IRTF: Data Reduction

How the data is reduced.
In short, we use a program developed specifically for reducing spectra from SPEX called SPEXTOOL.  It is written in a commonly used programming language for Astronomers called IDL.  I'm going to skip over a lot of the details, but in SPEXTOOL there are basically two steps:

1. Construct Calibration Frames (for wavelength calibration, flat-field subtraction).
2. Extract the spectral orders.

As you may recall from my last post, we took a series of calibration frames called arcs.  I've extracted a section of the image in my last post as a reminder:

The bright lines in the image (here shown in red) are spectral lines from the Argon arc lamp.  The software automatically finds the spectral features, whose wavelengths are known, and generates a linear regression to determine the wavelength as a function of pixel value (i.e. a wavelength vs. pixel value graph) for each order in the image. Then, this data is cross-correlated with a stored model for where the spectral lines should appear and the pixel differences are recorded a wavecal file.  These data look like the below image:

I've scaled this image logarithmically and added false color to make things easier to see.  (Also note that these may not be the same order or even the same mode as I don't recall what image I used for the sample spectra in the last post, this is just for visualization).  After this is done, we extract the spectral orders using the calibration information we have just obtained.  We used the telescope in AB mode in which the star was moved up and down the slit, so our data is reduced in AB mode.  In terms of data reduction, this means that we feed in two files and use both to construct the spectra.  An example of this extraction is shown in the image below:

Now we have to remove atmospheric features from the data.  This is done with another IDL program distributed with SPEXTOOL called Xtellcor.  Sadly these steps take a long time and I forgot to take screenshots as I was going along so for now if you want to know the full details I'm going to refer you to the paper on Xtellcor. Xtellcor has a built-in model of the throughput of the instrument and a model of an ideal A0V star.  Our first step is to convolve our comparison (A0V) star with the model A0V star to account for any residual wavelength shifts in the data.  Then we fix the continuum levels of the spectra by scaling atmospheric absorption lines of our comparison star to match the built-in model.  Finally we apply these corrections to our target star spectra and we're (almost entirely) done reducing the spectra.  We can still do more (i.e. normalize the spectra), but for now we can at least look at the data and see spectral lines:

What happens next?

From here, where do we proceed?  It's mostly measuring equivalent widths and seeing if anything has changed and of that which has changed, trying to explain what is going on in terms of the eclipse and the hydrodynamics/composition/structure of the disk.  You might be wondering "did we see any CO in the disk" as we originally discussed?  The answer is, not yet.  As expected, CO shouldn't appear until after mid-eclipse.  Did we see anything else interesting in the spectra?  Yes, but we can't talk about it yet as we need additional confirmation.