A recent series of E-mail messages was sparked by the discovery of an eclipsing binary star by the STARE team; the new eclipser is close to the known variable star IO Aur. I decided to look at the Mark IV data for stars in a small region around these two stars. Below, I analyze the time-series photometry of Mark IV data in two ways:
Executive summary:
It takes quite a bit of work to push Mark IV data through an ensemble solution, but it clearly improves the quality, especially in V-band. Anyone planning to use Mark IV data to look for variables with an amplitude of less than, say, 0.25 mag, should be prepared to do the extra work.
Here is an image showing a square 0.6 degrees on a side, centered on IO Aur: RA=05:15:51.4 Dec=+38:34:37.
A closeup of the center of the field shows the known variable IO Aur, which was not present in the Mark IV database; it is a bit too faint. Star J = GSC 2896-1308 is the new eclipsing binary. This chart is 14 arcminutes on a side.
Note that in this relatively crowded field, many stars have neighbors less than one arcminute away. The Mark IV pixels are about 7.5 arcseconds across, which means that in this field, the TASS aperture photometry may have problems: if the seeing or PSF changes from one image to the next, the software may include more or less of the light from a neighbor.
Suppose we take magnitudes straight from the Mark IV database. I will refer to these as raw magnitudes. Running a search for all stars within a radius of 15 arcminutes of IO Aur, the Mark IV database returns 1025 measurement pairs (one V and one I magnitude per pair) from 182 distinct objects. I required at least 10 measurement pairs per object, and so winnowed down the dataset to 32 stars. The timespan covered is JD 2,452,902 (2003 Sep 19) to 2,452,999 (2003 Dec 25).
First, let's look at the light curves of 14 stars from this 32-star set. Remember that star J is the new eclipsing binary.
V-band:
I-band:
Even the bright objects have very messy light curves, especially in I-band. Some of the variations, however, may be common to all (or most) of the objects; perhaps ensemble photometry of this set will remove them.... Before we try that, look at the mean and standard deviation from the mean for all stars. I plot the reflection of the I-band scatter for clarity.
As one would expect, there is more scatter for fainter stars. However, the floor of the scatter is about 0.05 mag in V-band, and about 0.08 mag in I-band. That's much larger than one would expect for such bright stars.
Some of the variation we see in the Mark IV data may occur because of some condition which affects all stars (nearly) equally: clouds on one night, perhaps, or an error in flatfielding with a large spatial scale. Recall that the Mark IV images are over 4 degrees on a side. This set of stars would fit in a square less than 1/8-by-1/8 of the entire field. I have seen definite recurring errors as a function of position on the frame in other Mark IV data; if the scale of the errors is less than a half-degree, then we might be able to improve the relative photometry via Honeycutt's inhomogeneous ensemble photometry technique. This has been tried before, both by me and by Andrew Bennett and again by Andrew Bennett. How well would it work in this particular case?
After running the ensemble photometry code to make a solution, we can look at the "offset": the amount by which the magnitudes from each image must be moved up or down to bring them into best agreement with the other frames.
Note that some shifts are very large, with three over 0.30 mag in I-band. This is not good news for users of values straight from the Mark IV database. The shifts in V-band are consistently smaller.
Note also that the shifts in V and I are roughly correlated. This could be due either to the weather, would ought to affect both cameras, or due to position on the frame, since both cameras place a star at roughly the same spot in the field.
After one has shifted the images in the optimal way, one can re-calculate the mean and stdev of magnitudes from each star. Here is the scatter-vs-mag plot for the ensemble solution:
The ensemble solution has decreased the scatter in both sets of measurements. The V-band data now looks pretty good, with a floor of about 0.01 mag; the I-band data still has a larger-than-expected floor of 0.03 mag or so.
The eclipsing binary star "J" is the I-band outlier at mag I=9.5, with scatter 0.13 mag. In the V-band, it appears at V=12.3, with scatter 0.092 mag; this does not appear outside the locus for ordinary, faint stars.
We can also plot the light curves of the stars from the ensemble solution. First, V-band:
Most of the stars appear to be steady in light. Fainter stars, below V=12, have a random scatter. The new eclipsing binary "J" doesn't stand out in V-band.
Now, in I-band:
The scale is slightly different here, but the photometry does look a bit worse overall. Star "J" is much brighter in I-band, yet it has a very large variation. Is that due to eclipses, or just bad measurements?
Let's look at the light curves of star "J", the new eclipsing binary found by STARE, in both "raw" and "ensemble" photometry. I use the STARE period of 2.48059 days mentioned by Sebastian Otero in his E-mail message. However, since I don't know the epoch of minimum (or maximum), my phased light curves have an arbitrary phase shift; don't expect to see extrema at phases 0.0 and 0.5.
Below is a phased light curve in V-band, showing "raw" and "ensemble" data:
Hmmmm. Does this show any real variation? It's hard to tell. There isn't much difference between the raw and ensemble light curves. The amplitude of change of the ensemble data is about +/- 0.15 mag, but other, probably constant stars of similar V-band brightness have random scatter of +/- 0.10 mag in the ensemble solution. The STARE light curve (reproduced below) has an amplitude of only +/- 0.06 mag, so we would probably not expect to see the eclipses in the Mark IV V-band data.
Now, let's look at the I-band phased light curve. The star is much brighter in I-band, bright enough that other stars of similar brightness have scatter of only 0.03 mag or so in the ensemble solution. We might expect to see at least hints of the eclipses in the I-band ensemble photometry.
Once again, there is no clear evidence for the eclipses seen in the STARE dataset. There is one very faint point in the TASS dataset, which could be due to an observation exactly at eclipse. I would need to know the ephemeris for the star to check that. As it is, all I can say is that this star does show larger-than-typical scatter for its I-band magnitude, which is consistent with the fact that it is an eclipsing binary star.
