Meagan Morscher's REU homepage (Summer 2006)

Meagan Morscher
University of Wisconsin - Milwaukee

 

Astrophysics REU program-Summer 2006
University of Wisconsin - Madison

 

email me at meaganm2 at uwm.edu

 

Kitt Peak Sunset
Arizona sunset


About me...

I am completing my bachelors degree from the University of Wisconsin-Milwaukee. I am a physics major and I was interested in astrophysics; I found myself spending the summer at the University of Wisconsin in the astronomy department. Although I had not had much of a taste of astronomy before participating in this research program, I have grown to love it. The REU program turned out to be an experience that guided me in a wonderful new direction.

Quick overview of my research...


Hard at work in the WIYN control room at Kitt Peak

I was very lucky to work with Bob Mathieu this summer. Bob does spectroscopy, particularly stellar radial velocities, as part of the WOCS (WIYN Open Cluster Study) project. I worked on NGC 6819, an open cluster in the constellation Cygnus. Binaries, which are systems where two stars rotate about each other, make it difficult to determine radial velocities because their radial velocities are constantly changing. My job was to find orbits (or at least attempt to) for binaries in NGC 6819, because once an orbit is determined, we know the radial velocity of the center of mass of the binary system, which is the most important quantity. Along with my computer work, I also had the opportunity to go on an observing run at Kitt Peak National Observatory in Arizona; unfortunately, the bad weather during "monsoon season" made it very difficult to do much observing! The goal of the immediate project that I worked on was to get the results for NGC 6819's first radial velocity survey out to the astronomical community.

Did I mention that I also got to visit open cluster NGC 6819? Here's proof!

 

________________________________________________________

MY SUMMER RESEARCH PROJECT:

Orbital Solutions for Binaries in NGC6819

________________________________________________________

Introduction

WOCS

The goal of WOCS is to compile a database of fundamental open clusters, fundamental meaning that they span a range of cluster attributes, such as age and metallicity. To the right is a metallicity vs. age plot showing candidate WOCS open clusters. I have pointed out NGC 6819, which is the open cluster that I studied over the summer.

WOCS will combine photometric, astrometric, and spectroscopic data for the chosen fundamental clusters. I worked with Bob Mathieu doing spectroscopy. It is through spectroscopy that we determine stellar radial velocities (see Radial Velocities section). Spectroscopy can also be used to study the abundances of elements in a star's atmosphere.

 

 


WIYN Observatory, Kitt Peak, AZ

 

 

NGC 6819

To the right is a photo of NGC6819, an open cluster that has been regularly observed at WIYN and the cluster that I worked on for the summer. It is located in the plane of our galaxy in the constellation Cygnus. The cluster is at a distance of 2.4 Kpc from Earth, has an age of 2.4 Gyr, and has metallicity of - 0.05. NGC 6819 has a mean radial velocity (rv) of 2.3 km/s.

Since the stars in a cluster formed from the same cloud at the same time, all members of the cluster should have the approximately the same radial velocity. Therefore, we determine cluster membership by studying radial velocities.

 

 


WOCS candidate open clusters. Mathieu,
The WIYN Open Cluster Study.
NGC 6819 is 2.4 Gyr old and has Fe/H ~ - 0.05.

 

 

WIYN

WOCS is centered around the WIYN (WI, IN, Yale, NOAO) 3.5 m telescope at Kitt Peak National Observatory in Arizona. Observations from WIYN have provided a large database for stellar radial velocities. These radial velocities are obtained by taking spectra of stars and cross-correlating them against the solar spectrum (see section titled Template Star).

 

 

 

Open
Cluster NGC188
NGC 6819. Photo from jrhommes.com/Astro/NGC6819.htm

Spectroscopy

Spectroscopy is the study of the interaction between matter and light. We study absorption spectra, which show us the wavelengths where light is absorbed in a star's atmosphere. The atoms in the stellar atmosphere absorb light of particular wavelengths depending on the elements that are present. The absorbed wavelengths of light show up as dips in the spectrum that look like vertical lines; these are called absorption lines. The more light that is absorbed, at a certain wavelength, the deeper the absorption line.

A spectrum (see figure at right) is no more than a plot of intensity vs. wavelength. We see absorption lines because at certain wavelengths the intensity of light that we actually receive is lower.

Intensity

Wavelength
Solar spectrum. The icicle-like dips extending downward are the absorption lines.

Observing

WIYN

We use the WIYN 3.5 m reflecting telescope at Kitt Peak. Bob sets up 4-6 observing runs per year, each of which consists of 3-4 nights on the telescope. Weather permitting, we get multiple pointings for both faint and bright sets of stars (for a successful faint set, the sky must be very clear whereas it need not be as clear to do a bright pointing). Unfortunately, on my observing run we were only able to get a few integrations done on the first night; the rest of the nights were too cloudy and/or stormy to even open the dome of the telescope.


WIYN 3.5 m telescope

This photo that we took at Kitt Peak is an example of weather that is "not permitting."
We had to put up with weather like this nearly all night, every night!

 

Hydra

We use the Hydra Fiber Positioner (see figure at right), which can detect 80 objects over a 1-degree field (each fiber is positioned to detect light from one star or a multiple star system). Before we begin to observe, we compile a prioritized list of the stars that we hope to hit. We then setup the field use Hydra-Simulator (see figure below), which figures out where to position the fibers based on both our priority and also on Hydra's physical capabilities (for example, fibers must not overlap each other).


The incoming light is reflected into the Hydra fibers.
The lighter colored circular plate on the left section is where
the light-collecting fibers are located.

Once the simulation is complete, Hydra can begin to setup the field; this means actually placing the magnetized ends of the fibers onto a plate where the incoming starlight will be directed through reflection.


Hydra-simulator. This is what we see on the computer
screen when we are setting up an observing field


Hydra close-up. The ends of the fibers on the plate are where the incoming light from each object is collected.

The Bench Spectrograph, which is a multi-object spectrograph (MOS), uses diffraction to spread out the incoming light and separate it by wavelength. This is how we get an absorption spectrum for each object. We work in the visual band and use a central wavelength of 5130 Å (angstroms), which is the Magnesium B triplet.


Schematic layout of the Bench Spectrograph.
http://www.noao.edu/kpno/manuals/hydraman/node16.html

Reduction

In the reduction process, we make sense of the starlight that we collect when we point the telescope into outer space. We turn our raw data into meaningful spectra that can be used to extract information about our objects in the sky.

CCD

The spectra are read off of the charged coupled device (CCD), a tiny chip that operates by making use of the photoelectric effect. When photons of light hit the CCD, electrons are knocked off and collected; they are then unloaded and counted in order to figure out how much light came in at each pixel. In order to make sense of the raw data, we must first understand the CCD that are working with so that we can determine what each pixel is actually telling us.

CCD

A CCD chip. http://fuse.pha.jhu.edu/~wpb/
spectroscopy/figures/ccd.gif

dome flat screenDome flat screen

Before and after we do an integration on a field, we do dome-flats in order to take into account the inconsistencies in the CCD. In a dome-flat, we collect light from a uniformly illuminated screen (see photo at left). Although the CCD should be illuminated uniformly, some pixels will appear to have received more photons and others fewer because light does not affect the CCD identically at every pixel. We also take biases, which means that we unload the CCD and then see what is left. This remainder will be subtracted off when we reduce the data so that we only use what comes to us from our stars.

 

In order to identify the wavelengths represented in our absorption spectra, we take spectra of light from Thorium-Argon (Th-Ar) lamps. We know how the Th-Ar spectrum looks, and so we can later identify the different absorption lines appearing on the CCD. Through this process, each pixel on the CCD can be converted to its corresponding wavelength. We can then identify the wavelengths for any of our object spectra.

These operations are done at the telescope along with our observations. We make use of them later when we go through the reduction process. This is when the raw CCD data is converted to meaningful spectra. Only then can we begin finding radial velocities.

Radial Velocities

How do we use the spectra to determine radial velocities? The basic principle that we use is the doppler shift of light. If an object is moving radially (toward or away) relative to us, then the light we receive from it will be doppler shifted in some direction along the electromagnetic spectrum as the incoming wave is stretched (source moving away) or compressed (source moving toward us). This means that the incoming light will appear to be of different color than that which is emitted from the star. We use the terms redshift and blueshift to indicate spectral shifts in the direction of the red (lower energy, longer wavelength) or blue (higher energy, shorter wavelength) end of the spectrum. Correspondingly, every absorption line in the will appear at either a longer or shorter wavelength than it should based on the actual energy of the absorbed photons. As long as we know what we should see in the absorption spectrum, we can look at the actual observed spectrum and determine the radial velocity for the star at the time of the observation.

Template Star

How do we know what the absorption spectrum should look like? Spectra from other stars are very similar to that of our own sun. Therefore, to determine rv's for a star, we compare its spectrum to the sun's spectrum, since the sun is at rest relative to earth (at least in terms of radial motion). In other words, we use the solar spectrum as a template from which we can figure out how other stars are moving relative to our solar system. We call the process cross-correlation.

 

 

 

Cross-correlation function
Cross-correlation function and gaussian fit for an SB1. The y-axis represents the correlation and the x-axis is the pixel shift. This peak has a correlation height of about 0.825, where 1.0 would be a perfect match.

Object spectrum and solar spectrum
Top: object spectrum. Bottom: template (solar) spectrum. The
two spectra have similar absorption lines; the intensity of the solar spectrum is greater because the sun is so much closer to us.

 

 

The idea is to determine how we must shift the object spectrum in order to get it to best match up with the solar spectrum, line for line. What we actually look at is called a cross-correlation function (see figure at left) . Through the procedure just outlined, we expect to see a peak at the particular shift that best matches the two spectra; this peak represents the particular pixel shift that makes the correlation of the two spectra the greatest. Recall that in the reduction process, each pixel is converted to a wavelength, which is in turn converted to a radial velocity shift through the doppler shift equation. All of this is done in Iraf, a very useful data reduction and analysis program.

Binaries

In a binary system two stars orbit each other, and so the rv's of the stars are constantly changing as they move through their cycle; this makes it more difficult to extract the rv for a binary system. In order to determine the center of mass velocity of the system, we must first understand how the stars are moving about their center of mass, or their orbit. Once an orbit is found, the binary system's rv comes with it.

Follow the link to a great simulation of orbiting binary stars.

Identifying Binaries

Binaries are picked out by looking for variation in the observed radial velocities. We look for stars with a standard deviation of at least four times our internal error, or precision, which is 0.4 km/s. Although a binary system, by definition, involves two stars, we may not always be able to detect them both if the secondary is too faint. This is called a single-lined spectroscopic binary, or an SB1, because we receive light from only the brighter (primary) star (we can currently observe stars to magnitude V=17). Correspondingly, an SB2 is a double-lined spectroscopic binary where both the primary and the secondary are bright enough to be observed. In either case, we can use the rv's for the primary to identify candidate binaries.

STAR: 12006

    10976.94   -2.4962
    11352.93   -33.6237
    11447.71   -22.5795
    11448.66   -25.4613
    11449.65   -27.6287
    11716.96   -31.8105
    11736.90     9.3225
    12108.88    -9.2295
    12121.82   -12.3689 

Sample of radial velocities for object 12006 in NGC 6819:
julian date (left column), rv for the julian date (right).
The rv's range from +9.3225 km/s down to -33.6237 km/s.

Why should we care about binaries?

By studying binaries we get into stellar dynamics: since the stars are relatively close to one another, stellar collisions and mergers are more likely. Blue stragglers are stars that do not follow the normal path of stellar evolution, but instead live on an extension of the main sequence; they are bluer (hotter) than they should be, hence the name blue straggler. Theoretical work is being done to model different types of stellar collisions in order to determine if they can explain the formation of blue stragglers. In off-axis collisions, the angular momentum that the product star has should cause the star to fly apart; if blue stragglers are formed by collisions, then they must lose the excess angular momentum in some way (Sills et al, 2002). For more details on the subject, see the 2001 paper by Alison Sills, et al on high-resolution simulations of stellar collisions.

Through binaries, we can also study cluster evolution; one example is tidal circularization. In a binary system, although the total angular momentum is conserved, it can be converted to different forms. For a short period binary, tidal interactions between the stars can cause the transfer of some of the rotational angular momentum (each stars rotation about its own axis) into orbital angular momentum. This causes the orbit to become more circular (the eccentricity of the orbit decreases). For a given cluster, there seems to be a cutoff period, called the tidal circularization period, where all binaries with periods shorter than the cutoff tend to be circular. Further, as a cluster ages, binaries with longer periods will become circularized. In other words, the tidal circularization cutoff period is directly related to the cluster's age. For more information on tidal circlarization, check out the website of Bob's previous REU student, Sylvana Yelda.

My Work

We want to know the center of mass velocities for our stars, which includes binaries. The center of mass velocity of a binary, however, can be obtained only if we can nail down the orbit of the system. I used correlation-functions and previously written programs SB1, SB2, and SBCOR, in order to find new orbits and to improve and/or update old obits for nearly 100 candidate binaries in NGC 6819

Correlation Functions

Some stars have only a few observations while others have more than thirty. We used, however, only those observations that peak height h>0.4 and full-width-half-max (FWHM) <150, which represent the initial attempt to choose only "good" observations. For each observation we have a correlation function, from which we get one rv measurement (or two for an SB2).

For an SB1, a gaussian is automatically fit to the peak and the rv is determined. Sometimes we have to manually fit the gaussian for some observations if the auto-fit doesn't match the peak well enough (this sometimes happens when the peak is asymmetric).For SB2's, we must manually fit gaussians to the double-peaked correlation functions in order to obtain each star's rv, and so it is important that we can recognize the primary peak from the secondary in each observation. In some systems, a tertiary (third) star can be detected, but it is usually too faint to produce a strong peak of its own.

cross-correlation functions
Cross-correlation functions for two objects: one is an SB1, the other an SB2. In the SB2, the primary and secondary stars are clearly identifiable by the differences in their correlation peak heights. For brighter stars, we get a better ratio of signal to noise, and so the correlation to the solar spectrum tends to be higher.

 

When the stars are moving nearly perpendicularly relative to us (when their radial velocities are close to zero), the peaks are very blended making it more difficult to distinguish the two peaks and fit the gaussians (see figure at right). When the stars pass each other in our line of sight, they both have zero radial velocity. If the primary is in front of the secondary it will eclipse it and block its light from us.

It can be even more tricky when the two stars are of comparable magnitude, because it can make it nearly impossible to distinguish between the primary and secondary peaks in each observation.

blended correlation-function
Blended correlation function for an SB2. The primary and secondary peaks are of comparable heights, yet still identifiable. The secondary peak is blended into the primary peak because the stars are close to the passing point in their orbit.

Orbits

Once the gaussians are fit, we have rv measurements for the star at different julian dates (a particular date system that we use) as it went about its cycle. The next step is to work toward finding its orbit. We use a program called SB1 that searches through possible periods based on the stars rv's at each observation and picks out the period which fit the data the best. (This process is a little more complicated than just described: we have to manually decrease the period search interval in order to get closer and closer to the best period. There is always the chance that the best period might actually be excluded when the interval is decreased). If all goes well, we find a period with a low chi-squared value and we can then have SB1 produce the orbit based on the period that is found. We then have the period, the center of mass velocity of the system, the rv amplitude, and the eccentricity of the orbit. For SB2's the story is very much the same: we do a period search for the primary star (both stars in the binary have the same period), find the orbit of the primary, and feed the information to SB2, which spits out an SB2 orbit. Below are some examples of SB1 and SB2 orbits that I found.

 

circular orbit
SB1 orbit. The curve looks almost like a sine wave because
the orbit is nearly circular. The y-axis is the radial velocity
and the x-axis is the orbital phase.

SB1 orbit
More eccentric SB1 orbit

SB2 orbit
SB2 orbit. The point in the orbit when the two stars are the closest is called
periastron; the point where they are the farthest apart is called apastron

There were cases for SB1's and SB2's where the orbit simply could not be found. Also, for several binaries, although we found an orbit, a large uncertainty one or more of the parameters (e.g. eccentricity) meant that we didn't have the orbit nailed down yet. In these cases, we are not ready to publish the center of mass velocity for the binary, and the average the individual rv measurements for the star must be used instead.

My results

For some stars, the process of fitting the correlation finctions and finding the orbit is fairly quick, but for others this is not the case. Often it takes multiple tries and a bit of time (and perhaps even luck) before any orbit can be found. Sometimes it is necessary to go back and refit the blended correlation functions and perhaps switch the primary and secondary, if, for example, the peaks have similar heights. For 44 out of the 99 binaries that I worked on, we have at most an incomplete orbit. For some of these, it may simply be we are uncertain on one or two of the parameters, and a couple more observations is all we will need to complete the orbit. For others we don't yet have a logical, halfway-decent orbit, and it may take going back to the correlation functions and perhaps several more observations before we can make sense of the data. For the other 55 binaries, however, we have found good orbits. Many of these are systems that needed either another attempt at an orbit or an update to include new observations, while the rest of them I have found the first orbit myself. We are well on our way toward publishing the radial velocities for the binaries in NGC 6819.

What's Next?

The next step for WOCS is to work toward bringing together the photometric, astrometric, and spectroscopic data for the fundamental open clusters. By combining a proper motion survey with the radial velocity survey, we can understand the three-dimensional motion of our stars. The proper motion survey should come in the near future.

Useful Links

University of Wisconsin Astronomy Department

Kitt Peak National Observatory

WIYN Observatory

Alison Sills paper

Sylvana Yelda's website

Webda Site devoted to open clusters

SIMBAD Astronomical Database

Research projects of other REU students

 

Thanks to Bob Mathieu, Ella Braden, Aaron Geller, and Mike DiPompeo
for being such a wonderful research group!

 

Thanks also to Ed Mierkiewicz for being a great program director
and for putting up with our shenanigans!

References

Mathieu, R.D., The WIYN Open Cluster Study

Hole, T.K., Mathieu, R.D., Meibom, S., Latham, D.W., Platais, I., WIYN Open Cluster Study. XXIII. Stellar Radial-Velocity Measurements in NGC 6819