Week 6

Hello there, avid blog readers, I’m Adam, the group coordinator, and this will be the final update from the SHREDS team before our eagerly awaited paper is published in just a couple of week’s time. This record of our exploits over the past 6 weeks will undoubtedly sit beside the great literary works of authors such as Dickens, Austen, and Bronte in a library of timeless classics. There have been highs, and there have been lows. There have been laughs, and quite possibly tears. Python has worked first time, and QTI has crashed without saving. Group meetings were forgotten, and entire lab sessions wasted. Nothing has been quite so emotional for young British adults since Bake-off moved to Chanel 4. All of this was done not because the IOP said we had to (yeah it was), but because we lay awake at night longing to determine the properties of high redshift galaxies in order to see what local galaxies might have looked like when they were younger. I therefore dedicate this final testament to my fellow members of the SHREDS team, may our time together not be quickly forgotten by a couple of Jager bombs in Sugar.

This week, the team wrapped up the data collection phase, began writing the report, and started thinking about their results and what the significance of them are. Throughout the project, my main area of investigation has been star formation rates (SFR), and how it changes with redshift and stellar mass in galaxies. There are a number of different methods to determine the SFR in a distant galaxy at our disposal, but throughout this project, we have used simple scaling relations, which have been calibrated in previous work. Essentially, we measure the flux of light at certain wavelengths emitted by each galaxy, then convert this flux to a luminosity using the redshift of that galaxy. Luminosity is a measure of the amount of light emitted by a galaxy and is therefore information about how many stars are forming. However, different wavelengths are absorbed by dust and the interstellar medium (ISM) to a greater or lesser degree, for example, UV light is absorbed by dust, and re-emitted in the far infrared (FIR), whereas radio (tracing mostly supernovae) is not attenuated. Thus, radio gives us an upper limit to SFR, and explains why SFRs calculated using UV are so much lower. I have included one of the firsts plots I made of SFR against redshift (figure 1: left), and as you can see, a lot of progress has been made since week one.

Left (end of first week of labs): Mean SFR against redshift, calculated using Lyman-Alpha luminosity. Right (final plot and result of the project!): SFR evolution with redshift using Lyman-alpha, UV, Radio, and FIR. Some results from the literature of typical SFR (SFR*) are also plotted.

Figure 1 shows that across all our data, observational SFRs tends to increase with redshift. Specifically, we find that Lyman-Alpha SFR increases at a rate of 0.140 ±0.004 solar masses per year per redshift (logged values), and that UV SFR increases at a rate of 0.099 ±0.004 solar masses per year per redshift. It is important to keep in mind with these results (the increase in SFR with redshift) are a consequence of a selection bias. As we observe galaxies at higher redshift with a more or less constant flux limit we end up having a higher and higher luminosity limit, which means that we are only able to observe galaxies with high enough SFRs, which will shift the average SFR up with redshift. The reason that UV SFRs are generally lower than Lyman-Alpha SFRs is most likely due to dust (Lyman-alpha SFRs include a dust correction while UV are observed) and the ionisation efficiency, which means that the UV SFRs themselves can be off if Lyman-alpha emitters are experiencing bursty SFRs.

The results for Smit et al. (2018), Sobral et al. (2012), and Reddy et al. (2018) are results for the SFR function, not (like our results) for the average SFR of their sample. At lower redshift, we are able to observe a larger number of galaxies forming stars at a much lower rate than SFR*, thus our results are lower at this redshift (SFR < SFR*). According to figure 1, we find that at redshifts around 4-5, the flux limit is such that we observe the typical SFR. Above this redshift, our selection bias pushes the observed SFRs up, and hence at higher redshifts we find SFR>SFR*.

SFRD against redshift calculated using Lyman-Alpha and UV. Some results from the literature are also plotted.

In addition to looking at SFR evolution with redshift, I also looked into the star formation rate density (SFRD). SFRD is essentially a measure of the number of stars being produced per unit volume and can be a better indication of how SFR has changed since the early universe. It is important to note that the SFRDs shown in figure 2 were not calculated by simply dividing the SFRs shown in figure 1 by the volume of the corresponding redshift slice. For example, the Lyman-Alpha SFRDs below were calculated by integrating Lyman-Alpha luminosity functions, to give a Lyman-Alpha luminosity density, which we converted to a SFRD, see Sobral et al. (2018) for more detail.

Figure 2 shows that in general, SFRD decreases with redshift. In a similar fashion to figure 1, we see that Lyman-Alpha SFRD and UV SFRD diverge, thought to be due to dust and the ionisation fraction increasing with redshift. This is in good agreement with previous work by Sobral et al. (2018), where a more detailed analysis of the increase in (Lyman-Alpha SFRD/ UV SFRD) with redshift is presented. The results from Gruppioni et al. (2013) are calculated using FIR data, and Rowan-Robertson et al. (2016) data is for starburst galaxies (galaxies forming stars at a very high rate), and hence form upper limits which are consistent with our results. 

For more detailed insight into SFRs, how they were calculated, how they vary with stellar mass in galaxies, and everything else discussed in this blog, look out for our paper: “Properties of high redshift galaxies and their evolution with cosmic time: morphologies, SFR, and AGN”. I hope you have enjoyed reading this blog, but for now, goodbye.

SHREDS out (mic drop).

Week 5

Week 5 is our final lab session of the project, and as such was primarily focussed around finishing any remaining tasks or little odd-jobs that had yet to be done, as well as collating the data from the three sub-projects so the relationships between the different data could be investigated.

Once the data from all three sub-projects had been merged into one large catalogue, it became trivial to examine the trends and relationships between the properties of the objects, simply by plotting the data against each other and analysing the spread of the data points. We were particularly interested in the relationship between the black hole accretion rate from the AGN sub-project and properties from the other sub-projects, as how active a galaxy’s central black hole is can have a big effect on the development and structure of the galaxy, and it is intriguing to discover which properties are either affected by a high accretion rate, or in fact cause it. The property we were most keen to analyse is the stellar mass of the galaxy, as this is often one of the most important characteristics of a galaxy, and is usually an important factor in the structure and development of said galaxy. Figure 1 shows the relation between the log of the galaxies’ masses and accretion rates.

Figure 1: Log-log plot of the black hole accretion vs the stellar mass of the objects.

Unfortunately, it is rather difficult to determine any meaningful trend from this graph, and as such this area of the project may require additional investigation. The other key relationship that needs further analysis is the mass of the galaxies at each redshift, as this is perhaps the most important property of an object and determines much of its other characteristics.

The result of the week comes from the Star Formation Rate sub-project, and is essentially a culmination of their entire project, as it illustrates the relationship between the average SFR and stellar masses of the galaxies at each redshift bin ranging from z=2 to 6 (see Figure 2).

Figure 2: Log-log plots of the average star formation rate against the average stellar mass of all the galaxies at each redshift.

Figure 2 also includes data from other papers for comparison, and the last plot (bottom right) shows a side-by-side comparison of all the different redshift bins. From this last plot it is possible to deduce that the star formation rate increases both with increased mass and higher redshift, as there is generally a positive correlation between log SFR and log mass, and the SFR data from higher redshifts tends to be greater than that at lower redshifts even at roughly equal masses.

SHREDS: Week 4

During the fourth week of our project the group continued working on the three areas of the project, investigating the Star Formation Rates, Sersic profiles and activity of galaxies. In the Active Galactic Nuclei (AGN) section of the project the main area of investigation is the activity of the supermassive black holes in active galaxies.

AGN are highly active supermassive blackholes of over 10⁶ times the mass of our sun, at the centre of galaxies known as active galaxies. Dust and stars in the accretion disc of the AGN is heated as it falls towards the supermassive blackhole, which causes high energy radiation to be emitted in the form of X-rays. Using telescopes such as the Chandra space telescope we can observe the X-rays from the accretion disc of AGN and from the brightness (measured as the luminosity) of these galaxies we can find the accretion rate (used as a measure of the activity) of the AGN. The Black hole accretion rate is calculated using the equation shown below (equation 1) then converted into units of solar masses per year.

Where BHAR is the black hole accretion rate, epsilon is the accretion rate set to 0.1 for our calculations, c is the speed of light and L_bol^AGN is the bolometric luminosity of the AGN.

Due to the energy of the matter falling into the black hole some particles are launched into space from the edge black hole forming huge jets that extend out beyond the active galaxy. These jets of high energy particles emit light of both X-ray and Radio wavelengths and in some AGN form lobe structures as the particles collect in a region towards the end of the jets.

The first step in this part of the project was to identify all AGN in our catalogue, which was done using images from the Chandra Telescope and the VLA (Very Large Array), to then find the brightness of these active galaxies in our catalogue in both X-ray and Radio wavelengths of light. First the background noise was removed from the images and the signal of the active galaxies in our catalogue, which appear in the X-ray and Radio regions, was measured using a program called SExtractor. From the measured signal values the luminosity (a measure of brightness) of all galaxies was calculated, and the blackhole accretion rate found from the X-ray luminosity. Plotting the fraction of the galaxies in our catalogue which emit X-rays or Radio Waves at different redshifts we can see that the fraction of these galaxies which are AGN (or emit Radio or X-rays) peaks at around a redshift of 3 and is more level at higher redshifts, showing there was a peak in the number of AGN during this period of cosmological time (see figure 1). This plot is also the result of the week for this week!

Figure 1: Result of the Week! A plot showing the fraction of galaxies in our catalogue which emit X-rays or Radio Waves or are AGN (emit in both X-rays and Radio waves) at different redshifts.

Figure 2: Radio and X-ray image of SC4K-IA427-47810 showing distinct radio jets.

We then wrote a program using python to extract images of all our AGN in the catalogue from the Chandra and VLA images. This allowed us to obtain some interesting images of the AGN in multiple wavelengths. The result of the week from week 3 was one of the clearer AGN found and the images showed clear radio jets emanating from the supermassive black hole of the galaxy SC4K-IA427-47810 (see figure 2). The aim with these Radio and X-ray images is to create contour plots of the AGN using another python program which will hopefully show the structure of these galaxies more clearly than the noisy images we have at the moment.

With the accretion rate of our AGN calculated we where able to start investigating the relation between the activity of the supermassive black hole in AGN with the Star Formation Rate and Morphology of the galaxy using the data obtained by the other two sub-groups. Collecting the morphologies data and the AGN data we were able to make plots which show the trends in the shape of the galaxy about AGN. Plotting the fraction of AGN in our catalogue with different mean visual morphologies we found visual that there is a peak at a morphology of between 0 and 1, this shows that most AGN appear as point-like and elliptical when viewed in optical wavelengths (see figure 2).

Figure 2: Plot showing the fraction of our catalogue which are AGN, with different visual morphologies. 0 is point-like, 1 is elliptical, 2 is disky, 3 is irregular and -1 means the galaxy did not appear in Hubble images thus could not be visually classified.

The Morphologies sub-project also calculated the size of our galaxies at which different fractions of the total light is emitted, using two different methods. The radii from which 20%, 50% and 80% of the light of our AGN are emitted plotted against the Black Hole Accretion Rate shows there is a slight increase in the light radius with the accretion rate, indicating larger galaxies are likely to have more active AGN (see figure 3). Plotting the Sersic profile of the active galaxies shows that most AGN have a low Sersic profile, which suggests most of the light produced by them is from the central region of the galaxy, where the Supermassive Black hole is located, hence the emission of an active galaxy is dominated by the black hole at it’s centre (see figure 4).

Figure 3: Plot of the AGN activity (measured as the accretion rate) against the 20%, 50% and 80% light radii of the AGN in our catalogue.

Figure 4: Plot of the AGN (measured as the accretion rate) against the Sersic profile (a measure of the morphology) of the galaxies.

As AGN emit strongly in Radio and X-ray wavelengths the Star Formation Rates of AGN cannot be found using their brightness in UV or Radio and IR will underestimate the SFR. So one of our next tasks is to plot the Lyman-α  star formation rate against the black hole accretion rate of our AGN. During the next week we will also begin linking the three sub-projects back together in order to show how galaxies evolve over cosmic time, and how our own galaxy the Milky-Way may have looked earlier in its life.

SHREDS: Week 3

Our third week of labs was setup mostly the same as last week, with each sub-project already split into teams and each team having a solid plan to work with. In the morphologies sub-project, analysis was done by fitting a surface brightness profile to a stacked image of Lyman α emitters. This is useful as most galaxies follow a certain function depending on their Sersic index number which means general insights into these galaxies can be made. The Sersic index was calculated last week as part of the Sersic data which was found using Galfit. The Sersic profile follows the equation: 

Where I(r) is the intensity at a radius r, I₀ is the intensity at the centre of the galaxy, α is the scale length (radius at which the intensity has dropped by e¹) and n is the Sersic index. 

Another profile, which is a special case of the Sersic profile, is De Vaucouleurs profile which is when n=4 and follows the equation:

Where I_e is the surface brightness at R_e, which is the radius that contains half the total luminosity. 

Figure 1 shows an image with different sized masks that were used to find the surface brightness profile of the image. The number of pixels was counted inside the area of each mask.

We then ran a Python script on the stacked image which counted the number of pixels at radii away from the centre of the stacked image. For example, this is the original stacked image and examples of different sized masks in which the number of pixels were counted.

The surface brightness of the stacked image against the radius can now be plot with the new data.

Figure 2 shows a plot of the surface brightness within the mask against the radius aways from the centre of the image.

Clearly, the data mostly fits the exponential fit, which is when n is 1 in equation 1. This suggests that Lyman α emitters tend to be disky galaxies, however elliptical are still common. However, when r>=4 the exponential fit no longer agrees with the data but a combination of the De Vaucouleurs fit and the exponential fit does. This could suggest that there is more than one object that is being detected in the surface brightness profile. 

Further evidence of more than one object being detected can be found by plotting the residuals from a sample fit of the surface brightness against the radius in arcseconds. 

Figure 3 shows a plot of the residuals from a sample fit of the surface brightness against the radius in arcseconds.

When the masks reached the edge of the galaxy it would be expected that the residual noise would be random since they are measuring background noise. However, since a pattern can clearly be seen, this could suggest that more than one physical region is being detected by the surface brightness profile. Possible candidates for these regions are the galaxy itself and a galactic halo. 

Result of the Week:

The result of the week for this week is the combined Radio and X-ray image of one of the galaxies, SC4K-IA427-47810, in our catalogue which was found to have an active galaxy, with a supermassive black hole at its centre. The image shows the jets of high energy particles thrown off the active galactic nucleus (AGN) and the bright region of the accretion disk of the supermassive black hole.

Figure 4 shows this week’s result of the week, a Radio and X-ray image of the AGN SC4K-IA427-47810.

SHREDS: week2

In our second week we split into three pairs with one sub-project each, these are the Morphologies (Michael and Joe), Star Formation Rates (Adam and Oliver) and Active Galactic Nuclei (Cass and Dan) sub-projects. This will allow us to get more work done in the same time span and means we can each specialise on our own projects and get more in-depth research done rather than having to all know everything about each sub-project.

Within the Morphologies sub-project, we first used our mean visual classification values to check how well we each agree with it and each other. As it turns out there was a high correlation between our results and thus no individual adjustments were needed. This having been confirmed, we checked the distribution of our results. Using our mean values for the galaxies not classified as unreviewable, roughly 39% were classified as elliptical, 30% as disky, 20% as point-like and 11% as irregular (non-integer mean classifications were rounded to the nearest integer, the results were however roughly grouped about the integer values, so this seems a fair approximation to use). This would suggest that while irregular galaxies are less common, those with the more regular elliptical or spiral structure are much more commonly found in our universe (at least in the area we were observing).

We then decided to see if there was any obvious relationship between our classifications and the redshifts of each of the galaxies we looked at. As can be seen in our graph comparing the two there was a slight trend towards lower values the further away a galaxy is which is plausible as a younger galaxy is likely to be smaller and more spherical whereas an older one is more likely to have grown and developed into a spiral or irregular shape over its evolution. However, there is a very large spread about the fitted line, which is only slightly sloped so possibly this is not the most reliable of results. This isn’t particularly surprising as in using the brightest 1000 galaxies in our catalogue there are over 2000 galaxies left out of these classifications, all of which will be dimmer than those we did use, which likely skews the results and doesn’t provide a particularly accurate depiction of the full data set.

This graph shows the possible relationship between morphologies (VC Mean) and redshift (z) with the visual classification values defined as in last week’s blog post with the addition if -1 for galaxies we could not classify either because they could not be discerned or because they could not actually be imaged.

This problem can be worked around by instead using the galaxy’s Sersic profiles, which are mathematical functions that describe how the intensity of a galaxies changes with respect to the distance from its centre. The profile fitting program Galfit was used to produce our data and amongst other things it gives values for each galaxy’s half-light radius and its 20%, 50% and 80% light radii (the latter three are calculated with a different fit than the former and thus give slightly different but still useful data). Once we had calculated the means and standard deviations per redshift slice for these results were able to plot them against redshift to produce a more representative picture of how the morphologies of galaxies change with age (higher redshifts mean we are seeing the galaxies earlier in their lifetimes). As can be seen below there is a clear trend in both graphs showing that as a galaxy grows older its radius increases and it becomes larger (however the very large errors on some of the half-light radii in the latter graph should be taken into account and so the first 20%, 50% and 80% graph is probably more reliable). This would seem to confirm the conclusion obtained from the less representative results of our manual visual classifications.

The above graphs show the relationship between the 20%, 50% and 80% light radii and redshift (z), and between the half-light radius and redshift (z) respectively.

Our result of the week however comes from the Star Formation Rates sub-project who produced the graph below which shows how the lyman alpha star formation rate density changes with redshift. We have chosen this result as it agrees with established theory, showing a decrease in star formation rate density as redshift increases. It should however be noted that this is not strictly the real star formation rate density as it overlooks the fact that space is mostly empty and many fainter galaxies, with lower star formation rates, will have been missed by the survey so the real average star formation rate will be lower than that which we have calculated. This problem can be overcome next week by using the distribution of star formation rates and producing and integrating a star formation rate function (number of galaxies per volume per star formation rate).

Our result of the week shows a clear trend that as redshift (z) increases the star formation rate density (SFRD_Lya) decreases, in agreement with established theory.