I want to linger

A little longer, little

longer here with you

Aw jeez, where to start. I guess I’ll go with the prompts:

I think overall the event went well. My presentation to start went nicely, that one really had me nervous. Not that I felt more weight about that work or anything, it was just first and that always makes me nervous. But I thought that presentation went well. The rest of the talks were great as well. I thought that the 341 presentation about the class was nice. My folks were tuned in so it was cool for Kim to explain to them what the class was. It’s been tough to really convey what it is and why it’s special, so hearing it from Kim was nice for them I think. I thought our presentation accomplished what it wanted to in five minutes. The support from the faculty was great the entire time. The encouragement in the chat was really enjoyable, both in Zoom and our group chat in slack.

It was surreal when it wrapped up. It feels like the culmination of an entire year, after 337 and now 341. I thought it was really nicely put together by those who put the meeting together. It was lots of fun and overall successful in my opinion. I thought James Young had an interesting question. I find it hard though to answer questions from faculty. It is nice to know that we are being listen two and our work being thought about when we are presenting it, but when the question is asked I sometimes feel like a give a lackluster answer not deserving of the question which makes me feel a little bad. But other than that I think i understood what James was asking and we gave the best answer we could. After rehearsing with Cassidy so many times prior to the meeting I think I was on autopilot. I got what I wanted to get across and we didn’t go over time (I think), so that was a success.

What I learned in this course I think has provided me with information, and savvy, and know-how, and life experience, and so many other things that I would not have otherwise. Going to Arizona and KPNO speaks for itself, what an incredible once in a lifetime thing I could do and i think that trip has inspired me when it comes to the kinds of paths my career can take in this field. Being in that control room and doing that work really solidified that this is what I want to do with my life. The reading for the trip and our discussion about it gave me a little context about where we were going but also a little insight on the bureaucratical side of our field, which is super valuable in my opinion. (Also watching arrival at 2am on the top of a mountain looking at space was a visceral experience, and something I will cherish forever)

After coming back I think doing the work with our data was done really organically. I felt that if you were organized and prepared for the next step week in week out it made making progress that much easier and very natural. At least in my experience. I think working on low-mass stars was a nice escape from my galaxy work with Alex, and was nice to diversify my astronomy dictionary and toolbox in a lot of ways. Working with Cassidy was great. I remember our first year being in the same astro seminar, and ending our careers with each other on this project was a nice touch.

This class has been nothing but a positive experience for me. It has given me so much, from skills I will use down the road in my career to snacks in class at 9 at night (Thank you, Kim) I just wish I could have given something back.

Thank you, Kim and Sarah, for everything. Without the two of you this class would not have been what it was. What you have provided for me in this course, I will take with me onward. And I will try my very best to be a good disciple of what the both of you have taught me. I appreciate ya’ll, and I with that I give you my best regards. Farewell.

Signing off,

Owen H.

I’m so sorry

Something struck me in the rear

I just ended up here

— Sokka

This week has been decently productive, it is indeed coming together in a inexplicably short amount of time. On the analysis side i’ve been mostly trying to chug out my analysis plots, namely the R and I band rotation period vs mass plot. Which I have produced! But let me not get ahead of myself. It could not have been done without some wrestling of lightcurves. I inspected the periods of each of the targets that I had masses from for the TESS input table and tried to weed out those that were in the noise if there was a peak of considerable amplitude at longer periods, and folded the data over that period to do a “chi-by-eye” goodness of fit assessment on the new period. If I liked what I saw I edited the period I had recorded. I did this seperatley for each band, not looking at the results from the other band so I could refrain from becoming bias and trying to match them up intentionally. This was the result I produced:

The blue points are indeed on the graph. The periods are just so similar the data are on top of one another. I prevented bias is every way I could. I in no way had the periods of the targets in my head of either band while doing the other. This was the result of my determination of the periods independently. I had reported last week that the best periods were different from one another, but since then the result has improved after I went through and determined the new best periods. Next steps for analysis is making this plot for the brown dwarf targets, and then making other calculations for analysis plots such as percentage amplitude change, which I think i will simply determine to be the amplitude of the LS model. I think this is the least expensive in terms of time and should be simple enough to put together. I’d also like to make a R-I vs. time plot for my targets, as well as compare my data with Cassidy and hopefully combine our data and make more comprehensive plots of what I’ve just discussed above.

I decided to take a closer look at the abstract and intro of this paper: https://arxiv.org/pdf/2002.09135.pdf

I’ve kind of been focused on this paper the whole semester because the author has a similar scientific motivation behind their data and their purpose for it. The methods are similar as they also use differential photometry and periodograms to measure periods. Some of my notes on the background that I’d like to remember for the presentaion are as follows:

• Young stars are born very active and thus variable. We look at object on their surface such as sunspots which is also an indicator of the stars variability as they arise from magnetic activity around the star.
• Skumanich 1972 find that Ca II H & K line intensities decay as t^1/2
• Photospheric activity has been able to be observed with a higher fidelidty since then with the Kepler K2 and TESS space telescope missions
• They take data from these missions to measure clusters of different ages to see how properties like rotation period depend on stellar lifetime
• The smoothed amplitude is a quantity derived from difference of the max and minimum of the phase-folded lightcurve (after a gaussian kernel is applied). –>We will use a quantity like this for our analysis.
• Morris 202 finds that stars become less variable as they get older. As well as targets in older fields have a longer rotation period on average.
• Must consider that a single lightcurve can be produced by a myriad of models. The group puts it nicely in their own words so I’ll quote them: ” (1) that stars of similar age, mass, and rotation period should have similar spot distributions; and (2) the inclination angles of stars are nearly randomly distributed as observed from Earth. If these assumptions are true, then an ensemble of light curves of stars in a young cluster can be used to constrain their spot distributions.” (we cannot make these assumptions of the TESS data.

In terms of the presentaion what i’ve been doing is trying to boil down our text blocks into digestable bullet points, and making it look aesthetically pleasing. This is what I have so far for the first section:

Motivation: 

• Detecting variable low-mass stars (masses) and brown dwarfs (masses) in the Praesepe field in order to characterize relationships between stellar mass, age, variability, and rotation period. 
• (things) like these have been explored in the past by groups such as BLAH, BLAH for objects in the intermediate (solar) and high mass regimes. It is still unclear weather these relationships extend down to the lower mass regime
• We explore (name) R and I band in order to measure variable targets in fields of differing ages: the Praesepe star forming region in the Beehive Cluster, and objects encompassed in the TESS BD-21 scan conducted in January/February of this year

It is difficult trying to pin down what should be on the poster and what shouldn’t. It is valuable real estate that we can’t really afford to be generous with. I’d also like to move away from the more traditional headers, so that the poster can be more digestable as a whole. The feedback was very thorough so I think chugging through that is a good gameplan for now.

Scratch that plot from the beginging, I was doing something silly. Here is prot v. m for I band m-dwarfs:

Although not completely disheartening it is still not as gorgeous as the result before. I was plotting it such that the data was sorted and it would naturally plot that way, now i have adjusted it. I am making fixes R-band now but I think my computer has been working hard for too long because Jupyter notebooks is getting confused.

Until next time.

Shapeshifter shaking

with strange hands, some say I am

Fresher than Maine clams

A lot of progress was made this week in terms of data reduction (as far as I know). I made stacks per viewing cycle, and ran some wholesale photometry on my lightcurves from 1 night stacks. My goal right now is to plot the figures I have generated a la Noah’s suggestion in slack. It is not yet successful. In terms of data what needs to be done next is modify my aperture settings, loop thorough the periodograms for the all night stacks, and then plot a gallery of those at R and I band. Then rinse and repeat for the cycle stacks.

So far I’ve made a gallery of lightcurves. It was a little tricky to get Noah’s plotting routine going but now it is up and running. I’ve edited it a little to suit my data and provide a little more information, but it is mostly from his slack post. The link to the file is below.

All night:

Cycle:

I have also been doing some LS periodogram analysis and the first thing that strikes me is that the best period for a given target bewteen cycle and all night data are pretty different. For example, this is a periodic looking source in Tess R-band:

And then below it is the same target with more data points. The periods are quite different. My understanding is that when we add data points we are making out previous result with less data more accurate. Do we expect it on the order of 6 hours, is that comprable within error?

I can reconcile, however, why a given target in two different bands might appear different. There could be a astrophysical explanation explaining that discrepancy (that also can be difficult to diagnose, though right?), as oppose to the adding points issue where I find that to be more under a statistics concern.

Does it make sense to test the goodness of the fits like the ones above with a chi squared analysis? Would it make more sense to do that the more data there are? Or is the nature of the LS providing a model of a good chi sq value?

Accessing this data is my next step. I need to go through and really nitpick what I have taken. Compare with AIJ and Cassidy to assure the lightcurves are good, and then start looking at the LS periodograms. As well as continue to work on the poster draft for tomorrow. Some good suggestions that I received from Mikey last week in class was that to make sure that the plots are telling as much of the story, if not more, than the text on the poster. They shouldn’t be some random analysis plots, they should help carry on the narrative and connect back to the project goals presented in the introduction.

I’m glad I look at

the stars for school or else I’d

spend less time outside

This week has been fairly productive. I did a dry run of some photometry in python and got some results that match up with my findings from AIJ last week. My results are shown in the gallery below. sThus far I have only done lightcurves for I band-Brown dwarf data but it is looking promising. There is a gap in our data which is unfortunate because it definitely appears to be some ramping up in the flux that we miss in the gap. I feel the peridograms would be cleaner if there was more information about the period. Alas, weather is weather. That being said my one thing I did find strange was the detrend step in the python photometry. It is looking like my data is not detrended at all.

• 

• 

• 

• 

• 

• 

• 

• 

• 

• 

I think the best match was for the most central target in our frame (from the BD list). The AIJ curve is the bottom right of the first gallery, and the python curve is the first from the next gallery. That being said all of them look similar within the errors. The strange thing to me is the second curve in the python gallery. It looks like it went up almost a whole magnitude, while there is no other AIJ curve to match. I think this could be a mishap because there really isn’t any variation of that kind from the AIJ curves. The good thing though is that there is variation across the board. We are indeed looking at young, volatile objects and not something a little less interesting. In terms of how variable they actually I are I think the LS analysis will start to shed a light on the legitimacy of these periodicities. The errorbars on the AIJ curves are larger across the board so it’s hard to make any sort of conclusions/comparison about that data.

I have not yet had a chance to look at Python curves of R-Band data to compare with the AIJ curves that I reported last week, but I would guess that the variances between those AIJ I/R band data could be due to the sensitivity in those bands to procesees that require or produce energy at peaking in those wavebands. For instance the lower mass, cooler objects will appear brighter in I-band than R-Band. That being said these are both ‘red’ or long wavelength bands (relatively; to the other data being examined by my colleagues in other groups), so I think it’s more a matter of the magnitude and less of probing different activity if that makes sense.

It is my understanding that in the figure directly above the data should be continuous. I could be wrong but that is what I thought the detrending step did: get rid of the implicit sampling bias in the data. Since my first dry run at photometry in python I have complied a ‘super’ region file with all of the targets, not just low-mass or vice versa. I ran the targets through the program but now I am confused as to how to plot my data. .npz files are strange so I’m not so sure how to parse it but as far as I know I have taken data for all of my targets in the I-Band. Scratch that, I have now learned how to plot individual lightcurves but now I am having trouble reading them into the periodogram analysis routine. What we have done in class is reading in a csv file, but how do we read in a npz file. When trying to open it from finder, it is incomprehensible. So I assume there is some function to read it.

I’ve created a property table for the brown dwarf targets cross checked with the TESS input catalog v8.0: https://vizier.u-strasbg.fr/viz-bin/VizieR?-source=IV/38. It is posted below:

I’ve since discovered that you read in .npz files with the function np.load. However, I am still trying to wrap my head around the contents of the file, and furthermore my results from photometry. The map of my sources and comparison stars is below:

There are a total of 41 target stars, but the length of the data read in from the npz file is only 7. Why this is I do not know.

Starting to make more sense of it now. I’m thinking if I we’re to put it into an array it would be a 5 x 41 rectangle with the positions in ra/dec or x/y on the top and the data following is flux +/- error for each of the 5 days of observation (in my case this is TESS I band data so there is only 5 nights, for other bands/fields it would be 6 x 41) . I am now digging into the periodogram analysis for the 3rd star in the list. I am doing it for this target right now because it seems most periodic.

Here are some results, a little more promising than I had anticipated [I think (hope?)]:

• 

I was excited

To wear my birkenstocks with

Socks when it got warm

The week so far has not been too productive. After OOC meeting last week I went to revise my alignment code as I thought I was stacking, but in fact I was not. However my edits, as simple as they appeared, proved to be a little more involved than I had anticipated. My output_image function seems to be creating .fits files of entirely nans, which is frustrating to say the least. What I confuses me the most is why my procedure works for some data, but not the other. It mostly works for the TESS data. Which may be okay because it just occurred to me that is the data that I will be dealing with for the most part as Cassidy and I are dividing and conquering on our work.

After OOC meeting things started to roll. In a good way. Kim conjured up a solution to my nan image problem. After seems like inf or nan pixel values was blowing everything up. It could be a problem with the some bad pixels as I do not use the masking procedure. However, after implementing it in a night of reduction It helped, but the problem was not completely resolved. The shifting procedure only worked for some images and not others. Kim said it makes sense for it to work with mostly no problems for the I band sets as there isn’t as high a rate of pixel misbehaving. All it took was this fix:

def shift_image(image,xshift,yshift):
    '''
    shift_image
    -------------
    wrapper for scipy's implementation that shifts images according to values from cross_image    inputs
    ------------
    image           : (matrix of floats) image to be shifted
    xshift          : (float) x-shift in pixels
    yshift          : (float) y-shift in pixels    outputs
    ------------
    shifted image   : shifted, interpolated image. 
                      same shape as input image, with zeros filled where the image is rolled over    '''
    image2 = np.copy(image)
    image2[np.isnan(image) | np.isinf(image)] = np.nanmedian(image)
    return scipy.ndimage.interpolation.shift(image2,(xshift,yshift))

After meeting I hunkered down to align and stack everything for astroimageJ lightcurves using the solution above. My astroimageJ results are below.

• 

• 

• 

• 

• 

• 

The attempt went as well as it could. I think there is a lot of guess and checking going on, which isn’t the worst thing. It would feel nice, though to have a more concrete procedure for making curves. Hopefully the Python routine alleviates some of that. I notice that the BD sources are far fainter in the R-band compared to I. These curves are being done with 1-night stacks. Is there any other way to try and boost the signals of these objects. At what point do we say no more stacking because now there is not enough data.

The other documents requested are attached below:

This is an outline of a properties table:

Do we have the property information for the TESS field. I know we got what we needed for Praesepe from the in-class vizier exercise, but there is no analogous catalog for the TESS field. So how will we get this information for those targets? After running the Buedrolt low-mass targets through the Kraus there is information about the overlapping targets that could be interesting to report on such as the proper motion as well as IDs from 2MASS that could possibly provide other information.

Zoom zoom zoom zoom zoom

zoom zoom zoom zoom zoom zoom zoom

zoom zoom zoom zoom zoom

Spring break was a relaxing as it could of been. The sun was out mostly which is nice. Trying to distance. Things are a bit strange, but that’s alright. I’m excited to start focusing on some work. It’ll take my mind off of everything else for a bit. I’m trying to be on my Isaac Newton vibes and try to do some great work during quarantine. Really zen out and avatar state on some science. Anyways, I digress. Now to updates:

I didn’t do much over break but I did do a bit. I made a region file of the VLM targets in Praesepe outlined in the Boudreault et al. 2010 paper that Noah presented in the paper talk a couple weeks ago. The ADS link has the paper and the data products from the work: https://ui.adsabs.harvard.edu/abs/2010A%26A…510A..27B/abstract. From there I cross checked these targets with the the ones that we extracted even a further back in class in the Astrometry exercise. A screenshot of DS9 of the two files is shown below. The green is the Boudreault targets. None of them overlap with the other stars extracted from the Krauss catalog.

Besides that I watched the Fourier transform video. I thought it was SUPER informative. I feel like I’ve always known at a very general level what a FT does, but I was not very fluent at all in the math behind what it does. Rather, the intuition behind it all didn’t make sense. The ‘unmixing the paint’ analogy was nice, and his use of this 2-D plane where the circled up the signal using the ‘winding frequency’ in order to build up to the reason that we utilize imaginary numbers in FTs, and the elegance/usefulness of the complex plane. The integral for FTs make a lot more intuitive sense to me. That could also be a product of me being a lot more familiar with math concepts at this stage in my college carrer. Who knows. I was a little foggy about his comments about why we did not want that 1/(t1-t2) term in that expression that he presented. Also the big motivating idea of the ‘center of mass’ of the rings created by the wrapped signal puzzels me. I understand the purpose it serves in the transform, but I’m confused as to what it represents mathematically? Or say we were talking about sound waves, does that quantity represent anything physically? But I think another watch will clear it up. I’ve subscribed to the channel. This video is cool: https://www.youtube.com/watch?v=OkmNXy7er84.

The Templeton reading was also interesting. The formalism is nice and solidifies a lot of the ideas from the video. Puts the FT more in the context of statistics than interpreting it visually. Does the Nyquist frequency represent the ‘center of mass’ that I discussed above? The discussion of cautions with sampling was informative as I believe that we might run into similar problems. I think that our data could fall into the ‘gappy’ regime. I guess that would depend on the stack sizes, but I think our data will have to be stacked for each night as the objects we are observing are small and red. The discussion of autocorrelation functions for a more statistical approach to time series analysis was super interesting. Mostly because my last experiment for i-Lab last semester utilized one to try and measure light scattered off of a solution, and I did not really understand what is was at all. But now I do that that is cool. However learning about how it can be used in the context of variable stars was also informative. Figure 4 was a nice visual to represent that discussion. Overall a nicely written, digestible paper. I’m not 100% on everything, but maybe I can answer my own questions with some more reads.

Here are some recent papers that I think are relevant to my project:

• The TESS light curve of AI Phoenicis: https://arxiv.org/pdf/2003.09295.pdf
  • The group aims to measure accurate and radii for stars in an eclipsing binary system AI Phe. They compare models with TESS photometry and spectroscopy. I thought it was relevant because they are analyzing some lightcurves in the same way we are. Examining plots like % amp difference vs. time similar to the Morris et al paper.
• Spectral library of age-benchmark low-mass stars and brown dwarfs: https://arxiv.org/pdf/1912.02806.pdf
  • In this work the group uses data from the VLT X-Shooter spectrograph in order to more deeply understand ‘youth spectral characteristics.’ They take sample of brown dwarf members in different aged clusters: Chameleon I, Upper Scorpius Pleiades, and Praesepe. Even though they are using spectroscopic data and trying to measure different things, I think that the motivation and methods are similar to our project.
• The nature of the photometric variability of HgMN stars; A test of simulated light curves of $\phi$ Phe against the TESS data: https://arxiv.org/pdf/1912.07265.pdf
  • Again this is not very similar to the data that w have, or necessarily what we are observing. But their measurements of light curves and they ways in which they analyze them using FTs and %amp vs wavelength graphs is somewhat similar to what we will be doing. They aim to compare their models of mecury-manganese star variability with observations from TESS.

In terms of recent data reduction stuff: I thought I was all set but after the OOC meeting on Monday, Kim let us know that we were in fact not stacking any of our images, simply aligning them and calling them a stack. That could also point to why the light curve I made a couple weeks ago looked funny. But anyways I have to go back. I think I will just align all of the images and then combine them how I see fit afterward. It might be a little expensive on the memory side. I’m running out of space on my SD card. Is it alright to delete all of the old images. Like the ones background subtracted but not flatfeilded? Anyways, it should be a quick fix, and will give me a chance to polish a couple things up.

My next steps are to:

• Properly shift images and then combine them
  • Make new function to accomplish this
  • Make stacks PER NIGHT –> 6 data points
• Make some lightcurves now that I have astroimageJ installed.

Justin Timberlake

Where are the new tunes, I need

Some new JT please

So far this week I’ve reduced the rest of the data and attempted to make a light curve with my Praesepe I-Band stacks. I got astro-IMGJ to work on the UMass computers following the steps from class, but I am not picking proper references or targets. Right now I am just freestyling as I don’t really know where the brown dwarfs are. When trying to cross-check with my region file of the cluster created a couple weeks ago, i could not successfully open it in DS9. I think it has to do with the region ‘uploader’ is trying to upload the file in physical/image coordinates as oppose to FK5 which the file is in. Right now I can’t find a way to change the settings, I hope that my OOC meeting can help in this regard.

OOC meeting was very helpful. I have learned that my shifts are too big. I am aligning them to a different reference image as I am typing. Tonight I think I will make all night stacks and run one through astrometry to I can make a proper light curve tomorrow.

These are the results for last weeks class that Savio and I got:

• 

• 

• 

These results are really interesting. There is a super regular period shown in the lightcurve to the right. Considering this was kinda a shot in the dark I was surprised it worked so well on the first try. Hopefully I will be as fortuitous tomorrow making the lightcurve on my own data. As we can see the difference is magnitude is only about 0.025 magnitudes, but a very clear oscillation. AstroimageJ is a super nifty program that takes in a series of images and performs ensemble photometry for you. It takes into consideration all of the technical notes of the instrument that you took data on, as well as pixel value saturation/non-linearity values etc. in order to run differential photometry on a target and as many comparisons as you’d like to plot a lightcurve. Even though it is seemingly intimidating for our purposes it is pretty straight forward to use. Frankly it is impressive this was put together by one or two astronomers. My cap off to them.

Making a lightcurve went alright. My test one is shown below. There appears to be some variability, but nothing too regular. My first instinct is that the comparisons were not chosen too wisely. I was trying to pick sources outside of the cluster members to get an object that is relatively constant in magnitude across our image set. Alas, i have made a curve and seen what I saw. I know that this is not our method to perform photometry on our data so I’m not gonna trip about it. My image with the cluster members circled is also shown below.

Aligning so far is going well. There is still some issues with shifts greater than 100 pixels in an axis for some stacks. Should I pick another reference image. I’m realigning another night right now so we will see if it remains to be a significant drift. Some TESS images have huge shifts. We might need to consider changing reference images.

Cassidy and i have compared our reduction and it seems they resemble eachother well enough to reproduce data products. Which is all we need. Our shifting methods are the same, but our trimming differs a bit. I trim to a square 4096×4096 while Cassidy trims to a rectangular 4110 x 4095. This is something that we should keep in mind but i don’t think will produce too many roadblocks in themselves. We have a similar median count for our test images of about 150 a piece. Overall it appears we are on the same page and should proceed as we have for the last few weeks. The next step in our reduction is to make deep stacks. I have been tasked with the TESS field as it has less data and Cassidy’s machine has more computing power than mine so she will take the brunt of the force. Then hopefully we can do lightcurves in Python and start making some measurements.

“Listen good, I don’t

have nobody” – unfinished

Kid Cudi lyric

So far my work this week has consisted of organizing my pipeline. I have just aligned and combined the images for night 4. They are in a folder below. All in all they look good. There is one outlier that looks a little strange. The cycle 3 Praesepe I-band combination looks like it has a brighter background than the other combined images. Two images for comparison are shown in the comparison. The counts are overall higher in that frame. It doesn’t make sense for this to be similar to the TESS images taken on night two as these frames we’re taken around ‘2020-01-23T11:54:11’ as shown by the header. I’d like to further stream line my code and start using paths instead of cd’ing in my notebook. I am going to try and see Sarah in her office hour to try and figure out how or if this should be implemented in my pipeline. Besides that I’d like to somehow have the centriod coordinates somehow be recovered from the img_prime automatically instead of me going into the file in DS9 and eyeballing it. If I’m able to this I’d argue that the rest of the reduction will be less laborious. Other than that everything seems to be working properly, my stacks look good.

• 

• 

• 

One questions I have is: I have been stacking my images by cycle, but the stacks themselves are offset quite a bit (some more than others), but they are definitley not aligned. It is my instinct that we are going to want every stack to fall on the same pixel as this is how we will identify our cluster members with the region file from the target extraction in class. Should I also try to align these images to one another in order for the lightcurves to be made properly?

After going to Sarah’s office hours most of my questions are answered. The stacks DO all have to be lying on the same pixels in order for us to do differential photmetry properly. This is what I expected but was hoping it wouldn’t be the case (as this requires some editing of the pipeline. The fix, however, kills two birds with one stone. Having all of the images aligned to one makes it such that the centriod coordinates in the cross_image function all the same. No need to automate. No need to keep fishing for coordinates in DS9. Blamo. That being said after I implemented this method some of my stacks look strange. for images with large shifts I have black bars on the left and bottom sections of the image like so:

I suspect this is because the image is being moved by so many pixels that there is not data extended to these regions of the image. All of the sources are lying on the same points so it seems that we can do photometry as long as our sources have not been booted off the frame in the top right. That being said, stuff is working and science can be done soon. I am going to try and implement the path method, but also the returns may be diminishing: I only have a couple of nights left to reduce and my pipeline as is seems pretty efficient. So we will see.

Next steps include actually making lightcurves. We have the sources extracted for Praesepe and not TESS. In class we were told to leave TESS alone for now, but i suspect we will somehow need to retrieve the sources we have been looking at. For Praesepe we recovered about 70 sources +/- 1 depending on the sigma threshold and FWHM allowance. My instinct is that we will plop these regions down on our stacks and start trying to look at the flux (not calibrated) through them. Then we can start making lightcurves and hopefully coming to conclusions. I suspect the code will look similar but not the same as our photometry routine from 337, but to be honest I don’t know for sure.

The paper I’ve been digging into the angular momentum review paper written by Bouvier+ in 2013 and the Bailer-Jones+ in 1999. I think that the review will provide some good foundational info, and the Bailer-Jones will be good to provide some context, motivation and methods for our work. The reading’s haven’t been in too much detail since I’ve been tidying up my pipeline, but I hope to spend more time with these papers and get some good edits and revisions in before Friday.

Ran out of ideas

It is only the third blog

This does not bode well

After meeting with Kim on Monday and discussing with Cassidy I have got both my Bias subtraction and flatfielding issues resolved. Running through the check data set was successful, yielding counts for science images of the Orion nebula ranging from ~ -10 – 250 or so. There are some negative values in the background sky for those frames, but arithmetically it makes sense for them to be there. I hope it is easily applicable to our data. I anticipate there will have to be some additional sorting routines, as the check set only had one band of observation to deal with. Other than that by class next week I’d like to have all eight nights of data bias subtracted and flatfielded.

I’ve just finished sorting and calibrating the data from night 2. There were some hiccups, in particular sorting by the RASTRNG in order to get the data in the same field in separate folders. This involved a simple (and elegant if I don’t say so myself) edit to a line in the file-sorting function to make it a little more flexible. I also had some issues with trimming, initially I was just trimming until pixel column 4100. I think that left some of the overscan behind so now I trim images to 4095 x 4095. I’d like to rethink and reorganize the pipeline a little so the sorting happens a little more hierarchal as oppose to the more sporadic ‘cross that bridge when we come to it approach.’ I wonder if there is a way to make another edit to the file-sort function that will throw the files not needed in the trash. Alas, computers cannot do everything for us.

In terms of the quality of the data products through the pipeline I’d say everything is making sense. One of the only things I had to go back and double check was what was going on with the TESS field. The bias subtracted counts were much larger than those from Praesepe. If I recall correctly, however, the TESS data was the last data taken of the night and I believe the sun was coming up. So the sky would be a good deal brighter. Python is compiling a histogram of count values for me that I will include below. I am also confident that the subtraction is working because of the overscan count histogram shown with the aforementioned total count histogram. The total count historgram is taking very long. The mean of one of the images is 1256.459, higher than that of the Praesepe image which are about 200.

• 

• 

• 

The next step is to align and stack the images. How to do an elegant way of combining the image per image set seems to be a little challenging to me at the moment. The only way to do this I can think of is manually putting the data in directories based off of their groupings. I am pondering some code that sort them according to consecutive image numbers in the file name string. But the solution has not yet come to me.

After reading the shift_methods.py and the function documentations makes the shifting process make a little more sense to me. Each output of the functions feeds into the next one. The centroiding function allows you to look at a target, guess it’s center position and the function will return to you a more accurate one. The next function cross_image uses cross correlation to return to you the shift for each image, and then that result is fed into a function that actually shifts the images (and stacks them?). What is still somewhat a mystery to me is what image we should decide to shift towards. What is our ‘image prime.’ My first instinct is to shift to the image that looks the most resolved, but there may be more technical rhyme or reason to that choice.

As far as I know I’ve aligned my images properly. The counts make sense and they look good. The only discrepancy is the cycle 3 R band Praesepe images look oblong. Why this is I’m not sure. In general the images being stacked are fairly blurry. Otherwise the images look good. The gallery below shows the final product for Praesepe images. I have not yet stacked the TESS images for night two as there are only two images and they have irregularly high counts as described above. The science usefulness of them is not so certain to me. For next steps reduction wise I’d like to organize my script a little in so it runs top to bottom. And then I’ll let it loose on the rest of the data.

• 

• 

• 

• 

• 

• 

Next steps for the proposal will be a little more concrete after looking at the feedback, but after OOC meeting yesterday I have a better idea as to what the technical justification looks like and what it will contain. I think some inclusion of figures from other papers will be interesting, and discussion of how our data will change/update those figures. Probably will also discuss/make outlines of origional figures that we will produce.

Overall been a productive week.

Late night dinners are

Nesecesary when day

Hours go away

So far I have been writing a cleaner version of the bias sub/flatfielding code so that it runs with no issues when applied to the test data set. So far things are running but the counts I am getting were initially puzzling but after some inspection makes some sense. The raw frames on average have a lower pixel count value than the bias frames, so for the bias subtracted science frames I was getting a decent amount of negative values. After considering this however I did some analysis on the subtracted overscans and was getting some strange results from my tests. I have included a plot below. The central value is 120 not zero which is concerning. I will do some further investigation into this.

Looks like there could be something wrong with the scaling of the master bias, and a separate issue with flatfielding. I might have blindly used some old code so hopefully that is a quick fix. Otherwise I still have some digging to do.

The Nakajima et al 1995 paper has been interesting thus far. It’s cool that this work came out of the Palomar observatory at Caltech, which was discussed in our Giant Telescopes reading. An initial question I have is as to why the difficulty of detecting energy from gravitational contraction makes it hard to find brown dwarfs? What is the energy from this process a diagnostic of that helps us find BDs?

The team sets out to look for stars with a luminosity less than 10^-4 solar luminosities which correspond to an age of 10^9 years. BDs can be detected with analysis of proper motions in relation to a companion star. That is how the BD being discussed in this work, G1229b, was discovered. This arises the question for me: why is it that a star’s kinematics lead it to be classified as a ‘young disk star?’

• 

• 

After a meeting with Kim interpreting the right plot is a little easier. Simply put it is a spectrum plotted in frequency space. The higher the frequency the bluer we get. The solid curve is the BD being discussed in the paper and the other two are low mass M-dwarf field objects. As we can see there is a systematic shift in luminosity space as we move to the less massive objects. We can also see the spectrum fall off quite a bit as we move to higher energy wavebands. It is also below some threshold that the field stars are above. Further indicating it’s low mass not adequate enough for hydrogen burning. If we wanted we could put the panels of the left plot on top of their corresponding locations in frequency space as those panels are different images of the BD in different wavelengths. It seems like the group never takes real measurements of the mass or luminosity, but infer from models. As a supplementary presentation I will also discuss this paper that has cited Nakajima+1995: https://ui.adsabs.harvard.edu/abs/2009AJ….137….1F/abstract

In terms of plans for data reduction the coming week I hope to iron out the issues that was discussed in the OOC meeting and apply it broadly to the whole data set. Ideally it will run top to bottom in every data directory no problem, realistically this probably will not be the case. I expect my week in data reduction will be a lot of debugging and polishing the pipeline for more seamless use. Besides that I’ve only started to wrap my head around differential photometry and the procedure that it requires. Conceptually it makes sense and is maybe easier/more dependable than the calibration photometry that we did in 337? I guess that would be a product of the kind of information we’d like to tease out of our data. Regardless, I can foresee what the future looks like photometry wise. Things are sorting now though, which is definitely progress. I’m confused because the bias sub/flat routines work fine for the actual data but are behaving strange for this test sample. Hopefully the problems will be solved by tomorrow.