1. Developing a New Method to Detect Galactic Winds to Understand Galaxy
Evolution
Executive Summary
In order to understand the detailed evolution and history of galaxies, it is first essential to
acknowledge the role that galactic winds plays. Galactic winds, in general, are defined as the
movement of large amounts of gas particles into and out of galaxies. These are primarily caused
by light pressure from stars or by the force of supernovas. For most galaxies after the Big Bang,
the more distant ones tend to have a higher star-forming rate, leading to brighter stars and more
resulting supernovas in these galaxies. Thus, surveys of very distant galaxies, in this case DEEP2
and DEEP3, were examined as they are likely show the presence of winds their resulting
impacts.
Even for distant galaxies though, it is difficult for telescopes to detect these winds due to
their faint light emission when compared to the surrounding light of the entire galaxy. This report
describes a continuation project on detecting faint wind emissions on the outskirts of very distant
galaxies. In this project we developed a novel method of detecting wind emissions in various
galaxies through the implementation of a graphical user interface for inspection of galactic
spectra. When looking through the DEEP3 galaxy survey of approximately 1800 spectra at the
Mg II doublet at wavelengths 2796 Å and 2803 Å and the Fe II doublet at wavelengths 2612 Å
and 2626 Å, we found that there were no large wind emission visible across the 20 kiloparsecs
distance from the center of the galaxy for both the Mg II and the Fe II doublets. Despite these
findings, the procedure developed is significant as there has never before been an accurate and
rigorous method to detect the presence of wind emission across multiple galaxies. Previous
detections have only examined one galaxy or galactic object as opposed to hundreds of galaxies
as is described in this report. Thus, this newly developed procedure can serve as the foundation
for future research in wind emission detection across several galaxies.
2. Developing A Novel Method of Detecting Galactic
Wind to Understand the Evolution of Galaxies
Abstract
A detailed comprehension of galactic wind is critical in the understanding of galactic
evolution, but the field of galactic winds has remained relatively unexplored. This is be-
cause long amounts of exposure time are necessary to clearly detect galactic winds, which
is impractical given the tens or hundreds of thousands of galaxies that still need to be ob-
served. Consequently, we present a novel graphical user interface (GUI) that enables the
detection of wind emission without the use of long-exposure spectra by stacking multi-
ple spectra to improve the signal, with the GUI also accounting for skylines, bad pixels,
serendipitous objects, and various other errors in the process. With this GUI, galaxies from
the DEEP2 and DEEP3 surveys were examined to detect wind emission for the Oxygen
II, Magnesium II, and Iron II doublets. There were no signs of extended wind emission
found in either survey, but this has large implications. First, our work built and expanded
on the paper conducted by Huang and Shekhar, previous Siemens regional finalists in 2014,
and with our more expansive analysis we established that their first observational detection
of wind emission over a large data set was false. This demonstrates the necessity for our
procedure to be implemented. Second, we present spatial constraints of wind emissions
with the largest data set ever used that wind emission is not present within 15 kiloparsecs
of galaxy’s centers for galaxies z ∼ 1.
3. 1. Introduction
1.1 Galactic Wind
Galactic wind, or extended wind emission, describes the movement of gaseous particles
through intergalactic space and its surrounding areas after a high-energy event releases ki-
netic energy into the gas. This gaseous flow is usually caused either by the radiation pressure
from newly formed stars or by the force of supernova explosions, both of which infuse ionized
gas particles with kinetic energy to push them out of the galaxy. (Johnson and Axford 1971,
Matthews and Baker 1971, Vellieux et. al 2005, Westmoquette et al. 2007) . A sample image
of galactic wind is shown in Figure 1.
Figure 1. Sample Galactic Wind An image of galactic wind in the Messier 082 galaxy taken
by the WIYN telescope in H alpha and HST in BVI continuum colors. Taken from Veilleiux et
al. 2005
Understanding this phenomenon known as galactic wind is essential for current astrophysics
research in galaxies, because a proper theoretical understanding of galaxy evolution and for-
mation requires a comprehensive understanding of the physical processes inside the galaxies
as well (Veilleux et. al 2005, Zahid et. al 2013). For example, current research suggests that
1
4. galactic winds act as a mechanism that decreases the star formation rates in galaxies (Voort et. al
2011). Consequently, galactic winds would then be able to account for how the observed baryon
fraction of low-mass halos is lower than the cosmic value (Papastergis et al. 2012). Galactic
winds are also thought to be a mechanism for enriching the intergalactic medium (AGUIrre et
al. 2001, Creasy et al. 2014, Springel & Henrquist 2003) and the circumgalactic medium (Barai
et al. 2012), which would explain the existing chemical makeup of galaxies today. More specif-
ically, galactic winds would be able to account for the presence of metals in the Lyman alpha
forest (Springel & Henrquist 2003). As research into the proeprties of galactic winds increases,
more information can be determined about the relationships between galactic winds and the
evolution of galaxies. However, confirming these existing hypotheses and developing new re-
lationships between wind emission and galaxy evolution require more long-exposure spectra
of galaxies like that in Figure 1, which is difficult given the thousands or tens of thousands of
galaxies that would need to be observed. Thus, we conducted this report to determine a novel
method of detecting extended wind emission in distant galaxies without requiring such long
galaxy exposures.
1.2 Wind Emission Detection
In the past, researchers have usually attempted to examine the effects of extended wind
emissions through simulations (Voort et al. 2011) or by analyzing individual long-exposure
spectra of galaxies (Martin et al. 2013, Rubin et al. 2011). Some scientists have instead looked
for wind emission by stacking multiple galaxies together to improve the signal (Weiner et al.
2009, Tang et al. 2014). However, when stacking these spectra there hasnt been a method yet
of determining the spatial extent or mass loss of the wind. For example, the Weiner et al. 2009
study was only able to determine that the wind was present for Mg II through the Deep2 galaxy
set. The Martin et al. 2013 and Rubin et al. 2011 studies were alternatively able to successfully
2
5. determine the spatial extent of the wind emission , but they had to use very long-exposure
spectra to attain this result, which isnt efficient when looking through multiple galaxies.
The process of stacking spectra with medium to low exposure times also largely increases
the likelihood of error, because several erroneous spectra when stacked could either disGUIse or
mimic the presence of wind. These sort of errors arent accounted for in studies stacking multiple
spectra like Tang et. al. To solve this problem, we developed a graphical user interface(GUI)
to establish a novel method of easily detecting wind emission across short-exposure galaxies
while simultaneously removing the chance of any possible errors.
2. Methods and Procedure
2.0 DEEP2 and DEEP3
The newly released DEEP2 data and the DEEP3 survey together contain large amounts
of data in the form of spectra from redshift 0.7 to 1.45,which were collected using the DEep
Imaging Multi-Object Spectrograph (DEIMOS) of the Keck II telescope in Mauna Kea, Hawaii
(Newman et al 2013). However, most of our data collected through DEIMOS must still go
through multiple procedures and methods before actually being visually represented on the
screen and in the GUI. We used the DEEP2 surveys new data release to detect only Mg II
extended wind emission because the spectra didnt cover the wavelengths that Fe II is expected
to be at, while we used the entire DEEP3 data set to look for both Mg II and Fe II extended
wind emission.
3
6. 2.0.1 Red and Blue Sides of the Spectra
For DEEP2 and DEEP3, the data was encoded by DEIMOS separately as red and blue sides
of the full spectra. In order to get this complete image, we had to stitch the red and blue sides
of the spectra together and overlapped them onto each other. However, since sometimes the
red side of the spectra was longer or shorter than the blue side of the spectra, this would lead
to problems in the overlapping that would make those data files unable to be used. This is one
of the primary reasons we only stacked around 1800 slits in DEEP3 and 400 slits in DEEP2
analyzed and stacked, because many of the slits werent able to be stitched together properly.
2.0.2 Restoring Wavelength Solution
Each 2D spectra in the DEEP2 and DEEP3 data sets in its basic form represents the bright-
ness measured along a slit given a wavelength on the x-axis and a distance on the y-axis. While
storing this data, the measurers of the DEEP2 and DEEP3 data set stored it using two arrays, 0
and d, to avoid any losses of precision while allowing the files to be more readily compressed,
but this also meant that any basic spectra plotting would be severely tilted or altered. This is
because the wavelength information of the slit was encoded using Equation 1.
Λ(x,y) = Λ0[x]+dΛ[x,y] (1)
This means that any 2D spectra initially plotted isnt visually accurate as the pixels in a column
of the wavelength axis do not actually share the same wavelength. As a result, Equation 2 had
to be used to translate the data to assign the proper value to each pixel in the DEEP2 and DEEP3
data files before an accurate 2D spectra could be plotted.
4
7. 2.0.3 Mask and crMask Filtering
In the DEEP2 and DEEP3 surveys, the measurers of the data assigned Mask and crMask
variables to each brightness, or flux, value in the spectra, with accurate pixels having a value of
0 and noted abnormal pixels having a value of 1. More specifically, the Mask variable describes
if a pixel is vignetted or not, which refers to a phenomenon where light is attenuated away from
the center of a spectra (Zheng et al. 2008), while the crMask variable describes if a pixel was
impacted by cosmic rays. In both cases, certain flux values in the spectra would be altered from
their true value which would affect the end result when stacked. Thus, when running through
the data preliminarily, the flux values at every point with the Mask or crMask variables equal
to 1 were set to 0 to prevent any distortions later on in the GUI inspection stage or the stacking
stage.
2.0.4 Interpolation
For many spectra, it is possible due to the large quantity of data that a flux value is shifted
one pixel left or right from its real position. To solve this problem, each spectra was linearly
interpolated into a grid format along the wavelength axis with each box one kiloparsec long
and 0.5 ˚A wide in order to account for any possible discrepancy pixels. This is because through
interpolation, where a function is mapped based on every pixel in a row, each flux value in
that row is assigned the designated function value, thus removing any large outliers from being
stacked.
2.1 Graphical User Interface Development
When analyzing the galaxies in the DEEP2 and DEEP3 surveys, the possibility of skewed
spectra is large among them. Many of the spectra can contain skylines covering and distorting
5
8. the wind emission, missing portions of the galaxy all together, other serendipitous galactic
objects in the spectra, abnormally bright fluxes from certain pixels, or too faint a signal from
the galaxy for any wind emission to be detected at all. If these galaxies were stacked alongside
other correctly imaged galaxies, they could cause spikes of fluxes at certain points outside of the
galaxy that would be mistaken for extended wind emissions. To ensure that false results dont
occur, each spectra must first be inspected for these flaws. In order to efficiently go through the
inspection of the 1781 galaxies in the DEEP3 galaxies and 341 galaxies in the DEEP2 galaxies,
a graphical user interface (GUI) as shown in Figure 2 was developed for quick inspections of
Figure 2. GUI for spectra analysis The GUI that was developed for the inspection of errors
and stacking. The upper left image shows the O II doublet emission, which contains no visible
errors. However, the upper right image representing the Mg II emission has black vertical lines
representing spectra and the
each galaxy to find these possible errors as well as to provide more accurate measurements of
distances to the various functions used.
6
9. 2.1.1 Inspection
The GUI was formatted to provide images of the spectra at each of the predicted wavelengths
for O II, Fe II, and Mg II. This way, when running through each of the data files in the DEEP2 or
DEEP3 surveys, the runners of the GUI could flag any files that contained skylines that covered
the emission, random missing portions of the spectra, or other serendipitous galactic objects.
However, in instances when the entire slit was cut off, for example for the Mg II emission, the
data file was still used if the O2 emission didnt show any skylines or wasnt too faint. This is
because when the entire slit was cut off for the Mg II emission in this case, the invariance and
fluxes at those values are both 0, meaning that they will have no effect when stacking all of the
the Mg II emissions. This scenario occurred mostly for the DEEP2 galaxies and not for DEEP3.
2.1.2 Determining Distance to Galaxy
After two test runs of going through the DEEP3 survey, a fixed procedure was developed
to optimize the implementation of the GUI. First, by looking at the theoretical O II flux area
at wavelengths 3726 ˚A and 3729 ˚A, the observed distance along the slit of the O II flux was
manually inputted into the GUI as an initial estimate of the distance to the Fe II and Mg II
doublets. This was necessary since the Fe II doublet and Mg II doublet both have much weaker
fluxes in the spectra, making it impossible to observe their distance along the slit when looking
at the spectra. Since the galaxy is at most 10 kiloparsecs wide for both the DEEP2 and DEEP3
galaxies though, the distance to the O2 emission line provided an accurate first measurement of
the predicted distance to Iron and Magnesium doublets for DEEP3 and Magnesium for DEEP2.
Using this first distance, an even more accurate estimate was calculated by collapsing the spectra
along the wavelength axes to generate a 1D version of the spectra as flux versus distance. Then,
by finding the distance of the maximum flux value in a range 10 kiloparsecs up and down the O
7
10. II derived distance, a proper centering could finally be conducted.
2.1.3 Gaussian Fitting
Alternatively, the fitting of a two dimensional double-gaussian function was also used sep-
arately to approximate the distances to the O II, Mg II, and Fe II emissions in DEEP3 galaxies.
This method involved fitting this general two-dimensional elliptical Gaussian function with a,
b, and c defined as follows in Equation 2 and 3,
f(x,y) = A∗exp(−(a(x−xo)2
+2b(x−xo)(y−yo)+c(y−yo)2
)) (2)
a = −
cos2 θ
2σ2
x
+
sin2
θ
2σ2
y
,b = −
sin2θ
4σ2
x
+
sin2θ
4σ2
y
,c =
sin2
θ
2σ2
x
+
cos2 θ
2σ2
y
(3)
In total, the gaussian function has six parameters with σx representing the spread of the
model along the x-axis, σy representing the spread of the model along the y-axis, A as the am-
plitude of the model, x0 and y0 as the x and y coordinates of the mean of the model respectively,
and θ as the angle of the emission.
When fitting this function to the O II emission, it was necessary to provide an estimate of
the correct parameters that would describe the gaussian model fitted to the O2 emission. This
is to prevent the fitting function from being too inefficient to run practically, given that it has to
go over each case for six parameters. This is important especially for the x0 and y0 variables,
because the function without a guess would have to scan through the full spectra until it finds
the O2 emission. As a result, to ensure the accuracy of the guesses specifically for x0 and y0,
the guess for x0 was observed using the GUI while the guess for y0 was based on the theoretical
wavelengths of 3726 ˚A and 3729 ˚A for O II emissions. For the rest of the parameters, they were
all guessed to be 1 with theta guessed to be 0 since the fitting function doesnt need any guesses
8
11. for these less flexible parameters. Through this method, an even more accurate distance to the
O II emission was able to be calculated, which then extends to a more accurate Fe II and Mg II
distance measurement as well. When doing the 2D gaussian fit though, the estimate was already
so close that this alternate method did not have any large effect on the end results and thus was
not used for Deep2 galaxies.
2.1.4 Centering
Determining the distances along the slit to the O II, Fe II, and Mg II emissions is vital in
order to stack the various emissions on top of each other. This is because different spectra have
the emissions at different distances along the slit, and in each spectra itself the distance along
the slit to one type of emission is different from the distance to other emissions. Thus, each
emission must have its own individual distance calculated or estimated so that they can all be
accurately shifted to a baseline of 0, which allows them to then properly overlap when stacking
the spectra.
2.2 Stacking Data Sets
After the GUI has finished being run through all the spectra in the DEEP2 and DEEP3 data
sets, the files can properly be stacked. However, instead of just adding all of them together, for
each spectras individual flux value there is a corresponding invariance value which determines
the accuracy of the flux. By using Equation 4
flux =
Σ(flux∗invariance)
Σinvariance
(4)
when stacking the finished emission data, the final result is much less sensitive to outliers in the
process as they are given less weight than more accurate data.
9
12. 2.2.1 GUI Stacking
During the centering and collapsing process when the GUI was being run, each time a
spectra was completed it would automatically be stacked in a separate display. This way, we
could see the effect that each data file has on the overall stacked result beforehand, preventing
any outliars from disturbing the data set. For example, during inspection if we missed a skyline
on the spectra, the effect of the skyline would still be immediately noticed when it distorts the
final stacked result and then be properly properly flagged.
2.2.2 Normalizing
After stacking the various types of emissions and the continuums behind them, the final
curves were all then normalized so that the maximum flux value would be one. This is because
there might have been different intensity profiles among the DEEP2 and DEEP3 surveys, which
would represent false outflow emission. Thus, by normalizing the final curves, the shapes of
both the continuum and the emission with the continuum can be overlapped and analyzed to
understand whether any wind emission exists or not.
2.2.3 Bootstrapping
Through the method of bootstrapping, the accuracy of the final stacked results could be
measured efficiently regardless of the amount of spectra available. For example, by randomly
taking a finished stacking data slit from a set of 2000 and replacing it, and by doing this ten
thousand times, a much more accurate measure of the error and the distribution of the final
result can be measured, even if the data set is relatively small.
10
13. 2.2.4 Multi-Variable Binning
When each spectra was being processed by the GUI to finally be stacked, the GUI also
saved the redshift, inclination, O II Emission Strength, mass, and zQuality for DEEP3 galaxies,
while only the redshift and zQuality were available in the data files be saved for DEEP2. After
acquiring the end result, by binning the final set of stackable spectra under the constraints of
many of these variables, potential relationships could be found between extended wind emission
and galaxies of a specific constraint.
3. Results and Discussion
After stacking the spectra of the DEEP2 data set for Mg II emission lines and the DEEP3
data set for Fe II emission as well, it was found that there were no large signs of extended wind
emission for either doublet. Extended wind emission is shown when there is positive flux when
the Mg or Fe emissions are subtracted from either the continuum or the O II tracer.
3.1 DEEP2 Wind Emission
With 156 out of 341 spectra being flagged for various errors in the DEEP2 data set, the
remaining 185 high-quality spectra were stacked with invariance weighing and plotted in one
dimension. For DEEP2 specifically, only the Mg II doublet was examined for wind emission
because the slits didnt contain the 2612 ˚A to 2626 ˚A wavelengths where Fe II is present. The
final stacked result for DEEP2s Mg II is shown in Figure 3.
11
14. Figure 3. Bootstrapped DEEP2 Mg Emission Line and Continuum This figure shows the
final result of the normalized and stacked Mg II emission plotted on top of the interpolated
continuum. It has been bootstrapped 1000 times to better represent the final result’s error.
The Mg II emission in Figure 3 does extend past the continuum in some areas, but not by an
expected amount of 0.2 that would indicate galactic-scale outflow. Thus, it was concluded that
for the DEEP2 data set that there was no extended wind emission present. In order to see how
much the Mg II extends past the interpolated continuum more clearly, the difference was plotted
in Figure 4. The standard deviation calculated for the DEEP2 Mg II emission was 0.0108.
Figure 4. Bootstrapped DEEP2 Mg II Emission Line and Continuum In this graph, the
interpolated continuum behind the Mg II emission line and the Mg II emission line were
normalized and plotted against the distance from the center of the emission. It was
bootstrapped 1000 times to show the span of the error.
12
15. 3.2 DEEP3 Wind Emission
Similar results were found for the DEEP3 Data set, where the Mg II and Fe II emission
lines both didnt go a significant amount past the interpolated continuum. It is important to note
as well that the DEEP3 data set had almost six times as many galaxies used in the data set
as DEEP2, thus representing a larger accuracy as well. In addition, the DEEP3 Data set was
analyzed differently using the O II emission as a tracer. This is because as shown in Figure 5,
the O II emissions normalized flux shape is almost exactly the same as the continuums shape
as shown in this picture. By using this tracer as well, the detection of wind emission becomes
much more accurate, because here O II serves as a closer measure of the continuums shape than
the interpolated continuum around Mg II.
Figure 5. Bootstrapped DEEP3 O II Emission Line and Continuum In this graph, the
interpolated continuum behind the O II emission line and the O II emission line itself were
normalized, stacked, and then plotted against the distance from the center of the emission. It
was bootstrapped 1000 times to show the span of the error.
13
16. Figure 6. Bootstrapped DEEP3 Mg II Emission Line and Continuum In this graph, the
interpolated continuum behind DEEP3’s O II emission lines and the O II emission line itself
were stacked and normalized. Then they were plotted against the distance from the center of
the emission. It was bootstrapped 1000 times to show the span of the error.
In Figure 6, the Mg II emission line plotted on top of the continuum is similar to the DEEP2
data sets stacked graph as well, where the Mg II emission overlaps the Mg II continuum in
some places, but overall there is no galactic-scale sign of extended wind emission. The Mg II
emission for DEEP3 was this time subtracted from the O II emission because the O II emission
acted as a good enough tracer for the curve of the galaxy. As shown in Figure 7, there were no
traces of extended wind emission across multiple objects, with the standard deviation for Mg
here calculated as 0.0109 flux.
Regarding the Fe II emission in the DEEP3 data set as well, there was no large extension
past the interpolated continuum as shown in Figure 8, with the standard deviation calculated as
0.016.
In whole, we can conclude that for both Fe II and Mg II, wind emissions in the DEEP3
galaxies were not found when stacking all the galaxies on the surveys.
14
17. Figure 7. DEEP3 Mg II Subtracted from O II In this graph, the Mg II emission line had the
O II emission line subtracted from it. It was bootstrapped 1000 times to show the span of the
error.
Figure 8. Bootstrapped DEEP3 Fe II Emission Line and Continuum In this graph, the
interpolated continuum behind DEEP3’s O II emission lines and the O II emission line itself
were stacked and normalized. Then they were subtracted from each other to show the presence
of wind emission. It was bootstrapped 1000 times to show the span of the error.
3.3 Discussion
Given the robust nature of the procedure conducted on the DEEP2 and DEEP3 galaxy sets
to find wind emission, it was concluded that there was no wind emission detected across these
galaxies and that the previous work of Sekhar and Huang is incorrect. We thus provide a spatial
constraint on the DEEP2 and DEEP3 galaxies that wind emission does not exist within 15 kpc of
15
18. the center of the galaxy using the largest data set ever conducted for this purpose. Additionally,
it was already predicted that the DEEP2 wind emission was ubiquitous throughout the DEEP2
galaxies, but the study couldnt identify the spatial extent or the amount of mass lost for the
wind emission. (Weiner et al. 2009) However, with this constraint future researchers will have
a much better chance of observing the place of the DEEP2 wind emission and where it is indeed
present. However, it is important to note that this project was limited by the number of quality
spectra contained in the DEEP2 and DEEP3 data set that passed through our entire procedure.
After running through the inspection with the GUI, approximately half of both sample sets of
data were removed in order to ensure the highest quality galaxies free from the six possible
errors that could occur when stacking them on top of each other. Thus, even though our galaxy
set was still much larger than studies like Tangs study and the single galaxy wind detection
studies, it could be better improved if there were a much larger set of galaxies analyzed.
Conclusion
Currently, research into extended wind emissions is an expanding field, and efforts to dis-
cover them have always been difficult. Given this difficulty, it has usually only been accurately
detected in one object, while looking for extended wind emission over many galactic objects
has thus far remained a relatively new topic. Therefore, a novel, more robust method of detect-
ing the spatial extents of wind emission was created for the purpose of finding wind emission
in distant galaxies. By conducting our procedure, the detection of any wind emission within 15
kiloparsecs from the center of the galaxy can now be determined across multiple galaxy sets
efficiently. Additionally, through the results of our analysis with the GUI on the DEEP2 and
DEEP3 data sets, it was interpreted that Mg II and Fe II present no observable wind emission 15
kiloparsecs from the center of galaxies for redshift 1 galaxies. More importantly, the effects of
16
19. serendipitous objects, skylines, and outlier pixels in galaxies can be determined to have a mim-
icking effect on the detection of wind emission in galaxies. When looking for the Fe II emission
in the DEEP2 galaxy set, without the flagging of spectra for these serendipitous objects, sky-
lines, and outlier pixels, there was a much more significant signs of wind emission than with
the bad spectra flagged. Thus, it is essential for current and future researchers trying to detect
extended wind emissions to recognize the importance of analyzing for serendipitous objects in
the data set instead of stacking them automatically. In the future, this report hopes to look to-
ward other galaxy sets in efforts to detect wind emission in those regions. More specifically, it
is worth looking toward galaxies with redshifts between 3 to 4 for at least Fe II emission seeing
as the galaxies from redshifts near 0.7 to 1.5 dont show any signs of extended wind emission.
After going through multiple sets of galaxies and compiling them all into a large list, future
researchers will hopefully be able to determine how galactic winds have had varied effects on
galaxies depending on their redshifts, masses, or other important characteristics.
17
20. References
1. Veilleux, S., Cecil, G. & Bland-Hawthorn, J. ARA&A. 43:769-826 (2005).
2. Zahid, H. J., Torrey, P., Vogelsberger, M., Hernquist, L., Kewley, L. & Dave, R. Ap&SS.
873-879 (2014).
3. Martin C. L., Shapley, A. E., Coil, A. L., Kornei, K. A., Murray, N. & Pancoast, A. ApJ.
770:41 (2013).
4. AGUIrre, A., Hernquist, L., Katz, N., Gardner, J. & Weinberg, D. ApJ. 556:L11-L15 (2001).
5. Creasey P., Theuns, T. & Bower R. G. MNRAS. 446:2125-2143 (2014).
6. Weiner, B.J., Coil, A.L., Prochaska, J. X., Newman, J. A., Cooper, M. C., Bundy, K., Con-
selice, C. J., Dutton, A. A., Faber S. M., Koo, David. C., Lotz, J. M., Rieke, G. H. & Rubin, K.
H. R. ApJ. 692:187-211 (2009).
7. Tang, Y., Giavalisco, M., Guo, Y. & Kurk, J. ApJ. 793:92 (2014).
8. Voort, F., Schaye, J., Booth, C. M. & Vecchia, C. D. MNRAS. 415:2782-2789 (2011).
9. Rubin, K. H. R., Prochaska, J. X., Menard, B., Murray, N., Kasen, D., Koo, D. C. & Philips
A. C. ApJ. 728:55 (2011).
10. Westmoquette, M. S., Smith, L. J., Gallagher, J. S. III & Exter, K. M. MNRAS. 381:913-931
(2007).
11. Newman, J. A., Cooper, M. C., Davis, M., Faber, S. M., Coil, A. L., Guhathakurta, P., Koo,
D. C., Philips, A. C., Conroy, C., Dutton, A. A., Finkbeiner, D. P., Gerke, B. F., Rosario, D.
21. J., Weiner, B. J., Wilmer, C. N. A., Yan, R., Harker, J. J., Kassin, S. A., Konidaris, N. P., Lai,
K., Madgwick, D. S., Noeske, K. G., Wirth, G. D., Connolly, A. J., Kaiser, N., Kirby, E. N.,
Lemaux, B. C., Lin, L., Lotz, J. M., Luppino, G. A., Marinoni, C., Matthews, D. J., Metevier,
A. & Schiavon, R. P. ApJS. 208:57 (2013).
12. Johnson, H. E. & Axford, W. I. ApJ. 165:381-390 (1971).
13. Matthews, W. G., Baker, J. C. ApJ. 170:241-259 (1971).