New Tools and Observing Modes for Hubble

Jay Anderson (jayander[at]stsci.edu), Varun Bajaj (vbajaj[at]stsci.edu), Charles R. Proffitt (proffitt[at]stsci.edu), Jenna Ryon (ryon[at]stsci.edu) Elena Sabbi (sabbi[at]stsci.edu), Kailash Sahu (ksahu[at]stsci.edu, Ravi Sankrit (rsankrit[at]stsci.edu), Clare Shanahan (cshanahan[at]stsci.edu), and Daniel E. Welty (dwelty[at]stsci.edu)

Abstract

More than nine years after the most recent servicing mission to Hubble, the instrument teams at the Institute are continuing to update the capabilities of the four operating instruments with new observing modes and tools.  A new searchable database in MAST gives users the ability to construct detailed point-spread function (PSF) libraries to create accurate models of the PSF behavior on the UVIS channel on WFC3. A new observing mode on STIS enables high signal-to-noise, moderate-resolution spectra of bright targets. New observing modes on COS improve the observing efficiency and permit background-limited science at short wavelengths. Careful study of the pixel stability in the ACS Wide-Field Channel mitigates the effect of accumulated radiation damage and enables more pixels to be used for science.

WFC3 Team Enables Lookup Search of 22 Million Point-Spread Functions

The WFC3 team has enabled a lookup tool to aid observers seeking detailed information about the point-spread function of the UVIS channel. The team has identified nearly 22 million high signal-to-noise images of non-saturated point sources in WFC3/UVIS observations. The dataset was collected between May 2009 and May 2017, and can now be downloaded from the MAST portal using the “WFC3 UVIS PSF advanced search” interface. The database is updated yearly to include new sets of non-proprietary observations, with the last update in Spring 2018. Users can search the database to select sources based on different parameters, such as filter, telescope focus level, position on the detector, signal to noise, background value, exposure time, date, Right Ascension and Declination, FGS lock, and UVIS aperture. Figure 1 displays the WFC3 UVIS PSF search interface.

In a single submission, users can download up to 50,000 sets of PSFs, each set consisting of 21 × 21 pixel cutouts from the original UVIS (_raw,_flt, and _flc) files. They can also download a table containing the relevant details (rootname, filter, x/y position, flux, obs-date, etc.) of up to 500,000 PSFs in a single submission. The images are intended for those projects that would benefit from an accurate understanding of the UVIS point-spread function and its dependence on the telescope focus on a specific location, but do not have enough sources to create an accurate model. In the coming year, the team will release a similar database for the WFC3/IR channel, along with python-based data-analysis tools to assess the focus level of an image and to perform focus-dependent, PSF-fitting analysis in crowded fields.

WFC3 UVIS PSF Database
Figure 1: Example database search of the WFC3 UVIS Point-Spread Function Lookup in MAST. Filtering columns can be selected on the left, while selection parameters for each field can be specified on the right.

STIS Provides New Observing Mode for Spatial Scanning

Spatial scanning with the STIS CCD is a recently enabled, available-but-unsupported mode for obtaining moderate resolution, high S/N ratio spectra of relatively bright targets. This technique trails the target in the spatial direction within one of the long STIS apertures. Although high S/N-ratio spectra can also be obtained by deliberately saturating the CCD detector at a fixed pointing (e.g., Gilliland et al. 1999), spatial scans provide several distinct advantages. In particular, spatial scans should allow better averaging over flat-field variations, more robust removal of bad pixels and cosmic rays, and more complete correction for the fringing seen in the CCD spectra beyond 7000 Å. There are several possible scientific applications for this new observing mode. For example, spatial scans can enable more reliable detection of weak stellar and interstellar absorption features—particularly at red and near-IR wavelengths, where ground-based observations can be severely compromised by strong telluric absorption and where the fringing in pointed CCD observations can be difficult to remove. Spatial scans may also enable the very accurate monitoring of stellar fluxes, both broad-band and in narrower spectral intervals, that is needed for characterizing transiting exoplanets and their atmospheres.

Several programs have begun to explore the technical aspects and scientific capabilities of this new observing mode. In one calibration program (PID 15383; PI C. Proffitt), long-trailed images of the white dwarf GRW+70 5824 and trailed G750L spectra of the well-studied exoplanet host star 55 Cnc were obtained in order to check the alignment of the trails with the apertures and the alignment of the fringing pattern in the trailed spectra with that in the flat-field exposures. In addition, two sets of ten shorter trailed G750L spectra of 55 Cnc were obtained through the wide 52 × 2 aperture, in order to assess the consistency of the fluxes derived from those spectra. Preliminary analyses of these data have yielded some encouraging results regarding the alignment and flux stability of the trailed spectra; see the 2018 July STIS STScI Analysis Newsletter for more details. Three GO programs (PIDs 14705, 15429, and 15478; PI M. Cordiner) have obtained trailed G750M/9336 spectra of a number of heavily reddened stars, in an attempt to detect and characterize several weak diffuse interstellar bands attributed to C60+. The absence of telluric absorption and the effective removal of the fringing pattern in these trailed STIS spectra (Figure 2) have enabled very high S/N ratios in the final extracted spectra (e.g., S/N ~ 600–800 for BD+63 1964; see Figure 1 in Cordiner et al. 2017).

Example of spatial scanning with STIS.
Figure 2:  Trailed G750M/9336 exposure of a bright star from program 14705 illustrating the technique of spatial scanning for STIS. In these images, wavelength increases to the right, and the star was scanned in the vertical direction along the length of the narrow 52 × 0.1 aperture. At the left is the raw image.  In the center, the fringing pattern has been effectively removed by dividing by a contemporaneous flat-field image—allowing reliable detection of the weak absorption features (vertical dark bands). Instability in the trail rate and the effects of jitter can produce noticeable flux variations as a function of position along the trail direction, however. At the right, those variations have been removed (along with most cosmic rays and hot pixels). The corresponding 1D extracted spectrum is shown in Cordiner et al. (2017).

COS Creates New FUV Modes

Improved Observing Efficiency and Background-Limited Science

Starting in Cycle 26, observers using the Cosmic Origins Spectrograph will have two new far-ultraviolet (FUV) observing modes: G160M/1533 and G140L/800 have recently been commissioned. The G160M/1533 mode extends the G160M coverage by 44 Å towards shorter wavelengths to overlap with the longest wavelengths of the G130M/1222 mode. This allows for high signal-to-noise (S/N), medium-resolution spectra to be obtained over a broad range of wavelengths using only two settings, G130M/1222 and G160M/1533. Reaching high S/N requires that observations be obtained at all four FP-POS, which is possible for the G130M/1222 and G160M/1533 modes, but not for the existing G130M/1291 mode (Figure 3, top panel). Covering the full FUV bandpass at high S/N without the new mode required an additional observation using the G130M/1327 setting. Therefore, by reducing the number of settings needed, G160M/1533 offers a substantial saving in exposure time.

The G140L/800 mode, originally explored in a Cycle 19 Guest Observer Calibration Program (PID: 12501, PI: S. McCandliss), places a broad range of wavelengths (800–1950 Å) on segment A of the COS FUV channel. With the spectrum on a single segment, there is no gap in the wavelength coverage as happens with the G140L/1280 mode. The G140L/800 mode has been optimized to reduce the astigmatic height of the spectrum in the region below ≈1100 Å, allowing for a reduced extraction height, and thereby a decreased detector background. Correspondingly higher S/N is achievable at these wavelengths compared to spectra obtained on segment B using the G140L/1280 mode. Spectra obtained using these two modes are in Figure 3 (bottom panel), and show the difference in astigmatic heights, which are 15–20 pixels for G140L/800, and about twice that (30–40 pixels) for G140L/1280 in the 900–1100 Å wavelength region. The goal is to obtain a flux calibration accuracy of 10–20% at these short wavelengths with the new mode.

COS wavelength coverage
Figure 3: Top panel – COS FUV wavelength ranges for each setting, with the new modes labeled in bold. The four horizontal lines for each setting indicate the four FP-POS settings. The diagonal black lines are the wavelengths that fall on gain-sagged regions of the detector due to Ly-α on G130M/1291. The overlap region between the G130M/1222 and the new G160M/1533 mode is indicated by the green rectangle. Note that the long-wavelength limit of 1800 Å has been artificially imposed for display purposes. Bottom panel – COS 2D spectra of white dwarf WD0308-565 obtained using G140/800/FUVA (left) as part of the flux calibration program for the new mode (PID 15483), and G140L/1280/FUVB (right) as part of the LP4 flux calibration program (PID 14910). The spectra are shown in detector coordinates, and the display ranges have been chosen to align the profile centers along the cross-dispersion (YFULL) axis. The smaller astigmatic height for G140L/800 is apparent.

ACS Saves Pixels for Science

As radiation damage to the ACS Wide Field Channel (WFC) detectors accumulates over time, pixels with anomalously high dark current rates become increasingly prevalent. These “hot” and “warm” pixels are flagged in the DQ extensions of science images during calibration and typically discarded by users. In 2017, an analysis of pixel stability in dark images over the history of ACS determined that the vast majority of hot and warm pixels are stable over time (Borncamp et al. 2017). In recent data, about 1.3% of total detector pixels are considered hot, whereas 0.002% of the total are hot and unstable. While hot, stable pixels are noisier than normal pixels due to elevated dark current, they can be subtracted accurately from science images during dark correction. Therefore, they can be “saved” for scientific use, i.e., no longer treated as unusable. Pixels that are determined to be unstable cannot be reliably removed during calibration, and are now assigned the DQ flag 32. Unstable pixels, which make up about 0.12% of the detector, should be ignored during scientific analysis. Figure 4 shows the stability metric calculated for each pixel as a function of pixel intensity in dark images from two anneal intervals.

Hot pixels can also give rise to hot columns during readout of the ACS/WFC detectors. This occurs because the readout time of a full-frame ACS/WFC image is about 100 seconds, allowing significant dark current to be deposited by hot pixels during each parallel transfer of charge packets. A column-based stability analysis, similar to that for individual pixels, properly identifies stable and unstable columns (Ryon et al. 2017b). Just as with hot/warm individual pixels, the stable hot columns can also be accurately subtracted during bias correction, even though they are noisier than normal columns. Unstable hot columns cannot be accurately subtracted during bias correction, and are now assigned the bias structure DQ flag 128, which allows users to discard these columns during later analysis.

Stability of pixels in ACS WFC.
Figure 4: WFC dark current stability analysis for calibration dark images taken during month-long periods between anneals prior to (left) and after (right) Servicing Mission 4 (SM4). The vertical axis shows the stability metric, the ratio of measured to Poisson-estimated variance of each pixel’s intensity in the set of dark images; the horizontal axis shows the mean pixel intensity. Each point is color-coded according to the density of pixels located in a given region of parameter space. The differences between the panels stem from the addition of post-flash electrons to the dark images from January 2015 onwards. The dotted green curve is the stability threshold: above this curve (red arrows) pixels are rejected as unstable in the data quality (DQ) extension during calibration. All other pixels, including most hot pixels (green vertical line), are now retained as usable (albeit noisy). Note that saturated pixels (indicated by purple arrows) are nominally stable, but are independently flagged as unusable.

References

Borncamp, D., Grogin, N., Bourque, M., & Ogaz, S. 2017, Pixel History for Advanced Camera for Surveys Wide Field Channel, ACS ISR 2017-05, STScI

Cordiner, M. A.,  et al. 2017, ApJ, 843L, 2

Gilliland, R. L. et al. 1999, PASP, 111, 1009

Ryon, J. E., Grogin, N. A., & Coe, D. 2017, Accounting for Readout Dark in ACS/WFC Superbiases, ACS ISR 2017-13, STScI