Vera C. Rubin Observatory (LSST)

A telescope built to make a movie of the sky

Almost every previous optical observatory has been designed to point at one target at a time. The Vera C. Rubin Observatory is the opposite: a wide-field survey machine, fast enough to slew between fields in a few seconds, with a camera large enough that a single exposure covers about forty full Moons of sky. Over a single night it acquires roughly two thousand exposures. Over three nights it has covered the whole visible Southern sky. Over ten years it will have covered each patch about eight hundred times in each of six filters, building a deep static map of the static universe and a movie of everything that changed.

The instrument is named for Vera C. Rubin (1928-2016), who measured the rotation curves of disk galaxies with Kent Ford in the 1970s and showed that they remain flat at large radius — among the most direct dynamical evidence that ordinary matter is far outweighed by an invisible component. The name is fitting because the survey's headline science is testing the nature of the dark sector: dark matter, dark energy, and the growth of cosmic structure across thirteen billion years.

The hardware: a fast 8.4-metre and a 3.2-gigapixel camera

Rubin sits at 2,647 metres on Cerro Pachón in north-central Chile, sharing the ridge with the Gemini South and SOAR telescopes. The atmospheric seeing at this site has a long-term median of about 0.67 arcseconds in r-band, comfortably matched to the 0.2-arcsecond pixel scale of the focal plane. The dome is unusually large because the telescope inside is unusually compact: the optical assembly weighs 350 tonnes and slews at four degrees per second, settling on a new field in under five seconds.

The optical design is a three-mirror anastigmat in the Paul-Baker form. The 8.4-metre primary M1 and 5.0-metre tertiary M3 are figured into a single monolithic substrate — the so-called M1M3 — fabricated at the University of Arizona Mirror Lab. The 3.4-metre secondary M2 is a conventional concave element. Combined, they deliver a final focal ratio of f/1.234 and a 3.5-degree-diameter unvignetted field. The fast beam is what lets a single short exposure go deep enough to detect a 24th-magnitude star at signal-to-noise of ten — fainter than anything visible to the naked eye by a factor of a million.

The light is fed into LSSTCam, the largest digital astronomical camera ever built. Its 3.2-gigapixel focal plane is tiled from 189 thinned, back-illuminated 16-megapixel CCDs grouped into 21 rafts, each raft an independent vacuum module with its own electronics. The imaging area is a 64-centimetre disk that captures 9.6 square degrees of sky per exposure. A six-position filter changer rotates the ugrizy filters through the optical path during the night; the seventh slot is empty for engineering. The whole camera assembly is the size of a small car, weighs 3,200 kg, and is cooled with cryogenic loops to keep dark current and read noise at the few-electron level required for photon-limited imaging.

Cadence: scanning the whole sky every three nights

The Legacy Survey of Space and Time is structured around the wide-fast-deep main survey, which covers about 18,000 square degrees of the Southern sky to a depth of r ~ 24.5 per visit and r ~ 27.5 in the ten-year coadd. Each visit is a pair of 15-second exposures in one filter, with the telescope dithering between them to bridge gaps in the focal plane. Visits are issued by a feedback scheduler that balances airmass, sky brightness, weather forecasts, science priorities, and filter rotation to maximise survey uniformity.

Survey component	Time fraction	Cadence / depth target
Wide-fast-deep main survey	~85 %	~825 visits per field over 10 yr; r ~ 27.5 coadd
Galactic plane / bulge	~3 %	Lower cadence, more crowding tolerance
North-ecliptic spur	~2 %	Solar System / NEO completeness
South-celestial-pole region	~1 %	SMC / LMC science
Deep-drilling fields	~5 %	5 deep fields, ultra-deep transient science
Twilight / engineering	~4 %	Calibration, short-period asteroid recovery

Each LSSTCam exposure is roughly 6.4 GB of raw pixels. With about 1,000 fields per night and two snaps per visit, the nightly raw take is about 15 TB. Multiply by 250 photometric nights per year and a ten-year mission and the cumulative raw archive crosses 35 PB. The reduced data products — coadds, difference images, object catalogues — push the final archive to about 200 PB.

Six bands: ugrizy

Rubin observes in six broad filters that together span 320 to 1080 nanometres — from atmospheric-cutoff near-ultraviolet through visible to near-infrared, before the silicon-CCD response falls off in the y-band.

Filter	Central λ (nm)	FWHM (nm)	5σ depth (single visit)	10-yr coadd depth
u	368	62	23.8	26.1
g	478	140	25.0	27.4
r	620	140	24.7	27.5
i	755	134	24.0	26.8
z	870	124	23.3	26.1
y	1004	120	22.1	24.9

The six bands let Rubin estimate a photometric redshift for every catalogued galaxy. The lensing-quality redshift uncertainty is approximately σ_z/(1+z) ~ 0.05, which is the floor that drives most of the dark-energy systematic budget. The bands also act as a coarse spectral classifier: stars of different temperatures and metallicities trace different loci in u−g vs. g−r colour space, allowing main-sequence stars, red giants, white dwarfs, M dwarfs and brown dwarf candidates to be separated by photometry alone.

Data volume and the alert stream

An exposure has to be reduced and difference-imaged within 60 seconds of the shutter closing — fast enough that astronomers anywhere in the world can react to a transient while it is still bright. Each new image is registered to the deep coadd template for that patch of sky, point-spread-function matched, and subtracted; the residual difference image is scanned for sources above 5 sigma. Each surviving difference source becomes an "alert" — a packet that contains the cutout image, photometry, history, classification scores, and astrometric solution.

About ten million alerts are issued every night. They are too many for any single human to read, so the stream is broadcast to a small set of community brokers — ALeRCE, ANTARES, Fink, Lasair — which subscribe to the full stream, run their own classifiers, and serve filtered sub-streams to downstream users. Brokers turn the firehose into a working list of, say, the fifty most likely young Type Ia supernovae and the three best kilonova candidates near a LIGO-Virgo localisation.

Per night             ~ 2,000 exposures × 6.4 GB    ≈  15 TB raw
Reduction latency       < 60 s per visit
Alert rate              ~ 10⁷ per night
Survey total raw         ~ 35 PB    (10 yr, ~ 250 nights/yr)
Final archive incl. catalogues, coadds, DIA  ~ 200 PB
DR1 (released ~6 mo after start of operations)  full pipeline, year 1
DRn (annual)            cumulative survey to date
Object catalogue (final) ~ 20 × 10⁹ galaxies, ~ 37 × 10⁹ stars
Solar System object catalogue (final) ~ 6 × 10⁶ bodies

Dark energy: four probes, one survey

Constraining the dark-energy equation of state w(z) is Rubin's headline cosmology goal. No single observable nails w on its own — each probe is sensitive to a different combination of the expansion history and the growth of structure, with its own systematic floor. Rubin is designed to run four largely independent probes off the same data and then combine them, breaking degeneracies that each probe alone cannot.

• Weak gravitational lensing. The shapes of about three billion source galaxies are measured to ~0.3 % accuracy in shear; the cosmic-shear power spectrum from z ~ 0 to z ~ 3 traces the growth of structure σ_8(z) and the matter density Ω_m.
• Galaxy clustering & BAO. Spatial clustering of ~20 billion catalogued galaxies, including the characteristic 150-Mpc baryon-acoustic-oscillation feature, provides a standard-ruler constraint on the angular-diameter distance D_A(z) and the Hubble expansion H(z).
• Type Ia supernovae. Difference-imaging the 5 deep-drilling fields delivers ~10⁵ photometrically classified Type Ia SNe with light curves out to z ~ 1.5 — a standardised-candle Hubble diagram with ten times the supernovae of all current samples combined.
• Cluster abundance. Optical-richness catalogues, calibrated by stacked weak lensing, count ~10⁵ clusters with M > 10¹⁴ M☉ out to z ~ 1.2; cluster counts versus redshift test the exponential tail of structure growth.

Each probe constrains a different ellipse in the (w_0, w_a) plane, where w(z) = w_0 + w_a z/(1+z). The product of the joint posterior — the "Dark Energy Task Force figure of merit" — is forecast to be roughly an order of magnitude better than the combined DES, KiDS and HSC constraints.

Milky Way structure and stellar populations

Rubin will photometrically classify a few billion stars out to galactocentric distances of 100 kpc. The combination of multi-band photometry, sub-percent astrometric repeatability, and ten years of cadence yields parallaxes for nearby stars, proper motions across the entire Milky Way halo, and a near-complete census of low-mass and white-dwarf stars in the local volume.

• Stellar streams and dwarf galaxies — direct probes of small-scale dark matter, sensitive to dark-matter substructure with halo masses down to ~10⁶ M☉.
• RR Lyrae and other variable distance indicators — a clean 3-D map of the halo to ~120 kpc using y-band cadence.
• Faint white dwarfs — a complete cooling-age sample within 100 pc, anchoring the local star formation history.
• Brown dwarfs and free-floating planets — y-band depth makes the local brown-dwarf census essentially complete out to 25 pc, and microlensing toward the Galactic bulge detects free-floating Earth-mass objects.

The Solar System catalogue

Rubin will roughly quintuple the known Solar System object inventory, from about 1.4 million catalogued bodies today to about 6 million. The cadence is well-matched to the typical inverse motion of main-belt asteroids (~0.5 arcsec/min), allowing reliable detection and orbital linkage by the Moving Object Processing System (MOPS). Different parts of the Solar System will be sampled differently:

Population	Current census	Rubin 10-yr expected	Headline science
Near-Earth objects ≥ 140 m	~ 11,000	~ 70 % of total population	Planetary defence, congressional NEO mandate
Main-belt asteroids	~ 1.3 × 10⁶	~ 5 × 10⁶	Size distribution, family ages, Yarkovsky drift
Jupiter Trojans	~ 12,000	~ 280,000	L4/L5 asymmetry, capture mechanism
Trans-Neptunian objects	~ 4,000	~ 40,000	Cold-classical / hot population census, Planet 9 tests
Comets	~ 4,000	~ 10,000	Activity statistics, dynamical lifetimes
Interstellar visitors	2	1 – 2 per year	Population statistics, composition

For the near-Earth population, Rubin is the dominant contributor to the U.S. Congressional mandate to catalogue 90 percent of NEOs larger than 140 metres. Predictions place the survey completeness at roughly 65-75 percent of the 140-metre population by the end of year 10, with the residual coverage gap filled by the planned NEO Surveyor space mission.

The transient sky

The static images get most of the press, but it is the difference-imaging output — the transient and variable sky — that defines Rubin's reach as a discovery engine. With ten million alerts per night, Rubin will roughly double the rate at which the optical transient sky is discovered. The science drivers span the full range of variability timescales:

• Multi-messenger counterparts. Rubin can tile the typical 100-square-degree gravitational-wave localisation in a single hour, then return every few hours to track the colour evolution. Kilonovae — the kind detected as GW170817 — drop two magnitudes per day; Rubin is the only optical survey wide and deep enough to find them across the full LIGO-Virgo-KAGRA horizon.
• Type Ia supernovae. A photometric sample of 10⁵ Type Ia events, with rest-frame ugriz light curves and host-galaxy spectroscopic priors from companion surveys, builds the deepest standardised Hubble diagram ever constructed.
• Tidal disruption events. A few hundred TDEs per year, with light curves out to peak — a clean probe of dormant SMBH demographics.
• AGN variability. Damped-random-walk light curves for millions of AGN over a decade — the cleanest reverberation-mapping dataset ever taken.
• Galactic variability. Eclipsing binaries, M-dwarf flares, cataclysmic variables, microlensing toward the bulge — at depths and cadences inaccessible to OGLE, ASAS-SN, ATLAS or ZTF.

Vera Rubin and the namesake

Vera Rubin's rotation-curve work with Kent Ford between 1970 and 1980 established that disk galaxies have approximately flat rotation curves out to many disk scale lengths. If the only mass present were the visible stars and gas, the orbital speeds would decline as v ∝ r^(-1/2) beyond the optical disk; instead they remain at 200-300 km/s out to whatever radius Hα emission can still be detected. The simplest interpretation — and the one now overwhelmingly accepted — is that the galaxies are embedded in extended halos of non-luminous matter whose mass exceeds the visible matter by an order of magnitude.

Rubin's work joined a growing thread that included Zwicky's 1933 cluster virial argument and the 21-cm rotation curves of M31 published by Roberts and Whitehurst earlier in the 1970s, but it was Rubin's systematic survey across dozens of galaxies that made the dark-matter case impossible to dismiss. The observatory that bears her name is now the leading instrument for testing the very hypothesis her measurements imposed on the field.

LSST Data Management: the software half of the project

Building the telescope is half the story. The LSST Data Management (DM) effort is the other half: a custom pipeline written in Python and C++ that ingests raw exposures, applies instrument signature removal, performs astrometric and photometric calibration, runs source detection, difference-images against templates, links sources into objects across visits, and produces the annual data releases. It runs in three places: at the U.S. Data Facility hosted by SLAC, the French data facility at CC-IN2P3, and the U.K. data facility at the Royal Observatory Edinburgh.

The two-minute alert latency is the tightest engineering constraint. Each exposure has to be reduced, difference-imaged, source-extracted and packaged into alerts before the next exposure arrives. With about 2,000 exposures per night and a finite number of processing nodes, every step in the pipeline has been profiled and optimised for predictable wall time.

How Rubin compares to previous and contemporary surveys

Survey	Aperture	FoV (deg²)	Single-visit depth (r)	Cadence / coverage
SDSS (2000-2008)	2.5 m	1.5	22.2	One-pass, ~14,000 deg² N. sky
Pan-STARRS PS1	1.8 m	7.0	22.0	Multi-pass N. sky, 5 bands
ZTF	1.2 m	47	20.5	2-3 night cadence, N. sky, transients
Dark Energy Survey (2013-2019)	4.0 m	3.0	23.6	5,000 deg² S. sky, 5 bands, weak lensing
HSC-SSP (Subaru)	8.2 m	1.8	26.0 coadd	1,400 deg², deep survey
Vera Rubin / LSST	8.4 m	9.6	24.7	3-night cadence, 18,000 deg², 6 bands

The combination of aperture, field of view and cadence is what isolates Rubin. ZTF has a comparable field of view at a comparable cadence but stops two magnitudes brighter than Rubin in a single visit. HSC matches Rubin's depth but covers about one-tenth the area and lacks a transient cadence. DES covered the same hemisphere with the same number of filters, but at a fifth the area-per-exposure and twice-shallower depth. Rubin replaces the bespoke surveys of the previous decade with a single facility that supersedes them all.

Timeline and operations

• 2003. Concept paper for an 8.4-metre wide-field survey telescope, then known as the LSST.
• 2014. Construction begins on Cerro Pachón with NSF and DOE funding.
• 2019. Renamed in honour of Vera C. Rubin.
• 2024. First photons on LSSTCam; system-level engineering imaging.
• 2025. Full operations begin; year-1 survey under way.
• 2026. DR1 (first annual data release) expected to public archive.
• 2035. Ten-year primary mission ends; final DR10 begins production.

Common pitfalls and misconceptions

• "LSST" is the survey, not the telescope. Since 2019 the telescope is the Simonyi Survey Telescope and the observatory is the Vera C. Rubin Observatory; LSST now refers strictly to the Legacy Survey of Space and Time being executed on those facilities.
• It is not a Hubble replacement. Rubin is a ground-based wide-field optical survey; Hubble's resolution and UV access are not addressed by Rubin. Rubin is closer in spirit to a souped-up SDSS than to a flagship space telescope.
• The alert stream is not a public stream of images. The 10⁷ nightly alerts are packets of metadata plus small cutouts; full-resolution images are released annually, not in real time.
• "3.2 gigapixels" refers to the camera, not a typical exposure on disk. Each raw exposure is roughly 6.4 GB once detector overheads and per-amplifier metadata are included; the 200 PB total archive size includes reduced data products, not just raw pixels.
• Photometric redshifts are not spectroscopic redshifts. Rubin's six-band photo-z accuracy of σ_z/(1+z) ~ 0.05 is impressive given the absence of spectra, but it is not a substitute for follow-up spectroscopy from facilities such as 4MOST, DESI or PFS.