7. Discussion and conclusion

Although astronomers have for a long time agreed that a uniform, complete, and realistic stellar library is urgently needed, it must be emphasized that this goal has remained too ambitious to be achieved in a single concerted effort through the present epoch. We have thus attempted to proceed in well-defined steps, with priorities set according to the availability of basic data and following the most obvious scientific questions that would likely become more tractable, or even answerable. Therefore, we briefly review the present achievement to clarify its status in the ongoing process toward a future standard library of theoretical stellar spectra for photometric evolutionary synthesis.

1. Completeness. The unification of the massive K-library spectra with those for high-luminosity M-star models (the -library) is most important, because even as a small minority of the number population of a given stellar system, the late-type giants and supergiants may provide a large fraction of this system's integrated light at visible and infrared wavelengths. This fact was recognized early on (e.g., Baum 1959), and eventually also co-motivated the effort leading to the existence of the B-library used in this work (Bessell et al. 1988).

   While the K-, B-, and F-libraries have been used to remedy incompleteness (in either wavelength or parameter coverage, or both) of available observed stellar libraries before (e.g., Worthey 1992, 1994; Fioc & Rocca-Volmerange 1996), our first goal here has been to join them as a purely theoretical library, providing the main advantages of physical homogeneity and definition in terms of fundamental stellar parameters - which allows direct use with stellar evolutionary calculations.

   But even so, the present library remains incomplete in several respects. First, stellar evolution calculations (e.g., Green et al. 1987) predict that high-luminosity stars with temperatures near or below 3500 K may also exist at low metallicities, , and their flux contributions at visible-near ultraviolet wavelengths (where metallicity produces significant effects) may not quite be negligible in the integrated light of old stellar populations. Therefore, in order to provide the fuller coverage required for an adequate study of this metallicity-sensitive domain, new calculations of B-library spectra extending the original data to both shorter wavelengths and lower abundances (Buser et al. 1997) will replace the current hybrid spectra and make the next library version more homogeneous.

   Secondly, even though the low-temperature, low-luminosity M-dwarf stars do not contribute significantly to the integrated bolometric flux, they are not negligible in the determination of mass-to-light ratios in stellar populations. Thus, a suitable grid of (theoretical) M-dwarf spectra calculated by Allard & Hauschildt (1995) is being subjected to a similar calibration process (Paper II) and will be implemented in the present library as an important step toward the intended standard stellar library.

   Finally, the libraries of synthetic spectra for hot O- and WR-stars which were recently calculated by Schmutz et al. (1992) will allow us to extend the calibration algorithm to ultraviolet IUE colors, where such stars radiate most of their light.
   

2. Realism. In view of its major intended application - photometric evolutionary synthesis -, the minimum requirement that we insist the theoretical library must satisfy, is to provide stellar flux spectra having (synthetic) colors which are systematically consistent with calibrations derived from observations. How else could we hope to learn the physics of distant stellar populations from their integrated colors, unless the basic building blocks - i.e., the library spectra used in the synthesis calculations - can be taken as adequate representations of the better-known fundamental stellar properties, such as their color-temperature relations?

   Because the original library spectra do not meet the above minimum requirement (Sect. 3), we have developed an algorithm for calibrating existing theoretical spectra against empirical color-temperature relations (Sect. 4). Because comprehensive empirical data are unavailable for large segments of the parameter space covered by the theoretical library, direct calibration can be effected only for the major sequences of solar-abundance models (Sect. 5). However, we have also shown that the present algorithm provides the desired broad-band (or pseudo-continuum) color calibration without destroying the original relative monochromatic fluxes between arbitrary model grid spectra and solar-abundance calibration sequence spectra of the same effective temperature. This conservation of original grid properties also propagates with useful systematic accuracy even through most differential broad-band colors of the corrected library spectra. Thus, to the extent that differential broad-band colors of original library spectra were previously shown to be consistent with spectroscopic or other empirical calibrations of the UBV-, RGU-, and Washington ultraviolet-excess-metallicity relations (Buser & Fenkart 1990; Buser & Kurucz 1992; Lejeune & Buser 1996), the corrected library spectra are still consistent with the same calibrations.
   

3. Library development. At this point, we feel that some of the more important intrinsic properties required of the future standard library have already been established. Of course, many more consistency tests and calibrations will now be needed that can, however, only be performed for local volumes of the full parameter space covered by the new library. For example, we shall use libraries of observed flux spectra for individual field and cluster stars to better assess - and/or improve - the performance of the present library version in the non-solar-abundance and non-visual wavelength regimes. Eventually, we also expect significant guidance toward a more systematically realistic version of the library from actual evolutionary synthesis calculations of the integrated spectra and colors of globular clusters (Lejeune 1997).

   Last, but not least, we would like to emphasize that, while we here present the results taylored according to the general needs in the field of photometric evolutionary synthesis, the library construction algorithm has been designed such as to allow flexible adaptation to alternative calibration data as well. As we shall ourselves peruse this flexibility to accommodate both feed-back and new data, the reader, too, is invited to define his or her own preferred calibration constraints and have the algorithm adapted to perform accordingly.

Acknowledgements

We are grateful to Michael Scholz and Gustavo Bruzual for providing vital input and critical discussions. Christophe Pichon is also aknowledged for his precious help with the final implementation of some of the figures. We wish to thank warmly the referee for his helpful comments and suggestions. This work was supported by the Swiss National Science Foundation.