Cosmic rays are energetic particles that travel through space and strike planetary atmospheres, spacecraft, and interstellar matter. Most are atomic nuclei, especially protons, accompanied by electrons, positrons, and small quantities of heavier particles and antiparticles. Despite their historical name, cosmic rays are primarily matter rather than rays of electromagnetic radiation. Their energies extend from millions of electronvolts to more than electronvolts, far beyond the energy attained by individual particles in human-built accelerators.
Cosmic rays connect astrophysics, particle physics, and planetary science. They carry evidence of stellar explosions and other powerful accelerators, help determine the physical and chemical state of interstellar gas, and produce cascades of secondary particles in Earth’s atmosphere. Their interpretation is difficult because magnetic fields deflect charged particles, obscuring their origins, while nuclear collisions alter their composition during travel. Research therefore combines measurements of particles, atmospheric showers, gamma rays, and neutrinos with models of acceleration and transport.
Discovery and the emergence of particle astrophysics
Cosmic-ray research began with a puzzle about the spontaneous discharge of insulated electrical instruments. Around the beginning of the twentieth century, penetrating radiation continued to ionize air even when known radioactive materials were removed. Radiation from radioactive substances in the ground explained part of the effect, but measurements at different heights did not consistently show the decrease expected if the ground were its only source.
In 1911, Domenico Pacini measured ionization underwater and found that it decreased compared with measurements near the surface. His experiments supported the existence of radiation originating outside the immediate terrestrial surroundings. The decisive altitude measurements were made by Victor Hess during balloon flights in 1912. On 7 August, Hess reached an altitude of about 5.3 kilometres and observed a substantial increase in ionization at high altitude. A flight during a solar eclipse had also shown that the radiation did not disappear when direct sunlight was blocked. Werner Kolhörster subsequently extended the measurements to greater heights. Hess received half of the 1936 Nobel Prize in Physics for discovering cosmic radiation.
The name “cosmic rays,” promoted by Robert Millikan during the 1920s, reflected the initially influential interpretation that the radiation consisted of very penetrating photons. Evidence gradually established that much of the primary radiation was electrically charged. Variations with geomagnetic latitude, investigated by Jacob Clay and others, showed that Earth’s magnetic field affected the incoming flux. Measurements of an east–west asymmetry subsequently demonstrated that positively charged primaries predominated.
Cosmic rays also provided a natural laboratory before high-energy particle accelerators became available. In 1932, Carl Anderson identified the positron in cloud-chamber photographs of cosmic-ray tracks. Anderson and Seth Neddermeyer reported evidence in 1936 for the particle now called the muon. In 1947, photographic-emulsion work by Cecil Powell, Giuseppe Occhialini, César Lattes, and collaborators established the charged pion, a meson important in nuclear interactions.
During the 1930s, coincidence circuits showed that separated detectors could register particles from the same atmospheric cascade. Work by Pierre Auger and collaborators in 1938 demonstrated extensive showers spreading across large areas. Cosmic-ray research consequently developed along two complementary lines: direct examination of individual particles and indirect reconstruction of extremely energetic primaries from their atmospheric descendants. After accelerators became the principal instruments for controlled particle discovery, cosmic rays retained a distinctive role as probes of energies, distances, and astrophysical environments inaccessible in laboratories.
Particle composition and the energy spectrum
The composition of cosmic rays depends on energy, location, and whether primary or secondary particles are being considered. Among the nuclei measured near Earth at roughly gigaelectronvolt energies, about nine-tenths are hydrogen nuclei and most of the remainder are helium nuclei. Heavier nuclei account for a much smaller number fraction, although they carry important information about sources and propagation. These percentages are approximate: they change with energy, solar activity, and the convention used to compare particles.
The elemental abundance pattern resembles ordinary astrophysical matter but is not identical to it. Lithium, beryllium, and boron are relatively abundant in cosmic rays compared with their abundances in the Solar System. Much of this material is generated when heavier nuclei, especially carbon and oxygen, fragment in collisions with interstellar gas. This process, called nuclear spallation, also produces some lighter isotopes. Ratios between secondary and primary species are therefore central diagnostics of cosmic-ray transport.
Cosmic-ray energy is usually expressed in the electronvolt and its multiples: GeV, TeV, PeV, and EeV correspond to , , , and eV. For nuclei, energy per nucleon is often more useful than total energy when comparing nuclear interactions. Magnetic effects are naturally described using rigidity,
where is momentum, is the speed of light, and is the particle’s charge. Particles with the same rigidity follow similar trajectories in a given magnetic field, even if their masses differ.
Across broad energy intervals, the differential intensity approximately follows a power law,
Below the “knee,” near a few eV, the all-particle spectral index is roughly . Above the knee, the spectrum steepens, with further structure near eV. The “ankle,” a hardening near several eV, occurs in the region associated with the changing balance between Galactic and extragalactic particles. A pronounced suppression appears above several eV.
These features are not sharp universal boundaries. Their measured positions and shapes depend partly on energy calibration and on how different nuclear components are combined. At the highest energies, the flux becomes extraordinarily small: around eV, the integral rate is of order one particle per square kilometre per century. Large collecting areas and long exposures are therefore essential. Electrons, positrons, and antiprotons have separate spectra whose shapes reflect their distinct production mechanisms and energy losses.
Astrophysical sources and their evidence
The Milky Way is thought to supply most cosmic rays below the transition to a predominantly extragalactic population. Expanding remnants of supernova explosions are leading candidate sources. A typical supernova releases about ergs of kinetic energy, equivalent to joules. Given a Galactic explosion rate of a few per century, conversion of roughly a tenth of this mechanical power into energetic particles would be sufficient, at order-of-magnitude level, to sustain the observed Galactic cosmic-ray population.
The evidence is stronger than an energy-budget argument alone. Radio and X-ray observations show that supernova-remnant shocks accelerate electrons to relativistic energies. Gamma-ray observations of remnants including IC 443 and W44 reveal the characteristic spectral signature of neutral-pion decay, providing direct evidence that energetic protons and nuclei collide with surrounding gas. However, these observations do not establish that every remnant accelerates protons to the energies required to explain the knee.
Other Galactic environments may contribute substantially. Winds from massive stars, clusters containing many such stars, and large cavities shaped by repeated explosions can sustain shocks and turbulence over extended periods. Pulsars and their surrounding nebulae efficiently accelerate electrons and positrons. These objects are plausible contributors to the local high-energy positron population, but their importance for the overall nuclear cosmic-ray population is a separate question. Compact binary systems and regions near the Galactic centre are also investigated as energetic-particle sources.
At the highest energies, plausible accelerators lie outside the Galaxy. Candidate environments include jets powered by black holes in active galactic nuclei, gamma-ray bursts, and energetic outflows associated with intense star formation. Large size, strong magnetic fields, and sufficient power are necessary conditions, but none alone proves that an object produces the observed particles.
Identifying a source requires several kinds of evidence. Gamma rays can indicate collisions of accelerated nuclei, but can also arise from energetic electrons. Neutrinos are particularly valuable because they can escape dense environments and travel without magnetic deflection. Their production in pion decays links them to hadronic interactions, although detecting a neutrino source does not automatically establish that it supplies a large fraction of cosmic rays reaching Earth.
Arrival directions offer another test. At sufficiently high rigidity, deflections may become small enough for broad source associations to emerge. Nevertheless, unknown nuclear charges, intervening magnetic fields, and limited event numbers complicate interpretation. A population can satisfy the required energy budget without yet being uniquely identified as the dominant source.
Acceleration mechanisms and energy limits
A magnetic field bends a charged particle’s trajectory but does not, by itself, increase its energy: the magnetic part of the Lorentz force acts perpendicular to velocity. Acceleration requires electric fields or interactions with moving electromagnetic structures. In astrophysical plasmas, such fields arise naturally in shocks, turbulent flows, rotating magnetized objects, and regions of magnetic reconnection.
Enrico Fermi proposed in 1949 that particles could gain energy through repeated encounters with moving magnetic irregularities. In its original form, now called second-order Fermi acceleration, the mean fractional energy gain scales approximately as , where is the characteristic speed of the scattering structures. Encounters can either increase or decrease a particle’s energy, but head-on encounters are statistically favoured. Related stochastic processes remain relevant to turbulent astrophysical environments.
A more efficient mechanism is diffusive shock acceleration, developed in its modern form during the 1970s. Particles scatter from magnetic irregularities on both sides of a shock and repeatedly cross between converging plasma flows. The average gain per cycle is first order in the flow speed divided by . Competition between repeated energy gain and the probability of escape naturally produces a power-law distribution.
For a strong, nonrelativistic shock in an ideal monatomic gas, the simplest test-particle theory gives a compression ratio close to four and a relativistic differential energy spectrum approximately proportional to . The spectrum observed near Earth is steeper because transport, escape, losses, and departures from the ideal assumptions modify the source spectrum. When accelerated particles exert appreciable pressure, they can alter the shock itself and amplify magnetic disturbances, making acceleration a coupled nonlinear process.
A basic confinement requirement is that the particle’s gyroradius fit inside the accelerating region. For an ultrarelativistic particle,
in SI units. Requiring , where is the source size, gives the basis of the Hillas criterion. For acceleration associated with a characteristic flow speed , a commonly used order-of-magnitude estimate is
This is a necessary estimate, not a sufficient guarantee. Acceleration must also outpace escape, source evolution, and energy losses. Electrons can lose energy rapidly through synchrotron radiation and inverse Compton scattering; nuclei may collide with gas or fragment in intense photon fields. Magnetic reconnection, which converts stored magnetic energy into particle and plasma energy, provides another route to acceleration, particularly in relativistic outflows and compact-object environments.
Transport through the Galaxy and intergalactic space
Charged cosmic rays do not usually travel directly from their sources to Earth. The Galactic magnetic field contains both large-scale structure and irregular components. Scattering by magnetic fluctuations changes particle directions, making transport approximately diffusive over sufficiently large distances and times. The observed intensity at a particular location therefore depends on both the source population and the history of propagation.
A schematic transport equation for a differential particle density is
Here represents injection, spatial diffusion, bulk transport, continuous energy losses, and an effective removal timescale. More complete models include momentum-space diffusion, radioactive decay, and production of one nuclear species through fragmentation of another. The equation is a framework rather than a unique description: several combinations of transport conditions can reproduce parts of the data.
The boron-to-carbon ratio is an especially important observable. Carbon is predominantly injected by sources, whereas much cosmic-ray boron forms during propagation. Its relative abundance constrains the integrated material traversed, often expressed as grammage in grams per square centimetre. The decline of this ratio with rigidity indicates that higher-rigidity particles generally escape more readily. Radioactive species such as beryllium-10 provide additional information about residence times. At gigaelectronvolt energies, characteristic Galactic residence times are of order ten million years, with model-dependent values.
Cosmic rays are not merely passive tracers. Their energy density near the Sun is of order eV per cubic centimetre, comparable in broad magnitude to several other components of the local interstellar environment. Their pressure can influence gas dynamics, and their streaming can generate waves that scatter subsequent particles. Interactions among cosmic rays, turbulence, and partially ionized gas are important in models of Galactic winds and star-forming environments.
Extragalactic propagation introduces additional processes. Ultrahigh-energy protons interacting with the cosmic microwave background can produce pions; this is the basis of the Greisen–Zatsepin–Kuzmin effect, predicted in 1966. Heavy nuclei can undergo photodisintegration on background radiation, losing nucleons rather than simply losing energy continuously. Pair production and the expansion of the Universe also reduce particle energy. Consequently, the most energetic observed particles predominantly sample a limited cosmic neighbourhood rather than the entire observable Universe. Their accessible distance depends on energy, nuclear species, and background photon fields.
Solar modulation and geomagnetic access
The cosmic-ray spectrum measured near Earth differs from the spectrum in the local interstellar medium. The solar wind, an expanding plasma flow from the Sun, carries a magnetic field through the heliosphere. Incoming Galactic particles scatter within this environment, undergo drift, and lose energy through the expansion of the flow. These effects, collectively called solar modulation, are strongest at lower energies and become progressively less important at high rigidity.
Galactic cosmic-ray intensities near Earth generally vary inversely with solar activity. Around solar maximum, the disturbed heliospheric environment usually reduces their intensity more strongly. The approximately 11-year activity cycle is accompanied by a 22-year magnetic-polarity cycle, which influences charge-dependent drift patterns. Electrons, positrons, protons, and antiprotons therefore need not respond identically. Passage of large solar-wind disturbances can cause temporary reductions called Forbush decreases, often developing rapidly and recovering over several days.
The Sun also generates its own energetic particles during eruptive events. Solar energetic particles are accelerated in environments associated with flares and coronal mass ejections. Their composition, timing, and energy spectra differ among events, reflecting different acceleration and transport conditions. Some eruptions accelerate particles to energies high enough for their atmospheric secondaries to increase ground-based neutron-monitor counts. Such episodes are called ground-level enhancements. Solar energetic particles are physically distinct from the continuously arriving Galactic population, even though both are often discussed under the broad heading of cosmic radiation.
Voyager 1 and Voyager 2, which crossed the heliopause in 2012 and 2018 respectively, provided direct measurements of low-energy particles outside the main solar-wind cavity. These measurements constrain the local interstellar spectrum and help separate source and Galactic-transport effects from heliospheric modulation. Connecting them to measurements closer to Earth requires time-dependent transport models rather than a single universal correction.
Earth adds a further magnetic filter. Its magnetosphere excludes many low-rigidity particles, particularly near the magnetic equator, while allowing easier access at high magnetic latitudes. The cutoff depends on particle rigidity, arrival direction, and location. It is not simply a fixed energy threshold, because different charges and nuclear species correspond to different rigidities at the same energy.
These filters explain why cosmic-ray measurements differ between polar and equatorial stations, between ground level and orbit, and across the solar cycle. They also distinguish cosmic rays from the trapped populations of Earth’s radiation belts: trapped particles circulate within the geomagnetic environment, whereas most Galactic cosmic rays enter it from outside and pass through or interact.
Atmospheric interactions and extensive air showers
When a high-energy primary nucleus enters the atmosphere, it can collide with an air nucleus and produce many secondary particles. Further collisions and decays create an extensive air shower, a branching cascade that develops over kilometres of atmosphere and can spread across a large ground area. The atmosphere functions as a massive, naturally occurring particle detector, although reconstructing the primary requires an understanding of the entire cascade.
The hadronic component contains nucleons and newly produced mesons, especially pions and kaons. Neutral pions decay rapidly into photons, feeding an electromagnetic cascade. Charged pions can either collide again or decay, producing muons and neutrinos. At high energies, relativistic time dilation increases decay lengths, allowing interactions to compete more effectively with decay. This competition shapes the energy spectra and angular distributions of atmospheric muons and neutrinos.
The electromagnetic component grows mainly through photon conversion into electron–positron pairs and electron or positron bremsstrahlung. Successive generations divide the available energy among increasing numbers of particles. Growth eventually ends when ionization losses and other low-energy processes dominate. The atmospheric depth at which the shower reaches its greatest particle population is called , usually measured in grams per square centimetre.
At equal total energy, showers initiated by heavier nuclei generally reach maximum development higher in the atmosphere than proton showers and fluctuate less from event to event. A useful approximate picture treats a nucleus of mass number and energy as several nucleon subshowers with characteristic energies near . Actual interactions are more complicated, so composition estimates rely on numerical simulations and high-energy hadronic models.
Muons are especially important at ground level because they penetrate much farther than most shower electrons and photons. Their relatively long lifetime, amplified by time dilation, allows many to reach the surface from production altitudes of several kilometres. A frequently used rough rule for the integrated sea-level muon flux is about one muon per square centimetre per minute, although the value depends on energy threshold, orientation, altitude, and local conditions.
Atmospheric showers also emit detectable light and radio signals. Excited nitrogen molecules produce ultraviolet fluorescence approximately related to energy deposited in the air. Relativistic charged particles emit Cherenkov radiation when their speed exceeds the phase velocity of light in air. Coherent radio emission arises mainly from geomagnetic separation of shower electrons and positrons, together with the shower’s net negative charge. These different signals provide complementary measurements of shower energy, direction, and development.
Instruments and reconstruction methods
Direct cosmic-ray instruments operate above most of the atmosphere, on satellites, space stations, and high-altitude balloons. They can identify individual particles before atmospheric interactions erase much of the primary information. Magnetic spectrometers measure track curvature to determine rigidity and charge sign. Silicon trackers, scintillators, time-of-flight systems, Cherenkov detectors, and calorimeters supply complementary information about charge, velocity, trajectory, and energy.
The Alpha Magnetic Spectrometer, AMS-02, has operated on the International Space Station since 2011. PAMELA, launched in 2006, made important measurements of charged particles and antiparticles. Instruments including CREAM and ISS-CREAM, CALET, and DAMPE have extended direct observations into high-energy regimes. Different designs emphasize nuclear charge resolution, electron identification, precision rigidity measurements, or calorimetric energy determination. No single configuration optimizes all of these simultaneously.
At sufficiently high energies, direct detection becomes impractical because the flux is too low for the available collecting area. Ground arrays instead sample extensive air showers. Scintillation counters and water-Cherenkov detectors measure particles arriving at separated locations. Relative arrival times constrain the shower direction; the lateral signal distribution and total signal provide energy estimators. Dedicated muon measurements help separate electromagnetic and hadronic components and constrain primary composition.
The Pierre Auger Observatory in Argentina combines a surface array covering approximately 3,000 square kilometres with fluorescence telescopes. The Telescope Array in Utah uses a related hybrid strategy over a smaller surface area. Fluorescence observations provide a nearly calorimetric measurement of deposited atmospheric energy but generally require dark, clear conditions. Surface instruments operate with a much higher duty cycle. Events observed by both systems connect surface signals to the fluorescence energy scale.
Other techniques occupy different energy and observational ranges. Imaging atmospheric Cherenkov telescopes are chiefly used for gamma-ray astronomy, but their cameras record showers generated by cosmic rays as well. High-altitude arrays such as HAWC and LHAASO exploit the larger surviving particle population near shower development. Radio antennas can observe large shower footprints and provide an additional route to measuring energy and longitudinal evolution.
Reconstruction is inseparable from calibration and measurement uncertainty. Atmospheric density affects shower development; aerosols and clouds affect optical transmission; detector gains and timing must remain controlled. Simulations translate measured signals into primary properties, introducing dependence on particle-interaction models. Experiments therefore report both statistical and systematic uncertainties and compare several composition-sensitive observables. Differences between energy scales can shift apparent spectral features, making cross-calibration important when combining results from separate observatories.
Environmental effects, applications, and unresolved problems
Cosmic rays are a persistent source of ionization in planetary atmospheres and interstellar gas. In dense molecular clouds, where ultraviolet radiation is strongly attenuated, energetic particles and their secondary electrons initiate reaction networks that help establish the abundance of molecular ions. Low-energy cosmic rays are especially important to this chemistry, but their distribution is difficult to infer because solar modulation restricts direct measurements and their limited ranges make their intensity vary between environments.
On Earth, cosmic-ray interactions produce cosmogenic nuclides. Neutron capture by atmospheric nitrogen generates carbon-14, the basis of radiocarbon dating. Beryllium-10 and chlorine-36 are among the nuclides used to investigate surface exposure, erosion, and past environmental changes. Interpreting their abundance requires accounting for production rates, shielding, transport, and radioactive decay. Records in tree rings and ice cores also preserve changes associated with solar activity and geomagnetic shielding, rather than recording a simple, unmodified measure of the Galactic flux.
Muon radiography uses penetrating cosmic-ray muons to investigate large structures. Differences in attenuation reveal variations in integrated density along different directions, enabling studies of volcanoes, archaeological structures, and industrial installations. Other applications exploit neutron scattering or absorption to estimate environmental water content. In electronics, energetic secondary particles can generate single-event effects, including memory upsets; reliability studies therefore consider altitude, shielding, device design, and the particle environment.
Cosmic radiation also contributes to human radiation exposure. Atmospheric shielding greatly reduces exposure at the surface, while aircraft crews and spacecraft occupants encounter different and generally higher cosmic-radiation environments. Galactic heavy ions and solar particle events present different challenges because their composition, timing, and energy deposition differ. Radiation assessments depend on particle transport and biological effectiveness, not merely on particle counts. Cosmic-ray ionization also produces atmospheric aerosol particles under some conditions, but ion formation alone does not establish a substantial cloud or climate response.
Several major physical questions remain unresolved. The relative contributions of supernova remnants and other Galactic accelerators are not fully established, especially near the knee. The energy and composition of the Galactic-to-extragalactic transition remain model-dependent. At ultrahigh energies, the observed spectral suppression may combine propagation losses with limits on source acceleration; the spectrum alone does not uniquely distinguish them.
Air-shower interpretation presents another important problem. Observatories have reported more muons in high-energy showers than several widely used interaction models predict for plausible compositions. Resolving this discrepancy requires improved measurements and modelling of hadronic interactions beyond energies directly accessible in accelerators. Likewise, positron and antiproton measurements constrain conventional sources and possible dark-matter processes, but an excess relative to one background model is not by itself evidence for a new particle. Progress depends on jointly explaining spectra, composition, arrival directions, and secondary radiation with physically consistent source and transport models.