Mapping the Ionized Component of High-Velocity Cloud Complex A

Allison Duncan
Rose-Hulman Institute of Technology

Astrophysics REU - Summer 2005
University of Wisconsin-Madison
Advisor: Matt Haffner

Part I: Fitting and Subtraction

In the first stage of my work this summer, I performed fitting and subtraction of the spectra data obtained from WHAM using a GUI program called whamspect2.

I began by using a set of four blocks (b290, b289, b237, and b236) to calibrate the atmospheric template. These blocks were off the main Complex A structure and had very little emission in HI. I adjusted two of the Gaussian components of the template to best fit the atmospheric lines in the four blocks. This adjusted template was then used to subtract atmospheric emission.

Then I began the process of subtracting and fitting all of the data. For each block, an averaged spectrum was created and fit. The atmospheric level and fit parameters from the averaged spectrum were then used as a basis for fitting all of the pointings within the block. Since the observations in each block are taken within a short time of each other, the atmospheric parameters should change very little throughout the block.

Each fit included a constant polynomial background, atmospheric template, and a Gaussian fit to the geocoronal ghost. These components were subtracted from the spectra. In some cases, another Gaussian was fit to the lower-velocity Galactic emission at the right end of the spectra, when it was needed to set the background level. An example of this appears in the spectrum on the right. The subtracted spectra and parameters of all Gaussian components were saved for each pointing.

Fitting the block 236 averaged spectrum

Fitting a pointing within block 236

In the screenshots above, the green component is the atmospheric template, the large orange Gaussian is the geocoronal ghost, and the constant dashed blue line is the background. The component present at the very right of the pointing spectra is the Galactic emission previously mentioned.

The preliminary results from this fitting procedure showed some structure of H-alpha corresponding to the regions of HI in Complex A. An IDL program was used to map the integrated intensity of each pointing, and compare an overlay of the two spectra.

Before I continued to fit the remainder of the blocks, we decided to add another aspect to the fitting process. Since the H-alpha intensities and signal to noise ratio of the data were very low, I included another Gaussian component to fit the H-alpha emission. The Gaussian was fit within a velocity range of 0 km/s to 100 km/s in raw velocity, where H-alpha emission was expected, and fixed with a width of 30 km/s. We chose to restrict the width to have more control over the fitting process. A value of 30 km/s was chosen because it had been previously found to be a typical line width of H-alpha emission, including two previous observations made of Complex A. Those observations had widths of 24 km/s and 34 km/s (Tufte, et al. 1998).

Fitting a pointing with the additional H-alpha Gaussian

I did not include an H-alpha Gaussian in all fits, but only where an added Gaussian had positive area and improved the chi-squared of the fit. I used this fitting procedure on all of the data, and created a short IDL program to extract the fitting parameters of the H-alpha Gaussian components.

I was then able to map the integrated H-alpha emission using the subtracted H-alpha. The H-alpha map appears on the following page with the rest of my results.

Part II: Automated Fitting

In the next stage of my research, I used two automated fitting programs to fit Gaussian components to the spectra from both H-alpha and HI data. The first program used the subtracted H-alpha spectra and corresponding averaged spectra for the HI pointings within the H-alpha pointing beam. It attempted to fit Gaussian components to both spectra within a specified range, and only for pointings with greater than a specified H-alpha intensity integrated across the range. The second program fit Gaussians to each individual HI pointing, again over a specified range of velocities.

For both automated programs, bad fits were rejected based on certain criteria. These criteria included:

Negative area
Chi-squared greater than 10
Mean outside of fitting range
Width less than 10 km/s or greater than 100 km/s

I ran these programs several times with variations in intensity level in an effort to determine the best cutoff value. I took into consideration the number of resulting good fits out of the fits attempted, the averaged chi-squared value of the fits, and the number of outliers. The value which I decided to use is 0.03 Rayleighs (R).

Using the fit parameters output from these programs, I compared the results with my earlier H-alpha component fits. The following plots show the mean and area of the good fits from my initial component fitting (HAFIT) and the automated component fitting.

The H-alpha components from the two methods show close correlation in their means and more scatter in the areas, which is likely due to the difference in widths. The automated fitting program had a variable width between 10 and 100 km/s, whereas the H-alpha Gaussian was fixed at 30 km/s in width.

The parameters of the H-alpha and HI components can also be compared. These plots and further discussion can be found in the next section.

Continue to my results and conclusions