Variations in D/H and D/O from new FUSE observations1 Cristina M. Oliveira2, H. Warren Moos2, Pierre Chayer2,3, and Jeffrey W. Kruk2 ABSTRACT We use data obtained with the Far Ultraviolet Spectroscopic Explorer (FUSE) 6 0 to determine the interstellar abundances of DI, NI, OI, FeII, and H along the 2 0 sightlines to WD1034+001, BD+393226, and TD132709. Our main focus is on 2 determining the D/H, N/H, O/H, and D/O ratios along these sightlines, with log n a N(H) >20.0, that probegaswell outside oftheLocalBubble. Hubble Space Tele- J scope (HST) and International Ultraviolet Explorer (IUE) archival data are used 5 to determine the HI column densities along the WD1034+001 and TD132709 1 v sightlines, respectively. For BD+393226, a previously published N(HI) is used. 4 We find (D/H)×105 = 2.14 ± 0.53, 1.17 ± 0.31, and 1.86 ± 0.53, and (D/O)×102 1 0.45 0.25 0.43 1 = 6.31 ± 1.79, 5.62 ± 1.61, and 7.59 ± 2.17, for the WD1034+001, BD+393226, 1.38 1.31 1.76 1 and TD132709 sightlines, respectively (all 1σ). The scatter in these three D/H 0 6 ratios exemplifies the scatter that has been found by other authors for sightlines 0 / with column densities in the range 19.2 < log N(H) < 20.7. The D/H ratio h p toward WD1034+001 and all the D/O ratios derived here are inconsistent with - o the Local Bubble value and are some of the highest in the literature. We discuss r the implications of our measurements for the determination of the present-epoch t s a abundance of deuterium, and for the different scenarios that try to explain the : v D/H variations. We present a study of D/H as a function of the average sightline i X gas density, using the ratios derived in this work as well as ratios from the lit- r erature, which suggests that D/H decreases with increasing gas volume density. a Similar behaviors by other elements such Fe and Si have been interpreted as the result of depletion into dust grains. Subject headings: ISM: Abundances — ISM: Evolution — Ultraviolet: ISM — Stars: Individual (WD1034+001, BD+393226, TD132709) 1BasedonobservationsmadewiththeNASA-CNES-CSAFarUltravioletSpectroscopicExplorer. FUSE is operated for NASA by The Johns Hopkins University under NASA contract NAS5-32985. 2Department of Physics and Astronomy, The Johns Hopkins University, 3400 N. Charles St., Baltimore MD 21218 3Primary affiliation: Department of Physics and Astronomy, University of Victoria, P.O. Box 3055, Victoria, BC V8W 3P6, Canada. – 2 – 1. INTRODUCTION The present day abundance ratio of deuterium to hydrogen places important constraints on Big Bang nucleosynthesis (BBN) and the chemical evolution of galaxies. Because deu- terium is only produced in appreciable amounts in primordial BBN and destroyed in stellar interiors, the measurement of D I/H I in the interstellar medium (ISM) places a lower limit on the primordial abundance of deuterium. In addition, by comparing the ISM abundance of deuterium to its abundance in high-redshift intergalactic gas we should be able to better understand the effects of astration and chemical evolution of galaxies. Measurements of the D/H ratio in intervening clouds of gas seen toward distant quasars have yielded a range of valuesD/H=(1.65–4.0)×10−5 (O’Mearaetal.2001;Pettini&Bowen2001;Levshakovetal. 2002, and references therein). Kirkman et al. (2003) measured D/H = (2.42 ± 0.35)×10−5 0.25 toward Q1234+3047 (z = 2.526). These authors calculate the value they believe is the best estimate of the primordial D/H abundance, D/H = (2.78 ± 0.44)×10−5 (1σ errors in the prim 0.38 mean), by taking the weighted mean of five D/H measurements toward QSOs ranging from z = 2.079 – 3.572. This value is in good agreement with that determined from the cos- mic microwave background measurements performed with WMAP and previous missions (Spergel et al. 2003). Sembach et al. (2004) found D/H = (2.2 ± 0.7)×10−5 for Complex C, a high velocity cloud falling into our galaxy, which has low metallicity and has presumably experienced more stellar processing than the gas seen toward QSOs. Measurements of D/H in the local ISM have been made with Copernicus (e.g. Rogerson & York 1973), HST, (e.g. Linsky et al. 1995), IMAPS, (Jenkins et al. 1999; Sonneborn et al. 2000), and more recently FUSE (e. g. Moos et al. 2002, and references therein). A nearly constant ratio of D/H = (1.5 ± 0.1)×10−5 (1σ on the mean) has been obtained in the Local Interstellar Cloud (LIC) by Linsky (1998); recent measurements inside the Local Bubble (LB, log N(HI) ≤ 19.2 Sfeir et al. 1999) appear to be consistent with a single value for D/H in the LB (Moos et al. 2002; Oliveira et al. 2003). From a compilation of measurements from the literature Wood et al. (2004) derive (D/H) = (1.56 ± 0.04)×10−5. Using (D/O) LB LB = (3.84 ± 0.16)×10−2 from H´ebrard & Moos (2003) and (O/H) = (3.45 ± 0.19)×10−4 LB from Oliveira et al. (2005) one can derive indirectly (D/H) = (1.33 ± 0.09)×10−5. The LB direct and indirect determinations of (D/H) are only consistent when one considers the LB 2σ uncertainties in both quantities. A detailed discussion of the possible causes for this discrepancy can be found in H´ebrard & Moos (2003). Outside the Local Bubble, however, there still is not a consistent picture of the D/H behavior. Measurements performed along sightlines probing gas outside the LB suggest variations of the interstellar D/H ratio beyond the Local Bubble, at the distance of a few hundredparsecs. D/Hvariesfor19.2≤logN(HI)≤20.7(e.g.Jenkinsetal.1999;Sonneborn – 3 – et al. 2000; Friedman et al. 2002) and apparently remains constant for log N(HI) ≥ 20.7, albeit with a lower value. The possibility of a trend towards low D abundance at high H column densities was noted in the D/O survey by H´ebrard & Moos (2003). Wood et al. (2004), using additional measurements and values in the literature, confirmed the trend for D/H. So far, only five measurements for sightlines with log N(HI) ≥ 20.7 have been published. Hoopes et al. (2003), Wood et al. (2004) and H´ebrard et al. (2005) measured D/H along five extended lines of sight (d ≥ 500 pc). The weighted mean of these five measurements yields D/H = (0.87 ± 0.08)×10−5, in clear disagreement with the LB value. The value of the present-epoch Milky Way abundance of deuterium, (D/H) , is still PE an issue subject to debate. Two opposite explanations have emerged. (H´ebrard & Moos 2003) argue for a value of the current-epoch D/H ratio lower than D/H , on the basis of the LB generallylowvaluesfoundformoredistantsightlines. Ontheotherhand,Linskyetal. (2005, in prep) argue that an important fraction of deuterium along many sightlines is trapped in a population of grain material and hence the current-epoch deuterium abundance (gas plus grains) is (D/H) ≥ (2.19 ± 0.27)×10−5. The astration factors, (D/H) /(D/H) , PE prim PE derived from both of these scenarios (∼4 and ∼1.2, respectively) challenge current models of galactic chemical evolution. Because the O/H variations in the diffuse ISM from the LB out to 1 Kpc, are small, consistent with the measurement uncertainties (Meyer et al. 1998; Andr´e et al. 2003; Car- tledge et al. 2004; Oliveira et al. 2005), OI has been used as a proxy for HI (for a more detailed discussion see H´ebrard & Moos 2003, and references therein). Thus D/O and D/H are expected to trace each other in the diffuse ISM. However, how well this works at high column densities is a subject of discussion (H´ebrard 2006, and §7.1 of this paper). At the present time, definitive conclusions are prevented by the limited number of D/H and D/O measurements along high column density sightlines. In this work we present D/H andD/Omeasurements along threesightlines (WD1034+001,BD+393226,andTD132709) with log N(HI) ≥ 20.0. Using data obtained with FUSE, IUE, and the Goddard High Resolution Spectrograph (GHRS, onboard HST) we derive column densities of atomic and molecular species which are then used to determine important ratios (D/H, N/H, O/H, etc.) that can be compared to values in the literature. This paper is organized as follows. The three targets used for the analyses are described in §2. The observations and data processing are presented in §3. In §4 we determine the column densities of HI, DI, NI, OI, FeII, and H along the WD1034+001, BD+393226, 2 and TD132709 sightlines. Lower limits to the column densities of other elements are also reported. The D/H, D/N, D/O, N/H, O/H, and O/N ratios for the three sightlines are pre- sented in §5 and discussed in §6, where evidence for above average D abundance is presented – 4 – for the WD1034+001 and TD132709 sightlines. The D/O ratios for the three sightlines are substantially higher than the mean LB value, and thus they considerably increase the scatter in the published D/O ratios. Implications of the derived ratios for hypothetical mechanisms, such as dust depletion and infall of metal poor gas, that try to explain the D/H variations in the Galactic disk are considered in §6.4. Our derived D/H, D/O, and O/H ratios are com- pared with previously published ratios in §7.1. In §7.2 we consider the relationship between D/O and Fe/O. In §7.3 we perform a study of D/H as a function of the average sightline gas density (N(H)/d). Our study indicates that D/H is constant up to densities of 0.10 cm−3 and decreases with increasing density after this point, similarly to what has been observed for other elements such as Fe and Si. We summarize our findings in §8. All uncertainties are quoted at the 1σ level unless stated otherwise. 2. THE TARGETS The properties of the three stars are listed in Table 1. Below we discuss each star in detail. 2.1. WD1034+001 WD1034+001 is a hot DO white dwarf with T =100,000 ± 15,000 K, log g = 7.5 eff 10,000 ± 0.3, at a distance of 155 ± 58 pc (photometric, Werner et al. 1995) in the direction l 43 = 247.55◦ and b = +47.75◦. This star was first observed in the ultraviolet with IUE by Sion et al. (1985), who identified several photospheric features (CIV, NV, and OV) as well as interstellar features (NI, CII, SiII, MgII, SiII, and SiIII). These authors derived v − v ∼50 km s−1 (v and v correspond to stellar and interstellar absorptions, PH ISM PH ISM respectively). Non-stellar absorption by CIV was also detected along this sightline by Sion et al. (1985), however the resolution of the data did not allow them to decide whether this absorption had an interstellar or circumstellar origin. Werner et al. (1995) used GHRS onboard HST to observe this target at a resolution of ∼20,000. In addition to stellar absorption by HeII, CIV, NV, OV, SiIV, FeVI, and FeVII, they also identified absorption by highly ionized species of CIV, NV, and SiIV. With resolution similar to that of Sion et al. (1985), Werner et al.(1995) could not determine whethertheabsorptionhadaninterstellarorcircumstellarorigin. UsingHILyαobservations obtained with GHRS in conjunction with NLTE model atmospheres, Werner et al. (1995) determined log N(HI) = 20.05 ± 0.20. In this work we reanalyze these HI Lyα observations – 5 – together with FUSE data to place tighter constraints on N(HI) (see §4.3.1). Using Sloan Digital Sky Survey data, Hewett et al. (2003) discovered a region of ionized gas around WD1034+001, seen in H , H , [NII], [OIII], and [SII] emission, with a diameter α β greater than 2◦ and identified it with a planetary nebula, Hewett 1. This is one of the largest planetary nebula in the sky, and also the first to be unambiguously associated with a DO white dwarf. Using data obtained with the Southern H-Alpha Sky Survey Atlas, Rauch et al. (2004) found two extended emission structures surrounding WD1034+001, in addition to theplanetary nebula discovered by Hewett et al.(2003). The inner halohasa linear diameter of 16.2 ± 6.1×24.3± 9.1 pc while the wider shell has a linear diameter of 27×43 pc (Rauch 4.5 6.8 et al. 2004), at the quoted distance for WD1034+001 (see Table 1). During its evolution, WD1034+001 is likely to have passed through a much hotter phase with T = 150,000K, eff roughly 30,000 years ago (Rauch et al. 2004). These authors estimate that WD1034+001 is most likely the exciting star of the nebula and the halo given that the recombination time scale in such a low-density (n = 3 cm−3) nebula is longer than 30,000 years. e 2.2. BD+393226 BD+393226, first identified as a sdO by Berger & Fringant (1978), lies at a distance of 290 ± 140 pc (Hipparcos) in the direction l = 65.00◦, b = +28.77◦. Bluhm et al. (1999) 70 determined the abundances of the atomic (HI, DI, etc.) and molecular species (H ) along 2 the line of sight using data obtained with IUEand ORFEUS II,and determined v −v ∼ PH ISM −255 km s−1. However, the spectral resolution of only ∼30 km s−1 for both the IUE and ORFEUS IIdatasetsledtofairlylargeuncertainties onthecolumndensities determinedwith those data. The new FUSE data with a resolution of∼20 km s−1 andbetter S/N ratioallows us to determine more accurate column densities and in addition, offers us the possibility of comparing column densities determined from data obtained with different instruments. InthisworkweusetheFUSE datatorevisit thecolumndensities alongtheBD+393226 sightlineforthespeciesthathavetransitionsintheFUSE bandpass. SincetheLyαtransition of HI is not covered by FUSE we use the value reported by Bluhm et al. (1999). These authors used Lyα observations from IUE and ORFEUS II, in conjunction with a stellar model with T = 45,000 K and logg = 5.5, to determine log N(HI) = 20.08 ± 0.09. eff – 6 – 2.3. TD132709 TD132709 (UVO 0904−02) was first identified as a sdO by Berger & Fringant (1980) using data obtained with the Ultraviolet Sky Survey Telescope S2/68. Using optical data in conjunction with NLTE stellar atmosphere models, Dreizler (1993) determined T = 46,500 eff ± 1,000 K and log g = 5.55 ± 0.15. The corresponding photometric distance is d = 520 ± 90 pc (z = 245 ± 40 pc). IUE observations covering the Lyα region in combination with FUSE data and stellar models allow us to determine N(HI) along this sightline (see §4.3.2). From our work we determine v −v = −13 km s−1 (see §4.2.3). PH ISM 3. OBSERVATIONS AND DATA PROCESSING 3.1. FUSE Observations The FUSE observatory consists of four coaligned prime-focus telescopes and Rowland- circle spectrographs that produce spectra over the wavelength range 905 – 1187 ˚A, with a spectral resolution of ∼15 – 20 km s−1 (wavelength dependent) for point sources. Details about the FUSE mission, its planning, and on-orbit performance can be found in Moos et al. (2000) and Sahnow et al. (2000). Table 2 summarizes the FUSE observations of the three targets. The data were ob- tained through the medium size aperture (MDRS, 4′′×20′′), in histogram (H) or ttag modes (T). The two-dimensional FUSE spectra are reduced using the CalFUSE pipeline v2.4.1 or v2.4.21. The processing includes data screening for low quality or unreliable data, ther- mal drift correction, geometric distortion correction, heliocentric velocity correction, dead time correction, wavelength calibration, detection and removal of event bursts, background subtraction, and astigmatism correction. The spectra are aligned by cross-correlating the individual exposures over a short wavelength range that contains prominent spectral features and then coadded by weighting each exposure by its exposure time, using the CORRCAL software developed by S. Friedman. All the spectra are binned to three pixel samples, or∼20 m˚A, for analysis (the line spread function, LSF, is about 11 pixels or ∼70 m˚A wide). For each target, all the observations were coadded in order to increase the S/N of the dataset. The S/N per unbinned pixel in the SiC 1B channel around 940 ˚A is 8, 26, and 14, for WD1034+001, BD+393226, and TD132709, respectively. Figures 1, 2, and 3, present the FUSE spectra of the three targets. The most prominent 1TheCalFUSEpipelinereferenceguideisavailableathttp://fuse.pha.jhu.edu/analysis/pipeline reference.html – 7 – interstellar and photospheric lines are labeled; photospheric lines are indicated by [], and H absorption by dashed vertical lines. 2 4. ANALYSIS Whenever possible we use apparent optical depth, curve of growth and profile fitting methods (hereafter AOD, COG, and PF, respectively) to determine column densities (see e.g. Oliveira et al. 2003, and references therein for a further discussion of these methods). AOD is used on weak transitions to determine column densities of species that have multiple transitions in the FUSE bandpass, which are then compared to the values derived with the other methods. For species that only have saturated transitions in the FUSE bandpass we use the AOD technique to place lower limits on the column densities of those species. We use the COG method to determine the column densities of species that have several transitionsintheFUSEbandpass(oneormoreofthesetransitionsmustbeunsaturated). We measure the equivalent widths of all the non-blended transitions, in all the FUSE channels where those transitions are covered, and compare them to check for inconsistencies (due to fixed pattern noise, for instance). A single component Gaussian curve-of-growth is then fit to the measured equivalent widths of the atomic species, allowing us to determine N and b, for each species. For H we fit a single-component Gaussian COG to the equivalent widths 2 of all the J levels, simultaneously. We determine then N(J) and b, common to all J levels. With PF we fit a single absorption component to one or more non-saturated transitions of the species for which N is being sought. We use the profile fitting code Owens, developed by Martin Lemoine and the French FUSE Team. Details of how the PF method is used for the analysis of each sightline studied in this paper are given below. More information about the fitting code can be found in Lemoine et al. (2002) and H´ebrard et al. (2002). We follow the procedures outlined in Oliveira et al. (2003) to use these techniques and determine the uncertainties in N associated with each method. We introduce a modification from Oliveira et al. (2003) regarding the PF technique. A single-Gaussian with a FWHM of 10.5 pixels (constant across all wavelengths and channels and during the fitting procedure) is used here to describe the FUSE line spread function (LSF), whereas previously the LSF was a free parameter of the fitting procedure during the initial stages of the fit (see Oliveira et al. 2003, for more details). However, since several studies have shown that column densities obtained with fits using variable and fixed LSFs are similar (e.g. Williger et al. 2005) we prefer to decrease the number of degrees of freedom of the profile fitting routine by using a fixed LSF. – 8 – All the atomic lines used inthe AOD,COG, andPF analyses are shown in Table 3 along with log fλ for each transition. We use the compilation of Morton(2003)for the atomic data and the ones by Abgrall et al. (1993a) and Abgrall et al. (1993b) for the H data. For each 2 star, A, C, and P denote transitions that are used with the apparent optical depth, curve of growth, and profile fitting methods, respectively. Table 4 presents the equivalent widths of the atomic lines used with the COG method for each sightline. After column densities and uncertainties are determined with these different techniques the results are examined to check the consistency between the different methods. The adopted results are a subjective compromise between the values obtained with the several methods used; the uncertainties adopted are such that they in general include the extreme values obtained with the different methods. No attempt was made to combine the different results in any statistical way. The results of the individual methods are also presented (see § 4.2). In § 4.1 we determine the H column densities along the three lines of sight, atomic 2 column densities are discussed in § 4.2. Determination of the HI column density along the WD1034+001 and TD132709 sightlines is discussed in §4.3. 4.1. Molecular Hydrogen 4.1.1. WD1034+001 Along this sightline we detect absorption arising from the H rotational levels J = 0 2 – 3. We use AOD, single-b COG and PF to derive the column densities of the J = 0 – 3 levels. From the single-b COG fit we derive b = 3.6 ± 0.4 km s−1. Profile fitting of the COG 0.3 different J levels is done simultaneously with the atomic species along this sightline, under the assumption that the molecular gas does not trace precisely the atomic gas (i.e. atomic and molecular species are fit in different absorption components). Although the velocities of the atomic and molecular component are a free parameter of the fit, no significant velocity differences are found between the two components. The adopted column densities for each J-level are summarized in Table 5. We adopt log N(H ) = 15.72 ± 0.13. The excitation 2 0.12 diagram for H along this sightline is presented in the top panel of Figure 4. We derive T 2 02 = 341 ± 75 K. 4.1.2. BD+393226 For this sightline we detect absorption arising from the H rotational levels J = 0 – 5. 2 We use AOD, single-b COG and PF to derive the column densities of the J = 0 – 3 levels. – 9 – The J = 4, 5 levels have very low column densities and we use only PF to determine the column densities associated with these two J levels. From the single-b COG fit to the J = 0 – 3 levels we derive b = 4.3 ± 0.3 km s−1. COG 0.2 Similarly to WD1034+001, profile fitting of the J = 0 – 5 levels is done simultaneously with the atomic species, but in different absorption components. Table 5 presents the adopted column densities for this sightline. We adopt log N(H ) = 15.65 ± 0.06. The excitation 2 0.07 diagram for H along this sightline is presented in the middle panel of Figure 4. The popula- 2 tions of the levels with J ≥ 2 are larger than expected if the different levels were populated according to a Boltzmann distribution with an excitation temperature corresponding to T 01 (indicated by a dashed line in Figure 4). This nonthermal excitation is the result of pump- ing by UV photons followed by cascading transitions down through the various rotational levels (Black & Dalgarno 1976). The high J lines can typically be fit by a single excitation temperature. However, in the case of this sightline a single temperature fit to J = 2 – 5 clearly overestimates the column densities of the J = 3 and J = 4 levels. We find that the distribution of H through the different J levels is better described by several temperatures. 2 We derive then T = 104 ± 27 K, T = 200 ± 15 K, and T = 953 ± 158 K. 01 13 35 4.1.3. TD132709 We use AOD and PF to determine the column densities of the H rotational levels J 2 = 0 – 3. For this sightline, the low column density of each J level implies that the number of absorption lines that could be used with the COG method is small, and hence we do not performaCOGfitforH alongthissightline. TheconsistencyofPFresultsischeckedagainst 2 that of the AOD method. Table 5 presents the adopted N(J) for this sightline. We derive N(H ) = 14.48 ± 0.12. The bottom panel of Figure 4 presents the H excitation diagram for 2 0.11 2 this sightline. All the J levels are consistent with a single temperature corresponding to T 02 = 292 ± 83 K. No HD is detected along any of these sightlines. Typical HD/H ratios along these types 2 of diffuse sightlines are of the order of 1×10−5 or less. Considering N(H ) quoted in Table 2 5 we expect log N(HD) < 11. 4.2. Atomic Species In this section we discuss the determination of the atomic column densities along the three sightlines. However, determination of N(HI) along the WD1034+001 and TD132709 – 10 – sightlines will be discussed separately in §4.3. Tables 6, 7, and 8 present the column densities for DI, NI, OI, and FeII obtained with the different methods discussed above. The adopted column densities for the atomic species along the three sightlines are summarized in Table 9. 4.2.1. WD1034+001 We use the AOD method to derive lower limits to the column densities of CII∗, CIII, NII, SiII, PII, and ArI, because all the transitions of these species in the FUSE bandpass are saturated. The AOD method is also used for weak transitions of DI, NI, OI, and FeII, to determine column densities which are then compared to the column densities determined with the other methods. We fit a single-Gaussian COG to the measured equivalent widths of DI, NI, OI, and FeII. The four COGs yield b-values in the range b = 5 – 6 km s−1. Figure 5 presents the curve of growth for OI along this sightline. With PF we fit a single absorption component to non-saturated lines of DI, NI, OI, and FeII. In order to model the continuum in the vicinity of the DI lines an extra absorption component with HI only, is also included in the fit. This component is not used to derive N(HI) along this sightline. No stellar model is used in the fitting process because the photospheric spectrum is generally flat in the regions of interest. As discussed above, H is 2 also included in the fit, in a different absorption component. The DI lines used in this analysis have a minor blending with H , hence N(DI) is not affected by the assumed N(H ). 2 2 Along this sightline we also detect absorption by OVI (λ1032, 1038), SIII (λ1012), and SIV (λ1063). OVI absorption has two components, separated by ∼ 50 km s−1. The strongest component falls at the stellar velocity derived by Sion et al. (1985) and is likely of stellar origin. Absorption by SIII, SIV, and the weaker OVI absorption component fall at the velocity of the atomic gas (DI, NI, OI, and FeII) along this sightline. However, we cannot conclude with certainty whether these features have an interstellar or circumstellar origin. Note that as discussed in §2.1, two large ionized halos have been discovered in the vicinity of WD1034+001. The OVI lines are weak and blended with stellar OVI absorption; we derive N(OVI) with PF, following the method described in §4. The SIII absorption line is strong and likely saturated. We derive only a lower limit to its column density using the AOD technique. The SIV line is blended with an unidentified feature on the blue side, and is likely saturated. We derive a lower limit to its column density using the AOD technique. Figure 6 presents fits to some of the lines used in the analysis of the WD1034+001