COS FUV Dispersion Solution Verification at the New Lifetime Position

SPACE TELESCOPE SCIENCE INSTITUTE Operated for NASA by AURA Instrument Science Report COS 2013-06(v1) COS FUV Dispersion Solution Verification at the New Lifetime Position Paule Sonnentrucker 1, Julia Roman-Duval 1, Justin Ely 1, Cristina Oliveira 1, Charles Proffitt 2, and Alessandra Aloisi 1 1 Space Telescope Science Institute, Baltimore, MD 2 Computer Science Corporation, STScI, Baltimore, MD 10 June, 2013 ABSTRACT Program 12805, Second COS FUV Lifetime Position: Wavelength and Resolution Calibration (PI: Roman-Duval), was executed in the Summer and Fall 2012 in order to determine the spectral resolutions (Calibration program FCAL2) and to cross-check the dispersion solutions of the COS FUV modes (Calibration program FCAL1) at the second lifetime position. In this report, we describe the analysis we performed to verify the dispersion solutions of the M gratings using data collected in 12805 and of the L grating using data obtained in enabling program 12796 (FENA3: Second COS FUV Lifetime Position: Focus Sweep Enabling Program ; Oliveira et al. 2013). Our results indicate that the dispersion solutions for all FUV modes fulfill the operation requirements described in Osten et al. (2013) at LP2 as well. The analysis pertaining to verification of the spectral resolution will be discussed in Roman-Duval et al. (2013). Contents Introduction (page 2) COS FUV Dispersion Solution Requirements (page 2) Operated by the Association of Universities for Research in Astronomy, Inc., for the National Aeronautics and Space Administration

Observations and Data Analysis (page 6) Results (page 7) Conclusions (page 12) Change History (page 13) References (page 13) 1. Introduction Since the beginning of its on-orbit operations the COS FUV detector is experiencing sustained gain sag at the location used for science operations. Starting in Cycle 18, HST users were urged to use multiple FP-POS positions when observing with the COS FUV detector in order to minimize the effects of localized gain sag on the science data quality at the original lifetime position (aka lifetime position 1 or LP1). Owing to the accelerated rate at which gain sag affected the FUV data on local scales and to maintain optimum science data quality, a series of exploratory and enabling programs where executed during Cycle 18 (see Osten et al. 2013; Oliveira et al. 2013; Proffitt et al. 2013) in order to determine the number of additional lifetime positions available for FUV science operations and to select the next best lifetime position at which to continue science operations (aka lifetime position 2 or LP2). The move to LP2 took place on July 23rd, 2012. To verify that all COS science operations in the FUV fulfilled the missions science operation requirements three calibration programs were executed shortly thereafter: program 12805 (FCAL2/FCAL1), program 12806 (FCAL3) and program 12807 (FCAL4). Program 12805 was designed to verify both the spectral resolution (FCAL2) and the dispersion solutions (FCAL1) using external spectra obtained with the COS G130M and G160M gratings. This ISR describes the analysis performed to cross-check the COS FUV dispersion solutions for the M and L gratings at the new lifetime position and discusses the accuracy achieved for those solutions. Verification of the dispersion solution for the G140L grating at LP2 is also performed and reported here using data obtained in enabling program 12796 (FENA3: Oliveira et al. 2013) 2. COS FUV Dispersion Solution Requirements at LP2 Analysis of thermal vacuum data taken in 2003 (TV03) and 2006 (TV06) revealed that the dispersion solutions of the spectra obtained through the primary science aperture (PSA) were different from those obtained through the wavelength calibration aperture (WCA). As a result, on-orbit data obtained through the WCA cannot be used directly to derive the on-orbit dispersion solutions for spectra obtained with the PSA. To calibrate on-orbit science data, the dispersion solutions derived in TV03 are used instead but need to be placed in the on-orbit reference frame. To do so the WCA-to-PSA offsets need to be derived for all gratings and central wavelengths. During SMOV, Program 11487 was executed to derive the WCA-to-PSA offsets for all COS FUV configurations in order to Instrument Science Report COS 2013-06(v1) Page 2

place the TV03 dispersion solutions in the on-orbit frame at lifetime position 1 (see Oliveira et al. 2010, for details). The requirements for calibration of the COS FUV dispersion solutions at lifetime position 2 (LP2) were based on the results obtained at LP1 during SMOV, and are as follows (see also Osten et al., 2013). 1) The zero-point offsets (PSA-to-WCA) in the dispersion solutions for each FUV grating for each central wavelength setting shall be obtained as needed. 2) Internal wavelength calibration spectra for each FUV grating at each central wavelength (cenwave) setting shall be obtained. 3) The location of the spectra for each FUV grating shall be measured. This is done by observing an astronomical target and acquiring a spectrum using G130M, G160M and G140L gratings. Program FCAL1: FUV Dispersion Solution Verification (PI: Sonnentrucker) was designed to address those requirements. Since internal wavelength calibration spectra for each FUV grating at each cenwave were being obtained through TAGFLASH as part of lifetime calibration program 12806 (FCAL3) COS FUV Lifetime Position: Flux and Flat Field Calibration and TDS transfer (PI: Massa), Requirement 2 of FCAL1 was achieved using FCAL3 data alone. As a result, no request for internal orbits was attached to program FCAL1. The selection of an appropriate external target to address Requirement 1 and 3 was based on the following criteria: a) a FUV-bright continuum source showing unblended, optically thin or partially optically thick ISM absorption lines; b) S/N ratio of at least 30 throughout the bandpass; c) as high an ISM line density as possible throughout the bandpass; and d) existence of archival STIS E140M data for line identification and dispersion solution verification. The early type star AV75 (O5III) located in the Small Magellanic Cloud (SMC) was identified as an appropriate target to derive the spectral resolution of the FUV modes at lifetime position 2 (FCAL2; PI; Roman-Duval). AV75 was also used in enabling program 12796 (FENA3: Second COS FUV Lifetime Position: Focus Sweep Enabling Program ; see Oliveira et al. 2013), hence providing G140L/1105 spectra useful to verify the disperison solution of the L grating at LP2. After inspection of the AV75 archival STIS E140M, AV75 was also adopted as external target for FCAL1. Owing to the similiarities in observing requirements between FCAL2 and FCAL1, the possibility to merge both calibration efforts under one program was considered in order to maximize the return of the lifetime calibration undertaking while minimizing the cost in terms of external orbits. The FCAL2 requirements (see Roman-Duval et al. 2013) could be verified with observations of AV75 with the M gratings at the extreme central wavelengths alone. Hence, one needed to verify that the zero-point offsets at the intermediate cenwaves could be derived based on the zero-point offsets measured at the extreme cenwaves alone to fulfill the FCAL1 requirements. A comparison of the PSA-to-WCA offsets derived from TV03 and SMOV was thereby performed (see Béland et al. 2011; Oliveira et al. 2010). Figures 1 and 2 show the results of the TV03-SMOV comparisons for the G130M and G160M gratings, respectively. For both gratings, the data indicate that estimating the zero-point offsets at the intermediate cenwaves from a linear fit to the Instrument Science Report COS 2013-06(v1) Page 3

zero-point offsets measured at the extreme cenwaves would lead to uncertainties at the intermediate cenwaves no greater than half a resolution element for the standard FUV settings. Since the latter accuracy is consistent with the operational requirements, program FCAL2 and FCAL1 were merged together under program 12805 (PI: Roman- Duval.) A 2-orbit contingency visit was attached to FCAL1 (PI: Sonnentrucker), however, in order to account for the possibility that the PSA-to-WCA offsets measured in 12805 would be greater than half a resolution element (or 3 pixels), thus requiring execution of observations using the M gratings at the intermediate cenwaves to measure the PSA-to-WCA offsets at LP2. Instrument Science Report COS 2013-06(v1) Page 4

Distance (pixel) Cenwave Cenwave Figure 1: Measurements of the distance between the PSA and WCA spectra at each central wavelength of the G130M grating for segment B (left panel) and segment A (right panel) for data taken in TV03 (green) and data taken in SMOV (red). The separations are in units of pixel. Linear fits to those distances based on the measurements obtained at the extreme central wavelengths alone (λ1291 and λ1327) are also overplotted in green (TV03) and red (SMOV) in each panel. Reported at the bottom of each panel are the central wavelengths showing the largest deviation from the linear fits and their corresponding deviation value in pixels for TV03 (green) and SMOV (red). Distance (pixel) Cenwave Cenwave Figure 2: Measurements of the distance between the PSA and WCA spectra at each central wavelength of the G160M grating for segment B (left panel) and segment A (right panel) for data taken in TV03 (green) and data taken in SMOV (red). The separations are in units of pixel. Linear fits to those distances based on the measurements obtained at the extreme central wavelength alone (λ1577 and λ1623) are also overplotted in green (TV03) and red (SMOV) in each panel. Reported at the bottom of each panel are the central wavelengths showing the largest deviation from the linear fits and the corresponding deviation value in pixels. Instrument Science Report COS 2013-06(v1) Page 5

3. Observations and Data Analysis 3.1 Observations Program 12805 consisted of a 4-orbit visit to acquire COS FUV G130M and G160M spectra of the external SMC source AV75 using the extreme central wavelength (λ1291, λ1327 for G130M and λ1577, λ1623 for G160M) alone. Spectra at all 4 FP-POS positions were obtained at each central wavelength in order to achieve a S/N ratio in excess of 60 per resolution element in the FP-POS combined spectra. Such high S/N ratios are warranted to verify both the LSF profiles at LP2 (Roman-Duval et al. 2013) and cross-check the wavelength solutions for the M gratings. Visit 01 executed on July 27, 2012. Data in the G130M/1291 configuration was obtained successfully. However, a guide star re-acquisition failure resulted in the loss of all subsequent observations using the G130M/1327, G160M/1577 and G160M/1623 configurations. A repeat of the latter observations was granted (HOPR#71903) and the remaining observations executed successfully on Sept 12, 2012 as part of visit 51. The data for G140L was taken as part of program 12976 (FENA3: See Oliveira et al. 2013 for details). 3.2 Data Analysis The data obtained with the M gratings were processed with the standard CalCOS pipeline version 2.18.5 which uses the zero-point offsets and dispersion solutions appropriate for LP1 operations at each central wavelength. To validate the use of these solutions at LP2 and determine whether the zero-point offsets needed updating, a crosscorrelation analysis was performed in 3 steps. The methodology described below for the G130M data was also applied to the G160M and G140L data. 1) For each central wavelength, a cross-correlation of the exposure taken at FP-POS=3 with the exposures taken at FP-POS=1, 2 and 4 was performed to verify that the FP-POS combination (that produces the x1dsum spectra) did not introduce localized line smearing that could cause artificially large wavelength offsets, unrelated to the dispersion solution accuracy itself. The cross-correlation was performed by slicing each spectrum in 3Å-wide consecutive windows over the entire wavelength range and applying the cross-correlation function to each window. There is no a-priori window selection when the cross-correlaiton is performed. The IDL procedure cross-correlate.pro was used for this exercise. The test showed that localized offsets measured between FP- POS=3 and FP-POS=1, 2 and 4 where smaller than half an resolution element (or 3 pixels) in all cases. As a result, the x1dsum spectra were used for the remainder of the analyses. To estimate the accuracy of the cross-correlation function, each x1dsum spectrum was cross-correlated with itself as described above. For all 3Å-wide windows, the offset distribution over the entire spectral range shows a mean value of 0.0 pixel and standard deviation of 0.5 pixel. An accuracy of 0.5 pixel was thus adopted for the remainder of cross-correlation analysis. 2) The FP-POS combined spectrum (x1dsum) taken at central wavelength λ1327 was cross-correlated with the x1dsum spectrum taken at central wavelength λ1291 by stepping through the spectra in 3Å-wide windows and applying the cross-correlation function to each window, as above. For each window, the maximum of the cross- Instrument Science Report COS 2013-06(v1) Page 6

correlation obtained from a quadratic fit to the function (named corcoef hereafter) was recorded separately along with its corresponding pixel position, wavelength value and offset value (named Offset hereafter). 3) The G130M dispersion solutions and zero-point offsets were independently verified by cross-correlating the x1dsum spectra obtained at λ1291 and λ1327 with archival STIS data of AV75 that were collected with the E140M grating in Cycle 7 program 7437. The STIS archival data were coadded and convolved with the COS LSF derived at LP2 and optimized for each central wavelength (Roman-Duval et al. 2013) before crosscorrelation. 4. Results 4.1 G130M Gratings Figure 3 display the results of the cross-correlation of the G130M/1327 x1dsum spectrum with G130M/1291 x1dsum spectrum for both segments. The top panel displays the Offset (in pixel) versus corresponding wavelength for all windows where the maximum of the correlation function is greater or equal to 0.70 (corcoef 0.70). The Offset represents the shift in pixels that needs to be applied to the data at λ1327 so that they align to the data in the λ1291 spectrum for all windows where corcoef 0.70. The red-filled diamonds point out the windows where the cross-correlation was driven by the presence of ISM absorption lines alone. For all the other windows, the crosscorrelation was driven by the presence of stellar absorption lines (black diamonds). The two horizontal dotted lines mark the Offset range consistent with the operational requirements of 3 pixels (Section 2) modulo the 0.5 pixel uncertainty in our analysis. Because of the variable nature some stellar lines can exhibit for AV75- a variability readily seen in the scatter around the mean value of Offset in the top panel as the observations were taken at different epoch due to the failure in Visit 01- the windows showing ISM absorption lines alone were used to verify that the operation requirements are met at LP2 as they constitute a more robust analysis sample. For clarity, the middle panel displays the measured Offset versus wavelength between λ1327 and λ1291 for the windows with ISM lines alone. Note that lowering corcoef below a value of 0.7 does not increase the number of windows containing ISM lines but only increases the number of windows dominated by stellar lines or noise, hence, increasing the scatter around the mean offset. The bottom panel displays the G130M/1291 x1dsum spectrum in black and the G130M/1327 x1dsum spectrum in red. Figures 4 and 5 display the results of the cross-correlation analysis of the λ1291 spectrum and the λ1327 spectrum with the archival STIS E140M spectrum, respectively. As in Figure 3, the middle panels shows the Offset distribution for the windows containing ISM absorption lines alone. The combined results displayed in Figures 3, 4 and 5 pertaining to the analyses performed on the G130M data obtained at LP2 in program 12805 lead to the following conclusions. 1) The cross-correlation analysis using COS data obtained at the extreme central wavelengths (λ1327 vs λ1291) at LP2 yields Offset 3.0 ± 0.5 pixels for most lines, as Instrument Science Report COS 2013-06(v1) Page 7

predicted in Section 2. The 3 discrepant points originate from the unresolved saturated O I line (λ1301) and from the blend between the N I line system around 1200Å and Lyman α. The proximity of the N I line system to the bottom of the Lyman α absorption introduces errors larger than average that are routinely seen. As a result, the zero-point offsets for the dispersion solutions for data obtained at the intermediate central wavelength (λ1300, λ1309 and λ1318) can be derived from a linear fit to the zero-point offsets measured at the extreme central wavelengths. 2) Cross-correlations of the COS G130M λ1291 and λ1327 data with the STIS E140M archival data show that the accuracy of the dispersion solutions at LP2 is within the mission s requirements of 3 pixels or half a resolution element (resel). As a result, (i) the dispersion solution relationships used to calibrate the COS FUV G130M data at LP1 are adequate to calibrate COS FUV G130M data at LP2 and (ii) the zero-point offsets for the G130M/1291 and G130M/1327 dispersion solutions do not need to be updated at LP2. Subsequently, the zero-point offsets of the dispersion solutions at the intermediate central wavelengths (λ1300, λ1309 and λ1318) do not need to be updated either. Figure 3: Results of the G130M λ1327-λ1291 cross-correlation analysis. Top Panel: Offset distribution for all windows with corcoef 0.70. Middle Panel: Offset distribution for the windows containing ISM absorption lines alone. Bottom Panel: G130M/1291 spectrum of AV75 taken at LP2 with program 12805. The G130M/1327 spectrum is overplotted in red. Offsets greater than 5 pixels measured for N I line system around 1200Å are introduced by the proximity of this line system to the bottom of the Lymanαabsorption where errors larger than average are routinely seen. Instrument Science Report COS 2013-06(v1) Page 8

Figure 4: Results of the G130M/1291-STIS E140M cross-correlation analysis. Top Panel: Offset distribution for all windows with corcoef 0.70. Middle Panel: Offset distribution for the windows containing ISM absorption lines alone. Bottom Panel: G130M/1291 spectrum of AV75 taken at LP2 with program 12805. The STIS E140M spectrum convolved with appropriate COS LSF for LP2 is overplotted in red. Instrument Science Report COS 2013-06(v1) Page 9

Figure 5: Results of the G130M/1327-STIS E140M cross-correlation analysis. Top Panel: Offset distribution for all windows with corcoef 0.70. Middle Panel: Offset distribution for the windows containing ISM absorption lines alone. Bottom Panel: G130M/1327 spectrum of AV75 taken at LP2 with program 12805. The STIS E140M spectrum convolved with appropriate COS LSF for LP2 is overplotted in red. Offsets greater than 5 pixels measured for N I line system around 1200Å are introduced by the proximity of this line system to the bottom of the Lymanαabsorption where errors larger than average are routinely seen. Instrument Science Report COS 2013-06(v1) Page 10

4.1 G160M Grating Figure 6 displays the results of the cross-correlation of the G160M/1577 x1dsum spectrum with G160M/1623 x1dsum spectrum for both segments. The top panel displays the Offset (in pixel) versus corresponding wavelength for all windows where the maximum of the correlation function is greater or equal to 0.70 (corcoef 0.70). The Offset represents the shift in pixels that needs to be applied to the data at λ1623 so that they align to the data in the λ1577 spectrum. As previoulsy, the red-filled diamonds point out the windows where the cross-correlation was driven by the presence of ISM absorption lines alone. The two horizontal dotted lines mark the Offset range consistent with the operation requirements (Section 2) modulo the 0.5 pixel analysis uncertainty. Note that lowering corcoef below a value of 0.7 does not increase the number of windows containing ISM lines but only increases the number of windows dominated by stellar lines or noise, hence, increasing the scatter around the mean offset. Figures 7 and 8 display the results of cross-correlation analysis of the G160M/1577 spectrum and the G160M/1623 spectrum with the archival STIS E140M spectrum of AV75, respectively. As in Figure 6, the middle panel shows the Offset distribution for the windows containing ISM absorption lines alone. The combined results displayed in Figures 6, 7 and 8 pertaining to the analyses performed on the G160M data obtained at LP2 in program 12805 lead to the following conclusions. 1) The cross-correlation analysis using COS data obtained at the extreme central wavelengths (λ1577 vs λ1623) at LP2 yields Offset 3.0 ± 0.5 pixels, as predicted in Section 2. As a result, the zero-point offsets for the dispersion solutions of data obtained at the intermediate central wavelength (λ1589, λ1600 and λ1610) can be derived from a linear fit to the zero-point offsets measured at the extreme central wavelengths. 2) Cross-correlations of the COS G160M λ1577 and λ1623 data with the STIS E140M archival data show that the accuracy of the dispersion solutions at LP2 is within the mission s requirements of 3 pixels or half a resolution element (resel). As a result, (i) the dispersion solution relationships used to calibrate the COS FUV G160M data at LP1 are adequate to calibrate COS FUV G160M data at LP2 and (ii) the zero-point offsets for the G160M/1577 and G160M/1623 dispersion solutions do not need to be updated at LP2. Subsequently, the zero-point offsets of the dispersion solutions at the intermediate central wavelengths (λ1589, λ1600 and λ1610) do not need to be updated either. Instrument Science Report COS 2013-06(v1) Page 11

Figure 6: Results of the G160M/1577-G160M/1623 cross-correlation analysis. Top Panel: Offset distribution for all windows with corcoef 0.70. Middle panel: Offset distribution for the windows containing ISM absorption lines alone. Bottom panel: G160M/1577 spectrum of AV75 taken at LP2 with program 12805. The G160M/1623 spectrum is overplotted in red. Instrument Science Report COS 2013-06(v1) Page 12

Figure 7: Results of the G160M/1577-STIS E140M cross-correlation analysis. Top Panel: Offset distribution for all windows with corcoef 0.70. Middle Panel: Offset distribution for the windows containing ISM absorption lines alone. Bottom Panel: G160M/1577 spectrum of AV75 taken at LP2 with program 12805. The STIS E140M spectrum convolved with appropriate COS LSF for LP2 is overplotted in red. Instrument Science Report COS 2013-06(v1) Page 13

Figure 8: Results of the G160M/1623-STIS E140M cross-correlation analysis. Top Panel: Offset distribution for all windows with corcoef 0.70. Middle Panel: Offset distribution for the windows containing ISM absorption lines alone. Bottom Panel: G160M/1623 spectrum of AV75 taken at LP2 with program 12805. The STIS E140M spectrum convolved with appropriate COS LSF for LP2 is overplotted in red. 4.3 G140L Grating The G140L/1105 data were obtained as part of enabling program 12796: Second COS FUV Lifetime Position: Focus Sweep Enabling Program (FENA3) (See Oliveira et al., 2013). Contrary to the M gratings, data for the L grating were obtained at FP-POS=3 alone for the same external target AV75. We used the data obtained at the focus value closest to that adopted at LP2 and recorded in exposure: lbx503kpq. The crosscorrelation analysis is identical to that performed for the M gratings. However owing to the lower resolving power, the G140L spectrum was cross-correlated with the archival STIS E140M convolved with the G140L LSF by slicing the spectrum in 10Å-wide consecutive windows over the entire wavelength range. There is no a-priori window selection when the cross-correlaiton is performed. In order to reduce the scatter due both to stellar line variability in AV75 and to the lower data quality compared to the M gratings, Figure 9 displays the results of the cross correlation analysis for all windows where the maximum of the function is greater or equal to 0.80 (corcoef 0.80). Lowering corcoef below a value of 0.8 did not increase the number of windows Instrument Science Report COS 2013-06(v1) Page 14

containing ISM lines but only increased the number of windows dominated by stellar lines or noise, hence, increasing the scatter around the mean offset. As in previous figures, the red-filled diamonds indicate the windows for which the crosscorrelation was driven by the presence of ISM absorption lines alone. The analysis yields Offset 3 ± 0.5 pixels, a result consistent with mission specifications. Considering the limited data-points available, this analysis suggests that, as for the M gratings, the zero-point offset and the G140L/1105 dispersions solutions derived at LP1 are adequate to perform science operations at LP2. Figure 9: Results of the G140L/1105-STIS E140M cross-correlation analysis. Top Panel: Offset distribution for all windows with corcoef 0.80. Middle Panel: Offset distribution for the windows containing ISM absorption lines alone. Bottom Panel: G140L/1105 spectrum of AV75 taken at LP2 with program 12796. The STIS E140M archival spectrum convolved with the G140L LSF is overplotted in red. 5. Conclusion Cross-correlation analyses between the data obtained in COS programs 12805 and 12796 and archival STIS E140M data showed that the dispersion solutions used to calibrate the COS FUV G130M, G160M and G140L data at lifetime position 1 (LP1) are also adequate to calibrate the data obtained at lifetime calibration 2 (LP2) and are accurate to within half a resolution element (or 3 pixels). Our results further show that the zero-point offsets of the dispersion solutions for the M and L gratings do not need to be updated at LP2. As a result, the 2-orbit contingency visit built-in FCAL1 did not need to be executed. Considering the proven stability of the dispersion solutions when the spectra were moved from LP1 to LP2 on the COS detector, we recommend that a similar Instrument Science Report COS 2013-06(v1) Page 15

calibration approach be followed when moving operations to future lifetime positions. LP2 should be used as the reference position for calibration of operations at the next lifetime position (LP3). Acknowledgements Change History for COS ISR 2013-06 Version 1: 10 June 2013 Original Document References Osten, R., et al. 2013, Techncal Instrument Report COS 2013-02, Requirements and Preparations for a COS FUV Lifetime Position Move. Oliveira, C., et al. 2010, Instrument Science Report COS 2010-06, SMOV: COS FUV Wavelength Calibration. Oliveira C. et al. 2013, Instrument Science Report COS 2013-01, Second COS FUV Lifetime Position: Focus Sweep Enabling Program (FENA3) Béland, S., et al. 2011, Technical Evaluation Report COS-11-0050, PSA-WCA Separation Measurements for TV 2003 and 2006. Instrument Science Report COS 2013-06(v1) Page 16