In 1835 the philosopher Auguste Comte gave an example of knowledge that humans could never have: what the stars are made of. Their distance, he argued, put chemistry out of reach for ever. Within 25 years he was wrong. In 1859 Gustav Kirchhoff and Robert Bunsen showed that each chemical element prints its own pattern of lines in light, and those same lines were already known as dark gaps in sunlight. In 1868, during a total solar eclipse, astronomers saw a bright yellow line in the Sun's edge that matched no known element. It was named helium, after the Greek for Sun, and it was found on Earth only in 1895.
That is the astronomer's situation in one story. You cannot touch a star, so you collect its light, spread it into a rainbow, and read it. A graduate of this course knows how big a mirror must be to see a faint thing, how sharp it can ever be, why colour tells temperature, and why the lines tell composition and motion. Telescope pictures are never just pictures.
A telescope does not mainly make things bigger. It collects more light than your pupil can. Light collected goes as the area of the mirror or lens, which is the square of its diameter D, the aperture. A dark-adapted pupil is about 7 mm across. A 70 mm lens collects 100 times as much light. A 2.4 m mirror such as Hubble's collects about 117,000 times as much.
This is why astronomers build ever bigger mirrors. A faint galaxy sends so few photons that a bigger bucket is the only way to collect enough in a night. The 6.5 m James Webb Space Telescope has 18 hexagonal gold-coated beryllium segments and about 25.4 m² of collecting area. The Extremely Large Telescope, being built by ESO on Cerro Armazones in Chile, will have a primary mirror of about 39 m; ESO's plan, as listed in 2025, gives technical first light in March 2029.
Ideal circular aperture, ignoring central obstructions and gaps between segments (JWST's real area is 25.4 m², less than the ideal circle). Magnitude gain compares with a 7 mm pupil.
Without JavaScript: a 200 mm telescope collects about 820 times as much light as a 7 mm pupil (a gain of 7.3 magnitudes); Hubble's 2.4 m collects about 117,000 times (12.7 magnitudes); a 39 m mirror collects about 31 million times.
Aperture is not the only thing that counts: detectors matter as much. Photographic plates, the astronomer's detector for a century, record only a few percent of the light that hits them. The charge-coupled device (CCD), invented at Bell Labs in 1969 by Willard Boyle and George Smith (Nobel Prize 2009), records most of it. Modern astronomical CCDs have quantum efficiencies above 90% at their best wavelengths. Switching from plates to CCDs was like building a telescope several times larger for free.
Light collected is half the story. The other half is sharpness: how close two points can be and still look like two. Light is a wave, so a perfect lens still smears a point of light into a small disc with rings, the Airy pattern. The smallest angle you can separate is set by the wavelength λ and the aperture D.
Two things follow. A bigger aperture is sharper, and a longer wavelength is blurrier. A 200 mm telescope in green light has a diffraction limit near 0.7 arcsecond. Hubble at the same wavelength: 0.058 arcsecond. A radio dish at 21 cm needs to be enormous to match: even a 40 m dish at 21 cm resolves only about 22 arcminutes, two-thirds of the width of the full Moon.
On the ground there is a second limit. Air is turbulent, and turbulence blurs a star's image to about an arcsecond ("seeing") at a good site, whatever the mirror's size. Past roughly 12 cm of aperture, a ground-based optical telescope is limited by the sky, not by its mirror. Two cures exist: go above the air (Hubble, JWST), or measure the blur many times a second and flex a deformable mirror to cancel it (adaptive optics), often with a laser creating an artificial guide star.
Seeing is taken as 1 arcsecond, a typical good site, combined with the diffraction limit in quadrature. Adaptive optics is treated as ideal, which real systems approach only partly and mostly in the infrared.
Without JavaScript: in visible light a 12 cm telescope has a diffraction limit of 1.2 arcseconds, about equal to ground seeing; Hubble reaches 0.058 arcsecond and the Moon feature it could resolve is about 110 m across.
The way round diffraction is to join telescopes. Combine signals from dishes thousands of kilometres apart (very long baseline interferometry) and the "aperture" becomes the separation between them. The Event Horizon Telescope used dishes across the Earth at 1.3 mm, giving a diffraction-limited resolution of roughly 25 microarcseconds: enough to image the shadow around the black hole in the galaxy M87 (announced 10 April 2019) and the one at the centre of our own galaxy (12 May 2022). The radio telescope FAST in Guizhou, China, completed in 2016, has a 500 m dish (about 300 m used at once) and a beam about 2.9 arcminutes wide at 21 cm, a good example of a very big collector that is not a sharp one.
A telescope has two lenses or mirrors: the objective that gathers light and makes an image at its focal length F, and an eyepiece of focal length f that magnifies the image. Magnification is just their ratio. Galileo's spyglass of 1609 had about 20 times; he used it to see the Moon's mountains and four moons of Jupiter, published in Sidereus Nuncius (1610). Hans Lipperhey of Middelburg had applied for a patent on a spyglass in the Netherlands in 1608, and who really invented it is argued over.
Refractors bend light with a lens and suffer from colour fringes, because glass bends blue more than red; the largest ever, at Yerkes (102 cm, 1897), is close to the practical limit because a big lens can only be held at its edge and sags. Isaac Newton built a reflecting telescope about 1668, using a curved mirror, which has no colour fringes and can be supported from behind. Every large research telescope since is a reflector.
Higher power is not better. Beyond about twice the aperture in millimetres, the image just gets larger and dimmer, because diffraction and seeing set the detail. A very short focal-length eyepiece may need more power than the sky allows.
Assumes an eyepiece with a 52° apparent field, typical of a simple design. The rule of "twice the aperture in millimetres" is a common guideline, not a law; steady air can allow more.
Without JavaScript: a 200 mm, 1,200 mm focal-length telescope with a 25 mm eyepiece gives 48 times, an exit pupil of 4.2 mm and a field of about 1.1°; with a 4 mm eyepiece it gives 300 times, near the practical limit of about 400 for this aperture.
Visible light is a narrow slice of the electromagnetic spectrum, which runs from gamma rays (wavelengths smaller than an atom) to radio waves (longer than a football pitch). Different parts need different telescopes, and the atmosphere lets only some through. The window is open for visible light, near-infrared patches and radio waves from about a centimetre to ten metres or so; gamma rays, X-rays and most ultraviolet and infrared are absorbed. That is why we launch telescopes, and why ALMA sits on a dry Chilean plateau at 5,000 m.
Frequency = c / wavelength; photon energy = hc / wavelength. Band edges are conventional and differ slightly between books; atmospheric transmission depends on altitude and weather.
Without JavaScript: green light at 550 nm has a frequency of about 545 THz and a photon energy of 2.3 eV; the 21.1 cm hydrogen line is 1.42 GHz; an X-ray at 1 nm has about 1,240 eV.
Every hot, dense object glows with a smooth spectrum whose shape depends on one thing, its temperature. This is blackbody radiation, which Max Planck explained in 1900. Two simple results hold for stars. Wien's law (1893) says the peak wavelength is inversely proportional to temperature, so hotter means bluer. The Stefan–Boltzmann law says the power per unit area goes as T⁴, so a star's luminosity is its surface area times T⁴.
The swatch integrates the Planck curve against fitted CIE 1931 colour-matching functions (Wyman, Sloan and Shirley 2013) and converts to sRGB, brightest channel scaled to full. It is an approximation; screens differ and your eye adapts, so the Sun looks white. Presets use approximate literature values.
Without JavaScript: the Sun (5772 K) peaks near 502 nm; a 3,600 K red supergiant such as Betelgeuse peaks near 805 nm in the infrared and, at about 760 solar radii, shines roughly 90,000 times brighter than the Sun; a 25,000 K white dwarf peaks in the ultraviolet near 116 nm.
Spread the light through a prism or grating and a star's rainbow is crossed by dark lines. Joseph von Fraunhofer mapped hundreds in sunlight in 1814. Kirchhoff and Bunsen explained them in 1859 with three rules. A hot, dense body glows with a continuous spectrum. A hot, thin gas glows only at certain wavelengths: an emission spectrum. And a cooler gas in front of a hot source takes those same wavelengths out of the continuum: an absorption spectrum. Every element has its own set.
The reason came from quantum mechanics (see Quantum Mechanics, 34/83). An atom has only certain energy levels, and a photon is emitted or absorbed when an electron jumps between them. For hydrogen the jumps down to level 2 make the visible Balmer series, found by Johann Balmer in 1885 and explained by Bohr in 1913:
Wavelengths in air (nm). Hydrogen lines are computed from the Rydberg formula with a standard air correction; the others are standard laboratory values. Colours are approximate.
Without JavaScript: hydrogen's Balmer lines fall at about 656.3, 486.1, 434.1 and 410.2 nm; sodium's two D lines at 589.0 and 589.6 nm; calcium's H and K lines near 397 and 393 nm.
Line strengths give more than composition. A star's temperature and pressure decide which ionisation states exist and which levels are occupied, so the pattern of lines orders stars into the spectral classes O, B, A, F, G, K, M (hot to cool), worked out at Harvard around 1900 and refined by Annie Jump Cannon. In 1925 Cecilia Payne showed from the lines that stars are mostly hydrogen and helium. Nebulae show strong lines of doubly ionised oxygen, a green glow first mistaken for a new element, "nebulium", until Ira Bowen explained it in 1927.
A source moving toward you has its light squeezed to shorter wavelengths (a blueshift); one moving away is stretched (a redshift). For speeds well below light, the fractional shift equals the speed along the line of sight divided by the speed of light.
Because the rest wavelength of every line is known from the lab, the shift gives a speed to a precision no ruler could match. Two big results rest on it. In 1995 Michel Mayor and Didier Queloz found a planet around 51 Pegasi (Nobel Prize 2019, shared) from the star's tiny back-and-forth wobble: Jupiter makes the Sun move at about 12.5 m/s, an Earth at about 9 cm/s, which at 550 nm is a shift of only about 2 × 10⁻⁵ nm for Jupiter. And in the 1920s Edwin Hubble and others found that nearly every distant galaxy is redshifted, more so the farther it is.
That second kind of redshift is not quite Doppler. In an expanding universe the wavelength is stretched along with space, and the redshift z gives the factor (1 + z) by which the universe has grown since the light left. The most distant galaxy confirmed with a spectrum, as of 2025, MoM-z14, has z = 14.44, so its light has stretched by a factor of about 15.4.
Rest wavelength of H-alpha taken as 656.3 nm (air). The lower slider multiplies wavelengths by 1 + z; it is not a speed. JWST covers 0.6 to 28.5 µm.
Without JavaScript: a source receding at 300 km/s moves H-alpha from 656.3 to 657.0 nm, and Earth's own 30 km/s orbit moves it by 0.07 nm. At z = 2 H-alpha is at 1.97 µm and at z = 14.44 it is at 10.1 µm, deep in the mid-infrared where JWST's MIRI instrument works.
This is the practical reason for infrared telescopes. The light from the first galaxies, emitted as visible and ultraviolet, has been stretched into the infrared; to see it you need a cold telescope above the atmosphere, which is what JWST (launched 25 December 2021, first science images released 11 July 2022) is. The Hubble tension and what the lines say about the universe's expansion belong to Astrophysics and Cosmology (34/88).
Light is no longer the only messenger. On 14 September 2015 the LIGO detectors recorded gravitational waves from two merging black holes (announced 11 February 2016; Nobel Prize 2017 to Rainer Weiss, Barry Barish and Kip Thorne). On 17 August 2017 detectors saw a merger of two neutron stars, and telescopes across the spectrum saw the gamma-ray burst and the afterglow: the first event caught in both gravitational waves and light. Neutrinos from the supernova SN 1987A were seen at Kamiokande in Japan on 23 February 1987, and IceCube reported a high-energy neutrino traced to a blazar in 2018. Astronomers call the combination multi-messenger astronomy.
Light itself is being collected in new ways. The Vera C. Rubin Observatory in Chile, with an 8.4 m mirror and a 3.2-gigapixel camera, released its first images on 23 June 2025, and its ten-year survey of the southern sky was scheduled to start in mid-2026. The data volumes there are why Astronomical Data & Computation (36/44) exists.
Light carries almost everything we know about the sky. Aperture sets how faint, wavelength over aperture sets how sharp, the continuum sets temperature, the lines set composition, and their shifts set speed. Each bigger mirror, better detector and new window has changed what astronomers could say.
Try this: look at a bright star and a planet on the same night. Notice the colours of two bright stars (Betelgeuse and Rigel in Orion, or Antares and Arcturus), and write which is cooler. Then look at a street lamp through a CD held at an angle: you are using a spectrograph.
Sources. Rayleigh criterion and Wien, Planck, Stefan–Boltzmann laws as in standard texts (Carroll and Ostlie 2017); Kirchhoff and Bunsen 1859–60; Fraunhofer 1814; Balmer 1885; Bohr 1913; Edlén's 1966 formula for the refractive index of air; Wyman, Sloan and Shirley, "Simple analytic approximations to the CIE XYZ color matching functions," Journal of Computer Graphics Techniques 2013; IAU 2015 Resolution B3 (nominal solar values); CODATA 2018 constants; Mayor and Queloz, Nature 1995; Boyle and Smith 1970 and the 2009 Nobel Prize in Physics; Abbott et al. (LIGO and Virgo), Physical Review Letters 2016 and 2017; NASA and ESA mission pages for JWST (launch 25 December 2021; first images 11 July 2022; 25.4 m²; 0.6 to 28.5 µm); ESO pages for the ELT (39 m; first light planned March 2029); the Event Horizon Telescope Collaboration 2019 and 2022; Vera C. Rubin Observatory (first look 23 June 2025); the MoM-z14 redshift (z = 14.44, 2025). All numbers in the interactives were computed in node. This piece takes no side in open questions such as the Hubble tension, which it only names.