The raw images from Mini-Mo must be reduced before any scientific analysis can be done. The images are processed using the Image Reduction
and Analysis Facility (IRAF). Since Mini-Mo is a CCD mosaic, some addition reduction steps are necessary in addition to those described in
2.1 Charged-Coupled Devices. The most notable additional reductions include merging amplifiers and removing
other mosaic gaps. Mini-Mo has four separate amplifiers for two CCDs separated by a physical gap. We briefly name and discuss the key IRAF tasks
that are used to reduce our raw Mini-Mo images.
The main IRAF package for mosaic images is MSCRED developed by Francisco Valdes. The basic mosaic reduction task is CCDPROC. This task
allows you to measure and subtract the bias level for each image from the overscan region, trim the overscan region, merge amplifiers,
subtract off a zero frame, divide by a
flat-field, and interpolate over bad pixels and create a bad pixel mask. CCDPROC is first run on all the zero images correcting for only the bias
and trim, which are then combined into a master zero frame using ZEROCOMBINE. Next, CCDPROC
is run on the dome flat images correcting for the bias, trim, and zero level. These reduced dome flat images are combined using FLATCOMBINE,
which will make a master flat for each filter. Finally, CCDPROC is run
on all the science images with correcting for the bias, trim, zero level, gain
variations, and bad pixels. In addition, the amplifiers are merged. CCDPROC
merges amps 1 and 2 and amps 3 and 4 but cannot remove the physical gap between
the two CCDs.
To remove the cosmic rays from our images, we use the IRAF task LACOSMIC developed by Pieter Dokkum. The task is part of the
STSDAS package and may be downloaded at http://www.astro.yale.edu/dokkum/lacosmic/.
Typically, cosmic rays are removed by
combining dithered images because the likelihood that the same pixel will be hit by a cosmic ray in more than one image in a dithered set is low. However,
since the readout time of Mini-Mo is approximately three minutes, taking more than two images of an object is not efficient. Instead, LACOSMIC
uses a Laplacian rejection algorithm to remove cosmic rays. More information on the task can be found here. LACOSMIC is not designed to work
with mosaic data. We must use MSCSPLIT and run the task on each individual
To accurately combine dither images and consistently perform other analysis such as photometry, we apply a world
coordinate system (WCS) to each science image.
This amounts to finding a conversion between pixel coordinates on the image and the equatorial coordinate system.
The IRAF task CCMAP takes as input a list of pixel
and celestial coordinates for a few objects in an image and outputs a conversion between pixel and
celestial coordinates. The solution is stored in a .db file and can be applied to the image from
CCMAP if the update parameter is
set to yes.
For each amplifier section we choose approximately five unsaturated stars. We check these stars for saturation by using IMEXAMINE to make radial
profiles. We choose stars that are common between images from different filters. Once an appropriate set of stars has been
chosen, we compute their pixel
coordinates by centroiding using IMEXAMINE and the 'a' key. Next, we use NED
to search for our object. From NED we can access the
Aladin Java applet image from the Palomar 48in Schmidt telescope of our object from which we can search for the stars in our set and obtain USNO A2.0 catalogue coordinates for these stars.
We repeat this and run CCMAP on each amplifier section for each filter.
Now that each amplifier section has a WCS applied to it, we are ready to rejoin the amplifier sections. To do this, we simply use the MSCRED task
MSCJOIN . The task has a clobber option, which will delete the individual files after they are joined.
I do not use this option because IRAF
likes to do things I don't want it to (like delete files it wasn't supposed to).
The last step in the astrometric calibration is a rotation. The raw images from Mini-Mo have North pointing to the left
and East point down. We would like our
image to be oriented such that North is up and East is left. Thus, we must apply a 90 degree clockwise rotation. We use IMTRANSPOSE instead of
ROTATE because the former will actually rotate the WCS as well. The syntax for the 90 degree
clockwise rotation in IMTRANSPOSE is input=name_image.fits[-*,*].
Our images now have the standard reductions and a WCS applied to them. For single CCD images, this would probably be enough.
However, since we are dealing
with a CCD mosaic, we must remove the mosaic gap since objects frequently lie across the gap.
Mosaic images are not regular FITS files. They are actually multi-extension fits files (MEF). For Mini-Mo, the MEF file is actually four
FITS files (two after we merged amplifiers with CCDPROC). We would like to change our MEF images into single FITS images.
This is accomplished using the MSCRED task MSCIMAGE. This task will produce a single image with the CCD gaps visible, but set to zero. For more
on how MSCIMAGE works, refer to Francisco Valdes's paper
"The Reduction of CCD Mosaic Data.", which can be found at
Our goal is to remove mosaic gaps by combining dithered images. When combining dithered images,
we must first match the intensity levels of the two images.
This is done by comparing the background sky level.
The combined image will be poor if there is a sky gradient in our images. To
remove the sky gradient and measure the mean sky level, we use the MSCRED task MSCSKYSUB.
This task fits a surface to the sky and removes the gradient without subtracting off the sky level when output is set to residual. The mean sky level is
stored in the header file under SKYMEAN. This keyword will be used by
MSCIMATCH and MSCSTACK. Please refer to Valdes (2001) for
more on how MSCSKYSUB works.
MSCIMATCH is an IRAF task that assumes a linear relation between the
intensity levels of two dithered images and computes and stores the multiplicative and additive scaling factors
between the two in the header file under MSCSCALE
and MSCZERO. These two keywords will be used by the task MSCSTACK. Again, refer to
Valdes (2001) for more on MSCIMATCH.
MSCSTACK uses the header keywords SKYMEAN, MSCZERO, and MSCSCALE to first match the intensity level of the dithered images
and then uses the WCS to combine overlapping pixels by averaging. This will remove the physical gap between the CCDs of Mini-Mo. Once this is done for each
filter, our images are fully reduced and we are ready to perform scientific analysis. Below we show a raw and reduced image of the UGC 7576 field.
| RAW || REDUCED |
| (a) || (b) |
Figure 3. This figure shows a raw (a) and reduced (b) images of the UGC 7576 field from Mini-Mo of the WIYN 3.5m at KPNO.
Here we present reduced gray-scale images of the polar ring galaxies UGC 7576, NGC 2685, and NGC 3718.
In Figure 4 below, we show the reduced gray-scale B and R band images for UGC 7576. UGC 7576 is very similar to NGC 4650A,
which is the best studied polar ring galaxy. It contains a nearly edge-on disk and a very luminous, elliptical central host galaxy.
Assuming a value of 70 km s-1 Mpc-1 for the Hubble Constant and using the radial velocity measurement of 7000 km/s for
UGC 7576, we estimate that the galaxy is approximately 100 Mpc away.
From the B image, we estimate the diameter of the polar ring is 558 pixels and we convert
this number to arc seconds using 0.1'' per pixel. The diameter of the polar ring is approximately equal to the angular size of the polar
ring in radians times the distance of UGC 7576. Thus, UGC 7575's polar ring is approximately 27 kpc in diameter. Note that the polar ring
appears more luminous in B than R. We will explore this color difference with color maps later.
Figure 4. This figure shows the reduced B image (a) and R image (b)
of UGC 7576.
Reduced gray-scale U and B band images for NGC 2685 are shown below in Figure 5. Like UGC 7576, NGC 2685 have a very
luminous, elliptical central host galaxy. However, in this case the polar ring is not a flatten, edge-on disk but rather a helix that
encloses much of the central host and there is a diffuse cloud of dust surrounding the galaxy.
Note the dust lanes from the polar ring cutting around the face of the central host.
Figure 5. This figure shows the reduced U image (a) and B image (b)
of NGC 2685.
Images of NGC 3718 in the V and K bands are shown below in Figure 6. The optical images were reduced
by Elizabeth Wehner and the K image by Marc Verheijen. The central host of NGC 3718 has a pill-shaped appearance and we can see
a dust lane from the polar ring. The polar ring is flat and warped at the edges. Like NGC 2685, the galaxy is surrounded by a diffuse cloud of dust.
The polar ring does not appear in the K band image. This filter is in the infrared region of the electromagnetic spectrum and there is less
extinction or dust absorption at longer wavelengths. This makes the infrared observations ideal for sutdying regions with much dust.
Figure 6. This figure shows the reduced V image (a) and K image (b)
of NGC 3718.