This week, we have primarily been creating and analysing fake spectra. This involves creating a noise spectrum, with no obvious peaks that stand out. We also create a spectrum that is perfectly smooth aside from one gaussian peak at the wavelength of the ion at which we are looking. We then combine these two spectra together to show what a “perfect” spectrum would look like. By changing the flux on the gaussian curve, we can change the height of its peak. We continue to change the height until the peak blends into the background. This input flux is then our limiting flux. If we have an emission line coming from CR7 with a flux smaller than this limiting flux, we will not notice it, and therefore assume that the ion isn’t present. We repeated this for all the ions of interest that had emission lines in the correct wavelength range.
Using these results, we attempted to mimic the spectra of CR7 and BR3. This meant we could see what ions we could potentially be missing, and account for these in our end results.
Since this week is our penultimate week, we started writing up a presentation and report to summarise our findings. We will show the presentation at a mock conference, designed to emulate a professional conference environment. This means we have to successfully pull our results together, so Project 2 and Project 3 will be coalescing, since they are closely related.
This week has been mainly trying to make our data more accurate. We started off by matching the spectral resolution of KECK (a ground based telescope in Hawaii) and HST. This involved changing the smoothing factor in the data (essentially adding up the flux in adjacent pixels) to make any peaks stand out more and reduce the background noise. However, if this is done to the extreme, then it massively reduces the accuracy and precision. We have taken data from KECK, and altered the smoothing to make it have the same apparent resolution as the Hubble data. This was done on a second galaxy (BR3), to get the same spectral resolution as for Hubble so we can compare the two galaxies directly, as well as seeing what emission lines in BR3 (and CR7 using KECK data) are real, which are from the atmosphere, and which should be seen in CR7, but aren’t due to noise.
We also needed to convert the flux to luminosity, since CR7 is much further away than BR3, and so will have generally lower fluxes, despite the galaxies being similarly luminous.
We have also had a visitor to the university this week. Alyssa Drake works with MUSE at the Observatoire de Lyon in France, and gave us a talk on the “a-muse-ing” project she is doing there. She is looking at what caused the reionisation of the universe a few billion years ago by looking at Lyman-α haloes of galaxies from that era.
This week, most of the University Observational Astrophysics department went to Hull on a National Astronomy Meeting (NAM), so as a team we had to be more independent. To this effect, we had a meeting on Tuesday so that we could all discuss with each other what exactly we were doing, and what our results so far were. We could then give each other feedback on what we could do to improve, and what to do next.
I have been working on measurement improvements, and including errors in my results mainly. I also found several mistakes from last week which I needed to correct. David had sent us a code which included the flux error, contamination from other sources, and another set of data from which I could calculate the spectral resolution (how precise the emission measurements could be). This should remain roughly constant at all wavelengths, and whilst my values for resolution were not massively different, there was still more variation than I would have liked.
Background reading is a crucial part of any research placement, and so I have also read a few papers this week (one of which I have reviewed here). This is something I have not needed to do before in as much depth as I am doing now, so this is very good experience for me to gain. Understanding them takes up a lot of time, since most of them are written for people with more knowledge than me, and it is taking me a while to get used to the style of the language that the authors use. As part of our internship, we each need to present two papers in front of our supervisor and peers, and I have been preparing for that a lot this week as well. I had the aim of presenting it this week as well, but that shall have to wait until next Thursday now due to the NAM conference. However, it also contains lots of information relevant to Umar’s and my project, since it looks at another high redshift Lyman-alpha emitting galaxy, and there are many comparisons I can draw.
Starting a week later than most of the other interns, I was worried about being so far behind. However, I quickly settled in and realised that we were all on the same page.
On my first day, I had an introduction to the various software I would be using over the course of the internship. This software includes Topcat and ds9. Topcat is designed for analysing lots of astronomical data, and ds9 is designed for viewing images. I had never heard of either of these before, but was looking forward to becoming competent in them.
My partner for this project, Umar, had been working on his other project (CLOUDY simulations) over the previous week and so we were both starting at the same place for this project. The idea for this week was to get used to handling the data and become accustomed to the new software. I was looking at the spectrum of CR7, which can tell us what ions are present in the galaxy by looking at what wavelengths of radiation are present. However, there is a lot of background noise when trying to read values off the spectrum, and so the aim is to find the Signal-to-Noise ratio (SNR). This is done by dividing the apparent signal by the background noise. I started off by taking a spectrum along one line of pixels where the galaxy lay, and then another one with no apparent source, and using the second line as my background noise. I soon found out this was not a good choice of method, and so another way had to be found.
The next few days were spent trying to find the standard deviation of the noise by looking at several rows of pixels with no apparent source and finding the standard deviation at each relevant wavelength (for example, a high flux at 1484.4Å indicates NIV). This was then used as the background noise and from there, a more accurate SNR could be calculated for each wavelength.