Further observational constraints to models of the kHz QPOs in LMXBs

I present measurements of 3 different aspects of the kHz QPOs in the low-mas X-ray binaries Sco X-1, 4U 1608--52, 4U 1728--34, and Aql X-1: 1. Besides the two main peaks at ν1 and &nu2, models of the kHz QPOs predict other weaker high-frequency QPOs: the sidebands and overtones of the kHz QPOs. I use ~ 180 ks of data of Sco X-1 to search for other QPOs at 2 ν1, 2 ν2, ν2 - ν1, ν2 + ν1, 2 ν1 - ν2, 2 ν2 - ν1, 2 ν1 + &nu2, 2 ν2 + ν1, and 2 (ν2 - ν1). 2. For individual sources QPO frequency and count rate are well correlated on time scales shorter than a day, whereas on longer time scales the QPO covers the same frequency range despite count rate being a factor > 2 different. This has been interpreted in terms of two independent channels of mass flow onto the neutron star: A radial inflow that only affects the X-ray intensity and produces separate branches in the frequency vs. count rate diagram, and an inflow through the disk that affects both intensity and timing properties and is responsible for the correlated changes on short time scales. In this scenario, at a constant QPO frequency, the fractional amplitude of the kHz QPOs should decrease with count rate. I present this type of measurements for the kHz QPO sources 4U 1608--52, 4U 1728--34, and Aql X-1; I discuss the results in the context of the two-channel accretion scenario. 3. I measure the fractional rms amplitude of the two simultaneous kHz QPOs in 4U 1608--52 as a function of energy. I divide the data into 3 intervals according to the frequency of the lower-frequency kHz QPO: ν1 <700 Hz, 700 < ν1 <800 Hz, and ν1 > 800 Hz, respectively. I show that the energy spectra of the two QPOs are significantly different. I compare the QPO energy spectra measured at different QPO frequencies.