Large Fiber Array Spectroscopic Telescope (LFAST)

• LFAST is a spectroscopic telescope concept defined by modular fiber-fed mirrors that reformat light to enable extreme photon collection and high multiplex efficiency.
• It offers dual architectural forms: a 10–12 m class wide-field survey design and a scalable array configuration that mimics ELT-scale collecting area through modular sub-apertures.
• The design integrates advanced fiber positioning and microlens array techniques via two-photon polymerization to optimize spectral coupling for diverse astrophysical surveys.

Searching arXiv for LFAST and related spectroscopic facility papers. The Large Fiber Array Spectroscopic Telescope (LFAST) denotes a family of spectroscopic-facility concepts centered on extreme photon collection, fiber feeding, and massively multiplexed spectroscopy. In the recent literature, the name LFAST is used explicitly for a scalable array telescope in which many small mirrors each feed an individual optical fiber, and the fibers are then reformatted for optical and infrared spectrometers (Delamer et al., 29 Aug 2025). Earlier southern-hemisphere facility studies—especially the ESO Spectroscopic facility and the US SpecTel concept—describe a closely related design philosophy: a dedicated spectroscopic survey telescope in the 10–12 m class, optimized end-to-end for a very large fiber-fed focal plane, simultaneous high- and low-resolution spectroscopy, and optional panoramic integral-field capability (Pasquini et al., 2017, Ellis et al., 2019). Taken together, these works define LFAST as an architectural approach rather than a single finalized instrument.

1. Terminology and conceptual scope

The explicit use of the name LFAST appears in work on a scalable array telescope whose elements are multiple mirrors feeding fibers and, through those fibers, spectrometers (Delamer et al., 29 Aug 2025). In that formulation, the telescope is not a monolithic aperture but an aggregate photon-collecting system: many individual mirrors collect light, each mirror feeds one optical fiber, and the outputs from many fibers are combined into a single rectangular fiber that feeds one or more optical and infrared spectrometers.

A closely allied concept appears in the earlier literature under different names. The ESO Spectroscopic facility is presented as a novel facility dedicated to massively-multiplexed spectroscopy, with a very wide field Cassegrain focus optimized for fibre feeding, an 11.4 m pupil, and a gravity-invariant focus for a giant IFU (Pasquini et al., 2017). The later SpecTel study recommends a southern-hemisphere spectroscopic facility comprising a large diameter telescope, fiber system, and spectrographs collectively optimized for massively-multiplexed spectroscopy, with an explicit baseline of an 11.4-meter aperture, a five square degree field of view, and 15,000 robotically-controlled fibers (Ellis et al., 2019).

This suggests that LFAST functions in the literature both as a specific named project and as a broader design lineage. In both variants, the defining characteristics are the same: end-to-end optimization for fiber spectroscopy, very high multiplex, broad wavelength coverage, and explicit coupling to large imaging and astrometric surveys.

2. Architectural forms

Two architectural realizations recur in the cited work.

The first is the large-aperture survey telescope form. In the ESO and SpecTel studies, the facility is a 10–12 m class southern-hemisphere telescope dedicated uniquely to spectroscopic surveys, with a corrected 2.5 degree diameter field of view, a large Cassegrain focal plane, and a survey design in which the telescope, corrector, focal plane, fiber-positioning system, and spectrographs are conceived simultaneously (Pasquini et al., 2017). The baseline SpecTel formulation is explicit: an 11.4-meter aperture, optical spectroscopic survey telescope with a five square degree field of view (Ellis et al., 2019).

The second is the scalable array telescope form. In the 2025 LFAST paper, each element of the array consists of multiple mirrors each feeding to an individual fiber; each LFAST unit uses 20 primary mirrors, each with 0.76 m diameter; one 20-mirror unit is equivalent in collecting area to a traditional 3.5 m aperture; and the full planned system consists of 132 such 20-unit systems, giving a total collecting area of about $1200\,\mathrm{m^2}$ (Delamer et al., 29 Aug 2025). The motivation is explicitly economic as well as scientific: traditional telescope costs scale steeply with diameter, approximately as

$$\mathrm{cost} \propto D^n,\qquad n \approx 2.0\text{--}2.8,$$

while support structure costs can scale roughly as $D^3$ (Delamer et al., 29 Aug 2025).

A plausible implication is that the two forms address the same throughput problem from different engineering directions. The 10–12 m facility maximizes etendue and multiplex within a single large telescope, whereas the array concept attempts to reproduce ELT-scale collecting area through modular fiber-coupled sub-apertures.

3. Optical design, focal plane geometry, and multiplex

The large-aperture LFAST-style studies place exceptional emphasis on the optical train between sky and fiber. In the ESO concept, the telescope has a Field of View (FoV) of 2.5 degrees diameter and a 11.4m pupil, and is described as the largest etendue telescope (Pasquini et al., 2017). The corrected focal plane is about 1.43 m in diameter at F/2.86, with a plate scale of roughly 160 micron/arcsecond, described as an excellent aperture for fibre coupling (Pasquini et al., 2017). The three-lens corrector includes an ADC and has good performance in the 360-1300 nm wavelength range; the working principle is described as one in which the last two lenses and the focal plane tilt around a circle of 5m radius to compensate for atmospheric dispersion, with coma corrected by translating the secondary mirror (Pasquini et al., 2017).

SpecTel gives the same geometric core in more detail. The primary mirror consists of 78 ELT primary segments, the secondary is 4.2 m, and the telescope uses a three-lens corrector that also serves as an ADC (Ellis et al., 2019). The corrector provides a corrected 2.5 degrees, 1.43m diameter Cassegrain focal plane at F/2.86 and a platescale of 570 mm/deg (Ellis et al., 2019). The scale comparison with existing facilities is explicit: the focal plane has 3.1 times the area of DESI’s prime focus, and the authors state that it would be possible to duplicate DESI’s fiber positioner design to produce a 15,000 fiber system (Ellis et al., 2019).

The array-telescope LFAST uses a different front end but an equally fiber-centric backend. Each mirror feeds one optical fiber; each fiber has an 18 $\mu$m core diameter, corresponding to an on-sky footprint of about 1.4 arcsec; and the outputs from many fibers are then combined into a single rectangular fiber (Delamer et al., 29 Aug 2025). In this architecture, focal-plane geometry is displaced from a single large corrected field to the problem of reformatting many fiber outputs into a spectrometer-compatible slit.

4. Fiber system, positioners, and spectrographs

The fiber system is the central enabling technology in all three papers. In the ESO facility, the large focal plane can host up to 16.000 fibres, and using the DESI positioner mechanism, more than 15000 positioners can be hosted, just using existing technology (Pasquini et al., 2017). The same study proposes a two-tier multiplex allocation in which each point of the focal plane can be reached by three fibres, yielding a 2:1 ratio between high-resolution and low-resolution fibres, so that with 15000 fibers the facility could observe 5000 objects … at high resolution and 10000 at low resolution simultaneously (Pasquini et al., 2017).

SpecTel retains the 15,000 robotically-controlled fibers baseline and makes extensibility explicit. It states that miniaturized electric motors for a higher density of fiber positioners at a pitch of 5 mm would allow a 4X increase in fiber number, enabling instrumentation of 60,000 fibers (Ellis et al., 2019). The paper therefore frames the baseline as a first stage rather than a hard upper limit.

The spectrograph plans are similarly modular. SpecTel describes an alternative design inspired by MOONS in which 600 fibers of 160 micron diameter (1 arcsec projected aperture) feed the spectrograph through a 300 mm diameter collimating mirror; the IR arm covers 965 to 1330 nm with R~4000; and optical extension to 360 nm is possible through two optical channels with R~3000 or three optical channels with R~4000 (Ellis et al., 2019). The same paper states that high resolution spectrographs would naturally be considered as part of the conceptual design process, and cites use cases requiring high resolution (R ~ 20,000 - 40,000) spectra at S/N~80 for 30-85M stars to V~15.5 - 17 (Ellis et al., 2019).

The ESO study specifies the top-level spectroscopic envelope differently but compatibly: R=20-40000 for High Res, R=1000-3000 for Low Res, over 360-1000 nm (Pasquini et al., 2017). Its low-resolution optical design uses curved detectors and an F/1.1 camera, with $\sim 600$ one arcsecond fibres per spectrograph, and less than 20 such spectrographs to cover all low-resolution fibers (Pasquini et al., 2017).

For the array-telescope LFAST, the bottleneck is not focal-plane patrol but efficient reformatting into a slit. The 2025 study focuses on optimizing coupling between the fiber bundle and the spectrometer slit input, because losses at this stage directly reduce the usable photon budget (Delamer et al., 29 Aug 2025). In that sense, the spectrograph problem migrates from large-scale fiber positioner density to micro-optical coupling precision.

5. Integral-field capability and alternative observing modes

A notable feature of the large-aperture LFAST-style facilities is the inclusion of a central-field integral-field mode rather than exclusive reliance on pre-selected fiber targets. In the ESO concept, a gravity invariant focus for the central 10 arc-minutes is available to host a giant integral field unit (IFU), and the top-level requirements specify a Panoramic IFU: 3x3 arcmin FoV, with R∼5000 and coverage in the blue down to 325 nm (Pasquini et al., 2017). The low-resolution system must also be usable by the panoramic IFU, and simultaneous low- and high-resolution observations over the whole FoV are specifically required.

SpecTel states that by inserting mirrors ahead of the corrector, the central 10 arcminutes can be directed to a gravity-invariant Coude focus hosting a panoramic IFU, with the floor below the telescope rotating to compensate for IFU field rotation (Ellis et al., 2019). It also gives the scientific rationale for that mode: alongside traditional multi-fiber surveys based on targets pre-selected from deep photometry, a panoramic IFU offers the opportunity for statistically-complete deep emission line and other serendipitous searches (Ellis et al., 2019).

This dual-mode architecture is significant because it broadens the operational meaning of LFAST. The facility is not restricted to catalog-driven spectroscopy; it also admits contiguous-field spectroscopy in the central region, which is especially relevant for emission-line searches, dense environments, and studies where source pre-selection is incomplete.

6. Scientific programs and survey role

The science case attached to the large-aperture concepts is organized around four themes: assembly history of the Milky Way and dark matter, galaxy evolution in the cosmic web and intergalactic medium, cosmology, and the transient universe (Ellis et al., 2019). The strongest recurring driver is Galactic archaeology in the Gaia era. Gaia is providing radial velocities for only 10% of its sources, leaving more than a billion stars with excellent astrometry and photometry but no spectroscopy (Ellis et al., 2019). The ESO study similarly identifies high resolution spectroscopy of GAIA stars as a primary case and discusses surveys covering 85 million stars at limiting magnitude 17 (Pasquini et al., 2017). SpecTel states that a 15,000 fiber instrument on a 10–12 m telescope could observe 650 million stars in only a year for moderate-resolution kinematic work (Ellis et al., 2019).

For extragalactic astronomy and cosmology, the motivation is the conjunction of deep imaging and insufficient spectroscopic follow-up capacity. By 2025-2030, deep imaging surveys will provide 0.2 arcsec image quality over more than 15,000 deg2 from Euclid and multi-band photometry to AB~27 from LSST (Ellis et al., 2019). LSST is said to image the southern extragalactic sky in six filters, providing a photometric catalog of > 4 billion galaxies (Ellis et al., 2019). The proposed facility is intended to supply the missing combination of large aperture + wide field + high multiplex + better resolution/SNR relative to current or planned instruments.

Concrete cosmological survey modes are listed in SpecTel: photometric-redshift calibration with 20,000 galaxies to I~25.3, dense redshift surveys of 20,000 galaxies deg-2 over 14,000 deg2, and high-redshift clustering of at least one hundred million galaxies at 1.5<z<4 (Ellis et al., 2019). For cosmic web mapping and Ly$\alpha$ tomography, the paper argues that representative-volume mapping over 1 < z < 4 is a formidable challenge that can only be achieved with a dedicated wide-field 10-12m class facility, and gives the requirement of covering about 200 deg2 from AB~23 at z~1 to AB~25 at z~4 to span about 1 Gpc3 (Ellis et al., 2019).

Transient follow-up is tied particularly to LSST. SpecTel notes that LSST will find more than a million transient events in five years, and estimates that if roughly ten live events per field are followed while a galaxy survey studies ten fields a night to AB~23, then roughly 10% of all visible LSST transients could be observed via merging fiber allocations on the fly (Ellis et al., 2019). The ESO study similarly notes that for transients such as SNe, more than 400 events will be present in the FoV at any time (Pasquini et al., 2017).

The array-telescope LFAST paper does not elaborate a comparably broad survey program, but its framing is consistent with the same scientific logic: astronomy is often photon-starved, and scientific reach improves most directly by collecting more photons (Delamer et al., 29 Aug 2025). That formulation aligns the array concept with the survey-driven facilities, even though the cited paper is primarily an instrumentation study.

7. Enabling micro-optics and developmental status

The 2025 LFAST study shifts attention from facility-level architecture to the micro-optical details required to make a fiber-array telescope practical (Delamer et al., 29 Aug 2025). Its core problem is the efficient coupling of many fibers into spectrometer slits with minimal loss. The authors use a two-dimensional microlens array, with each microlens aligned to an individual fiber, and turn to two-photon polymerization (2PP) because standard fabrication approaches are insufficiently flexible for the required scale and complexity.

The fabrication platform is a Nanoscribe Photonic Professional / GT2-class system at Penn State, using a 780 nm laser with pulse duration 80–100 fs and repetition rate 80 MHz, a 25×, NA 0.8 objective, and a 25 mm × 25 mm × 0.7 mm fused silica substrate (Delamer et al., 29 Aug 2025). The transparent resin IPX-Clear has transmission from 350 nm to 1550 nm greater than 95%, with minimum transmission still greater than 88% for a 1 mm thick sample within the LFAST wavelength regime of 390\,\mathrm{nm} \text{ to } 1700\,\mathrm{nm} (Delamer et al., 29 Aug 2025). The abstract highlights that 2PP can produce microlenses with wavefront aberrations as small as $\lambda/20$.

The initial prototype is a $2 \times 10$ microlens array with pitch = 250 $\mu$m, radius of curvature = 400 $\mu$m, diameter = 245 $$\mathrm{cost} \propto D^n,\qquad n \approx 2.0\text{--}2.8,$$0m, sag = 19.2 $$\mathrm{cost} \propto D^n,\qquad n \approx 2.0\text{--}2.8,$$1m, and total designed height 25 $$\mathrm{cost} \propto D^n,\qquad n \approx 2.0\text{--}2.8,$$2m after adding a base (Delamer et al., 29 Aug 2025). The fabrication workflow includes acetone and IPA cleaning, O$$\mathrm{cost} \propto D^n,\qquad n \approx 2.0\text{--}2.8,$$3 plasma etch for 5 min, IPX-Clear thermalization for 10 min, post-print bakes at 65°C, 95°C, and 65°C for 2 min each, development in mr-dev 600 for 20 min, an IPA bath for 2 min, and UV cure for 10 min (Delamer et al., 29 Aug 2025).

The principal fabrication issue is dose optimization. The paper gives the simplified relation

$$\mathrm{cost} \propto D^n,\qquad n \approx 2.0\text{--}2.8,$$4

where $$\mathrm{cost} \propto D^n,\qquad n \approx 2.0\text{--}2.8,$$5 is laser power, $$\mathrm{cost} \propto D^n,\qquad n \approx 2.0\text{--}2.8,$$6 is write speed, and $$\mathrm{cost} \propto D^n,\qquad n \approx 2.0\text{--}2.8,$$7 is typically 2–4, while noting that the formula omits voxel-size dependence, overlap between slices, and time-dependent polymerization effects (Delamer et al., 29 Aug 2025). For the 25× objective, the approximate voxel size is

$$\mathrm{cost} \propto D^n,\qquad n \approx 2.0\text{--}2.8,$$8

The reported metrology shows incremental improvement. Adapted 2GL shell-like parameters produced a lens with a peaked profile deviating from a spherical best fit by nearly $$\mathrm{cost} \propto D^n,\qquad n \approx 2.0\text{--}2.8,$$9, although surface roughness in the center was as low as 30 nm (Delamer et al., 29 Aug 2025). Varying PiezoSettlingTime between 5 s and 30 s reduced deviation below $D^3$0 in the best case, but individual lens print time exceeded 2 hours, which is impractical for a final system requiring 2640 microlenses (Delamer et al., 29 Aug 2025). A later shell-and-scaffold approach, using a representative scaffold with 1 $D^3$1m slice, 0.5 $D^3$2m hatch, and 17.5 mW laser power, plus a shell at 0.2 $D^3$3m slice/hatch, yielded for the most recent $D^3$4 prototype an average deviation from spherical = $D^3$5 (Delamer et al., 29 Aug 2025). The authors plan iterative pre-compensation to reduce spherical deviation to < 0.5\,\mu\mathrm{m}, with an ideal target of < 0.25\,\mu\mathrm{m} (Delamer et al., 29 Aug 2025).

This developmental status is important for interpreting LFAST as a whole. The 2017 and 2019 facility studies are explicitly conceptual design exercises, including staged implementation from a 15,000 fiber system and moderate-resolution spectrographs to later high-resolution spectrographs and a large IFU at the Coude focus, and potentially a final stage with 60,000 fibers (Ellis et al., 2019). The 2025 paper shows that, in the array-telescope variant, critical backend technologies such as microlens-aligned fiber reformatting are still being optimized experimentally (Delamer et al., 29 Aug 2025). A plausible implication is that LFAST remains best understood as an evolving spectroscopic systems concept whose key components—collecting architecture, focal-plane strategy, spectrograph modularity, and coupling optics—are still under active design refinement.