So for the past (almost) month, we have all taken a holiday, but we are back this week for our final official week of the internship! In order to remind ourselves, and you, about what we’ve been up before our break, we have written this blog post, explaining what we got up to for the final week before our break, and what we have lying ahead of us this week.
After several weeks of trials and hard work, filled with moments of doubts and hope, we have finally managed to determine the optimal EW and Sigma cuts for each filter, which allow us to select line emitters out of all the sources. The next step is now to specifically select Lyman-alpha emitters out of all the line emitters. To do so, we need to apply two additional cuts: a photoz cut, and a colour cut: these cuts will allow us to eliminate sources that are either too blue or too red to be Lyman-alpha emitters. This time again, we want to determine which combination of cut is the most efficient.
These are two further aspects that are required to be done separately, as we don’t want the intersection between the sources selected with photometric redshift and colour cuts, instead we want the sum of them, as they select emitters in different ways and have different data available. The photometric redshift selection is done in a similar way to the spectroscopic redshift. There are more sources that have a photometric redshift available than spectroscopic, however, these redshifts are less accurate. Looking at figure 1 you can get an idea of the difference, and why a wider range of photometric redshift is needed.
We could of course test each combination individually. However, this would take a lot of time, as we would have to try a lot of combinations to get an accurate result. Therefore, we decided to create another algorithm, based on the one we used previously, in order to automate the process, and then plot the effectiveness of each cut/photoz width, to find which is best.
When running the code for the photometric redshift, we found that for a lot of filters we didn’t have enough data to draw conclusions on an appropriate width, so we decided that an average of the ones available would be the way to proceed, as the spread of photometric redshifts doesn’t vary wildly per filter, and an average would produce good results.
We are currently working on deciding appropriate colour cuts per filter, to make sure that we eliminate sources that are either too red or too blue to be able to be a Lyman Alpha emitter.
For example, the two colour cuts applied for the IA427 filter are (u > u3σ ∨ u − B > i1) & (B − r+ < i2), where 3σ is the estimated depth of broad-band data. Our objective is to determine the best combination of i1 and i2. We use an algorithm similar to the one used for the EW and Sigma cuts. The conditions tested are different, but the process is similar: for each combination, we calculate the corresponding purity, the completeness and the effectiveness, before calculating the relative effectiveness. Finally, for each filter we obtain a catalogue that lists an important number of combinations of cuts and their corresponding characteristics. We are then able to get a 3D plot for each filter on Topcat.
So now that we’ve done all this…. What’s next?
- Finish deciding colour cuts
- Create a code that creates a subset of the original catalogue that has all data points, the subset having only the sources that are selected with our methods (spectroscopic redshift, photometric redshift width, and colour cuts)
- Using this new subset/catalogue, visually inspect all the sources selected and eliminate any noise that may have managed to make its way into the final selection.
– Amaia, Louis, and Emily