Week 5: Final lab session

Luminosity bins

We suspected that there is a bias when looking at high redshift sources, much like there is a 3-sigma detection limit that limits the sources we see at high redshift, we predict that more distant sources must be more luminous for them to be detected by us. To eliminate this bias, we bin by luminosity.
We have a total of 5 bins of Lyman alpha luminosity ranging from 10^42.1 ergs^-1 to 10^44.6 ergs^-1 in intervals of 10^0.5 ergs^-1. Lyman alpha luminosity was used instead of X-ray or radio since all AGN are Lyman alpha emitters and this luminosity has not been used to calculate accretion rates.

Figure 1: Redshift vs BHAR. All AGN sources are plotted and have been placed into bins of Lyman alpha luminosity. The lowest accretion rates trail upward at higher redshift due to the 3-sigma detection limit.

If what we thought was correct, we would have expected the most luminous sources to be at high redshift. Figure 1 shows that this is not the case; our most luminous sources are represented by red points which are found between z ~ 2.5 and z ~ 3.5. In fact, our most distant sources are the second least luminous sources suggesting that there is no bias based on luminosity of the sources.

Most AGN sources are found around redshift 3 which agrees with the theory that AGN activity peaks between redshift 1 and 3.

We plotted Figure 2 to see whether there was any correlation between how luminous sources were in the Lyman alpha band and their BHAR. We find that there is a small, positive correlation for all the AGN sources.

Looking within the bins we different trends at different luminosities. What we found interesting is that at high luminosities (>10^43.1ergs^-1), all the classifiable sources are Type 1: Compact. This could suggest something about the structure of galaxies and how this relates to its luminosity, or it could be coincidence. Out of all the AGN, 232 are classifiable sources (morphology type 0 – 3), 63% of these sources are Type 1: Compact and the top two most luminous bins of Lyman alpha luminosity contain only 19% of the classifiable AGN. There is a good chance then that these results are just coincidental.

Comparing the data

After all our hard work, we thought it was important to check that we had reproduced the data correctly.

Figure 3: Data obtained from the reference paper against data calculated by CHARSS. The points lie on an x = y line showing that our calculated values are in complete agreement with the expected ones.

In the coming weeks we will be writing the report ready to publish our results.

Week 4: Binning the sources and continuing our analysis

Carrying on from last week’s preliminary plots, we dedicated this week’s lab session to looking at the data in more depth. To start we binned the sources based on red shift and morphology. We have six red shift bins of width 0.6 and three bins of morphology. For the morphology we have decided to combine type 0: point like, and 1: compact, into a single type called ‘compact’, this was done because the structures of point-like and compact galaxies are very similar, the only difference being their size, meaning if there is a correlation between galaxy structure and BHAR, 0 and 1 type galaxies should be the same. We also have five bins of Lyman alpha luminosity which we will be using next week. Lyman alpha luminosity was chosen over using either the X-ray or the radio luminosity because it was not used for the calculations of the BHARs and all AGN are Lyman alpha emitters so all have a value for this.

To make the plots we used a python script, this allowed us to plot all the data along with the median values of each bin on one plot. Taking an average would be more accurate provided we correctly identify outliers but using a median means that outliers do not dramatically affect our results.

X-ray BHAR

Figure 1: All AGN sources are plotted as blue squares. The median X-ray calculated BHAR for each type of morphology is shown for each red shift bin.

An upward trend of increasing X-ray BHAR with red shift can be seen. More interestingly, we seem to see a correlation between morphology and X-ray BHAR within each red shift bin, with only the final bin showing something different. It also seems that irregulars tend to have the lowest average X-ray BHAR, but since the sample size for ‘disky’ galaxies is so small I would be hesitant to draw any conclusion from this.

Radio BHAR

Figure 2: All AGN sources are plotted as blue squares. The median radio calculated BHAR for each type of morphology is shown for each red shift bin.

This plot paints a very different picture to the previous one. Here we can see that, for most of the red shift bins, compact galaxies have lower than average BHARs when they are calculated using radio data. There is also a stronger correlation between red shift and radio BHAR than X-ray BHAR showed.

It may seem odd that calculating BHARs using different wavelengths produces different results but we know that the X-ray and radio data span different time scales so the two accretion rates being different for any one source suggests that there is some evolution of the galaxy that causes the accretion rate to change over time.

It may be useful for us to look at each type of morphology and compare the two BHARs to see if there is a correlation between morphology and how much BHAR changes over time. This is interesting as it would give insight into the evolution of different types of galaxies.

The Outliers

Within our data we have two outliers; a type four, X-ray detected source with an X-ray BHAR calculated at 3547 solar masses per year, and a type three source which was detected in both X-ray and radio with a radio BHAR calculated at 9228 solar masses per year. To put this into perspective, our second largest accretion rates for X-ray and radio BHAR are 22.8 and 69.1 solar masses per year, respectively. For now these have been ignored in our plots as they would skew the results too much. It is possible that these anomalies are simply errors. For example, the type four source was found very close to the edge of the field where the background noise is much less than is accounted for. Another explanation is that these are very bright, very luminous AGN; quasars. Even still, these accretion rates seem too high as for an average quasar you would be looking at an accretion rate of a few hundred solar masses per year, not a few thousand.

Week 3: Valentine’s day special

Victory! During the third lab session Ben completed the program so that it calculates the black hole accretion rates for all the AGN using the X-ray and radio data previously calculated in the program, and the errors for them.

Accretion rates were calculated using the X-ray and radio data respectively using the following equations:

Long at last, we now have a complete catalogue of data for the 322 individual AGN sources, 57 of which are detected in both the X-ray and radio.

Every piece of data was from the Chandra COSMOS Legacy survey, the VLA COSMOS survey, and our own calculations were collated into one large catalogue using TopCat. This means we can easily compare the two data sets and the methods used in each study. Our values are expected to differ slightly due to the different cosmologies used in the two studies but these differences shouldn’t have much effect on the overall trends.

Now analysis of the data can begin. Our initial observations were that the highest black hole accretion rates are found between red shift 2.2 and around 3.5. This is promising as it is agreement with other papers we have looked at.

There are some sources in the catalogue of AGN are not checked as being in the radio or the Chandra full detect because they don’t pass the 3 sigma signal to noise cut threshold limit, however they still have been included as AGN since they are luminous to be classed as such.

In next weeks lab we plan to properly dive into analysis of the data by binning the sources based on morphology, red shift and luminosity. But for now we begun by plotting a few basic plots to investigate overall trends.

Not much of a correlation can be seen here in either the X-ray or the radio BHARs. We do however see a slight overall increase in both X-ray calculated BHAR and radio BHAR. BHAR is in units of solar masses per year.

Even though it seems that type 0 and 1 AGN have higher BHARs, most sources are either 0 or 1 and only a small proportion of these have higher than average BHAR. With very few sources being 2 or 3 it is hard to draw a definite conclusion about any correlation. 4 represents sources we cannot classify.

Seemingly there is no correlation between morphology and red shift. It is expected that the sources at higher red shift will be less luminous so may appear smaller i.e. look like point source rather than compact. It may be useful to combine type 0 and 1 when analysing as they are very similar morphologies.

Dealing with only the sources that are detected in both X-ray and radio, we see that the X-ray BHARs and radio BHARs are somewhat different, this is because they span different time scales.

Although there is a distinct lack of correlation in most of these plots, it was to be expected as these are the conclusions drawn by previous papers.

You’re probably a bit disappointed at the lack of correlation in our results so here’s something that has a strong correlation:

Week 2

A breakthrough in the code

After many painful hours of trying to ‘fix’ the python program so that it could reproduce the luminosities from count rate found in the reference paper, it became apparent that it wasn’t that our program didn’t work, but that the reference catalogue contained quantities calculated with a different cosmology. It’s safe to say we were more than thrilled to have a working program at last, all credit to the man of the hour, Ben, of course. The next course of action with the code was to calculate black hole accretion rates (BHARs) for the X-ray data which was achieved by the end of the lab session.

Continuing the classification of the sources based on morphology, we looked at the X-ray data in DS9 for the sources that were not in the Hubble field and tried to classify them based off this. However, it was pointed out to us that galaxies are going to look different in different wavelengths so classifying them based on how they look in X-ray or radio will not give comparable results to the ones that have been classified in the optical. We made the decision to put galaxies with no Hubble image into the ‘4’ category. All fours will be discarded when investigating the relationship between morphology and BHAR.

After classifying all the galaxies we have in X-ray and Radio we compared the two sets and found that they have very similar fractions of each type of morphology. A quick conclusion from this is that morphology does not affect/ is not affected by the type of radiation the galaxy is detected in.

Most of the sources are either compact (1) or point-like (0) in both radio and X-ray, type 4 galaxies are ignored in morphology analysis as they represent unclassifiable sources.

All the X-ray and radio detected sources are AGN, as their luminosities in their respective bands exceed by a large margin the typical luminosities of a star forming galaxy (SFG). In the X-ray for example, all of our sources have a luminosity exceeding 10^42 ergs^-1, this would correspond to a star formation rate of 1000 solar masses per year, far in excess of a typical SFG, meaning this X-ray emission must be from an AGN source.

There are a total of 258 X-ray sources and 121 radio sources (57 of these are also detected in the X-ray so represent the same source).

Week 1

Beginning to classify the sources

The first step in our project is to sort through the data and find the sources that are AGN. Looking at the SC4K data catalogue from the COSMOS survey we see that some sources are detected in the X-ray (258) and some are detected in the radio (121). We have 13 different slices of red-shift between z ~ 2 and z ~ 6.

This plot shows the data available, we are interested in the blue and green data detected in X-ray and radio respectively. As can be seen from the plot, there are more sources at lower red-shift, especially when looking at the radio data

We have decided to start with focusing on the X-ray data from the Chandra COSMOS legacy survey before moving on to the radio as the X-ray data is easier to work with and we have more sources in the X-ray. To find AGN we are looking for sources that have an X-ray luminosity of greater than or equal to 10^42 ergs^-1 within one standard error. To save time and reduce the chances of human error we are putting all of the sources through a python program which will tell us which of the sources are AGN.

Our coding lead, Ben spent the lab session writing and testing the program. The program was set up to change count rates to X-ray flux which is needed to find X-ray luminosity. Using TopCat, luminosity distance was calculated from red-shift which is going to be used along with X-ray flux to find the luminosity. As expected with any amount of programming, we ran into a few issues: there were unexpected steps and factor of 10 issues, meaning we left the lab without finding out which sources were definitely AGN and which weren’t.

The rest of us made a start on classifying galaxies based on their morphologies (their shape). We used the same classification scheme as the SHREDS team from last year’s group project for the sources. We added type ‘4’ as the category of sources where we saw nothing or found that the image is too distorted to draw any conclusions.

Examples of the different types of morphologies. Four members of the group assigned numbers to each of the sources and we took an average of the values to assign a source to a group.

This started as a bit of a filler exercise as practise for when the program was ready but we chose to classify all sources that are found in X-ray from the COSMOS survey since from just looking at the table of data with luminosities we think they may all be AGN. Using COSMOS Cutouts we downloaded 2” by 2” images of all the sources taken by the Hubble telescope. Most of the sources seem to be ‘compact’. Not all of the sources found in the X-ray from the SC4K catalogue are in the COSMOS Hubble coverage; the ones outside the field do not have images from Hubble. For the AGN we find outside of the field will try and classify them using X-ray data in ds9.