Observing with UNSWIRF

Wavelength Calibration

The first step in commissioning UNSWIRF was to calibrate the relationship between etalon spacing d and peak transmitted wavelength . This is complicated by a number of factors:

• the actual order n being passed by the blocking filter is not necessarily known in advance;
• the relationship between etalon spacing d, and the analog-to-digital units Z (-2048 < Z < +2047) employed by the control computer and CS-100 must first be calibrated;
• because of the wide tuning range demanded, and the multiple layer coatings involved, the effective depth where reflection occurs within the etalon coatings will change with wavelength, and so too will the apparent physical spacing of the plates. This manifests itself as a change in with respect to an assumed value of the physical plate spacing (or equivalently, by a non-integral order number n). Let us define such that when the angle of incidence , and the interplate medium is air at 1 atmosphere pressure, equation 1 becomes
   
  and thus
   
  Thus, measurements of the FSR over the full range in wavelength and for multiple orders allows us to derive the order number n and the wavelength variation of apparent plate separation .

First, the etalon and all blocking filters are withdrawn from the light path, and IRIS configured with the narrow () slit and its H+K band échelle grism. Images of the emission-line spectra produced by four separate lamps (Ar, Kr, Hg, and Xe) are then taken in order to wavelength-calibrate the four complete échelle orders (covering the range m after straightening) produced by this grism. Next, the Fabry-Perot is inserted into the beam, the plate spacing set to Z=0, and a continuum source (such as a quartz-iodine lamp) used to illuminate the system. Since there are no blocking filters in place, all orders passed by the etalon are imaged, resulting in the ``picket-fence'' appearance of Figure 1. Since wavelength as a function of position on the array is already known, the position of each peak, and their separations (i.e., ) can be determined. For improved accuracy, these measurements are repeated with etalon settings , which causes the etalon orders to shift position on the array, as illustrated in Figure 1.

  
Figure: The ``picket fence'' of orders produced using the H+K échelle and narrow slit of IRIS, a QI continuum lamp, and no blocking filters with UNSWIRF is shown for etalon Z=200 (top left), 400 (top right), 600 (bottom left), and 800 ( bottom right). Note also how the spacing between adjacent orders of the etalon () increases with wavelength (eqtn 2). The wavelength ranges covered by each complete échelle order are (top to bottom, and from left to right) 1.44-1.70 m, 1.62-1.91 m, 1.86-2.18 m, and 2.17-2.54 m.

As Figure 2 shows, the of UNSWIRF is indeed more complex than equation 2 would suggest for a simple dielectric coating. From equations (1) and (2), it follows that for normal incidence in air

The upper dashed line in Figure 2 corresponds to for a constant physical etalon spacing d of 52.0 m, while the lower dashed line is for m. For m, the effective spacing between the plates () is 52.0 m, but this grows rapidly at around 2 m due to the nature of the coatings, and is more like m out to the long wavelength cutoff of UNSWIRF.

  
Figure 2: Variation of Free Spectral Range with wavelength . The upper dashed line indicates a constant physical etalon spacing of 52.0 m, while the lower dashed line indicates a constant physical etalon spacing of 61.0 m.

The same data, analysed using equation 5, indicate that the change in spacing resulting from one step in Z is

Having determined over the full wavelength range of UNSWIRF, the accuracy of this calibration has been tested by comparing the predictions of equation 5 for the wavelengths of the arc-lines measured earlier with their known wavelengths. We find an r.m.s. accuracy of 0.04 nm, or % of the instrumental resolution.

Parallelism, Resolving Power, and Finesse

Having calibrated the relationship between d, Z, and , it now becomes possible to set the etalon spacing for the peak transmission of any required wavelength. The next step is to make the plates as parallel as possible, and maintain this parallelism over the full wavelength range, given the residual non-flatness of the plates and/or their coatings. This can most easily be done by scanning in wavelength across an unresolved emission line (e.g., from a discharge lamp or from the OH airglow) and building up a ``cube'' of images, with the third axis representing etalon spacing Z. For each spatial pixel (x,y) in the cube, the spectrum of intensity I(Z) is analysed with the UNSWFIT routine (Section 3.4) to determine the position and intensity of the emission peak. Owing to the dependence on of the condition for peak transmission (equation 1), should be a maximum near the centre of the array, and decrease towards the edges, as shown in Figure 3. From analysing such images, we find that the etalon surfaces are flat to at a wavelength of 1.65 m, well within the specification.

The way we have chosen to monitor the parallelism is to fit a tangent plane to this surface, and then use the measured x- and y-slopes to correct the parallelism settings passed to the etalon from the CS-100. The parallelism has been found to be weakly, but repeatably, dependent on Z (and thus on , due to the fact that the effective reflection at different wavelengths comes from different depths within the coatings, and the coating thicknesses vary slightly), and this is now accounted for by the observing control software. After optimising the parallelism, we have calculated the resolving power and effective finesse (equation 4) of UNSWIRF using a series of discharge lamp lines over the available wavelength range, as tabulated in Table 3. The high reflectivity R of the plates will make the dominant contributor to . In addition, the throughput of the etalon at each wavelength (except for 2.334 m) has been measured by comparing the peak intensities obtained with the etalon in and then out of the beam.

  
Figure 3: UNSWIRF scan of the Ar 1.6520 m line, incrementing the etalon spacing by 2Z units each time. Etalon spacing increases from left to right, and from top to bottom.

The shift in peak transmitted wavelength for the Kr 2.1165 m line, going from the centre to the edge of the etalon, is quite small compared with the instrumental resolution. As can be seen from Figure 4, the shift is over the inner 90 pixel diameter, and still over the entire usable field of view. In fact, owing to possible non-uniform illumination of the etalon by the discharge lamp, Figure 4 may slightly overestimate this shift. Thus, UNSWIRF is virtually monochromatic, and can in principle be used as a pure tunable line imaging filter, provided velocity gradients and dispersions are small (<50 km ), and the line centre wavelength is known in advance. Otherwise, more extensive scanning in wavelength will be necessary (but this of course furnishes, as a spinoff, the velocity field). Although the ability of UNSWIRF to resolve lines is limited by its instrumental profile width ( km , depending on parallelism), we have found that the profile fitting allows us to measure shifts in the position of the line peak equivalent to velocity changes of <10 km , depending on signal-to-noise of the data, profile shape, and plate parallelism.

  
Figure 4: Contour plot of the shift in transmitted wavelength with position on the IRIS array for the Kr 2.1165 m line, as a fraction of the instrumental profile width . The heavy black line around the edge marks the unvignetted field of view of the etalon. Beginning from the top right edge, the contours mark a shift of -80%, -40%, -20%, -10%, and -5% of relative to the mean peak wavelength at the array centre.

Observing strategies

Observing with UNSWIRF is much like normal narrow-band imaging in the near-infrared. As recommended by the IRIS manual (Allen 1993), readout method 4 is employed, which breaks up each exposure into a series of Non-Destructive Readouts (NDRs), enabling on-the-fly bias correction and linearisation, and yielding the lowest possible read-noise (typically  e). To guarantee background-limited performance, each exposure at a given etalon Z setting is normally 120 s in duration at K (180 s at H), broken up into 12 NDRs. In order to reduce overheads, a complete scan in Z on the object is normally done before moving the telescope to an offset sky position away, and repeating the sequence. Except when the scan crosses a strong, and rapidly varying, OH airglow line (usually more of a problem in H-band than K-band), sky subtraction is found to be quite adequate, even when sky frames are taken 15-20 minutes after the matching object frame.

An existing procedure for commanding telescope spatial offsets, written in the AAO DRAMA environment (Bailey et al. 1995), has been enhanced with the ability to request etalon spacing and parallelism changes from the CS-100. To guard against possible drifts in the UNSWIRF etalon parallelism (usually in response to changes in the ambient temperature and/or humidity), the parallelism is normally checked immediately prior to each night of observing, by scanning a calibration lamp line close to the region of interest.

Because of the monochromatic nature of the infrared radiation reaching the IRIS array from UNSWIRF, it is essential that matching sky exposures and dome flatfields be obtained for all of the etalon Z settings used on an object, as otherwise severe fringing can result. Similarly, bright spectroscopic standard stars need to be observed once, and preferably twice at these same Z settings. Although accuracy of the photometry is usually limited by the sky and the array to % at best, it is necessary to determine the intensity scaling of the continuum images relative to the line peak in order to ensure proper continuum subtraction.

Data Analysis

A schematic of the basic data reduction procedure for UNSWIRF is shown in Figure 5, and begins with subtraction of a matching sky frame, followed by division by a normalised matching dome flat. Monochromatic imaging of simple sources then requires just a scaling and subtraction of the off-line frame. For more complex sources, a ``cube'' is constructed from a sequence of such monochromatic images at a constant Z interval, aligned to a common spatial frame defined by field stars. The moments of this cube (integrated line intensity I(x,y), Z position of the line peak , and the line width ) are extracted by fitting a Lorentzian to the spectrum at each spatial pixel. In general, the line profile is usually too noisy to allow three free parameters (the base level having already been set to by the off-line subtraction). Since in most cases the emission-line profile will be unresolved by UNSWIRF, the line width can be assumed to be the same as the instrumental profile width, as mapped by the calibration lamp line scans, leaving only two free parameters in the fitting. Finally, all pixels in the vignetted corner regions of the moment maps, as well as any pixels which fall below a specified intensity threshold, or for which the fitted lies outside the actual Z range scanned (assuming that the observations did adequately span the line of interest) are blanked out.

  
Figure 5: Schematic diagram of data reduction steps, for monochromatic imaging, using on- and off-line images of the H emission around OMC-1.

In order to streamline the processing of UNSWIRF data, a suite of programs has been written using the IRAF environment. A listing of these programs and their functions is given in Table 4. With the exception of UNSWFIT, these programs are scripts written in the IRAF Command Language (CL) which execute a series of existing IRAF routines. The UNSWFIT task is a purpose-written SPP (Subset Pre-Processor) program that uses a Lorentzian-fitting algorithm supplied by F. Valdes.

© Copyright Astronomical Society of Australia 1997