D.K. Duncan, F. Primas, K.A. Coble, L.M. Rebull
University of Chicago, Department of Astronomy and Astrophysics, Chicago, IL 60637 USA
University of Hawaii Institute for Astronomy, Honolulu, HI 96822 USA
C. P. Deliyannis
Yale University, Department of Astronomy, New Haven, CT 06520-8101 USA
University of Chicago, Yerkes Observatory, Williams Bay, WI 53191-02458 USA
University of Texas, Department of Astronomy, Austin, TX 78712 USA
Anglo-Australian Observatory, Sydney, Australia
A straight line of slope 0.98 gives an excellent fit to the relationship of B and [Fe/H]. There is no indication of a change in slope between halo and disk metallicities. The B/Be ratio is typically 10. This data supports models of light element production by cosmic ray spallation of C,N,O nuclei onto protons and He nuclei, probably in the vicinity of massive supernovae in star-forming regions. It does not support the long-held view of light element production occurring in the general ISM from the collisions of high-energy protons onto C,N,O nuclei.
The evolution of Li, Be, and B is a powerful discriminant between different models of the chemical and dynamical evolution of the Galaxy. Light element production depends on the intensity and shape of the cosmic ray spectrum, which in turn depends on the supernova and massive star formation rates. It also could depend on the rise of the (progenitor) CNO abundances and the decline of the gas mass fraction, which is affected by rates of infall of fresh (unprocessed) material and outflow, e.g., by supernova heating. Detailed models incorporating these effects were first presented by Vangioni-Flam et al. (1990), and since then by many others including Ryan et al. (1992), and Prantzos et al. (1993). The present observations were obtained in order to test and constrain these models.
The Goddard High-Resolution Spectrograph (GHRS) of the Hubble Space Telescope (HST) was used with the G270M grating to obtain spectra of resolution 26,000 in the BI region (2500Å) of seven stars ranging in metallicity from [Fe/H] = -0.3 to -2.85. Typical S/N achieved was 25 per pixel (0.026Å), or 50 per diode. To this was added the data for BD -13^o 3442 ([Fe/H]=-2.96, the subject of a separate investigation), that of the Sun, and three stars observed by Duncan, Lambert, & Lemke (1992).
Boron abundances were determined via spectrum synthesis using the Synthe program distributed by Kurucz (1993) on CD-ROM, modified to run on Unix SPARCstations by Steve Allen (UC Santa Cruz). We used the line list of Duncan et al. (1996), which consists almost entirely of laboratory-measured lines, and which fits both the Hyades giants and metal-poor stars.
Considerable care was spent in determining the error of each of the B determinations. Sources of random error we considered included stellar effective temperature, metallicity, continuum placement, and photon statistics in the points defining the line itself. In more metal-poor stars the continuum is easier to define, but the B line is weaker and less certain. At disk metallicity, continuum errors are larger but make less difference since the line is deep. Suggestions of systematic NLTE effects which would increase B abundances for all the stars are beyond the scope of the present investigation, but are being studied by Kiselman (1995). We will also test the NLTE suggestion observationally in 1995-96 by HST observations of BI and BII.
A typical spectrum synthesis fit is shown in Figure 1. Errors for HD 76932 (one of the more metal-rich stars) and BD 26^o 3578 (one of the more metal-poor) are given as examples in the table.
Figure: Sensitivity of model to boron abundances. Shown are the best-fit model and ones with the B abundance changed by 0.3 dex, compared to the data (heavy solid line).
Figure: Variation of B abundance with [Fe/H]. Least-squares linear fit : [Fe/H] with a reduced chi square =0.66.
Figure 2 immediately shows the main result of this investigation: [B] is very linear with [Fe/H] over both disk and halo metallicities. A least squares fit to this data yields a slope of 0.98 and reduced chi square of 0.66, indicating an excellent fit.
Least-squares fitting with two line segments joined at any intermediate metallicity worsens the fit.
Figure: Variation of Li, Be, and B with [Fe/H]. Fit to the B data is that of Figure 2; fit to Be data is B fit - 1 dex. Fit to Li data from Duncan et al. (1992).
Figure 3 shows Be and Li abundances for the same stars. The least squares fit of Figure 3, reduced 10X, gives a good fit to the Be data. The two stars which show reduced Li abundances are the two coolest stars in the sample and, therefore, might be expected to have deeper convection zones and thus destroyed some of their Li.
These data are now being used to test models of galactic chemical evolution. A preliminary conclusion is that the B/Be ratio strongly supports cosmic ray (CR) spallation as the source of the B and Be observed; no evidence is found for direct contribution of B from supernovae (Woosley et al. 1990). The slope of nearly exactly 1 supports models in which B and Be are produced by spallation of C,N,O onto protons. This was originally suggested by Duncan, Lambert, & Lemke (1992), and has been modelled in detail by Casse, Lehoucq, & Vangioni-Flam (1995) and Ramaty, Kozlovsky, & Lingenfelter (1995). Casse et al. and Ramaty et al. find that winds from massive stars in star-forming regions and massive star supernovae produce a flux of C and O which through collisions with protons and He nuclei can reproduce both the magnitude and slope of B production seen in Figure 2. They further identify the recent detection of gamma ray line emission from the Orion Nebula as direct evidence for this process.
These models differ from the long-accepted picture (Reeves, Fowler, & Hoyle 1970) that spallation of CR protons onto C,N,O nuclei in the general interstellar medium is the primary source of light element production. The new models also predict a higher B/B ratio, and seem to solve the long-standing problem of isotopic composition seen for B in cosmic rays.
If the newer models are correct, one would expect the linear relationship to hold most accurately between B and O abundances, rather than B and Fe. Unfortunately, accurate O abundances for halo stars are difficult to obtain (e.g., King & Boesgaard 1995). We are currently seeking the most accurate O abundances for the present sample of stars to test the trend of B vs. O.
The authors wish to thank Michel Casse, Elizabeth Vangioni-Flam, and Stan Woosley for very helpful discussions.
This research was based on observations obtained with the NASA/ESA Hubble Space Telescope through the Space Telescope Science Institute, which is operated by the AURA, Inc., under NASA contract NAS5-26555.
Casse, M., Lehoucq, R., & Vangioni-Flam, E. 1995, Nature, 373, 318
Duncan, D., Peterson, R., Thorburn, J., Pinsonneault, M., & Deliyannis, C. 1996, ApJ, in press
Duncan, D.K., Lambert, D.L., & Lemke, M. 1992, ApJ, 401, 584
King, J.R. & Boesgaard, A.M. 1995, AJ, 109, 383
Kurucz, R.L. 1993, CDROM # 18
Prantzos, N., Casse, M., & Vangioni-Flam, E. 1993, ApJ, 403, 630
Ramaty, R., Kozlovsky, B., & Lingenfelter, R. 1995, ApJ, in press
Ryan, S.G., Norris, J.E., Bessel, M.S., & Deliyannis, C.P. 1992, ApJ, 388, 184
Vangioni-Flam, E., Casse, M., Audouze, J., & Oberto, Y. 1990, ApJ, 364, 568
Woosley, S., Hartmann, D., Hoffman, R., & Haxton, W. 1990, ApJ, 356, 272