OGLE-2007-BLG-224L: A Direct Test of Terrestrial Parallax

OGLE-2007-BLG-224 was observed on the night of 2014 April 20 using the MagAO instrument on the Magellan Clay Telescope (Morzinski et al. 2016). For these observations, we used the Clio camera in "wide" mode, which gives a pixel scale of 27.49 mas pix⁻¹ and a 14'' × 28'' field of view. The observations were taken using a four-point dither pattern. At each position, five 30 s exposures were taken. The observations were made in the H band. A sky field was also observed earlier in the night.

The data were flat fielded using a difference flat constructed from flats that were taken at twilight on 2014 April 21. They were then sky subtracted using the mean sky frame. Bad pixels were flagged using a custom bad pixel mask, based on pixels with either large standard deviations in the flat frames ("hot" pixels) or low counts ("dead" pixels). After sky subtraction, we found that the median counts for each frame varied as a function of time (by ≲1% of the raw counts). We measured an offset for each frame and subtracted it. Then, we resampled the images at a factor of 5 higher resolution, before shifting them and taking the median of the images. The final image is downsampled from this median image to the original resolution.

The location of the target was measured from observations taken with ANDICAM (DePoy et al. 2003) on the SMARTS 1.3 m telescope in Chile, while the event was highly magnified, i.e., on HJD' = HJD−2,450,000 = 4233.6. These observations were taken in both I and H bands. Only one star in the ANDICAM H-band data (other than the target) is bright enough to be detected and is also in the MagAO field. Thus, we matched the catalog for this frame to the VVV catalog (Minniti et al. 2017), and then the VVV catalog to a catalog of stars extracted from the MagAO field using Source Extractor (Bertin & Arnouts 1996) to determine the location of the target. We also checked the microlens source position by matching the ANDICAM I-band catalog directly to the MagAO field and found consistent results. The identified target matches the description of the field given by Gould et al. (2009).

For the astrometric solution, we used the full 14'' × 28'' image. However, the two chips of the Clio detector have different photometric responses. Thus, we restricted our flux measurements to the chip containing the science target.

Because the region surrounding OGLE-2007-BLG-224 is relatively crowded (Figures 1 and 2), we fit sources using an empirical point-spread function (PSF), constructed from four bright, relatively isolated stars surrounding the target. The empirical PSF is parameterized by the centroid position (X, Y) and a scaling factor (peak counts). We calibrated the MagAO fluxes by cross-matching to the VVV catalog (Minniti et al. 2017). All of the VVV sources are blends of multiple MagAO stars. For each VVV star, we fit for all sources in a 20 substamp around the target. We then summed the fluxes for all MagAO stars within 07 of the VVV source. In total, there were eight calibration stars spanning a magnitude range from H_VVV = 15.8 to 12.8. From these stars, we derived the flux relation:

$$\begin{eqnarray}&&{H}_{\mathrm{VVV}}={H}_{\mathrm{Clio}}+6.813\pm 0.015\end{eqnarray} \tag{ 2 }$$

where H_Clio is calculated from the peak counts in the PSF (i.e., ${H}_{\mathrm{Clio}}=18-2.5{\mathrm{log}}_{10}{f}_{\mathrm{peak}}$).

Zoom In Zoom Out Reset image size

To estimate the uncertainties in the measured fluxes, we divided the 20 exposures into two subsets and coadded each subset into new composite images 1 and 2. Then we measured the fluxes of stars in each composite image and took the difference of the two measurement, Δ_i = (flux_i,1 − flux_i,2), for each star, i. In the limit, as the number of stars at a particular apparent magnitude goes to infinity, the uncertainty in the flux in the original coadded image is

$$\begin{eqnarray}&&{\sigma }_{\mathrm{tot}}\equiv \displaystyle \frac{\sqrt{\langle {{\rm{\Delta }}}^{2}\rangle }}{2}\end{eqnarray} \tag{ 3 }$$

for that magnitude. In a real image, there is a finite number of stars spanning a finite set of discrete flux values. So to estimate 〈Δ²〉, we assume that the uncertainty in the flux is a smooth function of the peak flux. We fit a power law to the measured ${{\rm{\Delta }}}_{i}^{2}$ as a function of flux_i from the original image. Solving Equation (3) for σ_tot gives us the typical measurement error as a function of flux.

We began by fitting for sources in a rectangular (20 × 13) substamp (Figure 2) containing the source star of the event and its immediate neighbors. These fits use the empirical PSF described in Section 2.3 plus a constant background level for the image. Sources were recovered iteratively by examining the residuals to the fitted image and comparing them to the HST NICMOS images of the field. In total, 11 sources were recovered in the (20 × 13) substamp. All of them can be cross-matched to stars in the HST NICMOS imaging. Their properties are given in Table 1. The residuals to the 11-source fit are shown in Figure 2. There is no indication of an additional source within the white annulus.

The one exception to this could be if the lens were very near to source #3, which happens to lie at the expected lens–source separation. In that case, light from the lens might be absorbed into the fits for source #3. To check for this possibility we compared the fluxes of the stars to the fluxes measured from the HST image from 2008 (Figure 3). The fluxes for all the sources are consistent between MagAO and HST, with no significant excesses. Therefore, we conclude that source #3 is not hiding significant light from the lens.

Zoom In Zoom Out Reset image size

Finally, from the MagAO image, we measure H = 17.38 ± 0.04 for the source at the position of OGLE-2007-BLG-224 (source #1). This is slightly brighter than the microlens source flux, as reported in Gould et al. (2009; H_S = 17.50 ± 0.02), which is marginally inconsistent at 2.7σ. The excess would correspond to an H = 19.8 star. However, the measured source flux is also consistent with the flux measured for the same source in the HST imaging. Gould et al. (2009) state that the HST source is consistent with zero additional flux, aside from the microlens source. Even if the excess is real, it is in the wrong place to be the lens, although it could be a companion to the microlens source.

To confirm that there are no additional sources in the field, we conduct a blind search. We seed fits for a twelfth source at every pixel over a 135 × 135 square surrounding OGLE-2007-BLG-224S. The peak flux for the PSF is initialized at 50 counts, and then this source is fit freely, simultaneously with the 11 known sources. We consider a source detected if the final fit is inside the search grid, within 20 pixels of the starting location, and has a fitted peak flux of at least 10 counts. We tested our method for 11-star fits to confirm that the two faintest sources (10 and 11) are recovered.

There are two classes of "detections" that survive these cuts. First, there are "detections" very close to the centroids of known sources in the field. In these cases, the twelfth source partially supplants the known source; the sum of their fluxes remains the same as for the known source alone (in particular, this is true for source #3, i.e., there is no evidence for an additional source "hidden" in the wings of this star). Hence, we apply an additional cut where the new detection should be at least 1 FWHM from a previously known source to eliminate this class of "detections." Second, there are "detections" of a twelfth source along the bottom edge of the substamp. There are unexplained residuals in this region at the level of ∼20 counts, possibly from unresolved faint sources, associated with the wings of bright sources outside the substamp, or simply spurious. However, because this string of "detections" are also separated from the source by almost twice the expected distance traveled by the lens, even if they are real, they are unlikely to be the lens.

Thus, this blind search did not reveal any new sources that are plausible candidates for the lens of OGLE-2007-BLG-224.

To assess the limits on nondetections in the image, we inject fake sources into the image and attempt to recover them. Fake sources are injected at every pixel over the 135 × 135 grid used for the blind search. Then, fits are seeded on a 3 × 3 "mini-grid" (with spacings between points of 2.5 pixels) around the injected source. A source is considered to be recovered if more than five out of the nine fits initialized from the mini-grid around the injected source meet the following conditions: (1) the retrieved centroid is less than 1 pixel from the injected location and (2) the retrieved peak flux is within 50% of the injected value. Thus, we can place a detection sensitivity limit at every grid point around the microlensing source star. Note that this detection criterion is considerably more stringent than that used for the blind search. Hence, it results in a conservative lower limit on the brightness of recoverable sources at each location.

At the expected location of the lens, we find the 3σ flux limit to be H > 20.57. Along the annulus defined by μ_{rel, hel}, the limits are generally better except where the annulus intersects the PSF of stars 3 and 4. Although the limits nominally reach H ∼ 22.1 in some places, these limits are highly dependent on the exact level of the background in a particular pixel, and thus, may be overoptimistic. However, stars 10 and 11 (with magnitudes H = 21.6 and H = 21.4, respectively) are clearly recovered in the blind search. Hence, we can conservatively say that the limits on the lens flux range from 21.6 to 20.6, depending on the location along the annulus.