A scattering amplitude is a complex-valued quantity describing how an incoming physical state is transformed into a particular outgoing state by an interaction. Its absolute square, combined with appropriate kinematic and normalization factors, determines measurable scattering probabilities and cross sections. Its phase controls interference between alternative processes. Scattering amplitudes therefore contain more information than the event rates that experiments directly record.
The concept appears throughout quantum mechanics, wave physics, nuclear physics, and particle physics. In nonrelativistic potential scattering, an amplitude specifies the strength and phase of an outgoing spherical wave in each direction. In quantum field theory, an invariant amplitude connects incoming and outgoing multiparticle states and provides a central link between an interaction theory and experimental observations. Related amplitudes describe electromagnetic, acoustic, and other classical waves, although their normalization and physical dimensions depend on the problem.
Scattering amplitudes can be calculated from equations of motion, expanded into angular-momentum channels, constrained by symmetry and probability conservation, or reconstructed from their singularities. Their study joins practical calculations for experiments with questions about locality, causality, and the mathematical structure of physical theories.
Wave amplitudes and measurable cross sections
The basic nonrelativistic example is the elastic scattering of a spinless particle by a localized, time-independent potential. Far outside the interaction region, a stationary wave function with an incident beam along the -axis has the asymptotic form
Here is the incident wave number, is distance from the scattering center, and specify the outgoing direction. The first term is an incident plane wave. The second is an outgoing spherical wave, with scattering amplitude . This expression assumes an interaction sufficiently short-ranged for free-wave asymptotics to apply. The unscreened Coulomb potential requires a modified asymptotic form because its influence persists to arbitrarily large distances.
The relation to observation follows from probability current. For a particle of mass , the incident current density is , where is the reduced Planck constant. The outgoing radial current decreases as , while a detector subtending solid angle intercepts area . Dividing the scattered rate by the incident flux gives the differential cross section
Consequently, has dimensions of length in three spatial dimensions. A cross section has dimensions of area, but it need not represent the geometric area of an object: interference, resonances, and wave propagation can make it much larger or smaller than a classical estimate.
For transitions between different nonrelativistic channels, outgoing and incoming speeds need not agree. With a common wave-amplitude convention,
where and are the corresponding relative speeds. Other definitions absorb this flux ratio into the amplitude. Comparisons between formulas therefore require attention to state normalization, reduced masses, spin conventions, and whether the quoted quantity is an amplitude or an already normalized cross section.
The complex nature of an amplitude is essential. If two indistinguishable mechanisms contribute, the combined amplitude is , not a sum of probabilities. The cross section includes the interference term . A single angular distribution generally determines only , leaving phase information unresolved. Access to phases usually requires interference with a known contribution, polarization measurements, or additional theoretical constraints. An overall common phase is unobservable, but relative phases can decisively change angular distributions and transition rates.
Potential scattering and the Born expansion
For a nonrelativistic particle, scattering amplitudes can be derived from the stationary Schrödinger equation. For two interacting bodies, the relative-motion problem uses their reduced mass . Writing the Hamiltonian as , where describes free motion and the interaction, the incoming scattering state satisfies the Lippmann–Schwinger equation
The infinitesimal specifies an outgoing-wave boundary condition. It is not an extra physical interaction; it chooses the solution appropriate to an incident beam followed by outgoing scattered particles.
For a local potential in three dimensions, this equation gives
using an incident plane wave of unit amplitude. Elastic scattering has . The formula shows that the outgoing amplitude depends on the interaction acting on the full, distorted incident state, rather than on the potential alone. Even a simple potential can produce a complicated amplitude because the wave may undergo repeated scattering within the interaction region.
Replacing the exact wave function inside the integral by the incident plane wave yields the first Born approximation:
Thus, at this order, the amplitude is proportional to the spatial Fourier transform of the potential at the wave-vector transfer . The physical momentum transfer is . Large-angle scattering at fixed incident momentum generally probes more rapidly varying spatial structure than forward scattering.
For example, a Yukawa potential gives
The parameter sets the interaction range. For elastic scattering, , so the result predicts a forward-enhanced distribution. This example exhibits the connection between spatial range and momentum-transfer dependence: a long-range interaction produces a narrower concentration near the forward direction.
Repeated substitution in the integral equation generates the Born series, conventionally expressed through the transition operator as
Its terms describe successive interactions separated by free propagation. Weak potentials and suitable incident energies can make the first term useful, but the series is not universally convergent. Near a bound-state threshold or a narrow resonance, repeated interactions may be as important as the first one. Numerical wave-equation methods, variational approaches, or nonperturbative resummations are then needed.
Partial waves, threshold behavior, and resonances
For a central potential, rotational symmetry permits an expansion into channels of definite orbital angular momentum. For elastic scattering of spinless particles,
where is a Legendre polynomial and is the scattering-matrix element in the th channel. If no inelastic process is available, conservation of probability requires . One can therefore write , giving
The real quantity is the phase shift caused by the interaction relative to free radial propagation.
Orthogonality of the Legendre polynomials gives the total elastic cross section,
For distinguishable spinless particles, each partial wave consequently obeys the bound . At low momentum, the centrifugal term suppresses penetration into the interaction region for higher . Short-range scattering is therefore often dominated by the , or -wave, channel. Quantum statistics can alter which channels are allowed for identical particles.
The low-energy -wave amplitude is conveniently expressed through the scattering length and effective range :
For finite , sufficiently small , and a distinguishable single-channel system, the elastic cross section approaches . This compact expression is useful because many microscopic details enter only through a few low-energy parameters. A scattering length much larger than the interaction range signals a nearby threshold pole; depending on its location, this may correspond to a shallow bound state or a virtual state rather than to an ordinary resonance above threshold.
When inelastic channels are open, a diagonal partial-wave element can be written , with . The missing elastic flux goes into other channels, and the corresponding reaction cross section is
An isolated resonance often produces rapid phase motion and an approximately Breit–Wigner energy dependence. Its width is related to a characteristic lifetime by . Actual observed peaks can be displaced or distorted by thresholds, background amplitudes, and channel coupling. A peak alone is therefore not a complete definition of a resonance; the pole structure and phase behavior provide a more precise characterization.
The scattering matrix and relativistic amplitudes
The scattering matrix, or -matrix, provides a general framework for transitions between asymptotically free incoming and outgoing states. Its matrix element is
Writing separates unchanged free propagation from interactions. For a connected relativistic process, a common convention defines the invariant amplitude by
Here one-particle states have covariant normalization, and natural units are used. The four-dimensional delta function enforces conservation of total energy and momentum. Although and the nonrelativistic amplitude express closely related physics, they have different normalizations and generally different dimensions.
For two incoming particles with four-momenta and masses , the differential cross section is
where the Lorentz-invariant final-state phase space is
The bar conventionally denotes averaging over unobserved initial spin or color states and summing over unobserved final ones. Identical final particles require an appropriate symmetry factor when the integration counts equivalent configurations more than once.
For a two-body final state in the center-of-mass frame,
before any needed identical-particle factor. The quantity is the squared center-of-mass energy. This formula makes explicit how an interaction amplitude becomes an observable only after multiplication by kinematic and phase-space factors. The same amplitude formalism also describes decays, although a decay rate has a different initial-state normalization and flux factor.
Relativistic amplitudes are often calculated from correlation functions of quantum fields. The Lehmann–Symanzik–Zimmermann reduction formula relates scattering amplitudes to residues of single-particle poles in time-ordered correlation functions, with the external propagators removed. It provides a bridge between local field equations and asymptotic particle transitions.
This construction presupposes suitable asymptotic states. Unstable particles are generally characterized through poles and their contributions to processes involving stable observable products, rather than treated as exact external states. Likewise, quarks and gluons are useful perturbative external objects in short-distance calculations, but confinement prevents their appearance as isolated asymptotic particles in experiments.
Unitarity, analyticity, crossing, and symmetry
Probability conservation imposes the operator condition , known as unitarity. With , this becomes
Its significance is dynamical: the imaginary part of an amplitude is constrained by sums over physically allowed intermediate states. A calculation of one process cannot be chosen independently of all others if the complete theory is to conserve probability. In the nonrelativistic short-range normalization introduced earlier, the forward-scattering consequence is the optical theorem,
The total cross section includes both elastic and inelastic channels. Even an experiment counting particles scattered away from the beam is thus connected to a phase-sensitive property of the forward wave.
Relativistic two-to-two amplitudes are commonly expressed using Mandelstam variables,
which satisfy . Analytic continuation treats these variables as complex quantities. Stable exchanged particles produce poles, while multiparticle thresholds produce branch points and cuts. Resonances are associated with poles reached by continuation onto appropriate nonphysical sheets. This analytic structure records what states can propagate and what intermediate processes can become physically accessible.
Under suitable locality, causality, and growth assumptions, dispersion relations connect real and imaginary parts of amplitudes. Schematically, an amplitude can be reconstructed from integrals over its discontinuities, supplemented by subtraction constants and explicit pole terms. Subtractions are needed when high-energy behavior prevents an unsubtracted integral from converging. These relations are especially valuable when absorptive information can be constrained by measured cross sections, even if a direct microscopic calculation is difficult.
Crossing symmetry relates processes obtained by exchanging an incoming particle with the corresponding outgoing antiparticle. For example, scattering and pair-annihilation amplitudes may arise as different physical regions of one analytically continued function. Crossing is not simply the assertion that their measured cross sections are equal: the allowed momenta, spin assignments, flux factors, and physical thresholds differ between channels.
Other symmetries constrain the amplitude’s tensor structure and allowed transitions. Rotational and Lorentz symmetry restrict dependence on momenta and spins. Exchange symmetry requires amplitudes for identical bosons to be symmetric and those for identical fermions to be antisymmetric under complete particle exchange. In a gauge theory, Ward identities ensure that unphysical polarization components do not affect observable scattering. These constraints frequently eliminate candidate terms before any integral is evaluated, and they provide powerful checks on both analytic and numerical results.
Feynman diagrams, loops, and finite predictions
In perturbative quantum field theory, amplitudes are organized as expansions in interaction strengths. A Feynman diagram represents a contribution to this expansion: external lines specify particles, internal lines represent propagators, and vertices encode interaction terms in the Lagrangian. Tree diagrams contain no closed loops. Loop diagrams involve integrations over undetermined internal momenta and describe quantum corrections. Individual diagrams are generally not observables; only appropriate sums possess the required symmetry and physical interpretation.
Electron–muon scattering in quantum electrodynamics illustrates the structure. Its leading contribution is photon exchange. Apart from a convention-dependent overall phase, the amplitude contains
The two spinor bilinears are electromagnetic currents, and is the exchanged four-momentum. The propagator contributes the dependence. Squaring the complete amplitude, summing or resolving spins, and integrating over the desired kinematic region produces a cross section. More complicated processes combine many such terms, with interference between diagrams essential to the result.
Loop integrals can contain ultraviolet divergences arising from arbitrarily large internal momenta. Renormalization introduces a regulator and counterterms, expressing predictions in terms of consistently defined masses, couplings, and field normalizations. The intermediate expressions depend on a renormalization prescription and scale. Exact observables do not depend on this arbitrary scale, although a calculation truncated at finite order retains residual dependence. That dependence is one diagnostic of omitted higher-order contributions, not itself a physical variation of the interaction.
Massless particles introduce a different issue. Virtual soft photons and real emission of photons with very low energy produce infrared singularities when considered separately. Probabilities for suitably inclusive measurements combine these contributions so that the singularities cancel. In high-energy quantum chromodynamics, collinear splittings add another source of singular behavior. Infrared-safe observables, such as appropriately defined jet quantities, are designed not to change under the addition of an arbitrarily soft particle or an exactly collinear splitting.
For collisions involving hadrons, perturbative partonic amplitudes are combined with parton distribution functions and other ingredients that encode long-distance physics. A short-distance amplitude is therefore only one component of a collider prediction. Fixed-order calculations may also require resummation when large logarithms make successive terms numerically important. Precision predictions ultimately depend on a consistent combination of virtual corrections, real radiation, phase-space integration, experimental event definitions, and nonperturbative input where necessary.
On-shell methods and the organization of amplitudes
Traditional diagrammatic calculations can become inefficient as the number of external particles increases. Many diagrams contain gauge-dependent terms that cancel only after extensive algebra. On-shell methods instead emphasize physical external states, pole residues, and factorization. When an intermediate momentum approaches a physical particle pole, a tree amplitude has the schematic behavior
up to overall conventions. The sum runs over the intermediate particle’s physical states. A complicated amplitude can therefore be constrained by simpler amplitudes and by the requirement that every allowed pole have the correct residue.
For massless particles in four dimensions, spinor-helicity variables encode a null momentum as
Lorentz-invariant spinor contractions are written and . This notation expresses helicity amplitudes compactly and avoids unnecessary components of polarization vectors. A famous example is the Parke–Taylor formula, obtained in 1986, for a color-ordered tree-level gluon amplitude with two negative-helicity gluons and all remaining gluons of positive helicity. With overall coupling and phase conventions suppressed, its kinematic form is
The compact result contrasts sharply with the number of diagrams appearing in a direct expansion.
Britto–Cachazo–Feng–Witten recursion, developed in 2005, reconstructs suitable tree amplitudes from lower-point on-shell amplitudes. Two external momenta are deformed by a complex parameter while preserving momentum conservation and their mass-shell conditions. The deformed amplitude becomes a complex function whose finite poles correspond to factorization channels. Cauchy’s theorem then relates the physical amplitude to residues at those poles, provided the contribution from infinity vanishes or is separately determined. The method’s applicability depends on the theory and the chosen deformation.
Generalized unitarity applies related ideas to loops. Putting selected internal propagators on shell exposes products of tree amplitudes, allowing coefficients of loop integrals or representations of loop integrands to be reconstructed. A complete treatment must account for terms that may be missed by an insufficient set of cuts, and for regulator-dependent contributions. These methods are central to many high-multiplicity and higher-order gauge-theory calculations.
Another structural connection is double-copy construction, in which gauge-theory ingredients generate gravity amplitudes. Relations between open- and closed-string amplitudes were established by Kawai, Lewellen, and Tye in 1986. Later color–kinematics duality organized gauge-theory numerators so that replacing color factors by a second set of kinematic numerators yields gravity amplitudes in many cases. The construction reveals relations between theories with very different field descriptions, while the scope of general all-order formulations remains an active research question.
Experimental reconstruction and physical applications
Experiments observe detector signals, event counts, and distributions, not complex amplitudes directly. For a collider process, an idealized expectation is , where is integrated luminosity. Real analyses must also account for detection efficiency, acceptance, resolution, and backgrounds. Since a cross section involves an absolute square, extracting amplitudes from data is an inverse problem: several complex amplitudes can reproduce the same limited set of intensity measurements.
Angular distributions and polarization observables supply complementary information. With spin present, a process generally has multiple independent helicity or spin amplitudes. Their interference generates correlations that disappear from an unpolarized total rate. Partial-wave analyses fit angular and energy dependence to infer channel amplitudes, often imposing unitarity and analyticity. Such analyses can retain discrete ambiguities or depend on assumptions about unmeasured channels. Resonance identification is consequently stronger when phase motion, several decay modes, and analytically consistent pole determinations support the same interpretation.
In atomic and ultracold-gas physics, low-energy amplitudes determine collision rates, interaction shifts, and thermalization. A magnetic Feshbach resonance occurs when a bound state associated with a closed channel is tuned near an open-channel threshold. Near an isolated resonance, the scattering length is often parameterized as
where is a background value, the resonance position, and a width parameter. This tunability has made scattering amplitudes central to experiments on Bose–Einstein condensates, interacting Fermi gases, and universal few-body behavior.
In nuclear physics, nucleon–nucleon amplitudes constrain nuclear forces through phase shifts and spin observables. Near thresholds, effective-range expansions and low-energy interaction theories reduce complicated short-distance dynamics to a controlled set of parameters. In hadron physics, coupled-channel amplitudes are important because several final states can share quantum numbers and influence the same resonance. Calculations using lattice quantum chromodynamics can relate discrete energy levels in a finite spatial volume to infinite-volume scattering information, rather than reading an ordinary scattering cross section directly from a Euclidean simulation.
Wave-scattering applications extend beyond collisions between elementary particles. In X-ray crystallography, waves scattered from different atomic positions add coherently, and their arrangement produces a structure factor. Measured intensities constrain its magnitude, while missing phase information creates the crystallographic phase problem. Neutron scattering probes nuclear positions and magnetic structure, with amplitudes depending on isotope and spin. Acoustic and electromagnetic scattering similarly use complex amplitudes to characterize targets, reconstruct spatial structure, or distinguish coherent propagation from incoherent intensity sums.
Historical development and limits of the framework
Scattering was already a powerful investigative tool before quantum mechanics. Ernest Rutherford interpreted large-angle alpha-particle scattering in 1911 using a concentrated positively charged nucleus. The resulting angular law belongs historically to classical Coulomb scattering, but it became a benchmark for quantum collision theory. Its continued importance also illustrates a subtle point: agreement of a cross section with a classical expression does not imply that the underlying quantum amplitude has no nontrivial phase structure.
Max Born’s work on collisions in 1926 helped establish the probabilistic interpretation of quantum mechanics. Outgoing wave coefficients were connected to probabilities for different collision outcomes, complementing the general Born rule for quantum measurement. Partial-wave methods subsequently gave systematic descriptions of atomic and nuclear collisions. The resonance analysis developed by Gregory Breit and Eugene Wigner in the 1930s connected collision-energy dependence with unstable intermediate states and remains foundational for resonance spectroscopy.
Werner Heisenberg promoted the -matrix as a central object in the 1940s. Postwar developments in quantum electrodynamics supplied practical perturbative calculations, while the reduction formalism of Harry Lehmann, Kurt Symanzik, and Wolfhart Zimmermann clarified the relation between field correlation functions and scattering. In the 1950s and 1960s, dispersion relations and the analytic -matrix program sought to determine strong-interaction amplitudes from general principles. Geoffrey Chew’s bootstrap approach emphasized consistency among observable hadronic processes rather than a privileged set of elementary constituents.
The establishment of quantum chromodynamics gave the strong interaction a successful gauge-field description, but amplitude-based reasoning remained valuable. Unitarity, crossing, dispersion relations, and factorization continued to constrain calculations. From the late twentieth century onward, compact helicity formulas, generalized unitarity, recursion relations, and double-copy constructions transformed both computational practice and the search for deeper mathematical organization. Modern amplitude research thus combines Lagrangian field theory with methods that use observable on-shell information more directly.
The framework has identifiable limits. A conventional -matrix assumes well-defined incoming and outgoing states separated by sufficiently simple asymptotic evolution. Long-range electromagnetic interactions complicate the use of bare charged-particle states, motivating inclusive observables or states accompanied by appropriate soft radiation. In general relativity and cosmology, the availability of an asymptotic scattering description depends on the geometry and its boundary conditions; a cosmological spacetime need not admit the same particle interpretation as flat spacetime.
Current problems include nonperturbative reconstruction of strongly interacting amplitudes, reliable analytic continuation from finite and uncertain data, and the treatment of overlapping resonances and many coupled channels. Positivity bounds derived from dispersive arguments constrain some low-energy interactions under stated assumptions, but require care in the presence of massless exchange. In gravity, amplitude methods expose unexpectedly strong cancellations and relations to gauge theories, while the ultraviolet behavior of candidate quantum theories remains dependent on their field content, symmetries, and perturbative order.