On the night of Sep 27/28, 2023, under good conditions, I acquired images of the occultation of a background star by the Trans-Neptunian Object (19521) Chaos. A very brief summary is that I measured the background star to disappear for about 33.9 seconds, as this animated GIF of my images shows.
Contents:
This occultation was predicted by the Lucky Star project , led by Bruno Sicardy. You can find the information they provided to observers at
As you can see from the predicted shadow path below, we in Rochester, NY
RIT Observatory: longitude -77.66423 -77:39:51
latitude +43.07538 +43:04:31
altitude 172 m
were outside the best guess at the shadow's extent -- but inside the +/- 1-sigma limits (marked by dashed red lines). So, I decided to try to observe it.
The target star
RA = 06:09:56.9 Dec = +31:21:23.2
is pretty faint, so it wasn't clear what sort of equipment might be best to capture the event.
I ran some tests on the previous night which indicated that using our 12-inch telescope, and ASI 6200MM camera without a filter ought to work. Specifically, a one-second exposure time of a small sub-frame region would provide adequate signal-to-noise ratio and a reasonable time resolution, with a sequence of
Here's a prediction from my CCD signal-to-noise calculator ; I've highlighted the V-band magnitude of the target star from the UCAC4 catalog.
Filter: none Tel_diam: 30 (cm) Overall QE: 0.5 Pixsize: 1.0 (arcsec/pixel) Readnoise 3 (electrons) Sky mag: 17 (mag/sq.arcsec) Airmass: 1.4 Ext_coeff: 0.2 Exptime: 1 (sec) FWHM: 3 (arcsec) Aper_rad: 3 (arcsec) mag 12.00: star 17514 sky 6842 read 254 -> S/N 111.64 mag 12.20: star 14568 sky 6842 read 254 -> S/N 98.97 mag 12.40: star 12117 sky 6842 read 254 -> S/N 87.42 mag 12.60: star 10078 sky 6842 read 254 -> S/N 76.90 mag 12.80: star 8383 sky 6842 read 254 -> S/N 67.38 mag 13.00: star 6972 sky 6842 read 254 -> S/N 58.78 mag 13.20: star 5799 sky 6842 read 254 -> S/N 51.07 mag 13.40: star 4824 sky 6842 read 254 -> S/N 44.18 mag 13.60: star 4012 sky 6842 read 254 -> S/N 38.07 mag 13.80: star 3337 sky 6842 read 254 -> S/N 32.67 mag 14.00: star 2776 sky 6842 read 254 -> S/N 27.94 mag 14.20: star 2309 sky 6842 read 254 -> S/N 23.81 mag 14.40: star 1920 sky 6842 read 254 -> S/N 20.22 mag 14.60: star 1597 sky 6842 read 254 -> S/N 17.13 mag 14.80: star 1329 sky 6842 read 254 -> S/N 14.47 mag 15.00: star 1105 sky 6842 read 254 -> S/N 12.20
The practice observations indicated that the actual scatter in repeated measurements of the target was between 0.06 and 0.10 mag, perhaps a bit more pessimistic than the expectations -- but then, some of the inputs to my calculator are just rough estimates.
Here's the field of the event. The target lies in the crosshairs inside the small blue box near center; the bright star is the Mira variable BU Aur. The field of view of this finding chart is about 15 x 15 arcminutes.
I defined a small sub-region on the camera just 116-by-116 pixels in size, large enough to encompass the target star as well as two nearby and slightly brighter references. The figure also shows the location of a "blank sky" region that I will use for reference photometry, as explained below.
label APASS9 B V
----------------------------------------------------
B 25613642 12.792 12.422
C 25613645 13.639 14.844
----------------------------------------------------
I started acquiring images at 03:05 AM EDT = UT 07:05, about nine minutes before the predicted time of occultation, and continued until 03:26 AM EDT = UT 07:26, well after the event. The skies weren't clear during this time, with light and not-so-light cirrus everywhere and light from the waxing gibbous Moon on the other side of the sky. Fortunately, the target area fell into a reasonably clear hole during the predicted time ... although clouds arrived just a few minutes later, as this graph of sky brightness shows.
I reduced the images in the following manner:
I didn't divide by a flatfield frame because the field was very small and close to the optical center of the field, and because the background sky levels were very low -- just 3 or 4 ADU, not much larger than the readout noise. There is no gradient larger than about 0.5 percent across this tiny area.
There are two possible problems with the times I have associated with each measurement of the target star's brightness.
Let me discuss them briefly in that order.
First, the round-off issue. I used the program MaximDL (version 6.16) to operate the camera and save images to my computer's hard drive. The program read the computer's system clock, set via NTP, and placed a value in the FITS header like this:
DATE-OBS= '2023-09-28T07:05:33' /YYYY-MM-DDThh:mm:ss observation start, UT
Note that the time is "starting time" for the exposure. Since all exposures were 1-second long, it was easy to add 0.5 seconds to yield the time of mid-exposure (which I did, a bit later in the process).
But note also that the time is recorded only to the nearest integer second. Since the dead time was about 0.6 seconds between successive exposures, this mean that a sequence of images might have times recorded as follow:
DATE-OBS= '2023-09-28T07:06:14' /YYYY-MM-DDThh:mm:ss observation start, UT DATE-OBS= '2023-09-28T07:06:16' /YYYY-MM-DDThh:mm:ss observation start, UT DATE-OBS= '2023-09-28T07:06:18' /YYYY-MM-DDThh:mm:ss observation start, UT DATE-OBS= '2023-09-28T07:06:19' /YYYY-MM-DDThh:mm:ss observation start, UT DATE-OBS= '2023-09-28T07:06:21' /YYYY-MM-DDThh:mm:ss observation start, UT DATE-OBS= '2023-09-28T07:06:22' /YYYY-MM-DDThh:mm:ss observation start, UT DATE-OBS= '2023-09-28T07:06:24' /YYYY-MM-DDThh:mm:ss observation start, UT
These do NOT accurately reflect the actual starting times of the images. So, I fit a linear function to the times recorded in the FITS headers for the entire dataset (733 images, covering about 20 minutes), then interpolated using this function to find the mid-point of each image. I then used that interpolated time -- which shows a smooth, linear progression -- in all subsequent analysis.
The second issue is that there may be some fixed systematic offset that affects all the recorded times, if the computer's clock was offset from the actual time. Several hours after the occultation, later in the morning, I went back to the observatory and logged into the computer which recorded the images. I displayed the time on the computer's screen in full HH:MM:SS format and watched the display change each second. I used my iPhone, which synchronizes its time in a completely different manner, as a reference for time. I held my iPhone up, next to the screen, and displayed its time in the same HH:MM:SS format. Watching the two displays, side-by-side, I saw no obvious offset.
I conclude that it is likely, at least, that the recorded times have no zero-point offset larger than +/- 1 second.
Let's first make a set of measurements in the ordinary way: in each image independently, detect each star, measure its position, and then perform aperture photometry at those positions, using a local sky background based on annuli around each star. I used parameters
I used the XVista software package to carry out all these reductions and measurements.
Next, I performed inhomogeneous aperture photometry to bring all the images to a common photometric reference point. Because star "B" was brighter than the others, it was given most of the weight in this procedure. The graph below shows the zero-point adjustment made to each image; points higher in the graph correspond to images in which all the stars appeared fainter. Note how well the changes mirror those in the background sky (shown in the graph above).
Basically, in the second half of the run, the target star was either un-detectable or just barely measured. It's a good thing that the occultation occurred just before the clouds arrived.
One of the results of the ensemble photometry is an estimate of the uncertainty in each star's average (instrumental) magnitude. The graph below shows those estimates for -- reading left to right -- stars "B", "C", and "target". The typical uncertainty in measurements of the target star was about 0.10 mag.
Finally, here are the light curves of the three stars based on this analysis. I've shifted all the instrumental magnitudes so that star "B" has a value of 12.4, which matches the APASS9 value of V = 12.4. First, all the measurements ...
... and now, a closeup of the region around the occultation.
The target star disappeared in the following manner:
frame_index mid-exp_UT mid-exp_JD target_star
--------------------------------------------------------------------------
323 07:14:08.29 2460215.8014848 present
324 07:14:09.90 2460215.8015035 missing
344 07:14:42.20 2460215.8018773 missing
345 07:14:43.81 2460215.8018960 present
--------------------------------------------------------------------------
Of course, when the target star is blocked by the asteroid, the light of the combination grows much fainter; so faint, in fact, that my software did not detect the source, and therefore made no measurement. It is possible that there might still be some light at that position, which might yield some interesting information about the details of the ingress or egress from the occultation. But the normal mode of analysis would not provide any values for very low levels of light at the target's position.
Therefore, I decided to carry out something like "forced photometry" at the target position, even if it wasn't detected as a source. The basic idea was
I used one hundred images (frame indices 200 - 299), all taken before the occultation when conditions were good, to measure the offset between "B" and the target. I also chose a location, marked "blank" in the figure above, where no stars appeared. By measuring the light at this "blank" location, I would learn how my measurement routines would behave when there was no real object present. Recall that the plate scale is roughly 1 arcsec per pixel.
object offset_x_(pix) offset_y_(pix) --------------------------------------------------------- target -65.7 -34.5 blank -35.7 -4.5 ---------------------------------------------------------
I then processed all images as before, but this time removed all real detections of the target star and replaced then with measurements at the target's offset position. I also measured light at the "blank" position as well. After extracting instrumental magnitudes as these positions, I carried out inhomogeneous ensemble photometry as before.
The resulting light curve is shown below. Note that the "blank" region is labelled as "D" in the graphs. First, the entire dataset ...
.... and now a closeup around the time of occultation.
The fact that the "measurements" at the position of the target star during the occultation have values (when they exist) which are very similar to those of the "blank" region indicate that there was no significant light detected from the target star during the entire occultation.
As far as my measurements can say, the star changed from "completely visible" to "completely invisible" between one frame and the next; and then reappeared in the same discontinuous manner.