SNAG: Studying Nearby AGN & Galaxies: Post 5

This week has been the final week of lab work, and as such we have been rounding up our findings and neatening up all the graphs we have made. Jonathan went over the graphs he had made and colour coordinated them to better link the graphs to the data each point came from. 

As you can see this more clearly illustrates where the relationship between the graph and the original data, and as was mentioned in last weeks post there were some clear links showing up between mass and temperature of a galaxy, as well as mass and metallicity as we would expect. John began the same process for the AGN galaxies, splitting the whole subsection down into smaller sections and calculating average values in each section. Below shows the AGN containing galaxies split down into the smaller sections used for analysis (CLOUDY data has not been overlaid in this image).

Now as all three active galaxy types had been analysed Tom was able to produce some graphs which showed trends across them, laid side by side for easy comparison. An example is shown below where you can see that the star forming and star burst galaxies show similar trends, which makes sense as star burst galaxies are very close to star forming, but the AGN galaxies show the complete opposite relation between their masses and star formation rates.

Charlie spent this week redefining how we identify passive galaxies in order to reduce the amount of overlap between galaxies already identified by the SDSS and ones we had identified as passive. She achieved this with much success, meaning that at the end of our investigation, 90% of our data has been classified, 41% was classified by SDSS and a further 52% we have now identified as being passive with an overlap of only 3%. Identification of passive galaxies was achieved using a combination of the error in the flux of the Hydrogen Alpha emission line, and the equivalent width of the emission line which looks at the area covered by the emission line.

 Because of this, it was possible for Pascale and Phoebe to re-plot the fractional percentage graph of the different galaxy types and their mass. This is shown below and follows all the trends we would expect of the different galaxy types. The passive galaxies take up the large portion of the high mass galaxies as these have all become quenched, with the lower masses dominated by star forming and star burst galaxies, with the relatively rare AGN galaxies showing a slight bias towards higher masses. This graph is shown below.

To round our data analysis off, Jonathan plotted the position of the AGN galaxies on the sky with the rest of the data. The AGN for the most part are located in the more densely packed regions of the cosmic filaments (represented by the black dots on the image below), which we observed in the first week of data analysis. This reaffirms what we believe about AGN formation as the denser regions contain more mass causing the local environment to be more volatile. It is both reassuring to have our data continually agreeing with what theories predict, and a nice way to end the investigative part of our project by linking our current knowledge back to what we saw in week one.

Pascale Desmet

SNAG: Studying Nearby AGN & Galaxies: Post 4

Missed out first blog post? Click here to read it.

It’s now week 4 and the project has been progressing well. We are now starting to tailor our investigations back together between the two groups. This will ensure that we will come to some conclusion and ensure that the two projects running side by side will combine together in the last week.
This led to a large planning session before the week even started outlining the direction which we wanted to go with these labs. The end of many of the projects from the previous week meant that we had to find investigations within the interesting results we had already had. This led to a plan and a manic whiteboard being produced on the way to the said plan.  

Group 2 (Jonathan, Tom & John) carried on using the CLOUDY simulation due to the familiarisation that they had developed in the previous weeks. Their aim for the lab was to find how the quantities which we had been studying about the galaxies such as mass, density, SFR etc., changes throughout each of the subsets.  This involved splitting the different galaxy regions(starforming, starburst and AGN), which they had been working on in week 2 up into sub sections within cloudy.  Due to the progress last week each subset had been created and the cloudy data which coincided with each subset was known and refined. This meant that the trends for each subsection could be found form the CLOUDY data points which lay in this section. 

The above graph is the starforming galaxies split up into seven sections. The below is the starburst data, this was harder to separate due to the shape and is in 12 sections. The AGN subset will be done in next week. 

From these subsets, created on TOPCAT, they will use the data in the CLOUDY simulation to extrapolate the average (sigma) value of each of the quantities in the subset. This will then produce values of parameter allowing errors  to be calculated and trends in parameters across each region 

This produced a mixed bag of results once average values had been calculated. There are some plots where the errors are large when compared to the spread of the data 

This allowed us to confirm some trends such as age/mass of a galaxy. However we were unable to confirm whether there was any correlation between some factors such as metallicity and temperature due to the large errors associated with these variables. The errors in solar metallicity are so large that we end up with negative metallicity values in some cases making the results  unusable. Over the course of the day we were able to find a large amount of data which supported the trend between mass and age of the galaxies. This trend was confirmed both in the star forming and starburst data sets (it was also confirmed by group 1, through the trends of passive galaxies) 

The CLOUDY models being used to find the trends have some different properties depending which model you use, such as age or temperature. This means that there are some properties which can only be found for some galaxy types. It also means that when the same properties are being calculated using two different CLOUDY models, a comparison between them can be made. This has allowed us to see differences between the models, we presume this is due to different assumptions made for the models.
It can be seen in the BB vs BPASS density figure, these are the predicted densities of the subsections made by the two different models plotted against each other. They have produced largely different destiny approximations depending on the model used. We may follow this by determining what it is that makes the CLOUDY models create such different values from different models. 

Group 1 – ( Charlie, Ciara, Pascale, Phoebe) We began investigating the passive galaxies which we had within our data set. These are galaxies which are no longer forming stars. We initially defined the passive galaxies as those who had a H Alpha flux line which was smaller than three times the error. In the process we had previously removed these data points to help remove noise from the data set, after adding them back in and marking them as passive galaxies we were able to split the data.  All the points in green we have classified as passive galaxies. Those in blue had already been classified. The points in red are still unclassified in our data set.  

We started the photo-metric analysis of the data using the colour lines. We did this by matching the data set with the photometric data in another set. From this we were able to retrieve the UGRIZ filter flux for each data point. U-ultraviolet G- visible (green)R – visible (red) I- infrared Z- far infrared Using this data we were able to plot two colour indexes against each other, e.g (U-R) in order to hopefully identify/confirm the passive galaxies in our data set as they should cluster in a particular area. 
In this figure the blue data is the passive galaxies in our sample. They tend to clearly cluster together agreeing that the 3 flux condition is a good indicator of passive galaxies. However there are classified (starforming and AGN) galaxies which lie under the passive galaxy points which increases the uncertainty in our definition of passive galaxies which we are hoping to improve on next week.  In this figure the U-R and R-I colour indexes were used. 

However when we plotted the newly defined passive galaxies on a plot which we know clearly defines the star forming and AGN galaxies, a clear grouping was not apparent.  It was hoped for a clear, defined group of passive galaxies which would help to confirm that a good choice of subset condition had been done. 

The final task of group 2 was to plot the percentage of a galaxy type against mass. This has produced the trend we were expecting with the passive galaxies dominating at high mass, and star forming galaxies at low. There is also the slight peak of AGN galaxies at the higher mass as is predicted from their properties. 

Looking Forward Next week we are hoping to define our passive galaxies better, ensuring that the selection which we have made is correct and is not including noise or non passive galaxies. We may also look at plots made in previous weeks and decided whether adding passive galaxies will tell us any more information. The team using CLOUDY will do the sectional analysis of the AGN galaxies to have a complete set of values to be able to analyse. Further investigation into the differences in the models and how this materialises in the results may also be possible. 

SNAG: Studying Nearby AGN & Galaxies: Post 3

Missed out first blog post? Click here to read it.

Week 3 saw a slight change to the groups, as Pascale joined Ciara, Charlie, and Phoebe to use python to plot graphs using the SDSS data. The day started with Dr Sobral giving the group a crash course in the python script we had been provided with, showing us how to change the variables plotted, format the graphs produced, and (most importantly) how to debug the script when it inevitably went a bit wrong. 

“The 2 Peas in a Pod”, as Pascale and Phoebe were beautifully christened by Charlie and Ciara, set about making graphs that showed the percentages of each galaxy type (AGN and SF) against mass, and then did another series of graphs that showed the galaxy fraction vs element ratios. The ratios used were [OIII 5007/Hβ flux], [NII 6584/Hα flux], [OIII 4959/Hβ flux], and [NeII 3869/Hγ flux]. As this is astrophysics, all these ratios were of course done using a logarithmic scale. These ratios are often used in astronomy and astrophysics as the difference in the wavelengths between the two elements is small, which reduces the possibility of different degrees of reddening due to dust in the local environment of the galaxy.  

The mass graph clearly showed that galaxies with higher stellar masses were almost always AGN galaxies, with SF galaxies having lower masses. Most of the element graphs showed clear trends in the way different galaxy types had different element distributions, but the NeII graph showed no strong trend at all. Phoebe and Pascale tried to replot the graph, thinking that there was a problem in the code, but the answer was a lot more fundamental: NeII line ratios were not a good method of distinguishing between AGN and SF galaxies.  

Meanwhile, the other half of the group had hit a snag: the star formation rate graphs that were being plotted suggested that 85% of galaxies with an SFR of 0 were in fact those that had been classified as star forming galaxies. As we would expect these galaxies to be contributing most to star formation, this graph obviously threw up some interesting questions. After checking the code for errors, it soon became clear that the calculations for star formation rate were not totally accurate, and that the difference in the number of SF galaxies and AGN galaxies was likely to be skewing the data. 

Dr Sobral then suggested that starburst galaxies were included in the conditions, which improved the graph, and Ciara and Charlie figured out that the unexpected shape was due to the fact that SFR is calculated from H-alpha lines, but that AGN can give off large amounts of H-alpha without forming stars. This meant that many AGN galaxies were shown as having high SFR even when this isn’t the case. To rectify this, AGN galaxies were removed, and plots were done of SFR against the fraction of SF galaxies and starburst galaxies instead. 

Phoebe Stainton

The star formation rate (SFR) of star forming (SF) and starburst (SB) galaxies 

For the third week of our investigation, we continued to work in 2 groups.Jonathan, Tom and John continued investigations into what information we could gleam from the CLOUDY simulation and our data set, with each person investigating something unique.

Jonathan began investigations into the physical limits of the CLOUDY simulation. CLOUDY works to predict line spectral data with various inputs such as Temperature, Ionisation power, metallicity, and density. Each CLOUDY Model, BB, BPASS and PLAW use different starting conditions for these simulations, and we expect that most of these starting positions will represent galaxies that are un-physical.An example of this is shown below: 

 In the graph above, we can see that in grey is all the value CLOUDY predicts using the PLAW model, and in blue are all the values that correspond to real AGN galaxies (determined via comparisons of predicted OIII/HBeta and NII/HAlpha flux ratios). There are many grey values visible here, and all of these correspond to simulation inputs that have no association with real conditions in our galaxy data. Tom looked a bit closer at our data set this week and noticed that there were a significant number of data points who were uncategorised by the SDSS survey. 

  The initial thought was that perhaps these galaxies were passive (little to no star formation) given that most of these galaxies span across the SF and AGN regions of our galaxy, though upon further investigation it was revealed that most of these values are just data points with an unusually high error within the measurements of their flux. By taking our data sample with a tolerance of the Flux must be greater than 50x the error of the flux, the following graph was created, which reduces the number of unclassified (yellow) galaxies significantly.

John investigated a large range of other quantities with relation to both CLOUDY, and our data set itself. From CLOUDY, it was noticed that there were clear patterns and relations in the Carbon/Oxygen ratios that seemed to match with our data relatively well. While initially the reason for this was unknown, it turns out that this is a good representation of the metallicity of a galaxy.

In addition to this, the SFR  (Star Formation Rate) across our data was considered due to the work of Group 1 last week. The results were not what was expected however, with a seemingly uniform distribution appearing across our data. This is likely because of the way the SFR is calculated, using the Hydrogen Alpha emission lines as the measurement. As H-Alpha lines can be emitted in ways other than just star formation, this is the reason that our data set looks uniform, as opposed to the SF regions having the higher SFR,

Jonathan Dixon

SNAG: Studying Nearby AGN & Galaxies: Post 2

Missed out first blog post? Click here to read it.

This week, the team split into two groups, in order to run two different projects side by side.

Charlie, Phoebe and Ciara worked with the SDSS catalogues, calculating values for characteristics to be studied, including recessional velocity, comoving distance, and star formation rate.Through manipulating the data in Topcat, we were able to add data for luminosity distance, comoving distance and recessional velocity. Topcat has built in functions for some of these, although the calculation for recessional velocity had to be triple checked… 35,000 km/s seemed way too high, but our calculations are correct! Redshift is a no-linear scale, so although we are working with relatively low redshift galaxies, they are hurtling away from us at thousands of kilometres every second.

The manual calculation of star formation rate was rather challenging, requiring several steps and using the flux of each galaxy, taking into account reddening from dust. After reaching some very unreasonable values (10^18 solar masses per year!), Dr Sobral tipped us off about the weird units used in some astronomical measurements that may be present in the catalogues. Although we had been wary of the distance conversions (Megaparsecs to cm), we weren’t aware of the use of “ergs” in the measurement of flux. A quick search of the SDSS Data Products website revealed flux was in fact measured in erg/s/cm^2 and scaled by a factor of 10^17 in the catalogue. Finally we calculated the actual star formation rate, and were able to plot this against the 50th percentile mass estimate for each galaxy…INSERT logMass IMAGEThis shows, on log scales, the Star Formation Rate (SFR) in solar masses per year, and the 50th percentile estimate of mass of each galaxy in log of solar masses (so on the graph it is logged again!). This graph is to be analysed in our lab session next week.

We were also able to make some lovely graphs using various flux ratios. The SDSS measures the flux of four Hydrogen lines (alpha, beta, delta and gamma), along with absorption line flux at 12 wavelengths for 8 different ions. To make ratios of metal to hydrogen as accurate as possible, the ratios should be taken with the closest hydrogen wavelength: for example, NII at 6548 Angstroms should be taken as a ratio with H-alpha at 6563 Angstroms; OIII at 4959 Angstroms should be taken as a ratio with H-beta at 4861, and so on. This reduces the effect reddening might have on the ratios, assuming similar reddening at similar wavelengths. Some examples are shown below.


Each has a characteristic curve with a ‘plume’ coming off it. The last graph has been shown broken down into galaxy types: the curve tends to consist of star forming (including starburst) galaxies, while the plume on the graph is due to the AGN. Similar graphs will be used by the CLOUDY group to overlay our data with the results of the simulation.

Ciara Lithgow

For the second week of our investigation of the nature of various galaxies we decided to split up into two groups. Pascale, Jonathan, Tom and John took finding correlations between our data and the simulations found using CLOUDY software as its main objective. The first part of this involved observing our data and splitting it up into equal sections to analyze each part. We ended up getting a star forming section, a starburst section and an AGN section. Before conducting our analysis, we observed that there were entire areas of our plotted data that weren’t at all covered by the simulations from CLOUDY. We deliberated with the whole group about what these plumes might have been. After copious speculation we concluded that these areas must have been noise in the data.  This was deduced from the fact that the data points in the areas were subject to very large errors making the points unreliable sources of information.

This is our data now that the noise has been removed. The 3 regions we will investigate are clearly visible.

With the data being reduced even further, it was time to move on to the comparison with the simulation. Each individual section was analyzed by a separate member of the subgroup. The starbursts section, concerning galaxies in which the star formation rate has momentarily increased considerably, was analyzed by Tom who observed doubly ionized oxygen within the region as well as a comparison between the alpha and beta lines in the Balmer series.

Upon inspection of these graphs several trends are already visible and could be the source of further speculation during the next couple of weeks.

Jonathan took a similar approach in the observations regarding the star forming galaxies. Holding a broad range of quantities to look at many plots were achieved comparing the data to various simulations in CLOUDY, specifically regarding density, metallicity, age of the galaxy and where possible temperature.

Amongst the most interesting graphs here we respectively have a plot of the comparison with the BB simulation observing temperature, a comparison with the BB simulation displaying metallicity and a comparison with the PLAW simulation displaying metallicity.

Finally, John focused on a similar task in the observation of AGN. An important observation was that the BPASS simulation, as we expected, had barely any common ground with the AGN data. Whilst this was expected due to the nature of the simulation it was satisfying to see that the simulations were working parallel to our expectations. An interesting comparison was drawn between the BB simulation and the PLAW simulation, of which the latter is supposed to be the most accurate for AGN.

The larger metallicity occupies a defined area in the PLAW simulation (the top plot) which is what we expect given the ratios we are plotting. Finally, a comparison was made regarding density and metallicity within the PLAW simulation.

John Pollard