Team West's Research Blog

Milky Way's Red, White, and Brown Dwarfs

Proper Motions (Part 1 of many)

I can’t believe it’s already been (about) half a year since the last post on the blog. Now that Summer is here, it’s time to get some research done! My primary goal of the Summer will be to create a comprehensive proper motion catalog for photometric low-mass stars using the SDSS DR10, WISE AllWISE, and 2MASS point source catalogs. This will be useful for people wanting to differentiate low-mass stars from other sources that have similar colors (QSOs, giants, etc.).

The first question we might ask ourselves if how good will the catalog be (what are the expected errors). The easiest way to estimate this is to use a sample of objects that don’t move on the sky, and see how much each catalog pair says they moved. Luckily for us, SDSS has compiled a large catalog of spectroscopically confirmed QSOs (166,583 QSOs, here). Using this catalog, I queried each catalog for positions, positional errors, and a bunch of quality flags that will help me pick objects with good detections in each survey. In a perfect world, we should hope that the proper motions for all of these quasars are normally distributed about a proper motion of zero, with a little bit of scatter due to the different resolutions between each survey. What we find is the following figure (normalized distributions between each catalog pair for each proper motion component):


I would like to say up front that this is a first pass, and these results will most likely change after I do some filtering and figure out potential contaminants. With that said, let’s talk about the figure. All three distributions appear to be roughly centered about zero, with small offsets. We can see that in general, the WISE/2MASS proper motions do the best, most likely due to the quality and depth of both of these surveys, as well as the long time baselines. The large RMS errors are due largely to the huge WISE PSF (~6 arcsec FWHM).

One of the interesting things I found was that the largest source of contamination was due to nearby, you guessed it, red dwarfs! 2MASS and WISE were more likely to pick up the red dwarf, which is much more luminous in the infrared, than the QSO. I don’t foresee this being an issue in the final catalog since we only care about the red dwarfs. Since we are using the QSOs to estimate the intrinsic error between each catalog pair, we should just pick the best point sources that don’t suffer from contamination by nearby red sources (maybe a useful high school project?). I expect this QSO analysis to be part of my final publication, so we’ll definitely be seeing more of this plot in the future once I decide which QSOs to use.


Screen sharing

In line with Jan Marie’s post on backgrounding, I thought I’d post a quick note on screen sharing with any of the beer-flavored Macs in the office via BU’s 802.1x network (or VPN if off campus).

From your own Mac’s Finder, head to the Go tab, then Connect to Server (or simply ⌘K from Finder). For my Sam Adams desktop, the address is ‘vnc://,’ then I connect As a registered user with the usual Kerberos credentials and voilà! Note: you might need to manually enable screen sharing in /System Preferences/Sharing/ for both computers.

I should also mention this setup’s usually a complete bust from Pavement due to their reliably unreliable internet, but if you’re anywhere else with a solid connection and want to avoid maneuvering too many ssh shells (sorry, RAS syndrome), it’s awesome.

Alright, a really quick note.


Although I am intimately familiar with ssh-ing, I realized yesterday that I still don’t know how to ssh into a remote computer and start running something in such a way that it will keep on running even after I turn my laptop off.  I asked a fellow student who didn’t know how off the top of his head either, and so I decided to Google it.  As I was making my way to Google I realized that I have quite an extensive network of astronomers on Facebook, and maybe I should throw that question out to them.  I posted a status update, and then had to run to a lunch talk.  When I returned an hour later there were loads of responses.  (I really need to remember that trick when I have other questions about science, IDL, etc.)

I distilled all the comments down to two easy methods: nohup, and screen.

Nohup is short for “no hangup”, or no HUP, and it basically will tell something to run, and ignore the HUP signal, which is how terminal warns dependent processes of logout.  I didn’t try using nohup, but one of my favorite astronomers, David Anderson, provided the following instructions for its use:

in tcsh:
( nohup <stuff> ) >& Log &
bash is more like:
( nohup <stuff> ) 2>&1 Log &

Because someone else told me “nohup is so last year” and now the cool kids are using screen, I went straight to that one (peer pressure, what can I say).  Screen is a little harder to google, but I managed to figure out how it works.  When you run the screen command it creates a window with a shell in it that you can then use to do whatever you want.  You can give the window a name, and then hide it, and the programs inside will continue to run (detached).  You can create as many windows as you want and they will run independently of each other.  You can also call up a list of window, kill windows, retrieve them, etc.  Screen is probably a lot more powerful than the simple ssh sessions I have been using it for, but it’s a good place to start.  Here’s a quick step-by-step:

Create a screen:

$ screen -S [name]

Then do something in this window.  For example, ssh into a remote computer and run something.

$ Ctrl+a+d will “hide” the window and save the task. It will say [detached]

$ screen -ls shows what screens are currently
There is a screen on:
10000.task (Detached) running

$ screen -r 10000 recover the screen

$ exit to exit screen
[screen is terminating]

So far, screen seems to be working well for me.  I will post an update with any new tips, tricks, or warnings that I learn after I have used it for a while.

Ford Conference

I’m going to do something a little different and talk about my experience at the Ford Conference. It was incredible to meet so many people across so many disciplines and it really put into perspective my responsibilities as a graduate student, a scientist, a scholar, and a teacher. I realized how important it is to build an academic network, not only within Astronomy, but across multiple disciplines. I feel reinvigorated and I’m ready to come back to Boston and hit the ground running! Bring on AAS!

Here’s the science part. I met two other astronomer’s within the Ford Family, Jedidah Isler (currently at Yale), and Charles “Chat” Hull (currently at UC Berkeley). Chat does polarization studies of protostars, and Jedidah studies the multi-wavelength variability in blazars (AGN). What do these things have in common? Let’s look at the following pictures:


Okay, so you don’t need to be a rocket scientist to notice there are a lot of similarities (in these watered down images). Fundamentally, these two classes of objects are very similar in terms of the phenomena we attribute to them (e.g. accretion disks, jets, outflows, etc.). Recently, Lynne Hillenbrand came and gave a lunch talk to talk about the multi-wavelength variability seen in protostars, and she talked with the blazar group here about the similarities seen in variability in AGN. It is clear that there are a lot of similarities, but I don’t think anyone has really tried to look at the data and find trends between the two classes of objects. I brought this up with the other two astronomers and they seemed intrigued. With the hordes of data available on both types of objects, it might be worth looking into further.

Thar be infrared excesses

Hey everyone, been a while since I’ve posted and I apologize profusely for that. Let me tell you what I’ve been up to. Now that my time using the spectroscopic M dwarf sample is coming to a close, I’ve started working with the photometric sample from DR10. Pulling all the M dwarfs using the criteria from Bochanski et al. (2010) returns something like 80 million M dwarfs. DR10 also has this new feature where you can join the WISE all sky catalog with DR10 and pull objects that have been cross-matched between the two. Doing this and making some signal-to-noise cuts reduces the sample to about 1.5 million M dwarfs. I also removed objects that had a contamination/confusion flag and this gave me ~1 million M dwarfs. Let’s look at the sample I currently have.

ColorPhotSamp2 ColorPhotSamp1

Sorry for axes on the histograms, but the only thing that is important is the morphology for now. The red dashed line represents the linear relation I developed in my (draft) paper to separate stars with excesses from stars with normal photospheric levels. In r-z color, we get a peak at about 1 with another bunch at about 2. This is consistent with results from the ‘field’ stars in Bochanski et al.’s paper. But if we look at the WISE colors, we notice something that should not be. There are WAAAY more stars with infrared excesses than there are stars without. This is probably due to 2 things. 1) There is a bias for things that have higher than normal infrared flux densities since I made a cut on signal-to-noise and most the the things in SDSS are at large distances. 2) There is probably still some contamination due to nearby objects. One of the things I did for the spectroscopic sample was filter out things with close neighbors (stars that had another photometric object within 6″). This is the step I am currently running, but it takes some time for the script to run since there are a lot of checks and balances. I’ll see how things come out after this filter and update you on the results.

Fast moving M dwarves

Hi everyone!

I am Andrej, an astronomy graduate student at the University of Maine. I have been working with Andrew on fast moving M dwarves in the Milky Way galaxy. The goal is to build a data set of the fastest moving of them, and see if any will leave the Milky Way. I have been using Andrew’s SDSS DR7 M Dwarf Sample.

Since our campus (deep in the woods here in Maine) doesn’t have a license to IDL, I have had to learn how to use R, to read in his data set, plot spectra from SDSS, and calculate a bunch of correlations (particularly the CCF) to study the fast movers.

A subset of Andrew’s sample consists of some rather fast movers (some 1280 M dwarves whose speeds relative to the galactic center are 500 km/s or faster). I have looked at the sodium peaks of all 1280 of the spectra. I compared the line-of-sight velocities of these stars (which translates to where their sodium peaks should appear) with where they actually appear. As a preliminary inspection by eye, I counted 517 fast movers whose radial velocities were over 100 km/s and whose velocities did not seem to agree with the spectra.

To address the radial velocity errors, Andrew has asked me to spline the Bochanski et al. (2007) M dwarf templates to the spectral resolution of the SDSS and put them all on the same log scale. I did this with R, then I selected intervals (equally spaced along the log wavelength metric) within about 6600-8000 angstroms, which is where the signal is greatest. I then calculated a series of re-splined CCFs along these intervals. The expectation is that the peak of each of the CCFs should all be in about the same location. I know that R is doing this correctly, because Andrew (using IDL) and I compared the CCF results for a K7 and we get the same clean peak.

I found, however, that of most of the 517 fast moving M dwarves, most tend to have significant discrepancies (that is, discrepancies within the intervals between 6600-8000) in the locations of the peaks where the CCF is a maximum. More specifically, the standard deviation in the location of the peaks is at least 0.5 spectral channel units, for 374 of the 517 fast movers. Here is an example to show what is happening. I’ve also considered smaller intervals, and choosing a bluer end of the spectra, but I still get some pretty big offsets. Perhaps the spectra is just noisy. What are your thoughts on this?



To keep the topic moving, I’m going to talk about binary pairs again. One of the things I did for my paper was to create combined models of an M dwarf and brown dwarf (close) binary. Close enough that the pair cannot be resolved with SDSS. This required using model photospheres (as I’ve talked about in numerous other posts), and scaling both the M dwarf and brown dwarf using luminosity relations. I used the luminosity relations for M dwarfs from New Light on Dark Stars, and the luminosity relations for brown dwarfs from Burrows et al. (1997). Then I added the models together and performed synthetic photometry on the combined model to produce flux ratios that are measurable. The result is posted here:


Okay, so what exactly is going on here. Well, each line represents a binary case. The temperature of the M dwarf is on the x-axis and each color represents a different case with an ultracool (<2001 K) companion. I have also plotted the case where there is no companion (black line). I have also plotted two different surface gravities for the M dwarf. The dots and errorbars represent my sample and the magenta point at 2400 K is a known low-mass binary. Things seem pretty reasonable, most of my sample fall on or below the lowest ratio cases, except for the J-band ratios. You shouldn’t expect to see anything above the line with no companion. What might be going on here? Possible explanations: 1) the photometry is bad. This is hopefully not the case for these many stars but I put it out there. 2) Variability. The photometry between SDSS, 2MASS, and WISE was taken over different epochs, and it was been shown that young, active low-mass stars have significant variations in their magnitudes. They found J-band variability on the order of 1 magnitude from high to low, with lower variability (~50% less) at I- and K- bands. Considering that, it is probably best to compare WISE fluxes, so let’s look at the bottom plot. At the 3-sigma level, the known binary falls into on the model tracks (this is true for each band). If we look at 3-sigma levels for the entire sample, we find that only 21 could be potential binaries. Of course, this is highly dependent on the stellar parameters (effective temperature and gravity). If these stars move to the left or right they could be more or less likely to host a companion. One last possibility. 3) Planet. If there is a Jupiter sized planet, this could also create a flux decrement on the order of 0.5-1 magnitude. This is the least likely considering the circumstances but I thought I would list it.

Back to Binarity

I wrote a previous post on examining binarity here, but now I am going to revisit it by a different approach. As I was looking through my candidates on ADS to see who had published what on them, I found this paper (both Andrew and Saurav are on it), saying that one of my candidates is a spectroscopic binary with a brown dwarf (T5) companion. I decided to see if the infrared excess I’m seeing could be attributed to the brown dwarf companion. First, let’s talk about the non-binary case and just look at the Spectral Energy Distribution (SED).

SED of an M9 with SDSS spectra and SDSS, 2MASS, and WISE photometry. The model photospheres are smoothed BT-Settl models.

SED of an M9 with SDSS spectra and SDSS, 2MASS, and WISE photometry. The model photospheres are smoothed BT-Settl models.

I made a mistake with the y-axis. Technically a Jansky is a unit of ‘flux density’ so we will overlook that for now. I could have put arbitrary units since it doesn’t matter so much for this argument. So what are we looking at here? Well, if a star was a perfect blackbody we would be looking at a blackbody curve, but alas the stars are not and so we are seeing all the intricacies (smoothed over) of the stellar spectrum. For your ease, I have plotted and labeled (on the top) all the wavebands for the photometry I am using. The WISE bands are listed as microns since there is no convention for their bandpasses. The large line going through each data point is actually the width of the bandpass. You can see that SDSS and 2MASS have much smaller bandpasses than WISE. This will be important to consider. This model actually fits the data extremely well, except for the g-band and blue end of the spectrum. These are both due to low signal-to-noise for a late type (M9) dwarf. I am claiming that the 12 micron WISE band is potentially probing a real effect, that there is more flux at 12 microns than can be attributed to the stellar photosphere alone. Although this is not a large increase above the photosphere, especially considering the uncertainties, it is a possibility. Let’s look a model with brown dwarf added into the mix.

Combined SED (blue) of a M9 dwarf (magenta) and a brown dwarf (red) with SDSS spectra and SDSS, 2MASS, and WISE photometry. The model photospheres are smoothed BT-Settl models.

Combined SED (blue) of a M9 dwarf (magenta) and a brown dwarf (red) with SDSS spectra and SDSS, 2MASS, and WISE photometry. The model photospheres are smoothed BT-Settl models.

As you can see above, this model seems to fit pretty well to the data. In fact, this was the model that fit best to the 12 micron data point that was within reasonable limits (brown dwarf radius ~ Jupiter radius). There are some interesting things to notice here. First, the brown dwarf does not really affect the visible wavelengths due to it’s low flux at those wavelengths, however, it does make a big bump at the J-band. From the first SED I showed with just the stellar photosphere, we are apparently not seeing a bump in the J-band. There is also a bump between 3.4 and 4.6 microns that we could in theory probe if there was a band there. This might be something interesting to consider for future missions if we want to find these low-mass binaries.

Back to the main question, is it a brown dwarf companion that we are seeing? The answer? Maybe. The best fit model by far is just the star without a companion. Mostly because we are not seeing the bump in the J-band, which has a small uncertainty. However, this (and a couple other) companion models all fall within 1 sigma so they are also possibilities. It would be helpful if we could probe that bump between 3.4 and 4.6 microns to rule with any certainty on what we are looking at. Either way, there are questions raised.

If this is a true excess that we are seeing at 12 microns, then that means there is dust in this close companion binary. In the paper I referenced above, they do see the bump in the NIR spectrum. The discrepancy could be due to bad photometry or spectrum. It is also possible that 2MASS caught this binary system when the brown dwarf was eclipsed (estimated inclination angle of ~60 degrees).  Since the WISE photometry is made up of stacked images, it is very possible that the brown dwarf was visible during many of the exposures, which could account for the excess we are seeing.

Jerkish gcirc

For the past few days I’ve worked on matching proper motions for stars in the Kepler Input Catalog (KIC) and the USNO CCD Astrograph Catalog (UCAC4). The whole 100 million+ star thing’s destroyed the computer and though I left it over the weekend, my preliminary run with (an IDL program to determine arc distances between objects) has yet to finish. A quick ‘top’ check in the Terminal reports IDL using 100.1% of the CPU. Hmph. Less than stellar, killing it now. I should have this resolved by the end of today and an update with potential pairs by the end of the week. Also, I just learned of Kepler’s call for white papers (thanks, Saurav) and hope to submit our identified binaries for rotation periods.

Flux Calibration Issues: H-Alpha imposters

So, I’ve been hastily reducing my data and I’ve noticed some unwelcome consequences of the way I’ve been doing my flux calibrations.  I’ll try to explain this visually.  Below is my non-flux calibrated standard.  Notice the features at ~6550, ~6900, the large telluric feature at ~7600, and the extensive fringing at redder wavelengths (the waviness, those aren’t real features).


Below is what the response function looks like when we take the above spectrum and divide it by the model spectrum.  All the other spectra are then multiplied by this response function to flux calibrate them.


Notice any potential problems with this response function as I have it?  Let me show you.  Here is a fairly high signal to noise spectrum of an M dwarf pre-flux calibration.


Looks pretty good right?  It has semblance of being an M dwarf, you can see some TiO and VO molecular absorption bands.  There doesn’t seem to be any H-alpha emission though at ~6560 Angstroms.  Now let’s look at the flux calibrated M dwarf spectrum using the above response function.


Hey!  Looks pretty decent!  The light green is the observed flux-calibrated spectrum and the dark green is the best fit template.  The fringing at the red end wasn’t fully corrected for but you can see a definite improvement over the non-flux calibrated spectrum.  So we definitely want the response function to have that “jaggedness” at the rend end to help correct for this fringing.  Looks like there is some telluric absorption between ~7600 and ~7700 Å that wasn’t quite corrected for…guess that’s okay.  Oh and it appears to be active!  Cool.  Except…we didn’t see any H-alpha emission in the non-flux calibrated spectrum.  It looks like the response function is adding in a “fake” H-alpha signature into the corrected spectrum.  Considering the nature of my study, we want to avoid messing with the H-alpha feature at all cost.

To help account for this , I’ve been taking my response function from earlier and smoothing the blue end.  I don’t want to completely smooth it over, because I want to somewhat correct for the horrible fringing on the red end.  Here’s what the smoothed response function looks like.


I’m hoping this improves the flux calibrations.  It’s still a work in progress.  If you have any suggestions that would be great!