Large Array Survey Telescope (LAST)

• LAST is a modular survey telescope array comprising multiple 28-cm, wide-field instruments designed for high-cadence, time-domain astronomy.
• Its innovative architecture emphasizes grasp and flexible pointing modes, allowing both expansive sky coverage and deeper imaging through co-mounted telescopes.
• The system integrates rapid, real-time data processing and calibration pipelines that yield high-precision photometry and robust transient detection.

The Large Array Survey Telescope (LAST) is a wide-field visible-light telescope array designed to explore the variable and transient sky with a high cadence. In the 2023 system papers, a LAST node is described as being composed of 48, 28-cm $f/2.2$ telescopes mounted on 12 separate mounts, each telescope providing a field of view of $7.4\,\mathrm{deg}^2$ with $1.25\,\mathrm{arcsec/pix}$, and the full system providing a field of view of $355\,\mathrm{deg}^2$ in $2.9$ Gpix; a later pipeline paper describes a completed array that will consist of 72 telescopes on 18 mount units, indicating that the published documentation spans different deployment stages and operational configurations (Ofek et al., 2023, Ofek et al., 2023, Konno et al., 30 Apr 2026). Across these papers, LAST is presented both as a scientific survey facility for time-domain astronomy and as a test-bed for cost-effective telescope design grounded in arrays of small, seeing-limited instruments rather than a single large telescope (Ofek et al., 2020).

1. Origins and design rationale

The intellectual basis for LAST is the argument that, for seeing-limited imaging sky surveys, recent technological advancements make it more cost effective to construct multiple small telescopes rather than a single large telescope with a similar grasp or etendue (Ofek et al., 2020). In this framework, the relevant performance metric is “grasp,” defined as the amount of volume of space in which a standard candle is detectable per unit time, while an associated “information-content grasp” summarizes the variance of all sources observed by the telescope per unit time (Ofek et al., 2020). The same paper argues that, in the background-dominated noise limit, the optimal exposure time is three times the dead time, $t_{\mathrm{E,optimal}} = 3 t_{\mathrm{D}}$ (Ofek et al., 2020).

This emphasis on grasp is carried directly into LAST’s science case. The science-goals paper states that survey grasp is the “amount of volume of space in which a standard candle is detectable per unit time,” and notes that etendue is comparable to ZTF, but that LAST emphasizes grasp over etendue as the relevant metric for transient science (Ben-Ami et al., 2023). A common misconception is therefore to treat LAST primarily as a small-aperture substitute for a conventional survey telescope. The published rationale instead treats it as an array architecture optimized for wide-area, high-cadence, seeing-limited survey work, enabled by large-format back-side illuminated CMOS detectors with small pixels and by modular replication of off-the-shelf components (Ofek et al., 2020, Ofek et al., 2023).

2. Array architecture and deployment state

In the 2023 overview, LAST is described as a modular array of 48 telescopes, with 32 already installed during commissioning, each telescope being a 28-cm Rowe-Ackermann-Schmidt system at $f/2.2$ and focal length $620\,\mathrm{mm}$, equipped with a full-frame backside-illuminated cooled CMOS detector (Ofek et al., 2023). Each camera is a QHY600M-PH housing a SONY IMX455 sensor with 60.8 MP, $6354\times9576$ sensitive pixels, rolling-shutter readout, dead time of about $0.7\,\mathrm{s}$ per exposure, and two-stage thermoelectric cooling (Ofek et al., 2023). The single-telescope field of view is approximately $7.4\,\mathrm{deg}^2$0, while the full 48-telescope system field of view is $7.4\,\mathrm{deg}^2$1 (Ofek et al., 2023). The total collecting area of a 48-telescope system is equivalent to a 1.9-m telescope (Ofek et al., 2023).

Mechanically, the 48 telescopes are mounted on 12 separate mounts, each carrying four telescopes (Ofek et al., 2023). This geometry is central to LAST’s operating flexibility. In “wide” mode, the four telescopes on a mount point to adjacent fields for maximal coverage; in “narrow” mode, they point to the same field for deeper imaging or related specialized modes (Ben-Ami et al., 2023). The science-goals paper describes this as a modular design that allows optimized parallel surveys, while the system paper notes that it provides significant flexibility in operating the system (Ben-Ami et al., 2023, Ofek et al., 2023).

The first LAST system is located in the Israeli Negev Desert, with commissioning at the Weizmann Astrophysical Observatory in southern Israel (Ofek et al., 2023, Ben-Ami et al., 2023). The system-overview paper gives the site as Neot-Smadar, Israeli Negev Desert, at $7.4\,\mathrm{deg}^2$2, with night sky brightness of about $7.4\,\mathrm{deg}^2$3 and median seeing of $7.4\,\mathrm{deg}^2$4 (Ofek et al., 2023). The science-goals paper reports median seeing of about $7.4\,\mathrm{deg}^2$5, a very high clear-sky fraction, and notes that the geographic location fills an “Asiatic gap” in global survey coverage for time-domain follow-up (Ben-Ami et al., 2023).

A later pipeline paper describes an expanded system state: 72 telescopes, 18 mount units, and 40 telescopes already operational, with simultaneous coverage up to $7.4\,\mathrm{deg}^2$6 or co-aligned operation equivalent to a 2.4-m telescope with $7.4\,\mathrm{deg}^2$7 field of view (Konno et al., 30 Apr 2026). This suggests that “LAST” in the literature refers both to the original 48-telescope node concept and to a subsequently enlarged operational configuration.

3. Survey modes, cadence, and scheduling

LAST’s baseline observing strategy is to obtain multiple consecutive short exposures of each field. The pipeline paper states that the basic strategy is to obtain multiple 20-s consecutive exposures of each field, called a visit, with the default visit consisting of 20 exposures and totaling 400 s (Ofek et al., 2023). The rationale is operational as well as scientific: the visit structure enables rejection of transient artifacts such as satellite glints, intra-visit discovery of asteroids via motion, and access to rapid stellar variability on 20-s timescales (Ofek et al., 2023).

The science-goals paper formalizes this into parallel survey programs. A high-cadence survey uses 8 of the 12 mounts, covers about $7.4\,\mathrm{deg}^2$8 per night, and obtains 8 visits per night per pointing, each visit consisting of $7.4\,\mathrm{deg}^2$9 exposures; a low-cadence survey uses 4 of the 12 mounts, covers about $1.25\,\mathrm{arcsec/pix}$0 per night, and obtains 2 visits per night per field (Ben-Ami et al., 2023). The total nightly coverage of these two surveys is about $1.25\,\mathrm{arcsec/pix}$1 (Ben-Ami et al., 2023). The system-overview paper also notes that 5% of observing time is reserved for Target-of-Opportunity observations in support of multi-messenger astronomy (Ofek et al., 2023).

Scheduling for telescope arrays of this class has been studied in a dedicated multilevel framework tested with parameters and simulated layouts inspired by ZTF, LSST, and the forthcoming LAST (Zhang et al., 2023). That framework includes a Global Scheduler, a Site Scheduler, an optional Telescope Scheduler, and a Record and Display Layer (Zhang et al., 2023). The scheduling problem is formulated as a Mixed Integer Linear Programming problem with objectives that maximize execution priority and cumulative scheduled observation time, together with a sky-coverage uniformity metric,

$1.25\,\mathrm{arcsec/pix}$2

where $1.25\,\mathrm{arcsec/pix}$3 is an adjustable constant, $1.25\,\mathrm{arcsec/pix}$4 is current time, $1.25\,\mathrm{arcsec/pix}$5 is the last observation time for the field, $1.25\,\mathrm{arcsec/pix}$6 is the total number of visits to the field, and $1.25\,\mathrm{arcsec/pix}$7 is the total observation time for the field (Zhang et al., 2023). Simulations up to 20 sites and 1000 fields yielded Time Allocated Ratio values exceeding 90% for large field arrays, with global planning completed in 1–2 hours (Zhang et al., 2023). Although this framework is generic rather than LAST-specific, it clarifies the algorithmic environment in which a distributed, high-cadence LAST-like array is intended to operate.

4. Data flow, calibration, and transient-detection pipeline

LAST produces a data stream of about $1.25\,\mathrm{arcsec/pix}$8 for the full array and about $1.25\,\mathrm{arcsec/pix}$9 of raw images per year, which are processed near real time at the observatory site using about 700 CPU cores (Ofek et al., 2023, Ofek et al., 2023). The data pipeline is divided into two major parts: processing and calibration of single images followed by coaddition of the visit exposures, and construction of reference images followed by image subtraction and transient detection (Ofek et al., 2023, Konno et al., 30 Apr 2026).

The first part of the pipeline performs dark subtraction, non-linearity correction, flat-field correction, gain normalization, masking, source detection, PSF modeling, astrometric calibration, photometric calibration, coaddition, and visit-level source merging (Ofek et al., 2023). Each full-frame image is split into 24 sub-images of $355\,\mathrm{deg}^2$0 pixels with overlaps of about 64 pixels (Ofek et al., 2023). Astrometry is tied to Gaia DR3 using an affine transform plus a 3rd-order polynomial, and photometric calibration is modeled per sub-image as

$355\,\mathrm{deg}^2$1

where $355\,\mathrm{deg}^2$2 is the instrumental magnitude, $355\,\mathrm{deg}^2$3 is Gaia BP, and $355\,\mathrm{deg}^2$4 and $355\,\mathrm{deg}^2$5 are the zeropoint and color term (Ofek et al., 2023). The products include calibrated single and coadded images, 32-bit mask images, source catalogs, PSF photometry, merged source catalogs, proper-motion and variability indicators, minor-planet detection, calibrated light curves, and cross-matching with external catalogs (Ofek et al., 2023).

The second part of the pipeline performs transient detection using proper image subtraction. The 2026 pipeline paper states that transient detection is based on ZOGY subtraction, combined with the Translient statistic for sub-pixel motion discrimination and a sequence of deterministic filtering steps, explicitly without the use of machine learning (Konno et al., 30 Apr 2026). The subtraction stage produces a difference image, a PSF for the difference image, and an $355\,\mathrm{deg}^2$6-statistic image; candidates are filtered using diagnostics that include $355\,\mathrm{deg}^2$7-function matched filtering, extended-PSF filtering, Gabor filtering, peak-valley tests, PSF-shape-based filtering, and cross-matching against known variables, minor planets, and external catalogs (Konno et al., 30 Apr 2026). Using commissioning data, this stage achieved a preliminary $355\,\mathrm{deg}^2$8 limiting magnitude of $355\,\mathrm{deg}^2$9–$2.9$0 mag, a single-epoch transient detection efficiency of about 80%, and a purity of at least 90% at signal-to-noise ratio $2.9$1 (Konno et al., 30 Apr 2026).

A specialized variability report generator has also been developed for LAST. It identifies candidate variable stars using adjustable thresholds for periodic and non-periodic variability, and outputs visual and tabular photometric reports from a given LAST sub-image (Zemenu, 2023). In that implementation, the default periodicity threshold is a periodogram significance of 13$2.9$2 (Zemenu, 2023).

5. Science program

LAST’s science goals are explicitly time-domain and multi-purpose. The science-goals paper identifies gravitational-wave electromagnetic counterparts, planetary systems around white dwarfs, and near-Earth objects as key drivers, while the system-overview paper adds supernovae, kilonovae, gamma-ray-burst afterglows, flare stars, asteroid and NEO detection, satellite glints, and exoplanet transits, including white-dwarf hosts (Ben-Ami et al., 2023, Ofek et al., 2023).

For gravitational-wave counterpart searches, LAST is designed to cover the uncertainty regions of next-generation GW detectors in a single exposure, with limiting magnitude sufficient to detect GW170817-like kilonovae beyond the O4 sensitivity horizon of $2.9$3 (Ben-Ami et al., 2023). Continuous and rapid coverage is intended to improve sensitivity to early optical emission within hours after merger, which is relevant to kilonova physics, $2.9$4-process nucleosynthesis, and GW cosmology (Ben-Ami et al., 2023).

For white-dwarf planetary systems, the high-cadence $2.9$5 visit structure is intended to detect minute-scale transits, pulsations, and eclipses in double white-dwarf systems (Ben-Ami et al., 2023). The science-goals paper states that, for a 1% planet occurrence rate, the expected yield is about 5 detections over the survey lifetime, and that the survey can provide order-of-magnitude improvement in upper limits for white-dwarf planet occurrence rates (Ben-Ami et al., 2023).

For Solar System work, LAST supports streak-based and proper-motion-based detection of asteroids and near-Earth objects. The science-goals paper states that the system can monitor about 2800 asteroids per night, including main-belt objects and NEAs, and can detect NEAs brighter than $2.9$6 in the stated streak-sensitivity framework (Ben-Ami et al., 2023). The pipeline paper further subdivides moving-object detection into several apparent-speed regimes, from very fast streaks to slow movers searched for as faint unlinked sources after differencing (Ofek et al., 2023).

Additional programs include extragalactic transients such as Type Ia supernovae, superluminous supernovae, tidal disruption events, fast blue optical transients, orphan GRB afterglows, AGN and blazar variability, microlensing as a potential network-enabled application, and polarimetry through dedicated filtered subsets of the array (Ben-Ami et al., 2023). The science-goals paper also states that calibrated object catalogs are to be publicly released every 12 months and that transient events are reported daily to TNS (Ben-Ami et al., 2023).

6. Measured performance and projected extensions

Commissioning measurements reported in the system-overview paper give a $2.9$7 $2.9$8 limiting magnitude of about 19.6 in a single 20-s exposure and 21.0 in a $2.9$9 coadd for a single 28-cm telescope (Ofek et al., 2023). Astrometric two-axes precision at the bright end is about 60 mas in 20 s and 30 mas in $t_{\mathrm{E,optimal}} = 3 t_{\mathrm{D}}$0, while averaging over multiple coadds reaches about 15 mas (Ofek et al., 2023). Absolute photometric calibration relative to Gaia provides about 10 millimag accuracy, and relative photometric precision at the bright end is about 3 millimag in a single 20-s image and about 1 millimag in a 320-s coadd, measured over a time scale of about 60 min (Ofek et al., 2023). Image quality after tip/tilt correction typically ranges from $t_{\mathrm{E,optimal}} = 3 t_{\mathrm{D}}$1 to $t_{\mathrm{E,optimal}} = 3 t_{\mathrm{D}}$2, and vignetting leaves 72% of the image area with at least 90% of peak response and 93% with at least 80% (Ofek et al., 2023).

These metrics are significant because LAST’s scientific case depends not only on coverage and cadence but also on millimagnitude photometry, tens-of-milliarcsecond astrometry, and real-time screening of a data stream that exceeds traditional small-survey regimes. The pipeline paper reports that source measurement is about 30 times faster than SExtractor and that a typical 20-exposure visit from one telescope can be processed in less than 5 minutes for low-density fields (Ofek et al., 2023). This suggests that LAST’s architecture is inseparable from its software system: the telescope array, visit cadence, on-site computing, and reduction pipeline are designed as a single operational unit.

Several extensions are described in the literature. The 2023 overview mentions PAST (Pan-chromatic Array for Survey Telescopes) and MAST (Multi-Aperture Spectroscopic Telescope) as follow-up facilities (Ofek et al., 2023). A 2026 paper proposes LAST-P, the Large Array Survey Telescope Polarization Node, consisting of 48 small telescopes on 12 co-mounted units with an instantaneous field of view of $t_{\mathrm{E,optimal}} = 3 t_{\mathrm{D}}$3 in polarization survey mode (Martins et al., 7 Jan 2026). In that design, a $t_{\mathrm{E,optimal}} = 3 t_{\mathrm{D}}$4-minute exposure reaches Gaia $t_{\mathrm{E,optimal}} = 3 t_{\mathrm{D}}$5, and the precision on linear polarization degree reaches 0.7%, 1.5%, and 3.5% for sources with magnitudes 17, 18, and 19, respectively, for seeing of $t_{\mathrm{E,optimal}} = 3 t_{\mathrm{D}}$6 and air mass of about 1 in dark locations (Martins et al., 7 Jan 2026). The same paper proposes an AGN strategy for long-term monitoring of about 200 AGN with less than 1-day cadence (Martins et al., 7 Jan 2026).

Taken together, these publications define LAST not as a single instrument in a fixed final form, but as a modular survey architecture. Its central features are repeated throughout the literature: many small $t_{\mathrm{E,optimal}} = 3 t_{\mathrm{D}}$7 telescopes, wide instantaneous coverage, repeated short exposures grouped into visits, near-real-time on-site reduction, and explicit optimization for transient and variability science through grasp, cadence, and automation (Ofek et al., 2020, Ofek et al., 2023, Ofek et al., 2023).