What am I doing here? (Part 1)
Ok, so this isn’t really the post about the curiously high metals content (that’s going to be a future part). Instead, this is the post about what I do, what spectra are, how we reduce them, and what we find out from them. If the post is still short enough, I’ll add in a bit about diffuse bands, and our results there so far (they’re published, so I’m quite safe in doing that).
First, I’m going to start you with a picture (stolen from Sara Ellison, my advisor) which shows what it is that I’m actually looking at. The basic idea is that, while we can see galaxies very far away (redshifts of 6 or even higher), at any significant distance (even redshift 0.5, for example) we see a very biased sample of galaxies - we can see bright galaxies, but not faint galaxies. The obvious result of this is that we end up with a biased view of what galaxies were like in the early universe, especially since faint galaxies are much more numerous than bright galaxies.
Now, with Damped Lyman-α (DLA) systems (the galaxy in the figure above, named because the Lyman-α line of neutral hydrogen (H I) is strong enough to display damping wings, more on that later), we don’t have that problem. These systems aren’t selected by brightness, by redshift, or by anything other than random chance - whether or not they happen to be on a sightline that intersects with a quasar. As such, we can see all types of galaxies, bright and faint, with this technique.
So, we aim the telescope at a quasar, send the light through a slit, pass it through a prism (actually, generally either a grating or a mirrored grating (called a “grism”), and record a spectrum in which wavelength is our x co-ordinate, and position on the sky (in one dimension, the one you happened to set when deciding how to orient the telescope) is your y co-ordinate. Well, you’re done, right? Sadly, not yet. You first have to reduce the spectrum, and then check to see if there is in fact a DLA there (since nature was not kind enough to provide us with a list).
Now, reducing the data (going from the raw data as the telescope recorded it to the processed data from which you can actually extract something useful) is different at every telescope (and every part of the spectrum), but I’ll walk you through a generic reduction of optical (near UV, visible light, and near IR) spectra (data with wavelength as one axis, as opposed to imaging, which actually takes a picture of the sky). Later on, I’ll go through the reduction procedure for low-frequency radio data. Note that this will be a very brief walkthrough, skip a few steps, and assume that everything works perfectly - things can, in reality, get a lot worse than this.
Now, the CCDs used in telescopes are very good compared to the CCDs in digital cameras. They’re more sensitive to low light, more linear, cooled (to avoid thermal jitter and like problems), etc. But they still have a basic amount of error that you can’t get rid of, including random noise in each pixel. Now, the problem with this is that, when your pixels don’t have very many counts (and, for a good astronomical CCD, the ratio isn’t too far off 1 count = 1 photon), your noise is always in the same direction - up (and easy to mistake for signal), since a pixel with only 1-2 counts in it can’t really read low. The solution to this is to run a current through all of your pixels, to set their zero value (the value they would (in theory) read out even with no light at all) to about 100 counts. That way, reading low won’t really matter. This is called the “bias voltage”, and to figure out what it is you take a bunch of short exposures with the shutter closed, average them together, and subtract them away from each data frame.
The next problem you face is that different pixels respond slightly differently to the light that hits them. Some will read a bit higher than others, and you want to be able to deal with that. Some will also respond differently to different wavelengths, and for spectra you want to deal with that too. So you take a few exposures of a bright, uniform source (like the sky at twilight, or the inside of the dome, or special lamps designed for the purpose), and then divide every frame by your “flatfields” (as these frames are called). Of course, this, well, flattens out the field, and gets rid of the per-pixel effects.
You then extract out your spectrum, combining as much light as you can together (while hopefully subtracting off any interference, including emission and absorption from the night sky, which can be really nasty). Now you have a rough idea of which wavelength is there (based on the settings you gave the telescope), but a rough idea isn’t good enough. So you take some more exposures, this time of special lamps that excite specific elements that give off light at known frequencies (like neon lamps, but usually mercury, copper, titanium, argon (etc.) instead of neon). Then you calibrate the wavelength scale on these images, and apply the calibration to your own images. Finally, you combine all the exposures you took together (since you’ve removed the noise on each one individually, this has the net effect of increasing your signal-to-noise), and you’re left with something like this (assuming you’re looking for DLAs):
This is, actually, a picture of the quasar spectrum the does have a DLA in it (right around 4000 Angstroms, can you spot it?). In fact, this is a very special DLA, for reasons that will become clear in a later part. Now, the DLA itself looks very much like this:
In this case, the black histogram is the same spectrum as the figure above, while the dashed line is a fit to a damped profile with H I column density of 2.0×1020 cm-2. You can see the broad shallow “wings” on either side of the central absorption. This is all based on the so-called “curve of growth” that relates line width and column density. In the low-density or “linear” region, column density and equivalent width are directly related, by
Column Density = 1.13×1020 × EW / (λ2 × f)
where EW is the equivalent width, f the oscillator strength (how relatively strong that line is), and λ the wavelength of the line. As lines become stronger (around a column density of 1×1015 cm-2, they enter the “saturated” regime, where column density is only related to the logarithm of EW (which is what you measure), and column density also depends on internal velocity (which is impossible to measure), making the actual column density impossible to determine. At a very high column density (around 1×1020 cm-2), the damping wings start to appear, and the column density is now based on the square root of the EW (easier to find), and velocity disappears again. This generally means that you can measure either very weak or very strong lines, but nothing in the middle.
So now that we’ve found our DLA (and now that you all know what the “damped” part of damped Lyman-α system means), you can do something with it. But, since this has turned out to be another long post, I’ll discuss that in the next part, hopefully in a couple of days.