The Kepler space telescope, during its first mission, rolled over four times per year in order to keep its heat shield facing the Sun. At what dates (in that BJD time format) was this done during the 4+ year duration of the primary mission?

I note that the three main periods of anomalies in the light curve of KIC 8462852 occurred with exactly two years intervals. About day 40, day 800 and day 1500+. This leads me to believe that the data is the result of some kind of instrument failure, maybe a degrading CCD element, a bad pixel.

  • Did the three main anomalies occur while the telescope was oriented the same way?
  • Obviously, the telescope had the Sun in the same direction during all of the three recorded anomalies. But was there some difference in orientation in the years in between every other year?
  • While the telescope was held extraordinarily still, could the remaining tiny wobbling be enough to make a dead pixel go in and out of the light from a star, in order to cause the dips in the light curve of KIC 8462852? Maybe more so in the end as the second reaction wheel was failing, explaining the greater anomaly after day 1500+.

Please note that Kepler's orbital period is slightly shorter than an Earth year. It is of course Kepler's orbital period which is of interest here.

enter image description here

Source of illustration:


  • 1
    $\begingroup$ How do the smaller anomalies in figure 4 fit in to your theory, around days 216, 378, and 1255? Why would a bad CCD element only produce its bad data at biannual intervals, instead of over an entire "quarter"? $\endgroup$ Oct 21, 2015 at 15:04
  • $\begingroup$ @RussellBorogove I'd love to have the raw light curve data. But at least the anomalies you mention are way smaller than 1% and would not have cause any alarm. Maybe one of them represents a real exoplanet transit, that head and shoulder shape maybe looks like a ringed planet candidate. The biannual variation is one of the things I'm asking about. Maybe every other time the telescope was oriented in a slightly different way for some obscure practical reason. 95,000,000 pixels towards 150,000+ stars. $\endgroup$
    – LocalFluff
    Oct 21, 2015 at 16:40
  • $\begingroup$ @RussellBorogove One potential explanation to why a bad pixel gives sporadic results over time, which I mention in my question, is that it is completely dead. It just passes through the star as the telescope wobbles. A single dead pixel would wobble in and out of the star's light more at the end as the reaction wheel gradually fails. $\endgroup$
    – LocalFluff
    Oct 21, 2015 at 16:45

2 Answers 2


Kepler rolled between the quarter dates found at this site. At first glance, they don't seem to correlate, although I need to do more work to actually line it up. As Kepler rolls the data between data releases, it's possible to look at whole datasets, and see if they occur at a common point in the cycle. I can tell you the following:

  1. A dip occurred at the start of KPLR008462852-2009259160929, ending at KPLR008462852-2009350155506
  2. No other dips correspond to transitions in quarters, except possibly one at the very last quarter (KPLR008462852-2013131215648)
  • $\begingroup$ The two last, and biggest, anomalies seem to occur very near quarter changes, if I do the date conversions right. $\endgroup$
    – LocalFluff
    Oct 21, 2015 at 12:39
  • $\begingroup$ @LocalFluff: Found a better way to prove, I've edited my answer accordingly. $\endgroup$
    – PearsonArtPhoto
    Oct 21, 2015 at 13:21
  • $\begingroup$ I can't quite make certain sense of the dates. However, Kepler seems to have changed its orientation days 787 and 1467, to the same orientation, which is just before the two major anomalies of 16% and 22%. And also day 49 which unfortunately is about 10 days before the first (<1%) anomaly. But the operations before that date were not regular since it was new in operation. Maybe the telescope was oriented in the same direction during the week before GO Cycle 1. $\endgroup$
    – LocalFluff
    Oct 23, 2015 at 0:35

Section 4.1 on page 8 of the original report on the phenomenon discards this possibility, albeit indirectly:

The Kepler light curve for KIC 8462852 is unique, and we have thoroughly explored the raw data for defects/instrumental effects, which could cause the observed variations in KIC 8462852’s flux. We use the P Y K E software tools for Kepler data analysis to check the data for instrumental effects. We check the following possibilities:

  • We checked that the same flux variations, i.e., the ‘dips’, are present in the SAP FLUX data set

  • We verified that data gaps and cosmic rays events do not co- incide with the dipping events, as they are prone to produce glitches in the corrected fluxes.

  • We verified at the pixel-level that there are no signs of peculiar photometric masks used in making the light curves.

  • We verified at the pixel level that the image light centroid does not shift during the ‘dipping’ events

  • We inspected light curves of neighboring sources and find that they do not show similar variability patterns in their light curves.

  • We determined that CCD cross talk and reflection cannot be the cause (Coughlin et al. 2014).

  • We verified with the Kepler team mission scientists that the data were of good quality.

This analysis concludes that instrumental effects or artifacts in the data reduction are not the cause of the observed dipping events, and thus the nature of KIC 8462852’s light curve is astrophysical in origin.

The light curves obtained by Kepler were consistent with data from the Nordic Optical Telescope (though just for the star, not for the times when there are dips in its light) - section 2.2, page 4 of original report:

we used the co-added FIES spectrum to determine the stellar effective temperature Teff , surface gravity log g, projected rotational velocity v sin i, metal abundance [M/H], and spectral type of KIC 8462852... The temperature we derive (Teff = 6750 ± 140 K) is consistent with the photometric estimate of Teff = 6584+178 −279 K from the revised Kepler Input Catalog properties (Huber et al. 2014), as well as with Teff = 6780 K derived from the empirical (V − K) color-temperature relation from Boyajian et al. (2013). The projected rotational velocity we measure v sin i = 84 ± 4 km s−1 is also well in line with the one predicted from rotation in Section 2.1, if the 0.88 d signal is in fact the rotation period

Several of the verification techniques would have revealed whether the dips were due to Kepler's positioning - the masks would have been out of alignment, other stars would have shown the same phenomenon. There is no known mechanism by which these readings could have been caused by Kepler itself.

The pixels of Kepler's CCD are calibrated according to procedures described in Pixel-Level Calibration in the Kepler Science Operations Center Pipeline:

the flight data is subject to numerous data processing steps that are performed in a complex automated pipeline... [it] is composed of a number of modules that operate sequentially (Figure 1)... Additional modules operate in parallel to monitor the instrument performance and provide target management... The raw data include photometric (target and background) pixels, along with a subset of the CCD termed “collateral data” which includes masked and virtual (over-clocked) rows and columns that are used primarily for calibration... Many of the primary corrections use external models of each CCD that were developed from pre-flight hardware tests and FFI data taken during commissioning... We discuss how these models are applied within CAL to correct for 2D bias structure, gain and nonlinearity of the conversion from analog-to-digital units (ADU) to photoelectrons, local detector electronics effects (undershoot and overshoot), and flat field (variations in pixel sensitivity). Other signals that are corrected include excess charge from saturated stars that leak into the masked and virtual regions, cosmic ray events, dark current and smear.

flowchart of automated procedures to calibrate Kepler's CCD

  • $\begingroup$ They don't address the issue of the anomalies all occurring during the same quarters if the telescopes orbit (AFAIK from the schedule link in PearsonArtPhoto answer above). Those were planned positionings to keep the sun shield towards the Sun, that should ring an alarm bell. The erratic signal during the final quarter could be an electrical glitch for pixels on an area covering the whole star. Since the anomalies seem to have been discovered manually (by zooniverse), and they were filtered away by the algorithms, maybe an algorithm to scan the data for anomalies could find more of interest. $\endgroup$
    – LocalFluff
    Nov 2, 2015 at 10:47
  • 1
    $\begingroup$ @LocalFluff but wouldn't the pattern have occurred with a bunch of stars if it was related to orientation and position in any way? $\endgroup$
    – kim holder
    Nov 2, 2015 at 16:31
  • $\begingroup$ Because only a few of the 95 million pixels caught the light from that particular star, one out of 0.16 million stars. Were'nt most pixels lightless all the time? Sure, if the centroid of the star didn't move in an anomaly, then I guess more than one pixel was faulty in that area of the CCD. $\endgroup$
    – LocalFluff
    Nov 2, 2015 at 16:36
  • 1
    $\begingroup$ @LocalFluff intermittently faulty in a way that couldn't be detected by the mission team? I suppose that wouldn't be more remarkable than other proposed explanations :P $\endgroup$
    – kim holder
    Nov 2, 2015 at 17:22
  • 1
    $\begingroup$ Please pay attention to section 2.2 and figure 5 in the original report. There were follow-up observations of KIC 8462852 done for high resolution spectroscopy and spectral energy distribution analyses by FIES spectrograph of the Nordic Optical Telescope, Instituto de Astrofisica de Canarias, La Palma, Canary Islands, Spain.Derived $T_\text{eff}$ and light curve for the two observations fit Kepler Input Catalog. That should help nail it that it isn't Kepler instrumental effects or data reduction artifacts. $\endgroup$
    – TildalWave
    Nov 4, 2015 at 16:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.