In total, 40 neutral diatomic molecules, 2 molecular ions, and 7 polyatomic molecules are known from observed photospheric stellar spectra. Line data for opacity computations (i.e., lists of line frequencies, intensities, and excitation energies) exist for 17 of these molecules, although the data are complete only for a handful of them. A detailed description of stellar photospheric molecules can be found in Tsuji (1986), and the existing opacity data have been reviewed by Jorgensen (1995).

Listed line frequencies in the data bases are either the measured values, or based on computed molecular constants obtained from fits to measured values. Attempts to compute ab initio line frequencies have so far resulted in lower accuracy than what is obtained by use of molecular constants. Published line strengths include measured values as well as ab initio values. For strong bands the ab initio intensities are as accurate as the laboratory values, whereas measured values for weak bands are generally more accurate than the ab initio values. The primary advantage of ab initio computations is therefore that the complete set of all transitions can be obtained. Exploratory studies have shown that completeness of the line data is crucial for the obtained stellar photospheric structure.

As an alternative to the ab initio computations of the line intensities, fits to experimental data have been attempted. The most promising method seems to be to fit the dipole function by use of a Padé approximant. Combined with a potential fitted to experimental energy levels, such a dipole function can in principle be used to predict the complete list of band intensities and line intensities for all bands with energies up to the molecular dissociation energy. The part of the dipole function which corresponds to the largest stretching (or bending) of the molecule is the most uncertain in such fits as well as in ab initio computations. This part is responsible for most of the many weak transitions, and large uncertainties are therefore to be excepted in the computed intensities of the weak spectral bands. As these are of major importance for the stellar photospheric structure (due to their huge number and their pseudo continuous appearance in the spectrum), a particularly large effort is desirable in comparing computed intensities with laboratory data for a representative sample of weak bands. Unfortunately, only few measurements of weak bands exist.