Chapter 6: Preparing the Profile
A spectral profile is a plot of wavelength versus intensity. We use a program called Visual Spec (Vspec) to produce the profile from the spectral image. There are three steps that must be done to the profile before it can be analyzed. These are (1) wavelength calibration, (2) radiometric calibration, and (3) normalization. This chapter tells how to make these corrections.
6.1
Visual Spec and Wavelength Calibration
Visual Spec (Vspec) is a freeware program for the analysis of astronomical spectral images. Vspec was designed by French amateur astronomer Valerie Desnoux as a means of offering other amateurs a user-friendly and inexpensive way of analyzing spectra. In the past, spectral analysis has been limited to overly advanced or overly expensive packages. This program is capable of handling most any task an amateur or undergraduate researcher would want to perform. Vspec is designed for Windows and comes with a reference manual in Word document format. Vspec is downloadable at http://valerie.desnoux.free.fr/vspec/.
A spectral image is the image recorded by our CCD chip, and is a three dimensional array where pixels are assigned x, y, and intensity coordinates. A spectral profile is a plot of the wavelength versus intensity. In our spectral images, the wavelength changes along the x-axis. Therefore, the spectral profile is a plot of the x-axis versus the intensity value. Because we are creating a two-dimensional array from a three-dimensional array, we lose the information in the y-axis. If the y-axis information is important, as is the case with extended objects, we may have to make multiple profiles at different ranges of y.
The process
of extracting a spectral profile begins with a command similar to the addition
of column values in Iris’s “>l_add” command. The command in Vspec is “Object Binning,” and is accessed through
this icon, 
. This command adds the pixel values along the
columns, increasing the overall intensity and increasing the signal to noise
ratio. Vspec will automatically
identify the region containing the object spectrum and bin that region only,
ignoring the majority of the background. Vspec will not remove the background sky spectrum, however; this is done
through Iris (see Chapter 5). If the
image has been processed with Iris prior to Vspec, then each row is identical
and it does not matter where Vspec bins. It is also possible to manually define the region for binning by
creating a box around the area with the right mouse button. This feature is useful for extended objects.
Object binning generates a profile document with the spectrum’s profile. A profile document can hold numerous spectral profiles. Vspec refers to an individual profile within the document as a series. Other series are generated when you add a reference series or perform operations on the original intensity series, such as filters or mathematical functions. When a spectral series is calibrated for wavelength, all the series take on the same calibration. A profile document can save only four of these series; the rest of them are temporary and lost when the profile document is closed. The four basic series are “intensity, reference 1, reference 2, and normalized.”
It is possible to calibrate a series directly from its own spectral lines. Some spectral lines are easily recognizable, such as hydrogen lines. However, calibrating from the spectrum itself means you cannot measure Doppler shifts. I believe it is always best to use the reference spectrum taken with the gas discharge lamp. Also, it is important that the reference spectrum has increasing wavelength from left to right. (Recall that the SGS produces spectra ordered from red to blue.) If the image has not been reversed, Vspec cannot perform radiometric corrections.
Open the reference spectrum and object bin it. Under the “Spectrometry” menu, select “Calibrate.” This brings up the calibration tool bar. Figure 6.1 shows the Vspec layout with the calibration tool bar and the profile of an argon reference spectrum.
Figure
6.1
In order to
calibrate, it is necessary to identify two reference lines. Appendix C has the identification of the
intense lines from an argon gas discharge lamp. Type in the wavelengths (in angstroms) that you would like to use
for calibration into the spaces titled raie 1 and raie 2, (“raie” is French for
line). Once these have been entered,
use the curser to isolate the first spectral line. Make sure the dotted box contains at least 95 percent of the
spectral line. Click the “line 1” icon,
. Repeat for the second line. Vspec assumes the spectrum is linear and
calculates the dispersion in angstroms per pixel. Make sure this dispersion is
positive and matches up with the theoretical dispersion of the
spectrometer. (The low-resolution
grating has a dispersion of 4.3 angstroms per pixel and the high-resolution
dispersion is 1.07 angstroms per pixel.) Save this profile document. We
will use this calibrated series to calibrate other profiles by cutting and
pasting, rather than going through the entire calibration procedure again.
Next, open the spectral image to be examined and object bin it. Vspec will ask you if you would like to replace the series that is currently open; in this case, it is the calibration series. By selecting “no”, another profile document opens with the intensity series of the spectral image. Return to the calibration lamp profile, select, and copy the series. Copying and pasting are done through the usual Windows commands. Select the spectrum’s profile and paste the calibration series into the document. This calibrates the spectrum’s series. In order to save the calibration series in this profile you must replace it into one of the basic series. When two or more series are present in a profile document, you can select a series to be the active series from the window on the right hand side of the tool bar. Under the “Edit” menu, replace this series into the “Reference 1” series. Save this profile.
6.2
Radiometric Correction
Once the spectrum is calibrated for wavelength, we can adjust for intensity variations due to our equipment and the atmosphere. This process falls under the category of radiometric corrections in spectrometry terms. It is analogous to flat fielding in direct imaging. The telescope, the optics of the spectrometer, the CCD chip, and the atmosphere all affect the intensity curve of our spectral image. The equipment responds differently to different wavelengths of light. The variation in wavelength sensitivity is a common problem with CCD chips; they are typically more sensitive to red light than blue. The optics of the telescope and the spectrometer can also affect the intensity. The upper atmosphere scatters light laterally, an effect known as Raleigh Scattering. Higher frequencies are scattered more than lower frequencies. The result is that relatively more blue is lost and the light we collect has been atmospherically reddened. The amount of blue light lost depends on the amount of atmosphere it passes through; therefore, this loss is dependent on the position of the object in the sky. It is difficult to isolate each variation, but there is no need because we can correct for all of them at once (Desnoux 71).
Making the radiometric correction is not vital for all applications. For example, it is probably not necessary for Doppler shift measurements. However, it is necessary if you are to accurately compare the spectral intensities of lines located at different sections of the chip. The radiometric correction is relatively easy to perform. Even if it is not crucial to your work, it often makes the spectral profile more recognizable and easier to understand. However, it is up to the researcher to decide if this correction is worthwhile.
To make the radiometric calibration, we collect the spectrum of a bright star of known spectral class. We then find the response curve of our equipment by comparing the spectrum to a standardized spectra of a star with the same spectral class. We can then use that response curve to correct other spectra within the same general region of the sky. Vspec has a library of 131 standardized spectrum of various spectral classes. The spectra come directly from the Centre de Donnees Sellaire of Strasbourg (Desnoux 71). Under tools/library, find the same spectral class as the sampled star and drag the file into the spectral profile. A new series of the standardized star will appear in dark purple. The spectral lines of the standardized star should match the spectrum of the sampled star. The spectra may appear to be off horizontally; this may be due to the radial velocity of the star or problems with calibration. As long as the shift is not too great, it will not pose a problem for determining the response curve. Figure 6.2 shows the sampled star intensity series in blue and the standardized star series in dark purple.
Figure
6.2
Next we apply a spline filter to the standardized series. The filter is accessed through “operations/spline.” This filter softens the spectra, and if we apply a strong enough coefficient we can use it to model the continuum of the standardized spectrum and essentially remove its spectral lines. A slider bar appears to adjust the coefficient. Adjust the slider until prominent spectral lines are masked. The new filtered series appears in pink and is updated as the coefficient is changed. Once the coefficient is adjusted to satisfaction, simply close the slider window. Figure 6.3 shows the original standardized series with the filtered series.
Figure
6.3
Select the original intensity series of the sampled star spectra to make it the active series. Divide this series by the splined standardized series.[1] This is done through the “Operations” menu. Select “Divide” and then select the filtered series; this will be the standard series title with the prefix “sp,” for splined. A new series appears in green. The divided series looks similar to the intensity series of the sampled spectra. Spectral lines are the same; however, the overall curve of the series is different. Next, we want to spline filter this series.[2] Follow the procedure for spline filtering described before. When filtering this series you may find that the coefficient has to be rather large. On the spline filter window there is an option to increase the coefficient value by 10x. This splined version of the divided series is the response curve of the instrument. Make this the active series and replace it into the Reference 2 series. Figure 6.4 shows the final response curve.
Figure
6.4
This response curve can be applied only to targets within the same region of the sky. To calibrate an unknown, it is necessary to image a nearby star and use its response curve. To apply it to other profile documents, copy and paste it just as we did for the calibration spectra. For the purpose of demonstrating how to complete the correction once the response curve is found, we will continue with this same spectrum. To finish the calibration, we make the original sampled intensity the active series and divide this by the response curve, which is now the Reference 2 series. This gives us our adjusted series. To save this series, replace it into the Normalized series. If you choose to normalize the series after calibration, it will automatically replace itself into the Normalize series. The next section discuses normalization.
Figure 6.5 shows the standard star series from which we based our ideal curve on. Also shown is the final divided series. Notice that the spectral lines are similar and the overall continuum curve is similar. The original intensity series is also shown. Some of the spectral lines are marked by bars. These are atmospheric absorption lines produced in our own atmosphere, thus explaining their absence in the standardized spectra.
Figure
6.5
6.3
Normalization
The last calibration to do is normalization of the series. Normalizing the series refigures the intensity levels relative to a continuum with a value of one. Spectra of an object taken on different nights may have different intensity levels because of different exposure times or atmospheric conditions. The profiles will have similar lines; however, in order to compare them we need to put their intensity levels on a relative scale. This is done through normalization. In order to compare two spectra by simply overlaying them, dividing them, or measuring equivalent width, it is necessary to normalize them.
Normalization
in Vspec is relatively simple. First,
it is necessary to select a section over the continuum that will serve as the
reference continuum. Under
options/preferences/continuum, enter the beginning wavelength of the continuum
and then the end wavelength. This part
of the spectrum will be given a value of one. It is important that this area does not contain any spectral lines. Next use the normalize icon, 
. The normalized series automatically replaces
the basic Normalized series. Figure 6.6
shows the spectra on the original scale and on the normalized scale. Note that they both look identical; however,
the scale is much different.
Figure 6.6