A gravitational lens bends and focuses light from a distant source via the gravity of an intervening mass. It reveals dark matter, hidden galaxies and even the Hubble constant.
Imagine a perfectly taut trampoline surface. Place a bocce ball on it: it digs a well. Roll a marble across the surface. Its trajectory, instead of staying straight, curves as it passes close to the well, as if pulled. Replace the ball with the Sun, the marble with a light ray: you've just seen the principle of gravitational lensing.
In general relativity, mass doesn't 'pull' light Newtonian-style — mass curves spacetime, and light follows the curves of that spacetime (geodesics). Near a massive object, geodesics bend, and the image of a background source is distorted, amplified, sometimes doubled, or even multiplied into an arc or a ring.
Einstein calculated the deflection as early as 1915 in his nascent theory. For a ray grazing the solar surface, his prediction: 1.75 arcseconds — exactly twice the Newtonian value. The decisive test came on May 29, 1919. Arthur Eddington and Frank Dyson organised two expeditions to West Africa and Brazil to observe a total eclipse: for ~6 minutes, stars near the eclipsed Sun became visible. Comparing their positions to those measured six months earlier (when the Sun was far away), they confirmed the exact shift Einstein had predicted. On November 6, 1919, Eddington announced the results publicly. It made worldwide headlines — The Times ran 'REVOLUTION IN SCIENCE'. Einstein became famous overnight.
Since then, gravitational lensing has moved from theoretical curiosity to frontline observational tool. It lets us:
• Weigh galaxies and clusters, including their invisible dark matter. • See beyond the farthest directly observable galaxies, thanks to amplification. • Measure the Hubble constant independently of other methods. • Detect massive compact objects in the Milky Way (MACHOs, rogue exoplanets) via microlensing. • Constrain the nature of dark energy via statistical weak lensing.
Three distinct regimes, depending on effect intensity and source-lens alignment.
Strong lensing. Near-perfect alignment. The source is split into 2 to 5 images, stretched into arcs, or transformed into a complete ring (Einstein ring) when alignment is exact. The ring radius is:
θ_E = √((4GM/c²) · (D_LS)/(D_L · D_S))
with M the lens mass, D_L observer-lens distance, D_S observer-source distance, D_LS lens-source distance. For a typical galaxy lens and a source at z = 1, θ_E ≈ 1 arcsecond.
Weak lensing. Arbitrary alignment, subtle effect: background galaxies are slightly flattened in a preferential direction (shear). Detection requires averaging over thousands of galaxies. The cosmological tool par excellence: it maps total mass (including dark matter) over vast sky areas. Euclid (ESA, 2023), LSST/Rubin (2025) and Nancy Grace Roman (2027) are projects heavily dedicated to this.
Microlensing. A foreground star passes in front of a background star: the source temporarily brightens, peaking for days to months depending on the lens mass. No visible multiple image (except interferometrically), just amplification. Ground programs: OGLE (Poland, since 1992, galactic centre), MOA (New Zealand), KMTNet. Detects compact objects from planetary to stellar mass in the Milky Way disk, including exoplanets very far from their stars.
Fun peculiarity. Multiple images of a lensed quasar don't all show the same temporal phase — light follows paths of different lengths, introducing time delays from several days to several years. These measured delays (H0LiCOW project) yield H₀ directly.
A few lenses that became astrophysical icons.
Q0957+561. The first lensed quasar discovered (Walsh, Carswell, Weymann 1979). Two images (A and B) 6 arcseconds apart, from a single quasar at z = 1.41 lensed by a galaxy at z = 0.36. 417-day time delay between A and B.
Einstein Cross (Q2237+0305). A z = 1.69 quasar lensed by a nearby galaxy (z = 0.04) forming four images arranged in a cross around the galactic nucleus. Iconic image published by HST in 1990. A visual jewel.
Complete Einstein rings. Configurations with perfect source-lens-observer alignment. SDSS J0946+1006 (2008) is the first 'Horseshoe' observed. JVAS B1938+666, B1608+656, SDP.81 (ALMA, 2015) are among the most complete ever observed.
Abell 1689. A z = 0.18 galaxy cluster in Virgo, one of the most massive and most efficient lenses known. HST detected ~160 multiple images of background galaxies, building a mass map of unmatched precision. Its field reveals galaxies at z > 7.
Abell 370 and the 'Frontier Fields'. HST (2013-2018), then JWST (2022-), use six massive clusters as cosmic telescopes to see the most primitive galaxies. Local amplification factors ×10 to ×100. Thanks to this method, JWST-NIRCam detected multiple galaxies at z > 12 as early as 2022.
Icarus star (MACS J1149). First individual star ever observed at z = 1.49 (9 billion light-years away), via an amplification factor of several thousand from a stellar microlens superimposed on the cluster macro-lens. Detected by HST in 2018.
And for amateurs? Visually observing a gravitational lens is beyond amateur reach — separations are typically under 5 arcseconds and objects extremely faint (mag 20+). That said, photographing the Einstein Cross with a 300+ mm amateur telescope and a good sensor is possible (it's a favourite target of advanced deep-sky imagers).
Four main methodological streams.
High-resolution direct imaging. To resolve multiple images of strong lenses. HST (0.05'' resolution), JWST (0.03'' at 1 µm), radio interferometers (VLBI, EHT) for quasars. The largest ground telescopes (Keck, VLT, GTC) with adaptive optics.
Wide-field photometric surveys. To hunt lenses on the ground, among the billions of galaxies in surveys. SDSS, DES, HSC, then today Euclid (eventually 1.5 billion galaxies) and LSST/Rubin. AI now plays a growing role in automatically identifying lens candidates.
Microlensing alert networks. OGLE, MOA, KMTNet continuously monitor millions of stars toward the galactic centre. When a microlensing event begins, a worldwide alert triggers intensive follow-up to characterize the light curve and look for planetary perturbation signatures.
Weak-lensing cosmology. Measuring the statistical distortion of shear over millions of background galaxies, cell by cell, yields a 3D map of total matter. This map tightly constrains cosmological parameters (Ω_m, σ_8, dark-energy equation of state). Key surveys: CFHTLenS (2012), KiDS (2014-2019), DES (2013-2019), HSC (2014-), Euclid (2023-2029), LSST (2025-2035).
H₀ from time delays. The H0LiCOW project measured H₀ = 73.3 km/s/Mpc by observing time delays in six lensed quasars over several years (Nature, 2020). Independent of the distance ladder and the CMB, it lands on the local-measurement side (SH0ES) in the Hubble tension.
Several cosmic optical phenomena are related but distinct.
Dark matter. Gravitational lenses detect dark matter (by revealing more mass than luminous) but are not it. The tool is not the target.
Atmospheric refraction. Our terrestrial atmosphere also bends light (~0.5° at the horizon), but via a change of optical index in a material fluid. Totally different from gravitational deflection, which occurs in pure vacuum and involves spacetime itself.
Adaptive optics. Real-time compensation of atmospheric turbulence by a deformable mirror. Classical optical engineering, with no link to general relativity.
Scintillation. Rapid brightness variations of a star due to atmospheric turbulence (naked-eye) or interstellar-medium inhomogeneities (radio pulsars). Different phenomena — no massive lens involved.
Black holes and accretion disks. A black hole does lens background-source light (the 'shadow' of a black hole is seen via this mechanism — EHT image of M87* in 2019). But the lensing is an effect around the black hole, not the black hole itself.
Classical optical lenses. Glass lenses focus light by refraction (media of different indices). A gravitational lens doesn't focus the same way: it has no unique focal point, and its effect depends on the overall geometry of the mass field. It's not a cosmic magnifying glass in the strict sense, even if that phrase is sometimes used.
Because it validated Einstein's general relativity four years after publication. On May 29, 1919, Arthur Eddington's team (Príncipe, Africa) and Andrew Crommelin's team (Sobral, Brazil) measured stellar deflection by the Sun during a total eclipse: 1.75 arcseconds at the limb, exactly Einstein's prediction and twice the Newtonian one. The public announcement at the Royal Society on November 6, 1919, made world headlines and propelled Einstein to international scientific celebrity.
Four major uses: (1) weighing galaxies and clusters, including their invisible dark matter; (2) amplifying very distant sources we couldn't otherwise see (this is how JWST detects galaxies at z > 12); (3) measuring the Hubble constant independently via time delays in lensed quasars (H0LiCOW project: H₀ = 73.3 km/s/Mpc); (4) mapping total matter over large sky areas via statistical weak lensing — the great mission of Euclid and LSST.
A perfectly symmetric strong-lensing configuration in which source, lens and observer are rigorously aligned. Light from the source goes around the lens on all sides at once, forming a complete ring around the deflecting mass. Mathematically predicted by Einstein in 1936. The first half-ring was observed in 1988 (MG1131+0456), the first near-complete ring in 1998. SDSS J0946+1006 'Horseshoe' and SDP.81 (ALMA, 2015) are among the most spectacular.
Visually in an eyepiece, nearly impossible: objects are extremely faint (mag > 18-20) and separations under 5 arcseconds. But CCD/CMOS photography with a 250-300 mm telescope and long exposures can resolve the Einstein Cross (Q2237+0305, integrated mag 16.8) or the Horseshoe Einstein ring. It's one of the ultimate targets of advanced deep-sky astrophotography — a trophy for seasoned imagers.