Analysis of the OGLE microlensing candidates using the image subtraction method

Abstract

The light curves of the OGLE microlensing candidates have been reconstructed using the image subtraction method. A large improvement of the photometric accuracy has been found in comparison with previous processing of the data with DoPHOT. On the mean, the residuals to the fit of a microlensing light curve are improved by a factor of 2 for baseline data points, and by a factor of 2.5 during magnification. The largest improvement was found for the OGLE #5 event, where we get an accuracy 7.5 times better than with DoPHOT. Despite some defects in the old CCD used during the OGLE I experiment we obtain most of the time errors that are only 30 % to 40 % in excess of the photon noise. Previous experiment showed that with modern CCD chips (OGLE II), residuals much closer to the photon noise were obtained. The better photometric quality enabled us to find a low amplitude, long term variability in the OGLE #12 and OGLE #11 baseline magnitude. We also found that the shape of the OGLE #14 candidate light curve is fairly inconsistent with microlensing of a point source by a point lens. A dramatic change in the light curve of the OGLE #9 candidate was also found, which indicates that very large biases can be present in data processed with DoPHOT. To conclude we made a detailed analysis of the blending issue. It is found that OGLE #5 and OGLE #6 are very likely highly blended microlensing events. These events result from the magnification of a faint star that would have been undetectable without microlensing.