Be Abundance in Li-Rich Giants

ASTRONOMY AND ASTROPHYSICS Astron. Astrophys. 345, 249–255 (1999) Beryllium abundance in lithium-rich giants⋆ B.V. Castilho1 , F. Spite2 , B. Barbuy1 , M. Spite2 , J.R. De Medeiros3 , and J. Gregorio-Hetem1 1 2 3 Universidade de São Paulo, CP 3386, São Paulo, SP, 01060-970, Brazil Observatoire de Paris-Meudon, DASGAL, URA 335 CNRS, F-92195 Meudon Cedex, France Universidade Federal do Rio Grande do Norte, Natal, RN, 59072-970, Brazil Received 26 November 1998 / Accepted 2 February 1999 Abstract. About 2% of the Population I red giants show lithium abundances significantly larger than expected by dilution due to mixing by classical convection and, in some of them, the lithium abundance reaches values similar to (and even larger than) the Pop I value (meteoritic, open clusters etc.) around logN(Li) = 3.3. The classical convection predicts also a (smaller) dilution of beryllium and (an even smaller one) of boron. Two main interpretations of the Li-rich giants are possible: – the initial Li has been somehow preserved (perhaps by the inhibition of the classical mixing) – on the contrary, a mixing, deeper than the classical one, took place in some (or all?) giants, leading to a dilution of the superficial lithium but sometimes overcompensating for it by an internal production of lithium, with transport to the surface (by the Cameron-Fowler mechanism). In the first interpretation, Be (more robust than Li) is a fortiori preserved, in the second one, a mixing deeper than the classical one should dilute Be more than the classical Be dilution. We have observed the Be II λ 3130.420 and 3131.066 Å lines in two Li-rich giants and three reference stars: one Li-poor giant and α Cen A and B. The observations were carried out at the ESO 3.6m telescope using the CASPEC spectrograph. By comparing the observed spectra with spectrum synthesis calculations we show that for the three giant stars the Be abundance is low, suggesting that, in these giants, Be is very depleted (>90%) from the initial Pop I value (logN(Be)∼1.4). This result implies that the original Li in these stars must have been almost completely destroyed, and that the high Li abundances in the Li-rich red giants is most probably due to Li production in these stars. Key words: nuclear reactions, nucleosynthesis, abundances – stars: abundances – stars: late-type Send offprint requests to: B.V. Castilho ⋆ Observations collected at the European Southern Observatory – ESO, Chile 1. Introduction The internal structure of the evolved stars is not well known. Lithium, Beryllium and Boron are fragile elements, selectively destroyed in the deep (hot) layers of the stars. Lithium is the most fragile element, and is preserved in a thin external layer. Beryllium is less fragile, and is preserved in a thicker zone, and boron in an even deeper zone. A mixing of the external layers with deeper ones will lead to a depletion of the fragile elements in the external layers. A shallow mixing will deplete essentially lithium, a deeper one will deplete also Be, and a still deeper one will even deplete B. Gathering simultaneous data on both Be and Li is a powerful tool for a better understanding of stellar structure. Such a tool will be useful for the analysis of the intriguing problem of the spread of the lithium abundance in supergiants and giants: this spread is not understood and its study should help to check the processes of lithium depletion, and the possible lithium production in giants and supergiants. Observation of BeII lines in two Li-rich giants using IUE data of limited resolution and S/N gave encouraging indication that Be is not highly preserved in these giants (De Medeiros et al. 1997). The galactic lithium production itself is an important pending problem (Matteucci et al. 1995, Lemoine et al. 1998), linked with the primordial nucleosynthesis. Sackmann & Boothroyd (1999) presented the first models for the production of Li in low mass normal red giants. The intriguing problem of the lithium abundance spread in giants has been widely discussed in the literature since the work of Brown et al. (1989), without any satisfactory conclusion. A dilution by a factor of 60 is expected from the standard convective mixing model (Iben 1967) but the observations show that lower values of the lithium abundance are often found (even down by an additional factor of 50). In general terms, lithium is observed to be severely depleted in most giants. More recent work (see Castilho et al. 1998 and references therein) has been successful in the discovery of a number of Lirich stars. Let us note, that in some of them, the lithium reaches a value similar to (and even larger than) the lithium abundance generally observed for the Population I objects; those giants, through their wind could contribute to the enrichment of the interstellar medium. 250 B.V. Castilho et al.: Beryllium abundance in lithium-rich giants Several interpretations may be considered: 1- A combination of destruction in the MS phase with further dilution in the giant phase: An interpretation of the low Li abundance found in most giants could be that the depletion is the standard dilution, by classical convective mixing (Iben 1967), but that most stars have in addition suffered a previous lithium destruction in the main sequence phase: a)- in the domain of the Boesgaard-Tripicco’s dip (Boesgaard & Tripicco 1986) Li may be strongly destroyed (Balachandran 1990), but Be is only slightly depleted by the lithium dip’s process (Garcı́a López et al. 1995) so that the final Be abundance would be only slightly lower than predicted by the classical convection, provided that the line is observable. b)- in a way similar to the (not yet well understood) solar lithium depletion on the main sequence. However, very few giants in the solar neighborhood will have solar masses, and even then the solar depletion of Be is moderate if existing at all (cf. Balachandran & Bell 1998). Here again, Be would be observable. 2- Li production: Another interpretation would be that lithium is severely depleted in all giants, but that some variable Li production occurs in a few giants. The severe depletion would be produced by a mixing deeper than the classical convection (first dredge-up, standard model). This problem has been tackled by Charbonnel (1994; 1995) who finds a deep mixing in giants of small masses (smaller than about 2 M⊙ ): this mixing should also dilute Be (more moderately); so that the Be observation could confirm (or infirm) this interpretation. The problem of the higher mass giants remains, but most Li-rich giants in the RGB seem to have small masses (da Silva et al. 1995). 3- Li preservation: An opposite interpretation would be a variable inhibition of the deep mixing: if the low level of Li observed in most of the giant stars is due to a mixing larger than the one predicted by the first dredge-up standard model, an inhibition of this mixing in a few of the giants could preserve their lithium. The Be would then be a fortiori preserved and observable. A simple preservation would not by itself explain the few red giants with a lithium abundance larger than the Pop I abundance. The aim of this work is the determination of the Be abundance in Li-rich and Li-poor giants, in order to discriminate between the possible interpretations, obtaining also some information about the mixing in the red giants. In Sect. 2 the observations and data reduction are reported, in Sect. 3 the calculations are described, in Sect. 4 the results are given, in Sect. 5 a discussion is presented and in Sect. 6 the conclusions are drawn. Table 1. Log of observations Target mv Spec. type Date Exp. S/N Comments (min.) α Cen A α Cen B HD 220321 HD 148650 HD 787 η Cen −0.01 1.33 3.97 5.94 5.25 2.31 G2 V K1 IV K0 III K3 III K4 III B1 V 26.06.97 26.06.97 25.06.97 25.06.97 26.06.97 25.06.97 2×5 2×12 1×90 6×90 4×55 2×10 180 40 45 20 25 40 ref. star ref. star ref. giant Li-rich giant Li-rich giant B star Table 2. Adopted stellar parameters for program stars Star Te f f log(g) [Fe/H] vm t ref. α Cen A α Cen B HD 220321 HD 146850 HD 787 5800 5350 4490 4000 3890 4.40 4.50 2.73 1.50 1.74 0.10 0.10 −0.40 −0.30 0.03 1.0 1.0 1.3 1.6 1.5 1 1 2 3 2 Note: 1 Primas et al. (1997), 2 McWilliam (1990), 3 Castilho et al. (1995) were used. In all observations we employed the filter ESO # 4 (UG5) which is 90% transparent in the UV, and cuts light redder than 4200 Å in order to avoid light contamination by saturation in the redder orders. The CCD ESO # 37, a Tektronix TK1024 of 1024x1024 pixels, with pixel size of 24 µm was used as detector. The slit was chosen to have 180x350 µm, corresponding to 2.” 4 on the sky, and (with a scale of 0.33′′ /pixel in the direction of the dispersion) to a resolving power R ∼ 32000. The slit was aligned in the direction of atmospheric dispersion (parallactic angle), corresponding to the position of the star at half the exposure time. An external Th-Ar lamp was necessary since the light of the internal lamp crosses a prism opaque in the UV. The data were reduced using the IRAF ECHELLE package. Flatfield corrections were not applied because there was no lamp available for the UV; but we find by inspecting the spectrum of the B1 star that the reduction of such images can be performed without flatfield division, since the CCD response is smooth (see also Molaro et al. 1997). Spectra from different exposures were added by weighting them with (S/N)2 . The final S/N ratio for the giants goes from 40 to 20, due mainly to the low atmospheric transmission and low CCD sensitivity at these wavelengths. The spectrum of η Cen was inspected for telluric lines. The log of observations and final S/N ratios are given in Table 1. 3. Calculations 2. Observations and data reduction 3.1. Stellar parameters and models Observations were carried out using the 3.6m ESO telescope with the CASPEC Spectrograph equipped with the long camera centered at λ 3640 Å (order 156), covering the orders 131 to 190. The BeII λ 3130 Å lines are found at the order 181. The échelle grating of 31.6 l/mm and a cross-disperser of 300 l/mm Stellar parameters for the program stars were adopted from the literature, as indicated in Table 2. Model atmospheres employed have been interpolated in tables computed with the MARCS code by Gustafsson et al. (1975) and Edvardsson et al. (1993). The α Cen A and B synthetic spectra were also computed with B.V. Castilho et al.: Beryllium abundance in lithium-rich giants models by Kurucz (1992) and we obtain a very good agreement with the above calculations. 251 1.0 0.8 3.2. Atomic and molecular line lists Several authors discussed the list of the absorption lines which are present in this crowded region, and chose different ways of fitting, correcting and merging the incomplete data available in the literature. Our list of atomic lines was built trying to adopt only the most reliable data, using the list of identified lines by Moore et al. (1966), and including the updated list and accurate data for Fe I by Nave et al. (1994). Oscillator strengths for atomic lines are adopted from Nave et al. (1994), Wiese et al. (1969), Fuhr et al. (1988) and Martin et al. (1988) whenever available otherwise they were obtained by fitting the solar spectrum. We used the solar atmospheric model by Edvardsson et al. (1993) and the Kurucz solar atlas (Kurucz et al. 1984) and the abundances are adopted from Grevesse et al. (1996). The molecular lines of the following molecules were taken into account in the calculations: MgH (A2 Π-X2 Σ), C2 (A3 ΠX3 Π), CN blue (B2 Σ-X2 Σ), CH (A2 ∆-X2 Π), CH (B2 ∆-X2 Π), CH (C2 Σ-X2 Π), OH (A2 Σ-X2 Π), NH (A3 Π-X3 Σ). Franck-Condon factors with dependence on the rotational quantum number J as given in Dwiwedi et al. (1978) and Bell et al. (1979) were computed and adopted where possible. For vibrational bands for which such values were not available, we adopted a constant value kindly made available to us through computations by P. D. Singh (Singh 1998, unpublished). For all molecular systems the line lists by R. Kurucz (CD ROM 18) were adopted, where we recomputed the ‘molecular oscillator strengths’, by recomputing their Honl-London factors using the formulae by Kovacs (1979), using the FranckCondon factors as indicated above, and employing literature values for the electronic oscillator strengths fel . We have adopted the fel (CN blue) = 0.0338 (Duric et al. 1978), fel (C2 ) = 0.033 (Kirby et al. 1979), fel (CH) = 5.257E-3 for the (A2 ∆-X2 Π) system (Brzozowski et al. 1976), fel (CH) = 2.5E-3 for the CH (B2 ∆-X2 Π) system (Grevesse & Sauval 1973), and fel (CH) = 5.95E-3 for the CH (C2 Σ-X2 Π) (Lambert 1978), fel (OH) = 8.0E-4 and fel (NH) = 8.0E-3 (Grevesse & Sauval 1973) and dissociation potentials Do (CN) = 7.65 eV, Do (C2 ) = 6.21 eV, Do (CH) = 3.46 eV, Do (OH) = 4.392 eV, Do (NH) = 3.47 eV (Huber & Herzberg 1979). To fit the solar spectrum with the logN(Be)=1.15 (Grevesse et al. 1996), it was necessary to make slight modifications of the derived log gf values of the BeII lines. The adopted values are -0.308 for the λ 3130.420 Å line and -0.508 for the λ 3131.066 Å line. By calculating the synthetic spectrum with the Kurucz (1992) solar model, the BeII lines are adequately fitted with log gf -0.168 and -0.468 respectively. To fit the blue wing of the BeII λ 3131.066 Å (avoiding large modifications of the log gf of the surrounding lines, as found in the literature) we chose to include a FeI line at λ 3130.995 Å with χex =3.00 eV and log(gf)=-3.30. The behaviour of this line in the fits of α Cen A and B proved to be appropriate. 0.6 0.4 0.2 Be II 0.0 3129.0 3129.5 3130.0 3130.5 Be II 3131.0 3131.5 3132.0 3132.5 3133.0 Wavelength (Å) Fig. 1. Our synthetic spectra for the Sun (- - -), overimposed to the Kurucz Solar Atlas (—-). The damping constants for the neutral element lines was calculated using tables of cross sections by Barklem et al. (1998 and references therein), and for the other lines they were determined by fitting the solar spectrum. Therefore we have built an atomic and molecular data base, available upon request. In Fig. 1 we show our synthetic spectra for the Sun, overimposed to the Kurucz Solar Atlas (Kurucz et al. 1984). No multiplicative or additive scaling was used in the Solar Atlas or the computed spectra. 3.3. Spectrum synthesis An updated version of the code by Spite (1967), extended to include molecular lines by Barbuy (1982), where LTE is assumed, is used for the spectrum synthesis calculations. This code (FSYNTH) now includes atomic and molecular lines from UV to near-IR and the computations are made faster. For α Cen A and B we calculated the synthetic spectra using the solar chemical composition, scaling to the adopted metallicity ([M/H]=0.1). Small changes for some element abundances (≤0.15 dex) were applied in order to fit the spectra. For the giants we used published abundances available for some elements (Castilho et al. 1995 for HD 146850, and McWilliam 1990 for HD 787 and HD 220321) and solar composition for the other elements. A few elements for which the abundances were not available were slightly modified. The uncertainties in Be abundance, as given in Table 3, are due to the S/N ratio. The intrinsic error in the abundance calculation is mostly due to the log gf and is estimated to be of 0.10 dex. We estimate that the uncertainties due to NLTE effects on the formation of the Be lines are small. The analysis of these effects for dwarfs by Garcı́a López et al. (1995) shows that the NLTE corrections are lower than 0.1 dex, and for the subgiant HD 140283 it is even smaller, so that the NLTE cannot explain the difference between a dwarf like α Cen A and a giant. The main source of uncertainty is due to the relatively low S/N ratios. 252 B.V. Castilho et al.: Beryllium abundance in lithium-rich giants Table 3. Be abundances (present work) and Li abundances (literature) HD 146850 Star logN(Be) logN(Li) α Cen A α Cen B HD 220321 HD 146850 HD 787 Sun 1.20±0.1 0.80±0.2 0.40±0.3 −0,50±0.4 0.00±0.4 1.15±0.1(6) 1.37(1) ≤0.4(2) ≤-0.2(3) 1.6 (1.9nlte)(4) 2.2 (3.1nlte)(5) 1.16±0.10(6) HD 787 HD 220321 Note: 1 King et al. (1997b), 2 Chmielewski et al. (1992), 3 Brown et al. (1989), 4 Castilho et al. (1995), 5 de la Reza & da Silva (1995), 6 Grevesse et al. (1996) α Cen A Be II Be II Table 4. Expected Li abundances from Be depletion 3130.0 Star present Be fraction logN(Li) from solar [Be/Li] Fig. 2. Comparison between α Cen A (bottom) where we can see the Be II lines and the other three giants where the lines are nearly absent. Sun α Cen A α Cen B HD 220321 HD 146850 HD 787 0.0071 0.0079 0.0020 0.0013 0.0001 0.0005 – 1.21 0.62 0.41 −0.49 0.01 4. Results In Table 3 we report the presently derived Be abundances and literature values of Li abundances, including for reference the Be and Li abundances of the Sun according to Grevesse et al. (1996). We observed α Cen A and B as comparison stars. The derived Be abundances agree within the uncertainties (illustrated in Fig. 3) with previous determinations of N(Be) = 1.21 and 0.61 (Primas et al. 1997) or N(Be) = 1.35 and 1.20 (King et al. 1997a). The determinations for the giants were made using the same techniques of data reduction, the same atomic and molecular constants, and the same grid of MARCS models. The Be abundance estimated for the three red giant stars (Table 3) show that the Be abundance has been very depleted (by >90% i. e. by a factor of 10 or larger) from the initial Pop I value (logN(Be)= 1.4). By computing the Be abundance for HD 220321 and HD 787 using the atmospheric parameters used by Brown et al. (1989) Teff = 4510 K and 4220 K; log(g)=2.3 and 1.5 and [Fe/H]=-0.32 and 0.07, we find Be abundances of logN(Be)= 0.1 and -0.2 respectively. In Table 4 we show the present observed Be fraction (from Primas et al. (1997) for α Cen A and B and from Table 3 for the giants) with respect to the meteoritic value (2nd column), and the corresponding Li abundances (3rd column) deduced from the observed Be depletion by assuming the same depletion ratio of Be relative to Li as in the Sun. These estimations show large Li depletions for the three observed giants, that agree with the observed Li abundance for the low Li abundance giant HD 220321 but are in complete discordance with the two Li-rich giants. 3130.2 3130.4 3130.6 3130.8 3131.0 3131.2 3131.4 Wavelength (Å) It is important to note however that this Li/Be depletion ratio is correct for dwarfs of solar mass (as shown by the agreement of the computations with the observed values for α Cen) - the ratio may be different for giants -, and that Balachandran & Bell (1998) have found that the solar Be abundance is, in fact, not depleted with respect to the meteoritic value of N(Be) = 1.42 dex. In this case the Li depletion with respect to Be will be even larger. Our results for Be show good agreement with recent calculations by C. Charbonnel, carried out for giants of [Fe/H] = 0.0 and -0.5 and masses between 1.2 and 2.0 M⊙ , where mean values of Be/Be⊙ ∼ 0.1 and Li/Li⊙ ∼ 0.05; our depletions for Li are larger than predicted by the models, but an extra Li depletion is expected. In Fig. 2 we compare the observed spectrum of α Cen A, where we can see the Be II lines, to the other three giants where the Be II lines are essentially absent. In Figs. 3 to 7 we show the Be spectral region together with synthetic spectra for different Be abundances respectively for α Cen A and B, HD 220321, HD 146850 and HD 787. 5. Discussion The small Be abundance found in the Li-rich giants suggests a Be depletion, since the Be depletion found in the giants is larger than the one found in the region of the dip. Moreover, it is not likely that these two stars had by chance a low Be initial abundance, in spite of the fact that the Be abundance in Pop I stars shows some spread. Therefore, Be is not preserved, and its depletion should be due to a rather deep mixing, implying that the original Li in these stars must have been strongly depleted, as in the case of HD 220321. As a consequence, the high Li abundance found in the two Be-poor Li-rich giants analysed here (and in all Li-rich giants) is due to a further Li production (cf. Sackmann & Boothroyd 1999). Recently, two Hyades giants have been analyzed by Duncan et al. (1998). They have a lithium depletion by a factor of 120, B.V. Castilho et al.: Beryllium abundance in lithium-rich giants 253 0.8 0.6 0.6 0.4 0.4 0.2 0.2 Be II 3130.0 3130.5 Be II 3131.0 Be II 3131.5 3132.0 0.0 3130.0 3130.5 Be II 3131.0 3131.5 3132.0 Wavelength (Å) Wavelength (Å) Fig. 3. α Cen A (+ + +), synthetic spectra with logN(Be) = 1.10, 1.20, 1.30 (—–). The best fit with logN(Be)=1.20 agrees with logN(Be)=1.21 from HST observations (Primas et al. 1997). Fig. 6. HD 146850: Observed spectrum (+ + +) and synthetic spectra (—–). computed with Be abundances: logN(Be) = −1.0, −0.5 (best fit), 0.0, and 1.2. 0.8 0.4 0.6 0.4 0.2 0.2 Be II 0.0 3130.0 3130.5 Be II Be II 3131.0 3131.5 3132.0 0.0 3130.0 3130.5 Be II 3131.0 3131.5 3132.0 Wavelength (Å) Wavelength (Å) Fig. 4. α Cen B (+ + +) together with synthetic spectra for three Be abundances: logN(Be) = 0.50, 0.80 (best fit) and 1.10 (—–). Fig. 7. HD 787: Observed spectrum (+ + +) and synthetic spectra (—–) computed with Be abundances: logN(Be) = −0.4, 0.0 (best fit), 0.4 and 1.2. 0.8 0.6 0.4 0.2 Be II 0.0 3130.0 3130.5 Be II 3131.0 3131.5 3132.0 Wavelength (Å) Fig. 5. HD 220321: Observed spectrum (+ + +) and synthetic spectra (—–) computed with Be abundances: logN(Be) = 0.1, 0.3, 0.5 and 0.7. A best fit is for logN(Be) = 0.4 similar to the solar depletion. In fact, in the sample of Brown et al. (1989) the depletions are generally even larger. The boron is found depleted by a factor of 10: the mixing responsible for the lithium depletion has been deep enough for depleting significantly the boron, implying a Be depletion by an even larger factor. In fact, a NLTE analysis of the lithium abundance shows a depletion of only ≈ 80. This analysis suggests that the relatively moderate lithium depletion in the Hyades giants has been obtained by a deep mixing, deep enough for depleting Boron and, a fortiori Be (the three depletions of Li, Be and B are found compatible with the Yale model of stellar evolution by Pinsonneault et al. 1992). Both results concur to suggest that rather deep mixing occurs generally in Li-poor giants, implying depletions of Li, Be and B. The occurrence of a high Li abundance in Be-poor giants suggests some lithium production, compensating (and sometimes overcompensating) the previous depletion. 254 B.V. Castilho et al.: Beryllium abundance in lithium-rich giants Except for the Li abundance, the Li-rich giants are normal red giants. No correlation was found between mass, rotation or 12 C/13 C ratio and Li abundance (da Silva et al. 1995; De Medeiros et al. 1996a). Observations with the CORAVEL spectrometer indicate that HD 146850 is a binary system, very probably with a double-lined behaviour, noting however that this was not detectable in our spectroscopic observations (Castilho et al. 1997). A projected rotational velocity V sin i of 2.8 km s−1 is found for this star, which seems to be the first known Li-rich giant presenting indication of binarity. The rotational velocity for HD 146850 can be considered normal in comparison with the observed rotation for giant stars. As shown by De Medeiros et al. (1996b) the mean rotational velocity for K3 III giants (the assigned spectral type for HD 146850) is near 2.0 km s−1 . For the other two lithium-rich stars of our sample, CORAVEL observations show no sign of binarity. These facts suggest that the Li-rich giants are not a particular class of stars, but ordinary low mass stars, observed during a short phase of their evolution, when Li is created (Castilho 1995). Sackmann & Boothroyd (1999) demonstrated that 7 Li can be created in low mass red giant stars via the CameronFowler mechanism, due to non-classical deep mixing and the associated “cool bottom processing” yielding photospheric abundances like the ones observed in the Li-rich red giants. Observed Li abundances in red giants can even reach values larger than the Pop I value, such as the case of HD 19745 (de la Reza & da Silva 1995). If indeed all low mass red giants go through a phase of Li production in the RGB (that could be cyclic), together with an increase in mass loss seen in the IRAS colour diagram (Gregorio-Hetem et al. 1993, de la Reza et al. 1996), and since some of them reach a lithium abundance larger than the Pop I abundance, they could be an important source of Li enrichment in the Galaxy. 6. Conclusion The observations presented here suggest that the lithium abundance distribution observed in the red giants is best explained by a classical mixing, deep enough for depleting B, Be and Li, and probably deep enough for inducing (sometimes, or in delimited evolutionary phases) a variable lithium production, confirming the computations of extra-mixing in low-mass giants published by Charbonnel (1998) and of cool bottom processing in such stars recently presented by Sackmann & Boothroyd (1999). However, in the limited telescope time attributed to this project, it was possible to observe only two Li-rich giants, and owing to the importance of a better understanding of extra mixing processes in stars, and lithium production in the Galaxy, the observation of some more Li-rich giants would be desirable. Acknowledgements. We are grateful for useful discussions with C. Charbonnel, F. Primas, and J. Melendez. We thank Matthew Shetronne and the ESO 3.6m team for the preparation of the instrument setup for these observations, and the referee Dr. Y. Chmielewski for the careful reading of the manuscript. We acknowledge partial financial support from CNPq and Fapesp (Brazil) and CNRS (France). BC acknowledges the FAPESP PhD fellowship no 97/007814-4. References Balachandran S., 1990, ApJ 354, 310 Balachandran S., Bell R.A., 1998, Nat 392, 791 Barbuy B., 1982, Ph.D. Thesis, Univ. Paris VII Barklem P.S., O’Mara B.J., Ross J.E., 1998, MNRAS 290, 102 Bell R.A., Dwiwedi P.H., Branch D., Huffaker J.N., 1979, ApJS 41, 593 Boesgaard A.M., Tripicco M.J., 1986, ApJ 303, 724 Brzozowski J., Bunker P., Elander N., Erman P., 1976, ApJ 207, 414 Brown J.A., Sneden C., Lambert D.L., Dutchover E., 1989, ApJS 71, 293 Castilho B.V. 1995, MsC Thesis, IAG, Universidade de São Paulo Castilho B.V., Gregorio-Hetem J., Barbuy B., 1995, A&A 297, 503 Castilho B.V., Gregorio-Hetem J., Barbuy B., 1997, Journal of Astronomical Data 3, 1 Castilho B.V., Gregorio-Hetem J., Spite F., Spite M., Barbuy B., 1998, A&AS 127, 139 Chmielewski Y., Friel E., Cayrel de Strobel G., Bentolila C., 1992, A&A 263, 219 Charbonnel C., 1994, A&A 282, 811 Charbonnel C., 1995, ApJ 453, L41 Charbonnel C., 1998, Space Science Rev. 84, 199 da Silva L., de la Reza J.R., Barbuy B., 1995, ApJ 448, L41 de la Reza J.R., da Silva L., 1995, ApJ 439, 917 de la Reza R., Drake N., da Silva L., 1996, ApJ 456, L115 De Medeiros J. R., Lèbre A., de Garcia Maia M.R., Monier R., 1997, A&A 321, L37 De Medeiros J.R., Melo C.H.F., Mayor M., 1996a, A&A 309, 465 De Medeiros J.R., Da Rocha C., Mayor M., 1996b, A&A 314, 499 Duncan D.K., Peterson R.C., Thorburn J.A., Pinsonneault M.H., 1998, ApJ 499, 871 Duric N., Erman P., Larson M., 1978, Phys. Scripta 18, 39 Dwiwedi P.H., Branch D., Huffaker J.N., Bell R.A., 1978, ApJS 36, 573 Edvardsson B., Andersen J., Gustafsson B., et al., 1993, A&A 275, 101 Fuhr J.R., Martin G.A., Wiese W.L., 1988, Atomic Transition Probabilities: Iron through Nickel. Journal of Physical and Chemical Reference Data vol. 17, suppl. no. 4 Garcı́a López R.J., Rebolo R., Pérez de Taoro M.R., 1995, A&A 302, 184 Garcı́a López R.J., Severino G., Gomez M.T., 1995, A&A 297, 787 Gregorio-Hetem J., Castilho B.V., Barbuy B., 1993, A&A 268, L25 Grevesse N., Sauval J., 1973, A&A 27, 29 Grevesse N., Noels A., Sauval J., 1996, In: Holt S.S., Sonneborn G. (eds.) ASP Conf. Ser. 99, p. 117 Gustafsson B., Bell R.A., Eriksson K., Nordlund A., 1975, A&A 42, 407 Huber K.P., Herzberg G., 1979, Constants of Diatomic Molecules. van Nostrand Reinhold, New York Iben I. Jr., 1967, ApJ 147, 624 King J.R., Deliyannis C.P., Boesgaard A.M., 1997a, ApJ 478, 778 King J.R., Deliyannis C.P., Hiltgen D.D., et al., 1997b, AJ 113, 1871 Kirby K., Saxon R.P., Liu B., 1979, ApJ 231, 637 Kovacs I., 1979, Rotational Structure in the Spectra of Diatomic Molecules. Am. Elsevier Pub. Co. Kurucz R.L., 1992, In: Barbuy B., Renzini A. (eds.) IAU Symp. 149, Kluwer Acad. Press, p. 225 Kurucz R.L., Furenlid I., Brault J.W., Testerman L., 1984, Solar Flux Atlas from 296 to 1200 nm, NSO, Tucson, Atlas no. 1 Lambert D.L., 1978, MNRAS 182, 249 Lemoine M., Vangioni-Flam E., Casse M., 1998, ApJ 499, 735 B.V. Castilho et al.: Beryllium abundance in lithium-rich giants McWilliam A., 1990, ApJS 74, 1075 Matteucci F., D’Antona F., Timmes F.X., 1995, A&A 303, 460 Martin G.A., Fuhr J.R., Wiese W.L., 1988, Atomic Transition Probabilities: Scandium through Manganese. Journal of Physical and Chemical Reference Data vol. 17, suppl. no. 3 Molaro P., Bonifacio P., Castelli F., Pasquini L., 1997, A&A 319, 593 Moore C.E., Minnaert M.G., Houtgast J., 1966, The Solar Spectrum from 2935 to 8870 Å. Second Revision of Rowland’s Preliminary Table of Solar Spectrum Wavelengths, N.B.S. Monog. 61, Government Printing Office, Washington, D.C. 255 Nave G., Johansson S., Learner R.C.M., Thorne A.P., Brault T.J., 1994, ApJS 94, 221 Pinsonneault M., Deliyannis C.P., Demarque P., 1992, ApJS 78, 181 Primas F., Duncan D.K., Pinsonneault M.H., Deliyannis C.P., Thorburn J.A., 1997, ApJ 480, 784 Sackmann I.-J., Boothroyd A.I., 1999, ApJ 510, 217 Spite M., 1967, Ann. Astrophys. 30, 211 Wiese W.L., Martin G.A., Fuhr J.R., 1969, Atomic Transition Probabilities: Sodium through Calcium. NSRDS-NBS 22