Molecules at =0.89:
Key Words.:quasars: absorption lines - quasars: individual: PKS 1830211 - galaxies: ISM - galaxies: abundances - ISM: molecules
We present the results of a 7 mm spectral survey of molecular absorption lines originating in the disk of a =0.89 spiral galaxy located in front of the quasar PKS 1830211. Our survey was performed with the Australia Telescope Compact Array and covers the frequency interval 30–50 GHz, corresponding to the rest-frame frequency interval 57–94 GHz. A total of 28 different species, plus 8 isotopic variants, were detected toward the south-west absorption region, located about 2 kpc from the center of the =0.89 galaxy, which therefore has the largest number of detected molecular species of any extragalactic object so far. The results of our rotation diagram analysis show that the rotation temperatures are close to the cosmic microwave background temperature of 5.14 K that we expect to measure at =0.89, whereas the kinetic temperature is one order of magnitude higher, indicating that the gas is subthermally excited. The molecular fractional abundances are found to be in-between those in typical Galactic diffuse and translucent clouds, and clearly deviate from those observed in the dark cloud TMC 1 or in the Galactic center giant molecular cloud Sgr B2. The isotopic ratios of carbon, nitrogen, oxygen, and silicon deviate significantly from the solar values, which can be linked to the young age of the =0.89 galaxy and a release of nucleosynthesis products dominated by massive stars. Toward the north-east absorption region, where the extinction and column density of gas is roughly one order of magnitude lower than toward the SW absorption region, only a handful of molecules are detected. Their relative abundances are comparable to those in Galactic diffuse clouds. We also report the discovery of several new absorption components, with velocities spanning between 300 and +170 km s. Finally, the line centroids of several species (e.g., CHOH, NH) are found to be significantly offset from the average velocity. If caused by a variation in the proton-to-electron mass ratio with redshift, these offsets yield an upper limit 4, which takes into account the kinematical noise produced by the velocity dispersion measured from a large number of molecular species.
More than 150 molecules have been discovered in space, as a result of targeted investigations or spectral surveys toward Galactic sources such as circumstellar envelopes of evolved stars (e.g. IRC+10216, Cernicharo et al. 2000; Patel et al. 2011), massive star-forming clouds (e.g. Sgr B2, Nummelin et al. 2000), or cold molecular clouds (e.g. TMC-1, Ohishi et al. 1992; Kaifu et al. 2004). The study of the rotational and vibrational spectra of these molecules is particularly useful to probe the physical conditions and chemical content of the interstellar gas. Most importantly, the large number of molecules found in various environments demonstrates the richness and variety of interstellar chemistry.
As the sensitivity of millimeter instruments has increased, a large number of molecules have become observable in galaxies in the Local Universe (see e.g. Wang et al. 2004; Martín et al. 2006 for NGC4945 and NGC253, respectively) and molecules are still detected at high redshifts (currently up to =6.42), though only a few species such as CO, HCN, HCO, and HNC (e.g. Walter et al. 2003; Wagg et al. 2005; Riechers et al. 2006; Guélin et al. 2007). Those detections are mostly related to ultra-luminous galaxies, which are potentially unrepresentative of all galaxies at these epochs. Moreover, line emission from these high- galaxies is dominated by the warmest and densest regions of the molecular component.
Observations of molecular absorption lines toward a bright radio continuum source allow rare molecular species to be detected even in distant galaxies, where they cannot be observed in emission because of distance dilution. This technique can thus be used to investigate the chemistry and its complexity at high redshifts and trace the chemical enrichment history of the Universe. In addition, high- molecular absorbers can serve as cosmological probes of the cosmic microwave background (CMB) temperature and have attracted much interest in constraining variations in the fundamental constants of physics (see e.g., Henkel et al. 2009 and references therein).
Unfortunately, only a handful of high-redshift radio molecular absorbers 111That is, not including the H/CO absorbers detected from optical/UV spectroscopy, such as those discussed in Noterdaeme et al. (2011). have been discovered to date (see the review by Combes 2008), despite numerous searches (e.g. Wiklind & Combes 1996a; Drinkwater et al. 1996; Murphy et al. 2003; Curran et al. 2004, 2006, see also Curran et al. 2011). The molecular absorption originating in the =0.89 and =0.68 intervening galaxies located in front of the quasars PKS 1830211 (=2.5) and B 0218+357 (=0.94) respectively, were discovered by Wiklind & Combes (1995, 1996b) and have many similarities: both absorbers appear to be (faint) nearly face-on spiral galaxies (Winn et al. 2002; York et al. 2005); both galaxies act as gravitational lenses and split the image of their background quasar into two main compact components and an Einstein ring, as seen at radio wavelengths (Jauncey et al. 1991; Patnaik et al. 1993); they both harbor large column densities of absorbing molecular gas (H cm); and in both cases the radio continuum of the background quasar is bright enough to permit sensitive observations with current millimeter instruments. A dozen molecular species have indeed been successfully detected in these absorbers, as well as absorption by their rare isotopologues (cf e.g., Wiklind & Combes 1995, 1996b, 1998; Menten et al. 1999; Muller et al. 2006).
The multiplicity of lensed continuum images of the background quasar makes it possible to explore as many different lines of sight through the intervening galaxy. In the case of B 0218+357, molecular absorption is detected only toward one image of the quasar (Menten & Reid 1996; Muller et al. 2007). Toward PKS 1830211, molecular absorption is detected, however, for both of the bright and compact images, located one arcsecond apart on the north-east and south-west side, respectively, of the Einstein ring (Wiklind & Combes 1998; Muller et al. 2006). The two lines of sight intercept the disk of the =0.89 galaxy (hereafter referred to as FG0.89, where FG stands for foreground galaxy) on either side of its bulge, at galactocentric distances of 2 kpc (to the SW) and 4 kpc (to the NE). An image of the galaxy, obtained with the Hubble Space Telescope in the I band is shown in Fig.1.
The absorption toward the SW image (hereafter FG0.89SW) has a large column of absorbing gas ((H cm) and is taken as the reference 0 km s in velocity (=0.88582, Wiklind & Combes 1996b). Remarkably, the full width at zero power of the HCO(2-1) line reaches nearly 100 km s for this component, although the galaxy is seen almost face-on. We note the relatively high kinetic temperature of 80 K (Henkel et al. 2008) and moderate density of a few times cm (Henkel et al. 2009) toward FG0.89SW, which imply that the excitation of the rotational lines is subthermal and mostly, but not necessarily totally, coupled with the CMB. Rotation temperatures derived for several molecules are consistent with the value =5.14 K at =0.89 (Combes & Wiklind 1997; Menten et al. 1999; Henkel et al. 2009), as predicted for an adiabatic expansion of the Universe.
The absorption toward the NE component (hereafter FG0.89NE) is located 147 km s blueward and is narrower (FWHM15 km s), so that both velocity components can easily be resolved kinematically. Nevertheless, the column of molecular gas toward the NE image is about two orders of magnitude less than in FG0.89SW, and the corresponding absorption lines are significantly weaker. Little is known about the excitation of and the physical conditions toward FG0.89NE. However, based on the analysis of time variations in the NE absorption, Muller & Guélin (2008) argued that it should be caused by a small number (5) of clouds of size 1 pc and of density a few 100 cm, hence resembling Galactic diffuse clouds.
Molecular absorption has also been detected in the galaxies hosting the radio sources B3 1504+377 at =0.67 (in CO, HCO, HCN, and HNC; Wiklind & Combes 1996b), PKS 1413+135 at =0.25 (in CO, HCO, HCN, HNC, and CN; Wiklind & Combes 1997), and PKS 0132097 at =0.76 (although only in OH; Kanekar et al. 2005). However, the weakness of the background radio continuum sources, narrower linewidths and smaller column densities of absorbing gas than the two absorbers toward PKS 1830211 and B 0218+357, ensures that it is difficult to perform a deep inventory of their molecular constituents with current instruments.
In this paper, we present the first unbiased spectral-line survey toward a high- molecular absorber, located at =0.89 in front of the quasar PKS 1830211. Observations, conducted with the Australian Compact Array and covering the whole 7 mm band from 30 GHz to 50 GHz, are presented in §.2. Our results are presented in §.3, and then discussed in §.4, in terms of isotopic ratios (§.4.1), the chemical composition relative to other objects (§.4.2), and constraints on the variation in fundamental constants of physics (§.4.3). Notes for the different molecular species detected in the survey are given in appendix.
2 Observations and data reduction
Observations were carried out with the Australian Telescope Compact Array (ATCA) over three nights, on 2009 September 1 and 2 and on 2010 March 17 (see Table 1). The array was in the EW352 and H168 configuration, respectively, and we used the five inner antennas, discarding the sixth one located 4 km away as we obtained unsatisfactory pointing solutions and data quality. The maximum baseline length was then up to 352 m and 192 m in either configuration, respectively. With this configuration, PKS 1830211 was mostly unresolved, and is further considered as a point-like source. The weather was clear in both runs.
For each tuning, the bright quasar 1923293 was observed for several minutes to help us perform bandpass calibration, and the pointing was checked and updated against PKS 1830211, before integrating. We did not observe gain calibrators, as PKS 1830211 visibilities were self-calibrated and the continuum level normalized to unity.
The Compact Array Broadband Backend (CABB) system was configured to provide a 1 MHz spectral resolution per 2 GHz window with dual polarization. Two windows were used simultaneously, with central frequencies set as reported in Table 1. The whole 7 mm band (30–50 GHz) could then be covered in a fairly small number of individual tunings. Each part of the band was covered in at least two tunings, in order to prevent potential defectuous spectral channels. Some parts of the spectral band were indeed affected by a series of bad channels, as can be seen for example in the final spectrum between 40 and 40.2 GHz in Fig.12. Some frequency intervals were observed in more tunings, to increase the sensitivity, for example at the frequencies of the deuterated species DCO and DCN. The spectral resolution of 1 MHz corresponds to 10 km s at 30 GHz and 6 km s at 50 GHz.
The data reduction was performed in two steps. We used the MIRIAD package for a first bandpass calibration adopting the solution derived from 1921-293 and the self-calibration of the PKS 1830211 data. At this step, the analysis of each 2-GHz spectral window revealed a residual bandpass, with amplitudes roughly varying by an order of one percent over several 100 MHz. To remove this bandpass drift residuals – the final sensitivity being of about a few times only –, data were then exported to FITS format and subsequently processed with a customised ITT/IDL pipeline.
The pipeline is based on an iterative boxcar drift removal. Each iteration was designed to construct a model spectrum (i.e., a spectrum with any slow variations in the bandpass removed) from the raw data following several steps:
first, for each observed tuning, each polarization and each baseline, we flagged outliers in time by assuming that the visibilities were constant on timescales longer than ten seconds;
a drift component was then computed and removed, in both time and frequency with different windows widths, taking proper care of any flagged or missing data;
a first crude spectrum, after evaluating the time-averaged visibilities, was obtained and its residuals were used to find frequency channels with large deviations from a random Gaussian distribution;
at this stage, we also examined the spectra for each antenna, to check for possible antenna-based problems;
the standard deviation of each baseline was then used to compute a weighted mean spectrum and the uncertainty for each tuning and polarization;
finally, a Doppler track correction was applied according to the observatory velocities given in Table 1, obtained from the ATCA frequency calculator 222http://www.narrabri.atnf.csiro.au/observing/obstools/velo.html. All velocities in this paper refer to the Solar System barycenter rest-frame. We note that the Doppler correction is nearly constant during the short observing intervals for individual tunings. Data were accordingly re-gridded onto a common frequency scale using a one-dimensional (1D) drizzling algorithm (Fruchter & Hook 2002). We tested different frequency resolutions and chose a channel width of 0.5 MHz. A weighted mean was finally used to combine all spectra.
A model of the spectrum was derived by finding all frequency channels above a given threshold. This model was then removed from the data, and the whole process was iterated on the residual datasets.
We started with a large width for the bandpass residual filtering, corresponding roughly to 1 GHz, and a high threshold of 10 to derive the model. The pipeline was run until the variance of the resulting spectrum was found to have reached a minimum. The model threshold was lowered iteratively in steps of two down to 4 in order to correct for the negative corrections of the boxcar drift removal on strong lines. Following the same scheme, the filtering width was successively divided by two until it reached a width of 16 MHz.
This whole procedure ensures that the second order bandpass is removed, without making any a priori assumptions about the presence of strong absorption lines. The resulting spectrum is shown in Fig.12.
|2009 September 1|
|31 & 33||6:49||0.8|
|37 & 39||8:14||1.2|
|44 & 47.5||10:00||1.5|
|42 & 46.5||11:56||1.6|
|35 & 38||13:32||0.9|
|2009 September 2|
|32 & 34||6:44||0.8|
|41 & 45||8:08||1.2|
|46.5 & 49||10:19||1.4|
|43 & 47||11:55||1.0|
|36 & 38.5||13:10||0.7|
|2010 March 17|
|37.25 & 40||19:16||3.0|
|46.1 & 48.1||22:22||3.0|
3 Results and analysis
An overview of the 20-GHz-wide resulting spectrum is shown in Fig.2. The frequency coverage corresponds, in the rest frame of the =0.89 galaxy, to the range 57–94 GHz. The lower end of this frequency interval is relatively unexplored because of poor atmospheric transmission caused by a forest of fine-structure lines of O around 60 GHz. The rms noise level is not uniform over the whole band, and varies owing to the higher system temperatures at higher frequencies and the different integration times. The full survey is presented in Fig.12.
Only a handful of absorption lines reach an optical depth of 10% (see Eq.1) or more. They originate in the SW absorption component and correspond, in order of decreasing optical depth, to: HCO, HCN, HNC, CH, NH, HCO, c-CH, and HCO. All these species have been detected in previous observations (Wiklind & Combes 1998; Menten et al. 1999). The weaker NE absorption at km is detected only for HCO, HCN, HNC, CH, c-CH, and HCO.
The major result of this survey, though, is the detection of a collection of weak absorption lines toward the SW component, down to a few times 10 of the continuum intensity, as illustrated in Fig.3, where the cumulative number of detected lines is plotted as a function of opacity. These lines were identified using the on-line Cologne Database for Molecular Spectroscopy444http://www.astro.uni-koeln.de/cdms (CDMS, Müller et al. 2001), and Jet Propulsion Laboratory Molecular Spectroscopy Database 555http://spec.jpl.nasa.gov/home.html (JPL, Pickett et al. 1998). The corresponding transitions are listed in Table 11.
All detected transitions (except the transitions of CHCN, see §.B.14) are ground state or have low energy levels. Most molecules have high dipole moments, and the gas density is too moderate for collisions to play an efficient role in the excitation. Rotational transitions are therefore nearly in radiative equilibrium with the CMB, which has a temperature of 5 K at =0.89. Several species are detected in more than one transition, and a rotation diagram analysis indeed yields rotation temperatures in agreement with this value. A low energy level is therefore a good filter to identify transitions/species in the molecular databases, and this is why we believe that our identifications are robust, even for species with a single detected transitions. The confusion is also low, with a rough average of four clear lines per GHz over the 20-GHz-wide band.
The census of molecules detected to date in the =0.89 galaxy located in front of PKS 1830211 is given in Table 3. It is quite remarkable that, despite its distance, this =0.89 galaxy has now become the object with the largest number of detected molecules outside the Milky Way. A large fraction of the total number of these species (28 out of 34) are detected in our survey, among them 19 for the first time toward this source. As far as we know, this work also represents the first extragalactic detection for SO, l-CH, l-CH, HCCN, HCCO, CH, CHNH, and CHCHO. In addition, isotopologues of HCO, HCN, HNC, and SiO were also detected. Elemental isotopic ratios of C, N, O, and Si are derived and discussed in §4.1.
|2 atoms||3 atoms||4 atoms||5 atoms||6 atoms||7 atoms|
including HCO, HCO, and HCO; including HCN and HCN; including HNC and HNC; including SiO.
(a) Wiklind & Combes (1996b); (b) Gérin et al. (1997); (c) Wiklind & Combes (1998); (d) Chengalur et al. (1999); (e) Menten et al. (1999); (f) Muller et al. (2006), including isotopologues; (g) Henkel et al. (2008); (h) Menten et al. (2008); (i) Bottinelli et al. (2009); (j) Henkel et al. (2009).
3.1 Fitting procedure
We fitted a Gaussian profile to all the transitions of a given species with a single velocity and a single linewidth. The ortho and para forms, when relevant, were considered as the same species for the fit. The relative intensities of all species with resolved or marginally resolved hyperfine structure were fixed, as it was observed that, in the case of CH and HCO (for which the hyperfine structure is clearly resolved), they follow the predicted ratios assuming local thermal equilibrium (LTE). The velocity and linewidth of isotopic variants were fixed relative to the main isotopologue, except for HCO and HCN, the lines of which are likely saturated. For these, the corresponding isotopic variants were instead fixed relative to HCO and HCN. For some species with data for the various transitions of low signal-to-noise ratio (S/N), we decided to fix the velocity (e.g. l-CH) or the linewidth (e.g. SO) after initial free fits. Finally, the fit for CHCN is degenerate because of the blending of the =0 and =1 transitions, and for this species, we therefore chose to adjust a synthetic spectrum by eye. The fitting results are given in Table 11.
|Rotation temperature and column density obtained from Monte Carlo simulations of rotation diagrams. Uncertainties correspond to a 95% confidence interval.||Rotation temperature and column density obtained from Monte Carlo simulations of rotation diagrams. Uncertainties correspond to a 95% confidence interval.||Column density, assuming a rotation temperature of 5.14 K and LTE.||Note|
|(Debye)||For species with ortho/para symmetry, the two forms were considered as distinct species and the partition function Q was thus calculated accordingly, with the weight factors =3 for ortho form, and =1 for para form. Hyperfine structure was taken into account when necessary.||Number of lines detected or resolved (in the case of hyperfine structure).||(km s)||(km s)||(K)||( cm)||( cm)|
$e$$e$footnotetext: Possibly underestimated due to the large opacity of the line.
|(km s)||(km s)||(10 km s)||( cm)|
|(MHz)||(D)||(K)||(10 km s)||( cm)|
3.2 Time variations
A monitoring of the HCO =2-1 absorption profile between 1995 and 2008 revealed significant time variations (Muller & Guélin 2008), in particular with the quasi-disappearance of the km s absorption component between 2006 to 2007. In the HCO =2-1 spectrum acquired by Muller & Guélin in October 2008 with the Plateau de Bure interferometer, this component had returned with an absorption of 15% of the total continuum intensity. Variations were found to be correlated between the km s component and the blue wings of the 0 km s absorption, hence should originate from morphological changes in the background quasar.
From Very Long Baseline Interferometry continuum measurements, the relative distance between the NE and SW lensed images of the quasar was found to vary by as much as 200 micro-arcseconds within a few months (Jin et al. 2003). Nair et al. (2005) proposed that this apparent motion is caused by the sporadic ejection of plasmons along a helical jet by the quasar. Consequently, the continuum illumination scans different pencil beams through the foreground absorbing clouds, where the apparent displacement of the quasar lensed images corresponds to a projected distance of about a parcsec, i.e. a scale comparable to the size of Galactic molecular-cloud cores.
The relative differences (normalized to the averaged spectrum) for the spectral regions covered in both runs (September 2009 and March 2010) are shown together with the averaged spectrum in Fig.12. They are at most 5% for the HCO, HCN, and HNC lines, and barely visible for all other lines. Given the small relative changes, data from both observing sessions were combined to improve the S/N of the data.
3.3 Background continuum illumination
Absorption intensities (I), measured from the continuum level, were converted to optical depths () according to
where is the intensity of the background continuum source, and the source covering factor. At radio wavelengths, the continuum emission of the quasar PKS 1830211, lensed by the =0.89 intervening galaxy, is dominated by two bright and compact components (to the NE and SW), separated by 1” and included in an overall fainter Einstein ring. The flux ratio of the NE to SW compact components is 1.5–1.6 at cm wavelengths as measured by Subrahmanyan et al. (1990). The remaining emission (Einstein ring and fainter other components) contributes only to a few percent of the total continuum intensity.
Molecular absorption in the =0.89 galaxy is observed toward both the SW and NE continuum images. The HCO =2-1 absorption toward the SW image shows a flat-bottom profile, indicating saturation. Assuming that, at this wavelength, all the continuum emission comes from the NE and SW images of the quasar, it is then possible to derive the relative flux NE/SW of both images from the level of this saturation, without actually resolving them. The NE/SW ratio was measured to be about 1.7 with a dispersion of 0.3 over several observations between 1995 and 2007 (Muller & Guélin 2008), i.e. roughly similar to the flux ratio directly measured at cm wavelengths. Some of the fluctuations might be due to morphological changes in the quasar (plasmon burst, microlensing) and the time delay between the two line of sights.
Our 7 mm survey is limited in terms of spectral resolution, and the flat-bottom part of the HCO/HCN =1-0 lines (supposing that they are also saturated) is smooth, preventing the same analysis. Nevertheless, as the HCO absorption reaches an apparent depth of 35% and 8% toward the SW and NE components, relative to the total continuum level, we can certainly state that and . If, in addition, we impose the ratio NE/SW, we get the stronger constraints of and , which gives (NE)= and (SW)= with =1.6. We adopt these values. The rough uncertainty in the illuminating background continuum for both NE and SW components is therefore less than 10%. The residual contribution from components other than the NE and SW images is de facto less than , that is 9% with =1.6.
This is valid, of course, if no changes happened in the quasar within the time corresponding to the time delay between the two lines of sight (24 days) before our observations. Unfortunately, we do not have flux monitoring data corresponding to these epochs. Nevertheless, that we do not see significant variations in the absorption line profiles between the observations of September 2009 and March 2010 is reassuring.
We can now investigate how this uncertainty in the background continuum illumination propagates to the derivation of opacities, by means of Eq.1. In the case of weak lines (), we get , and our opacities are therefore estimated more accurately than 10%. For the other lines, we get , where the function , and tends to infinity for , although the rise is slow, as (that is ) is still . For example, the corresponding uncertainty in HNC opacity (the third strongest absorption line in our survey, after HCO and HCN) is less than 20%.
Finally, we assume here (as in the rest of the paper) a source covering factor of unity for all absorption components. It is however possible that and changes with frequency, as the continuum emission probably becomes more extended at lower frequencies (see also the discussion by Henkel et al. 2009). As a result, the optical depths derived from Eq.1 are, strictly speaking, lower limits. We note however that the HCO and HCN =2-1 lines toward the SW absorption are optically thick and saturated. The source covering factor for this component is thus certainly close to unity. Toward the NE absorption, the situation is not so clear. The formal observational constraints from the HCO 1-0 line are and . While the non-detection of HCO (with HCO/HCO71) suggests that the line is likely optically thin, the source covering factor is still only poorly constrained.
3.4 LTE analysis and rotation diagrams
For optically thin lines, and assuming a Boltzmann distribution characterized by a rotation temperature , column densities can be derived as
where is the partition function, the energy of the lower level with respect to ground state, the dipole moment, the frequency, and the line strength. Under these conditions, the column density of a detected molecule can be calculated directly with the knowledge of only one physical parameter: the rotation temperature .
Since some species are detected in several transitions, an excitation analysis can be done. Under the Rayleigh-Jeans (RJ) approximation (), Eq.2 can be re-arranged as
Plotting the quantity on the left side of this equation versus the energy of the lower level for each observed transition then allows the rotation temperature to be measured as the inverse of the slope of a linear fit to the data points in the rotation diagram. Column densities were also derived, following Eq.3, from the intercept of the y-axis in the rotation diagram. Because the uncertainties in the data points in the rotation diagrams do not follow a normal distribution, the rotation temperatures, column densities, and their corresponding uncertainties were estimated using Monte Carlo simulations. We note that the RJ approximation is still valid for K at 94 GHz, the highest rest-frame frequency reached in our survey.
The rotation diagrams for molecules detected toward FG0.89SW in several transitions are presented in Fig.4, and the resulting measurements of rotation temperatures and column densities are listed in Table 4. These rotation temperatures are plotted in Fig.5 as a function of the molecular dipole moment. The results for column densities toward FG0.89NE are given in Table 5.
Most of these species have high dipole moments, and consequently high critical densities (10 cm). For low density gas (10 cm), the population of the different energy levels is not efficiently thermalized by collisions, but is coupled with the ambient radiation field, which, in the absence of any other radiative excitation (UV or IR pumping), consists of CMB photons. In this case, rotation temperatures then nearly equal the CMB temperature, which is expected to be =5.14 K at =0.89, assuming a scaling proportional to . The derived rotation temperatures are consistent with this value. A thorough excitation analysis is beyond the scope of this paper, as it would require a careful treatment of collisional rates and partners (H and e) for each molecule. This will be the subject of a future paper, in which we will also study the excitation in FG0.89NE with new multi-transitions observations, in order to derive an accurate and robust measurement of the CMB temperature at =0.89 through the two independent lines of sight.
As a first step to interpreting the observations, we produced a LTE synthetic spectrum resulting from the combination of absorption lines from the molecular species listed in Table 4. We assume a common rotation temperature ==5.14 K for all species and for the sake of simplicity, the absorption profile is set to be Gaussian with a FWHM of 20 km s centered at =0 km s for the SW absorption (some lines are slightly offset from these values, see Table 4 and §4.3) and at = km s for the few species for which NE absorption is detected. The partition functions were calculated for =5.14 K by performing a direct sum over all energy levels. Their values are given in Table 4. Following Eq.2, the only free parameter is then the column density of each species. A kinetic temperature of 50 K was adopted to take into account the -ladder population of the symmetric-top molecule CHCN (see §B.14).
The resulting LTE spectrum, obtained with the LTE column densities listed in Table 4, is overlaid on top of the observed spectrum in Fig.12. Despite the simplicity of the model, it reproduces relatively well the observations.
In addition, we list in Table 6 the upper limits to the column densities and abundances obtained for some interstellar species of interest that have low-energy transitions within the frequency coverage of our survey. The upper limits to the integrated intensities were calculated as , where is the optical depth (Eq.1) noise level, is the velocity resolution, and the linewidth at half-maximum, which we fixed to 20 km s. The upper limits to the column densities were calculated from Eq.2, again assuming LTE and =5.14 K.
3.5 Additional velocity components
Previous observations have shown that the molecular absorption profile consists of two main velocity components, separated by 147 km s and located in front of the SW and NE images of the quasar (see Introduction). However, this picture should now be updated with the discovery of additional absorption components. Close to the redshifted frequencies of both the HCN and HCO 1-0 lines (redshifted to 47 and 47.3 GHz respectively) in Fig.12, several additional weak lines can be seen, which we failed to identify with other species. Those lines clearly arise at the same velocities for both HCN and HCO 1-0 transitions, as shown in Fig.6. They therefore correspond to additional velocity components of HCN and HCO 1-0 in the =0.89 galaxy. The new components are located at , , , and +170 km s (in addition to the previously known ones at 0 and km s). They probably arise in front of secondary continuum peaks along the Einstein ring (see for example Fig.3 by Chengalur et al. 1999). It would be interesting to determine their locations to constrain the kinematics of the galaxy, as well as their background continuum illumination to derive the associated opacities and column densities of gas along different sightlines through the disk. This is unfortunately impossible with the current data, which are limited in terms of angular resolution and sensitivity. Interestingly, the and +170 km s components have velocities beyond the range covered by HI absorption (Chengalur et al. 1999; Koopmans & de Bruyn 2005), and are nearly symmetrical with respect to the central velocity (80 km s from the velocity of the SW absorption). These two components may arise from rapidly rotating gas near the center of the galaxy, e.g., in a circumnuclear ring.
The advantages of our survey are manifold. First, the 4-mm rest-frame window (redshifted to 7 mm at =0.89), includes numerous ground state and low energy level transitions of a large range of intermediate-size interstellar molecules, and is therefore particularly interesting for an absorption study. Lighter molecules, e.g. hydrides, on the one hand, and heavier and more complex molecules (with atoms) on the other hand, generally have these transitions at higher (mm/submm window) and lower (cm wavelengths) frequencies, respectively. The achieved sensitivity, of a fraction of a percent relative to the continuum intensity, and the large spectral coverage (20 GHz band) then allows us to detect many weak lines of rare molecules (i.e. with abundances of a few 10 relative to H) and in several cases, to build their rotation diagrams. For both runs (September 2009 and March 2010), observations were carried out within one or two days, i.e. a timescale shorter than the time delay of the lens (24 days, Lovell et al. 1998; Wiklind & Combes 1999) and much smaller than the timescale for variations in the absorption profiles, of the order of several months (Muller & Guélin 2008). The comparison of lines in the frequency overlap between the two datasets shows that the relative variations between September 2009 and March 2010 are smaller than 5% for the strong HCO/HCN 1-0 lines and yet smaller for other lines. Time variations are therefore negligible, and moreover, would affect only the few lines in the small frequency overlap between both observing runs. Finally, the absorption technique provides a high spatial resolution, equivalent to the angular size of the quasar continuum images and corresponding to a few parsecs in the plane of FG0.89.
On the other hand, several limitations have to be kept in mind. First, the detection of molecular species in our survey (for a given abundance) is obviously biased towards those that have i) a relatively high dipole moment, ii) a relatively low partition function, and iii) low energy transitions within the frequency coverage (57–94 GHz, rest-frame). The molecular inventory is then limited by a combination of these properties. Second, the angular resolution of our observations is insufficient to resolve the background continuum emission. While the different absorption components are well-resolved kinematically, this introduces an uncertainty in the continuum illumination, hence also in the line opacities and column densities. We argue however that the associated uncertainties are only 10% for the bulk of the lines (see §.3.3). Moreover, the spectral resolution is only a factor of a few smaller than the width of the line (20 km s). This reduces the absorption line depths, likely causing opacities to be slightly underestimated. We expect that this affects weak lines the most. The limited spectral resolution also prevents us from measuring the saturation level of the HCO/HCN SW lines, which could help us to derive the NE/SW flux ratio. Finally, we assume a source covering factor of unity for each absorption component, implying that the derived optical depths and column densities are, strictly speaking, lower limits.
4.1 Isotopic ratios
Isotopic abundances are directly affected by nucleosynthesis processes in stellar interiors, and isotopic ratios are thus good probes of the chemical enrichment history of the Universe. For this purpose, molecular absorption offers interesting prospects, owing to the rotational transitions of different isotopologues being easily resolved at mm wavelengths, such that low column densities of the rare isotopologues are still detectable e.g. for most isotopes of C, N, O, and S elements, and that abundances can be inferred directly from absorption optical depths (if the line of the most abundant isotopologue is not saturated).
The C/C, N/N, O/O, O/O, and S/S isotopic ratios were derived in FG0.89SW by Muller et al. (2006), based on observations of the different isotopologues of HCO, HCN, HNC, CS, and HS at 3 mm with the Plateau de Bure interferometer. These ratios, particularly O/O and S/S, reveal significant differences when compared to sources in the Local Universe (see Table 7 from Muller et al. 2006). Interestingly, comparable values are found using the same absorption technique in the =0.68 galaxy located in front of the quasar B 0218+357 (Muller et al. in prep.). These results are essentially consistent with the expectation that low-mass stars (1.5 M) had no (or only a short) time to play a major role in the gas enrichment of such young (a few Gyr old) galaxies.
Our 7 mm survey now allows us to check the previous results obtained at 3 mm in FG0.89SW, in particular for lines with different optical depths. In addition, time variability is not a concern for this 7 mm dataset. Unfortunately, the limited spectral resolution of the survey prevents us from measuring ratios across the lines, which is a thorough test of possible saturation, and we simply derive the isotopic ratios from the ratios of integrated opacities.
The spectra of different isotopologues of HCO, HCN, HNC, and SiO as observed in our 7 mm survey are shown in Fig.7. Results for the corresponding isotopologue abundance ratios are given in Table 7, where we also report the ratios previously measured at 3 mm. Our final estimates for the various isotopic ratios in FG0.89SW are given in Table 8. The values obtained from 7 mm and 3 mm data are in very good agreement. We hereafter discuss how we derived those ratios.
We do not detect lines from any of the deuterated species DCO/DCN, down to levels of a few 10, which is higher than the D/H cosmic abundance of 2.510 (e.g. Spergel et al. 2003). This is consistent with no strong fractionation enhancement, as expected in a relatively warm gas component (50 K) with a moderate density, (H) of the order of a few 10 cm.
It is not straightforward to go from the abundance ratio C/C, where C is a given carbon-bearing molecule, to a value of the C/C isotopic ratio, because the measured abundance ratio can be affected by opacity effects, chemical fractionation and/or selective photodissociation. Hence, observations of several C- and C- bearing molecules are highly desirable.
The interstellar gas phase abundance of CO and C relative to CO and C, can be affected by the isotopic fractionation reaction
In cold molecular clouds, C then becomes mostly trapped in CO, and the abundance of the C ion decreases. The C/C ratio for species formed from CO should thus reflect the C/C isotopic ratio, while for others C/C C/C should be observed. This was first investigated by Langer et al. (1984) for various carbon-bearing species. They found that the C/C isotopic ratio should be within the limits set by the CO/CO and HCO/HCO ratios. Since HCO can be produced from both CO and other species, it can be enhanced in either C or C, depending on the physical conditions.
The C/C abundance ratios are listed in Table 7 for relevant species. The HCO/HCO ratio might be underestimated because of the large opacity of the C isotopologue. The opacity of HNC, on the other hand, is lower, and the HNC/HNC abundance ratio might more accurately reflect the C/C elemental ratio. We do not detect transitions from either the CCH or CCH isotopomers, which implies that the lower limits to the abundance ratios are CH/CCH58 and CH/CCH43. Sakai et al. (2010) found C abundance anomalies in the C-isotopomer of CH toward TMC 1 and L1527, with the C-species underabundant relative to the interstellar C/C ratio. In addition, the two carbon atoms do not appear to be equivalent in the formation pathways of the molecule, as Sakai et al. (2010) measured an abundance ratio CCH/CCH = 1.6 toward both sources. They conclude that carbon-chain molecules are not indicated to determine the C/C elemental ratio, because of the positional differences and heavy C dilution. The non-detection of HCO points to a ratio HCO/HCO173, suggesting that fractionation could affect our estimate of the C/C isotopic ratio. On the basis of all these results, we choose to adopt the average ratio C/C=, obtained from the HCO, HCN, and HNC species.
This value of C/C is less than half the ratio measured in the Solar System (89, Lodders 2003), but closer to that derived in the local interstellar medium (, Lucas & Liszt 1998) and especially close to that in the starburst galaxies NGC253 and NGC4945 (40–50, Henkel et al. 1993; Curran et al. 2001; Wang et al. 2004). It is, however, much lower than the value that would be expected in a poorly processed environment (e.g. Kobayashi et al. 2011). We note that Martín et al. (2010) used CH and CO isotopologues to revisit the C/C ratios in NGC253 and M82 and found values larger than the ratios measured in previous studies. The interstellar C/C isotopic ratio clearly remains difficult to determine.
Toward the NE component, we do detect neither the HCO nor the HCN 1-0 absorptions, with lower limits HCO/HCO71 and HCN/HCN41 (at 3 level), respectively. As the NE absorption is located at an even larger galactocentric distance than the SW, thus in a region likely to be even less processed, we expect a C/C ratio larger than that toward FG0.89SW.
There are two ways of estimating the N/N isotopic ratios from our data: either from the direct HCN/HCN and HNC/HNC abundance ratios, or, indirectly from the HCN/HCN and HNC/HNC ratios, after correcting for the C/C isotopic ratio.
We obtain the ratios HCN/HCN=, and HNC/HNC=340 . The former ratio might be underestimated because of opacity effects on the HCN line, but the second ratio has a large uncertainty. Using the C variants, we obtain HCN/HCN= and HNC/HNC=. The use of a large C/C ratio (50) gives a N/N ratio close to the value observed in the local interstellar medium (=, Lucas & Liszt 1998) and in the Solar System (=272, Lodders 2003). Using our estimate C/C= and the ratio HCN/HCN=, we derive N/N=.
Whether this value actually reflects the true N/N isotopic ratio is however uncertain if fractionation is an issue for nitrogen-bearing species (Rodgers & Charnley 2008a, b). The recent measurements of Lis et al. (2010) toward the two cold dense molecular clouds Barnard 1 and NGC1333 do not suggest a high N enhancement.
We note that the N isotopologues of NH are not detected, implying formally that N/N .
As for N/N, the O/O isotopic ratio can be estimated either directly from the HCO/HCO abundance ratio, or indirectly from HCO/HCO. The former method yields a ratio O/O=, while the latter implies that O/=. Even considering that our value of C/C could be underestimated, the O/O isotopic ratio derived in FG0.89SW is much lower than that in both the local interstellar medium (=, Lucas & Liszt 1998) and the Solar System (=499, Lodders 2003), in likely connection with the youth of the galaxy.
Remarkably, the O/O isotopic ratio, measured directly from the HCO and HCO =1-0 optically thin lines, also differs significantly from values for sources in the Local Universe (, see e.g. Table 7 of Muller et al. 2006). While a value of O/O= was measured from the =2-1 lines, the marginal detection of HCO at 7 mm (Fig.7) yields HCO/HCO=. This new estimate of the O/O isotopic ratio confirms the large value previously obtained from the =2-1 lines at 3 mm.
The detection of the SiO and SiO =2-1 line allows us to estimate the isotopic ratio Si/Si=11 . Both SiO lines are optically thin, unless the source covering factor is much lower than unity. This Si/Si ratio at =0.89 is nearly half the value measured in the Solar System, 19.6 (Lodders 2003).
The situation is similar to that for the S/S isotopic ratio, estimated to be from the CS/CS and HS/HS abundance ratios in FG0.89SW (Muller et al. 2006), whereas it is 22 in the Solar System. Interestingly, the magnesium isotopic ratios were derived from UV-spectroscopy for a =0.45 absorption line system (Agafonova et al. 2011), and also suggest that there is a significant relative overabundance of heavy Mg isotopes. Silicon and sulfur (and magnesium) are both produced by massive stars. That the neutron-enriched isotopes appear to be more abundant at =0.89 than in the Local Universe might provide some constraint on their nucleosynthesis.
|Abundance ratios||@7 mm||@3 mm|
|Isotopic ratios||7 mm||3 mm||Averaged|
4.2 Comparative chemistry
It is interesting to compare the molecular abundances in the =0.89 galaxy with those measured for various sources in the Local Universe, to characterize the type of clouds and chemistry. For this purpose, we selected a sample of archetype sources with a large number of detected molecular species, usually covered by large spectral surveys. Molecular abundances in circumstellar envelopes around evolved stars, such as IRC+10216, were not included because of the distinctive chemical segregation within the envelope. We preferred to use data obtained from a single team/telescope, for the sake of homogeneity and to minimize beam effects.
Despite the low density and poor shielding from the interstellar radiation field, a variety of molecules is already present in Galactic diffuse clouds (): about 15 different species have been detected so far (see e.g. Liszt et al. 2008), and the limit of their chemical complexity remains unclear. We mention here that column densities in diffuse clouds are most often determined through mm-wave absorption toward extragalactic continuum sources, similar to the work presented in this paper.
Translucent clouds (25) are expected to have a richer chemistry, with about 30 molecules detected so far, including highly unsaturated and very reactive hydrocarbon chains. We adopt the molecular abundances compiled by Turner (2000).
Cold dark clouds are the sites of low-mass star formation, where an intricate gas-phase ion-molecule interstellar chemistry takes place. We adopt the abundances determined by Ohishi et al. (1992) toward TMC-1, which is a prototype of dark clouds.
Hot and dense cores, associated with massive-star formation harbor the most complex interstellar chemistry. There, the chemistry is characterized by the evaporation of dust grains, releasing large molecules in the gas phase. We adopt the abundances derived by Nummelin et al. (2000) in the region Sgr B2N toward the Galactic center as reference.
Finally, we also include in our comparison three extragalactic sources for which molecular abundances have been derived for a significantly large number of species: the nuclear regions of the starburst galaxies NGC253 and NGC4945 (Martín et al. 2006 and Wang et al. 2004, respectively), as well as the star forming region N113 in the Large Magellanic Cloud (LMC, Wang et al. 2009).
The comparison of abundances derived in FG0.89SW with those observed toward these various sources is illustrated in Fig.8. Molecular abundance ratios, normalized to FG0.89SW and detailed by species, are also given in Fig.9. Several species in our survey, such as HCO, HCN, CH, and c-CH are commonly observed in Galactic diffuse clouds. More complex molecules (e.g. carbon chains) are only seen toward cold dark clouds or in warm and dense clouds.
Comparing molecular abundances relies on the validity of various hypotheses. First, molecules have to be co-spatial and share the same source covering factor. The size of the SW and NE compact continuum components corresponds to a few pc in the plane of the =0.89 galaxy. We assume a common source covering factor of unity for all molecules. However, the absorbing gas could be the mix of (most likely several) dense cores embedded in a more diffuse component within the continuum beam illumination. Our derived column densities and molecular abundances are thus lower limits. In addition, opacity effects, if not correctly taken into account, could easily lead to erroneous molecular abundances. For this reason, we do not use the abundances of HCO, HCN, and CH from Nummelin et al. (2000) toward Sgr B2N, as they are probably largely underestimated. Finally, a reference species needs to be assigned for normalization, in order to compare the relative molecular abundances. We choose to normalize the column densities to that of molecular hydrogen, as commonly done in the literature. For this, we assume a value (H)= cm toward the SW component, as estimated by Wiklind & Combes (1998). This value is consistent with the measurement of (H)=10 cm obtained from X-ray absorption by Mathur & Nair (1997) assuming that all absorption is due to the SW component. The absorbing gas is indeed mostly molecular along this line of sight (Chengalur et al. 1999), so that (H) should be close to 2(H). The NE line of sight, on the other hand, intercepts the disk of the =0.89 galaxy at a galactocentric distance of 4 kpc. There, the extinction is roughly one order of magnitude lower than toward the SW image of the quasar (e.g. Falco et al. 1999, see also Fig.1), and the column density of molecular gas is likely to be similarly lower. We assume a H column density of 110 cm toward FG0.89NE. This choice makes the fractional abundances similar to those in Galactic diffuse clouds (Table 9 and Fig.9). While the normalization to H introduces an uncertainty in the absolute molecular abundances, abundance ratios for a given source are not affected, and should therefore bear greater significance (again provided that the species are co-spatial and have an identical source covering factor).
To compare the overall trend between the various sources, and inspired by Martín et al. (2006), we have calculated an abundance estimator , defined as the average of the logarithm of abundance ratios between two sources and
where stands for the abundance relative to H of molecule in source , and the sum runs over all molecules detected in common for the two sources and . A dispersion in the estimator was also calculated to reflect potential large variations in the abundance ratios between species. We do not take into account non-detections in this estimator. Abundance estimators were also calculated in the same manner for the following sub-samples of molecules: carbon chains (i.e. with more than two chained carbon atoms, such as l-CH, CH, c-CH, l-CH, and HCN), sulfur-bearing species (i.e. SO, NS, HCS, CS, and SO), and saturated molecules (CHOH and CHNH), presumably all formed on dust grains. The value of was calculated between FG0.89SW and the other sources only if the number of common species for two objects was 2. The abundance estimators are given in Table 10 and illustrated in Fig.10.
On the basis of the abundance estimators including all species, we obtain a basic sequence of increasing molecular abundances such as LMC diffuse clouds FG0.89NE NGC253 FG0.89SW NGC4945 TMC 1 translucent clouds Sgr B2N, indicating, unsurprisingly, that the chemical abundances in FG0.89SW are in-between those in typical Galactic diffuse and translucent clouds.
The dispersions for Sgr B2N are large, though it is clear that sulfur-bearing and especially saturated species have much higher abundances in this molecular-rich object. In TMC 1 and translucent clouds, the abundance of carbon chains is globally a factor of a few higher than in FG0.89SW, while that of sulfur-bearing species is one order of magnitude higher.
The overall similarity between the molecular abundances of FG0.89SW and the nuclear region of the starburst galaxy NGC253 is surprising. Martín et al. (2006) found good agreement between the molecular abundances in NGC253 and those at the position Sgr B2(OH) in the Galactic center molecular cloud Sgr B2. Shocks are thought to play an important role in the heating and chemistry toward these nuclear clouds, triggering the disruption of dust grains, and releasing the products of grain-surface chemistry into the gas phase. It is tempting to invoke similar chemical processes in FG0.89SW, as they could explain all of the similarities between the molecular abundances, the high kinetic temperature of the gas, and the relatively large linewidth (see also discussions in Henkel et al. 2008; Menten et al. 2008). Moreover, the elemental composition of the gas in both galaxies is expected to be dominated by the nucleosynthesis products of massive stars, because of their shorter timescales. The major difference, though, is that FG0.89SW clouds are located in a spiral arm at a galactocentric distance of about 2 kpc, and gas densities are likely lower. Abundances in NGC4945 show some differences, albeit small, when compared to those in NGC253. They are interpreted as different starburst evolutionary states (Wang et al. 2004; Martín et al. 2006). The systematically lower molecular abundances in the LMC-N113 region are attributed to a photon dominated region (PDR) in a nitrogen deficient environment (Wang et al. 2009). Nevertheless, we emphasize that, because of the limited angular resolution of these extragalactic observations, the corresponding molecular abundances reflect the average of the different types of cloud gas properties and excitation conditions within the beam. This is much less the case for absorption line studies toward quasars, where the spatial resolution is set by the small size of the background continuum emission.
It will be interesting to investigate the chemical complexity toward FG0.89SW when searching for more species, especially light hydrides and larger molecules. The molecular absorber located at =0.68 in front of the quasar B 0218+357 would also be a suitable target for these investigations.
|Species||FG0.89SW||FG0.89NE||Diffuse||Translucent||TMC 1||Sgr B2(N)||LMC||NGC253||NGC4945|
Diffuse clouds: Lucas & Liszt (2000a, b); Liszt & Lucas (2001); Lucas & Liszt (2002); Liszt et al. (2004); Translucent clouds: Turner (2000); TMC-1: Ohishi et al. (1992); SgrB2(N): Nummelin et al. (2000); LMC-N113: Wang et al. (2009), using H)= cm; NGC253: Martín et al. (2006); NGC4945: Wang et al. (2004).