Creep is the progressive deformation of a material under sustained mechanical stress. Unlike an instantaneous elastic response, it develops over time and depends strongly on temperature, loading history, and material structure. It can occur at stresses below the short-term yield stress and may eventually produce excessive distortion, cracking, or rupture. In metals, creep becomes especially important at elevated temperatures; in polymers, concrete, wood, ice, and some other materials, significant creep occurs under ordinary service conditions.
Creep governs the usable life of equipment such as turbine blades, steam pipes, and high-temperature pressure vessels, while also influencing the deflection of buildings, the settlement of ground, and the movement of glaciers. Its underlying mechanisms range from atomic diffusion and dislocation motion to molecular rearrangement, moisture-assisted deformation, and sliding between grains. The term therefore describes a class of time-dependent responses rather than a single microscopic process.
Physical meaning and controlling variables
Mechanical strain is the change in a body's dimensions relative to its original dimensions. In a uniaxial test, engineering strain is , where is the initial gauge length. Creep strain is the time-dependent component that develops while a stress remains applied. The creep rate, , is its rate of accumulation, conventionally expressed in reciprocal seconds or reciprocal hours. A rate that appears extremely small in a laboratory test may become important over decades: a constant rate of , for example, produces about 0.32% strain in ten years.
Stress, temperature, and elapsed time are the principal external variables, but they are insufficient to specify creep behavior by themselves. Grain size, crystallographic orientation, prior deformation, chemical composition, moisture content, and environmental exposure can all change the response. A material's current internal state also matters. Two specimens under identical present conditions may creep differently if one has previously experienced a higher temperature or a different loading sequence.
For crystalline materials, temperature is often expressed as the homologous temperature , where both the operating temperature and melting temperature are measured in kelvin. Appreciable creep in many engineering metals becomes important above roughly –, although this is a practical guide rather than a universal boundary. Lead, with a melting point near 601 K, is already at approximately at room temperature. Tungsten at the same temperature is at only about . Their room-temperature creep behavior is consequently very different.
Creep is related to, but distinct from, stress relaxation. In creep, stress is maintained and strain evolves; in stress relaxation, total deformation is held approximately fixed while stress decreases. A stretched bolt constrained by a joint may lose clamping force through relaxation, whereas a suspended weight may progressively elongate its support through creep. Both behaviors can arise from the same constitutive properties.
The permanence of creep deformation depends on the material. High-temperature creep of metals is usually treated as irreversible plastic deformation. In viscoelastic materials, some delayed strain can recover after unloading, while another portion may remain. Creep also differs from fatigue, which is associated with repeated loading, although sustained loading and cyclic loading frequently interact in actual components.
Creep curves and the development of rupture
A conventional tensile creep test applies a load to a specimen at controlled temperature and records its extension with time. The resulting creep curve often contains three recognizable stages: primary, secondary, and tertiary creep. An initial elastic strain, and sometimes an immediate plastic strain, occurs when the load is applied. These initial contributions are distinguished from the subsequent time-dependent strain.
During primary creep, the creep rate decreases. In many metals this occurs because deformation creates additional obstacles to dislocation motion, producing strain hardening. At the same time, elevated temperature enables recovery processes that rearrange or remove those obstacles. Primary creep reflects the evolving balance between hardening and recovery. Its duration varies widely: it may occupy a brief initial interval or a substantial fraction of a component's service life.
Secondary creep is characterized by an approximately constant rate. In a simple physical interpretation, hardening and recovery balance each other, allowing deformation to continue without a large change in the material's resistance. The lowest measured rate is often called the minimum creep rate. This quantity is especially useful for comparing materials and correlating deformation with rupture life. An exactly steady state is not always reached, however; many modern alloys show a continuously changing rate with only a shallow minimum.
During tertiary creep, the rate accelerates until rupture. Several processes can cause this acceleration: the cross-section may decrease, internal cavities may grow and link, grain boundaries may crack, or strengthening particles may coarsen. Under constant tensile force, a reduction in cross-sectional area raises the true stress, producing a geometric feedback that increases deformation. Material damage can accelerate the rate even when true stress is controlled.
The familiar three-stage curve is not universal. Some materials show little primary creep; others have no clearly defined secondary stage. Brittle creep rupture may occur at small total strain, while highly ductile materials can elongate substantially before failure. Compression creep also lacks the tensile necking instability characteristic of many tensile tests, although internal damage and microstructural changes can still occur.
Creep rupture and creep deformation are therefore separate design concerns. A turbine blade can become unusable because its extension closes a clearance long before it fractures. Conversely, a weld region may develop damaging cavities with relatively little visible deformation. A complete characterization normally includes strain-time behavior, minimum rate, rupture time, and rupture ductility rather than relying on a single measure of “creep strength.”
Atomic and microstructural mechanisms
In crystalline solids, creep results from several competing mechanisms whose importance depends on stress, temperature, grain size, and crystal structure. The principal categories are dislocation creep, diffusion creep, and grain-boundary sliding. These mechanisms can operate simultaneously, and the slowest necessary step may control the overall deformation rate.
Dislocations are line defects that allow crystals to deform without requiring an entire plane of atoms to move at once. At elevated temperature, dislocations can glide through the lattice and climb out of their original slip planes by absorbing or emitting atomic vacancies. Climb allows them to bypass obstacles that would strongly impede low-temperature deformation. Thermally assisted recovery also permits dislocations to rearrange into lower-energy structures, including subgrain boundaries. The balance between dislocation production, obstacle bypass, and recovery underlies much engineering-metal creep.
Diffusion creep occurs through stress-directed transport of atoms and vacancies. A stress field changes the chemical potential around a grain, encouraging atoms to move in directions that gradually alter its shape. In Nabarro–Herring creep, transport occurs through the crystal lattice; in Coble creep, transport occurs mainly along grain boundaries. Their idealized rates vary approximately with the inverse square and inverse cube of grain size, respectively. Fine-grained material can therefore resist ordinary yielding well yet deform relatively rapidly by diffusion creep.
Grain-boundary sliding is relative tangential movement between neighboring grains. Sliding alone cannot generally produce compatible deformation throughout a tightly packed polycrystal: diffusion or deformation within grains must accommodate the changing geometry. If accommodation is inadequate, stress concentrations develop at boundary ledges, particles, or triple junctions. Cavities can then nucleate and grow. Boundary sliding is also important in superplastic deformation, where a fine, stable grain structure permits very large elongations under suitable conditions.
Deformation-mechanism maps organize these processes using normalized stress and homologous temperature, usually for a specified grain size. Regions indicate where particular rate laws are expected to dominate, while contours show estimated strain rates. Such maps are interpretive models, not sharp physical phase diagrams: boundaries between mechanisms are gradual, and changing grain size shifts them.
Real service materials are more complicated than ideal crystals. Solute atoms can impede dislocations, precipitates can obstruct their motion, and dispersoids can stabilize the microstructure. Meanwhile, oxidation, irradiation, phase transformations, and particle coarsening can alter the controlling mechanism during exposure. A mechanism inferred from a short test need not remain dominant over a service life of many years.
Constitutive laws and mathematical descriptions
A creep constitutive law relates deformation rate to stress, temperature, and variables describing the material's internal state. The simplest widely used metallic-creep relation is the Norton power law:
Here, is a material coefficient and is the stress exponent. Because the dimensions of depend on , fitted coefficients must be used with their stated stress and time units. A temperature-dependent form introduces an Arrhenius factor:
where is an apparent activation energy, is the molar gas constant, and is absolute temperature.
The stress exponent and activation energy can help identify a mechanism, but neither is a unique fingerprint. Ideal diffusion creep is approximately linear in stress, whereas dislocation creep often produces exponents of several units. Strengthened alloys can show much larger apparent exponents, particularly when an effective threshold stress is used to describe obstacle resistance. Changes in the microstructure during testing may also distort fitted parameters.
Primary creep is often represented by a time-dependent power law, sometimes called a Norton–Bailey form:
For , its strain rate decreases with time. This equation is convenient under constant conditions, but extending it to changing loads requires additional assumptions. Time-hardening and strain-hardening formulations can predict different responses after a stress change because they represent loading history differently. More advanced formulations use internal variables for dislocation density, hardening, recovery, or damage.
For multiaxial loading, a common model for approximately incompressible isotropic creep uses the deviatoric stress tensor and an equivalent stress based on the von Mises criterion:
This flow rule can be useful for deformation calculations, but it does not imply that rupture depends only on equivalent stress. Tensile hydrostatic stress, boundary orientation, and stress triaxiality can strongly influence cavity growth.
Linear viscoelastic creep requires a different description. Under a constant uniaxial stress , strain is written as , where is creep compliance. Under changing stress, the response depends on the full stress history through a hereditary integral. Such models are useful within a linear-response range; they do not automatically represent large-strain yielding, microstructural damage, or rupture.
Metals and high-temperature alloys
Creep resistance is a central concern in metallurgy because high-temperature machinery must retain shape and load-bearing capacity for long periods. Steam systems commonly operate for tens or hundreds of thousands of hours, whereas aircraft-engine components experience repeated high-temperature missions. These different duty cycles create different requirements for creep strength, ductility, oxidation resistance, and resistance to thermal cycling.
Ferritic and martensitic steels are widely used in boilers, steam piping, and power-generation equipment. Their resistance comes from combinations of solid-solution strengthening, fine precipitates, and stable dislocation and subgrain structures. Chromium provides oxidation resistance, while additions such as molybdenum, tungsten, vanadium, niobium, and nitrogen can contribute to high-temperature strength. During prolonged exposure, precipitates may coarsen and subgrain structures may recover, progressively reducing resistance to deformation.
Welded components require particular attention because welding creates zones with different thermal histories and microstructures. In some creep-resistant ferritic steels, long-term cracking develops in a fine-grained or intercritical region of the heat-affected zone, commonly termed Type IV cracking. The weld metal, parent metal, and heat-affected zone cannot always be represented by a single creep law. Local differences in deformation produce constrained, multiaxial stress states that affect damage accumulation.
Nickel-based superalloys are major materials for gas turbine hot sections. Many are strengthened by a high volume fraction of ordered gamma-prime precipitates, commonly based on , within a gamma matrix. These precipitates impede dislocation motion while retaining useful stability at high temperature. Directionally solidified blades reduce transverse grain boundaries, while single-crystal blades remove conventional grain boundaries altogether. This suppresses important paths for boundary sliding and intergranular failure, but it does not eliminate dislocation creep or microstructural degradation.
Blade temperature is also controlled through internal cooling and thermal-barrier coatings. In advanced engines, combustion-gas temperatures can exceed the melting temperature of the uncooled blade alloy, so the metal's actual temperature, rather than gas temperature alone, determines its creep exposure. Small changes in local cooling effectiveness can substantially alter accumulated deformation and remaining life.
Creep-resistant alloy development involves trade-offs. Increasing strength may reduce ductility or complicate fabrication, while refractory additions may increase density or encourage undesirable phases. Additive manufacturing introduces further variables, including texture, residual stress, pore populations, and heat-treatment response. Creep properties must consequently be established for the relevant processing route, not inferred solely from nominal chemical composition.
Polymers, glasses, and ceramics
In a polymer, creep arises largely from time-dependent rearrangement of molecular chains and segments. The response depends on molecular weight, chain entanglement, crystallinity, cross-link density, and temperature. Relative to most structural metals, polymers can show substantial creep at low stresses and ordinary ambient temperatures, making the duration of loading essential to their mechanical specification.
The glass transition is particularly important. Below the glass-transition region, large-scale segmental motion is restricted, although local molecular motions still produce delayed deformation. Near and above that region, segmental mobility increases sharply. An amorphous thermoplastic may pass from a relatively stiff glassy response to a much more compliant rubbery response. A semicrystalline polymer retains crystalline regions that act as physical constraints until melting, so its behavior cannot be inferred from the glass transition alone.
Mechanical analogues help distinguish different contributions. A Maxwell element, consisting of a spring and dashpot in series, gives an instantaneous elastic strain plus continuing viscous strain under constant stress. A Kelvin–Voigt element, with spring and dashpot in parallel, gives delayed strain approaching a limiting value. More elaborate combinations represent a spectrum of retardation times. Cross-linked networks can limit long-range chain flow, whereas uncross-linked thermoplastics may undergo substantial irreversible deformation over sufficiently long periods.
Time–temperature superposition combines tests at several temperatures into a master curve by shifting them along a logarithmic time axis. The approach exploits the fact that increasing temperature often accelerates molecular relaxation. It is reliable only when the relevant processes shift consistently. Crystallization, moisture uptake, chemical degradation, or a change of deformation mechanism can invalidate a single shift factor. Long-term polymer predictions therefore require attention to physical aging and environmental history as well as short-term test data.
Glasses and ceramics also creep, but their mechanisms differ from ordinary polymer-chain relaxation. At sufficiently high temperature, many glasses deform by viscous flow, with viscosity strongly dependent on temperature and composition. In crystalline ceramics, diffusion, dislocation processes, and grain-boundary sliding can contribute. Intergranular glassy phases may dominate the response of a ceramic even when the crystalline grains themselves remain highly resistant.
Ceramic matrix composites introduce load transfer between fibers, matrix, and interfaces. Creep can redistribute stress among these constituents, while oxidation or interface degradation changes their interaction. Their behavior is therefore not simply the creep rate of the fiber or matrix considered separately. Similar constituent-level distinctions apply to polymer-matrix composites, in which the matrix may creep substantially while stiff fibers carry an increasing share of the sustained load.
Concrete, wood, and long-duration structural deformation
Creep is a major part of the service behavior of concrete. Under sustained compressive stress, hardened concrete continues to shorten, with the response governed by its hydrated cement paste, aggregate restraint, moisture state, age at loading, and stress level. The microscopic origins include rearrangement and sliding within calcium silicate hydrate, moisture-related processes, and, at sufficiently high stresses, microcracking. No single elementary mechanism explains every observed regime.
Concrete creep is commonly divided into basic creep, measured without moisture exchange, and drying creep, associated with loading during drying. Drying creep is not merely the sum of basic creep and unloaded drying shrinkage: mechanical loading and moisture transport interact. Since shrinkage occurs without applied stress, experiments and structural calculations must separate it from load-induced deformation. Temperature-induced strain is another distinct contribution.
Age at loading matters because cement hydration and microstructural development continue after casting. Concrete loaded at an early age generally shows greater creep than comparable concrete loaded later under otherwise similar conditions. Structural size also matters: thick members dry more slowly than thin ones, creating internal humidity gradients and spatially variable shrinkage and creep. Aggregate stiffness and volume fraction constrain deformation of the more compliant paste.
A conventional creep coefficient is the ratio of creep strain to an associated initial elastic strain. Its numerical value depends on how the elastic reference strain and concrete age are defined. For reinforced concrete, long-term deflection also depends on cracking, reinforcement ratio, changing stiffness, and stress redistribution between concrete and steel. In prestressed concrete, concrete creep contributes to loss of prestressing force alongside shrinkage and relaxation of the prestressing steel.
Wood is strongly anisotropic, so creep differs along and across the grain. Its cellular structure and moisture-sensitive constituents produce a response that depends on both sustained stress and humidity history. Mechano-sorptive creep is the additional deformation associated with moisture changes while wood is loaded; it can exceed the deformation observed under constant moisture conditions. Repeated humidity cycles can therefore be important even when average temperature and load remain nearly unchanged.
In buildings and bridges, these effects influence floor deflections, column shortening, connection slip, and the distribution of internal forces. Prestressed girders may develop long-term camber or sag depending on the balance between prestress and gravity loading. Tall concrete buildings can experience differential shortening between columns and walls. Construction sequence matters because individual members begin carrying sustained loads at different ages and for different durations.
Creep in ice, rock, and the Earth
Geological creep is the slow deformation of natural materials under sustained stress over timescales ranging from hours to millions of years. It contributes to glacier motion, deformation around underground openings, salt movement, and flow within the Earth's interior. The very long duration of geological loading allows substantial deformation at rates far below those easily measured in ordinary engineering tests.
Glacier ice is a crystalline solid that deforms through dislocation motion, grain-boundary processes, and changes in grain structure. Its bulk behavior is often approximated by Glen's flow law:
where is an effective strain rate, is an effective deviatoric stress, and describes temperature sensitivity. A stress exponent near 3 is widely used in glacier modeling, but it is not universal. Grain size, impurities, crystal orientation, and water content influence the effective response.
Ice-sheet motion also includes basal sliding and deformation of underlying sediments, neither of which is identical to creep within the ice. Separating these contributions is important when interpreting surface velocities. In deep ice, accumulated deformation produces crystallographic fabrics that make the material anisotropic. Consequently, a flow law calibrated for randomly oriented laboratory ice may require modification for an ice sheet with a strong preferred orientation.
In the mantle, high temperature permits solid minerals to deform over geological timescales. Diffusion creep and dislocation creep are both important, with their relative contributions controlled by stress, grain size, temperature, pressure, and water content. The mantle need not be molten to flow. Its slow solid-state deformation underlies mantle convection and is linked to plate tectonics. Inferring deep-Earth flow laws nevertheless requires extrapolation from laboratory experiments conducted at much higher accessible strain rates.
Rock salt creeps readily compared with many other geological materials and can progressively close underground cavities. This behavior influences salt cavern storage and the assessment of repositories for geological disposal. Convergence of openings, redistribution of stress, and sealing of fractures can occur together, while temperature and brine affect the rates and mechanisms.
Near the surface, slow slope movement is also called soil creep, but it should not automatically be equated with steady crystalline creep. Wetting and drying, freezing and thawing, particle rearrangement, biological disturbance, and gravitational shear can all contribute. Likewise, engineering descriptions of clay creep or secondary compression distinguish time-dependent skeleton deformation from primary consolidation caused by dissipation of excess pore-water pressure.
Experimental measurement and historical development
Systematic study of creep developed alongside the need for reliable high-temperature machinery. Eugène Andrade's 1910 work helped establish quantitative descriptions of time-dependent deformation in metals. Frederick Norton's 1929 treatment of high-temperature creep contributed the stress power law associated with his name. Later microscopic theories connected macroscopic measurements to dislocations and atomic transport: Nabarro and Herring developed lattice-diffusion creep descriptions in the mid-twentieth century, and Robert Coble described grain-boundary diffusion creep in 1963.
A conventional uniaxial test uses a specimen with a controlled gauge section, a loading system, a furnace when required, and an extensometer that measures strain over time. ASTM E139 and ISO 204 are established standards covering metallic creep testing. Accurate temperature measurement and uniformity are critical because a small temperature error can produce a substantial rate error. Alignment is also important: unintended bending can cause premature local damage and misleading rupture results.
Constant-load and constant-stress tests are not identical. With a constant tensile force, true stress rises as the specimen thins. Maintaining constant true stress requires adjustment of the applied force as geometry changes. Test reports must also distinguish total strain from creep strain and record interruptions, thermal excursions, specimen orientation, and relevant heat treatment. The precision required can be demanding when useful creep strains accumulate over thousands of hours.
Stress-rupture tests emphasize time to fracture under specified conditions and may use less detailed strain measurement than full creep tests. Interrupted tests examine the developing microstructure before failure. Electron microscopy reveals dislocation structures, precipitate changes, cavities, and cracks; surface replication permits repeated examination of accessible component surfaces. Small-punch, impression, and miniature-specimen techniques can investigate limited material samples, although translating their results into conventional uniaxial properties requires suitable calibration and modeling.
The principal experimental difficulty is duration. A power plant may operate for several decades, but a material-development program cannot always wait for equivalent laboratory exposures. Accelerated testing at higher stress or temperature shortens the experiment, yet it can activate a different mechanism or produce a different microstructure. Extrapolation is strongest when supported by long-duration data and evidence that the operative mechanisms remain comparable.
Modern programs increasingly combine mechanical tests with microstructural measurements and measurement uncertainty analysis. Variability among heats, welds, processing routes, and specimen positions must be distinguished from instrument error. A fitted mean curve alone does not describe the full range of behavior needed for engineering assessment.
Life prediction, damage assessment, and engineering design
Creep design addresses allowable deformation, rupture, instability, and interaction with other damage processes. A member can satisfy a rupture-life criterion yet fail a functional limit because it has elongated, sagged, or lost preload. Conversely, local intergranular damage can threaten integrity without a large overall dimensional change. Engineering assessments therefore use several material properties and failure criteria rather than a single short-term strength value.
Long-term rupture data are often correlated using time–temperature parameters. The Larson–Miller parameter, introduced in 1952, is commonly written as
where is rupture time expressed in a specified unit, is absolute temperature, and is an empirical constant. The parameter correlates temperature and rupture time at a given stress. Its constants and fitting range are material-dependent, and the method does not guarantee reliable extrapolation across changes in mechanism or microstructural stability.
For service with changing conditions, a simple time-fraction damage estimate is
where is time spent at a particular condition and is the corresponding rupture life. Failure is often idealized as occurring near . Actual damage can depend on the sequence of conditions, recovery, prior deformation, and stress redistribution, so the linear rule is an approximation rather than a physical conservation law.
Finite-element analysis incorporates creep constitutive equations into component geometry and boundary conditions. It can predict redistribution of stress near notches, bends, welds, and changes in section. Local elastic stresses may relax through creep, but relaxation is not necessarily beneficial: strain can concentrate in a weaker region, and hydrostatic tension may promote cavitation. Material descriptions must therefore distinguish deformation from damage and account for multiaxial effects where relevant.
Creep–fatigue interaction is important when equipment starts, stops, or changes temperature. Sustained holds allow creep and oxidation, while thermal cycles impose repeated strain. The combined damage can differ from a simple addition of separately measured creep and fatigue damage. High-temperature design rules, including provisions in the ASME Boiler and Pressure Vessel Code, address allowable stresses, service duration, and relevant failure modes within their specified scope.
In-service assessment combines operating records with dimensional surveys, temperature history, nondestructive testing, and, where justified, material sampling. Cavities, crack populations, wall thinning, and microstructural degradation provide different kinds of evidence; none is a universal clock for remaining life. The interpretation depends on component geometry, material condition, loading history, and the damage mechanism actually present.