After flat fielding, the V band frames are ready for measurement. The I band frames, however, suffer from fringing, as shown in Figs. 3.1 and 3.2, and these must be removed. These fringes are caused by thin film interference of monochromatic atmospheric emission lines.
Some fraction of the photons which pass through the front surface of a CCD’s silicon layer are absorbed as they pass through the interior, generating photoelectrons. The remain- der are not absorbed, and can be reflected off of the back surface back into the interior. Some will undergo multiple reflections before absorption or permanent escape from the CCD. Interference between incoming and reflected light generates a fringe pattern. Unlike an image of a monochromatic source, an image of a continuum source will show either no, weak, or negligible fringing, as the positions of the peaks and troughs of the individual wavelengths in the continuum are slightly different.
Images taken through infrared bandpasses suffer from severe fringing for two reasons. First the quantum efficiency of silicon is lower at these wavelengths than at bluer wave- lengths, resulting in less absorption, more reflection, and more interference. Secondly and more importantly, the far red and infrared region of the electromagnetic spectrum contains
a large number of atmospheric emission lines, especially from OH.
Removing these fringes involves producing a fringe frame from images of the dark night sky and fitting and subtracting that fringe frame from one’s program frames. The fitting consists of subtracting a constant sky from the fringe frame and then scaling it by a multiplicative factor which accounts for both the difference in exposure times between fringe and program frame as well as any variation in the amplitude of the fringes between the two. This sky subtracted and scaled fringe frame is then subtracted from the program frame. All of this is accomplished in one step with IRAF taskrmfringeof packagemscred. The procedure outlined below for producing master fringe frames actually removes the sky, so the sky subtraction above is redundant, though it is built intormfringe.
In order to find the correct fitting of fringe to program frame, one must feedrmfringe an object mask, a frame in which pixels in the program frame to be ignored are marked out. These are to be ignored because they record astronomical objects. One wants the fit to based on the fringes, not stars and galaxies. This mask is produced in a separate task.
Producing the fringe frame is a time-consuming effort. One must have a large total exposure time in order to capture enough signal. Because the night sky is full of stars, the best way to capture this signal is not to take a single exposure, but to take multiple exposures, dithering each frame so that no star falls on any given pixel in a majority of frames. The frames are then combined in some fashion while employing some kind of rejection algorithm to filter out stars. Thus, while one would ideally like to construct a fringe frame for each night, in practice one typically does so a few times a year at most and uses the same fringe frame across nights.
Fringe frames were constructed for four nights : May 22 and 23 of 2007, and June 11 and September 23 of 2009 using SOAR archival data. Combining the fringe frames of each night into a good master fringe frame for that night turned out not to be straightforward.
After experimentation and consultation with a guide formerly located on the website of the Kapteyn Astronomical Institute, but which has since been taken down (its URL was formerly http://www.astro.rug.nl/∼lugt/main.html#fringes), it was found that the best results were obtained by first fitting a two dimensional linear function to the background sky of each fringe frame and subtracting away that sky. The sky subtracted fringe frames were then median combined into a master fringe frame. No rejection algorithm was employed to filter out stars, as combining by taking the median inherently filters out outlying data.
Fitting and subtracting the sky of each fringe frame is an absolutely crucial part of this process. The master fringe frames obtained by combining the individual fringe frames without fitting and subtracting the sky are simply not very good. In order to obtain an even better sky subtraction, some authors advocate taking the master fringe frame obtained by the process above and using it to defringe the individual fringe frames that were combined to produce it. A better fit to the sky for each frame is then obtained from these defringed frames. That fit to the sky is then subtracted from the original un-defringed frames, and these are then combined into a new, better master fringe frame. The process is iterated until a satisfactory master fringe frame is obtained. For all four of the above nights, one additional iteration was performed. As there was found to be no difference between one additional iteration and no additional iterations, the master fringe frame produced from no additional iterations was adopted as the master fringe frame for each night.
Each program I frame was fringe subtracted using each of the four master fringe frames. The one that showed the most complete subtraction of the fringes was the one that was used in the photometry. Each of the four fringe frames turned out to be the best fringe frame for at least some of the program frames. There were times when the performance of two or more of the fringe frames was so close as to be indistinguishable. In those cases, one was simply picked.
of packagemscred. Subtraction of that sky background was done with the basic IRAF task imarith. mscskysub andimarith need to be called by taskmsccmd of package mscred in order to work on MEF files (note that this is true ofmscskysubeven though it is also part of packagemscred.) Each MEF file was then joined into a single FITS file by tasksoimosaic of packagesoibefore combining. Combination of individual fringe frames was then done with task imcombine of package immatch. Frames were scaled by exposure time before combining. The combine operation utilized was median combining. No rejection algorithm was employed, as median combining in and of itself tends to ignore outlying points. Object masks for each program frame were produced with taskobjmask of packagenproto. And finally, removal of fringes from program frames was performed bymscredtaskrmfringe.
A program frame after overscan fitting and subtraction, bias subtraction, and flat fielding is shown before defringing in Fig. 3.2 and after defringing in Fig. 3.3. The fringe frame used in the defringing is shown in Fig. 3.6.
Figure 3.1:I band raw image of Boötes II, taken on the night of May 16, 2007, at UT mid- exposure of 02:30. Notice not just the fringing, but also artifacts such as dark dust donuts. SOI’s detector consists of two CCDs each read out by two amplifiers. Note the slightly different bias level of the quarter image corresponding to each amplifier.
Figure 3.2:The image of Fig. 3.1 after overscan fitting and subtraction, subtraction of the bias, and division by the flat. Note the disappearance of image artifacts.
Figure 3.3:The image of Fig. 3.1 after overscan fitting and subtraction, subtraction of the bias, division by the flat, and defringing. Some residue of fringing is still visible, but the fringes are by and large gone.
Figure 3.4:Master bias frame for the night of May 16, 2007, used in the reduction illustrated in Figures 3.1 through 3.3.
Figure 3.5:Master I band dome flat for the night of May 16, 2007, used in the reduction illustrated in Figures 3.1 through 3.3.
Figure 3.6:Fringe frame used in the reduction illustrated in Figures 3.1 through 3.3. The frame was produced by combining fringe frame data taken on the night of June 11, 2009. The band down the middle is the gap between the two CCDs of SOI. This was not visible in Figures 3.1 through 3.5 either because the versions shown in those figures were mosaics of the quarter images taken from the four amplifiers of SOI’s detector displayed in a manner that did not show the gap, or the version shown was one with the gap trimmed out.
CHAPTER 4 : PHOTOMETRY