The latest version of the python notebook I am working in has been pushed to GitHub. The initial threshold values I decided to apply were max x-range = 70.0 => log(70.0)=1.845 and R min values = 6.0 => log(6.0)=0.778.
This narrowed down the DR13 list from 163,661 stars to 8,301 stars that satisfy these thresholds.
The following histogram is produced.

However, it would be more beneficial to look at this along side the original DR13 histogram.

In the yellow box is the boundaries of this potential binary list. It appears that these potential binaries lie in the second lower tail of the primary cluster of stars shown on the far left.

Let’s try to refine this list of potentials.
I performed some basic analysis of what the threshold values should be based off of our known binaries of DR12. I am still looking primarily at the relations between max x-range and minimum R values for now. Better quantities to use for such analysis of potential binaries is in the works such as bisector properties (slope differences from visit to visit,etc).
After examining DR12’s Drew’s List and SB9 SB2s the following can be said:
a) Each list has an average max x-range value of 53.62 and 84.8 respectively
b) Each list has an average R min value of 7.77 and 18.6 respectively.
c) The minimum value a binary in Drew’s List possessed for max x-range was 0.579 and 0.87 for R minimum value.

I chose to focus more on analyzing Drew’s List for the sake of having more confidence with that particular list classification. Drew’s List was visually confirmed binaries by CCFs. The majority of this list holds CCFs that contain duo peaks. They are more obvious binaries than SB9s. But! I am aware that the majority of binaries are also tricky to classify due to their CCFs having merged peaks. So SB9 can’t be neglected in these threshold considerations.

d) The minimum value a binary in SB9 list possessed for max x-range was 0.289 and for R min was 0.46.
Due to SB9 containing a lot of merged peaks in the CCFs it only makes sense that the range for SB9 in terms of max x-range and R min values be greater than Drew’s List.

Now we are ready to apply a new threshold. The initial guess wasn’t too off from what this new one will be.
This new group of stars must meet the following:
1) max x-range cannot be below 68 (I took the mean of the averages for Drew’s list and SB9)
2) R min values must be smaller than 12

This should give us more potentials to take a look at. Sure enough, we receive more of the tail than we began with. We are now looking at 15,000.

In addition to looking at the histograms, I have also generated the CCFs for this plot. The first set of CCFs are taken from the most populated bins that appear to be located at (2.33,0.66). The majority of stars at that location yield max x-range values of: 216,217,218 and 219 with associated R min values of: 4.5,4.6,4.7 and 4.8.

Some of these are a tad unexpected… I was anticipating at least one of them to have well separated peaks but some of them do show signs of merged peaks. This is just a small sample of the most populated bins after all. The other CCFs may hold different information.
For this batch of CCFs, I am looking at stars that hold log(max x-range) values of 1.8-2.2. Also, I had the IDL routine report the corresponding visit that hold these max x-range and R min values so that the appropriate CCF could be reported.

log(max x-range) = 1.858
log(R min) = 0.4828

log(max x-range) = 1.93
log(R min) = 0.50

log(max x-range) = 1.945
log(R min) = 1.035

log(max x-range) = 2.17
log(R min) = 0.71

log(max x-range) = 2.269
log(R min) = 0.828

Thus far, it doesn’t seem like there are duo peaks residing in this batch of potential binaries CCFs. However, that alone is not enough to start eliminating some of these candidates. Additionally, these were only a selected few to look at, perhaps looking over more over the range of 1.8 – 2.5 may be useful (this is the x-range). The next step will be to look at the bisector slopes and determine the variation of those values from epoch to epoch. I plan to run this list through the python routine to look at the slopes of the bisectors.

Figure 1: This is all of DR13 mapped onto the 2D Histogram.

This image displays all of the stars in DR13 max log(max x-range) values and log(min R) values. It looks very similar to the previous post because there was only a couple thousand that needed to be added into this histogram. I just realized that Kevin left a comment on the next step to take in this direction. However, I made a step in a similar-isn direction before reading the comment.

I decided to see if I could isolate stars that have potential binary signatures by applying thresholds on the dataset as a whole. Meaning, I arbitrarily decided to assign values for max x-range and min R quantities that I have noticed binaries tend to reside in. For this particular case, I said that I want stars whose max x-range value is greater than 60.0 and whose R min value is less than 10.0. This is just a starting point to see just how many potentials I can identify. Sure enough, I ended up with 15,245 out of 163,669 stars. This is a little over 9% of the DR13 dataset that are claiming to be potential binaries.
Next, I wanted to see how these would appear in the 2D Histogram. Once plotted I obtained the following figure:

I am not sure that I trust this confidently but, it is a starting point.
The plan for the week is to apply the recommended step that was mentioned in regards to this dataset.

Unfortunately, the computers were shut off when I was at star 157,532/164,563 so this 2D histogram is not for the entire set of DR13. But, we got kind of close! A couple of features on the graph seem a little off and I will discuss them below.

Fig.1: Notice that the location of the majority of stars appears to be towards small max x-range values and large R min values, which is expected of non-binaries.
However, the tail appears to be level with the bulk of stars regardless of the variations in maximum x-range values.

So a couple of things appear to be happening here.
Not surprisingly, there are majority of the stars that appear to be demonstrating non-binary system behavior. In our case, this means that the stars fall within the region where they possess small max x-range values and large corresponding minimum R values. However, there does appear to be a tail of stars that fall within a behavior that may be binary (large max x-ranges and small R min values).
I think a concern of mine would have to be that the upper tail of the 2D histogram (where the majority of the trailing stars lie) rest at a near exact R min value as the non-binaries. However, they have large x-max values. Particularly the batch of stars that lie at ~ (2.4, 1.9) on the very far right towards the top.
Interestingly, there appears to be another tail coming from the main cluster that holds fewer stars but seem to fit the binary system behavior better than the upper tail. It appears that the minimum R value goes down to 0.45 and the max x-ranges go towards 2.5.

There also appears to be an upper limit of R minimum value and max x-ranges so far. Almost like a threshold the data can’t seem to cross where the threshold for R min ~ 2.0 and for max x-range is ~ 2.4. I am also curious what the two tails’ CCFs look like (I’m hoping to see either merged peaks or two peaks for the lower tail) so those are in the works of being generated.

Another key feature to notice that is rather odd is the vertical stripes that are rather apparent in the large cluster of stars. It almost looks periodic which makes me suspicious of it’s validity. What I mean by ‘it looks periodic’ is that the lines appear to be occurring roughly every 0.4 units up until 0.0. And the line intervals begin smaller towards -1.1. This may just be me being overly cautious but one can never be too careful. I will try to look over the code again before going on my vacation but this may just be a task I do when I return on the 5th of September.

When the IDL program encountered the first error only a small amount of the data was processed from the large list.
This is the 2D histogram of the first 60,456 stars.

It appears that the common trend of the values for a star thus far is to have small max x-ranges and large minimum R values. This trend describes non-binary behavior. However, since this is a mere sample size from the master DR13 list we will have to wait and see how the all the data looks.

It should be noted that the reason this 2D histogram holds 60,456 stars only is because another error was encountered just today. That error had to deal with the aploadStar.pro itself. Oddly enough, the star that the IDL program claimed was unable to load its file, due to an “undeclared variable”, when ran independent of the list as read perfectly. Further debugging is being performed but it’s rather odd that the list encountered such an error in the package known as apload.

Here is a glimpse of tables that contain max x-range values and minimum Rs for the DR13 sample.

Loading...

Taking too long?

Reload document

| Open in new tab

Download

The CCFs will need to be generated after the DR13 master list is finished processing. When attempting to run the programs at the same time the programs lag slightly and it adds to the total run time. My guess is that the CCFs for these will look like single peaks. Additionally, I will double check that the values being generated by the IDL code are correct in the sense that there are no “junk values” of -9999 in the tables generated.

Just a quick post in regards to finding the problem with the output of the first set of 2D Histograms generated.
Apparently the input for the grid of the 2D histograms in python must be modified in order to yield the correct output. In our case, it was a matter of flipping and rotating the dimensions of the grid.

Below are the updated histograms that do in fact agree with our intuition of where the binaries should be located.

For our binary sample that holds a total of 1249 candidates from Drew’s List and SB9 we have the following:

Figure 1: Follows the trend of having high max x-ranges and small R minimum values.

For our controlled sample that contained 713 stars from APOGEE-DR12:

Figure 2: Follows the intuition that non-binaries should have relatively high R values and small max x-ranges.

And lastly, when these two are combined into a summary 2D histogram we achieve the following:

Figure 3: The boundary between the two main clusters is a little fuzzier than anticipated but we can still distinguish which is the binary candidates and which are not.

The next step is to run this identification code through the DR13 dataset and see how many binaries we can classify. Additionally, it would be handy to have the code provide the CCFs from the visit that hold the minimum R value as well as the maximum x-range value of a given star so we have some visuals for the 2D histograms. Not always do these two values fall within the same visit of a star.

For DR13,the following will be generated:
1) Bisector points from the CCFs
2) Associated statistics for the bisectors (R, x-range, CCF width, CCF height, x median, x mean, etc)
3) Slopes of the bisectors and differences in their slope from visit to visit (for a given star)
4) 2D Histograms for the relations of max x-ranges vs R min values