Hello everyone,
I'm looking for some advice regarding how to go about submitting a star to VSX if the data from each survey is all out of phase with each other. To see what I mean, refer to the attached plot. As you can see, the data from ROTSE1:Orphans, ASAS-SN, and SuperWASP are all out of phase with each other. Please note that in the plot, I intentionally shifted the data sets up and down, so that they are easy to see individually, and are not on top of each other. The point is they are all out of phase. Also, please note that there is a nearby star of similar magnitude, so ROTSE1:Orphans and SuperWASP data is blended for this object.
I have triple checked, all of the data are in HJD, and I have not converted between MJD and HJD more than once for any of them.
I also do not think this is a physical phenomenon. Aware that the three surveys I used took data years apart from one another, I plotted the epoch of maximum eclipse for each of the three data sets against the assigned phase value, and it formed a pretty good line. However, one can hardly make conclusions based on 3 data points, and I can't think of a physical phenomenon that would change the phase of the maximum without also changing the period. The period 0.410087 seems to work equally well over for each survey.
Furthermore, I have 2000 days of ASAS-SN data. If this was a physical phenomenon, one would expect the data to look "smeared" as the phase of the maximum changes over time.
NSVS, ASAS3, and CRTS all have problems with their data on this object, and I couldn't make a good phase plot out of them.
My question is, can I make a submission to VSX anyway? I was thinking that maybe I could just submit multiple plots, one of each survey each using the same period, with a 0-point calibration anchored to ASAS-SN?
Thanks for any response,
Grant
So there is no color variation I guess that could explain this, right? It seems a bit far fetched, I guess all three surveys use unfiltered photometry, so in theory if the CCDs had very different responses ... I dunno, just thinking aloud. Note tthat also the amplitude of the variation is different for the light curves.
BTW, what is the x-axis of the plot, and in what units? A period of 0.41 (days?) does not seem to match the scale of the plot (more like 0.5). And what is the error estimate for the period? If you are off only by 5 seconds for a 0.5 day period, this will translate to an offset of an hour per year of difference in the epochs, right?
CS
HB
This is a case of a period changing contact binary.
The time span of 2000 days is too short to show obvious smearing in the ASAS-SN data-set, while the time span covered by ROTSE, SWASP and ASAS-SN together is much longer (~7000 days?) and shows the phase shifts due to the changed period much more clearly.
To show it with an example (note that I don't claim that this star exibits a constant period change, that would need a more detailed investigation):
Typical values for constant period change rates for W UMa stars are in the order of 10-7 days per year.
If you take a constant period change rate of 5 . 10-7 days then you get for a period of 0.4 days a phase shift of about 2% over a time span of 2000 days and 20% over a time span of 7000 days.
Eric
Hi Eric,
Thanks for your response. How does one go about making a submission of a system with a changing period? How do you make a phase plot with data from multiple surveys? Could I just upload multiple plots of different survey's data and have a comment that the period is changing?
Grant
Hey Grant,
Period change is very tricky and hard to pin down well, especially as the changes tend to be subtle. You might have a period change, but if it really is that noticeable, then you should be able to tell. For each dataset, calculate the best fit period. There should be an obvious difference (including errors) between each set. This may very well be the case, but it's unlikely.
What can cause a similar effect, (and is in general more likely) is that the period you are using to phase the data isn't precise enough. For one dataset it looks fine, but there is enough error in the value, that if you have a long baseline that error becomes obvious. The best test for this would be to try and calculate the period using all three datasets together. If you can come up with a coherent solution, then this is likely your problem. If however, you cannot, then I would explore the idea of period change again. If you have any questions about what I have said, please don't hesitate to ask.
Thanks,
Bert Pablo
Staff Astronomer, AAVSO