2.3. Stellar IMF
The upper part of the HRD is degenerate in terms of the relation between stellar parameters and mass, even for a single stellar population. The situation is very different from, e.g., a globular cluster isochrone where most of the luminosity comes from a small mass interval close to the turn-off mass. In a typical massive-star population of single age, stars with vastly different zero-age-main-sequence masses can have similar Teff and L and contribute to the integrated light. The SED is therefore dependent on an assumption on the IMF because of the existence of a steep mass-to-luminosity relation in the upper HRD.
The royal way of deriving an IMF for young populations is via clusters. This is feasible only in few cases so that IMF data are often uncertain and restricted to a limited mass range (). Since massive stars are rare and have much larger luminosities than low-mass stars, it is difficult to observe the full stellar mass spectrum in one and the same star cluster. The low-mass star content of the closest massive-star formation region in Orion has been studied down to ~ 0.1 M. This region, however, contains only a handful of stars with masses above 10 M. In contrast, the R136 cluster in the LMC is rich enough for meaningful statistics above 10 M, but it is difficult to push the low-mass star detection limit below 1 M. The IMF in nearby star-forming galaxies has been derived using various indirect observational techniques using the integrated light. In these cases, the accessible stellar masses are between 100 M, where stochastic effects due to small-number statistics set in, and 1 M, where velocity dispersion measurements are used ().
Almost all studies suggest a rather uniform IMF in star-forming galaxies. Considering the very different observational techniques used and the prevailing theoretical uncertainties, this result suggests that any existing variation and its dependence on the environment are smaller than the current measuring uncertainties. The classical Salpeter IMF and its modern variant defined by Kroupa appear to provide the best match to the data ().