© ESO, 2010

1. Introduction

The Sun was usually assumed to be formed from the material characteristic of the local physical conditions in the Galaxy at the time of its formation and, therefore, to represent a standard chemical composition. However, this homogeneity hypothesis has often been questioned because of many improvements in the observation techniques and data analysis. With the discovery of extrasolar planetary systems, the study of heterogeneity sources (e.g. stellar formation process, stellar nucleosynthesis and evolution, collisions with molecular clouds, radial migration of stars in the Galactic disc) has gained a new perspective and brought new questions.

It is now undisputed that dwarf stars hosting giant planets are on average richer in metal content than stars in the solar neighbourhood for which no planet has been detected (see e.g. Fischer & Valenti 2005; Gonzalez 2006; Santos et al. 2001, 2004). Two hypotheses have been suggested in an attempt to explain the origin of this anomaly: i) primordial hypothesis: the chemical abundances presently observed represent those of the protostellar cloud from which the star was formed; ii) self-enrichment hypothesis: a significant amount of material enriched in metals is accreted by the star during the planetary system formation and evolution.

It has been speculated that abundance anomalies between dwarf stars with and without planets may not only involve the metal content of heavy elements, but also the abundance of light elements such as carbon and oxygen, which are measured by an overabundance in the ratio [X/Fe] of one stellar group compared to another in a given metallicity range. Gonzalez & Laws (2000) found that [Na/Fe] and [C/Fe] in stars with planets are on average lower than in stars without detected planets, for the same metallicity. Numerical simulations performed by Robinson et al. (2006) predicted an overabundance of [O/Fe] in planet-host stars. The same result for this element was obtained by Ecuvillon et al. (2006), although the authors noticed that it is not clear if this difference is caused by the presence of planets.

In other publications, however, no difference was found in the abundance ratios of light elements when comparing stars with and without planets (Ecuvillon et al. 2004a,b; Luck & Heiter 2006; Takeda & Honda 2005). In particular, in more recent studies, Gonzalez & Laws (2007) and Bond et al. (2008) do not confirm the overabundance of [O/Fe] in planet-host stars obtained by Ecuvillon et al. (2006) and Robinson et al. (2006), showing that a solution for the problem is not simple.

Most of the studies cited above are inconclusive and the discussion about this problem remains open. New tests are encouraged, using more precise and homogeneous data, with as many stars as possible. We show the results of our analysis of high-quality spectra, in which we determined photospheric parameters and carbon abundances for 172 G and K stars, including 18 planet hosts. Although this kind of investigation is not new, the spectra analysed here offer the possibility to perform an homogeneous and accurate study of the chemical anomalies that have been proposed in the literature and will surely help to distinguish the different stellar and planetary formation processes.

The abundance distribution of light elements in stars more evolved than the Sun, hosting planets or not, were also studied. Takeda et al. (2008) analysed a large sample of late-G giants, including a few stars hosting planets. For a solar metallicity, the authors found an underabundance of [C/Fe] and [O/Fe] and an overabundance of [Na/Fe] in the atmosphere of their sample stars compared to previous results for dwarf stars, which they attributed to evolution-induced mixing of H-burning products in the envelope of evolved stars.

Our sample comprises 63 dwarfs (of which 7 have planets), 13 subgiants (4 with planets), and 96 giants (7 with planets). This allowed us to investigate possible anomalies in the abundance ratios of stars more evolved than the Sun. Except for HD 7924, whose planet has a minimum mass of 9 Earth masses (about half that of Neptune), the planet-host stars in question come from systems that have at least one giant planet.

The data and the reduction process are presented in Sect. 2. The determination of the photospheric parameters and their uncertainties are described in Sect. 3. In Sect. 4 we describe the spectral synthesis method used to obtain the carbon abundances and their uncertainties. Our results are presented and discussed in Sect. 5, and final remarks and conclusions are made in Sect. 6.

2. Observation data and reduction

Our sample consists of 172 G and K dwarf, subgiant, and giant stars in the solar neighbourhood (distance  <100 pc) observed with the ELODIE high-resolution échelle spectrograph (Baranne et al. 1996) of the Haute Provence Observatory (France). The analysis was done based on spectra that were publicly available in the ELODIE archive (Moultaka et al. 2004) when the work started. The spectra have resolution R = 42   000 and cover the wavelength range 3895−6815 Å. The resulting sample stars were selected according to the following criteria:

• i)

  stars for which the averaged spectra have S/N ≥ 200; among all individual spectra available in the database, only those with a S/N ≥ 30 and with an image type classified as object fibre only (OBJO) were used;

• ii)

  stars within a distance ≤100 pc (parallax π ≥  10 mas) and with spectral type between F8 and M1; earlier type stars have a small number of spectral lines, whereas dwarfs later than M 1 are too faint to provide good quality spectra, and they are also quite cold (exhibiting a lot of strong molecular features, such as the TiO bands), which renders the determination of precise abundances difficult;

• iii)

  stars for which no close binary companion is known, since these objects may contaminate the observed spectra; we used the information of the angular separation between components (rho) available in the Hipparcos catalogue (ESA 1997), choosing only the cases with rho > 10 arcsec.

• iv)

  stars for which the determination of the photospheric parameters (Sect.  3 ) is reliable;

• v)

  stars with (B − V) values measured by Hipparcos and with spectral cross-correlation parameters available in the ELODIE database; both (B − V) and the width of the cross-correlation function are required in the estimate of the stellar projected rotation velocity vsini (see Sect. 4); and

• vi)

  stars that passed the quality control of the spectral synthesis (see Sect.  4).

The selected sample is plotted in the HR and colour-distance diagrams of Fig. 1, which shows separately the subsamples of dwarfs, subgiants, and giants. The transition boundaries between dwarfs and subgiants and also between subgiants and giants are not clearly defined on an observational HR plane. We chose to classify as subgiants those stars situated 1.5 mag above the lower limit of the main-sequence and having M_v > 2.0 mag. Note that the distance of dwarfs and subgiants is not limited to 100 pc, but to about 60 pc. This is not imposed by our selection criteria, but represents a selection effect of the ELODIE observation surveys.

For each sample star, the spectra available in the ELODIE database were processed using IRAF¹ routines. First, they were normalised (a general pre-normalisation) based on continuum windows selected in the wavelength range. Then the normalised spectra were corrected for the Doppler effect, i.e., transformed to a rest wavelength scale taking the solar spectrum as reference, with a precision of better than 0.02 Å in the correction. After these two first steps, the spectra were averaged to reduce noise. Finally, a more careful normalisation was done, this time only considering a small wavelength region around the spectral features analysed here: for the molecular bands the range 5100−5225 Å was used, and for the C atomic line the wavelength range was 5330−5430 Å. At this point, the stellar spectra were ready to be used by the synthesis method.

3. Determination of photospheric parameters

A precise and homogeneous determination of chemical abundances in stars depends on the calculation of realistic model atmospheres, which in turn depends on accurate stellar photospheric parameters: the effective temperature T_eff, the metallicity [Fe/H], the surface gravity log g, and the micro-turbulence velocity ξ. We developed a code that uses these four parameters as input and, iteratively changing their values, the code tries to find a solution for the model atmosphere and metal abundance that are physically acceptable.

The abundance yielded by different spectral lines of the same element should not depend on their excitation potential (χ) or their equivalent width (EW). Also, neutral and ionised lines of the same element should provide the same abundance. Therefore, the effective temperature was computed through the excitation equilibrium of neutral iron by removing any dependence in the [Fe i/H] versus χ diagram. Additionally, by removing any dependence of [Fe i/H] on EW, the micro-turbulence velocity was estimated. The surface gravity was computed through the ionisation equilibrium between Fe i and Fe ii, and the metallicity was yielded by the EW of Fe i lines. In other words, the photospheric parameters were determined following the conditions below: $\begin{array}{ccc}\begin{array}{c}\\ & \\ & \\ & \\ & \end{array}& & \end{array}$where c₁, c₂, c₃, and c₄ are arbitrary constants as small as one wishes. If at least one of the first three conditions is not satisfied, then T_eff, ξ, and/or log g are changed by a given step. In the fourth condition, the value of metallicity [Fe/H] used as input is compared to the one provided by Fe i lines and, if this condition is not satisfied, the code defines [Fe/H] = [Fe i/H]. Therefore, the code iteratively executes several cycles until these four conditions are satisfied at the same time.

Atomic line parameters (wavelength, oscillator strength gf, and lower-level excitation potential χ) for 72 Fe i and 12 Fe ii lines used in our analysis are listed in Table 1. They were all taken from the Vienna Atomic Line Database – VALD (Kupka et al. 2000, 1999; Piskunov et al. 1995; Ryabchikova et al. 1997), though the gf values were revised to fit the EW measured in the Kurucz Solar Flux Atlas (Kurucz et al. 1984), along with a model atmosphere for T_eff = 5777 K, log g = 4.44, ξ = 1.0 km s^-1, and log ϵ_⊙ = 7.47 (the solar Fe abundance). Concerning the calculations of the van der Waals line damping parameters, we adopted the Unsöld approximation multiplied by 6.3.

The stellar model atmospheres used in the spectroscopic analysis are those interpolated in a grid derived by Kurucz (1993) for stars with T_eff from 3500 to 50 000 K, log g from 0.0 to 5.0 dex, and [Fe/H] from −5.0 to 1.0 dex. These are plan-parallel and LTE models, computed over 72 layers. For each layer, the quantities column density (ρx), temperature (T), gas pressure (P_g), electronic density (N_e), and Rosseland mean opacity (κ_Ross) are listed. The models also include the micro-turbulence velocity, the elemental abundances in the format log ϵ (where log ϵ _star = log ϵ_⊙  +  [Fe/H]), both assumed to be constant in all layers, and a list of molecules used in the molecular equilibrium computation. Although the Kurucz models used here were computed for a micro-turbulence velocity ξ = 2 km s^-1, in our iterative computation of the photospheric parameters, ξ was set as a free parameter instead. In our opinion this does not significantly affect the chemical analysis performed here, since their uncertainties are dominated by the errors in the other photospheric parameters.

The equivalent widths were measured using the Automatic Routine for line Equivalent widths in stellar Spectra – ARES (Sousa et al. 2007). In order to test the reliability of the automatic measurements, we performed a comparison between EW measured in the solar spectrum (the sunlight reflected by the Moon) using ARES and those measured one by one using IRAF tasks (see Fig. 2). Notice that both procedures provide EW that are consistent with each other within the uncertainties (the absolute differences are smaller than 1.5 mÅ).

3.1. Uncertainties in the photospheric parameters

We developed a routine that iteratively estimates the uncertainties in the computed photospheric parameters of each star. The procedure is as follows:

• i)

  first, the micro-turbulence velocity is increased (decreased) by a given step and new model atmospheres are computed; the change proceeds iteratively until the angular coefficient of the linear regression in the [Fe i/H] versus EW diagram is of the same order as its standard error; the absolute differences between the increased (decreased) and best values provide the ξ upper (lower) limits; the uncertainty σ(ξ) is an average of lower and upper values;

• ii)

  next, a similar procedure is used for the effective temperature; it is iteratively changed until the angular coefficient of the linear regression in the [Fe i/H] versus χ diagram is of the same order as its standard error; since micro-turbulence velocity and effective temperature are not independent of each other, the uncertainty σ(ξ) estimated above is taken into account before changing T_eff; thus, the absolute differences between changed and best values of T_eff provide the uncertainty σ(T_eff), with the effect of σ(ξ) properly removed;

• iii)

  the uncertainty in the metallicity σ([Fe/H]) is the standard deviation of the abundances yielded by individual Fe i lines;

• iv)

  finally, the uncertainty in the surface gravity σ(log g) is estimated by iteratively changing its value until the difference between the iron abundance provided by Fe i and Fe ii is of the same order as σ([Fe/H]).

4. Carbon abundance from spectral synthesis

Spectral synthesis was performed to reproduce the observed spectra of the sample stars and thus determine their carbon abundance. The technique was applied to molecular lines of electronic-vibrational head bands of the C₂ Swan System within spectral regions centred at λ5128 and λ5165 as well as to a C atomic line at λ5380.3. The atomic line at λ5052.2 is also commonly used as a C abundance indicator, but we preferred not to use it because this line is blended with a strong Fe line, which may affect the abundance determination, particularly for C-poor stars. Using the MOOG spectral synthesis code (Sneden 2002), synthetic spectra based on atomic and molecular lines were computed in wavelength steps of 0.01 Å, also considering the continuum opacity contribution in ranges of 0.5 Å and line-broadening corrections, and then fitted to the observed spectra.

To compute a theoretical spectrum, the MOOG requires besides a model atmosphere for each star some parameters of atomic and molecular spectral lines, which come from the VALD online database and from Kurucz (1995), respectively, and some convolution parameters related to spectral line profiles.

In addition to C₂, another molecule that contributes to the spectral line formation in the studied wavelength regions is MgH, although its contribution is relatively small. The gf of C₂ and MgH lines from the Kurucz database were revised according to the normalisation of the Hönl-London factors (Whiting & Nicholls 1974). The gf values of atomic and other molecular lines were also revised when needed to fit the solar spectrum, taken as a reference in our differential chemical analysis.

The parameters of all atomic and molecular lines used to compute the synthetic spectra of the studied regions are listed in Tables 2 and 3, which are available in electronic form at the CDS. Table 2 contains the following information: the wavelength of the spectral feature, the atomic and molecular line identification, the lower-level excitation potential, the oscillator strength, and the dissociation energy D₀ (only for molecular features). Table 3 (strong atomic lines) contains the same information of Table 2, excepting the dissociation energy parameter.

The convolution parameters responsible for spectral line broadening that are important to our analysis are: i) the spectroscopic instrumental broadening; ii) a composite of velocity fields, such as rotation velocity and macro-turbulence, which we named V_broad; and iii) the limb darkening of the stellar disc. We estimated the instrumental broadening by measuring the Full Width at Half Maximum (FWHM) of thorium lines in a Th-Ar spectrum observed with ELODIE. As a first estimate of V_broad, the projected rotation velocity vsini of the star was used, which was computed according to Queloz et al. (1998). Then, small corrections in V_broad based on an eye-trained inspection of the spectral synthesis fit were applied when needed. Concerning the stellar limb darkening, an estimate of the linear coefficient (u) was performed by interpolating T_eff and log g in Table 1 of Díaz-Cordovés (1995), and it ranges from 0.63 to 0.83 for the stars in our sample.

Figure 3 shows the spectral synthesis method applied to the observed data. In this illustrative example, the spectrum of the sunlight reflected by the Moon is plotted, showing how reliable the reproduction of the Sun’s spectrum is. Synthetic spectra were computed for the molecular band heads around λ5128 and λ5165, and the C atomic line at λ5380.3, in steps of 0.01 Å, but resampled in steps of 0.05 Å in order to consistently match the observed spectrum wavelength scale.

4.1. Uncertainties in the C abundances

The three wavelength regions investigated provide an independent determination of C abundance and its respective uncertainty. In order to estimate the uncertainties due to the errors in the photospheric parameters, we developed a routine that takes into account the error propagation of input parameters used by the MOOG spectral synthesis code, namely, T_eff, log g, [Fe/H], ξ, and V_broad. Each one in turn, the MOOG input parameters are iteratively changed by their errors, and new values of the abundance ratio [C/Fe] are computed. The difference between new and best determination provides the uncertainty due to each parameter. The uncertainty σ([C/Fe]) is a quadratic sum of individual contributions. The error in V_broad was estimated to be of the order of 1 km s^-1 or smaller. The error in the limb darkening coefficient was not considered since its contribution can be neglected. In any case, the uncertainty in the C abundances obtained here is dominated by the errors in T_eff, log g, and [Fe/H].

5. Results and discussion

The photospheric parameters and C abundances obtained in the present work and their uncertainties are listed in Tables 4–6 for dwarfs, subgiants, and giants, respectively. Table 7, available in electronic form at the CDS, contains the individual [C/Fe] determinations provided by the three abundance indicators: the star name, the [C/Fe] abundance ratio and its uncertainty yielded by the C₂ molecular band indicator around λ5128, [C/Fe] and its uncertainty computed from the λ5165 indicator, and [C/Fe] and its uncertainty computed from the λ5380.3 indicator.

Both molecular and atomic indicators agree quite well, providing a rms of 0.06 dex in the C abundances. When comparing the two molecular indicators λ5128 and λ5165, a rms of only 0.02 dex is found. The atomic indicator λ5380.3 is a quite weak line (EW ~ 20 mÅ in the Sun’s spectrum) and could explain the larger dispersion. Nonetheless, we notice that the final C abundances listed in Tables 4–6 are the result of a weighted average of the abundances yielded by the three C abundance indicators, and that the weights are inversely proportional to the individual uncertainties in each determination.

Figure 4 shows a comparison of the photospheric parameters obtained in this work to those published by other works having stars in common. Effective temperature, metallicity, and surface gravity of dwarf, subgiant, and giant stars are compared with the results of Hekker & Meléndez (2007), Luck & Heiter (2006, 2007), Takeda et al. (2008), and Valenti & Fischer (2005). The errors in the difference this work  −  comparison paper are a quadratic sum of the errors in our photospheric parameters and those published by the papers. Typical values are represented by error bars plotted in each panel of the figure. The micro-turbulence velocity comparisons are not shown in Fig. 4, but our determination is consistent within the uncertainties with the publications above, for which an estimate of this parameter was also performed.

The photospheric parameters in the comparison papers are also based on an homogeneous analysis of high signal-to-noise ratio and high-resolution data. We can observe in Fig. 4 a very good agreement between our estimates and different determinations. An exception are the surface gravity values of Takeda et al. (2008), which are systematically lower than in our work. The authors also found that their log g determination is systematically lower than in previous studies, which appears to be caused by a different set of spectral lines used, as they suggested.

For the Sun, our estimate for the photospheric parameters is: T_eff = 5724  ±  38 K, log g = 4.37  ±  0.10, [Fe/H] = −0.03 ± 0.03, and ξ = 0.87 ± 0.05 km s^-1. The broadening velocity was set to V_broad = 1.8 km s^-1, and the linear limb darkening coefficient, u = 0.69, was obtained in the same way as for the other stars. These values were used to compute the solar model atmosphere, which in turn was used to obtain the solar value of the C abundance: [C/Fe] = 0.01  ±  0.01. Again, as for the other stars, this is the result of a weighted average of the abundances yielded by the three C abundance indicators, where the weights are inversely proportional to the individual uncertainties in each determination.

Figure 5 shows a few examples of spectral synthesis applied to different stars: a cold giant star, a hot dwarf star, and a Fe-poor and high-rotation giant star. The good fit of the synthetic spectra to the observed data in these examples demonstrates that the spectral synthesis method used here provides reliable results for the different spectral types of our stellar sample.

Figure 6 shows our results of the carbon abundance plotted in the form of abundance ratios: [C/H] and [C/Fe] in function of metallicity, and [C/Fe] distributions. Both diagrams and histograms compare the C abundance of planet-host stars with the abundance of stars for which no planet has been detected. To clarify the visualisation and simplify the discussion, the three subsamples are presented separately: dwarfs in the top panels, subgiants in the middle panels, and giants in the bottom panels. Choosing a metallicity range in which both the stars with and without planets are equally represented (−0.4  < [Fe/H]  < 0.4) and computing the [C/Fe] mean and standard deviation, we have: i) −0.02 ± 0.04 and −0.03 ± 0.05 respectively for dwarfs with and without planets; ii) −0.06 ± 0.07 and −0.03 ± 0.05 respectively for subgiants with and without planets; and iii) −0.13 ± 0.07 and −0.17 ± 0.07 respectively for giants with and without planets. Although it seems that planet-host giants are, on average, richer in [C/Fe] than giants without planets (especially regarding the histogram), according to these values, there is no indication that in all subsamples stars with and without planets share different C abundance ratios. In addition, applying a t-test for unequal sample sizes and equal variance, we obtain that the [C/Fe] distributions are indistinguishable with respect to the presence or the absence of planets. These results support the primordial hypothesis discussed in Sect. 1 instead of self-enrichment.

Figure 6 also shows that [C/Fe] is clearly depleted (by about 0.14 dex) in the atmosphere of giants compared with dwarf stars. This agrees with the results of Takeda et al. (2008) and Liu et al. (2010), which they attributed to evolution-induced mixing of H-burning products in the envelope of evolved stars in the sense that carbon-deficient material, produced by the CN-cycle, would be dredged up to the stellar photosphere.

Our [C/Fe] determinations for dwarfs are somewhat lower than in other works (Ecuvillon et al. 2004b; Gonzalez & Laws 2000; Reddy et al. 2003). We found that the Sun is slightly overabundant in carbon compared to other dwarfs in the same metallicity range, which is the opposite situation found by those authors. Luck & Heiter (2006, 2007) analysed a large sample of stars also using the spectral synthesis method in the determination of C abundances. In our study we have several stars in common with their papers, and a comparison is shown in Fig. 7. A systematic difference can be observed: our C determination passes from overabundant to underabundant with increasing [C/Fe], at least for dwarfs and giants. We notice, however, that the differences appear mostly from −0.1 to  +0.1 dex, and are compatible with typical uncertainties.

On the other hand, other recent studies corroborate our results in the sense that the Sun seems to be overabundant in carbon with respect to other solar metallicity dwarfs (Ramírez et al. 2009). We have only five stars in common with this work, which are also shown in Fig. 7. In addition, Fig. 1 in their paper for the [C/Fe] abundance ratio is very similar to Fig. 6 for the [C/Fe] distribution of dwarfs in this paper.

It is not unexpected to find systematic differences between samples analysed by different methods: different model atmospheres, or a different set of atomic and molecular lines could produce offsets and trends with regard to other works. Here, stars with planets were compared with their analogues without planets, and dwarfs and subgiants were compared with giants, and they were all analysed using the same method. Therefore, any possible offset that may exist among different works will not affect the analysis of our differential comparison among these subsamples or the conclusions that we draw.

5.1. Kinematics properties

Kinematic properties of the entire sample was considered to investigate the Galaxy population membership. Computation of the kinematics required astrometry (parallaxes and proper motions) and radial velocities. The astrometry was taken from the new reduction of the Hipparcos catalogue (van Leeuwen 2007) and the values of radial velocities were measured from the spectra. The space velocities (U, V, and W) were computed with respect to the local standard of rest (LSR), where the solar motion (U, V, W) = (10.0, 5.3, 7.2) km s^-1 was adopted (see Dehnen & Binney 1998).

With the kinematic data, we have grouped the entire sample into three main populations: thin disc, thick disc, and halo. The probability that a given star in the sample belongs to one of the three populations is computed based on the procedure outlined in Reddy et al. (2006) and references therein. A star whose probability P_thin, P_thick, or P_halo is greater than or equal to 75% is considered as thin, thick, or halo star, respectively. If the probabilities are in-between, they are considered as either thin/thick disc or thick/halo stars. Out of 172 stars in the sample, the vast majority (162) are of the thin disc population and a few metal-poor stars are of the thick disc (just 4). HD 10780 is the only halo star in the sample. With the exception of one thick disc giant, all planet hosts in our sample are thin disc members. Population groups are indicated in Tables 4–6 and also in Fig. 6. The C abundance results seem to be indistinguishable among the different populations.

6. Conclusions

The results presented here represent an homogeneous determination of photospheric parameters and carbon abundances for a large number of G and K stars, comprising 63 dwarfs, 13 subgiants, and 96 giants, of which 18 have already had at least one detected planetary companion at the time of developing this work (mostly giant planets indeed). Our analysis used high signal-to-noise ratio and high-resolution spectra that are public available in the ELODIE online database. We derived the photospheric parameters through the excitation potential, equivalent widths, and ionisation equilibrium of iron lines selected in the spectra. In order to compute the C abundances, we performed spectral synthesis applied to two C₂ molecular head bands (λ5128 and λ5165) and one C atomic line (λ5380.3).

The photospheric parameters estimated here (T_eff, log g, [Fe/H], and ξ) agree very well with several works that have stars in common with our sample. These comparison samples were also analysed based on high signal-to-noise ratio and high-resolution data. Our estimates are the result of a precise and homogeneous study, both required conditions to compute reliable model atmospheres used in abundance determinations. Concerning the C abundances, our results point out that:

• i)

  regarding the subsamples of dwarfs, subgiants, and giants, there is no clear indication that stars with and without planets have different [C/Fe] or [C/H] abundance distributions;

• ii)

  [C/Fe] is clearly underabundant (by about 0.14 dex) in the atmosphere of giants compared with dwarf stars, which is probably the result of carbon-deficient material, produced by the CN-cycle, dredged-up to the envelope of evolved stars; subgiant stars, although in small number, seem to follow the same behaviour of dwarfs; and

• iii)

  the Sun is slightly overabundant in carbon in comparison to other dwarf stars with the same metallicity.

The first of the above results are based on small-number statistics. In order to draw more reliable conclusions, a larger number of planet-host stars is required, covering a metallicity range as large as possible. Adding more elements to the study, e.g. nitrogen, oxygen, and some refractory metals, would also expand the analysis to a larger context. Indeed, the investigation of volatile and refractory elements with respect to the distribution of their abundances in function of the condensation temperature (T_C) will shed light on recent controversies aroused by Chavero et al. (2010). The flat distribution found by these authors should be confirmed with more precise abundance determination for T_C ≲ 300 K (which includes C, N, and O).

The systematic differences in [C/Fe] found between this and other works are probably related to different analysis methods employed to compute the abundances: model atmospheres, the atomic and molecular lines used, etc. Nevertheless, this will not affect our main results because they were based on differential comparisons among subsamples analysed with the same approach.

In this work, we also considered the kinematic properties of our sample to investigate C abundances among different population groups. The stars were separated according to their Galaxy population membership: thin disc, thick disc, or halo stars. We found that most of these stars are members of the thin disc. Moreover, appart from one thick disc star, all planet-host stars are thin disc members. This is probably related either to the fact that giant planets are normally not much detected in less metal-rich stars (the thick disc members indeed) or to the fact that the observation samples are usually limited in distance, which naturally selects thin disc stars.

Online material

---

Acknowledgments

R. Da Silva thanks the Instituto Nacional de Pesquisas Espaciais (INPE) for its support.

References

All Tables

All Figures