Spectroscopy of luminous infrared galaxies at 2 microns .3. Analysis for galaxies with log (L(IR)/L(circle dot)) greater than or similar to 11.2

We have obtained spectra across the K window for the first large sample of luminous galaxies selected from the IRAS survey. This paper contains the principal analysis of the 43 systems in our sample with luminosities of 11.2 less than or similar to log (L(IR)/L(.)) less than or similar to 11.9. The spectra themselves were presented in a companion paper by Goldader et al. (Paper II). The Br gamma luminosities are proportional to L(IR), at levels similar to those of star-forming regions. This strongly suggests that star formation accounts for the bulk of the energy production in these objects, in general agreement with previous studies. Good agreement is found for the continuous star formation models of Leitherer & Heckman with upper mass cutoffs well below 100 M(.). The models accommodate a range in starburst ages of similar to 10(7) to 10(9) yr. Instantaneous starburst models fit the data but imply an unrealistically short range of ages for the entire sample. It is difficult to avoid concluding that the initial mass functions are deficient in stars of less than similar to 1 M(.). Strong emission lines from molecular hydrogen are detected. The H-2 v = 1-0 S(1) line luminosities are proportional to L(IR); the correlation extends through the ultraluminous infrared galaxies. The H-2 emission in the galaxies tends to be more spatially extended than the Br gamma emission. Measured values and upper limits for the ratios of the various H-2 lines visible in our spectra indicate that the H-2 emission at 2 mu m is consistent with being shock excited. However, other mechanisms, operating at sufficiently high densities that the H-2 energy levels are thermalized by collisions, cannot be excluded. Based on energy considerations, we suggest that the shocks are due to supernova remnants expanding into the interstellar medium. The frequency of Type II supernovae necessary to account for the H-2 line emission agrees with frequencies deduced from the starburst models and the radio/far-infrared correlation. However, there remain a number of galaxies that cannot be made to fit this model. A decade after its discovery, a universal explanation of the strong H-2 emission in luminous infrared galaxies continues to elude us. No previously unrecognized broad-line active nuclei were discovered in our survey; either they are weak or absent or the true optical depths at 2 mu m are much higher than indicated by conventional extinction measures. However, there are clear differences between the K-band properties of galaxies that contain broad-line active nuclei and those that do not. The differences seem to be due to the presence of strong nonstellar continuum emission coming from the active nuclei themselves. With the addition of the 13 ultraluminous galaxies with log (L(IR)/L(.)) greater than or similar to 12 from Goldader et al. (Paper I), the number of systems observed in this program totals 56. We have incorporated these ultraluminous galaxies in some parts of the analysis to examine properties across the entire luminosity range of our sample.