3D-DASH: The Widest Near-infrared Hubble Space Telescope Survey

Drift And SHift (DASH) is an efficient technique to cover an extensive area using the NIR channel of Wide Field Camera 3 (WFC3) on the Hubble Space Telescope (Momcheva et al. 2017). In standard HST observations, guide stars are acquired for each new pointing. Acquiring a guide star takes approximately 10 minutes, which means that short exposures are only possible with huge overheads. This limits the total number of pointings that can be obtained within an orbit's visibility window. ¹⁸

The guide star acquisition limitation can be circumvented by acquiring a guide star for only the first pointing of an orbit and guiding with the three HST gyros for the remainder. This technique makes it possible to observe up to 8 WFC3 pointings in a single orbit, significantly increasing HST's large-scale mapping capabilities. Successfully used for the first time in the COSMOS-DASH survey (Mowla et al. 2019b) to image 0.49 deg² of the three stripes of the UltraVISTA field (McCracken et al. 2012), DASH remains the most efficient method for wide-area observations with WFC3.

During a standard guided exposure, the three HST gyros receive continuous corrections from the Fine Guidance Sensors (FGS). Turning off guiding stops the stream of corrections from the FGS, and the telescope begins to drift with an expected rate of 0001 − 0002 per second. In CCDs, this would lead to detrimental smearing of the image in a typical 5-minute exposure. However, the WFC3/IR detector can perform multiple nondestructive, zero overhead reads throughout the exposure. Setting the time between reads to 25 s or less lowers the drift between reads to ≤012—around a pixel (pixel scale =0129). The data obtained between the reads can be treated as independent 25 s exposures that can be drizzled to restore the full resolution of WFC3. In Momcheva et al. (2017) and Mowla et al. (2019b) we demonstrated that the resolution of the WFC3 camera is preserved in this process, and the measured structural parameters of the galaxies are consistent with those measured in guided observations (Cutler et al. 2022).

The DASH method was used in the Cycle 23 and 28 3D-DASH programs (Program ID: GO-14114 and GO-16259) to obtain 1256 WFC3 H₁₆₀ pointings in 157 orbits, covering an area of 1.35 deg² in the COSMOS field. The design of the mosaic was constrained by the available orientations. To ease scheduling no initial ORIENT constraints were imposed. After scheduling windows were assigned, the final mosaic was designed, constrained by the ORIENTs that were allowed within each of the windows. A consequence of this strategy is that the final mosaic has gaps, as the wide range of orientations precluded the design of a fully contiguous mosaic that also avoids severe overlaps between pointings. The Cycle 23 data were obtained between 2016 November and 2017 June, while the Cycle 28 data were obtained between 2020 December and 2022 March.

Each orbit consists of a guide star acquisition, followed by a single guided exposure. Next, the stream of corrections from the fine guidance sensor is turned off, and the telescope is moved to the second exposure. Guiding remains off for the next six exposures. The exposure times, and hence the number of independent reads within each exposure, were determined by the requirement to fit eight pointings in a single orbit.

Following the 2018 failure of one of the HST gyros, a new, previously dormant gyro was turned on. The new gyro exhibited extremely high bias at the start of operations, which necessitated certain changes in the observatory controls which led to differences between the GO-14114 (executed prior to the gyro switch) and GO-16259 (executed post-switch) observations. The orbital visibility in COSMOS post-switch is shorter by 290 s, the duration of guide star acquisitions and commanded offsets is increased and the observatory is not allowed to execute an offset during the 5 minutes after the guide star acquisition. The combined effect is a shorter length of the science exposure. With the additional constraints of the SPARS25 sampling sequence (exposure times must change by integer factors of 25 s) and avoiding buffer dump interruptions, we find a new optimum sequence of exposure times. The total exposure time in the orbit is 350 s shorter than in GO-14114 (1774 versus 2124 s). Figure 5 and Table 2 show the APT graphical representation of a single GO-16259 orbit and the sequence of events, respectively. These can be compared to Figure 2 and Table 1 in Momcheva et al. (2017).

Zoom In Zoom Out Reset image size

The 3D-DASH WFC3/F160W imaging will be supplemented by WFC3/G141 grism spectroscopy, executed in Cycles 28 and 29 in 159 orbits (see Figure 4 for the grism observation footprint). 3D-DASH will increase the area observed with infrared grism spectroscopy by almost an order of magnitude and map the spatially resolved emission lines of over a thousand star-forming galaxies at cosmic noon, providing crucial information on the buildup of disks and bulges of galaxies. Similar to the imaging, the grism spectroscopy will cover the ACS COSMOS field, observed efficiently in 1272 pointings using the DASH technique.

Owing to the shifts during the exposures, the reduction of DASH data is more complex than that of guided exposures. Here we provide a summary of the reduction procedures; we also refer to Momcheva et al. (2017) where the reduction of a subset of the 3D-DASH ¹⁹ data were first described.

Each DASH orbit consists of one guided and seven unguided exposures with offsets of roughly 2' between each. Momcheva et al. (2017) describe the overall mosaic strategy and detailed commanding instructions for the DASH visits. For uniformity of the data reduction, we process both guided and unguided exposures using the same analysis pipeline as outlined below.

The raw WFC3/IR images were downloaded from the Mikulski Archive for Space Telescopes (MAST) ²⁰ and processed into calibrated exposure ramps ("IMA" products) with the calwf3 pipeline after disabling the pipeline cosmic ray identification step (CRCORR=False). The IMA files provide a measure of the total charge on the detector sampled every 25 s over the duration of the exposure (253 or 278 s for exposures with NSAMP = 12 and 13, respectively). To reduce the degradation of the image quality of a given exposure due to the telescope drifts, we take image differences up the ramp and generate NSAMP − 2 essentially independent calibrated exposures with drifts now integrated over 25 s rather than the full exposure duration. ²¹ The properties of these difference images are essentially identical to normal calibrated WFC3/IR "FLT" products, though with slightly different noise characteristics. Taking image differences increases the effective read noise by a factor of $\sqrt{2}$ and the read noise of two adjacent image differences will be anticorrelated as the measured (noisy) flux of a given read appears as negative in the first difference image and positive in the second. In the case of guided exposures with no drifts, the image differences are equivalent to taking the pixel values of the last read minus the first, with read noise from just those two reads. The differences do not cancel out for sequences with drifts, where the pixel indices of the difference images are effectively shifted when they are combined into the output mosaic.

The DASH observations produce N × 7 × (NSAMP − 2) difference image "exposures," with N the number of orbits. These exposures are processed in an identical way to the N-guided exposures that are taken at the start of each orbit, and with all guided archival observations in the COSMOS field. We first compute an internal alignment of the visit exposures using sources detected in the images (both stars and galaxies), which corrects the DASH drifts between samples and small pointing errors typical of the guided sequences. We generate a small mosaic of the visit exposures to detect fainter sources and align these mosaics to galaxies in the F814W catalogs provided by the COSMOS collaboration (Koekemoer et al. 2007). Point sources are excluded from the catalog alignment as stars can have significant proper motions between the ACS COSMOS and DASH epochs. Since we do not identify cosmic rays in the DASH exposures at the pipeline level as with normal WFC3/IR exposures, we detect and mask the cosmic rays using the standard tools of the AstroDrizzle (Gonzaga 2012) package when creating the combined visit/pointing image (turning on cosmic ray identification is useful even for the guided exposures to mask unflagged hot pixels and weaker cosmic rays missed by calwf3). In the pilot COSMOS-DASH program, several sequences of eight pointings were broken up due to South Atlantic Anomaly (SAA) passages. There was typically a large offset of 10''–15'' between the "before" and "after" exposures, due to the spacecraft drift during the SAA passage; nevertheless, no observable degradation of the PSF was found in these sequences. However, the interruption of DASH orbits by SAA passages lowered the efficiency of the observatory because the orbit following the SAA passage was only 1/4 full. For this reason the planning team recommended that we avoid SAA passages and schedule all eight pointings within a single orbit.

The requirement of three functioning HST gyros is crucial for the optimum execution of the DASH technique. In recent years, multiple gyro failures have made DASH observations challenging. Here we show that the average drift between reads has increased from 002 in the pilot COSMOS-DASH survey in 2017 to 015 in 2021 and is over 05 per read for ∼5% of pointings as shown in Figure 7. The exact cause of this increase is not clear. This increased drift has resulted in a more irregular overlapping of pointings and some noticeable gaps in the mosaic, as can be seen in Figure 3. It has also resulted in increased background noise and thus a lower point-source depth than the pilot COSMOS-DASH program (see Section 3.4).

The full set of ∼5000 aligned DASH exposures, together with all other existing H₁₆₀ data in the COSMOS field, are then drizzled into a single, large mosaic using Astrodrizzle. The weight map scales with both the variations in exposure time and the impact of drifting across the final mosaic, where lower weight corresponds to higher drift and/or shorter exposure times. Both the science image and weight maps are drizzled to a pixel scale of 01 using a square kernel and with pixfrac = 0.8. The final mosaic of the 3D-DASH image is 53248 × 53248 pixels and is centered at RA = 10:00:28.6, DEC = +02:12:21.0. The image is shown in Figure 2 and the point-source depth map created from the weight map (using the average noise in 03 diameter apertures, details in Section 3.3) is shown in Figure 3. All our data products are available at MAST as a ²² High Level Science Product and ²³ here . The science images and their corresponding weight maps are available in three formats:

• 1.  
  Full 3D-DASH mosaic ²⁴
• 2.  
  3D-DASH mosaic (see footnote 22) divided into 256 (16 × 16) tiles
• 3.  
  Full DASH-only mosaic.

We created a grid of point-spread functions (PSF) for the 3D-DASH mosaic using Grizli ²⁵ to shift and drizzle HST empirical PSFs (Anderson et al. 2015) at the positions of 12335 stars brighter than I < 25 from the Hyper Suprime-Cam Subaru Strategic Program (Aihara et al. 2021). Some example position-dependent empirical PSFs and the stars at their positions are shown in Figure 6. These 12,335 stars sample the entire 3D-DASH mosaic area (see right panel of Figure 7), including the CANDELS region. We also provide a PSF generator tool ²⁶ to determine the PSF at any location within the 3D-DASH footprint.

While the position angle of each pointing is largely aligned, there are subtle variations visible in Figure 3 that motivate our decision to adopt a position-dependent PSF. When comparing the orientation of star spikes across the mosaic in Figure 5(b), we further see several cases where multiple orientation angles are coadded. The curves of growth, which show the fraction of light enclosed as a function of aperture size, for the DASH H₁₆₀ PSFs, normalized at 2'', are shown in the left panel of Figure 6 for both the synthetic PSF and the empirical PSF from the stars. The PSFs are in agreement with each other and with the encircled energy as a function of aperture provided in the WFC3 handbook when also normalized to a maximum radius of 2''. This demonstrates that the DASH technique does not introduce significant smoothing.

Zoom In Zoom Out Reset image size

To check the effect of the drift on the shape of the point-spread functions, we measured the axis ratios in the 12335 stars using GALFIT (Peng et al. 2010). While most of the stars have axis ratios >0.8, we notice that stars in certain regions demonstrate flattening see 7. These regions coincide with regions of higher drift, which affected some orbits of Cycle 28. The regions of high drift (>05 per pointing) account for 5.4% of the entire 3D-DASH area. This is reflected in the shape of the PSFs, with 4.3% of PSFs having an axis ratio <0.5.

Zoom In Zoom Out Reset image size

The depth of the mosaic is determined by the background noise, which is expected to scale with the size of the aperture that is used for photometry. To determine how the background noise scales with aperture size, we measure the distribution of counts in empty regions of increasing size within the noise-equalized F160W image. For each aperture size, we measure the flux in 10,000 apertures placed at random positions across the combined 3D-DASH mosaic with both the DASH and the "standard" depth pointings. DASH depths are selected with a weight <100, and "standard" depths have a weight >100. Apertures that overlap with sources in the detection segmentation map are randomly moved until they no longer overlap with a source. Figure 8 shows the distribution of flux counts for increasing aperture size ranging from 0.2 to 2'' diameters in the 3D-DASH image. Each histogram can be well described by a Gaussian, with the width increasing as aperture size increases. The increase in standard deviation with linear aperture size $N=\sqrt{A}$, where A is the area within the aperture, can be described as a power law. A power-law index of 1 would indicate that the noise is uncorrelated, whereas if the pixels within the aperture were perfectly correlated, the background noise would scale as N². The right-hand panel of Figure 8 shows the measured standard deviation as a function of aperture size in the noise-equalized 3D-DASH image. We fit a power law of the form

$\begin{eqnarray}&&\sigma ={\sigma }_{1}\alpha {N}^{\beta },\end{eqnarray} \tag{ 1 }$

where σ₁ is the standard deviation of the background pixels fixed to a value of 1.5 here, α is the normalization and 1 < β < 2 (Whitaker et al. 2012b). The fitted parameters are shown in Figure 8. The power-law fit is shown by the solid line in the figure.

Zoom In Zoom Out Reset image size

We also separate individual DASH apertures based on the program they were taken in (Cycle 23 COSMOS-DASH or Cycle 28 3D-DASH) to determine if the more recent DASH observations differ significantly in noise properties from the pilot program. On average, the 3D-DASH data with low drift rates similar to the pilot COSMOS-DASH data are deeper with a slightly steeper power-law index that supports more correlated noise (see Figure 8). Those pointings with the highest drift instead have shallower depths (noisier data) and β ∼ 1, suggesting uncorrelated noise. This makes sense as the drift rate in these cases is larger than the PSF itself between reads.

With the PSF and noise in hand we can calculate the photometric depth of the mosaic. In the right panel of Figure 8 we show the 5σ depth as a function of aperture size. Apertures of a given size are placed at random positions within the mosaic, and we then calculate the 1σ variation in the measured fluxes within these apertures. A multiplicative aperture correction is applied based on the flux that falls outside that aperture for a point source and multiplied by 5 to estimate the 5σ depth (see Skelton et al. 2014). For very small apertures the S/N is suppressed because of the large aperture corrections that need to be applied, and for very large apertures the S/N is suppressed because of the high noise within the aperture (see, e.g., Figure 7 in Whitaker et al. 2011). The optimal aperture for HST NIR imaging is 03, and the 5σ depth within that aperture is 24.72 ± 0.17 for 3D-DASH pointings (depth at 07 is 24.4 ± 0.2). This can be compared to the expected depth for guided exposures of the same exposure time, as determined by the Exposure Time Calculator. The expected depth is ≈ 25.2, which means that the use of the DASH technique imposes a penalty of ≈36% on the image depth due to increased gyro drift.

The Cycle 28 3D-DASH has an average point-source depth ∼0.4 shallower than COSMOS-DASH. The cause of this increased noise and shallower imaging is likely increased drift rates in the DASH observations due to the new gyros. To check for this, we split the new DASH pointings into three equal-sized bins based on drift rates: dr < 0024 (low drift), 024 < dr < 013 (medium drift), and dr > 013 per read (high drift) (see Figure 8). The pointings with low drift rates reach a point-source depth (H₁₆₀ ∼ 25.1) similar to that of COSMOS-DASH, while the pointings with high drift rates produce shallower images (H₁₆₀ ∼ 24.3).

In Figure 9 we show massive close pairs, many of which are expected to be mergers in progress that lead to the buildup of today's giant elliptical galaxies (see, e.g., De Lucia & Blaizot 2007; Bezanson et al. 2009; Naab et al. 2009; Cooke et al. 2019). For this example, we show 3D-DASH images of galaxies selected from the ground-based UVISTA catalog of Muzzin et al. (2013) that satisfy the following conditions: $\mathrm{log}({M}_{\star }/{M}_{\odot })\gt 10.5$ at z < 3.0 with H < 22 and separation less than 20 kpc (2D separation with an additional cut of z_diff < 0.05).

These galaxies are just-resolved in ground-based data, but it is not possible to measure the properties of the individual objects. In the 3D-DASH data set they are clearly distinct, showing a variety of colors and morphologies. The minimum separation of close pairs in the UVISTA is set by the ground-based resolution, and Marsan et al. (2019) find that up to 1/3 of the most massive single ground-selected massive galaxies are in fact close pairs at HST resolution. This is mitigated by the 3D-DASH images, which provide a resolution high enough to deblend the close pairs at low redshifts. This has significant implications for the high-mass end of low-z ground-based mass functions, as well as the inferred evolution of galaxies from techniques such as abundance matching (e.g., Behroozi et al. 2013) or empirical methods (Moster et al. 2018; Behroozi et al. 2019). In addition, current measurements of the massive galaxy merger rate at z ≫ 1 are still uncertain and model dependent (e.g., Duncan et al. 2019; Pena & Snyder 2021). A 3D-DASH-selected catalog will identify a complete sample of apparent massive galaxy close pairs, further constraining the contribution of mergers to the buildup of massive galaxies.

A second example is shown in Figure 10. These are relatively bright and large galaxies at z < 0.5 that are easily detected and studied from the ground. While these low redshift galaxies are spatially resolved enough from the ground for simple morphological measurements such as size and compactness, they are not resolved enough for detailed morphological analysis pertaining to their growth. The combination of H₁₆₀ and I₈₁₄ from 3D-DASH and ACS COSMOS provides spatially resolved morphological information at two distinct wavelengths and enables the identification of star-forming knots and dust lanes.

Zoom In Zoom Out Reset image size

The increased spatial resolution also allows for the identification and analysis of clumpy star formation (Guo et al. 2018; Huertas-Company et al. 2020). The characteristics of clumps, including numbers and lifetimes, as well as morphological features like tidal tails have important implications for the strength and mechanisms of stellar feedback and galactic winds (Elmegreen et al. 2021; Dekel et al. 2022).

A third example is shown in Figure 11, which are the intermediate redshift counterparts of star-forming massive galaxies shown in Figure 10. These are relatively bright and large galaxies at 0.5 < z < 1.0 that have deep spectroscopic observations from the ground in the LEGA-C survey (van der Wel et al. 2016) but which are not spatially resolved in the deep ground-based NIR imaging (UVISTA; Muzzin et al. 2013). At z > 0.5 the I − H colors are critical for interpreting the morphologies, as they enable the identification of dust, star-forming regions, and old stellar populations and—more generally—the conversion of light-weighted to mass-weighted images (Förster Schreiber et al. 2011; Wuyts et al. 2012; Suess et al. 2019). These images can be combined with the myriad of spectroscopic programs in the COSMOS field (such as van der Wel et al. 2016), for a comprehensive description of the buildup of the Hubble sequence over the past 5–8 Gyr.

Zoom In Zoom Out Reset image size