Wednesday 10Jul19 - MAJOR UPDATE
The dataset I'm going to work on has arrived, and it's time to begin the next steps of my research. This is going to be remarkably challenging. Dr. Micaela Bagley stopped by this afternoon so that we could talk through the concepts I'll need to understand in order to begin.
The first step will of course be to build my data tables. I have the output of Rebecca's code ('CGN7_lines.txt') and the translation file ('GOODSN_plus_translation_all.txt') which I will need to search through. I still need the catalogue file which contains the photometric redshift best guess.
Step two will be begin finding these lines in the spectra. Dr. Bagley and I discussed two possibilities for how to accomplish this:
1. Line Fitting - assume that the spikes Rebecca's code found are at given prominent wavelength. Fit a gaussian curve to that, and fit gaussian curves to nearby spikes that we might see if that were the correct fit. Compare this to Rebecca's data for each possibility. Take the most likely one and posit that as a guess. Be sure to account for pixel size as this will be the ± error for the spikes' locations; G102 has an error of ~ 25 Å, while G141 has about double that at ~ 47 Å. For each obj with lines detected, there is a directory. That directory contains a file ('filename_flat.txt') which shows the 1D spectrum with the continuum removed. This is what I'll need to use to build and fit a model of the data. Dr. Bagley has done this in the past, and has agreed to send me the code she used so that I may use it as a model.
2. Convolution - step each potential profile past the actual spectrum and see what the best fit is. For this I need Rebecca's data ('CGN&_lines.txt') & the wavelengths from GRISM ('G102_G141_lines.png'). For reference, the strongest lines in GRISM data usually are
If more than one line can be detected, it is vastly more helpful in terms of identifying the redshift of the galaxy. The ratio of the observed wavelength and at rest wavelength is also crucial. Let λο ≡ the wavelength we observe a line at in space, λr ≡ the wavelength we see at rest (here on Earth), and z ≡ the redshift (how much it has changed as a result of being far away). Then
λο1 / λr1 = z + 1 = λο2 / λr2
We know that a spike is caused by Oxygen III at 4363 Å and Oxygen II at 3727 Å on Earth. Imagine we observed a spike in a distant galaxy at λο = 17452 Å and guess that it is Oxygen III at z = 3. This is because the formula for observed vs rest wavelengths is λο = (z + 1)λr. Now we can guess that 3 + 1 = λOII / 3727 Å from our ratio, and therefore we should see a characteristic spike of Oxygen II at the wavelength λOIIz = 14908 Å. If we see that spike, we can know that we guessed well.
It begins to be clear that if we can only detect one spike in a spectrum, we may not be able to guess particularly well as to the redshift of the galaxy. This is a distinct possibility. But we may be able to rule out what it is not through this method.
The first step will of course be to build my data tables. I have the output of Rebecca's code ('CGN7_lines.txt') and the translation file ('GOODSN_plus_translation_all.txt') which I will need to search through. I still need the catalogue file which contains the photometric redshift best guess.
Step two will be begin finding these lines in the spectra. Dr. Bagley and I discussed two possibilities for how to accomplish this:
1. Line Fitting - assume that the spikes Rebecca's code found are at given prominent wavelength. Fit a gaussian curve to that, and fit gaussian curves to nearby spikes that we might see if that were the correct fit. Compare this to Rebecca's data for each possibility. Take the most likely one and posit that as a guess. Be sure to account for pixel size as this will be the ± error for the spikes' locations; G102 has an error of ~ 25 Å, while G141 has about double that at ~ 47 Å. For each obj with lines detected, there is a directory. That directory contains a file ('filename_flat.txt') which shows the 1D spectrum with the continuum removed. This is what I'll need to use to build and fit a model of the data. Dr. Bagley has done this in the past, and has agreed to send me the code she used so that I may use it as a model.
2. Convolution - step each potential profile past the actual spectrum and see what the best fit is. For this I need Rebecca's data ('CGN&_lines.txt') & the wavelengths from GRISM ('G102_G141_lines.png'). For reference, the strongest lines in GRISM data usually are
- HeI 10830Å
- SIII 9069, 9532Å
- SII 6725Å (blended doublet)
- Ha 6563Å
- OIII 4959,5007Å(mostly blended)
- OII 3727Å (blended)
This data has a lot of noise. It is not resolved well, and so the first method will be more effective, but convolution can give some very sure guesses when it works. There is an argument to be made for using both and comparing them with the photometric redshift guess. Perhaps "best 2/3" might be the strongest guess we can make.
If more than one line can be detected, it is vastly more helpful in terms of identifying the redshift of the galaxy. The ratio of the observed wavelength and at rest wavelength is also crucial. Let λο ≡ the wavelength we observe a line at in space, λr ≡ the wavelength we see at rest (here on Earth), and z ≡ the redshift (how much it has changed as a result of being far away). Then
λο1 / λr1 = z + 1 = λο2 / λr2
We know that a spike is caused by Oxygen III at 4363 Å and Oxygen II at 3727 Å on Earth. Imagine we observed a spike in a distant galaxy at λο = 17452 Å and guess that it is Oxygen III at z = 3. This is because the formula for observed vs rest wavelengths is λο = (z + 1)λr. Now we can guess that 3 + 1 = λOII / 3727 Å from our ratio, and therefore we should see a characteristic spike of Oxygen II at the wavelength λOIIz = 14908 Å. If we see that spike, we can know that we guessed well.
It begins to be clear that if we can only detect one spike in a spectrum, we may not be able to guess particularly well as to the redshift of the galaxy. This is a distinct possibility. But we may be able to rule out what it is not through this method.
Comments
Post a Comment