4 Be in stars with planets

There are many evidences indicating that the depletion of Be is connected with the rotational history of a star (e.g. Stephens et al. 1997). Although still not completely established, this history may be related to the presence or not of a (massive) proto-planetary disk (Edwards et al. 1993; Strom 1994; Stassun et al. 1999; Barnes et al. 2001; Rebull 2001; Hartmann 2002). If true, this may result in different depletion rates for stars which had different disk masses, and thus may or not have been able to form the now discovered giant planets. This could, in fact, have been the case for the pair of very similar dwarfs 16 Cyg A and B (the latter having a detected planetary companion), for which the Li abundances seem to be quite different (Cochran et al. 1997; King et al. 1997; Gonzalez et al. 1998), but show similar amounts of Be (García López & Perez de Taoro 1998). On the other hand, if pollution plays any role, we might also expect to detect some differences between stars with planets and stars without planets concerning light element abundances, and in particular concerning Be.

Stars with planets might thus present statistically different Be contents when compared with stars without (detected) planetary systems.

In Fig. 3 we show a plot of Be abundances for the stars presented in this paper as a function of $T_{\rm eff}$. Filled symbols represent stars with planets, while open symbols denote dwarfs with no planet companions found by radial-velocity surveys (i.e. measured in the context of the CORALIE survey (Udry et al. 2000), and having no clear radial-velocity signature of a planet). We also included three objects from the study of García López & Perez de Taoro (1998) - triangles, namely 16 Cyg A (HD 186408), 16 Cyg B (HD 186427) and 55 Cnc (HD 75732 A), for which the atmospheric parameters have been computed using the same technique as for the stars presented here (Gonzalez 1998). In order to keep this work as an homogeneous comparative study, we have preferred not to introduce measurements for stars "without'' planets taken from other works (e.g. Stephens et al. 1997), derived using different sets of atmospheric parameters and chemical analysis (e.g. equivalent widths).

In the figure, stars with a surface gravity $\log{g}\le4.1$ (probably slightly evolved) are represented by the squares. These include HD 10697, HD 38529, HD 117176, HD 143761, and HD 168443 (all planet hosts). Note also that HD 114762, a quite metal-poor dwarf ( $\rm [Fe/H]=-0.6$) with a brown-dwarf companion was also included amongst the planet hosts.

Figure 3: Be abundances for stars with planets (filled symbols) and stars without detected planetary companions (open symbols) as a function of $T_{\rm eff}$. Circles represent the stars studied in this work, while triangles denote the three objects taken from the study of García López & Perez de Taoro (1998). Stars with surface gravities $\log{g}\le4.1$ are denoted by squares. Superposed are the Be depletion isochrones (case A) of Pinsonneault et al. (1990) for solar metallicity and an age of 1.7 Gyr. The Sun is represented by the Solar Symbol. From top to bottom, the lines represent a standard model (solid line), and 4 models with different initial angular momentum (Tables 3 to 6). Typical error bars are represented in the lower-right corner of the figure.

Superposed with the measurements are a set of Yale beryllium depletion isochrones from Pinsonneault et al. (1990) for solar metallicity and an age of 1.7 Gyr. Given that it is not possible to know what was the initial Be abundance for each star, we have considered an initial $\log{N(\rm Be)}=1.28$, i.e., between Solar ( $\log{N(\rm Be)}=1.15$ - Chmielewski et al. 1975) and meteoritic ( $\log{N(\rm Be)}=1.42$ - Anders & Grevesse 1989).

Although still preliminary, a look at the figure shows that the current results argue against pollution as the key process leading to the metallicity excess of stars with planets (see also Santos et al. 2001a, 2001b; Pinsonneault et al. 2001). Adding $\sim$50 earth masses of C1 chondrites to the Sun, for example, would increase its iron abundance by about 0.25 dex (a value similar to the average difference observed between stars with and without detected giant planets), and its Be abundance would increase by a slightly higher factor. No such difference seems present in our data. In the same way, the results also do not support extra-mixing due to an eventual different angular momentum history of the two "populations'' of stars. We note, however, that at this stage we are limited by the small number of comparison stars analyzed in this work.

In this context, one particular star (HD 82943) deserves a few comments. This planet host was recently found to have a near meteoritic ⁶Li/⁷Li ratio (Israelian et al. 2001a), better interpreted as a sign that planet/planetary material was engulfed by this dwarf sometime during its lifetime. Israelian et al. have suggested that the Li isotopic ratio of HD 82943 could be explained by the fall of either the equivalent of a 2 jupiter mass giant planet or a 3 earth mass terrestrial planet (or planetary material). Considering that HD 82943 had initially a solar Be abundance, the addition of this quantity of material would increase its Be abundance by about 0.1 dex. A small "excess'' of Be would thus be observed, but this value would be within the error bars of our measurements. As we can see from Fig. 3, the Be abundance of HD 82943 is not particularly high when compared to other stars in the plot (although it seems to occupy a position near the upper envelope of the Be abundances). No further conclusions can be taken.

4.1 The log N(Be) vs. T $_\mathsf{eff}$ trend

According to the standard models (without rotation - e.g. Table 1 of Pinsonneault et al. 1990), basically no Be depletion should occur for dwarfs in the temperature region plotted in the figure. This is clearly not compatible with the observations. However, it is already known that standard models are not able to explain most of the observed behaviors of Li and Be (e.g. Stephens et al. 1997). On the other hand, models with rotation (Pinsonneault et al. 1990) were shown to be quite satisfactory for stars with $6500~{\rm K}\le T_{\rm eff}\le 5600$ K (Stephens et al. 1997). But as we can see from the figure, they do not predict significant burning at temperatures around 5200 K. We note that according to these isochrones, Be depletion is done at moderate rates once the star reaches the main-sequence (about 0.06 dex from 0.7 to 1.7 Gyr for a 0.9 $M_{\odot}$ dwarf, and at a considerably lower rate for higher ages following the Fig. 11 of Stephens et al. 1997), and thus we do not expect a crucial difference with the presented curves.

In contrast with the models, one very interesting detail is seen. If, for temperatures between about 5600 and 6200 K, the dispersion in the $\log{N({\rm Be})}$ is remarkably small, and the abundances seem to be consistent with the model predictions, the general trend for lower temperatures follows a slow decrease with decreasing temperature. Under the assumption that no difference exists between stars with and without giant planets, either the trend is due to some metallicity or age effect, or it is simply telling us that the models are not able to reproduce the observations for temperatures bellow $\sim$5600 K. This problem was also noted by Stephens et al. (1997), but no such extreme cases had been found, maybe because their "solar metallicity sample'' did not go down to temperatures lower than $\sim$5500 K. Note that the iron abundances for the stars in the plot are in the range from -0.6 to +0.5, and the variation of the initial Be abundance with stellar metallicity seems to be quite small in this interval (Rebolo et al. 1988; Molaro et al. 1997; García López et al., in preparation).

Particularly interesting are the 4 cases for which only upper values for the Be abundances were found: HD 38529, HD 145675 (14 Her), HD 13445 (Gl 86), and HD 75732 (55 Cnc), all in the temperature interval between 5150 K and 5700 K. A detail of the spectral synthesis for one of these stars can be seen in Fig. 2. While for HD 38529 the Be depletion may be explained by the fact that this star is already leaving the main-sequence (its low surface gravity and its position in the HR-diagramme - e.g. Gonzalez et al. 2001 - show it to be evolved) for the three remaining objects no simple explanations seem to exist.

One possible explanation for Be abundances of the most discrepant objects could be their ages. For example, 55 Cnc and Gl 86 seem to be quite old, with isochrone ages in agreement with the oldest stars in the galaxy (ages were determined from theoretical isochrones of Schaller et al. 1992, and Schaerer et al. 1993a, 1993b). This is not the case for 14 Her, with an age around $\sim$6 Gyr. Furthermore, for 14 Her, there is a star (HD 222335, without detected giant planet) with about the same temperature, similar age ($\sim$5-6 Gyr) but much higher Be abundance.

Figure 4: Lithium and Beryllium abundances for the stars studied in this paper. We note that for some of the stars, no Li abundances were found in the literature (see Table 2). Symbols as in Fig. 3. Typical error bars are represented in each panel (0.16 for Be and 0.12 for Li).

Can a metallicity effect explain the observed difference? An increase in the metallicity (and thus in the opacities) has the effect of changing the convective envelope depth of a star. It has been shown by Swenson et al. (1994) that even small changes in the oxygen abundances may have a strong effect on the Li depletion rates. Although it is unwise to extrapolate directly, a similar effect should exist concerning Be. In what concerns our case, most of the objects in the plot having temperatures lower than 5400 K are metal-rich with respect to the Sun. For example, 14 Her is one of the most metal rich stars in the sample ( $\rm [Fe/H]=+0.50)$. But some counter examples exist, like Gl 86, only about 100 K cooler, with a value of $\rm [Fe/H]=-0.20$. Unfortunately, no theoretical isochrones for Be depletion are available for these high metallicities, and it is not clear whether a difference of +0.5 dex in metallicity can or cannot induce significantly different Be depletion rates for this temperature range. On the other hand, if some unknown opacity effect could lead to systematic errors as a function of temperature - e.g. the missing UV opacity (Balachandran & Bell 1998) - we could maybe expect to measure some trend. However, we have no special reason to believe that this effect can lead to the observed trend. Furthermore, the missing UV opacity problem is far from being solved (if it exists at all); indeed, recent results seem to show that we can obtain a good fit of the solar spectrum without taking into account this extra-opacity (Allende Prieto & Lambert 2000).

Although not conclusive, the evidences discussed above suggest that a bona-fide explanation for the observed Be abundances may pass either by some inconsistency in the models, or by some metallicity effect. But given the much higher number of planet hosts in the plot when compared with the non-planet hosts, we still cannot discard that the presence of a planet might be the responsible for the observed trend. Else, since Be depletion in the Pinsonneault et al. (1990) models is strongly related to the angular momentum lost, we would have, for example, to consider that there is some negative correlation between initial angular momentum and/or angular momentum loss and stellar mass (which does not seem very plausible). Unfortunately only the addition of more objects to the plot, and in particular the determination of Be abundances for a larger sample of dwarfs with $T_{\rm eff}<5400$ K will permit to better settle down this question. Such observations are currently in progress.

It is also worth mentioning that on the other side of the $T_{\rm eff}$ plane, both HD 120136 ($\tau$ Boo) and BD -103166, positioned in the Li (and Be) dip region (e.g. Boesgaard & King 2002), have a Be abundance that may be compatible with the models.

4.2 Lithium vs. beryllium

Our data can also be used to further investigate the issue of Li and Be depletion in main sequence stars. Figure 4 compares Li and Be abundances for the same stars, and shows, as expected, that Li is burned much faster than Be, and its decline with temperature is much more clear. This is clearly expected from the models for a middle age solar type dwarf in this temperature regime. Only two points seem to come out of the main trend (in the Li panel). These are the cases of HD 10697 and HD 117176 (70 Vir). These two stars deserve some particular discussion in Sect. 4.2.1.

In Fig. 5 we plot Be vs. Li abundances for all the stars with $T_{{\rm eff}}<6300$ K for which both Li and Be abundances are available. The plot reveals no clear trends. Even 70 Vir and HD 10697 follow the main trend, probably attesting the normality of these two stars. Again, no significant difference is seen between the two groups of stars (planet hosts and non-planet hosts).

One interesting but expected detail in Fig. 5 is that there are no stars in the lower right part of the plot, a region that would correspond to stars having depleted much of their Be but that are still Li rich. This "gap'' in the figure, plus the few stars in the upper right corner, suggests that there might be a correlation between Li and Be depletions (like the one found by Deliyannis et al. 1998; Boesgaard et al. 2001 and more recently by Boesgaard & King 2002 for their sample of hotter dwarfs, but absent in the study of Domínguez Herrera 1998, and in the sample of old cluster members of Randich et al. 2002). However, there is a significant number of stars in the upper left region of the plot having Be determinations but only upper limits for the Li abundance (objects that are not present in Fig. 14 of Boesgaard et al.). This particular point in quite interesting, since it is, to our knowledge, the first time such a "population'' of stars is discussed in the literature.

Figure 5: Li vs. Be abundances for the stars in our sample having temperatures lower than 6300 K and for which both Li and Be abundances were available. Symbols as in Fig. 3. Typical error bars are represented in the lower-right corner of the figure.

Figure 6: Same as Fig. 5 but using different symbols for three temperature intervals: $T_{{\rm eff}}\le 5400$ K (filled circles), $5400~{\rm K}<T_{\rm eff}\le5900$ K (open circles), and $T_{{\rm eff}}>5900$ K (filled squares). See text for more details.

In Fig. 6 we have done the same plot as in Fig. 5, but this time using different symbols as a function of the temperature of the object. As we can easily see from the plot, the objects in the three different temperature regimes chosen occupy clearly different positions in the diagramme. Stars with $T_{{\rm eff}}\le 5400$ are positioned in the lower-left corner. These correspond basically to objects having burned some Be, and that only have upper values for the Li abundance. In the upper-right corner of the figure are positioned stars in the range $T_{{\rm eff}}>5900$ K. These correspond to objects that have essentially not burned any Be, having depleted only moderately their Li content. An intermediate population, with $5400~{\rm K}<T_{\rm eff}\le5900$ K, having already considerably depleted its original Li but not its Be content is present in the mid-upper part of the plot.

Except for HD 38529, a sub-giant star (the object with the lowest Be abundance in the plot), all the other objects are well grouped according to their temperature range. This is an interesting result, since it is giving us important clues about the temperature at which the onset of Li and Be depletion occurs in our metal-rich stars: if for Li depletion occurs already for temperatures around 5900 K, for Be this only seems to happen at $T_{{\rm eff}}\sim5400$ K. Note, however, that given the dispertion these limits have to be seen as approximate.

Note also that we did not include $\tau$ Boo in Figs. 5 and 6, since its the only object in the sample that is clearly in the Li-dip region.

   
4.2.1 HD 10697 and 70 Vir

As mentioned above, HD 10697 and 70 Vir do not fit into the mean trend in the lower panel of Fig. 4. Considering as initial abundances of Li the "cosmic'' value ( $\log{N({\rm Li})}=3.31$; Martin et al. 1994) and for Be the average between meteoritic and solar ( $\log{N(\rm Be)}=1.28$, as above), then the observed values for 70 Vir indicate that Be ( $\log{N(\rm Be)}=0.82$) is depleted about 3 times, while Li ( $\log{N(\rm Li)}=1.76$) is depleted by by a factor of 22. Nevertheless, such a pattern of Li and Be depletion is not so unexpected (see e.g. Deliyannis & Pinsonneault 1997; Stephens et al. 1997), and is compatible with models of depletion including turbulent induced rotational mixing (see e.g. discussion in Stephens et al. 1997). But 70 Vir is not a young object (it seems to be slightly evolved, as confirmed by its low surface gravity $\log{g}=3.90$), and it seems strange that an evolved star with its temperature may still have so much Li.

One possibility would be to invoke Li and Be dredge up from a "buffer'' below the former main-sequence convective envelope (Deliyannis et al. 1990). But as discussed in e.g. Randich et al. (1999), this scenario does not seem to be supported by the observations. However, the fact that this star is evolved means that its current temperature is probably different to the one it had during the main-sequence phase. With a near solar metallicity ( $\rm [Fe/H]=-0.03$) and a mass of $\sim$1.1 $M_{\odot}$, the temperature of 70 Vir would have been more close to 6000 K. A dwarf with this temperature is not supposed to burn much of its Li (as we can see from the plot). So, we speculate that this star may have just left the main-sequence, and started to dilute (and/or burn) the Li (as well as Be) content present in its convective envelope. If the size of the convective envelope has increased "quite fast'', maybe it still did not have time to deplete/dilute all the Li, but is already depleting part of its Be (Charbonnel et al. 2000).

The case of HD 10697 is in fact not very different from the one of 70 Vir. Both have the same mass (1.10 $M_{\odot}$), and similar surface gravities. The main difference resides in the fact that HD 10647 seems to have, within the errors, preserved most of its Be content, while Li is depleted about 23 times. But given the uncertainties in our Be abundance determination, and the unknown initial Li and Be content of the star, we cannot be certain whether the Be is really intact in the atmosphere of this object. The same arguments discussed for 70 Vir may thus be valid in this case.

It is worth mentioning that another possibility to explain the high Li abundances of these two stars would be to invoke a planet engulfment (like in the case of HD 82943 - Israelian et al. 2001a), that could simply result from the migration of a former close-in planet due to tidal interactions with the evolved star (Rasio et al. 1996). Unfortunately, no measurements of the Li isotopic ratio are available for 70 Vir. But recent models by Siess & Livio (1999) show that the absorption of a planet by a giant star would have the effect of increasing considerably its Li abundance, but only slightly the Be content, a situation that could be compatible with the observed Li and Be abundances for this star.

There is another interesting point about these two stars. Although they have very similar Li abundances, their Be contents differ by 0.56 dex. Given that both stars have similar metallicity (-0.03 and +0.16, for 70 Vir and HD 10697, respectively) their initial Be content is not expected to be significantly different (Rebolo et al. 1988; Molaro et al. 1997; García López et al., in preparation). However, we do not know up to which extent Be and Li abundances are uniform in the Inter-Stellar Medium. A different initial Li/Be ratio, together with the uncertainties in the Li and Be determinations, could probably lead to the observed effect. Else, we would need to invoke that the two stars have burned Li at the same rate, while the opposite happened for Be. But given that these two stars are sub-giants, we cannot exclude that we are simply observing some evolutionary effect, eventually combined with slightly different initial Li and Be abundances. Note that the former object is a bit cooler, being positioned in the temperature region where the Be depletion seems to occur (see previous section).