SpringerLink
Forum Springer Astron. Astrophys.
Forum Whats New Search Orders


Astron. Astrophys. 355, L39-L42 (2000)

Previous Section Next Section Title Page Table of Contents

3. LMC data analysis

The analysis of the LMC data set was done using a program independent from that used in the SMC study, with largely different selection criteria. The aim is to cross-validate both programs (as was already done in the analysis of EROS 1 Schmidt photographic plates, Ansari et al. 1996) and avoid losing rare microlensing events. Preliminary results of the present analysis were reported in Lasserre (1999). We only give here a list of the various steps, as well as a short description of the principal new features; details will be provided in Lasserre et al. (2000).

We first select the 8% "most variable" light curves, a sample much larger than the number of detectable variable stars. Working from this "enriched" subset, we apply a first set of cuts to select, in each colour separately, the light curves that exhibit significant variations. We first identify the baseline flux in the light curve - basically the most probable flux. We then search for runs along the light curve, i.e. groups of consecutive measurements that are all on the same side of the baseline flux. We select light curves that either have an abnormally low number of runs over the whole light curve, or show one long run (at least 5 valid measurements) that is very unlikely to be a statistical fluctuation. We then ask for a minimum signal-to-noise ratio by requiring that the group of 5 most luminous consecutive measurements be significantly further from the baseline than the average spread of the measurements. We also check that the measurements inside the most significant run show a smooth time variation.

The second set of cuts compares the measurements with the best fit point-lens point-source constant speed microlensing light curve (hereafter "simple microlensing"). They allow us to reject variable stars whose light curve differs too much from simple microlensing, and are sufficiently loose not to reject light curves affected by blending, parallax or the finite size of the source, and most cases of multiple lenses or sources.

After this second set of cuts, stars selected in at least one passband represent about 0.01% of the initial sample; almost all of them are found in two thinly populated zones of the colour-magnitude diagram. The third set of cuts deals with this physical background. The first zone contains stars brighter and much redder than those of the red clump; variable stars in this zone are rejected if they vary by less than a factor two or have a very poor fit to simple microlensing. The second zone is the top of the main sequence. Here we find that selected stars, known as blue bumpers (Alcock et al. 1997a), display variations that are always smaller than 60% and lower in the visible passband than in the red one. These cannot correspond to simple microlensing, which is achromatic; they cannot correspond to microlensing plus blending with another unmagnified star either, as it would imply blending by even bluer stars, which is very unlikely. We thus reject all candidates from the second zone exhibiting these two features.

The fourth set of cuts tests for compatibility between the light curves in both passbands. We retain candidates selected in only one passband if they have no conflicting data in the other passband. For candidates selected independently in the two passbands, we require that their largest variations coincide in time.

The tuning of each cut and the calculation of the microlensing detection efficiency are done with simulated simple microlensing light curves, as described in Palanque-Delabrouille et al. (1998). For the efficiency calculation, microlensing parameters are drawn uniformly in the following intervals: time of maximum magnification [FORMULA] within the observing period [FORMULA] days, impact parameter normalised to the Einstein radius [FORMULA] and timescale [FORMULA] days. All cuts on the data were also applied to the simulated light curves.

Only two candidates remain after all cuts. Their light curves are shown in Fig. 1; microlensing fit parameters are given in Table 1. Although the candidates pass all cuts, agreement with simple microlensing is not excellent.

[FIGURE] Fig. 1. Light curves of candidates EROS-LMC-3 and 4. The plain curves show the best point-lens point-source fits; time is in days since Jan. 1, 1990 (JD 2,447,892.5).


[TABLE]

Table 1. Results of microlensing fits to the two new LMC candidates; [FORMULA] is the Einstein radius crossing time in days, [FORMULA] the impact parameter, and [FORMULA] the [FORMULA] blending coefficients.


The efficiency of the analysis, normalised to events with an impact parameter [FORMULA] and to an observing period [FORMULA] of two years, is summarised in Table 2. The main source of systematic error is the uncertainty in the influence of blending. Blending lowers the observed magnifications and timescales. While this decreases the efficiency for a given star, the effective number of monitored stars is increased so that there is partial compensation. This effect was studied with synthetic images using measured magnitude distributions (Palanque-Delabrouille 1997). Our final efficiency is within 10% of the naive efficiency.


[TABLE]

Table 2. Detection efficiency in% as a function of the Einstein radius crossing time [FORMULA] in days, normalised to events generated with [FORMULA], and to [FORMULA].


Previous Section Next Section Title Page Table of Contents

© European Southern Observatory (ESO) 2000

Online publication: March 21, 2000
helpdesk.link@springer.de