The Gaia space mission is crafting revolutionary astrometric, photometric and spectroscopic catalogues that will allow us to map our Galaxy, but only if we know the completeness of this Gaia-verse of catalogues: what stars does it contain and what stars is it missing? We argue that the completeness is driven by Gaia's spinning-and-precessing scanning law and will apply this principle to the Gaia-verse over this series. We take a first step by identifying the periods in time that did not contribute any measurements to Gaia DR2; these gaps create ribbons of incompleteness across the sky that will bias any study that ignores them, although some of these gaps may be filled in future data releases. Our first approach was to use the variable star photometry to identify the 94 gaps longer than 1% of a day. Our second approach was to predict the number of observations of every point on the sky, which in comparison to the reported number of detections revealed additional gaps in the astrometry and spectroscopy. Making these predictions required us to make the most precise, publicly-available determination of the Gaia scanning law. Using this scanning law, we further identified that most stars fainter than G=22 in DR2 have spurious magnitudes due to a miscalibration resulting from a thunderstorm over Madrid. Our list of gaps and precision scanning law will allow astronomers to know when Gaia's eye was truly on their binary star, exoplanet or microlensing event during the time period of the second data release.
Determination of Gaia scanning law
We used the times at which Gaia observed each of the variable stars to determine the scanning law that Gaia was following. Our model was in terms of a time-varying rotation of the nominal scanning law and manifests as shifts in the locations of the fields of view perpendicular to the scanning direction. This model was fit to the locations of all the variable stars with epoch photometry at their times of observation and these data points are shown in the bottom left and bottom middle panels. The bottom row of panels zoom in to the indicated time periods in the top panel, with the curves in the bottom left panel having been linearly offset by the indicated values to maximise the dynamic range of the y-axis.
Gaps in Gaia data-taking
We identified a number of periods which did not contribute any measurements to the Gaia DR2 data products. The gaps in the astrometry (green), colour photometry (blue) and spectroscopy (red) data-taking are visible as dips in these three deviancy time series. Deviancy is defined as the mean fractional difference between the maximum observed number of detections minus the predicted number of observations of stars in the pixels nearest to the two fields-of-view at each time. We have applied a smoothing filter with a one-revolution window to make the trend visible and have applied scaling and offsets for visualisation purposes. The gaps identified from the epoch photometry are shown in red, while the gaps given in the DPAC papers or identified by this work are hatched. Textual annotations give explanations for some of the major gaps.
Comparing Predictions to Observations
We expect that the detection efficiency for bright stars should be near 100%, and thus that the maximum number of detections of any source in each pixel should be approximately equal to the number of observations of that pixel. For the astrometry, the number of detections is taken from astrometric_matched_observations, for the photometry we use the maximum of phot_bp_n_obs and phot_rp_n_obs and for spectroscopy rv_nb_transits. Here we show the fractional difference between the maximum number of detections and the predicted number of observations, before and after removing the gaps in Gaia data-taking identified in this work. Both plots in each row are scaled by the interquartile range of the plot in the left-hand column.
We note that there does appear to be another gap visible in the fourth panel, that crosses from the bottom right to the top middle. We were able to isolate this gap to the period OBMT = 1389.2 - 1391.7 rev. It is visible as a dip in the deviancy shown in above. This is not a genuine gap in photometric data-taking, but rather a drop in the efficiency of the colour photometry data-taking. Gaia Helpdesk informed us that this was caused by a thunder storm over Madrid on 11th October 2014 which interrupted transmission of certain data packets.
There are some sources with more reported detections used by the astrometric pipeline than we predict could have been taken by Gaia during DR2, i.e. there is a detection excess. We show the magnitude distributions of the sources with excess detections for increasing excess. The reference distribution shows detections we identified in the epoch photometry as spurious by being too close in time to the previous or subsequent detection. This demonstrates that the cause of the excess detections are spurious duplicate detections which are being wrongly counted as detections of genuine sources. The smoking gun that alerted us to the issue of duplicates was a bright G=8.77 star at the North Ecliptic Pole. This star has an impossibly large 361 matched observations in DR2. No source should have more than 264 observations in DR2, implying that almost one hundred of the detections of this star are spurious duplicates.
Spurious Dim Sources
There are sources in Gaia with reported mean G-band magnitudes as faint as G=23.5. We identified that an overwhelming majority of these sources were observed during two narrow time windows and so are likely due to missing calibration data packets preventing an accurate magnitude determination. The histograms show the faint magnitude distribution of all sources in Gaia DR2 (blue), of sources predicted to have been observed at least once during the two time periods OBMT = 1388 - 1392 rev and OBMT = 2211 - 2215 rev (red), and of those sources with no predicted observation during those periods (green). Almost all of the extremely faint sources can be traced to those time periods.