The POINT-AGAPE survey: comparing automated searches of microlensing events towards M31

Searching for microlensing in M31 using automated superpixel surveys raises a number of difficulties which are not present in more conventional techniques. Here we focus on the problem that the list of microlensing candidates is sensitive to the selection criteria or 'cuts' imposed, and some subjectivity is involved in this. Weakening the cuts will generate a longer list of microlensing candidates but with a greater fraction of spurious ones; strengthening the cuts will produce a shorter list but may exclude some genuine events. We illustrate this by comparing three analyses of the same data set obtained from a 3 yr observing run on the Isaac Newton Telescope in La Palma. The results of two of these analyses have been already reported: Belokurov et al. obtained between three and 22 candidates, depending on the strength of their cuts, while Calchi Novati et al. obtained six candidates. The third analysis is presented here for the first time and reports 10 microlensing candidates, seven of which are new. Only two of the candidates are common to all three analyses. In order to understand why these analyses produce different candidate lists, a comparison is made of the cuts used by the three groups. Particularly crucial are the method employed to distinguish between a microlensing event and a variable star, and the extent to which one encodes theoretical prejudices into the cuts. Another factor is that the superpixel technique requires the masking of resolved stars and bad pixels. Belokurov et al. and the present analysis use the same input catalogue and the same masks but Calchi Novati et al. use different ones and a somewhat less automated procedure. Because of these considerations, one expects the lists of candidates to vary and it is not possible to pronounce a candidate a definite microlensing event. Indeed we accept that several of our new candidates, especially the long time-scale ones, may not be genuine. This uncertainty also impinges on one of the most important goals of these surveys, which is to place constraints on the massive compact halo object (MACHO) fraction in M31. Such constraints depend on using Monte Carlo simulations to carry out an efficiency analysis for microlensing detection, and the results should be relatively insensitive to the selection criteria provided the simulations employ the same cuts as the pipelines. Calchi Novati et al. have already derived the constraints associated with their analysis and we present here the constraints associated with the most recent analysis. The constraints are similar if we neglect our long time-scale events and comparable to those found for MACHOs in our own Galaxy by earlier microlensing surveys of the Magellanic Clouds. However, our constraints are different from those of Calchi Novati et al. if we include our long time-scale events.