Version 5
Guide to processing slitless spectra

Translated by Robin Leadbeater

Version française

1. Pupose

This document is a guide to processing spectra acquired with spectrographs which do not have an entrance slit ("slitless spectrographs"). The most popular of these are probably those based around the StarAnalyser transmission grating. This component is most often simply placed in the converging beam of a telescope, a few centimeters in front of the detector. There are variations in use, in particular combining with a prism in order to reduce optical aberrations (GRISM mounting). The spectra used for this tutorial are from a somewhat more sophisticated instrument with  the GRISM is arranged in a collimated beam, giving higher quality spectra. The common factor of all these instruments is that we do not have the help of a spectral calibration lamp to calibrate the spectra in wavelength. This paper shows a method to circumvent this difficulty using ISIS software.

It is strongly advised before reading this page to spend some time on the Lhires III guide - available here. This guide actually addresses the use of ISIS from a broad perspective, describing many uses, concepts and tricks. Even if you do not have a Lhires III, reading this will be profitable. It  will also allow you to see the difference between slit and slitless spectrographs.

I suggest you also follow the ISIS processing described in this guide using the practical example given. You will progress much faster when confronted with a real practical situation. This is a step by step method that guarantees success with ISIS if you follow the procedure!

Procedure for changing ISIS langage interface.

2. Configuration

The set of images to accompany this guide are contained in the ZIP file downloadable from here:

After downloading (7.4Mb), uncompress the files into a directory of your choice (check that the subdirectory CALIB has been created).  In ISIS terminology, this is the working directory

In this case I chose the following path for the working directory (near the root of disc C):


simply because the images were aquired during the night of 12/13th August 2012

The slitless spectrograph in question is mounted at the focus of a small Takahashi FQS85ED refractor (it is an astrograph but it is quite possible to use it for spectroscopy as we will see). The imaging camera is an Atik314E+ model equipped with a Sony CCD with 6.45 micron pixels. The images were aquired using the program AudeLa. They were taken at  l'Observatoire de Haute-Provence (OHP) during the 2012 spectroscopy workshop.

Start the ISIS program, then open the tab "Configuration" and complete the following fields:

In the field "spectrograph model" it is important to select "Star Analyser". You will then get help specific to this type of spectrograph during processing.

Then provide the path to your working directory (here c:\nuit_20120812).

The ISIS spectroscopic database is installed in the directory C:\ISIS_DATA.

The rest is less critical. Consider however defining the limits which ISIS will use to normalise the spectra. 

It is also recommended that you include your name as observer in the file name.

3. Calibration

First we are going to process a sequence of spectra of the star Alpha Cephei. It is a bright star of spectral type A7iv.  We have 5 spectra in the series, each of 4 seconds exposure.

We will assume that you have a master offset (or bias) image for your detector as well as a map of the thermal signal of the same detector. To learn more about how to acquire these images, click here.

In the set of raw images that accompanies this guide, the offset image file is called master and the thermal signal image is (the latter name indicates that the image is  a composite of dark images each 120 seconds exposure  recorded at a CCD temperature of -5 ° C). These images are stored in a subdirectory of the working directory named CALIB (it is a reserved name).  There is also a file cosme.lst, which identifies hot pixels in the thermal image.

Here we are not going to use a flat field image (for normal work, the flat-field usage is hightly recommanded).

We will first look at the material that we will be working on, for example the first image in the Alpha Cephei series. The corresponding file names in the working directory are alphacep-1, alphacep-2, ..., alphacep-5. Note the hyphen "-". The generic name is "alphacep-" and the root name is "alphacep".

Open the "Image" tab and use Windows Explorer (folder and magnifying glass icon) to find the file or enter the name directly in the field at the top of the window.  Here we see the first image in the sequence:

Remember, it is easy to scroll through the images in the series by clicking on the right or left arrow buttons:

The images are displayed automatically each time you press an arrow button. Note that our 5 images are very similar (apart from the noise, the random component which will be reduced by summing the five individual spectra to form the final spectrum).

Click "General":

In the "General" tab that appears, if not already done, enter the name of the master images, and the pixel size (6.45 microns), and finally information about your instrument and place of observation:

Provide a name for the object as designated in the catalogues (CDS, your software chart, etc).  Here we choose "Alpha Cep" (an abbreviation recognized by CDS, but we could also have chosen the HD number "HD4544", for example):

Our first task will be to find  the dispersion law (the relationship between image pixels and wavelength). At this stage it is assumed this function is not known (for example you are using the spectrograph for the first time). Note that the "Calibration" field is inactive due to choosing the StarAnalyser as the instrument (Unlike the  LISA or Lhires III slit spectrographs you cannot use a lamp to calibrate your spectrum).

To determine the dispersion equation (x = f (lambda), where x = pixel number), we will use the lines in the spectrum of the star being studied (this step will not be needed when processing future objects). The choice of Alpha Cep is not insignificant in this context. This star has clear, easily identifiable lines (the Balmer series of hydrogen). Here's how to proceed.

Start by unchecking "Make the spectral calibration":

By doing this, ISIS will begin processing the image sequence, but the application will stop at the calibration step. We will get a good spectral profile, but it will be scaled in pixel number, not lambda (wavelength).

Before starting processing, open the tab "Calibration".

The first image in the sequence is automatically displayed (if necessary, adjust the display thresholds)

Double click with the mouse near the middle of the trace. ISIS calculates the Y coordinate of the spectrum (The vertical position) and automatically transfers the result to the field "Y coordinate of the spectrum trace" Here we see Y is around 101 (The value could be perhaps 6 or 7 pixels around this value):

Drag  the mouse, drawing a large horizontal rectangle enclosing a relatively intense region of the trace. Then click the button "Tilt". ISIS returns the angle of inclination of the spectrum trace relative to the horizontal (tilt) , here -0.35°:

Concerning the value of the slant angle (inclination of the spectral lines from the vertical), you can leave this at zero.

Select the "Reticule" option at the right of the tab. If you press the button "adjustment of the binning zone" you can adjust the position of the markers that indicate where ISIS will assess the level of the sky background (from both sides of the spectrum trace) and the vertical width of the binning area used to construct the profile from the image:

Here I have chosen the "option 2" predefined values

You can now process the spectrum. Open the "Go" tab and click the "Go" button:

Note: ISIS tells you that the treatment is not complete, it lacks the wavelength calibration.

Display the spectral profile by clicking on the button "View Profile":

The result (Note that if you move the vertical cursor along the profile, ISIS displays the number of the pixel, not the wavelength):

The displayed profile is stored as an intermediate file in the working directory, in the format .dat (editable using Notepad or Excel for example) The file is automatically named by ISIS as @alphacep (ISIS added the @ symbol to the name root).

We can easily identify up to 7 Balmer lines which can be used for spectral calibration:

Click the "Dispersion" button.

In the dialogue box that opens, remember to select the option "Lines in absorption":

Enter the precise wavelength of the Halpha line in the field "Line #1" :

Tip: To find the wavelength of the lines in the Balmer series, press  "Database" from the "Profile" tab, then "Balmer" in  "Reference wavelength."

With the "Calculating the Spectral Dispersion" window open, double click on the left side of the Halpha line, then on the right side (you can actually reverse this order ):

Tip: You can change the display scale to aid visibility.  Note also that the exact position of the cursor when you double click has little influence on the final result.

ISIS inserts the horizontal coordinate of the Halpha line in the calibration dialog box and prompts you to enter the next line wavelength and its position - up to 13 lines:

Repeat this sequence for all seven hydrogen lines (you can choose another number, or use other lines - it's your choice):

Tip: Entering the list of lines can be tedious.  It is perfectly possible to save the current list to a file and read it again as the need arises. For example:

Calculate the dispersion equation by clicking the "Calculate the polynoml" button after having selected the order (3 is sufficient here):

In this example the typical RMS error of the calculated terms in the equation is 0.55A. This is a good result considering that the dispersion is around 4A/pixel. Close the "Calculate the spectral dispersion" dialogue box

Now that we know the dispersion equation, we can complete the processing of the Alpha Cep spectra.  Open the "General" tab and check  "Make the spectral calibration." This is now the normal way ISIS is used.  You must also tell ISIS to use the dispersion law already calculated. Select "Use the calculated dispersion" in "Spectral Calibration" section:

Return to the "Calibration" tab. Here we have two choices:

(1) Either one enters zero in the X coordinate field "X coordinate of the line wavelength ...". ISIS interprets this 0 value as :

"I have a calibration law perfectly matched to the star studied" (for proof, I used the star itself for that), so I can click "Go" directly, without indicating a reference line. The polynomial calibration is used as it is, without any special adjustments.

(2) I enter the "X coordinate of the line of wavelength ..."  A non zero value, the horizontal coordinate of a  high contrast line in the spectrum, and I also enter the specific wavelength of this line.

We will experiment with these two techniques. Firstly, the first and most simple:

Start the processing by clicking the "Go" tab :

View the resulting profile. It is indeed now calibrated in wavelength

Now let's try the second method.  From the "Calibration" tab, locate the X coordinate of a  high contrast line (within 5 or 6 pixels, you have a margin of error). Use for example the Hbeta line, at wavelength 4861.34 A.  We find with the help of the mouse pointer that this line is located at the approximate coordinate X = 335:

Important note: If you decide to select an emission line as reference line, you must select the option "Line in emission."  For example, if your line is  Halpha in star Gamma Cas (a Be star where this line is strongly in emission). This is also the situation if your reference line is the zero order image of a StarAnalyser spectrum . In this case, you must enter zero as the wavelength (by definition), the horizontal coordinate X of the zero order image, and finally check "Line in emission."

Re-run the calculation via the "Go" tab. The result is quite similar to the previous case.  What is the difference and in particular the value of the second technique which requires locating the position of a particular line? This latter method should be used when processing other objects observed in the same session.  With a slitless spectrograph it is very unlikely that you will precisely position the various objects studied in the same place on the detector. By identifying the position of a known line in the spectrum, the dispersion law already calculated can be applied (by simple translation) to any displaced spectrum. In practice, the A0 term in the dispersion equation is modified.

We must now find the instrumental response. This is done via the "Profile" tab.

Remember that Alpha Cep is an  A7V type star. Click  "Response" in the toolbar. In the dialog box that opens, click "Select a reference spectrum" . Choose from the basic Pickles database, the synthetic spectrum of a star type A7V, then click "Select". The location of the reference spectrum file then appears in the "Instrumental response" dialog box. You just have to click on "Response" and adjust the lower and upper display limits (click again on "Response" to confirm the selection.)  Here is the state of the "Profile" tab at this stage:

In blue is the processed spectrum (not corrected for instrument response as this is unknown at this point).

In red is the synthetic spectrum of a type A7V star.

In green is the ratio of the two spectra. This is the desired instrument response.

Click "OK": The instrument response is retained on the screen. It only remains to smooth the result to remove noise and residual lines.

Click on "Continuum", then adjust the slider so that the curve (blue) smooths out the noise in the raw spectral response (in red):

If necessary, erase unwanted features by double clicking on both sides of them. If you press the "OK" in the "Adjusting the continuum" tool, you will retain the smoothed response.

Tip: The response profile changes shape very rapidly near the edges and it may not accurately represent the response in these regions as the signal there is almost zero. It is a good idea to crop the useful part  before smoothing the spectrum. Use the "Cut" tool for this operation. The smoothing will then be easier and the results better. Here is the smoothing result under these conditions:

Save the instrument response in a FITS file called for example "reponse_alphacep":

To complete the processing of the alpha Cep spectrum including instrument response correction, select the "General" tab:

Then start the processing by going directly to the "Go" tab.  Here is the final result:

Compare this with the synthetic spectrum for an A7V type star (from the Pickles database):

In blue is our spectrum, in red is the theoretical spectrum for the corresponding spectral type. The agreement is very good, with a similar resolution even far into the ultraviolet.

4. Routine processing

Now let's process the spectrum of star delta Cas. The raw images are found in the downloaded zip file. There is a sequence of 5 images of 8 seconds exposure. View the first image:

Supply the catalogue name of the object:

Under the "Calibration" tab, update the Y coordinate of the spectrum and the X coordinate of the Hbeta line (for example):

Start the processing from the "Go" tab:

View the resulting profile to check it:

It is done, you can move on to the next object. It is that simple.

Another example...

Let's process a sequence of 8 images of the star P Cygni (34 Cyg, V=4.8) taken with an individual exposure of 14 seconds.  

The appearence of a raw spectrum image:

 Locate the very bright Halpha emission line (note, at this point select the "Line in emission" option :

The processed profile:

It is instructive to examine the resulting processed 2D image, here shown at high contrast to reveal the sky background:

The sky background is cluttered with spectra from faint objects. The richness of this type of low resolution spectrography can be see with a keen eye but it is also a disadvantage: Without a slit, objects in the sky background reduce the photometric quality and the detectability (photon noise of the sky background). This is a limitation of all slitless spectrographs.  During processing, it is possible to fine tune the choice of binning and sky area calculation zones to avoid the annoying spectra from other  objects.  Here is a typical example of the star WR136, magnitude V = 7.5: It is absolutely necessary to avoid including the nearby star in the binning and background areas. The adjustment of binning areas is critical, but it is made easier by seeing the results interactively:


5. Supplement

To finish, I suggest processing a spectrum of the star delta Cas acquired using a particular technique: During the 20 seconds exposure, the telescope declination drive was run at slow speed. As a result the spectrum drifts 90° (nearly) from the dispersion direction which therefore spreads the spectrum in this direction. Display the image deltacas2-1 acquired with this technique:

The visual appearance of the raw image is interesting.  We immediately can see many details in the spectrum, including fine lines which we did not suspect beforehand. This technique is useful for teaching spectroscopy "on site".

We continue with the "General" tab - no surprise:

Take care with the calibration ... We must  adjust the binning area intelligently to include the maximum amount of the signal:

The predefined option 4 is convenient here but remember that it is also possible to make a manual adjustment.

The Hbeta line serves as a wavelength reference:

The spectral lines are not perpendicular to the axis of dispersion, because the rows and columns of the detector are not properly oriented with respect to the equatorial frame.  We find by trial and error that the slant angle is -5.5 ° (see below). We enter this parameter:

We run the processing (tab "Go"), and then display the processed 2D image (button "View image" tab "Go")

The lines are now vertical. The -5.5° slant angle was correct (if not, enter a new value, and re-run "Go").

Finally the spectrum profile - I leave it as an excercise to compare the profile with the 2D trailed image: