PAPER 1
RESULTS
PAPER 2
Add SFRs to analysis. Focus on size, stellar mass, SFR relation
RESULTS
In [2]:
import numpy as np
from pylab import *
%matplotlib inline
import warnings
warnings.filterwarnings('ignore')
In [4]:
%run ~/Dropbox/pythonCode/LCSanalyzespirals.py
In [5]:
# updating sigma vs LX to include errorbars on sigma values
plotsigmaLx()
In [81]:
%run ~/Dropbox/pythonCode/LCSbiweight.py
In [82]:
getbiweightall()
In [83]:
plotall()
Bianca asked about the specifics of how I calculate local density. I need to dig through code to figure out exactly what I do.
Looks like I calculate local density in
LCSwritemasterNSA.py
The local density measurements are written to
homedir+'research/LocalClusters/NSAmastertables/LocalDensityTables/'+self.prefix+'_localdensity.fits'
These are read by
LCSReadmasterBaseNSA.py
as self.ld
When the columns are merged:
LCSmergespiralcats.py
calls
LCSReadmasterBaseNSA.py
for each cluster and the following columns are added to the merged table
ld_columns=['SIGMA_NN','SIGMA_5','SIGMA_10','RHOMASS']
The local density measurements use the SDSS spectroscopic sample, so all $r < 17.7$ galaxies are included. It would be more appropriate to use an absolute magnitude cut because we are sampling clusters at a range of redshifts. Using the apparent magnitude cut will make Coma's local density measurements greater than Hercules, just because of the different distances to each cluster. Therefore, I should redo the local density cuts and use an absolute r-band magnitude cut.
The appropriate absolute magnitude cut is $M_r$ that corresponds to $r=17.7$ at Hercules. I already refer to this in the paper.
The stellar mass limit is determined primarily by the SDSS spectroscopic limit of $r=17.7$. This corresponds to an absolute magnitude of $M_r = -22.3$ at the distance to Hercules ($m-M=35.97$), our furthest cluster.
Therefore, I should use $M_r = -22.3$ cut when calculating local density.
Checking distance modulus.
$$ m_2 - m_1 = 2.5 \log_{10}(f_1/f_2) $$$$ m - M = 2.5 \log_{10}\left( \frac{L}{4 \pi (10~pc)^2} \frac{4 \pi d_{pc}^2}{L} \right) $$$$ m - M = 5 \log_{10} (d_{pc}) - 5 $$For Hercules, $z = 0.037$ and
$$ d_{Mpc} = z * c /H_0 = (0.037)(3.e5)/(70~km/s/Mpc) = 158.6~Mpc$$Thus
$$ m - M = 5 \log_{10} (158.6e6) - 5 = 36.00 $$To get the absolute magnitude corresponding to $r = 17.7$
$$ M = m - 36 = 17.7 - 36 = -18.3 $$Run:
LCSCalcLocalDensity.py -m -18.3 -v 2500
LCSmergespiralcats.py -r
mergedata()
s=spirals()
In [ ]:
In [9]:
%run ~/Dropbox/pythonCode/LCSanalyzeblue.py
In [10]:
s.compare_cluster_field()
No serious systematics between field and cluster samples.
I updated LCSmaketileplot.py so that a given galaxy wouldn't be repeated when randomly selecting galaxies within a certain range of size ratios (using numpy.random.shuffle).
In [85]:
# run LCSmaketileplot.py
%run ~/Dropbox/pythonCode/LCSmaketileplot.py
# plot galaxies with size ratios between 0.1 and .3
# 0 < Re(24)/Re(r) < 0.3
paperplots()
#s.plotcolor_subset(size1=0,size2=0.3,showall=False)
#savefig(homedir+'research/LocalClusters/SamplePlots/cutouts_smallsize.png')
#savefig(homedir+'research/LocalClusters/SamplePlots/cutouts_smallsize.eps')
# plot galaxies with size ratios between 0.3 and 0.6
# 0.3 < Re(24)/Re(r) < 0.6
#s.plotcolor_subset(size1=0.4,size2=0.6,showall=False)
#savefig(homedir+'research/LocalClusters/SamplePlots/cutouts_medsize.png')
#savefig(homedir+'research/LocalClusters/SamplePlots/cutouts_medsize.eps')
# plot galaxies with size ratios greater than 0.75
# 0.75 < Re(24)/Re(r)
#s.plotcolor_subset(size1=0.75,size2=2,showall=False)
#savefig(homedir+'research/LocalClusters/SamplePlots/cutouts_largesize.png')
#savefig(homedir+'research/LocalClusters/SamplePlots/cutouts_largesize.eps')
As of 1/6/16, need to make one more pass through the sample and remove galaxies that are blended with nearby companion. Not sure if people think the numbers in each panel are useful.
Galaxies that are blended with a nearby companion are:
ALSO running this from ipython and using paperplots().
Changed in May 2016 to normalize the size of the 24um emission using the size of the r-band disk only. The disk sizes are from Simard+11 B/D decompositions. The sense of the results are the same, but the average ratio of $R_e(24)/R_e(r)$ is now close to one; when I normalized using $R_e$ from the single-component Sersic fit, the average/median ratio of $R_e(24)/R_e(r) = 0.6-0.7$.
In [24]:
%run ~/Dropbox/pythonCode/LCSanalyzeblue.py
In [4]:
plotpositionson24()
In [5]:
plotRe24vsReall()
In [6]:
s.plotsizehist()
In [25]:
s.plotsizestellarmassblue()
In [28]:
print 'Field size ratios (Re(24)/Re(r)):'
flag=s.sampleflag & ~s.agnflag & ~s.membflag & s.gim2dflag
print 'mean = %5.2f'%(np.mean(s.sizeratio[flag]))
print 'weighted mean = %5.2f'%(ws.weighted_mean(s.sizeratio[flag],weights=1./s.sizeratioERR[flag]))
print 'Cluster size ratios:'
flag=s.sampleflag & ~s.agnflag & s.membflag & s.gim2dflag
print 'mean = %5.2f'%(np.mean(s.sizeratio[flag]))
print 'weighted mean = %5.2f'%(ws.weighted_mean(s.sizeratio[flag],weights=1./s.sizeratioERR[flag]))
The right panel shows the weighted mean vs. stellar mass. The size ratio for cluster galaxies lies below that of the field, but there is no trend between size and stellar mass.
Need to change this so that the solid and dashed lines show the weighted mean for the field. As of now, they are showing the mean.
also need to reset the y axis to reflect the higher size ratios that I get when using the disk size only.
In [8]:
s.plotsizeBTblue()
This shows $R_e(24)/R_e(r)$ vs. B/T for cluster and field galaxies. The left panels show the values for individual galaxies. The $\rho$ values indicates the 68% confidence interval for the Spearman rank correlation coeffienct, which I determine using Monte Carlo resampling. The Monte Carlo code assumes that the errors reported by galfit for $R_e(24)$ are gaussion - this is not likely to be true, but it's better than nothing.
Correlation is less significant among field galaxies.
I decided to put these in a separate notebook.
In [95]:
s.plotsizesb()
At first glance, this seems disconcerting, but the two variables are correlated b/c I use Re(24) to calculate surface brightness. So larger Re(24) should have fainter surface brightness.
R partial correlation analysis shows some very strong correlations between B/A and parameters I would expect to be correlated.
stellar mass
SFR
surface density
NUV24 color - this one is not surprising
In [6]:
figure()
plot(s.s.fcaxisratio1[s.sampleflag & ~s.agnflag],s.s.SIZE_RATIO[s.sampleflag & ~s.agnflag],'bo')
xlabel('B/A')
ylabel('Re(24)/Re(r)')
spearman(s.s.fcaxisratio1[s.sampleflag & ~s.agnflag],s.s.SIZE_RATIO[s.sampleflag & ~s.agnflag])
Out[6]:
In [7]:
s.s.columns
Out[7]:
In [14]:
figure()
plot(s.s.fcaxisratio1[s.sampleflag & ~s.agnflag],s.s.FLUX24[s.sampleflag & ~s.agnflag],'bo')
xlabel('B/A')
ylabel('FLUX24')
spearman(s.s.fcaxisratio1[s.sampleflag & ~s.agnflag],s.s.FLUX24[s.sampleflag & ~s.agnflag])
Out[14]:
In [13]:
figure()
plot(s.s.fcaxisratio1[s.sampleflag & ~s.agnflag],s.NUVr[s.sampleflag & ~s.agnflag],'bo')
xlabel('B/A')
ylabel('NUV-r')
spearman(s.s.fcaxisratio1[s.sampleflag & ~s.agnflag],s.NUVr[s.sampleflag & ~s.agnflag])
Out[13]:
I make the claim that the MIPS data is much lower resolution and SNR than typically used with galfit, but Greg said no.
Median redshift of my sample is $$z=0.03.$$ At this distance, there are $$0.604~kpc/arcsec.$$ The MIPS pixel size is 2.45 arcsec, so this means 1.48 kpc per resolution element.
The pixel size of Hubble is 0.1 arcsec/pixel. Therefore, to get 1.48 kpc per resolution element, we want to find the redshift where that angular diameter distance corresponds to $\sim$14.8 kpc/arcsec. Using Ned Wright's cosmology calculator to figure this out. Can't seem to get this to ever get that high. The max value peaks around $z = 1.6$ at 8.615 kpc/arcsec.
So, it looks like my statement is correct.
In [0]: