2004ApJ...611..846K

We have derived bias-corrected X-ray luminosity functions (XLFs) of sources detected in a uniformly selected sample of 14 E and S0 galaxies observed with Chandra ACIS-S3. The entire sample yields 985 pointlike X-ray sources, with typical detections of 30-140 sources per galaxy. After correcting for incompleteness, the individual XLFs are statistically consistent with a single power law of a (differential) XLF slope β=1.8-2.2 (with a typical error of 0.2-0.3). A break at or near L_X,Edd, as reported in the literature for some of these galaxies, is not required in any case. Given the uniform XLF shape, we have generated a combined, higher statistics XLF, representative of X-ray sources in elliptical galaxies. Although the combined XLF is marginally consistent with a single power law (with β=2.1±0.1), a broken power law gives an improved fit. The best-fit slope is β=1.8±0.2 in the low-luminosity range L_X=afewx10³⁷ to 5x10³⁸ ergs/s. At higher luminosities, the slope is steeper, β=2.8±0.6. The break luminosity is (5±1.6)x10³⁸ ergs/s (with an error at 90%), which may be consistent with the Eddington luminosity of neutron stars with the largest possible mass (3 M_☉), He-enriched neutron star binaries, or low-mass stellar-mass black holes. If the change in XLF slope at high luminosities is real and does not mask a step in the XLF, our result would imply a different population of high-luminosity sources, instead of a beaming effect. This high-luminosity portion of the XLF must reflect the mass function of black holes in these galaxies. We note that this high-luminosity population does not resemble that of the ultraluminous X-ray sources detected in star-forming galaxies, where no break in the XLF is present and the XLF is much flatter than in the older stellar system we are studying here. We use our results to derive the integrated X-ray luminosity of accreting low-mass X-ray binaries (LMXBs) in each sample galaxy. We confirm that the total X-ray luminosity of LMXBs is correlated with the optical and more tightly with the near-IR luminosities, but in both cases the scatter exceeds that expected from measurement errors. We find that the scatter in L_X(LMXB)/L_Kis marginally correlated with the specific frequency of globular clusters.