beust 2001), whicheventually plunge toward the star.Following the era of
IUE , observations with the Goddard High Resolution Spec-trograph (GHRS) (Brandt et al. 1994) and Space Telescope Imaging Spectrograph(STIS) (Woodgate et al. 1998) aboard
Hubble Space Telescope ( HST ), together withthe
Far Ultraviolet Spectroscopic Explorer ( FUSE ) (Moos et al. 2000 ; Sahnow etal. 2000), provided a vast improvement in the number and quality of the results.Absorption features associated with the debris disk of β Pic have been the most in- An unbiased survey of A-type stars by Thureau et al. (2014) indicates a somewhat higher percentage,ranging between 21 and 54%. A study by Mo´or et al. (2017) of A-type stars with dust-rich disksrevealed that 11 out of 16 of them exhibited measurable CO emission. ircumstellar Gas Surrounding 51 Oph σ Her (Chen & Jura 2003 - FUSE only), AB Aur (Roberge et al. 2001 - FUSE andSTIS), and HD 109573 (upper limits only) (Chen & Kamp 2004 - FUSE and STIS),HD 32297 (marginal S/N) (Fusco et al. 2013), HD 141569 (Malamut et al. 2014),HD 163296 (Tilling et al. 2012 - STIS medium res. echelle) and HD 172555 (Gradyet al. 2018 - STIS and COS).The origin of gas accompanying debris disks is a fundamental problem: is it primor-dial, is it ejected from the star (and slowed down by some process), or – more likely– does it arise from the orbiting planetesimals, comets, or dust grains that are eitherevaporating atoms or liberating gas as they collide with each other? An examinationof the relative abundances of different elements, which often differ appreciably fromthe solar abundances, can provide some guidance on this question (Xie et al. 2013).A distinctive property of the UV absorption spectra of gas associated with debrisdisks is the appearance of features arising from many different excited fine-structureand metastable levels of various atoms and ions. The most conspicuous absorptionsfrom levels of high excitation are those from Fe II , as we will illustrate later in Figure 2,which have radiative lifetimes ranging from 10 seconds to 50 hours (Quinet et al.1996) and must be populated by radiation pumping and/or collisional excitationswith electrons (the effects of collisions with neutral atoms and ions are small bycomparison). For densities below the critical densities of the levels, there can bedeviations from a straight-line relationship for the level populations in an excitationdiagram (ln( N/g ) vs.
E/k ), which in turn will reveal clues on the environmentalparameters that are responsible for the excitations. Such deviations are clearly evidentin this type of diagram for the Fe I level populations derived from a high-S/N spectrumof β Pic at visible wavelengths (Vidal-Madjar et al. 2017).Our current study is directed toward using absorption features in the ultraviolet togain information on the gas constituents that surround the star 51 Oph. This HerbigAe/Be star has been assigned a spectral classification B9.5Ve by Dunkin et al. (1997),and has the parameters M = 3 . M (cid:12) (Jamialahmadi et al. 2015), T eff = 10 ,
250 K,log g = 3 .
57, [M/H] = +0.10, and an estimated age 0 . +0 . − . Myr (Montesinos et al.2009). The distance to the star is 123 pc (Arenou et al. 2018 ; Luri et al. 2018). Aspectral energy distribution with an infrared excess (Malfait et al. 1998), the presenceof emissions at 10 µ m from silicates (Fajardo-Acosta et al. 1993 ; Meeus et al. 2001),7 . − . µ m from PAHs (Keller et al. 2008), several atomic excited atomic fine-structure levels (Meeus et al. 2012 ; Dent et al. 2013), and various gas-phase molecules(van den Ancker et al. 2001) all present clear evidence of gaseous and solid mattercirculating around this star. We are not the first to study absorption features in theUV spectrum of this star, and we will refer to a number of previous investigations Jenkins & Gry later in the paper. The UV spectrum of 51 Oph exhibits a rich assortment of linesfrom many different elements, and these elements have significant populations inthose metastable levels that cannot undergo rapid radiative decays. We had the goodfortune of locating relevant atomic data for N I , Fe II , and Ni II that allowed us toinvestigate in some detail the factors that governed the excited level populations. DATATable 1 summarizes some key parameters of the observations. Nearly all of theconclusions in this paper are based on observations taken at ultraviolet wavelengthswith the highest resolution echelle modes of STIS aboard
HST . These data are sup-plemented by observations by the GHRS over very limited wavelength ranges. Thespectra recorded by the GHRS are used only to make comparisons with STIS datato check for possible time variability of some spectral features. A limited number ofabsorption features relevant to our study appear at visible wavelengths. To measuresuch features, we made use of archived echelle spectra recorded by the Ultravioletand Visual Echelle Spectrograph (UVES) (Dekker et al. 2000) on the
Kueyen VeryLarge Telescope ( VLT ) operated by the European Southern Observatory.Observations of 51 Oph in the far ultraviolet recorded by
FUSE have been reportedby Roberge et al. (2002). We confined our use of the
FUSE spectra to probe only theelements N I , Cl II , P II and a metastable level of O I , which we could not measure inthe STIS spectrum. Finally, we used a spectrum recorded by the Hopkins UltravioletTelescope ( HUT ) (Kruk et al. 1995 ; Dixon et al. 2013) to validate our choice ofparameters for a theoretical stellar spectrum that we used to calculate the opticalpumping of levels, as will be discussed later in Section 6.1.1.It is well established that absorption features arising from circumstellar gas canvary with time, with the appearance and disappearance of velocity-shifted features.A large collection of 51 Oph spectra recorded by the
IUE satellite at different timesshow occasional appearances of features at very high negative and positive velocities,ranging from many tens of km s − to as much as 100 km s − (Grady & Silvis 1993).All of our observations with STIS were taken over a time interval of only two hours,so the possible influence of such variability is small. Nevertheless, we also made useof UVES spectra at visible wavelengths taken at a very different time for our analysisof Na I , Ca I , Ca II , Ti II , Cr II , Mn II , and Fe I .To gain some indication that the features at low velocities that we studied here arenot varying in time, we compared GHRS observations taken many years earlier thanour STIS ones (see Table 1) and found no evidence of any significant changes in theappearances of the features. Figure 1 shows the close match of features in the GHRSand STIS spectra taken at nearly the same wavelength resolutions. Of course, thisevidence is only circumstantial; we cannot rigorously rule out the existence of somevariability that could cause features in the UVES spectra to not match those in theSTIS ones. For instance, Roberge et al. (2002) reported some very small changes in ircumstellar Gas Surrounding 51 Oph Table 1.
ObservationsInstrument, Mode Aperture Resolution λ Range Observationarc-sec λ/ ∆ λ (˚A) DateSTIS E140H, E230H 0 . × . − . × .
03 200,000 1879.5 − − − − − − − AugCD − ×
20 15,000 992 − a HUT 20 500 800 − a We refrained from using exposures taken in 2000 with a larger entrance aper-ture. Also, various technical issues discussed by Roberge et al. (2002) made thereduction and interpretation of these earlier spectra less attractive. F l ux ( - e r g c m - s - Å - ) Figure 1.
Spectra recorded at different times by the GHRS ECH-B mode (black trace)and STIS E230H mode (red trace). The absorption on the left is from the ground state ofMn II , while the two features to the right are from metastable levels of Fe II . Except fora slightly enhanced absorption on the long-wavelength side of the very strong line line at2607.7 ˚A from an excited fine-structure level of Fe II in the GHRS spectrum, the spectraappear to be identical. These absorption features and their identifications can be seen inthe upper panel of Figure 2. Jenkins & Gry some profile shapes that occurred between their two
FUSE exposures separated by 6days in 2000. ANALYSIS OF THE SPECTRA3.1.
Absorption Features from Atomic Ground-State, Excited Fine-structure, andMetastable Levels
Table 7 in the appendix lists the different features that we analyzed in the UV andvisible spectra of 51 Oph. Most were strong enough to appear well above the noisein the continuum level and could yield column density values, while others were soweak or invisible that we could obtain only upper limits to column densities using aprocedure outlined below in Section 3.2.Features from the unexcited states of atoms probably have some contamination fromabsorption by the foreground interstellar medium, but all of the lines from the excitedlevels arise from just the circumstellar gas around 51 Oph. For the ground-state levels,we attempted to distinguish contributions from the foreground interstellar medium(ISM) from circumstellar ones, using the identifications of distinct radial velocitycomponents as a guide, as we will discuss in Section 4. To accomplish this goal, weused the profile-fitting software called
Owens.f that was developed in the 1990s byM. Lemoine and the French
FUSE science team. To obtain a more coherent pictureof the components, the analysis software allows us to fit simultaneously different linesfrom several energy levels and include several species in a single structure of velocitycomponents. A solution emerges where all central velocities and turbulent velocitywidths for all species of the same component are identical, but with thermal broad-ening widths that recognize the expected variations that change with atomic masses.In some cases, we sensed that two velocity components for purely circumstellar gaswere partly resolved, but in others, only one was present.
Owens constructs trial theoretical profiles that are then convolved with the instru-mental line spread function (LSF). The spectra have not been normalized to thestellar continuum; instead, the stellar profile is included in the fit as n + 1 free param-eters for an n -degree polynomial ( n is always ≤ . × . . × .
03 arc-sec.For both slits, the LSFs are composed of a broad and a narrow component whose fullwidths at half maxima (FWHM) are derived by performing a double-Gaussian fit tothe tabulated LSFs. As a result, we adopted for the low-amplitude, broad componentan FWHM of 4.9 pixels and 5.1 pixels for E140H and E230H data, respectively, andfor the taller narrow component, an FWHM of 1.26 pixels for E140H and 1.45 pixelsfor E230H. For the UVES data, we have adopted a Gaussian LSF as specified by Fig-ure 2.7 of the UVES User Manual (Sbordone & Ledoux 2019): FWHM = 2.4 pixels ircumstellar Gas Surrounding 51 Oph ∼ . λ/ , ∼ . λ/ , . − . − , effectively slightly less than 0.5 STISpixels (0.66 km s − for the E230H data). However, we note the existence of a generalshift of 0.6 − − between lines in the E230H spectrum and lines in the E140Hspectrum, and another 0.2 km s − shift relative to the UVES spectra. Because mostof the neutral lines are observed with the E140H spectrum (and UVES) whereas mostof the Fe II and Ni II lines are in the E230H spectrum, this could suggest a slightvelocity shift between neutrals and ions. However, we think it is not the case becausethe same velocity shift is observed for the few Fe II lines observed in the E140H spec-trum. In short, we regard that the absolute velocity in the STIS spectra can have anerror of at least 0.8 km s − , and we conclude that the main disk component velocitydoes not differ significantly for all studied elements.Figure 2 shows some representative pieces of the STIS echelle spectrum with lineidentifications. Table 2 lists our results for the column densities. We estimated ourmeasurement uncertainties for the lines using the ∆ χ method described by H´ebrardet al. (2002) . We then evaluated the uncertainties in log N by combining the f -valueand measurement uncertainties in quadrature. Many of the column density outcomeswere based on a collection of features with different uncertainties in their f -values. Insuch cases, we gave the most weight to the f -value uncertainties of the lines that wereneither very weak nor strongly saturated, as these medium-strength lines are the onesthat are most influential in determining the column densities. In addition, we made aworst-case assumption that the uncertainties of many f -values derived from a singlesource had systematic errors that all had the same direction and amplitude (i.e., wedid not consider that errors for different lines would tend to cancel each other).3.2. Determinations of Upper Limits for Column Densities
Several important species show either extremely marginal detections or no dis-cernible features. For such cases, we did not attempt to evaluate profile fits. Instead,we carried out formal measurements of equivalent widths W λ over the expected spansof the features and assumed that the column densities scaled in direct proportion to W λ , i.e. the absorptions were completely unsaturated. Our initial outcomes for upperlimits were expressed in terms of a formal value for W λ plus a 1 σ positive excursion.However, our final expressions went beyond these determinations and stated columndensity upper limits in terms of a modification of this initial upper bound that madeuse of a Bayesian prior that acknowledges that negative real line strengths are not Jenkins & Gry J = 1/2 J = 3/2 F e II F e II M n II M n II F e II ( ) F e II ( ) F e II ( ) F e II ( ) F e II ( ) F e II ( ) F e II ( ) F e II ( ) F e II ( ) F e II ( ) F e II ( ) F e II ( ) F e II ( ) F e II ( ) F e II ( ) S i II g r o und s t a t e Si II excited fine structure state N I ( ) N I ( ) N I ( ) C (cid:2) I g r o und s t a t e F e II ( ) Wavelength (Å) C (cid:2)(cid:1) I excited fine structure state (a)(b)(c) (d) J = 3/2 J = 1/2 F e II ( ) F e II ( ) F e II ( ) F e II ( ) F e II ( ) F e II ( ) F e II] ( ) F e II] ( ) Wavelength (Å) F l u x ( - e r g c m - s - Å - ) F l u x ( - e r g c m - s - Å - ) Wavelength (Å) F l u x F l u x Figure 2.
Noteworthy circumstellar features in the spectrum of 51 Oph. Absorptions frommetastable levels are identified with the relevant atom or ion, along with the excitationenergy (in cm − ) and the degeneracy of the lower level. Features identified in red textindicate absorptions from the ground states, which may suffer some contamination fromforeground interstellar gas. Panel (a) shows a portion of the spectrum dominated by linesfrom Fe II metastable levels. Panels (b)-(d) show features from the two fine-structure levelsof Si II , the metastable levels of N I , and the two fine-structure levels of Cl I , respectively. allowed. This avoids the absurd result that arises when random positive intensityfluctuations or a continuum placement that was too low creates a raw answer for W λ that is ≤ − σ uncertainty in its value. The details of this calculation and expositionson how it behaves are given in Appendix D of Bowen et al. (2008). This methodyields a seamless transition to a conventional expression of a measured value plus its1 σ upper error bar when the outcomes progress to higher levels of significance. ircumstellar Gas Surrounding 51 Oph Table 2 . Column DensitiesSpecies Excitation g log N Energy (cm − ) (cm − )C I a
43 5 < II < I ± . b ± . ± . ± . ± . I ± . > I ± . I < I < II ± . ± . I < II ± . b,c
164 3 12.73 ± . c
469 5 13.11 ± . c S I < I ± . ± . II < I < II ± . d Ti II ± . II ± . ± . ± . ± . ± . ± . ± . ± . ± . Table 2 continued on next page Jenkins & Gry
Table 2 (continued)
Species Excitation g log N Energy (cm − ) (cm − )20024 8 11.93 ± . ± . II ± . ± . ± . ± . ± . I < II ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . Table 2 continued on next page ircumstellar Gas Surrounding 51 Oph Table 2 (continued)
Species Excitation g log N Energy (cm − ) (cm − )26352 10 12.75 ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . II ± . ± . ± . II ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . < Table 2 continued on next page Jenkins & Gry
Table 2 (continued)
Species Excitation g log N Energy (cm − ) (cm − )Cu II ± . ± . ± . I ± . II ± . a Unlike other lines that we list in this table, we havedetermined from the discordant radial velocity of thisfeature that it arises purely from the foreground in-terstellar medium (ISM). The meaningful result forC I in the circumstellar gas can be understood fromour upper limit for the 3P state at an excitation of43 cm − , which leads to log N (C I) < .
40 for thesum of the three fine-structure levels of the groundstate; see the discussion in Section 7.1. b This ground-state column density was determinedfrom FUSE data, where we do not have the velocityresolution to separate the circumstellar contributionfrom that arising from the foreground ISM. Hence,this value represents the sum of the two sources. c This stated error refers to the measurement uncer-tainty alone, since the uncertainty in the f -value isnot available. d This value is consistent with log N =10 . . , − .
04) for the circumstellar Ca II de-termined by Crawford et al. (1997).4. OVERVIEW OF VELOCITY COMPONENTSIn the spectrum of 51 Oph at all wavelengths, all circumstellar features from all ofthe singly-ionized species show one predominant absorption at − . ± . − ,which we call Component 1. In the strongest lines, another absorption is clearlyvisible, broader, shallower, and displaced by − . ± . − relative to the maincomponent. We refer to it as Component 2. Therefore, a two-component fit is per-formed for all ions and gives a good fit to the lines as shown in a few examples offitted profiles for a selection of species in Figure 3. However, because Component 2is very broad ( b > . − in all cases) and kinematically very close to Compo-nent 1, in faint lines, its absorption is very shallow and difficult to distinguish fromthat of Component 1, even when its absorption amplitude is somewhat higher thanthe noise. Therefore, for species (or energy levels) that do not have strong lines, thedistribution of matter between the components is not reliable. On the other hand,for the abundant energy levels that do have strong lines as well as unsaturated lines,where the distribution of matter in the two components is more reliable, no significant ircumstellar Gas Surrounding 51 Oph Figure 3.
Examples of the absorption line profiles arising from gas in the circumstellar diskof 51 Oph. Excited lines for ions can be fitted with two components (excited lines of Si II andFe II ) consisting of a strong component that has width consistent with a thermal broadeningat T = 8000 K and turbulent broadening b , turb = 1 . − , which is accompanied by aweaker one that is broad ( b > . − for both thermal and turbulent broadening) anddisplaced to slightly more negative velocities. For neutral atoms (panels on the right), wedetect only the strong component. The central velocities of the components are markedwith the identification numbers discussed in the text. For absorption out of the groundstate of Si II (upper left panel), there is an additional contribution from the foregroundISM. In all of the panels, the black, histogram-style curves represent the observations, thered lines are the best fits, and the green lines depict the separate fits to the two individualcomponents. Stellar continua are shown in blue. differences have been noted in the relative populations. So, although the fits are donewith two components, we decided to list the results as the total column density inthe two circumstellar components taken as a whole.The profiles of the neutral atoms N I and Cl I that do have relatively strong lines insome of their metastable or fine-structure levels do not show a second, broad compo-nent. The profiles of the strongest O I lines (1302, 1304 & 1306 ˚A) are very difficultto analyze because of their strong saturation, as we will show in Section 8. However,there is evidence for some broad contribution, but at a small level, apparently of order1% of Component 1. This suggests that neutrals are proportionally far less abundantin the broad component.Except for Ca II and Na I , we found that for both neutrals and ions, our fits toComponent 1 yielded turbulent velocities b turb equal to 1 . ± .
17 km s − , assumingthat the contribution to the total width caused by thermal broadening was consistentwith T = 8000 K, as we found from our analysis of the metastable populations (whichwill be presented in Section 6.6). For Ca II and Na I we derived slightly highervelocity widths: b turb = 3 . − .4 Jenkins & Gry
Crawford et al. (1997) have reported the observation of the line of sight to 51 Ophin the Ca K line recorded at very high resolution ( R ∼ , II line from a metastable level in the visible. It had ameasured velocity of − ± − , in good agreement with the Ca II lines of ourComponent 1 in the UVES data. Their outcome for the width of this Ca II feature, b = 3 . . , − .
3) km s − , is compatible with ours.For the unexcited states, our ability to identify and disentangle the circumstellarcontributions from interstellar ones was greatly facilitated by the interpretation ofCrawford et al (1997). As shown in their Figure 1 and Table 2, four absorptioncomponents are detected in addition to the circumstellar one at − − : twovery narrow, very cold, components at − − − , and two broadercomponents at −
25 and −
29 km s − , corresponding to velocities of the diffuse localgas (the predicted velocity of the local cloud that surrounds the Sun in all directions is − − toward 51 Oph according to Gry & Jenkins (2014)). We indeed retrievedthis velocity structure in the two Ca II lines in our UVES spectra, but with our lowerresolution, each pair of interstellar components appeared as a single absorption, at v ∼ − . − for the cold gas ( b ∼ . − ) and v ∼ − . − forthe warm local gas ( b ∼ . − ). (The subscript “2” is kept for the second diskcomponent and is not included in the Ca II model, because it is undetectable in thefaint Ca II lines.)We subsequently apply this interstellar velocity model to fit all ground-state fea-tures. For some of them, the second broad disk component must be considered, whichmay have some overlap with the absorption of interstellar Component 3, resulting ina somewhat increased uncertainty. Nevertheless, to a good approximation, we wereable to determine the disk column density of all atoms in their ground states, exceptfor N I and P II because the unexcited states of these species were only availablein the FUSE spectrum, which does not have the resolution to separate the circum-stellar from the interstellar velocity components. We point out that for C I no diskcomponent is detected, the only feature clearly visible is due to the cold interstellarComponent 3. Component 3 also dominates the absorption in Na I , however, a diskcomponent is also detected. For the faint O I line at 1355 ˚A, only the circumstellarfeature is detected, which allowed us to obtain a reliable measurement for the disk N (O I ), which would be impossible with the sole strong 1302 ˚A line. LACK OF CO ABSORPTIONThe wavelength coverage of the spectrum recorded by STIS for 51 Oph coversthe locations of many absorption bands of CO, the strongest of which are in the A Π − X Σ + system (Morton & Noreau 1994). For the vibrational bands thatrange from 0 − −
0, we see no evidence of any absorption features at the ircumstellar Gas Surrounding 51 Oph II inmetastable levels. This lack of CO UV absorption is in stark contrast to observationsof far infrared emission from the rotation and vibration bands of CO that surroundsthis star (van den Ancker et al. 2001 ; Thi et al. 2005 ; Berthoud et al. 2007 ;Tatulli et al. 2008). It is difficult for us to quantify an upper limit for the columndensity of CO because our ability to detect this molecule depends on the broadeningcaused by rotational excitation of the ground vibrational state, which can be large.Nevertheless, even if much of the CO is so highly excited that the UV features arebroadened and many of the molecules are in excited vibrational states, our detectionthreshold is considerably below the column density estimates of order 10 − cm − in the models of Thi et al. (2005) and Berthoud et al. (2007). We also failed to detectfeatures from vibrationally excited H , such as those found by Meyer et al. (2001)toward HD 37903. Our lack of molecular absorption is consistent with the inability ofRoberge et al. (2002) to find any H absorption features out of the ground vibrationalstate in a far-UV spectrum of 51 Oph recorded by FUSE. Therefore, we conclude thatthe line of sight to the star does not pass through a zone near the central plane of thedisk where there is enough shielding from H and CO-dissociating radiation to allowappreciable concentrations of these molecules to accumulate. Essentially, we are notviewing the disk exactly edge on, even though the inclination angle could be not toofar from 90 ◦ . This conclusion is similar to that of Roberge et al. (2014), who failedto detect CO absorption in the spectrum of 49 Ceti, which has a circumstellar diskthat exhibits strong CO emission (Hughes et al. 2017). INTERPRETATION OF THE EXCITED LEVEL POPULATIONSThe excited fine-structure levels of the ground electronic states and the metastablestates at higher energies of the different atomic species are primarily populated byeither inelastic collisions with electrons and neutral atoms, or by optical pumping bythe light from the central star (or possibly a combination of all three). For now, wewill dismiss the importance of collisions with H atoms, and we will justify our neglectof H-atom interactions later in Section 6.5.2.If the electron density at a temperature T is above the critical density for excitingthe levels of a particular atom, we would expect the level populations to achieve athermodynamic equilibrium, where a plot that shows a relationship for ln( N i /g i ) vs. E i /k to be a straight line with a slope equal to − /T for all levels i . At lower electrondensities or within a dilute radiation field, deviations from this trend can be influ-enced by differing collision strengths and radiative decay rates of various levels, fromwhich one can attempt to solve for representative values of n ( e ), T , and the strengthof optical pumping. We acknowledge that the electron densities and temperaturesprobably vary throughout different zones in the circumstellar environment that con-tain appreciable concentrations of gas, so the outcomes of any analyses representaverages weighted according to the local densities of the atoms being investigated.6 Jenkins & Gry
We interpret the strength of optical pumping in terms of the distance of the gas fromthe center of the star R g , but the answers that we quote actually represent (cid:104) R − g (cid:105) − / (or to be more precise, 4 (cid:104) W (cid:105) , where a radiation dilution factor W is defined in Eq. 1as shown later in Section 6.1.2), again weighted according to atomic densities.In the following subsections, we will discuss the concepts that are important for theinterpretations of the excited level populations of N I , Fe II , and Ni II , where we havemeasurements of good quality for the column densities of many different metastablestates and for which we were able to find a broad range of relevant atomic datathat allowed us to solve for the relative populations. While our principal effort willbe to focus our attention on these three species, we will also present a rudimentaryinterpretation of an excited level of O I caused by inelastic collisions with electronsand H atoms to demonstrate the relative importance of these two collision partners.6.1. Optical Pumping
Stellar Flux Model
To estimate the effects of optical pumping, we must start with a representationof the flux from the central star at all of the relevant wavelengths for transitionsthat can populate the upper electronic levels of different atoms. We rely on modelsfor the stellar fluxes, as the wavelength coverage of our observations is not broadenough to cover all possible transitions. However, we can use UV observations toguide our choice for the best match of an effective temperature. The best observationfor this purpose is that obtained by
HUT , which provided continuous coverage ofthe critical wavelength range (see Table 1) where small differences in temperaturehave the strongest effect. The best match to the observations arose from the mean ofthe 11,000 and 10,000 K effective temperature models from UVBLUE (Rodriguez-Merino et al. 2005), and we adopted a surface gravity log g = 4 .
0. Both of theseparameters are consistent with the values listed for 51 Oph by Dunkin et al. (1997)and Montesinos et al. (2009). For λ < that hadthe same parameters, so that we could obtain fluxes out to 6500 ˚A. While Dunkin etal. (1997) determined that the projected rotation velocity v sin i = 267 ± − for51 Oph based on stellar spectral features at visible wavelengths, the lines appearing inthe STIS spectrum seem more consistent with approximately 200 km s − . We expectthat this difference arises from gravity darkening, which de-emphasizes equatorialfluxes at short wavelengths. Shortward of the Balmer break, we smoothed the modelspectrum with a kernel scaled to a rotation velocity of 200 km s − , while for longerwavelengths, we increased the assumed velocity to 267 km s − .In principle, an additional contribution to the radiation field can arise from the emis-sion from a shock within a magnetic accretion column. For 51 Oph, this contribution ∼ eps/uvblue/uvblue.html Available from http://cdsarc.u-strasbg.fr/viz-bin/qcat?J/A+A/618/A25 ircumstellar Gas Surrounding 51 Oph
Radiation Dilution Factor
The rapid rotation of 51 Oph distorts the photosphere into an oblate spheroid.Interferometric measurements by Jamialahmadi et al. (2015) indicated that the majoraxis θ eq = 0 . ± .
05 mas and the minor axis θ pol = 0 . ± .
01 mas, which is equivalentto R eq = 8 . ± . R (cid:12) and R pol = 5 . ± . R (cid:12) if the star is at a distance d = 123 pc(Arenou et al. 2018 ; Luri et al. 2018). As viewed in the equatorial plane at a largedistance from the star, the solid angle occupied by the star’s photosphere should beequivalent to that of a sphere with a radius R ∗ = [(8 . . R (cid:12) = 6 . R (cid:12) =0 . R g from the center of the star, the radiation has a dilutionfactor W given by W = − (cid:34) − (cid:18) R ∗ R g (cid:19) (cid:35) (1)(Viotti 1976). 6.1.3. The Attenuation of Radiation for Strong Transitions
Figures 1 and 2 show absorption features that reach zero intensity, indicating thatthere is sufficient gas around 51 Oph to create deficiencies of flux in the cores of thestrongest lines. Therefore, we must recognize that the starlight is shielded for manytransitions that can be influential in the optical pumping of the levels. The amountof this shielding varies from virtually nothing at small distances from the star toan almost complete loss of photons for atoms at great distances. When photons areabsorbed, they can be reradiated at the same or lower energies when the excited levelsdecay to ones at lower energies. We invoke the simplifying assumption that, exceptfor stimulated emission, such photons are lost and do not cause further pumpingelsewhere. This concept is probably valid for gas confined to a thin disk around thestar, because the photons from the spontaneous decays are emitted isotropically.An average level of attenuation for all of the atoms may be computed for a giventransition by measuring the equivalent width of the absorption and comparing it tothat which one would expect if the velocity dispersion were so large that the centraldepth of the line would not be far below the continuum level. Instead of using actualequivalent width measurement outcomes, we simplified our computations by invokinga standard Gaussian curve-of-growth behavior for all equivalent widths and assumeda single velocity dispersion parameter b that approximated the composite behaviorof the components described in Section 4. A spot check of equivalent widths for N I and Fe II indicated that good values for b are 4.5 and 4 km s − , respectively. Hence,we can derive values of the central optical depths of the lines, τ = 1 . × N f λ/b , (2)8
Jenkins & Gry where N is expressed in cm − , λ in ˚A, and b in km s − . It then follows that we cancompute the saturation factor s.f. (the ratio of the equivalent widths of a saturatedto unsaturated line), s . f . = 2 F ( τ ) √ πτ , (3)where F ( τ ) = √ π ∞ (cid:88) n =1 ( − n − τ n n ! √ n . (4)When we computed pumping rates, we multiplied the stellar fluxes by the appropriatevalues of s.f. for all of the transitions of N I and Fe II . All of the lines of Ni II wereweak and thus did not need corrections for saturation.6.1.4. Radiative Transition Rates
The level populations are influenced by the effects of absorption, stimulated emis-sion, and spontaneous decay. We follow a formulation given by Draine (2011, pp53-55) and express the rates of these processes in terms of a dimensionless local pho-ton occupation number for the starlight, n γ = c πhν u ν , (5)where the energy density u ν = 4 πc W B ν ( T )(s . f . ) . (6)The W term in this equation is the dilution factor defined in Eq. 1, and B ν ( T ) isthe radiation intensity at a frequency ν for a Planck distribution with a temperature T , and s.f. was defined in Eq. 3. It then follows that for interactions between anygiven pair of levels, the rate of downward transitions per unit volume is given bythe product of the number density n u of atoms in the upper level and the sum ofspontaneous and stimulated emission rates, dn (cid:96) dt = n u A u(cid:96) (1 + n γ ) , (7)where A u(cid:96) is the Einstein A coefficient for the spontaneous decay from level u to alower level (cid:96) . The reverse process that repopulates the upper level is given by dn u dt = n (cid:96) g u g (cid:96) A u(cid:96) n γ , (8)where g u and g (cid:96) are the degeneracies of the two levels. For T that applies to B ν ( T ),we adopted 10,000 K and modified the radiation density to reflect the small deviationsof the star’s model flux from this distribution. The A ∼ peter/newpage/. Not all possible transitions have values of A listed; missing values are generally semi-forbidden transitions or allowed ones at very long wavelengths. We have no choice but to assumetheir contributions are small and thus can be neglected. ircumstellar Gas Surrounding 51 Oph Collisional Excitation and De-excitation by Electrons
We operate under the premise that collisions with electrons dominate over those withneutral atoms. We will justify this approach by citing in Section 6.5.2 some specificexamples that demonstrate that the collisions by neutral partners have reaction ratesthat are considerably weaker than those associated with electrons.For N I we used the collision strengths Ω( T ) defined by the fits by Draine (2011,Appendix F) to the calculations by Tayal (2006). We obtained values for Ω( T ) thatapplied to Fe II and Ni II from Tayal & Zatsarinny (2018) and Cassidy et al. (2010),respectively. Collision rate constants C u,(cid:96) ( e, T ) for de-excitation of an upper level u to a lower one (cid:96) are given by C u,(cid:96) ( e, T ) = h Ω( T ) g u (2 πm e ) / ( kT ) / = 8 . × − Ω( T ) g u T / cm s − . (9)The reverse process (i.e., excitation from a low level to a higher one) is related to C u,(cid:96) ( e, T ) through the principle of detailed balancing, C (cid:96),u ( e, T ) = g u g (cid:96) exp( − ∆ E/kT ) C u,(cid:96) ( e, T ) , (10)where ∆ E is the energy separation of the two levels.6.3. Special Considerations for Nitrogen
We will show later that, along with electrons, there are significant amounts of bothionized nitrogen and neutral hydrogen in the gas (Sections 6.5.2 and 7.3, respectively).This raises the prospect that for neutral nitrogen atoms there may be additionalways to populate the metastable levels, which we will investigate in the following twosubsections. 6.3.1.
Charge Exchange
The ionization fractions of N and H are coupled to each other through chargeexchange reactions. In particular, the reaction N + + H → N + H + could create N atoms not only in the ground p S o state but also in the excited 2 p D o levels (withexcitation energies E = 19224 and 19233 cm − ). However, we compared the strengthof the collisional excitation rate constants against the charge exchange rate constantsto the excited level of N calculated by Lin et al. (2005), and we found that theformer dominates over the latter by a factor > for 4000 < T < K. Therefore,we feel that it is safe to ignore the contribution of charge exchange in populating the2 p D o level of N . 6.3.2. Recombination
Ignoring other processes, the volume density of neutral nitrogen atoms in an excitedstate n (N ∗ ) (where * denotes either D or P) is given by the equilibrium condition n (N ∗ )[ A u,(cid:96) + C u,(cid:96) ( e, T ) n ( e )] = n (N + ) n ( e ) α partial (11)0 Jenkins & Gry where α partial is the recombination rate from the ground state of N + to the relevantstate of N ∗ . Knowing that n (N ) ≈ n (N + ) n (H ) n ( e ) , (12)we can divide both sides of Eq. 11 by n (N ) and obtain the expression n (N ∗ ) n (N ) = n ( e ) α partial n (H )[ A u,(cid:96) + C u,(cid:96) ( e, T ) n ( e )] . (13)From our findings that will be presented in Section 7.3, we can substitute 2 . n ( e ) fora lower limit for n (H ) and simplify this expression to the form n (N ∗ ) n (N ) ≤ n ( e ) α partial . A u,(cid:96) + C u,(cid:96) ( e, T ) n ( e )] . (14)Later (Section 6.6), we will show that the temperature T = 8000 K is strongly fa-vored. At this temperature, the partial recombination coefficients α partial to thetwo D metastable states equal 1 . × − cm s − , which was obtained from thecompilation at https://open.adas.ac.uk/adf08, and for the two P states α partial =8 . × − cm s − . For what we will later consider to be the largest value for theelectron density, n ( e ) = 3 × cm − , we obtain the numerical result for Eq. 14 n (N ∗ ) /n (N ) = 1 . × − for the D states (independent of n ( e ) because the A u,(cid:96) term is so small) and 1 . × − for the P states. The respective observational coun-terparts to these two values are 4 . × − and 5 . × − . While in this developmentwe have disregarded collisional exchanges between P and D states, we can still safelyassert that the recombinations are of little consequence to our analysis.While the 3.5 orders of magnitude may seem like a very large margin of safety(i.e., observed excited fractions vs. our computed values arising from recombination),in our particular case it may be reduced to only 2 orders of magnitude (AhmadNemer, private communication) by an additional recombination mechanism calledRydberg Enhanced Recombination (RER), which in some cases can dominate overother recombination routes at temperatures below about 10 K (Nemer et al. 2019).6.4.
Solutions for the Level Populations
We return to the general interpretations of the metastable excitations and derivesolutions for the equilibrium concentrations of the n electronic levels (the groundstate and its excited fine-structure levels, plus all higher metastable levels includingones we could not see) by solving n simultaneous linear equations. For N I n = 159,Fe II n = 340, and Ni II n = 686. The occupation fractions of these levels that wewish to derive are expressed in the terms contained within a column vector f whichis multiplied by an n × n matrix R consisting of the radiative and collisional rates toform an n -element column vector b = [0 , , . . . , R · f = b . (15) ircumstellar Gas Surrounding 51 Oph n − b set the requirement that for each row of R the sum ofthe products of the rates and the unknown quantities in f balance each other so thata given constituent is in equilibrium with the others. Setting the last element in b and all elements in the last row of R equal to 1 insures that the sum of all constituentfractions is equal to 1.Terms in the diagonal elements of R represent the sums of losses of level i thatpopulate other levels j , which consist of spontaneous radiative decays, stimulatedemission, upward and downward collision rates, and absorption, R i,i = − (cid:88) j [ A ij, ( i>j ) (1 + n γ ) + C i,j ( e, T ) n ( e ) + g i g j n γ A ij, ( i
Foreground Contamination
Our ability to isolate components with different radial velocities in the STIS andUVES spectra and identify foreground contributions from the ISM helped us to focuson measuring just the contributions from the circumstellar gas. Nevertheless, we feltthat it was important to acknowledge that for the unexcited levels of the atoms, ourdeterminations might have been contaminated by contributions from some unrecog-nized portion of the ISM that had a velocity that overlapped that of the circumstellargas, as we indicated in Section 4. For this reason, we elected to not include the groundlevels of Fe II and Ni II in our analysis, because these two elements have many otherlevels that we can analyze. However, for N I we were able to measure only 4 excitedlevels, and thus we regarded the zero-excitation state as being critical for analyzingeffects of collisional excitations and optical pumping. Unfortunately, for determiningthe column density of the unexcited level of N I , we could access only a single unsat-urated feature that was detected in a FUSE spectrum, where the velocity resolutionwas insufficient to isolate the interstellar from the circumstellar contributions. Forthis reason, we included our basic measurement of the ground state but lengthenedthe downward error bar to ∆ ln N = − . Jenkins & Gry
The main benefit of including the ground state of N I is to disallow solutions thatrequire a column density much higher than the observed one.6.5.2. H Atom vs. Electron Excitations
Next, we will explore the issue of collisional excitations by H atoms and how theycompare to those by electrons in populating the upper levels. We start with a dis-cussion about O I to serve as an illustrative example, and later we will briefly touchupon N I and Fe II .We measured O I only in its lowest fine-structure level with a degeneracy g = 5,so our determination of the total column density of this atom should be multipliedby 9/5 to account for the realistic premise that at the large electron and hydrogendensities and temperatures of the circumstellar material (see Sections 6.6 and 7.1) theoccupations of the atoms in all three fine-structure levels are proportional to theirdegeneracies to within our accuracy of determining N (O I in the unexcited level.Hence, we estimate that the total column density of O I that we must consider forthe P , , state is about 4 . × cm − .Krems et al. (2006) addressed the problem of interpreting the emission lines at6300 and 6364 ˚A in different astrophysical contexts and published calculations of rateconstants for H atom collisions that could populate the D upper level u that has anexcitation of 15868 cm − . As indicated in Table 7, there is an allowed transition outof this level at 1152 ˚A. The FUSE spectrum shows that this line is strongly saturatedand it is poorly resolved by the instrument, so we are only able to derive a lower limitfor the column density equal to 2 × cm − . For our adopted value of N (O I ) inthe P , , state, we find that the ratio n (O u ) /n (O (cid:96) ) for the population of the upperlevel u relative to that of the lower level (cid:96) is given by n (O u ) /n (O (cid:96) ) > × − .We can perform a simple calculation for the equilibrium arising from collisions withH atoms alone that includes both upward and downward transitions, governed bythe rate constants C (cid:96),u (H , T ) and C u,(cid:96) (H , T ), plus spontaneous radiative decays (ata rate A u,l = 8 . × − s − for the sum of the two most important transitions). Inthis simple treatment, we ignore cascades from even higher levels and the effects ofoptical pumping. Therefore, we state that in equilibrium n (O (cid:96) ) n (H) C (cid:96),u (H , T ) = n (O u )[ n (H) C u,(cid:96) (H , T ) + A u,(cid:96) ] (19)which leads to a solution for the density of neutral H atoms n (H) = [ n (O u ) /n (O (cid:96) )] A u,(cid:96) C (cid:96),u (H , T ) − [ n (O u ) /n (O (cid:96) )] C u,(cid:96) (H , T ) . (20)From results that will be presented later on the outcomes from N I , Fe II , and Ni II ,we adopt a temperature T = 8000 K. At this temperature, C (cid:96),u (H , T ) = 3 . × − cm s − and C u,(cid:96) (H , T ) = 1 . × − cm s − (Krems et al. 2006). If we adopta value for n (O u ) /n (O (cid:96) ) equal to its lower limit of 5 × − , the result that we findusing Eq. 20 is n (H) = 1 . × cm − . ircumstellar Gas Surrounding 51 Oph .
235 at T =8000 K (Zatsarinny & Tayal 2003). For the collision rate constants, we apply Eqs. 9and 10 to obtain C (cid:96),u ( e, T ) = 1 . × − cm s − and C u,(cid:96) ( e, T ) = 4 . × − cm s − and where g u = 5 but g (cid:96) = 9 applies to the combined degeneracy of all 3 fine-structurelevels of the ground P , , state. If we apply Eq. 20 with the substitution of e for H,we obtain n ( e ) = 2 . × cm − .In this approximate treatment, we find that for O I we have derived lower limits foreither n ( e ) or n (H) (because n (O u ) /n (O (cid:96) ) is a lower limit). It is apparent that for anelectron concentration x e = n ( e ) /n (H) = 1 . × − the electrons have an excitationcapability that is equivalent to those of the H atoms. It is important to note thatthis statement is not strongly dependent on whether or not n (O u ) /n (O (cid:96) ) is above thelower limit that we derived.Heavy-element atoms that have first-ionization potentials below that of H shouldbe predominantly ionized (an exception is Cl, which will be discussed in Section 7.3).After allowing for some depletions of these elements to form solids, we would expectthat, in the absence of any other source of electrons, x e should be about 1 . × − ,which is close to the value that would equalize the excitation rates of O I by electronsand neutral H atoms.It is probably more realistic to assert that the value of x e is much higher than1 . × − . There appears to be significant ionization of H caused by some combinationof radiation beyond the Lyman limit emitted by the star; soft X-rays emitted by gasin an accretion shock, for which there is evidence of one for 51 Oph (Mendigut´ıa et al.2011), and by cosmic rays (Padovani et al. 2018). Support for this idea comes fromthe FUSE spectrum of 51 Oph that shows very strong, saturated absorption featuresfrom the three fine-structure levels in the ground state of N II . (We are unable toderive column densities because the features are strongly saturated and the signal-to-noise ratio is very low at the wavelengths covering the 1085 ˚A multiplet of N II .)The existence of a significant amount of N II indicates that there are many additionalelectrons contributed by the ionization of hydrogen, because the ionizations of N andH are coupled to each other by charge exchange reactions (Lin et al. 2005). Wewill return to this topic in more detail in Section 10. Thus, in short, we can justifyinterpreting the excitation of the metastable level of O I in terms of just the electrondensity.For Fe II we will be invoking a more comprehensive approach for understanding theways to populate the many different metastable levels that we observed. To sensethe relative importance of H atom collisions compared to those from electrons, weevaluated for the lowest 10 excited levels the quantity C (cid:96),u (H , /C (cid:96),u ( e, . × − , a4 Jenkins & Gry numerical outcome that duplicates the result that we found in our simple treatmentfor populating an excited level of O I .The derivations of the collisional rate constants for N I by Amarsi & Barklem (2019)indicate that within the context of their analysis, the rates that involve the lowestfour excited levels are virtually zero because the level crossings occur at internuclearseparations that are too small and beyond the regime of their model. Other rateconstants across levels above these levels are many orders of magnitude below the C (cid:96),u ( e, T ) values that we used, except for a few levels adjacent to each other in energy.6.5.3. Excitations of N I , Fe II , and Ni II We now move on to the consideration of more detailed calculations for N I , Fe II ,and Ni II . For any one of these elements, our goal is to determine acceptable matchesof the observed level populations to the expected equilibria that are influenced byboth collisional exchanges and optical pumping. The collisions are governed by thelocal physical quantities n ( e ) and T , while the strength of the pumping to upper levelsis controlled by the distance from the star R g . In this particular context, we considerthe collective abundances of all levels as a nuisance parameter that we will need tomarginalize. In total, we must explore the goodness of fit between the data and ourcalculations over a range of 4 free parameters. We undertake this task by employinga Markov Chain Monte Carlo (MCMC) analysis with a Gibbs sampling. We assign arelative probability of any trial based on the probability of obtaining a worse fit to thedata for a χ distribution with the degrees of freedom set to the number of observedlevels (we do not subtract the number of free parameters because for any sample, weare not actively solving for a minimum value for χ ). We assumed uniform priorsover reasonable ranges for the logarithmic variables, but subject to two restrictions:(1) log T < . R g > − .
4, i.e., R g > R ∗ .We have organized our investigation so that levels of Fe II and Ni II are analyzedjointly, but the levels of N I are treated separately. The rationale for this approachis that Fe and Ni are refractory elements that probably share a common location,whereas N is a volatile element that may be distributed differently within the circum-stellar environment.Figure 4 shows examples where the conditions exhibit favorable matches betweenthe observed level populations and the predictions arising from the analysis discussedin Section 6.4. As will be clear from the results presented in the next section, theseconditions are not the only ones that give satisfactory fits to the data.6.6. Outcomes
Figure 5 illustrates the coverages of high probabilities in the log n ( e ), log T andlog R g parameter space, which are depicted by the densities of successful trials inthe MCMC runs. As expected, the outcomes for the most probable values of theseparameters for N I and Fe II +Ni II sometimes differ from each other, yet there is ircumstellar Gas Surrounding 51 Oph × × × × × × E/K (deg K)-12-10-8-6-4-202 l n ( N / g ) × × × × E (cm - ) × × × × × × E/K (deg K)-12-10-8-6-4-2 l n ( N / g ) × × × × E (cm - ) × × × × × × E/K (deg K)-10-8-6-4-20 l n ( N / g ) × × × × E (cm - ) N I log n(e) = 5.00log T = 3.90log R g = 0.60 c = 2.81 Fe II log n(e) = 6.00log T = 3.90log R g = 0.80 c =17.34 Ni II log n(e) = 6.00log T = 3.90log R g = 0.80 c =12.49 Figure 4.
The populations of various fine-structure and metastable levels of N I , Fe II ,and Ni II divided by their degeneracies, as a function of the excitation energies. Valuesof ln( N/g ) are relative and not absolute. The parameters shown in the panels representphysical conditions that give reasonably good matches between observations (black pointswith error bars) and theory (green triangles). The two pairs of excited levels with nearlythe same energy for N I are artificially separated from each other by a small amount in ahorizontal direction for clarity. The unreduced values of χ are indicated in each panel. Jenkins & Gry - ) . . . . l og n ( e ) ( c m - ) l og R g ( AU ) N I - ) . . . . l og n ( e ) ( c m - ) l og R g ( AU ) Fe II and Ni II
Figure 5.
Outputs from the MCMC analyses that indicate the relative probabilities of thefundamental parameters that govern the populations of the excited levels of N I , Fe II , andNi II . The white crosses indicate the locations of the parameter choices shown in Figure 4. ircumstellar Gas Surrounding 51 Oph I favors a temperature T = 8000 K (log T = 3 . II +Ni II are consistent with this result. Aside fromtemperature, the four metastable levels of N I constrain the possible combinations ofthe other conditions rather poorly. The results from Fe II +Ni II indicate rather wellthe distance to the star, R g = 6 AU (log R g = 10 . ), and we can rule out values of n ( e )that are significantly larger than 3 × cm − . For these two ions, arbitrarily low valuesfor the electron density n ( e ) and temperature T are possible, which indicates thatexcitations due to optical pumping dominate over those from collisions by electrons.From our interpretation of our lower limit for the relative population of O I in the D level that we derived in Section 6.5.2, we concluded that n ( e ) ≥ . × cm − ,which is consistent with the regions exhibiting high probabilities for the excitationsof N I , Fe II , and Ni II . OTHER NEUTRAL ATOMS AND THEIR IONIZATION EQUILIBRIA7.1.
Recombinations with Free Electrons
In previous sections, we discussed our findings on the circumstellar abundances ofthe neutral forms of two elements, O and N. These elements are plentiful, and thelogarithm of the ratio of their abundances log N (O I) − log N (N I) = 17 . − .
60 =1 .
03, which is not far from the solar abundance ratio 0.86 (Asplund et al. 2009). Forthe neutral forms of species that should be predominantly ionized, a very differentpicture emerges, as we discuss below.We now explore the implications for electron densities by comparing the columndensities N of elements X in their neutral states N ( X ) to those of their singly-ionizedstates N ( X + ). The abundances of these two forms are influenced by an equilibriumbetween the starlight photoionization of the neutral atoms at a rate Γ balanced againstthe sum of the radiative recombination (rr) and dielectronic recombination (dr) of theions with free electrons with a combined rate constant α rr+dr at 8000 K.We consider the formula for the ionization equilibrium for the element X , n ( e ) = Γ N ( X ) α rr+dr N ( X + ) , (21)where we have substituted observed column densities N for the local densities n andset Γ = 4 Whc − (cid:90) hc/E (IP)920 ˚ A σ λ λF λ dλ . (22)The dilution factor that we defined in Eq. 1 W = ( R ∗ /R g ) / R g (cid:29) R ∗ . For R g = 6 AU, W = 6 . × − . This value is considerably larger than the interstellar The difference from the solar abundance ratio decreases a bit when we consider that log(N / O) =log(N / O ) + 0 .
07, as we will show later in Section 10. Jenkins & Gry
UV and visible radiation field at our location in the Galaxy, which is described interms of W = 10 − for a blackbody spectrum with T = 7500 K added to W = 10 − for T = 4000 K (Mathis et al. 1983) and even lower effective values of W for veryhot stars (Mezger et al. 1982). Hence, we can ignore the effects of the generalradiation emitted by stars elsewhere. The upper limit for the integral hc/E (IP) isthe wavelength of the ionization edge of the element in question (1 . × /E (IP) if E (IP) is expressed in eV), and σ λ is its ionization cross section. The stellar flux atthe surface of the star F λ is expressed in the form erg cm − s − ˚A − , which we definedin Section 6.1.1, and the factor 10 − in the equation converts from ˚A to cm. Thelower integration limit of 920 ˚A corresponds to the cutoff in the stellar flux caused byhigh members of the Lyman series and Lyman limit continuum absorptions.Numbers that are relevant for evaluating n ( e ) based on the ratios of neutrals toions of various elements are listed in Table 3. Values of Γ listed in Column (2) of thetable are based on a radial distance from the star R g = 6 AU (i.e., 10 . , which wasfavored by the Fe II +Ni II metastable excitations; we recall that the stellar radius R ∗ = 0 . T = 8000 K, indicated by the N I excitations, wefind the effects of such collisions to be very small compared to Γ. For instance, we listthe rate constants (cid:104) σv (cid:105) for such collisions in Column (3) of the table; (cid:104) σv (cid:105) n ( e ) (cid:28) Γfor the largest value of n ( e ) indicated by the metastable excitations. At T = 8000 K,the dielectronic recombinations are substantial for the elements Mg, Ca, and Zn, andthe inferred values for n ( e ) change rapidly as a function of temperature for Mg andCa. The quantity α g in Column (5) will be discussed in Section 7.2.We start by comparing the abundances of neutral and singly-ionized carbon. Forthe former, we stated in Table 2 a measurement log N (C I) = 12 .
33, but we concludethat the origin of this C I absorption is interstellar based on two arguments: first,it has a velocity of − . − with respect to the other strongest circumstellarfeatures (see Section 4), and second, we see no evidence of absorption out of either ofthe two excited fine-structure levels. In a very dense and hot gas, the populations ofall three fine-structure levels should be in proportion to their degeneracies. Thus, inour attempt to identify C I associated with the circumstellar gas, it is best to try tomeasure features out of the P level at 43 cm − , which has a degeneracy equal to 5(compared to 3 and 1 for the other two levels). After considering relative signal-to-noise ratios and the groupings of lines within the different multiplets, we chose to usethe two lines listed in Table 7 for this level. The absorptions cannot be seen in thespectrum, so we derived an upper limit equal to 1 . × cm − using the methoddescribed in Section 3.2. We then multiply this upper limit by 9/5 to account for C I in the other two levels to arrive at a total upper limit N (C I ) < . × cm − . Our determination is considerably lower than that of Lecavelier Des Etangs et al. (1997a) of log N =13 .
08 based on a GHRS G160M spectrum taken at a different time – see Table 1). ircumstellar Gas Surrounding 51 Oph Table 3.
Electron Densities from Ionization Equilibria at 8000 KElem. Γ (6 AU) (cid:104) σv (cid:105) α rr+dr α g a N ( X ) N ( X + ) n ( e ) b X (s − ) (cm s − ) (cm s − ) (cm s − ) (cm − ) (cm − ) (cm − )(1) (2) (3) (4) (5) (6) (7) (8)C 3.4e-03 6.7e-16 9.1e-13 1.9e-16 < c < d Na 1.4e-02 4.5e-11 1.8e-13 3.3e-18 3.0e+10 6.5e+14 c < c < < < · · · < < < c < · · · < > e Ca 1.1e+00 1.3e-11 6.2e-12 7.8e-18 < < < < · · · · · · References — σ λ used for computing Γ: C (Verner et al. 1996); Si, Fe: NORAD-Atomic-Data website, https://norad.astronomy.osu.edu/; Mg: (Wang et al. 2010); S (Bautista etal. 1998); Ca: (Verner et al. 1996); Na, P, Cl, Zn: (Verner & Yakovlev 1995), Collisionalionization from electron impacts (cid:104) σv (cid:105) : (Voronov 1997), Recombination α rr+dr : C, Na,Mg: (Badnell 2006); Si: (Nahar 2000); P, Cl: (Landini & Monsignori Fossi 1991); S, Ca:(Shull & Van Steenberg 1982); Fe: (Nahar et al. 1997), Zn: (Mazzitelli & Mattioli 2002),Grain recombination α g for all elements where a value can be calculated: (Weingartner& Draine 2001b). a Derived using ψ = 8 . × , T = 8000 K, and the use of the fitting equation, Eq. 8 ofWeingartner & Draine (2001b) for the element in question. See the text for details. b Solutions to Eq. 21. c This value is not measured; instead, we had to estimate it by relating it to O and N andusing a solar abundance ratio (Asplund et al. 2009). If this element is depleted relativeto O and N, the value or upper limit for n ( e ) will be higher. d The abundance of C II is probably lower than the value given in Column 7. For thisreason, the upper limit shown here is too low. See Section 11. e This outcome is misleading for reasons outlined in Section 7.3.
For C II , we must make a preliminary estimate for its column density because ourdetermination log N (C II) < .
21 from a semiforbidden transition is quite high (theallowed transition at 1334.5 ˚A is very badly saturated and not useful). If we assumethat the ratio of C to O or N in the circumstellar gas is identical to the solar ratio(Asplund et al. 2009), we arrive at an approximate value log N (C II) ≈ .
37. Ata radial distance from the star R g = 6 AU, our determination of Γ in Eq. 22 gives avalue 3 . × − s − when we use the ionization cross sections provided by Verner et al.(1996). For carbon ions, α rr+dr = 9 . × − cm s − at T = 8000 K (Badnell 2006).An application of Eq. 21 to our upper limit for N (C I ) yields n ( e ) < . × cm − .This limit is below the example for n ( e ) that we chose to represent for the excitation0 Jenkins & Gry of N I in Fig. 4, but setting n ( e ) equal to the limit does not significantly degrade thematch with the data and is consistent with our lower limit n ( e ) ≥ . × cm − thatwe derived from the population of the O I metastable level in Section 6.5.2.Our calculation for Γ only considered direct photoionization by the flux from thestar. From the presence of N II we know that ionized hydrogen must be abundant. Wewill argue later (Section 12.4) that H must be ionized by photons that are far moreenergetic than this atom’s ionization potential. When this happens, the electronsthat are released can produce secondary ionizations of other elements, including C.Because we did not include secondary ionizations in our calculations, it is possiblethat the outcome of our calculation of the upper limit for n ( e ) is too low because weunderestimated the true total value of Γ.In principle, another route for ionizing the carbon atoms is through the chargeexchange reaction C + H + → C + + H . However, this process is quantitativelyunimportant since even for a very large proton density, say perhaps 10 cm − , theeffect of this reaction is 7 orders of magnitude weaker than the photoionization ratebecause its rate constant is only 1 . × − cm s − at T ≈ K (Butler & Dalgarno1980). (In contrast to carbon, we find that charge exchange reactions that involvechlorine are important, as we will show in Section 7.3).We summarize in Column (8) of Table 3 the values of n ( e ) that we derived for C andvarious other elements using Eq. 21. For Fe I the combined degeneracies of the otherfour fine-structure levels of this atom (cid:80) g = 16 compared to g = 9 of the lowestlevel could mean that our upper limit for all of the Fe I should be increased by afactor of 25/9 to 3 . × cm − if the excitation temperature is large. Our limit for N (Fe I ) is considerably lower than the amount that appears in the spectrum of β Pic(Vidal-Madjar et al. 2017 ; Kiefer et al. 2019). The column density for Fe II includesall of the levels that we observed plus reasonable estimates for the populations ofintermediate levels that we did not observe (to be discussed later in Section 9).While most of the upper limits for n ( e ) are consistent with our derivation basedon C, there are two major inconsistencies with the C result. The most prominentdisagreement is the outcome of a lower limit for n ( e ) derived from chlorine. In Sec-tion 7.3 we will explain why we can disregard this result. The outcomes for Na andZn are substantially higher than that for C, which probably indicates that the actual N (C II ) is lower than our assumed value. We will discuss the possibility that neutralcarbon atoms might be expelled by an outward radiation pressure arising from thestellar flux in Section 11. That discussion will rely on conclusions derived here onthe ionization and recombination of carbon atoms and ions and how these processesshould affect any outward migration of the atoms. ircumstellar Gas Surrounding 51 Oph Interactions with Dust Grains
Up to now, we have not considered recombinations of the ions with electrons on thesurfaces of dust grains (Snow 1975 ; Weingartner & Draine 2001b). The efficiency ofthis process depends on the local density of grains and their charge. Our discussionwill proceed within the framework of the dust-to-gas ratio being identical to that ofthe general ISM. However, we note that in reality it may be lower in the circumstellaratomic gas region that we sampled – at least for grains in the size range that isimportant for reddening. The color excess E( B − V ) for 51 Oph has been reportedto be 0.04 by Malfait et al. (1998) and 0 . ± .
06 by Manoj et al. (2006). Usingour measured column densities of N I and O I and assuming that their depletionsare consistent with a moderate density ISM (i.e., F ∗ ≈ . N (H) = 1 . × cm − , as wewill explain in more detail in Section 10. This value is greater than what we wouldhave expected from the general result of Diplas & Savage (1994) for the ISM, where (cid:104) N (H I) / E( B − V ) (cid:105) = 5 × cm − mag − , which would yield N (H I) ≈ . ± × cm − for the two determinations of E( B − V ). This disparity, which is made evenworse by the presence of ionized hydrogen, may indicate that some fraction of thegrains have been driven out of the circumstellar atomic gas by an outward radiationpressure. Nevertheless, in the discussion that follows, we will frame our argumentswithin a perspective that the relative proportion of grains in the gas and their sizedistribution are the same as in the general ISM.The charge of the grains is regulated by a balance between the loss of electrons dueto photoelectric emission, which is governed by the local radiation density, and thecapture of free electrons, which depends on n ( e ) (Weingartner & Draine 2001a). In aformalism developed by Weingartner & Draine (2001b) for the neutralization of ionsby grains in the general ISM, a rate coefficient α g ( ψ [ G, T, n ( e )] , T ) is normalized tothe local hydrogen density n (H) and depends on the parameter ψ = G √ T /n ( e ) (23)and T , where G is the radiation density relative to that in the ISM as defined byMathis et al. (1983). The flux of 51 Oph at its surface corresponds to G ≈ for photons with λ < R g vs. log n ( e ) for N I inFigure 5, the right-hand edge of the diagonal, red-colored region showing the mostprobable combinations corresponds to n ( e ) = 10 R − g , where n ( e ) is expressed in cm − and R g in AU. Here, we have chosen the highest set of values for n ( e ) to explore themost favorable condition for promoting the neutralization of ions by their interactionswith grains. Because G ≈ ( R ∗ /R g ) , we obtain the simple outcome for the lowestvalue ψ = 8 . × (i.e., one that is the most conservative from the standpoint ofgiving the highest values for α g ).We do not have a firm number for a representative value of n (H), but if we assumethat the column density N (H I) ≈ cm − (from our measurements of N I and2 Jenkins & Gry O I ) and the thickness of the zone is no smaller than about 0 . R g , we obtain theestimate n (H I) = 6 . × cm − R g . When we combine this limit with our condition n ( e ) < R − g , we obtain an outcome for the electron fraction x e = n ( e ) /n (H) < / (6 . R g + 1), which equals 6 . × − if R g = 6 AU. As long as x e is not less than α g /α rr+dr , we can declare that recombinations onto grains are less important thanrecombinations with free electrons. (Even if x e is smaller than this example, oneshould recall that x e (cid:29) . × − from the discussion in Section 6.5.2.) The resultsshown in Table 3 indicate that for all of the elements, α g (cid:28) x e α rr+dr , so we can ignorethe influence of grains in the ionization equilibrium.7.3. An Interpretation of the Large Abundance of Neutral Chlorine
The ionization balance for chlorine is vastly different from the examples that appliedto the other elements listed in Table 3. We find a very strong presence of Cl I (log N = 13 .
67 for the sum of the two fine-structure levels that we observed), butwe were unable to detect an absorption feature out of the singly-ionized form of Cl.The outcome shown in the table indicates that the lower limit for n ( e ) is stronglyinconsistent with the values or limits derived for C, Na, Mg, S, and Zn. The onlyways to reduce the inferred value of n ( e ) are either to lower the temperature towell below 10 . K (thus increasing α rr+dr ( T )) or to consider that R g (cid:29) I excitations and the latter is ruled out by the Fe II +Ni II excitations.We now reach a point where we must consider alternate means for neutralizing Cl.One of them might be the reaction Cl + + H → HCl + + H followed by the dissociativerecombination with electrons HCl + + e → H+Cl to form neutral hydrogen and chlorine(Jura 1974 ; Smith et al. 1980). The possible importance of this channel is supportedby observations of the ISM that show a strong correlation between N (Cl I ) and N (H )and evidence that in a number of cases N (Cl I) > N (Cl II) (Moomey et al. 2011).However, we regard this process as unimportant for the material surrounding 51 Oph,because Roberge et al. (2002) found that N (H ) < × cm − in the J = 0 and J = 1 states in the ground vibrational level, which is far lower than typical amountsof H that were found in ISM sight lines that showed enhancements of Cl I .Still another pathway for neutralizing Cl ions is through charge exchange withneutral hydrogen, Cl + +H → Cl +H + . The reaction is endothermic, but not stronglyso because the first-ionization potential of Cl is 12.97 eV, which is only slightly belowthat of hydrogen. The rate constant for this reaction is C + , = 6 . × − (cid:18) T (cid:19) . exp( − /T ) cm s − (24)and the reverse reaction rate is C , + = 9 . × − (cid:18) T (cid:19) . exp( − /T ) cm s − (25) ircumstellar Gas Surrounding 51 Oph T = 8000 K, C + , = 3 . × − cm s − and C , + =9 . × − cm s − . We now examine the equilibrium of the two chlorine ionizationstates that takes into account these reactions, along with recombinations with freeelectrons and starlight photoionizations (at rates specified in Table 3): n (Cl + )[ n ( e ) α rr+dr + n (H ) C + , ] = n (Cl )[Γ(Cl) + n (H + ) C , + ] . (26)When solving this equation for n (H ) we find that the n ( e ) α rr+dr term is insignificantand obtain the numerical result: n (H ) = 2 . × − + 9 . × − n (H + )3 . × − [ n (Cl + ) /n (Cl )] cm − . (27)Because our observation indicates that n (Cl + ) /n (Cl ) ≤
1, the above expression givesa lower limit: n (H ) ≥ . × cm − + 2 . n (H + ) . (28)This outcome demonstrates that even with the large electron densities indicated bythe metastable level populations, the neutral fraction of hydrogen is significant, whichis a conclusion that is consistent with our observed abundances of N I and O I . Thefact that we could not obtain a column density for N II and compare it with N I meansthat our knowledge on the neutral versus ionized hydrogen abundances is limited toour finding expressed here and the less definitive estimate n (H ) < × cm − at R g = 6 AU derived in Section 7.2. THE ALLOWED TRANSITIONS OF O I TO THE S LEVELFigure 6 shows the spectrum of 51 Oph over a wavelength interval that covers thevery strong features from three fine-structure levels of the ground electronic stateof O I . The levels P and P have excitation energies equal to 158 and 227 cm − ,respectively, above the P state. The absorption features from the excited levelsof O I have cores that do not reach zero intensity, unlike the absorptions from theground states of both O I and Si II . The P / excited fine-structure level of Si II alsodoes not reach zero intensity, but it is closer to zero than the features of excited O I .While one might envision that the profiles do not reach zero intensity because theabsorbing medium does not completely cover the disk of star seen in projection,we propose that a more likely explanation is that some of the flux from the star isresonantly scattered by oxygen atoms in the far side of the circumstellar disk. Ifsome of this emission were able to bypass the foreground absorption in the near sideof the disk because it was somewhat displaced in projection, it would create excessflux contributions which fill in positive intensities superposed on the two saturatedline cores. Portions of the emitting gas that subtend less than 20 R AU should We can dismiss the notion that the emission could be caused by telluric O I emission, because if thiswere so, we would find the same effect in the unexcited O I line at 1302 ˚A. Jenkins & Gry be admitted through the 0 . (cid:48)(cid:48) × . (cid:48)(cid:48) R = 1 . ± .
04 (Arenou et al. 2018 ; Luri et al. 2018) is the distance of the starfrom us in units of 100 pc. This emission effect is not seen in the feature from the P unexcited state because it is blocked by foreground O I in the interstellar medium,whereas the interstellar absorption from the two excited levels is extremely weak ornonexistent.A plausible configuration for our being able to view the emission from resonantscattering is depicted in Figure 7. If the disk has a wedge-shaped cross section andthe vertex falls short of the star, both sides can be illuminated by radiation from the O I( P ) Si II( P )O I( P ) O I( P ) Si II( P )1300 1302 1304 1306 1308 1310 1312Wavelength (Å)-100001000200030004000 I n t e n s it y ( - e r g c m - s - Å - ) Figure 6.
A segment of the spectrum of 51 Oph covering a wavelength interval that includesfeatures from O I arising from three different fine-structure states, along with those of Si II from two fine-structure levels. Figure 7.
A schematic illustration of a condition that can create the emission that issuperposed on the cores of the absorption lines of excited O I . Light passing directly fromthe central star to the observer (red arrow) is intercepted by atoms in the disk and thusresults in an absorption feature. The disk has a wedge-shaped cross section that permitslight from the star to illuminate the edges of the disk. Photons that are within the passbandsof the O I features are resonantly scattered and can then reach the observer (yellow arrows)if the tilt of the disk is sufficient to prevent this light from being intercepted by the outerportion of the disk. ircumstellar Gas Surrounding 51 Oph velocities (km s - ) ® relative tov ô = -17 km s - Si II
O I ( P ) O I ( P ) N ( P ) = ( 1.39 ´ , 1.41 ´ ) cm - N ( P ) = ( 4.64 ´ , 4.69 ´ ) cm - v ô = (-16.2,-17.8) km s - b = ( 3.45, 8.50) km s - C on ti nuu m no r m a li ze d i n t e n s it y -100 -50 0 50 100 -100 -50 0 50 100 Figure 8.
Portion of the spectrum of 51 Oph that shows the two features of O I inthe excited fine-structure levels. The thick, black trace shows the intensities that havebeen normalized to our best reconstruction of the stellar continuum. The wavelength andvelocity scales shown in this plot have been shifted to a reference frame of −
17 km s − . Forcomparison, the two green traces show the shape of the absorption out of the unexcitedlevel but displaced by the differences in transition wavelengths. The ISM absorption in theunexcited level makes its profile broader on the blue wing. The absorption feature in thethree O I lines at +26.9 km s − is telluric. The red trace shows our reconstruction of theabsorption features according to our model discussed in the text, the parameters for whichare shown below the zero level. The blue trace shows the difference between the observedintensities and this absorption model, which we interpret as the resonantly scattered lightfrom the disk whose radial velocities span ±
20 km s − on either side of the line center. star. We know from the nondetection of CO discussed in Section 5 that we are notviewing the disk edge on. If the tilt of the disk with respect to the observer is sufficient,light scattered by atoms in the rear portion could bypass gas in the foreground partthat would otherwise absorb it. Under the right conditions, other geometries thatcould also create conditions for this phenomenon might include circumstellar rings orwarped disks.Figure 8 shows the spectrum in the vicinity of the excited lines after the intensitieshave been normalized to the stellar continuum (black trace). To estimate the strengthsand shapes of the emission profiles, we must first estimate what the absorption wouldlook like in the absence of emission. To recreate this absorption, we used our value for N (O I ) in the P level that we derived from the weak intersystem line at 1355.598 ˚A,as listed in Table 2, and assumed that each fine-structure level was populated inproportion to its degeneracy, which is approximately valid if the medium is denseand has an excitation temperature is of order 8000 K. Next, we used the velocityparameters for our neutral gas absorptions that we described in Section 4 to constructthe expected profiles. The red trace in the plot shows the absorption profiles thatshould arise from this model. The effect of smoothing by the STIS line spread functionis very small for such wide absorption features, but it has been included. As a check6 Jenkins & Gry on the validity of this reconstruction, we require that the two profiles do not extendbeyond the velocity span of the absorption from the unexcited P level, which shouldbe stronger because it has a higher degeneracy than either of the two excited levels.The green traces show the absorption by the unexcited level superposed on the twoexcited level absorptions. In both cases, the red traces fall inside or on the greenones.The blue traces in the figure show the differences between the observed intensitiesand our reconstruction of what the absorption features should look like in the absenceof any emission. The broadening of the emission should arise from the Keplerianmotion of gas around the star, and the increased levels of intensity away from theline center are probably created by limb brightening along the tangent points. Bothof the emission profiles reach zero intensity at about 20 km s − away from the linecenter, and this velocity offset is consistent with the expected Keplerian velocity v K = (cid:112) GM ∗ /R g = 22 km s − at R g = 6 AU from the star, where the star has amass equal to M ∗ = 3 . M (cid:12) (Jamialahmadi et al. 2015). The intensity dip seen onthe right-hand side of the P profile is caused by telluric O I absorption, which wepredict to occur at a heliocentric velocity of +9.9 km s − at the time the observationwas carried out (or +26.9 km s − for the scale shown in Fig. 8). If the gas wereoptically thin, we would expect the P emission intensity to be three times as strongas that from the P state, but here it appears to be only about twice as strong. Highoptical depths would tend to equalize the two intensities, but at some expense to thelimb-brightening effect.We touch briefly on the possibility that some of the emission might arise fromBowen fluorescence (Bowen 1947) produced by optical pumping by L β irradiation tothe levels in the 2 p d D o , , states, which then decay to produce radiation at 8446˚A.Eventually, there is a cascade to the 2 s p s S state that produces 1304 and 1306˚Aphotons as it decays to the ground fine-structure levels. Broad H α emission arisingfrom a shock in an accretion flow has been observed for 51 Oph (Manoj et al. 2006; Mendigut´ıa et al. 2011). If the medium is optically thin, the predicted value of I H α /Iλ P feature, the average strength of the emissionover the 40 km s − span centered on the systemic velocity ( v (cid:12) = −
17 km s − ) is 0.10times the continuum intensity over the same interval. We compute that the predictionfor the fluorescence would result in about one order of magnitude less flux than whatwe observe, but a higher intensity could result if the region is optically thick or thereis a sufficient flux of L β photons from the star. THE PATTERN OF ELEMENT ABUNDANCESFigure 9 is a representation of ln(
N/g ) vs. excitation energy for the key elementscovered in this study. For many of the elements, there are levels whose absorptions wedid not observe and yet would be expected to be populated. For the elements Cr, Mn, ircumstellar Gas Surrounding 51 Oph
51 Oph × × × × E/k (deg K)253035 l n ( N / g ) L og ( N / g ) × × × × × E (cm - ) O IN ISi IIFe IIP IINi IICu IICr IIMn IIZn IICo IITi II= observed levels= unobserved levels= calc level for n(e) » cm - Figure 9.
A plot that shows all of the column densities of elements in their preferredionization stages for all of the levels that we observed, expressed in the representationln(
N/g ), where N is the column density and g is the degeneracy of the level, as a functionof excitation energy above the ground state. Levels that had measured column densitiesare depicted with diamond symbols, while our estimates for unobserved states are shownby plus markers. See the text for details on the excited level of O that is shown by a circle.Some markers have been moved in a horizontal direction by a small amount to preventoverlaps with other ones. Fe, Ni, and Cu, we estimated the locations for the unobserved levels in this diagramby placing them on a straight-line best fit to the measured levels. Unseen levels thatwere considered to have populations worthy of note were those that had excitationenergies below 80 ,
000 cm − and that did not have electric dipole or intercombinationlines to lower levels that would cause rapid radiative decays. The elements N, O, Si,and Zn did not have any such unseen levels. For the elements P, Ti, and Co, theobservations were so incomplete that we had to adopt a representative slope, and we8 Jenkins & Gry
Table 4.
Observations of the ElementsLog Total Fraction ExcitationElement Column Density Observed Temperature (K)(1) (2) (3) (4)N 16.60 1.00 5180O 17.63 0.98 a · · · Si 15.62 1.00 890 b P 13.61 0.85 450 b Ti 12.24 0.05 · · ·
Cr 13.49 0.61 11,990Mn 13.27 0.78 9120Fe 15.15 0.94 3090 b , 11,840 c Co 13.28 0.27 · · ·
Ni 13.81 0.99 10,210Cu 12.49 0.88 9700Zn 13.61 1.00 · · · a We could measure only the ground P level, but the othertwo fine-structure levels should be populated in proportionto their degeneracies. While we observed an absorption outof the metastable level at an excitation of 15868 cm − wecould only obtain a lower limit for its column density. b Fine-structure levels only. c Metastable levels only. chose one that was equal to the mean of the Fe II and Ni II fits. The situation forO is a special one; aside from the fine-structure levels, the only noteworthy excitedlevel is the one at 15 ,
868 cm − that we know has an appreciable column density, butfor which we could derive only a lower limit to its value. The circle symbol in thediagram corresponds to our expectation for n ( e ) = 10 cm − and T = 8000 K.We may be mildly understating the abundances of nitrogen and oxygen, as somefraction of these elements may be singly ionized, i.e., a condition similar to that ofhydrogen. We know from the FUSE observations discussed in Section 6.5.2 that this isthe case for nitrogen, but we are not able to quantify the relative ionization. Later, inTable 5 within Section 10, we will show our estimates for the ratios of ions to neutralsfor these elements. The other elements probably have no appreciable concentrationsin their doubly ionized form, but we have no firm evidence that this is so.Table 4 shows some outcomes of our measurements of the elements. Column (2)lists the sum of observed and unobserved column densities, while Column (3) showsthe fraction of this combined column density that was actually observed, which canbe used as a measure of the reliability of the total column density. The excitation ircumstellar Gas Surrounding 51 Oph N O Si P Ti Cr Mn Fe Co Ni Cu Zn -1 12 L og A bund a n ce Figure 10.
Total column densities of the preferred ionization stages of the circumstellarelements associated with 51 Oph. . temperatures given in Column (4) are derived from the negative inverses of the slopesof the level populations shown in Fig. 9. From the conclusions that we derived in Sec-tion 6.6, it should be clear that these excitation temperatures should not be regardedas actual physical temperatures. We list these temperatures only for the purposeof empirical comparisons with results for the circumstellar populations around otherstars. In particular, we caution against using the populations of the fine-structurelevels of Si II and P II as an alternate means of determining n ( e ) and n (H ) becausethe uncertainties in the column densities are large enough to allow their excitationtemperatures to be equal to the kinetic temperature. For instance, the critical densityfor the excitation of the Si II levels is as low as n ( e ) ≈ cm − (Silva & Viegas 2002,Fig. 8).We present the pattern of the total logarithmic column densities in Figure 10.Our next step will be to explore some possible interpretations of these abundances.Figure 11 presents a number of choices. In panel ( a ) of this figure, we show a repre-sentation of our total logarithmic column densities listed in Table 4 after subtractingthe logarithm of the solar photospheric abundances (on a scale H = 12) (Asplund etal. 2009) to test the proposition that the gas that we measured is dominated eitherby mass loss from the star or by interstellar material that has virtually no depletionsof atoms onto dust grains. The lack of uniformity in the bar heights indicates thatthis is not a favorable interpretation. Next, we apply the same test for gas that hasa composition that is similar to that of the general ISM with three possible choicesfor the severity of atomic depletions. These choices are represented by the expecteddepletions for the parameter F ∗ (Jenkins 2009), which ranged from values of 0.0 (i.e.,very light depletions) in Panel ( b ), 0.5 (moderate depletions) in Panel ( c ), and 1.00 Jenkins & Gry
N O Si P Ti Cr Mn Fe Co Ni Cu ZnLog (N) - Log(solar photospheric) (a) -1 12
N O Si P Ti Cr Mn Fe Ni Cu ZnLog (N) - Log(solar photospheric) - depl(F * = 0.0) (b) -1 11 N O Si P Ti Cr Mn Fe Ni Cu ZnLog (N) - Log(solar photospheric) - depl(F * = 0.5) (c) -1 11 N O Si P Ti Cr Mn Fe Ni Cu ZnLog (N) - Log(solar photospheric) - depl(F * = 1.0) (d) -1 11 N O Si P Ti Cr Mn Fe Ni Cu ZnLog (N) - Log(dust abund. for F * = 0.0) (e) -1 11 N O Si P Ti Cr Mn Fe Ni Cu ZnLog (N) - Log(dust abund. for F * = 1.0) (f) -1 11 O Si P Cr Mn Fe Co Ni Cu ZnLog (N) - Log(Earth crust abund.)N ä Ti ¯ (g) -1 12 (h) Log (N) - Log(comet abund.)N O Si Ti Cr Mn Fe Co Ni Cu Zn
Figure 11.
Logarithms of our total abundance measurements given in Table 4 after sub-tracting the logarithms of various choices for constituents in the circumstellar gas associatedwith 51 Oph. Panel ( a ) depicts a subtraction of pristine solar abundance gas. Panels ( b − d )represent subtractions for depleted gas-phase ISM material with three levels of severity ofsuch depletions. Panels ( e and f ) represent various choices for the conversions of ISM dustsolids into gaseous form, as described in the text. Panel ( g ) represents the subtraction of amixture of elements in the Earth’s crust, and Panel ( h ) does the same for samples of dustfrom three different comets. In Panel ( h ), we have adjusted the overall abundances for eachof the three choices so that they match each other for Fe. ircumstellar Gas Surrounding 51 Oph d ). The moderate depletions ( F ∗ = 0 .
5) seem to showthe best uniformity in the bar heights.The lower four panels of Fig. 11 test the proposition that the original gas that hascondensed out of the ISM to form the circumstellar disk has been expelled and beenreplaced by atomic constituents arising from different possible forms of solid materialthat have been converted into gaseous form by collisions or ablation. Recall that fromour finding stated in Section 5 about the absence of CO features in our spectrum, weconcluded that we must be sampling gas at some distance from the midplane of thedisk. This gas might have been subject to erosion by photoevaporation or a stellarwind and replaced by elements arising from destructive processes for solids in themidplane. Panels ( e ) and ( f ) represent our abundances after subtracting the patternof abundances of ISM dust for F ∗ = 0 and 1, respectively. The bar heights are quiteuneven in these panels. The same kind of presentation is presented for Earth’s crustabundances (Anderson 1989 p. 150), which could stand for material ejected fromthe outer portions of chemically differentiated planets by collisions (Panel ( g )). Wehave also tested the pattern that could emerge from the destruction of objects thathave chemical makeups similar to those of CI chondritic meteorites (Lodders 2003).However, except for volatile elements, the abundances in such primitive meteoritesclosely follow the solar photospheric ones, so we would simply be duplicating whatwe see in Panel ( a ), except for the elements N and O. Finally, in Panel ( h ) we in-vestigate the relative differences between our abundances and those found for threedifferent determinations of some relative abundances for cometary dust: (1) dustfrom Comet 81P/Wild 2 collected in the aerogel on the Stardust mission (Flynn etal. 2006), (2) material observed with the impact-ionization time-of-flight mass spec-trometer in the PUMA-1 experiment on the
VEGA-1 spacecraft, which encounteredComet 1P/Halley (Jessberger et al. 1988), and (3) the sampling of particles fromComet 67P/Churyumov-Gerasimenko by the COmetary Secondary Ion Mass Ana-lyzer (COSIMA) on the
Rosetta mission (Bardyn et al. 2017). Comets 81P/Wild 2and 67P/Churyumov-Gerasimenko are Jupiter-family comets (with a low orbital incli-nation and perturbed by Jupiter into a short period), and 1P/Halley is a long-periodcomet with a high inclination.From the choices that we presented in Fig. 11, we conclude that ISM-type gasthat has undergone moderate levels of depletion gives a satisfactory match to theabundance pattern that we observed. Over a limited range of refractory elementsthat were reported, the results for 81P/Wild 2 and 67P/Churyumov-Gerasimenkoalso seem to show a reasonably good uniformity of bar heights, but for Halley and67P our abundances of N and O seem too large. The ISM dust solids and the Earth’s The depletions of P and Zn have been modified from the original description by Jenkins (2009) toaccount for the revised f -values for transitions of P by Brown et al. (2018) and Zn by Kisielius etal. (2015) that replace earlier ones reported by Morton (2003). Jenkins & Gry crust material outcomes show significantly uneven bar heights in their respectivepanels.
ESTIMATES FOR THE COLUMN DENSITIES OF H I AND H IIEven though we cannot directly observe hydrogen in its neutral or ionized forms, wecan make use of the good match between the abundance pattern of different elementsto that of a mildly depleted gas-phase ISM to propose that the volatile elements Nand O can be used as proxies for H after correcting for abundance differences. Againreferring to the ISM, we conclude that the circumstellar N and O should be depletedby about 0.1 dex below their protosolar abundances. From the evidence presentedin Section 6.6, we adopt the viewpoint that the neutral hydrogen atoms and protonsare distributed in a common volume, rather than being in separate locations.The ionization fractions of N and O are coupled to that of H by charge exchangereactions with large rate constants. As we will discuss in Section 12, the ionizationlevels for all three elements must be maintained by photoionizations caused by ener-getic radiation in excess of the flux from the star at wavelengths below the Lymanlimit (and perhaps supplemented to some extent by cosmic-ray ionization).In order to have a good understanding about the strength of the coupling of therelative ionizations of nitrogen and oxygen to that of hydrogen, we must know thevolume densities n (H ) and n (H + ) ( ≈ n ( e )). Our constraints on these quantitiescover some broad ranges in values. From the metastable excitations and the ratiosof some neutral atoms to their ionized counterparts, we feel that it is reasonable toadopt 10 < n ( e ) < × cm − . For neutral hydrogen, we described in Section 7.3how the abundance of neutral chlorine compared to an upper limit to its ionized formled to an expression given Eq. 28, but we must remember that the specified value for n (H ) is a lower limit. We have no direct way to constrain how far above this limitthe density could go, other than to argue that if it were to reach ∼ cm − , ourdetermination of the column density N (H ) (which we will derive shortly) divided by n (H ) would shrink to a thickness 0 . R g (i.e., 0.6 AU), which we can deem to be areasonable lower limit.To determine the ionization balance of nitrogen, we can use the charge exchangerate constants C + , = 1 . × − cm s − and C , + = 5 . × − cm s − de-rived by Lin et al. (2005) for T = 8000 K. The recombination coefficient for ni-trogen is α = 1 . × − cm s − at T = 8000 K (Shull & Van Steenberg 1982 ;Nussbaumer & Storey 1983). The value of Γ(N ) is about ten times the value ofΓ(H ) = α (2) (H) n ( e ) /n (H ) after one considers additional secondary ionization pro-cesses for both elements (Jenkins 2013). Because the region is optically thick toLyman limit photons, we adopt a recombination rate to levels n = 2 and higher α (2) (H) = 3 . × − cm s − at T = 8000 K. We now can substitute N for Cl inEq. 26 and solve for n (N + ) /n (N ) to obtain the estimates listed in Table 5 and com-pare them to their hydrogen counterparts n (H + ) /n (H ), also listed in the table. We ircumstellar Gas Surrounding 51 Oph Y (N) = log (cid:20) n (N + ) /n (N )1 + n (H + ) /n (H ) (cid:21) (29)From our observation of N I , we evaluate our estimate for log N (H I) using therelation log N (H I) = log N (N I) − log(N / H) (cid:12) + 0 . Y (N) = 20 . +0 . − . (30)where we use our determination of log N (N I) listed in Table 4, and the protosolarabundance log(N / H) (cid:12) = 7 . −
12 (Asplund et al. 2009). The term 0.1 in theequation is a correction that applies to our estimate of the depletion of N. The statedvalue given in Eq. 30 is for Y (N) = 0 .
10, which recognizes that the FUSE spectrumexhibited strong absorptions of the N II multiplet near 1085 ˚A, as we had indicatedearlier in Section 6.5.2. The uncertainty for the result includes both the range inpossible Y (N) values and the uncertainty in the measurement of log N (N I) added inquadrature.For oxygen, we may repeat the calculation that we carried out for nitrogen, againadopting rates at T = 8000 K. Here, we make use of Draine’s (2011, p. 155) charac-terization of the charge exchange reaction rates calculated by Stancil et al. (1999) be-tween O + and the three fine-structure levels of O , yielding C + , = 1 . × − cm s − for the three levels and C , + = (1 . , . , . × − cm s − , for the lev-els J = (2 , ,
0) of O . As with nitrogen, of Γ(O ) ≈ ). For oxygen, α = 4 . × − cm s − (Shull & Van Steenberg 1982 ; Nussbaumer & Storey 1983).An application of Eq. 30 for O with log(O / H) (cid:12) = 8 . −
12 and again assuminga depletion correction of 0.1 dex yields the result log N (H I) = 21 . ± .
06. Theuncertainty for this result is dominated by the measurement uncertainty for N (O I);possible values of Y (O) given in the last row of Table 5 are close to zero and smallcompared to the uncertainty in N (O I). Because the determination of the column Table 5.
Relative Ionizationsmin[ n ( e )] min[ n ( e )] max[ n ( e )] max[ n ( e )]Quantity min[ n (H )] max[ n (H )] min[ n (H )] max[ n (H )] n ( e ) (cm − ) 10 × × n (H ) (cm − ) 3 . × . × n (H + ) /n (H ) 0 .
285 10 − .
36 0 . n (N + ) /n (N ) 0 .
63 5 . × − .
82 0 . Y (N) 0 .
10 0 .
00 0 .
13 0 . n (O + ) /n (O ) 0 .
25 9 × − .
31 3 × − Y (O) − .
01 0 . − .
02 0 . Jenkins & Gry density of H I based on O is substantially more precise than that for N, we adoptthis value but note that the outcome for N is consistent with that for O within ouruncertainties of log(N I) and Y (N).If we treat the expression for n (H ) in Eq. 28 as an equality, rather than a limit,and then evaluate a characteristic length scale (cid:96) = N (H I) /n (H ) we obtain the result (cid:96) ≤ . n ( e ) = 3 × and 10 cm − , respectively. Because both ofthese values for (cid:96) are larger than our representative value R g = 6 AU, the inequalityin Eq. 28 must apply to some extent.Again we can use Eq. 28 to arrive at a column density for ionized hydrogenlog N (H II) ≤ log N (H I) + log[ n (H + ) /n (H )] = 20 .
45 and 20.56 for n ( e ) = 10 and 3 × cm − , respectively. A DEFICIENCY OF CARBON11.1.
Abundance Estimate
Unlike the preferred ionization stages of N and O, whose neutral forms we couldmeasure using weak transitions to obtain column densities, the dominant phase of C(C II ) exhibits only two transitions: (1) an allowed transition at 1334.5 ˚A that is fartoo strong to be useful for measuring N (C II ) and (2) an intersystem line at 2325.4 ˚Athat is so weak that it provides an upper limit for N (C II ) that is far too high to beof any use. To gain an understanding about the abundance of carbon, we must directour attention to the neutral form of this element and attempt to relate it to the totalabundance of carbon in the gas phase.As we discussed in Section 7.1, our inability to detect an absorption from one of theexcited fine-structure levels of C I sets a limit on the abundance of neutral carbonlog N (C I) < . × cm − . This limit for the amount of C I is substantially lowerthan the findings by Roberge et al. (2000) that N (C I) = (2 − × cm − for β Picand 4 . × cm − for 49 Cet (Roberge et al. 2014). A comparison of the 49 Cetresult against ours is especially noteworthy, since they found considerably more C I than our upper limit, yet their sum of results for N (O I) in the P and P levelsequaled 1 . × cm − compared to our result for the same sum of 1 . × cm − .(The A1V spectral type of 49 Cet is only slightly later than that of 51 Oph.) Fromthis disparity in the observations of the neutral forms of C and O, we might concludethat R g and/or n ( e ) for the gas around 49 Cet is larger than what we found for51 Oph, thus favoring a shift in the ionization equilibrium toward a higher neutralfraction for carbon, or alternatively, the ratio of the C to O abundance in the gassurrounding 49 Cet is larger than that associated with 51 Oph.A contributing factor that drives a deficiency of C I is the high value of Γ at R g = 6 AU, but our conclusion given in Table 3 that n ( e ) < × cm − relied onthe abundance of carbon (mostly ionized) being in accord with an expectation basedon solar abundance ratios relative to the neutral forms of oxygen and nitrogen. Tomake the calculation of n ( e ) consistent with our findings for the ionization equilibrium ircumstellar Gas Surrounding 51 Oph n ( e ) that we found from the metastable level populations, we would need toreduce our assumed abundance of C by at least a factor of 40.11.2. Expulsion by Radiation Pressure?
We now explore one possibility for depleting C in the gas. A common theme inthe study of circumstellar atomic gases is the consideration that the star’s outwardradiation pressure can weaken or overcome the gravitation attraction for atoms (andsome dust grains) in orbit (Beust et al. 1989 ; Lagrange et al. 1998 ; Olofsson etal. 2001 ; Liseau 2003 ; Brandeker et al. 2004 ; Fern´andez et al. 2006 ; Xie et al.2013). The parameter β describes the ratio of these two forces, F rad /F grav , which isindependent of the distance from the star since they both scale in proportion to R − g .For any transition (which we identify with a subscript i ) with an f -value equal to f i , the exposure of an atom to a local flux vector ( R ∗ /R g ) F λ,i (erg cm − s − ˚A − ) willabsorb photons at a rate γ i = πe m e c g i f i N ν,i = πe λ i hm e c (cid:18) R ∗ R g (cid:19) g i f i F λ,i , (31)where e and m e are the charge and mass of the electron, and N ν,i is the flux ofphotons per unit frequency ν for the transition being considered. The factor g i / λ i . Each absorption imparts a radial momentumimpulse h/λ i . Hence, we find that if the absorption lines for all transitions have lowcentral optical depths (i.e., there is no self-shielding of the radiation from the star),we obtain for neutral carbon β (C ) = hR g GM ∗ m C (cid:88) i ( γ i /λ i ) = 10 πe R ∗ m e c GM ∗ m C (cid:88) i g i f i λ i F λ i , (32)where R ∗ = 6 . R (cid:12) is the effective stellar radius (see Section 6.1.2), M ∗ = 3 . M (cid:12) isthe mass of the star (Jamialahmadi et al. 2015), and m C is the mass of the carbonatom (12 amu). We calculate that for 51 Oph β (C ) = 666 when we evaluate thesum in Eq. 32 over all C I transitions listed in the compilation of Morton (2003).This value is 20,000 times the value 0.033 computed by Fern´andez et al. (2006) for β Pic, largely due to the fact that UV flux from β Pic with T eff = 8000 K is about 4orders of magnitude weaker than that of 51 Oph at the wavelengths of the strongestC I transitions. A similar calculation by Lagrange et al. (1998) for β Pic gave β (C )that ranged between 0.011 and 0.095, depending on the amount of self-shielding inthe lines. Unlike our previous formulations for the effects of starlight, we do not consider a dilution factor W for the radiation at some distance from the star because the projections of the momentum transferevents along the radial direction are the relevant quantities in the equation, rather than the totalphoton interaction rates of the atoms. Jenkins & Gry
We now argue that even though β (C ) is quite large (considerably in excess of 0.5,which is the threshold for becoming unbound), the atoms are unlikely to migrateoutward. At 6 AU from the star, each atom experiences an average rate of absorption (cid:80) i γ i = 1 .
63 s − , while concurrently it is exposed to an ionization rate Γ = 3 . × − s − . After a carbon ion is neutralized, we expect that over the survival time ofits neutral form, 1 / Γ, the radiation pressure accelerates it in a radial direction andincreases its kinetic energy by m C ∆ v /
2, where∆ v = F rad m C Γ = β (C ) GM ∗ R g Γ . (33)(Recall from our discussion in Section 7.1 that the ionization of C by charge exchangewith protons is negligible compared to Γ, but Γ could be enhanced by secondaryionizations arising from energetic electrons produced by the ionization of H.) We notethat because Γ is proportional to R − g (for R g (cid:29) R ∗ ), the value of ∆ v does not changewith R g . The kinetic energy of the Keplerian motion E K = m C v m C GM ∗ R g (34)receives a relative boost v + ∆ v v = 1 + β (C ) GM ∗ R g Γ . (35)Numerically, we find that Eq. 35 yields a result for the fractional increase in E K equalto 1 + 2 . × − for R g = 6 AU. After receiving this small boost in kinetic energy,the neutral atom becomes ionized and remains in that state for a time interval thataverages [ αn ( e )] − = 1 . × s for n ( e ) = 10 cm − (or one-tenth that time for n ( e ) = 10 cm − ). The value that we assign for α at T = 8000 K still applies, becausethe suprathermal ∆ v = 0 . − arising from Eq. 33 is small compared to thethermal velocities. During that time interval, the strong coupling of the carbon ionsto other ions via Coulomb collisions (and possibly an embedded magnetic field) willforce them to surrender their enhanced velocity ∆ v and merge with the flow. Detailson the physics of this ionic braking process are given by Beust et al. (1989), Fern´andezet al. (2006) and Xie et al. (2013).We now conclude that even though β (C ) is large, there presently is little chancethat the carbon atoms can migrate outward. Instead, they must blend in with thecirculating river of ions in Keplerian orbits. At some much earlier time when theprotostar was emitting strong radiation but not at energies capable of ionizing themedium, the carbon may have had a chance to escape. However, it seems puzzlingthat carbon in an atomic form is currently not being replenished by dissociatingCO molecules at the boundary of the thin molecular disk. Also, the present-daydeficiency of carbon may present a challenge to the notion that the gas is composed of ircumstellar Gas Surrounding 51 Oph β values are con-siderably lower than that of C because they have fewer transitions in the UV andthese transitions are strongly saturated. (In computations of β , it is better to use theactual line equivalent widths that were or would have been observed instead of thosethat apply to unsaturated lines, as was done for Eqs. 31 and 32). DISCUSSION AND SUMMARYThe spectrum of 51 Oph has a rich assortment of absorption features arising fromatomic constituents in a region within about 10 AU of the star. We have made use of304 different transitions from 16 different elements for which we could measure thecolumn densities for a total of 98 different atomic levels, most of which are highlyexcited and can come only from the environment near the star and not the foregroundISM. For 9 additional species, we searched for 14 strong transitions but could onlydetermine upper (or lower) limits for the column densities. After determining columndensities from the spectral features, we were able to derive a number of conclusionson the nature of the gas in the vicinity of the star. These conclusions are summarizedvery briefly in Table 6.
Table 6.
Summary of Conclusions on the Gas in the SightlineProperty Value Method n ( e ) 10 − × cm − N I , Fe II , and Ni II metastable populations (Sections 6.5.3 and 6.6)Na I , Zn I ionization equilibria (Section 7.1) n (H) ≥ . × cm − Cl I and Cl II charge exchange (Section 7.3) N (H I) 1 . × cm − N (O I) and N (N I) (Section 9) combined withcharge exchange and protosolar abundances (Section 10) N (H II) ≤ . × cm − (Section 10) T I , Fe II , and Ni II metastable populations (Sections 6.5.3 and 6.6)Element A pattern consistent with Contents of Section 9Abundances a mildly depleted ISM and the best matches in Fig. 11or Jupiter-class comet dustRepresentative distance 6 AU Fe II , Ni II metastable populations (Sections 6.5.3 and 6.6)from the star R g Jenkins & Gry
Disk Orientation
In Section 5 we noted that the contrast between the absence of both CO and H absorptions in the UV spectrum of 51 Oph and the detection of CO in emission in thefar infrared indicates that we must be sampling regions away from the central planeof the disk where the molecules are shielded and thus protected from dissociatingradiation that can reduce the molecular fraction to a very low value. Thus, while thedisk may be seem to be nearly edge on from our viewpoint, it is not exactly so. Thisslight misalignment could explain why we can view resonantly scattered starlight inthe bottoms of the saturated absorption features of O I . Our results, when combinedwith a future, accurate determination of the disk inclination, would help to constrainthe scale height h/r of the molecular material.12.2. Ionization State
Our sight line to 51 Oph consists of significant portions of both neutral and ionizedhydrogen, which we could not observe directly, but instead could be sensed from theabundances of neutral and ionized nitrogen. Further support for the existence of bothkinds of gas arises from the metastable level populations of N I that indicate largeelectron densities, along with the surprisingly strong presence of neutral chlorine thatmust arise from charge exchange reactions with neutral hydrogen.12.3. Excitation Mechanisms
Using atomic data that we could find in the literature for N I , Fe II , and Ni II , we caninterpret our measurements of their excited metastable levels in a manner that offeredimportant clues on the condition and location of the gas. The populations of theselevels are maintained by a combination of optical pumping by starlight and collisionswith electrons. The findings for the strengths of these two effects are somewhatdegenerate with each other, but our MCMC analysis allowed us to focus on a restrictedrange for their possible combinations. For the conditions that we found, collisionswith neutral hydrogen atoms should have a minimal effect on such excitations.We acknowledge that we have oversimplified our descriptions of the metastablelevel excitations in terms of a gas at a particular distance from the star which has auniform electron density and temperature. Nevertheless, future interpretations thatinvoke interpretations of more extended gas structures having variable conditionsshould still be able to make use of our simplified conclusions on what we can regardas density-weighted values for these parameters as a starting point. Such modelingmay also include a picture where the N I is located in a different region than thatwhich highlights the presence of Fe II and Ni II . Some support for a difference inthe distributions of ions and neutrals arises from the fact that the ions show botha principal, narrow-velocity component (Component 1 described in Section 4) and a Determinations that we could find in the literature: 88 ◦ +2 − (Thi et al. 2005), > ◦ (Berthoud etal. 2007), and 80 ◦ (no uncertainty specified) (Thi et al. 2013). ircumstellar Gas Surrounding 51 Oph I .12.4. Electron Density
An important finding of our investigation on the metastable level populations isthat the electron densities fall somewhere in the range 10 < n ( e ) < cm − . Theneutral to ionized fractions of Na and Zn indicate a somewhat narrower range 10 Spektr-RG space observatory launched in 2019.0 Jenkins & Gry We are unable to test for the presence of X-ray or EUV radiation by looking forevidence of ions that can be created by photons with energies significantly greaterthan 13.6 eV because they will be neutralized by charge exchange with the neutralhydrogen that we know to be present.12.5. Gas Temperature Our interpretation of the populations of the N I metastable levels yields an outcomefor the temperature T = 8000 K, and this value seems to hold over very broad rangesof possible electron densities and distances from the star. As the level of ionization ofthe gas is large, we can expect that the high temperature is created by a significantheat input from the energetic electrons produced by the ionization of hydrogen. Itis interesting to note that this temperature coincides with the base of the “Lyman- α Wall,” which is a location where the cooling rate from neutral hydrogen excitationrises steeply with temperature above the various other forms of atomic cooling; see,e.g. Figure 2 of Dalgarno & McCray (1972). (However, we point out that the otheratomic cooling rates are much lower than indicated in that figure, because the gas issubstantially denser than the critical densities of the collisionally excited levels thatare most responsible for cooling at slightly lower temperatures.)12.6. Atomic Abundances In the circumstellar material, the excitations of different atoms to excited levels isso strong that our determinations of the abundances of various elements is heavilydependent on our measurements of the column densities of both ground-state andupper level populations, together with our estimates for the amounts of atoms inlevels that we could not observe. For any particular element, the ratio of amounts ofobserved to total levels is a driver for the reliability of our abundance measurement,and it varied from 5% to 100% across different elements. The outcomes of our columndensity measurements range from log N = 12 . 14 for Ti to 17.63 for O. If we excludecarbon, the pattern of abundances across different elements seems to be most con-sistent with the gas composition of the interstellar medium with a moderate level ofdepletion. However, we also find reasonably good matches to the abundance patternsof solid material associated with Jupiter-family comets. In the 51 Oph system, solidcometary material with a similar composition could have been converted to gaseousform from collisions or during close approaches toward the star. Such an outlook isconsistent with the observations of transient absorption lines associated with fallingevaporating bodies (FEBs) detected for 51 Oph and other stars with circumstellardisks (Vidal-Madjar et al. 1994, 1998 ; Beust et al. 1998, 2001 ; Karmann et al. 2001,2003 ; Th´ebault & Beust 2001 ; Roberge et al. 2002 ; Beust & Valiron 2007 ; Welsh& Montgomery 2013 ; Kiefer et al. 2014, 2019 ; Eiroa et al. 2016 ; Vidal-Madjar etal. 2017 ; Zieba et al. 2019).Debris disks around A-type stars typically span distances from tens to hundreds ofAU from the central star (Hughes et al. 2018, Fig. 3). One can imagine that the ircumstellar Gas Surrounding 51 Oph R g = 6 AU from the star is smaller than the general radial extent of thedebris disk (which, as far as we know, has not yet been well measured for 51 Oph).One prominent deviation in the abundances is that of carbon. From the lack of C I and comparing its ionization equilibrium with those of other species, we concludedthat C is relatively deficient (by more than one order of magnitude) compared to othervolatile elements such as O and N. Even though neutral carbon atoms are subjectto a repulsive radiation force that is considerably stronger than the gravitationalattraction, these atoms become ionized before they can acquire a significant outwardvelocity, and then their migration is halted by Coulomb collisions that cause them toblend in and become bound to the sea of circulating ions.12.7. The Role of UV Spectroscopy This paper has highlighted how UV absorption-line spectroscopy can reveal manyimportant conclusions on the physical nature and composition of gases in the cir-cumstellar environments in a late stage of development. The observations are mosteasily interpreted when one chooses a star that has a radial velocity that differs fromthat of any foreground material in the ISM. Features from metastable levels providepowerful insights on the condition and location of the gas, and they frequently ap-pear in the spectra of hot stars with nearly edge-on disk systems. The observationsof atomic element abundances are complementary to those that investigate infraredand millimeter emission, which reveal the dust and molecular constituents.2 Jenkins & Gry ACKNOWLEDGMENTSThis research was supported by an archival research grant nr. HST-AR-15029.001-A provided by NASA through the Space Telescope Science Institute (STScI), whichis operated by the Associations of Universities for Research in Astronomy, Incorpo-rated, under NASA contract NAS5-26555. All of the ultraviolet spectroscopic dataanalyzed in this paper were obtained from the Mikulski Archive for Space Telescopes (MAST) maintained by the STScI. Specific observations used in this paper can beaccessed via the following collections of GHRS, STIS and FUSE data on the MASTdoi: 10.17909/t9-b6j3-w085 Our determination of the distance to 51 Oph came fromdata from the European Space Agency (ESA) mission Gaia Gaia Gaia Multilateral Agreement. The UVES observations were collected from Eu-ropean Southern Observatory (ESO) Science Archive Facility from which we down-loaded observations taken under the ESO observing program 079.C-0789(A). We ob-tained information on atomic energy levels from the NIST catalog VI/74 availableon the Strasbourg VizieR website http://cdsarc.u-strasbg.fr/viz-bin/Cat?VI/74. Wethank B. T. Draine for his review and comments on a late draft of this paper. Facilities: HST (STIS, GHRS), FUSE, HUT, VLT:Kueyen (UVES) Software: Owens, developed by Martin Lemoine (Institut d’Astrophysique de Paris)and the French FUSE Science Team ircumstellar Gas Surrounding 51 Oph A. ATOMIC ABSORPTION LINESTable 7 lists the atomic lines that we measured to determine column densities,along with the f -values that we adopted and their references. All wavelengths arevacuum values, regardless of whether they are in the UV or visible. The logarithmicrelative uncertainties in the f -values listed in Column (5) of the table are from thepercentage classifications given in the NIST compilation (see note a in the table),except for the elements Cr and Ni where uncertainties were specified in the originalreferences. Strongly saturated lines are not included in the table, except for (1) onecase where a line from N I in the ground state could be used in checking for theaverage attenuation of the flux for optical pumping, (2) O I lines featured in theexposition in Section 8, and (3) absorption from a metastable O I level that was usedfor determining a lower limit for the column density in Section 6.5.2. Table 7 . Properties of Atomic LinesWavelength Species Excitation f -value f -value Relative f -value(˚A) (cm − ) Uncertainty (dex) Source a (1) (2) (3) (4) (5) (6)1560.3092 C I 0 1.28E − 02 0.01 (1)1656.9283 C I 0 1.40E − 01 0.01 (1)1328.8333 C I 0 6.31E − 02 0.04 (1)1329.571 C I 43 5.69E − · · · (1)1329.593 C I 43 1.89E − · · · (1)2325.4029 C II 0 4.78E − 08 0.003 (1)1134.9803 b N I 0 4.15E − 02 0.04 (1)1159.8168 N I 0 9.95E − 06 0.01 (1)1492.6250 N I 19224 6.93E − 02 0.03 (2)1492.8200 N I 19233 1.09E − 02 0.03 (2)1494.6751 N I 19233 5.80E − 02 0.03 (2)1310.9431 N I 28838 3.12E − 02 0.04 (2)1318.9980 N I 28838 1.20E − 02 0.03 (2)1319.6695 N I 28838 8.90E − 03 0.03 (2)1327.9172 N I 28838 2.50E − 03 0.04 (2)1411.9310 N I 28838 2.67E − 02 0.01 (2)1742.7192 N I 28838 1.93E − 02 0.04 (2)1745.2485 N I 28838 3.82E − 02 0.04 (2) Table 7 continued on next page Jenkins & Gry Table 7 (continued) Wavelength Species Excitation f -value f -value Relative f -value(˚A) (cm − ) Uncertainty (dex) Source a (1) (2) (3) (4) (5) (6)1310.5403 N I 28839 2.97E − 02 0.04 (2)1310.9498 N I 28839 4.51E − 03 0.04 (2)1319.0050 N I 28839 2.97E − 03 0.03 (2)1319.6762 N I 28839 1.50E − 02 0.03 (2)1326.5709 N I 28839 1.77E − 03 0.04 (2)1411.9387 N I 28839 3.03E − 03 0.01 (2)1411.9483 N I 28839 2.39E − 02 0.01 (2)1742.7309 N I 28839 4.80E − 02 0.04 (2)1745.2603 N I 28839 9.16E − 03 0.04 (2)1083.9937 b N II 0 1.11E − 01 0.01 (1)1084.5659 b N II 49 2.72E − 02 0.01 (1)1084.5841 b N II 49 8.30E − 02 0.01 (1)1085.5328 b N II 131 1.06E − 03 0.01 (1)1085.5511 b N II 131 1.61E − 02 0.01 (1)1085.7096 b N II 131 9.21E − 02 0.01 (1)1302.1685 c O I 0 4.80E − 02 0.01 (1)1355.5977 O I 0 1.16E − 06 0.04 (1)1304.8576 c O I 158 4.78E − 02 0.01 (1)1358.5123 O I 158 6.27E − 07 0.04 (1)1306.0286 c O I 227 4.78E − 02 0.01 (1)1152.1512 O I 15868 1.08E − 01 0.04 (3)5897.5581 Na I 0 3.201E − < − < − 01 0.01 (1)1845.5205 Si I 0 2.70E − 01 0.07 (1)1304.3702 Si II 0 9.17E − 02 0.08 (1)1808.0126 Si II 0 2.08E − 03 0.12 (1)1309.2758 Si II 287 9.13E − 02 0.08 (1)1816.9285 Si II 287 1.66E − 03 0.12 (1)1817.4512 Si II 287 1.29E − 04 0.12 (1)1381.4760 P I 0 3.16E − · · · (1)1152.8180 P II 0 2.45E − · · · (1)1155.0137 P II 165 6.10E − · · · (1)1153.9951 P II 469 1.86E − · · · (1)1425.0300 S I 0 1.25E − 01 0.03 (1) Table 7 continued on next page ircumstellar Gas Surrounding 51 Oph Table 7 (continued) Wavelength Species Excitation f -value f -value Relative f -value(˚A) (cm − ) Uncertainty (dex) Source a (1) (2) (3) (4) (5) (6)1335.7258 d Cl I 0 3.13E − 02 0.12 (1)1347.2396 Cl I 0 1.53E − 01 0.08 (1)1351.6561 Cl I 882 1.21E − 01 0.08 (1)1363.4475 Cl I 882 5.50E − 02 0.08 (1)1071.0358 Cl II 0 1.50E − 02 0.02 (1)4227.918 Ca I 0 1.77E+00 0.01 (1)3934.7750 Ca II 0 1.267E − < . 01 (1)3969.5901 Ca II 0 3.116E − < . 01 (1)3384.7304 Ti II 0 3.58E − 01 0.03 (1)2056.2568 Cr II 0 1.03E − 01 0.04 (1)2062.2359 Cr II 0 7.59E − 02 0.04 (1)2066.1638 Cr II 0 5.12E − 02 0.04 (1)2744.4530 Cr II 11961 1.730E − 01 0.04 (4)2669.5010 Cr II 12032 7.744E − 02 0.04 (4)2679.5840 Cr II 12032 1.294E − 01 0.03 (4)2742.8430 Cr II 12032 5.104E − 02 0.04 (4)2749.7965 Cr II 12032 1.318E − 01 0.04 (4)2666.8120 Cr II 12147 1.306E − 01 0.11 (4)2672.5980 Cr II 12147 7.778E − 02 0.03 (4)2751.5400 Cr II 12147 1.084E − 01 0.04 (4)2758.5350 Cr II 12147 7.583E − 02 0.04 (4)2664.2140 Cr II 12303 7.013E − 02 0.04 (4)2677.9560 e Cr II 12303 1.282E − · · · (4)2752.6780 Cr II 12303 6.323E − 02 0.04 (4)2763.4050 Cr II 12303 1.364E − 01 0.04 (4)2844.0850 Cr II 12303 2.870E − 01 0.01 (4)2677.9540 e Cr II 12496 2.244E − · · · (4)2767.3480 Cr II 12496 2.051E − 01 0.04 (4)2836.4630 f Cr II 12496 3.614E − 01 0.01 (4)3422.1906 Cr II 19528 9.53E − 02 0.03 (4)3404.2968 Cr II 19631 5.42E − 02 0.07 (4)3343.5421 Cr II 19797 2.575E − 02 0.02 (4)3359.4649 Cr II 19797 3.571E − 02 0.06 (4)3423.7209 Cr II 19797 6.29E − 02 0.02 (4)3369.0170 Cr II 20024 8.549E − 02 0.02 (4) Table 7 continued on next page Jenkins & Gry Table 7 (continued) Wavelength Species Excitation f -value f -value Relative f -value(˚A) (cm − ) Uncertainty (dex) Source a (1) (2) (3) (4) (5) (6)3409.7432 Cr II 20024 4.71E − 02 0.04 (4)3380.7907 Cr II 25046 3.40E − 02 0.05 (4)2299.6631 Mn II 0 4.81E − 04 0.12 (1)2305.7141 Mn II 0 1.15E − 03 0.08 (1)2576.8770 Mn II 0 3.61E − 01 0.01 (1)2594.4990 g Mn II 0 2.80E − 01 0.01 (1)2606.4620 Mn II 0 1.98E − 01 0.01 (1)3442.9710 Mn II 14325 5.000E − 02 0.03 (5)3461.3050 Mn II 14593 3.340E − 02 0.03 (5)3489.6730 Mn II 14901 3.85E − 02 0.03 (5)2610.9810 Mn II 27547 3.500E − 01 0.12 (6)3720.9928 Fe I 0 4.11E − 02 0.01 (1)1611.2004 Fe II 0 1.38E − 03 0.12 (1)2249.8768 Fe II 0 1.82E − 03 0.04 (1)2260.7805 Fe II 0 2.44E − 03 0.04 (1)1618.4680 Fe II 384 2.14E − 02 0.01 (1)2146.7218 h Fe II 384 2.35E − · · · (1)2252.2537 Fe II 384 5.58E − > − 03 0.08 (1)2269.5248 h Fe II 384 3.06E − 04 0.08 (1)2280.6202 Fe II 384 4.37E − 03 0.03 (1)2333.5156 Fe II 384 7.78E − 02 0.08 (1)2365.5518 Fe II 384 4.95E − 02 0.03 (1)2383.7884 Fe II 384 5.57E − 03 0.08 (1)2389.3582 Fe II 384 8.25E − 02 0.03 (1)2396.3559 i Fe II 384 2.88E − 01 0.03 (1)2599.1465 Fe II 384 1.08E − 01 0.03 (1)2612.6542 Fe II 384 1.26E − 01 0.03 (1)2626.4511 Fe II 384 4.41E − 02 0.03 (1)1625.9123 Fe II 667 6.08E − 03 0.08 (1)1629.1596 j Fe II 667 3.67E − 02 0.01 (1)2251.6338 Fe II 667 2.20E − 03 0.08 (1)2268.2878 Fe II 667 3.62E − 03 0.08 (1)2328.1212 Fe II 667 3.45E − 02 0.03 (1)2349.0223 k Fe II 667 8.98E − 02 0.08 (1) Table 7 continued on next page ircumstellar Gas Surrounding 51 Oph Table 7 (continued) Wavelength Species Excitation f -value f -value Relative f -value(˚A) (cm − ) Uncertainty (dex) Source a (1) (2) (3) (4) (5) (6)2381.4887 Fe II 667 3.38E − 02 0.04 (1)2396.1497 i Fe II 667 1.53E − 02 0.03 (1)2607.8664 Fe II 667 1.18E − 01 0.04 (1)2618.3991 Fe II 667 5.05E − 02 0.04 (1)2632.1081 Fe II 667 8.60E − 02 0.03 (1)2250.8727 Fe II 862 1.35E − 03 0.08 (1)2255.1048 Fe II 862 2.11E − > − 03 0.08 (1)2338.7233 Fe II 862 8.97E − 02 0.03 (1)2359.8266 Fe II 862 6.79E − 02 0.08 (1)2614.6044 Fe II 862 1.08E − 01 0.03 (1)2621.1905 Fe II 862 3.93E − 03 0.04 (1)2631.8312 Fe II 862 1.31E − 01 0.03 (1)2256.6869 Fe II 977 1.17E − · · · (1)2345.0011 Fe II 977 1.53E − 01 0.03 (1)2622.4518 Fe II 977 5.60E − 02 0.03 (1)2629.0777 Fe II 977 1.73E − 01 0.04 (1)2332.0233 Fe II 1872 2.07E − 02 0.03 (7)2348.8346 k Fe II 1872 4.30E − 02 0.03 (7)2360.7210 Fe II 1872 3.00E − 02 0.04 (7,8)2361.0159 Fe II 2430 3.90E − 02 0.03 (8)2362.7434 Fe II 2430 1.18E − 02 0.08 (7,8)2392.2069 Fe II 2430 4.05E − 03 0.04 (7,8)2355.6103 Fe II 2837 1.48E − 02 0.04 (7)2367.3166 Fe II 2837 8.50E − 03 0.04 (7)2369.3195 Fe II 2837 3.40E − 02 0.03 (7)2383.9718 Fe II 2837 3.06E − 02 0.03 (8)2385.7331 Fe II 2837 4.09E − 03 0.08 (7,8)2371.2226 Fe II 3117 1.46E − 02 0.04 (7,8)2375.9187 Fe II 3117 4.20E − 02 0.04 (7)2385.1148 Fe II 3117 2.75E − 02 0.08 (7)1562.2689 Fe II 7955 2.85E − 03 0.11 (9)1635.4003 Fe II 7955 7.22E − 02 0.2 (10)2715.2171 Fe II 7955 4.70E − 02 0.03 (10)2740.3577 Fe II 7955 2.49E − 01 0.03 (7,8) Table 7 continued on next page Jenkins & Gry Table 7 (continued) Wavelength Species Excitation f -value f -value Relative f -value(˚A) (cm − ) Uncertainty (dex) Source a (1) (2) (3) (4) (5) (6)2756.5507 Fe II 7955 3.06E − 01 0.04 (7,8)1641.7630 Fe II 8391 4.85E − 02 0.2 (11)2592.3176 Fe II 8391 5.80E − 02 0.08 (8)2725.6909 Fe II 8391 1.07E − 02 0.04 (7)2728.3465 Fe II 8391 6.98E − 02 0.03 (7,8)2747.7940 Fe II 8391 1.91E − 01 0.03 (8)2750.1341 l Fe II 8391 3.27E − 01 0.03 (7)2583.3560 Fe II 8680 8.80E − 02 0.03 (8)2611.8523 Fe II 8680 1.12E − 02 0.08 (8)2731.5424 Fe II 8680 3.13E − 02 0.04 (7)2737.7758 Fe II 8680 6.90E − 02 0.04 (7)2747.2957 Fe II 8680 3.48E − 01 0.03 (7)2749.9936 l Fe II 8680 1.37E − 01 0.03 (7,8)2769.7520 Fe II 8680 8.20E − 03 0.04 (10)2578.6936 Fe II 8846 1.20E − 01 0.2 (10)2594.5034 g Fe II 8846 3.30E − 02 0.04 (8)2744.0081 Fe II 8846 4.50E − 01 0.03 (7)2750.2987 l Fe II 8846 1.32E − 01 0.04 (7)2762.6287 Fe II 8846 3.20E − 02 0.04 (7)2019.4288 Fe II 15844 1.9E − 02 0.2 (11)2041.3455 Fe II 15844 2.9E − 02 0.2 (11)2064.3371 Fe II 15844 1.0E − 02 0.2 (11)2162.7011 Fe II 15844 1.8E − 02 0.2 (10)1876.8391 Fe II 16369 2.4E − 02 0.2 (11)1878.3873 Fe II 16369 1.8E − 03 0.3 (11)2033.0634 Fe II 16369 2.1E − 02 0.2 (11)2051.6908 Fe II 16369 2.3E − 02 0.2 (11)2176.1348 Fe II 16369 1.5E − 02 0.2 (11)2088.2055 Fe II 18360 4.5E − 02 0.2 (11)1877.4657 Fe II 20340 4.3E − 02 0.08 (11)2001.0254 Fe II 20340 4.8E − 02 0.08 (11)2221.0720 m Fe II 20340 3.10E − 02 0.2 (10)2383.6237 Fe II 20340 1.61E − 02 0.2 (10)2037.0908 Fe II 20516 2.6E − 02 0.2 (11)1880.9722 Fe II 20805 2.9E − 02 0.2 (11) Table 7 continued on next page ircumstellar Gas Surrounding 51 Oph Table 7 (continued) Wavelength Species Excitation f -value f -value Relative f -value(˚A) (cm − ) Uncertainty (dex) Source a (1) (2) (3) (4) (5) (6)1888.7342 Fe II 20805 7.6E − 02 0.08 (11)2011.3468 Fe II 20805 4.0E − 02 0.2 (11)2234.6115 Fe II 20805 3.0E − 02 0.2 (11)2214.3450 Fe II 21251 2.39E − 02 0.2 (10)2346.0553 Fe II 21251 5.48E − 02 0.11 (9)2224.1775 Fe II 21711 2.2E − 02 0.2 (10)2362.4468 Fe II 21711 2.01E − 02 0.08 (10)2619.8555 Fe II 22637 2.78E − 02 0.2 (10)2632.3929 Fe II 22637 8.20E − 02 0.08 (10)2621.4769 Fe II 22810 3.40E − 02 0.2 (10)2624.5078 Fe II 22939 1.99E − 02 0.2 (9)2630.3724 Fe II 22939 8.57E − 02 0.08 (10)2627.2838 Fe II 23031 5.4E − 02 0.2 (10)2630.8554 Fe II 23031 8.0E − 02 0.2 (10)1785.2720 Fe II 23317 7.40E − 01 0.2 (9)1786.7520 Fe II 23317 5.60E − 01 0.2 (9)1788.0039 Fe II 23317 2.10E − 01 0.2 (11, 12)1788.0780 Fe II 23317 1.50E − 01 0.2 (9)2712.6451 Fe II 25428 5.6E − 02 0.08 (10)2728.1907 Fe II 25428 2.9E − 02 0.08 (10)2770.1736 n Fe II 25428 2.78E − 02 0.08 (9)2709.8587 Fe II 25787 6.4E − 02 0.2 (10)2832.3938 Fe II 25787 1.38E − 01 0.08 (10)2836.5452 f Fe II 25787 9.3E − 02 0.08 (10)2713.1943 Fe II 25805 1.72E − 02 0.12 (13)2768.3197 o Fe II 26170 2.12E − 01 0.03 (10)2784.5119 Fe II 26170 1.02E − 01 0.03 (10)2754.1013 Fe II 26352 2.58E − 01 0.12 (13)2780.1194 Fe II 26352 9.3E − 02 0.08 (10)2638.4295 Fe II 26932 1.4E − 01 0.2 (10)2775.5049 Fe II 26932 6.3E − 02 0.2 (10)2841.4847 Fe II 26932 1.85E − 01 0.08 (10)2364.5330 o Fe II 27314 2.5E − 02 0.2 (10)2665.4560 Fe II 27314 2.54E − 01 0.03 (10)2704.7908 Fe II 27314 1.51E − 01 0.03 (10) Table 7 continued on next page Jenkins & Gry Table 7 (continued) Wavelength Species Excitation f -value f -value Relative f -value(˚A) (cm − ) Uncertainty (dex) Source a (1) (2) (3) (4) (5) (6)2667.4286 o Fe II 27620 2.65E − 01 0.03 (10)2717.0220 Fe II 27620 1.27E − 01 0.08 (10)2693.4008 Fe II 30388 1.83E − 01 0.03 (10)2685.5506 Fe II 30764 2.12E − 01 0.03 (10)2593.5595 Fe II 32875 3.16E − 01 0.03 (10)2626.2722 Fe II 32909 3.08E − 01 0.03 (10)2577.6353 Fe II 33466 1.58E − 01 0.03 (10)2588.7233 Fe II 33501 2.13E − 01 0.03 (10)2698.2611 Fe II 36126 9.0E − 02 0.08 (10)2607.2945 Fe II 36252 2.35E − 01 0.03 (10)2770.1502 n Fe II 36252 1.50E − 02 0.2 (10)2286.8640 Co II 3350 3.06E − 01 0.03 (15)2308.5700 Co II 4028 2.54E − 01 0.04 (15)2354.1423 Co II 4560 1.58E − 01 0.03 (15)1317.217 Ni II 0 5.71E − 02 0.05 (16)1370.1323 Ni II 0 5.88E − 02 0.05 (16)1741.5500 Ni II 0 4.27E − 02 0.04 (17)1751.9100 Ni II 0 2.77E − 02 0.04 (17)1381.2860 p Ni II 1506 1.15E − > p Ni II 1506 1.42E − > − 02 0.05 (19)1788.4858 Ni II 1506 2.50E − 02 0.05 (19)2131.9350 Ni II 8393 4.27E − 03 0.05 (19)2166.2300 Ni II 8393 1.66E − 01 0.03 (19)2217.1680 Ni II 8393 3.16E − 01 0.03 (19)2223.6420 Ni II 8393 7.59E − 02 0.03 (19)2316.7480 Ni II 8393 1.86E − 01 0.03 (19)2169.7720 Ni II 9330 1.06E − 01 0.03 (19)2175.3500 Ni II 9330 1.37E − 01 0.03 (19)2211.0680 Ni II 9330 4.14E − 02 0.03 (19)2225.5560 Ni II 9330 1.22E − 01 0.03 (19)2270.9140 Ni II 9330 1.54E − 01 0.03 (19)2303.7010 Ni II 9330 1.65E − 01 0.03 (19)2159.4150 Ni II 10115 1.70E − 02 0.03 (19)2175.8240 Ni II 10115 1.26E − 01 0.03 (19) Table 7 continued on next page ircumstellar Gas Surrounding 51 Oph Table 7 (continued) Wavelength Species Excitation f -value f -value Relative f -value(˚A) (cm − ) Uncertainty (dex) Source a (1) (2) (3) (4) (5) (6)2207.4000 Ni II 10115 1.49E − 01 0.03 (19)2227.0200 Ni II 10115 1.00E − 01 0.03 (19)2265.1610 Ni II 10115 1.55E − 01 0.03 (19)2297.8490 Ni II 10115 1.42E − 01 0.03 (19)2346.1630 Ni II 10115 1.70E − 02 0.03 (19)2185.2860 Ni II 10663 2.03E − 01 0.03 (19)2202.0920 Ni II 10663 1.51E − 01 0.03 (19)2254.5460 Ni II 10663 2.23E − 01 0.03 (19)2298.1970 Ni II 10663 1.17E − 01 0.03 (19)2279.4740 Ni II 13550 1.94E − 01 0.04 (19)2297.2580 Ni II 13550 1.73E − 01 0.03 (19)2335.3020 Ni II 13550 6.56E − 02 0.03 (19)2395.2520 Ni II 13550 1.85E − 01 0.03 (19)2287.7860 Ni II 14995 1.46E − > − 01 0.04 (19)2376.1450 Ni II 14995 7.28E − 02 0.03 (19)2221.0900 m Ni II 23108 2.27E − > − > − > − > − > − > − > − > − > − > − > − > − > − > q Ni II 29593 3.06E − > − > − > − > − 01 0.08 (1) Table 7 continued on next page Jenkins & Gry Table 7 (continued) Wavelength Species Excitation f -value f -value Relative f -value(˚A) (cm − ) Uncertainty (dex) Source a (1) (2) (3) (4) (5) (6)1367.9509 Cu II 0 1.79E − 01 0.08 (1)1472.3951 Cu II 0 2.17E − > − 01 0.01 (20)2192.9531 Cu II 22847 3.02E − 01 0.04 (20)2139.2477 Zn I 0 1.47E00 0.01 (1)2026.1370 Zn II 0 6.30E − · · · (21)2062.6604 Zn II 0 3.09E − · · · (21) Table 7 continued on next page ircumstellar Gas Surrounding 51 Oph Table 7 (continued) Wavelength Species Excitation f -value f -value Relative f -value(˚A) (cm − ) Uncertainty (dex) Source a (1) (2) (3) (4) (5) (6) References — a Except for reference (1), most of the transition f ∼ peter/newpage/. Original determinations are as follows:(1) Values and references listed in Morton (2003), (2) Tachiev & Froese Fischer (2002), (3) Butler & Zeippen(1991) (4) Nilsson et al. (2006), (5) Den Hartog et al. (2011), (6) Kurucz (1990) with additional data downloadedfrom http://kurucz.harvard.edu/linelists.html on December 11, 2012. (7) Bergeson et al. (1996), (8) Schnabelet al. (2004), (9) Tayal & Zatsarinny (2018), (10) Fuhr & Wiese (2006), (11) Raassen & Uylings (1998), (12)Johansson et al. (1995) (13) Sikstr¨om et al. (1999), (14) Mullman et al. (1998), (15) Salih et al. (1985), (16)Jenkins & Tripp (2006) (17) Fedchak et al. (2000), (18) Kurucz (2012), (19) Fedchak & Lawler (1999), (20) vanHoof (2017, 2018): no original reference specified. (21) Kisielius et al. (2015) b These strong transitions for nitrogen are very strongly saturated and not properly resolved in the FUSE spectrum.The equivalent width of the N I feature was used only to verify the derivation of the saturation factors (s.f.,see Eq. 3) for optical pumping, and the very strong absorptions by N II indicated the importance of ionizationprocesses beyond that provided by photons with energies below the ionization potential of hydrogen. c Very strongly saturated lines shown in Figure 6. The very weak intersystem line at 1355.598 ˚A was used toderive N (O I ). d This line is not useful in the present study, since it occurs near the bottom of a very strong absorption by the λλ II . However, in future investigations of other stars, this Cl I line may notsuffer from this interference. e The Cr II lines at 2677.956 and 2677.954 ˚A are severely blended. However, the simultaneous fitting of the variouslines from different levels performed by the Owens analysis allows for this superposition. f The Cr II line at 2836.463 ˚A and the Fe II line at 2836.5452 ˚A are blended, but our analysis fitted these featuressimultaneously. g The Mn II line at 2594.4990 ˚A and Fe II line at 2594.5034 ˚A are blended, but our analysis fitted these featuressimultaneously. These fits are substantiated by 3 other lines that are free from interference for both Mn II andFe II in the same energy levels. h This line is too weak to see in our spectrum. i This line was not considered in our analysis because it was too close to the edge of our spectrum. j This line coincided with a detector flaw, so it was not considered in our analysis. k The Fe II lines at 2348.835 and 2349.022 ˚A are blended, but our analysis fitted these features simultaneously. l The Fe II lines at 2749.994, 2750.134, and 2750.299 ˚A overlap each other, but our analysis fitted these featuressimultaneously. m The Fe II line at 2221.072 ˚A and the Ni II line at 2221.090 ˚A are blended, but our analysis fitted these featuressimultaneously. Other lines of Fe II in this level substantiate the fit. n The Fe II lines at 2770.1736 and 2770.1502 ˚A are blended, but our analysis fitted these features simultaneously.Added confidence in these fits comes from other lines of Fe II from the same levels. o The lines of Fe II at 2364.533, 2667.429, and 2768.320 ˚A are blended with lines of Fe II originating from muchhigher levels, which are too weak to matter. p These two lines for Ni II were not used because two other lines with more accurate f -values were available forthis level. q An absorption line is present at this location, but it indicates an unreasonably high column density for this level.There must be some other unidentified transition that coincides with this one. Jenkins & Gry REFERENCES Allende Prieto, C., Koesterke, L., Hubeny, I.,et al. 2018, A&A, 618, A25Amarsi, A. M., & Barklem, P. S. 2019, A&A,625, A78Anderson, D. L. 1989, Theory of the Earth(Boston: Blackwell)Arenou, F., Luri, X., Babusiaux, C., et al. 2018,A&A, 616, A17Arun, R., Mathew, B., Manoj, P., et al. 2019,AJ, 157, 159Asplund, M., Grevesse, N., Sauval, A. J., &Scott, P. 2009, ARA&A, 47, 481Badnell, N. R. 2006, ApJS, 167, 334Bardyn, A., Baklouti, D., Cottin, H., et al.2017, MNRAS, 469, S712Bautista, M. A., Romano, P., & Pradhan, A.K. 1998, ApJS, 118, 259Bergeson, S. D., Mullman, K. L., Wickliffe, M.E., et al. 1996, ApJ, 464, 1044Bergh¨ofer, T. W., Schmitt, J. H. M. M., &Cassinelli, J. P. 1996, A&AS, 118, 481Berthoud, M. G., Keller, L. D., Herter, T. L.,Richter, M. J., & Whelan, D. G. 2007, ApJ,660, 461Beust, H., Lagrange-Henri, A. M., Vidal-Madjar, A., & Ferlet, R. 1989, A&A, 223,304Beust, H., Lagrange, A. M., Crawford, I. A., etal. 1998, A&A, 338, 1015Beust, H., Karmann, C., & Lagrange, A. M.2001, A&A, 366, 945Beust, H., & Valiron, P. 2007, A&A, 466, 201Bowen, D. V., Jenkins, E. B., Tripp, T. M., etal. 2008, ApJS, 176, 59Bowen, I. S. 1947, PASP, 59, 196Brandeker, A., Liseau, R., Olofsson, G., &Fridlund, M. 2004, A&A, 413, 681Brandt, J. C., Heap, S. R., Beaver, E. A., et al.1994, PASP, 106, 890Brown, M. S., Alkhayat, R. B., Irving, R. E.,et al. 2018, ApJ, 868, 42Butler, K., & Zeippen, C. J. 1991, J Phys IVFrance, C1, 141Butler, S. E., & Dalgarno, A. 1980, A&A, 85,144Cassidy, C. M., Ramsbottom, C. A., Scott, M.P., & Burke, P. G. 2010, A&A, 513, A55Chen, C. H., & Jura, M. 2003, ApJ, 582, 443Chen, C. H., & Kamp, I. 2004, ApJ, 602, 985Cheng, K. P., & Neff, J. E. 2003, AJ, 125, 868Crawford, I. A., Craig, N., & Welsh, B. Y. 1997,A&A, 317, 889Dalgarno, A., & McCray, R. A. 1972, ARA&A,10, 375Dekker, H., D’Odorico, S., Kaufer, A., Delabre,B., & Kotzlowski, H. 2000, SPIE, 4008, 534Den Hartog, E. A., Lawler, J. E., Sobeck, J.S., Sneden, C., & Cowan, J. J. 2011, ApJS,194, 35Dent, W. R. F., Thi, W. F., Kamp, I., et al.2013, PASP, 125, 477Diplas, A., & Savage, B. D. 1994, ApJ, 427, 274Dixon, W. V., Blair, W. P., Kruk, J. W., &Romelfanger, M. L. 2013, PASP, 125, 431 Draine, B. T. 2011, Physics of the Interstellarand Intergalactic Medium (Princeton:Princeton U. Press)Dunkin, S. K., Barlow, M. J., & Ryan, S. G.1997, MNRAS, 286, 604Eiroa, C., Rebollido, I., Montesinos, B., et al.2016, A&A, 594, L1Fajardo-Acosta, S. B., Telesco, C. M., &Knacke, R. F. 1993, ApJ, 417, L33Fedchak, J. A., & Lawler, J. E. 1999, ApJ, 523,734Fedchak, J. A., Wiese, L. M., & Lawler, J. E.2000, ApJ, 538, 773Fern´andez, R., Brandeker, A., & Wu, Y. 2006,ApJ, 643, 509Flynn, G. J., Bleuet, P., Borg, J., et al. 2006,Sci, 314, 1731Fossati, L., Koskinen, T., Lothringer, J. D., etal. 2018, ApJ, 868, L30Fuhr, J. R., & Wiese, W. L. 2006, JPCRD, 35,1669Fusco, M., Redfield, S., Roberge, A.,Stapelfeldt, K. R., & Jensen, A. G. 2013,AAS Meeting Abstracts 221.144.20Grady, C. A., & Silvis, J. M. S. 1993, ApJ, 402,L61Grady, C. A., Brown, A., Welsh, B., et al. 2018,AJ, 155, 242Gry, C., & Jenkins, E. B. 2014, A&A, 567, A58H´ebrard, G., Lemoine, M., Vidal-Madjar, A.,et al. 2002, ApJS, 140, 103Hughes, A. M., Lieman-Sifry, J., Flaherty, K.M., et al. 2017, ApJ, 839, 86Hughes, A. M., Duchene, G., & Matthews, B.2018, ARA&A, 56, 541Jamialahmadi, N., Berio, P., Meilland, A., etal. 2015, A&A, 579, A81Jenkins, E. B., & Tripp, T. M. 2006, ApJ, 637,548Jenkins, E. B. 2009, ApJ, 700, 1299—. 2013, ApJ, 764, 25Jessberger, E. K., Christoforidis, A., & Kissel,J. 1988, Natur, 332, 691Johansson, S., Brage, T., Leckrone, D. S.,Nave, G., & Wahlgren, G. M. 1995, ApJ,446, 361Jura, M. 1974, ApJ, 190, L33Jura, M., & Young, E. D. 2014, AREPS, 42, 45Karmann, C., Beust, H., & Klinger, J. 2001,A&A, 372, 616—. 2003, A&A, 409, 347Keller, L. D., Sloan, G. C., Forrest, W. J., etal. 2008, ApJ, 684, 411Kiefer, F., Lecavelier des Etangs, A., Augereau,J. C., et al. 2014, A&A, 561, L10Kiefer, F., Vidal-Madjar, A., Lecavelier desEtangs, A., et al. 2019, A&A, 621, A58Kisielius, R., Kulkarni, V. P., Ferland, G. J., etal. 2015, ApJ, 804, 76Kondo, Y., & Bruhweiler, F. C. 1985, ApJ, 291,L1Krems, R. V., Jamieson, M. J., & Dalgarno, A.2006, ApJ, 647, 1531Kruk, J. W., Durrance, S. T., Kriss, G. A., etal. 1995, ApJ, 454, L1 ircumstellar Gas Surrounding 51 Oph Jenkins & Gry ∼∼