Introduction: Microlensing is a new technique capable of detecting extra-solar planets orbiting very distant stars in the Galaxy. Recently, the microlensing technique has received media attention when a press release was issued claiming the possible detection of a planet with a mass between that of Earth and Neptune orbiting a very distant low-mass Galactic star. This announcement was made by a team of astronomers, known as the MPS and MOA collaborations, based on their microlensing observations of an event known as MACHO 98-BLG-35, originally discovered by the MACHO team. Because the MPS/MOA group reported that the claimed planet orbited this low-mass star at radii somewhat larger than the Earth orbits the Sun, some even speculated that if the planet existed, it may harbor life. In their subsequent published paper on this microlensing event, the MPS and MOA teams ruled out high-mass planets orbiting in the lens at certain radii, and reported intriguing evidence for the low-mass planet. Since the signal of this possible low-mass planet was below their detection limit, however, it could not be considered a firm detection.
The PLANET collaboration: The PLANET collaboration is an international team of astronomers that since 1995 has been performing precise and frequent microlensing observations with telescopes around the world in order to study interesting features in microlensing light curves, including those caused by planets orbiting the lenses. Due to the media attention given this event, and the natural public interest, PLANET decided to display publicly its own "raw" observations of this event, which were taken as part of its normal operations in 1998, but to report results only after a thorough analysis. That analysis is now nearing completion. The PLANET light curve for this event, and its implications for possible planets in MACHO 98-BLG-35, are briefly described below. As shown in the next two figures, the PLANET data are quite consistent with a single lens (no planets). In particular, the planetary deviation reported by the MPS and MOA teams appears to be somewhat disfavored by the PLANET data.
Presented above is a portion of PLANET team light curve near the peak of the microlensing event MACHO 98-BLG-35. This is the portion of the light curve in which a possible planetary deviation was reported by the MPS and MOA teams. The points represent the rise and then drop in brightness of the background star measured by the PLANET team as a foreground star passes almost directly in front of it as seen from Earth. The effect is due to the refocusing of light rays from the background star by the gravitational field of the foreground star, called the lens. Microlensing is rare since exceedingly precise alignment of the two stars is required. The observed brightness of the background star falls when the foreground lens moves out of alignment. Each mark on the bottom scale represents 0.1 day (2.4 hours) of time. The green points were taken in Chile; the blue in Tasmania, Australia, and the red in South Africa. In this plot, the uncertainties in the brightness reported are smaller than the size of the dots. The black line is a model for the light curve assuming that the lens of MACHO 98-BLG-35 has no planet; this model provides a good match to all PLANET team data.
Studying the region of light curve where the possible planet detection was claimed: The plot below shows a comparison between the PLANET team data and models with and without a planet in the region of the light curve where the MPS and MOA teams have reported seeing a deviation in brightness that they attribute to a possible planet orbiting the lens. Each mark on the bottom scale represents 2.4 hours of time. The points are color coded by observing site (Chile, Tasmania, SAAO), and solid points are data taken with a red (I band) filter in front of the camera while open points were taken with a green (V band) filter. The vertical lines represent measurement uncertainties. The plot contains three portions, each of which is described below.
The top plot shows differences (residuals) between the data taken around the peak of the light curve by the PLANET team and predictions from a single lens (no planet) model. If the model (and data!) were perfect in every respect, the data points would lie on the dashed line. Where a point is above the dashed line, the light curve is a bit brighter than expected; where a point is below the line, the light curve is a bit fainter. Most of the PLANET data agree with the single lens model to within 1% or better (that is, the points lie between -1.01 and +1.01 on the plot). The data should follow the ``wiggly'' solid line if this single lens has a planetary companion with parameters matching the planet model that best fit the data of the MPS/MOA team. (MPS/MOA proposed other models as well, some of which have lower mass and would therefore generate smaller ``wiggles.'')
The middle plot shows residuals between the PLANET peak data and the lens+planet model, using the planetary characteristics given by the MPS and MOA teams and allowing the primary lens parameters to adjust to achieve the best match. If this binary lens model (and the data) were perfect, all the points would lie on the dashed line. The farther from the dashed line, the larger the disagreement between the model and the data. (In all panels, only data near the peak of the light curve are used in order to minimize the effect of faint data at low microlensing magnification.)
The bottom plot shows the difference in a mathematical criterion called chi-squared between the lens+planet model and the single lens model. Where this quantity is positive (above the line), the data prefer the single-lens model; where the quantity is negative, the data prefer the lens+planet model. Chi-squared differences between -4 and 4 for each data point are easily explained due measurement uncertainties, but larger values may bear on the relative viabilities of different models. The South African (red) data seem to ``prefer'' the single-lens model somewhat over the lens+planet model.
Need for care in the analysis: In order to be certain what this and other light curves might tell us about the lenses and any planets that may be orbiting them, a careful analysis must be performed. This is especially important because the kind of anomalies (wiggles) in the light curve caused by planets can be small in size and last for only a short period of time. This means that often more than one scientific interpretation can be attached to the same real anomaly, and also that observational difficulties can sometimes masquerade as real anomalies.
Difficult observations: Microlensing events are rare (any given star has little more than a one-in-a-million chance of being lensed) and events are searched for only where there are millions of stars. Most microlensing events have been found by looking in directions close to the center of our Galaxy, were the stellar fields are very crowded (see right). In poor observing conditions light from neighboring stars (or even the moon!) can be confused with that from the background lensed star.
Systematic effects: Care must be taken in the analysis to study, and wherever possible remove, ``systematic effects.'' These are conditions in the experiment itself that can cause some data points to always be higher or lower than they would be in a perfect experiment. Effects from atmospheric turbulence and moonlight are examples. If these sorts of systematic effects go unnoticed, deviations in the data caused by them might be attributed to the wrong cause. Data from different telescopes can also exhibit systematic differences from one another. The PLANET light curve shown above has been corrected (to the best of our ability) for these effects so that the disturbances caused by moonlight and turbulence in the air have been removed for each observing site. To read more about the importance of removing systematic effects, and PLANET team results for another microlensing event, read our paper on OGLE-1998-BUL-14, which is now published in the Astrophysical Journal, vol 534, page 894 .
Measurement uncertainties: As the object gets fainter, or the observing conditions worsen, the brightness is more difficult to measure and the error bars get larger. Usually, discrepant points that seem to lie above or below the general trend have the largest error bars. Because the observations are so difficult, however, the photometry program sometimes underestimates the uncertainty, and attaches an error bar that is too small. The reliability of the error bars must be understood before we can know whether points that lie above or below the trend are real microlensing effects, or just uncertain measurements. The vertical lines (error bars) attached to every point in the plots above represent the PLANET team's best estimate of the size of the measurement uncertainties at each individual observing site.
In conclusion: The PLANET team data for the microlensing event MACHO 98-BLG-35 are well-explained by a single lens; we do not confirm the evidence for a possible planet orbiting the lens reported by other researchers. To date, the PLANET team has detected no planetary signatures in 43 microlensing events, implying that no more than 1/3 of the most typical stars in the Galaxy can have Jupiter-mass planets orbiting between 1.5 and 4 AU. (One AU is the distance between the Earth and its Sun.) To read more about this result, download our technical paper.
The data presented here are preliminary and subject to further analysis. We kindly request that any use of this information or reproduction of these figures in the press or any other reports acknowledge the PLANET collaboration and this http address.