1. Figure 8.Color map of the geometric correction along the dispersion axis for segment A. Figure 4. Measured distortions for all PSA positions for segment A along the dispersion axis. X X (DE) X DX (DE) Correcting for Geometric Distortions on the COS FUV Detector Stéphane Béland (sbeland@colorado.edu), Steven Penton (spenton@colorado.edu)sbeland@colorado.eduspenton@colorado.edu Center for Astrophysics and Space Astronomy University of Colorado at Boulder, USA http://cos.colorado.edu Abstract The Hubble Space Telescope’s (HST) Cosmic Origins Spectrograph (COS) Far Ultraviolet (FUV) detector consists of micro-channel plates (MCP) and time-delay anodes to provide photon counting and imaging capabilities. The COS FUV detector does not contain physical pixels, instead it uses time-delay circuits to compute the photon arrival location (in digital elements (DE)). The map between the physical photon arrival position and DE is a function of the physical characteristics of the MCPs and their associated electronics. Non-linearities in this mapping are seen as geometric distortions in images taken with either of the FUV detector’s two independent segments (‘A’ and ‘B’). Repairs to the FUV detector in early 2003 affected its physical characteristics, making the initial geometric distortion correction maps unusable. A new method was developed during the October 2003 calibration phase of COS to measure the new geometric distortion correction maps and is presented here. This method uses the known spectral lines of Platinum-Neon (Pt-Ne) lamps to map the two-dimensional geometric distortions across the FUV detector. Introduction The first attempt at mapping the geometric distortions of the FUV detector consisted of placing slit and pinhole masks with regularly spaced holes in front of the detector (Figure 1). Images of the masks were taken using a monochromatic light source. Knowing the physical spacing (in X and Y) of each slit/pinhole, it was straightforward to predict the expected physical slit/pinhole image positions on the detector. By measuring the displacements of the X and Y slit/pinhole centroids (in DE) from their expected positions, two-dimensional maps of the geometric distortions were derived. This proved a very effective way of measuring and correcting the two-dimensional geometric distortions for the COS FUV detector (see Figure 2). Unforeseen problems with the detector forced partial disassembly of the unit. This changed some of its physical characteristics enough to make the previously measured geometric distortion correction maps unusable. It was impossible to repeat the previous slit mask imaging at that point because the detector unit was fully assembled and the mask could not be placed against the detector itself. A different method, described in this poster, was needed. Method Description Ray-trace analysis of the HST+COS FUV optics and gratings were used to determine the wavelength dispersions (in  /mm) of the FUV detector segments. During our October 2003 calibration campaign, we used COS’s G160M grating to obtain high- quality spectra of well-defined Pt-Ne wavelength calibration lamps. Using the wavelength dispersion solutions from the ray-trace analysis, we determined the offset in digital elements (DE) for each line relative to its predicted position in both the X (dispersion) and Y (cross-dispersion) directions. A series of spectra were obtained at four grating offset positions (FPSPLIT), each moving the spectra along the dispersion axis (X). This allowed for a better sampling along the dispersion axis ensuring that we obtained a good coverage of identified spectral lines over the whole detector. COS has separate apertures for external targets (Primary Science Aperture (PSA)) and for the internal Pt-Ne lamp (Wavelength Calibration Aperture (WCA)). The projected apertures illuminate the detector at different cross-dispersion (Y) positions. The micro-channel plates of the FUV detector become depleted with usage and become less sensitive. The sensitivity loss is proportional to the amount of light each specific region has seen. Both the PSA and WCA mechanisms are designed to move up or down (in Y) to illuminate different areas over the lifetime of the detector. Using a series of FPSPLIT images taken at a variety of aperture positions, we were able to map the geometric distortion corrections in X and Y for the full science regions of each FUV detector segment. Each FUV segment was analyzed separately and has its own unique geometric distortion correction map. Figure 1.Slit and pinhole mask images superimposed. Figure 2. Geometrically corrected flat-field image showing the fiber bundle boundaries. Analysis and Results We used the G160M grating, the internal Pt-Ne lamp (WCA), and an external Pt-Ne lamp (PSA) with a small mask with a series of uniformly spaced (in Y) pinholes to cast well-known spectra across the science areas of the FUV detector (Figure 3). Around 110 lines were identified for the segment A and around 190 for segment B. With the FPSPLIT, we were able measure ~300 sampling positions in the dispersion direction for segment A and ~500 for segment B. This provided us a better sampling than what was previously obtained with the slit/pinhole masks. The measured geometric distortions in the dispersion (X) and the cross-dispersion directions (Y) are shown in Figures 4 and 5. As can be seen by the histogram distribution of the identified line separations in Figure 6, we obtained a good sampling over the whole detector. Since we are re-sampling the data to correct for detector geometric distortions, it was decided to also force all corrected digital elements (DE) to represent exactly 24.0 µm in the cross-dispersion direction and 6.0 µm in the dispersion direction (compared to 25.9 µm and 5.9 µm, on average, as measured from the pinhole mask). To verify the results, different Pt-Ne spectra images were processed using the geometric distortion maps described above. The new line positions were then measured and compared to the expected positions in both X and Y. The calculated standard deviations of the residuals were ~0.9 DE in X (~5.4 µm, Figure 7) and ~0.4 DE in Y (~9.6 µm), significantly smaller than the spectral resolution element size of ~36x240 µm (6x10 DE). A detailed wavelength solution was performed on a set of corrected Pt-Ne spectra with more than 50 lines identified for segment A and more than 80 lines for segment B. The linear fit for the wavelength solution over the whole detector resulted in an RMS residual of 0.84 DE for the segment A and 0.54 DE for segment B. For the G160M grating, this corresponds to average RMS error of ~7 mÅ. The full geometric distortion correction maps for the COS FUV segment ‘A’ are shown in Figure 8 (X, dispersion direction) and Figure 9 (Y, cross-dispersion direction). Figure 3. Overlay of the Pt-Ne spectra for all FPSPLIT and PSA/WCA positions. Figure 10.Color map of the geometric correction along the cross-dispersion axis for segment A. Conclusions From the residuals obtained as described above, it is clear that this method for correcting geometric distortion works well. These residuals with our new method are smaller than what was originally obtained using the slit and pinhole masks. The one factor to keep in mind is that we are mapping the detector space to what is expected from the ray-trace analysis of the system (with the re-scaling of the digital element size). This is slightly different than the real physical space of the detector. For spectroscopic work, the data is normally rescaled to a linear wavelength dispersion and this method does just that, thus combining both re-samplings into one processing step. Figure 9.FUV segment A flat-field image. The top image is the uncorrected image, while the bottom image has been geometrically corrected. The cross-dispersion (Y) distortions are particularly obvious in the left third of the detector segment. The dispersion direction (X) distortions are more subtle, but can be noticed by the variable spacing of the grid wire shadows (mainly on the left side of the detector segment). Figure 5.Measured distortions for all PSA positions for segment A along the cross-dispersion axis. X X (DE) X DY (DE) Figure 6.Histogram of the separation between consecutive identified spectral lines. X Line Separation (DE) X Number of Lines Figure 7.Residuals of line centers from expected position after the geometric correction was applied (  =0.9 DE) X Digital Element (DE) Error X Number of Lines