The MACHO Project HST Follow-Up: The Large Magellanic Cloud Microlensing Source Stars

We present Hubble Space Telescope (HST) WFPC2 photometry of 13 microlensed source stars from the 5.7 year Large Magellanic Cloud (LMC) survey conducted by the MACHO Project. The microlensing source stars are identified by deriving accurate centroids in the ground-based MACHO images using difference image analysis (DIA) and then transforming the DIA coordinates to the HST frame. None of these sources is coincident with a background galaxy, which rules out the possibility that the MACHO LMC microlensing sample is contaminated with misidentified supernovae or AGN in galaxies behind the LMC. This supports the conclusion that the MACHO LMC microlensing sample has only a small amount of contamination due to non-microlensing forms of variability. We compare the WFPC2 source star magnitudes with the lensed flux predictions derived from microlensing fits to the light curve data. In most cases the source star brightness is accurately predicted. Finally, we develop a statistic which constrains the location of the Large Magellanic Cloud (LMC) microlensing source stars with respect to the distributions of stars and dust in the LMC and compare this to the predictions of various models of LMC microlensing. This test excludes at > 90% confidence level models where more than 80% of the source stars lie behind the LMC. Exotic models that attempt to explain the excess LMC microlensing optical depth seen by MACHO with a population of background sources are disfavored or excluded by this test. Models in which most of the lenses reside in a halo or spheroid distribution associated with either the Milky Way or the LMC are consistent which these data, but LMC halo or spheroid models are favored by the combined MACHO and EROS microlensing results.

💡 Research Summary

The paper presents a detailed Hubble Space Telescope (HST) follow‑up of thirteen microlensing events detected toward the Large Magellanic Cloud (LMC) during the 5.7‑year MACHO survey. Using difference image analysis (DIA) on the ground‑based MACHO frames, the authors derived precise centroids for each event and transformed these coordinates into the HST reference frame with sub‑arcsecond accuracy. This methodology eliminates the usual blending ambiguities that plague ground‑based photometry and allows an unambiguous identification of the true source stars in the high‑resolution WFPC2 images.

All thirteen sources are clearly resolved and none coincides with a background galaxy. This directly rules out contamination by supernovae, active galactic nuclei, or any other extragalactic variable that could masquerade as a microlensing event. Consequently, the MACHO LMC sample is confirmed to be essentially free of non‑microlensing interlopers, strengthening the reliability of the derived optical depth.

The authors then compare the HST‑measured V and I magnitudes of the source stars with the “unmagnified” fluxes predicted by the microlensing light‑curve fits (which provide amplification, timescale, and baseline flux). In the majority of cases the observed brightness matches the model prediction within the quoted uncertainties, demonstrating that the light‑curve modeling accurately captures the source flux. A few outliers exhibit modest discrepancies, which the authors attribute to residual blending, small systematic errors in the DIA centroiding, or intrinsic variability of the source itself.

A novel statistical test is introduced to locate the sources relative to the three‑dimensional distribution of stars and dust in the LMC. By combining the measured colors and magnitudes with a model of the LMC’s stellar and dust layers, the test yields the probability that a given source lies in front of, within, or behind the main stellar disk. Applying this to the thirteen events shows that models in which more than 80 % of the sources reside behind the LMC are rejected at > 90 % confidence. In other words, the data strongly favor the interpretation that the sources are embedded in or in front of the LMC disk, not far behind it. This result effectively disfavours exotic “background‑source” scenarios that have been proposed to explain the excess microlensing optical depth.

Finally, the paper evaluates several lens‑population models. Scenarios placing the lenses in the Milky Way halo or spheroid, as well as those placing them in an LMC halo or spheroidal component, are compared against the observed event rate and the source‑location constraints. Models that attribute most of the lenses to a dark halo or spheroid associated with either the Milky Way or the LMC remain consistent with the data, while those relying primarily on LMC disk self‑lensing are less favored. When combined with independent results from the EROS collaboration, the evidence points toward a substantial contribution from an LMC halo or spheroid population of compact objects.

In summary, the HST follow‑up confirms the purity of the MACHO LMC microlensing sample, validates the source‑flux predictions of the light‑curve fits, and provides a robust statistical constraint on the three‑dimensional location of the sources. These constraints rule out models that invoke a dominant background‑source population and support lens‑population models in which the lenses reside in halo or spheroidal components of the Milky Way or the LMC.