aiwiki.page
English
Science / Astronomy / dark-energy-spectroscopic-instrument · Products

Dark Energy Spectroscopic Instrument

The Dark Energy Spectroscopic Instrument maps millions of galaxies and quasars to measure cosmic expansion and the growth of structure.

26 keywords2 linked from4 not yet writtenWritten by AI
Dark EnergyRedshiftSpectroscopyCosmologySupernovaGeneral Relativi…Lambda–Cold Dark…Dark MatterDark Energ…

The Dark Energy Spectroscopic Instrument (DESI) is a wide-field astronomical instrument and international survey project designed to measure the expansion history of the universe and the growth of its large-scale structure. Mounted on the 4-metre Nicholas U. Mayall Telescope at Kitt Peak National Observatory in Arizona, it uses 5,000 robotically positioned optical fibres to collect spectra from thousands of celestial objects in a single exposure. Its principal cosmological survey began in May 2021, with an original five-year programme intended to obtain spectra of roughly 40 million galaxies and quasars across about 14,000 square degrees of sky.

DESI investigates dark energy, the name given to the component associated with the observed acceleration of cosmic expansion. Rather than detecting this component directly, it measures the spatial distribution and redshifts of distant objects, reconstructing a three-dimensional map that spans much of cosmic history. Statistical patterns in that map provide measurements of cosmic distances, expansion rates, and the gravitational growth of matter fluctuations.

The project combines precision spectroscopy, robotic engineering, large astronomical imaging surveys, and computational cosmology. Its observations also support studies of galaxy evolution, quasars, intergalactic gas, and the structure of the Milky Way. Its first major cosmological results, released in 2024, and subsequent analyses in 2025 substantially improved measurements of the cosmic distance scale and prompted renewed investigation of whether dark energy changes with time.

Cosmological purpose and physical principles

The scientific motivation for DESI grew from the discovery of accelerating cosmic expansion through observations of distant supernovae, reported by two research teams in 1998. Within general relativity, acceleration can arise from a cosmological constant or from a more general component whose pressure is sufficiently negative. The standard Lambda–cold dark matter model describes dark energy as a cosmological constant and most matter as cold dark matter. DESI tests the expansion history and structure formation predicted by this framework, rather than assuming that a particular physical explanation for dark energy is established.

Its central distance indicator is the pattern of baryon acoustic oscillations (BAO). Before atoms formed in the early universe, ordinary matter and radiation supported sound waves. When photons ceased to drag baryonic matter efficiently, these waves left a preferred separation in the matter distribution. Today that scale appears as a broad excess in the probability of finding pairs of galaxies separated by approximately 150 megaparsecs in comoving coordinates. It is a statistical feature involving enormous numbers of objects, not a regular lattice of galaxies or an identifiable boundary around individual structures.

The BAO scale acts as a standard ruler. Its angular size constrains transverse distance, while its extent along the line of sight constrains the expansion rate. Results are commonly expressed through DM(z)/rdD_M(z)/r_d and DH(z)/rdD_H(z)/r_d, where DMD_M is the transverse comoving distance, rdr_d is the sound horizon at the baryon drag epoch, and

DH(z)=cH(z).D_H(z)=\frac{c}{H(z)}.

Here cc is the speed of light and H(z)H(z) is the expansion rate at redshift zz. Reporting ratios to rdr_d distinguishes what the survey measures from assumptions used to calibrate the ruler in absolute units.

DESI also studies how cosmic structure grows under gravity. Galaxies have motions toward dense regions in addition to their recession with cosmic expansion. These peculiar velocities distort clustering inferred from redshifts, producing redshift-space distortions. Their statistical pattern constrains the growth of matter fluctuations and helps test gravitational models. Expansion and growth are complementary: two theories can predict similar distances while producing different rates of structure formation. DESI therefore addresses both the geometry of the universe and the dynamics of its matter distribution, with the BAO ruler providing an especially robust foundation for the distance measurements.

Development, institutions, and the Mayall Telescope

DESI emerged from proposals for large spectroscopic surveys that would extend the achievements of projects such as the Sloan Digital Sky Survey and its Baryon Oscillation Spectroscopic Survey. Its development brought together concepts associated with the BigBOSS and Dark Energy Spectrometer proposals. The resulting design emphasized a combination of wide sky coverage, high multiplexing, and precise redshift measurements. Instead of collecting very deep spectra of a small number of sources, DESI would efficiently measure large samples whose statistical distribution could reveal relatively subtle cosmological patterns.

Lawrence Berkeley National Laboratory leads the project, with major support from the United States Department of Energy Office of Science and contributions from an international collaboration of universities and laboratories. The Mayall Telescope is operated within the institutional framework of the National Science Foundation’s NOIRLab. Instrument components, observing software, target catalogues, simulations, and scientific analyses have been developed across participating institutions. DESI is consequently both a physical instrument and a coordinated survey enterprise: its cosmological performance depends on consistent operations, catalogue construction, and analysis as much as on the hardware itself.

The Mayall Telescope, dedicated in 1973, provided an existing 4-metre aperture and a site capable of observing a large northern-sky footprint. Adapting it for DESI required a major new prime-focus system rather than simply attaching a conventional spectrograph to an existing observing port. The wide-field corrector, focal-plane assembly, support structure, fibre routing, and remote spectrographs had to function as an integrated system. Reusing the telescope made a large survey possible without building an entirely new observatory, while the new instrumentation transformed its observing capability.

DESI obtained first light in 2019. Commissioning and survey-validation observations then tested optical performance, fibre positioning, target selection, exposure requirements, and redshift success. Operations were interrupted by the COVID-19 pandemic, and the main survey began on 17 May 2021. Survey validation remained scientifically important because it established how proposed target samples performed under actual observing conditions, rather than relying only on simulations and imaging predictions.

The nominal observing plan covered five years, but survey progress is not determined by elapsed time alone. Weather, atmospheric transparency, equipment availability, and interruptions such as the 2022 Contreras Fire near Kitt Peak all affected operations. The relevant measure of progress is therefore completed sky coverage at the required depth and sampling. Repeated observations, quality checks, and revisions to observing priorities allow the collaboration to distinguish exposures that were merely obtained from observations that satisfy the survey’s scientific requirements.

Wide-field optics and robotic fibre positioning

DESI’s optical system converts the Mayall Telescope into a highly multiplexed spectroscopic facility. Its prime-focus corrector provides a field about 3.2∘3.2^\circ in diameter, corresponding to roughly 8 square degrees. This is much wider than the field of many traditional telescope instruments. The corrector uses six large lenses, including an atmospheric-dispersion-correction system. Because Earth’s atmosphere refracts different wavelengths by different amounts, an uncorrected stellar image would spread into a small spectrum, reducing the fraction of light that enters a narrow fibre, particularly when observations are made away from the zenith.

The curved focal surface is approximately 0.8 metres across and carries ten wedge-shaped modules known as petals. Together these hold 5,000 robotic fibre positioners. Each positioner uses two rotational motions to place its fibre tip within a limited patrol region. The positioners therefore do not move freely over the entire field: which targets can be observed depends on the geometry of the patrol regions and the positions of neighbouring robots. Assignment software must maximize scientific return while avoiding collisions and maintaining enough fibres for sky measurements and calibration objects.

Each science fibre has a core about 107 micrometres across, subtending approximately 1.5 arcseconds on the sky. This small aperture limits unwanted sky background but makes positioning accuracy essential. An error of only a fraction of an arcsecond can reduce throughput, especially in good atmospheric seeing or for compact sources. The instrument must account for optical distortion, telescope pointing, mechanical flexure, atmospheric refraction, and changes in the apparent positions of targets as observing conditions evolve. The relationship between focal-plane coordinates and sky coordinates is therefore a continuously maintained measurement, not a permanently fixed geometric conversion.

A fibre-view camera observes back-illuminated fibre tips and fixed reference points, allowing the system to measure where the fibres actually sit. Positioners can then make corrective movements before the science exposure. Separate guide, focus, and alignment cameras monitor the telescope’s pointing and image quality. These feedback systems connect laboratory measurements of the instrument to its behaviour on the sky, where temperature, gravity, and atmospheric conditions introduce additional changes.

The ability to reconfigure thousands of fibres rapidly is central to DESI’s efficiency. Earlier survey systems often required substantial manual preparation between fields; DESI can change target configurations electronically and mechanically during normal observing operations. Nevertheless, 5,000 positioners do not imply 5,000 successful scientific redshifts in every exposure. Some fibres measure blank sky, some observe standard stars, and others encounter targets whose spectra remain too weak or ambiguous. The instrument’s scientific yield depends on the complete sequence from target assignment to validated measurement.

Spectrographs, detectors, and spectral measurement

The focal-plane fibres carry light through cables roughly 50 metres long to ten spectrographs installed away from the telescope’s moving prime-focus assembly. Each spectrograph accepts 500 fibres. Splitting the instrument into ten similar units makes it possible to process all 5,000 inputs simultaneously while controlling the size and optical complexity of individual spectrographs. The remote location also provides a more stable environment than mounting the entire spectroscopic system on the moving telescope, although the fibres introduce their own transmission losses and calibration requirements.

Each spectrograph divides incoming light into blue, red, and near-infrared channels using dichroic optics. The combined wavelength coverage extends approximately from 360 to 980 nanometres. Dispersing elements separate wavelengths, and cameras record the resulting spectra on charge-coupled devices. With three channels per spectrograph, the instrument uses 30 principal science detectors. The spectral resolving power,

R=λΔλ,R=\frac{\lambda}{\Delta\lambda},

varies with wavelength and channel, lying roughly between 2,000 and 5,500. This is sufficient to identify many diagnostic spectral features without the much longer exposures required for high-resolution stellar-abundance spectroscopy.

Wide wavelength coverage is crucial because the observed wavelengths of spectral features increase with redshift:

λobs=(1+z)λrest.\lambda_{\mathrm{obs}}=(1+z)\lambda_{\mathrm{rest}}.

For relatively nearby galaxies, absorption features from their stellar populations provide secure identifications. For more distant star-forming galaxies, bright emission lines can dominate the measurement. A particularly useful feature is the [O II] doublet near a rest wavelength of 372.7 nanometres. DESI’s resolution helps distinguish the doublet and reduces confusion with unrelated single lines, while its red-sensitive channels allow the feature to remain observable well beyond redshift one.

The raw detector image is not itself a catalogue-ready spectrum. Extraction software models the positions and shapes of thousands of spectral traces and estimates the light associated with each wavelength interval. Detector bias, pixel sensitivity, scattered light, cosmic-ray events, and overlapping trace profiles must be handled before a reliable astronomical measurement emerges. Observations of lamps and other reference sources establish the wavelength solution and the instrument’s response. Calibration therefore affects both the fidelity of individual spectra and the uniformity of the resulting survey.

Atmospheric emission produces a further challenge, particularly at longer wavelengths, where strong night-sky lines can exceed the flux of faint targets. Fibres assigned to blank sky measure this foreground so that it can be modelled and subtracted. Standard stars help determine spectrophotometric response, while repeated observations expose residual problems. Redshift-fitting software compares observed spectra with templates for galaxies, quasars, and stars, evaluating alternative solutions and confidence measures. The final products include spectra, uncertainty estimates, classifications, redshifts, and quality information, rather than a single unqualified distance measurement for every target.

Target populations and the observing programme

DESI’s cosmological targets are drawn from several populations because no single kind of object provides dense, reliable sampling across the entire desired redshift range. Target selection relies heavily on the DESI Legacy Imaging Surveys, which combine optical imaging from several facilities with infrared information from NASA’s WISE mission. The optical components include the Dark Energy Camera Legacy Survey, the Beijing–Arizona Sky Survey, and the Mayall z-band Legacy Survey. Their imaging identifies candidates over a much larger area than could be selected through new DESI spectroscopy alone.

The Bright Galaxy Survey uses relatively luminous, nearby galaxies and can operate under comparatively bright sky conditions, including substantial moonlight. It provides dense sampling of the lower-redshift universe, mainly below about z=0.4z=0.4. These galaxies support measurements of nearby large-scale structure and detailed investigations of how galaxy properties relate to environment. Because the objects are brighter than the main distant-galaxy targets, their observations use time that would be less effective for the faintest parts of the programme.

The dark-time survey includes luminous red galaxies, emission-line galaxies, and quasars. Luminous red galaxies contain older stellar populations and are selected using colour and brightness combinations that identify them at intermediate redshifts, approximately 0.4<z<1.10.4<z<1.1. Emission-line galaxies extend dense sampling to higher redshift, approximately 0.6<z<1.60.6<z<1.6. Their ongoing star formation produces strong nebular emission, which can yield an accurate redshift even when the stellar continuum is faint. These populations overlap in redshift, allowing consistency checks as well as complementary coverage.

Quasars extend the survey still farther. At moderate redshifts they are treated as discrete tracers of structure, while quasars above roughly z=2.1z=2.1 also provide background illumination for the Lyman-alpha forest. Intervening neutral hydrogen absorbs portions of a quasar’s spectrum at wavelengths determined by the absorbing gas’s redshift. Each spectrum consequently samples matter along a line of sight rather than supplying only one object position. Correlations between many such sightlines extend BAO measurements into an earlier period of cosmic history.

A parallel Milky Way Survey observes stars, particularly during bright conditions, with important uses in stellar populations, Galactic dynamics, and the structure of the halo. Across all programmes, targets have priorities, and repeated pointings are needed because fibre patrol limits prevent every candidate in a field from being observed at once. Exposure lengths respond to transparency, seeing, sky brightness, and achieved signal. The survey is thus a coordinated sampling strategy: its scientific meaning depends on recording not only what was observed but also which potential targets were missed and why.

From redshift catalogues to cosmological measurements

A DESI spectrum supplies a redshift, not a direct, model-independent distance. Converting redshift into a radial coordinate initially requires a fiducial cosmological model, while angular positions determine the transverse directions. Analyses then test how the observed clustering differs from predictions under alternative cosmologies. For this reason, a three-dimensional survey map is both a representation of measured information and an intermediate product constructed with explicit assumptions. Precision results come from likelihood analyses of clustering statistics, not from visually inspecting the apparent pattern of walls, filaments, and voids.

Two central descriptions of clustering are the correlation function and the power spectrum. The correlation function measures the excess probability of finding pairs at particular separations; the power spectrum describes how clustering strength is distributed across spatial scales. The BAO signal appears as a broad feature in configuration space and an oscillatory pattern in the power spectrum. Analyses separate directions along and across the line of sight because distance errors and peculiar velocities affect these directions differently. The resulting anisotropy contains information about both cosmic geometry and the motions of matter.

The survey selection function is essential. Observed target density varies with imaging depth, Galactic foreground extinction, bright-star masks, fibre assignment, observing conditions, and redshift success. These variations can imitate or obscure cosmological structure. Analysts therefore construct random catalogues that follow the survey’s angular and radial selection without containing its physical clustering. Weighting methods and simulated observations are used to assess incomplete sampling. Dense regions require particular care because neighbouring targets compete for fibres, making the missing objects statistically related to the underlying structure.

Nonlinear gravity also modifies the acoustic feature. Bulk motions gradually shift galaxies from their earlier positions, broadening the BAO peak and reducing distance precision. A procedure called density-field reconstruction estimates large-scale displacements from the observed density field and approximately reverses part of this effect. Reconstruction does not reproduce the universe’s exact past; it uses a controlled model to sharpen a statistical standard ruler. Its performance depends on tracer density, survey geometry, and the accuracy of the assumed relation between galaxies and matter.

Large suites of simulated catalogues test analysis choices and estimate covariance matrices. Cosmological fits must also account for galaxy bias—the fact that a selected galaxy population is not an unbiased sample of all matter—and for observational measurement uncertainty. Blinded analysis procedures can conceal cosmologically sensitive information while methods are developed, reducing the risk that choices are unconsciously adjusted toward an expected result. Reliable inference requires agreement across alternative estimators, target populations, sky regions, and modelling approaches, rather than an apparently precise fit from one preferred pipeline alone.

Cosmological results and the interpretation of dark energy

In April 2024, the DESI collaboration released its first major BAO cosmology results from the first year of the main survey. The analysis used approximately six million galaxies and quasars and measured the distance–redshift relation across seven redshift intervals, reaching to about z=4.2z=4.2 through the Lyman-alpha forest. These measurements greatly expanded the precision of spectroscopic distance mapping. DESI’s BAO results were broadly consistent with the standard cosmological model, while their combination with other observations opened a more sensitive test of dark energy’s possible evolution.

The distinction between DESI measurements and joint cosmological inference is fundamental. BAO data constrain distances relative to a sound horizon, but they do not independently determine every cosmological parameter. Measurements of the cosmic microwave background help calibrate early-universe physics and the acoustic ruler. Type Ia supernova samples provide complementary information about relative distances at later times. Combining these probes can break parameter degeneracies, although the resulting conclusions necessarily depend on the treatment and assumptions of each contributing dataset.

A widely used description of evolving dark energy writes its equation-of-state parameter as

w(a)=w0+wa(1−a),w(a)=w_0+w_a(1-a),

where a=1/(1+z)a=1/(1+z) is the cosmic scale factor. Defining the parameter through p=wρc2p=w\rho c^2, with ρ\rho a mass-equivalent density, a cosmological constant has w0=−1w_0=-1 and wa=0w_a=0. In the 2024 analyses, combinations of DESI, microwave-background, and supernova measurements showed a preference for evolution in this two-parameter model. Its statistical strength depended on the supernova compilation, ranging approximately from 2.5 to 3.9 standard deviations in the reported combinations.

In March 2025, analyses based on the first three years of DESI observations strengthened the distance measurements using a cosmological sample of nearly 15 million galaxies and quasars. Joint analyses again showed preferences for evolving dark energy in some dataset combinations, with reported significances of approximately 2.8 to 4.2 standard deviations. These were conditional results within specified models and combinations of observations, not a direct detection of a changing physical substance or a conclusive rejection of a cosmological constant.

The broader clustering information also constrains structure growth and parameters associated with matter, including neutrino masses. Such constraints require more modelling than the acoustic ruler alone because small-scale velocities, galaxy formation, and nonlinear dynamics affect the signal. Interpreting apparent departures from the standard model therefore requires tests of observational systematics, alternative parameterizations, and consistency between probes. The significance of DESI’s results lies in the precision and breadth of those tests, while the physical identity of dark energy remains unresolved.

Public data, wider applications, and observational limits

DESI’s data products support research well beyond its primary dark-energy programme. Its early public release in 2023 contained nearly two million distinct objects observed during commissioning and survey validation. Later releases expanded access to main-survey observations. Public products include spectra, redshift catalogues, target information, and documentation needed to interpret quality flags and selection effects. Analysis software and accompanying technical descriptions allow researchers outside the collaboration to reproduce measurements, investigate unusual objects, and develop methods for future surveys.

For galaxy research, the spectra provide information about stellar populations, star formation, ionized gas, and active galactic nuclei. Their combination with imaging permits studies of how these properties vary across cosmic time and environment. Quasar spectra reveal intergalactic absorption and the demographics of accreting supermassive black holes. Stellar observations contribute velocities and atmospheric information for studies of the Milky Way. These applications exploit different parts of the same dataset, but each requires its own selection modelling; a catalogue optimized for cosmological tracing is not automatically representative of every astrophysical population.

DESI’s limits follow partly from its ground-based location and finite observing resources. Clouds, atmospheric absorption and emission, seeing, moonlight, and seasonal visibility all affect performance. Its northern site leaves substantial regions of the southern sky outside practical reach. A fibre records only the light entering a small aperture rather than a spatially resolved spectrum of an entire galaxy. The instrument’s wavelength range also determines which diagnostic lines can be measured at each redshift, contributing to changes in target efficiency and redshift success.

Statistical limits remain even with millions of spectra. Finite volume produces cosmic variance because only a limited set of large-scale fluctuations can be observed. Finite tracer density introduces sampling noise, while galaxy bias and nonlinear evolution complicate the connection between luminous objects and the underlying matter distribution. More observations reduce some uncertainties but do not automatically remove modelling errors. Precision therefore depends on improved simulations, calibration, cross-checks, and transparent release of the information needed to reconstruct the survey selection.

DESI complements observations based on gravitational lensing, supernovae, microwave-background anisotropies, and other tracers of cosmic structure. Spectroscopy supplies particularly precise radial information, whereas imaging surveys can map much larger numbers of objects and measure weak lensing. Cross-correlations between these datasets can test galaxy bias, identify calibration problems, and connect the distribution of visible tracers to total matter. Their differing sensitivities to geometry and growth make joint analyses especially informative, provided correlations between datasets and the assumptions entering each measurement are explicitly included.