Welcome to ATMs Astrophysics Group. We’re not the money type, and surely not the richest, but instead we consider ‘Are They Metal-poor Stars?’ 

During this 10 week project we are going to study Population III stars, a hypothetical population of incredibly massive, luminous and hot stars with no metal content. We will bring you along with us during the process and each week we will post our most relevant discoveries.  

Group members and roles: 

Uma Cladellas Sanjuan – Coordinator 

Adam Howes – Administrator  

Georgia Lowes – Communications Lead 

Katherine Ormerod – Theory Lead  



Week 1 mostly began with some organisation. As there are only four of us in the group this year, we decided it was best to only allocate major roles and split the rest of the work between us as and when needed. This means we currently have our PI (Uma), Administrator (Adam), Communication lead (Georgia) and Theory lead (Katherine).  

As our group name suggests, we will be focusing on metal-poor stars. Ideally, we are hoping to continue the search for Population III stars – the first generation of stars which formed in our universe. We are particularly interested in the work of the Pristine survey (Starkenburg et al. 2017). In the past 3 years, the survey have published 11 papers detailing their search for metal-poor stars, a search which has turned out to be very fruitful! However, we are not planning on starting our own search from scratch. Our project has two natural predecessors: LAMPSS and WARP from the first generation astrophysics group project here in Lancaster. We noted that both projects used similar techniques to Pristine, using CaHK absorption lines to estimate the metallicity of a star. This made us wonder if we could progress the work of our predecessors, much like Pristine have refined their data across advancing papers. LAMPPS and WARP both succeeded in finding a selection of metal-poor candidates, restricting Population III candidates to those with [Fe/H]<-3. Under these criteria, WARP propose 7 candidates and LAMPSS propose another 16. We wish to continue studying these 23 candidates.  

We did note that LAMPSS and WARP only removed galaxies, meaning it is possible that more contaminants remain. In particular, we would like to remove white dwarfs and variable stars. White dwarfs have a very low CaHK absorption, meaning they give a false reading as a metal-poor main sequence star. Variable stars can also contaminate plots such as WARP’s heat map. Taking a measurement on only one occasion means that variable stars appear with a fixed magnitude, seeming like main sequence stars. WARP also note requesting time on the VLT to observe spectra for their candidates. If these spectra are available, they would be of a lot of use as they would allow us to determine an accurate value of metallicity. Spectra would also allow us to consider the formation route of our stars and whether they are carbon-rich, carbon-poor or carbon-normal.  

Next week we will begin taking our project forward. We will begin by considering how best to remove contaminants and analyse the candidates as well as considering how best to utilise other star surveys in order to identify variable stars and find spectra. 




Welcome to ATMs again! We have finally overcome the first three weeks of the project. We have gone through many moments of uncertainty, doubts and excessive thinking about what and how we were going to approach the experiment, nevertheless, these have been necessary, helpful and worth it because we are now on track!   

Georgia’s and Adam’s role this week has been determining. They have opened the data from WARP and LAMPSS in TOPCAT and worked with a star at a time. They have converted each magnitude of a specific filter into a wavelength (center of the filter) and a flux density by using the magnitudes from the AB system shown below. Tables 1 and 2 show the filters used by WARP and LAMPSS. Then they have plotted the wavelength against the flux density and tried to discard the stars that were far from a followed pattern by the others. In Figure 1, a plot for the WARP-20 star is showed. Finally, they have used the Planck curve to estimate what is the temperature needed for them to match and to determine whether they look like a Black Body.

Filter nameWavelengthWavelength (eff)FWHM (A)
Table 1. WARP filters
Filter nameWavelengthWavelength (eff)
Table 2. LAMPSS filters
Figure 1. Plot of effective wavelength vs flux density of WARP-20

 This plot will look more like a typical spectrum when we add in the rest of the filters. Similar plots can be found in files for the rest of the WARP candidates. (WARP-1 and WARP-9 look distinctly different, this could be an early indicator they aren’t stars but further analysis is required).  

After that they have calculated the errors by using the following equation in order to understand how accurate the effective wavelength calculation is. It corresponds to the FWHM.   

 Katherine has focused on the theory behind our data. She has first looked at what the other papers have written about and how best to utilize other star surveys in order to identify variable stars and find spectra. We concluded that discarding variable stars might not be realistic.  

WARP/LAMPSS don’t match up, slightly different things. Might be able to match up.  

Then she tried to join the tables for both WARP and LAMPSS together in Topcat using the concatenate tables feature, however this will need to be improved upon to make the data easier to work with and to rename all candidates with an ATM ID, and will be redone, likely in QtiPlot.  

Began visual analysis of the candidates from both the LAMPSS and WARP catalogues with a metallicity of -3 or below, using  

For each candidate, the image was saved as a .jpeg and .fits file to enable viewing as a .jpeg and enabling it to be used in the LaTeX file, and for further viewing in ds9 as a .fits file. The dataset used was HST-ACS Mosaic. The image size for every image is 15″ x 15″. Figure 3 shows WARP-1. Some candidates did not have an image (WARP-8, -15, -60) due to there being no overlap with data found for the requested cutout.  

Figure 3. Image of WARP-1

As for me, I have analyzed the data collected from the LAMPSS group in the TOPCAT Software. I have selected the stars with metallicities of -3 or less ([Fe/H] < or = -3) and copied them in a new table in QtiPlot. There are a total of 16 objects, 15 of them with metallicity of -3 and 1 with metallicity of -5.  

I wanted to focus my study on the temperature of these stars and use this quality to further determine the possibility of them being Population III stars. One way to classify stars is by the ratio of the flux at one wavelength to the flux at another wavelength, this ratio is a strong function of temperature. Therefore, by measuring the ratio of fluxes we can learn about the temperature of the star. We know from theory that hot objects have more B band flux than V band flux. This means that the B-V color index will be negative. This is verified in the flux-magnitude equation below where I have used 1 as B magnitude and 2 as V magnitude.  

m_1-M_2 = -2.5log_{10} \frac{f_1}{f_2} 

Because stars are not ideal black bodies, there exists a modified version of the temperature index relationship, determined empirically that is valid for 4,000 < T < 10,000, given as:  

T= \frac{8540}{(B-V) +0.865} 

This is next used to determine the Temperature of LAMPSS selected stars in Qtiplot. Conclusions have not been extracted yet.  

Then I have written down the introduction in the Overleaf document. 

During the incoming week lab’s session, I will continue working on colors and try to extract some conclusions from the data obtained. WARP/LAMPSS both use color to estimate surface temperature and spectra type using the bands, their plots will be recovered and then improved. It is important to note that we need a metal specific filter for the plots (metal-broadband) vs (broadband-broadband) for color-color plots. I will also write down the background theory into the Overleaf shared document.  

Georgia will continue working with the spectra and will try to get some useful information from it.  

Katherine will obtain pictures for candidates -3 and below and will continue reading past papers. 

Adam will work on the programming part with python and will start writing stuff into overleaf.  

Another task will be looking at the Pristine papers and comparing/checking with Sloan Digital Sky Survey. We can also fit blackbody curve in QTI plot (science note: will be differences in T/flux depending on distances/type etc.). If it is not a blackbody, might not be a star. We want to obtain main science results for all sources: spectral types, temperature, distance and metallicity. 

Until next week!  



Welcome back to ATMs blog! This week has been quite a productive week for the group. Here’s what we’ve been up to this week.  

Our first task this week was to cut our sample down into stars and galaxies, so we know which sources to focus our analysis on in later weeks of the project. We start with a total of 81 sources determined to a metallicity of [Fe/H] <-3 by the WARP (65 sources; Jenkins et al, 2019), and the LAMPSS (16 sources; Worrell et al., 2019) groups. We inspected the SED spectrums produced by Santos et al. (2020), and with the guidance of Dr. Sobral to interpret these spectra, we classified the sources into three groups – stars, galaxies, and uncertain sources. Stars were classified as sources whose observed luminosities follow a black body curve well, galaxies were classified as sources whose observed luminosities don’t follow a black body curve, and those which we unable to classify from the nature of their observed luminosities, for example by only partially following a black body curve. From this cut, we were able to remove 30 sources as galaxies. For the remaining 51 sources, we performed a visual cut using images from the HST-ASC Mosaic, to further reduce the sample and check the initial spectral cut. This cut was made by looking at each image for the source and determining its shape. Those which are circular are classed as stars, whilst those that are smeared are galaxies.  Not all the samples had images and we planning to finish visual checks using images in other wavelengths early next week. 

Figure 1. Example of the SED used in the first step of the cut, as by Santos et al.(2020). We are interested in the red dots on this plot.

Based on the observed data and models from the Santos paper, we have also started plotting initial spectrum for the different potential sources. This is done by plotting the flux density observed in each for each filter (as described in table 2 Santos et al., 2020), against the effective wavelengths for each of the different filters. These were compared to a range of test blackbody curves, and all the curves normalised to compare their shape, as shown in figure 3. 

Figure 3. The plotted spectrum for the flux density of the WARP-1 source, with four sample blackbody curves plotted over the graph.

As observed from figure 3, the models are all similar to each other, despite the range of temperatures, and to improve upon this, we plan to improve out models of fitting a blackbody curve to the data over the coming week. 

We are also able to build a spectrum using the techniques described in Santos et al. (2020), and again try to fit a blackbody curve, however, the same issues arise as with the filter and central wavelength method when constructing the curves. 

Elsewhere, we have also started drafting the background theory section of our report, and have started updating the methods section of report. 

Looking forward to the next week’s work, we are planning to complete the tasks already discussed in this blog post, before starting to determine a variety of different properties of the sources we’ve classified as stars. 



Welcome back to Week 5 of the ATMs blog! This week was quite an interesting week for us as we moved forwards with finding the properties of our candidate stars.  

We completed the visual cuts and were left with 40 remaining candidates, which we shared between us, giving us ten candidates each. We decided our method for calculating the temperature would be to use Wein’s Law, which is given by:

𝜆max = 0.0029Km / T

𝜆max = wavelength of maximum emission in meters 

T = temperature of object in Kelvin 

In order to find the peak wavelength, we calculated our temperatures by converting each filter’s magnitude into flux and calibrating blackbody curves. We then used the effective wavelength of the peak filter as our peak wavelength to give us a temperature for each candidate.  

As a group, we found that our temperatures were very cool compared to the temperatures given by the previous groups, LAMPSS and WARP. For example, using our method, we obtained a temperature of 2395K for WARP-63, compared to the WARP value of 4032K. 

We were concerned with our results at first but as Dr Sobral said, finding what you expect is not science! We then used the cooler temperatures to classify our stars, but this only gave us 3 candidates on the main sequence, in M spectral class. This led to the question – what are our other candidates? This was something we chose to investigate moving into the next week, focussing on dwarf stars.  

 I also figured out how to create a distance plot using Topcat, which will be created after we are confident with our temperatures,. Apart from the temperatures, Adam continued to add to his code, and we added more to our report.  

I also gave the group a tour of Tesco when I went to work, but I unfortunately couldn’t find any metal poor stars in stock here (or Little Moons). I also didn’t get any for my birthday which was a shame! 

Moving on to next week, we decided to focus on determining the distances to our candidates, using the luminosities and absolute magnitudes to calculate distance. We also decided to research white and brown dwarfs, to see whether we could determine if our candidate stars are indeed dwarfs.  

See you next week!  




Welcome to another week of ATMS! Remember our last update when we thought we had discovered 40 dwarf starts and potentially disproved the entire method of using CaHK absorption lines to determine metallicity? Yeah, we take that back… 

This week was slightly chaotic. After determining our temperatures last week, we focused on distances this week. These came out a lot lower than we expected. For example, we determined ATMS-70 (LAMPSS_627) to lie a distance of 240.3pc away from Earth whilst LAMPSS calculated its distance to be 27kpc. This raised a lot of questions – were our distances just smaller because we’d proven the stars to be cooler than initially thought? A closer, less luminous star can have the same apparent magnitude as a more distant and more luminous star so this did seem fairly plausible to use. Either that, or perhaps our temperatures and spectral classes were wrong? 

 I am afraid to tell you, dear reader, it was the latter. Our distances were calculated using a fairly standard equation: 

d = 10^[(m+5−M)/5]

In this equation, we took M as a constant, using magnitude values from This allowed us to break down our sample into spectral subtypes, which was something we had hoped to do from the start to improve upon LAMPSS and WARP data. However, since we were using the wrong subclasses, our distances seemed far too small!  

Upon consultation with Dr Sobral, we realised the error of our ways. Flux density, Planck’s Law and Wien’s Law all take two forms – one based on wavelength and another based on frequency. Without realising there was a discrepancy between their results, we consistently mixed and matched between the methods. This meant that, while we thought we had blackbody curves that supported our temperatures, they only did so as a result of plotting conflicting data. Thankfully, our distance calculations were still appropriate, which we confirmed by seeing if it gave us the expected distance to the sun.  

Fear not though, we have since recalculated our temperatures. We now have two methods of doing so. Our primary method is borrowed from LAMPSS and we use the formula which they derived from Pollux data: 

T = 6800−2930(g−i)+780(g−i)^2+203(g−i)^3−125(g−i)^4+15.4(g−i)^5

This is a colour-based method, with g and i representing the magnitudes of the g and i filters respectively. Our data is actually from g+ and i+ filters. This means they lie at a slightly different effective wavelength to g and i but this difference is very slight, meaning we are still able to use the relation. Unfortunately, not all of our sources have data available in these filters. For these stars, we used (a now correct form of…) Wien’s Law. In order to verify the accuracy of this, we also determined Wien’s Law temperatures for stars with g and i filter data present. In some cases, the temperatures agreed extremely well but in others they massively disagreed. This is one of our final points of investigation and we hope to figure out what is causing this divergence.  

Next week is our last official lab week, we have lots left to finalise. We now have temperatures, distances, absolute magnitudes and luminosities for each star. This will allow us to plot a HR diagram, as well as a sort of sky map for all our candidates. This should also allow us to determine most physical properties, should the need arise. Of course, this is an investigation on metal poor stars and we can’t forget their most important characteristic. Hence, we will conclude our research by determining our remaining candidates’ metallicity – hopefully finding lots of metal-poor candidates in the process!  

Stay posted to see how many we find!