{
  "id": "annealing",
  "title": "Annealing",
  "lang": "en",
  "source_lang": "en",
  "summary": "Annealing is a controlled heat treatment that modifies a material’s structure, defects, or internal stresses to obtain selected properties.",
  "category": "technology",
  "subcategory": "materials",
  "type": "concept",
  "aliases": [
    "Annealing heat treatment",
    "thermal annealing"
  ],
  "generated_at": "2026-11-05",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/annealing/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/annealing/",
      "json": "https://aiwiki.page/zh/annealing.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "Annealing is the controlled heating, holding, and cooling of a material to alter its internal structure or physical properties. In metallurgy, it commonly restores ductility after cold working, reduces hardness, relieves residual stress, or produces a microstructure suitable for subsequent manufacturing. The term also applies to treatments of glass, semiconductor devices, polymers, and other solids, although the underlying transformations differ substantially among these materials.\n\nAnnealing is not defined by a single temperature or cooling rate. A steel component may require slow cooling through a phase-transformation range, whereas an austenitic stainless steel may be annealed at high temperature and then cooled rapidly. Glass annealing involves controlled stress relaxation without crystallization, while semiconductor annealing may last only seconds or less. The appropriate cycle therefore depends on composition, previous processing, dimensions, and the property being controlled.\n\nAcross these applications, annealing links atomic-scale processes to manufacturing performance. It can change the population of crystal defects, the size and orientation of grains, the distribution of alloying elements, or the arrangement of molecular chains. Its practical importance lies in making these changes reproducible while avoiding distortion, contamination, excessive grain growth, and other unwanted effects."
    },
    {
      "heading": "Thermodynamic driving forces and transformation kinetics",
      "text": "A processed solid is often far from equilibrium. Plastic deformation stores energy in crystal defects; casting can leave chemical segregation; rapid cooling can retain metastable phases; and uneven heating can produce internal stress. Annealing supplies thermal energy that allows the material to move toward a lower-free-energy state. The final condition need not be complete equilibrium: industrial treatments are usually stopped when the required properties have been obtained.\n\nIn crystalline materials, diffusion is central to many annealing processes. Atoms move through vacancies, along grain boundaries, or through other available pathways. The temperature dependence of a diffusion coefficient is often represented by an Arrhenius expression:\n\n\\[D=D_0\\exp\\left(-\\frac{Q}{RT}\\right),\\]\n\nwhere \\(D\\) is diffusivity, \\(D_0\\) is a prefactor, \\(Q\\) is an activation energy per mole, \\(R\\) is the gas constant, and \\(T\\) is absolute temperature. Because temperature occurs in the exponential, a modest increase can greatly shorten the time needed for a diffusion-controlled change. A characteristic diffusion distance scales approximately as \\(\\sqrt{Dt}\\), although actual transformation rates also depend on geometry, interfaces, and composition.\n\nDifferent structural changes have different driving forces. The reduction of stored deformation energy drives recrystallization; the reduction of grain-boundary area drives grain growth; and chemical-potential differences drive homogenization and phase redistribution. Residual stresses can relax when thermally activated deformation redistributes strain. In glass, structural rearrangements and viscous flow allow stress to decay without requiring a crystalline transformation.\n\nFor metals, temperature is sometimes expressed as a fraction of the absolute melting temperature, \\(T/T_m\\), called homologous temperature. Recrystallization of many cold-worked metals occurs within a broad range of roughly \\(0.3\\)–\\(0.5T_m\\), but this is a heuristic rather than a specification. Very pure metals, heavily deformed materials, and alloys containing particles that obstruct boundary motion can behave quite differently. All temperatures in such ratios must be expressed in kelvin.\n\nTime and temperature are therefore not freely interchangeable. A longer low-temperature treatment may relieve stress without producing the same grain structure as a shorter high-temperature treatment. Raising the temperature may also activate precipitation, oxidation, melting of a low-melting constituent, or an unwanted phase transformation. Annealing schedules must account for competing reactions, not merely the speed of the desired one. The relevant framework combines thermodynamics, which identifies driving forces and possible states, with kinetics, which determines whether those states can be approached during the available processing time."
    },
    {
      "heading": "Recovery, recrystallization, and grain growth",
      "text": "The classical description of annealing a cold-worked metal distinguishes recovery, recrystallization, and grain growth. These processes can overlap, and their boundaries are not always sharply separated, but the distinction explains many changes in strength, ductility, conductivity, and texture.\n\nCold rolling, drawing, and other forms of plastic deformation increase the density of dislocations, line defects that accommodate slip within a crystal. Their interactions obstruct further motion, producing strain hardening. Deformation also elongates grains, develops preferred crystallographic orientations, and generates heterogeneous internal stresses. In heavily worked metals, dislocation densities can reach approximately \\(10^{14}\\)–\\(10^{16}\\ \\mathrm{m^{-2}}\\), although values depend strongly on material and deformation history.\n\nDuring recovery, point defects and dislocations rearrange or annihilate without wholesale replacement of the deformed grains. Dislocations may organize into lower-energy arrays and subgrain boundaries, a process often called polygonization. Electrical resistivity and residual stress can decrease substantially while hardness changes comparatively little. Recovery is particularly significant in metals in which dislocation cross-slip and climb occur readily. It can therefore restore some functional properties without removing all the strengthening produced by cold work.\n\nRecrystallization replaces the deformed structure with new grains having much lower dislocation density. These grains develop from suitable regions in the worked microstructure and grow by boundary migration. The resulting reduction in strength and hardness is commonly accompanied by a large increase in ductility. Greater prior deformation generally increases the stored-energy driving force and the number of potential recrystallization sites, often lowering the temperature or time needed for substantial transformation. Particle distributions, solute atoms, initial grain size, and strain heterogeneity can all modify this behavior.\n\nAfter recrystallization, continued heating may cause grain growth. Boundary migration reduces total interfacial energy, but excessive growth can impair properties. In many metals, finer grains increase yield strength according to the approximate Hall–Petch relation:\n\n\\[\\sigma_y=\\sigma_0+k_y d^{-1/2},\\]\n\nwhere \\(\\sigma_y\\) is yield strength, \\(d\\) is average grain size, and \\(\\sigma_0\\) and \\(k_y\\) are material-dependent constants. This relation has limits, especially at very small grain sizes and under conditions dominated by high-temperature deformation.\n\nAnnealing also changes crystallographic texture rather than merely grain size. Some orientations grow preferentially, altering directional formability or magnetic behavior. Abnormal grain growth occurs when a small population of grains becomes much larger than its neighbors. It is undesirable in many structural products but deliberately exploited in grain-oriented electrical steel. Thus, “fully annealed” does not mean structurally featureless: the resulting grain size, texture, and phase distribution remain essential characteristics."
    },
    {
      "heading": "Annealing of steels",
      "text": "Annealing of steel is strongly influenced by transformations among ferrite, austenite, cementite, and their microstructural combinations. The appropriate schedule depends on carbon content, alloying additions, and the starting structure. In the metastable iron–cementite diagram, the eutectoid composition is approximately 0.76–0.77 mass percent carbon and the eutectoid temperature is about \\(727\\,^\\circ\\mathrm{C}\\). Actual transformation temperatures during heating and cooling differ from equilibrium values and vary with composition and processing rate.\n\nIn full annealing of a hypoeutectoid carbon steel, the material is heated above the upper critical temperature, conventionally designated \\(A_{\\mathrm{c3}}\\) on heating, so that its structure becomes austenitic. Controlled slow cooling then produces ferrite and pearlite. Compared with faster cooling, slow passage through the transformation range generally produces coarser pearlite and lower hardness. This condition can improve machinability or prepare a component for later forming and heat treatment. Holding time must allow adequate temperature uniformity and the intended structural change without unnecessary grain coarsening.\n\nHypereutectoid steels require different treatment because excessive heating and subsequent cooling can promote unfavorable distributions of proeutectoid cementite. Partial or intercritical annealing schedules may therefore be used instead of complete austenitization. For high-carbon tool steels and bearing steels, spheroidizing annealing is especially important. Prolonged holding near, usually just below, the eutectoid range, or controlled cycling around it, encourages cementite to form rounded particles in a ferritic matrix. This structure is generally easier to machine and cold form than a lamellar pearlitic structure.\n\nProcess annealing commonly restores ductility in low-carbon sheet or wire between forming operations. It is usually performed below the lower critical transformation temperature and relies mainly on recovery and recrystallization rather than austenite formation. Stress-relief annealing similarly seeks to reduce residual stress while limiting major changes in the existing structure. Isothermal annealing instead includes cooling from an austenitizing temperature to a selected transformation temperature and holding there, allowing transformation under more nearly uniform conditions.\n\nAnnealing must be distinguished from normalizing and tempering. Normalizing generally involves austenitizing followed by air cooling and commonly produces a finer structure than full furnace annealing. Tempering treats a hardened, often martensitic, steel below its critical transformation range to adjust hardness and toughness; it is not simply another name for annealing.\n\nStainless steels illustrate why slow cooling is not a universal requirement. Austenitic stainless steels are commonly solution annealed at roughly \\(1{,}000\\)–\\(1{,}150\\,^\\circ\\mathrm{C}\\), depending on grade, to dissolve carbides and establish the desired structure. Rapid subsequent cooling limits carbide reprecipitation. Prolonged exposure in sensitizing temperature ranges can instead produce chromium-depleted regions near grain boundaries and reduce resistance to intergranular corrosion."
    },
    {
      "heading": "Nonferrous alloys and functional metallic materials",
      "text": "In nonferrous metals, annealing frequently controls the balance between deformation strengthening, formability, conductivity, and dimensional stability. The same treatment name can conceal very different mechanisms. A copper wire may be softened mainly through recrystallization, an aluminum alloy may undergo changes in both deformation structure and precipitates, and a titanium alloy may require control of its constituent phases.\n\nCopper and brass are commonly annealed between drawing, rolling, or deep-drawing operations. Repeated deformation raises strength but eventually limits further forming. Recrystallization restores ductility, allowing additional reductions in section or more severe shaping. In brass cartridge-case manufacture, different regions may require different mechanical properties: localized annealing can soften the neck while preserving greater strength elsewhere. Grain size must be controlled because excessively coarse grains can produce a visibly roughened surface, known as orange peel, during subsequent forming.\n\nIn commercially pure aluminum and many non-heat-treatable aluminum alloys, annealing principally removes strain hardening. In precipitation-hardenable alloys, however, heating may also dissolve, coarsen, or redistribute strengthening particles. The fully annealed condition is commonly identified by the temper designation O in wrought aluminum products. Solution heat treatment followed by quenching and aging is a distinct processing route, even though solution treatment is sometimes described more broadly as solution annealing. Its purpose is to create a supersaturated solid solution from which strengthening precipitates can subsequently form.\n\nTitanium alloys require attention to the relative stability and distribution of the hexagonal alpha and body-centered-cubic beta phases. Treatments below, near, or above the beta-transus temperature can produce substantially different grain morphologies and mechanical properties. The transus is composition-dependent, so a schedule suitable for one alloy cannot be transferred directly to another. Titanium’s strong chemical affinity for oxygen and nitrogen at elevated temperatures also makes atmosphere control important; surface contamination can create a hard, brittle oxygen-enriched layer.\n\nAnnealing can be designed to improve functional rather than primarily mechanical properties. Electrical steels used in an electrical transformer or electric motor are processed to reduce magnetic losses and establish suitable grain structures. Grain-oriented transformer steel depends on a strong preferred orientation developed through carefully controlled rolling and annealing. Final heat treatments may also reduce stresses introduced by cutting or fabrication.\n\nFor conductive materials, annealing can reduce the electron scattering associated with deformation defects, although impurities and alloying elements often dominate electrical resistivity. For precision components, stress relief can reduce dimensional changes during later machining. These examples show that softness is only one possible objective: an annealed metallic product may instead be specified by conductivity, magnetic permeability, texture, fatigue behavior, or dimensional stability."
    },
    {
      "heading": "Glass annealing and structural relaxation",
      "text": "Annealing of glass primarily controls residual stress created during cooling. Because glass does not ordinarily solidify through a single sharp crystallization event, its mechanical response changes continuously over a broad temperature interval. At sufficiently high temperature it can relax stress by viscous flow; at lower temperature, relaxation becomes so slow that stresses are effectively frozen into the article.\n\nTemperature gradients are the main source of these stresses. During cooling, the surface and interior contract at different times and rates. If the material becomes too rigid before the temperature field becomes sufficiently uniform, the resulting differential contraction remains as residual stress. Thick sections, abrupt changes in wall thickness, and joints between glasses with different thermal expansion characteristics are particularly challenging.\n\nIndustrial terminology uses reference viscosities to define useful temperatures. The annealing point is conventionally associated with a viscosity of approximately \\(10^{13}\\) poise, or \\(10^{12}\\ \\mathrm{Pa\\,s}\\), at which substantial stress relaxation occurs on a timescale of minutes. The strain point corresponds approximately to \\(10^{14.5}\\) poise, or \\(10^{13.5}\\ \\mathrm{Pa\\,s}\\), where stress relaxation is much slower. These are material-specific temperatures, not universal values. Typical soda–lime container glass has an annealing point in the vicinity of \\(550\\,^\\circ\\mathrm{C}\\), whereas other glass compositions can differ markedly.\n\nAn annealing cycle brings the article into an appropriate temperature range, allows temperature and stress to become sufficiently uniform, and then cools it at a controlled rate through the critical relaxation interval. Faster cooling becomes possible once the risk of generating unacceptable stress has diminished. Continuous glass production commonly uses a lehr, a temperature-controlled tunnel through which containers or other products travel. Large optical blanks and thick castings may require much longer schedules because thermal equalization time increases strongly with section thickness.\n\nGlass annealing differs from thermal tempering. Tempering intentionally creates a compressive surface layer by cooling the surface rapidly while the interior remains hotter. The resulting stress distribution increases resistance to certain applied loads. Annealing instead seeks to reduce unwanted residual stress. Neither process should be confused with devitrification, the formation of crystals within glass.\n\nResidual stress can be examined through birefringence using polarized light. Stress-induced differences in refractive index produce optical patterns that reveal nonuniformity. In high-quality optical components, even stresses too small to cause immediate fracture can impair image formation or polarization performance. The glass transition provides the wider framework for understanding these effects: the frozen structure and properties depend not only on composition but also on the material’s thermal history."
    },
    {
      "heading": "Semiconductor and thin-film annealing",
      "text": "In semiconductor manufacturing, annealing modifies defects, dopant configurations, interfaces, and thin-film phases. The objective is often highly localized at the atomic scale, while the allowable changes elsewhere in the device are extremely small. Consequently, semiconductor annealing places unusual emphasis on heating rate, peak temperature, exposure time, atmosphere, and the cumulative effects of all thermal processing steps.\n\nAfter ion implantation, dopant atoms and implantation-induced defects occupy nonequilibrium configurations. An anneal can repair lattice damage and move dopant atoms into electrically active substitutional sites. Where implantation has amorphized part of a silicon crystal, solid-phase epitaxial regrowth can restore crystalline order from the underlying undamaged lattice. Activation and repair must be balanced against dopant diffusion, which can broaden junctions and change device dimensions.\n\nRapid thermal annealing heats a wafer quickly, commonly using intense lamps, and limits the time near the maximum temperature. Depending on the process, silicon activation anneals may involve peak temperatures around \\(900\\)–\\(1{,}100\\,^\\circ\\mathrm{C}\\) with high-temperature exposure lasting seconds or less. Spike annealing minimizes the dwell near the peak, while flash-lamp and laser treatments can deliver much shorter pulses. Some laser processes heat or melt only a shallow surface region, allowing rapid recrystallization while limiting heating of the underlying structure.\n\nThe relevant constraint is often called the thermal budget: earlier layers, interfaces, and dopant profiles must survive every later heating step. A treatment that repairs one defect population may cause another problem through interdiffusion, stress, agglomeration, or undesirable chemical reactions. Implantation damage can also cause transient enhanced diffusion, so dopant movement cannot always be predicted from equilibrium diffusivity alone.\n\nAnnealing has additional roles in contact and interface formation. Metal–silicon reactions can create silicides used to reduce electrical resistance. Lower-temperature treatments in hydrogen-containing atmospheres can passivate electrically active defects at silicon–silicon dioxide interfaces. Here the atmosphere is a chemical participant, not merely protection against oxidation.\n\nThin-film electronics and solar cell production use annealing to control crystallinity, grain size, defects, and phase formation. In some compound semiconductors, volatile elements require special atmosphere or encapsulation strategies to prevent composition changes during heating. Deposited films may crystallize, densify, or develop preferred orientation, but excessive treatment can cause cracking, interfacial reactions, or loss of film continuity. Thus, a successful semiconductor anneal is judged not by maximum structural perfection alone, but by the combined electrical, dimensional, and reliability requirements of the completed device."
    },
    {
      "heading": "Polymers, ceramics, and additively manufactured materials",
      "text": "Annealing of a polymer depends on whether it is amorphous, semicrystalline, or cross-linked. In an amorphous thermoplastic, heating near or above the glass-transition range can permit molecular rearrangement and relaxation of molding-induced stresses. In a semicrystalline polymer, treatment below melting can also increase crystallinity, reorganize lamellae, or improve crystalline order. These processes can change stiffness, dimensional stability, transparency, and resistance to subsequent heating.\n\nThe response is not uniformly beneficial. Increasing crystallinity often produces shrinkage and can reduce impact resistance or optical clarity. An injection-molded component may distort when frozen-in molecular orientation relaxes. A useful annealing schedule therefore depends on the polymer grade, part geometry, moisture content, and prior cooling history. Polyethylene terephthalate and poly(lactic acid), for example, can undergo substantial crystallization under suitable heating conditions, but their rates and final properties depend strongly on formulation and processing.\n\nPolymer annealing should be distinguished from curing. Curing of a thermosetting polymer creates or extends a cross-linked molecular network through chemical reactions. A post-cure treatment may simultaneously advance those reactions and relieve stress, but the two mechanisms remain distinct. Prolonged heating can also cause oxidation, chain scission, or other degradation, particularly in the presence of oxygen or moisture. Physical aging below the glass transition is another separate, though related, form of structural relaxation.\n\nIn ceramics, annealing may modify oxygen content, point-defect populations, electrical properties, or residual stresses without the large density increase characteristic of sintering. Oxide ceramics can respond strongly to oxygen partial pressure because changes in oxygen-vacancy concentration alter ionic and electronic transport. Piezoelectric and ferroelectric ceramics may undergo changes in domain state, and subsequent polarization treatment may be necessary to establish their intended operating properties. The relevant reactions vary widely among ceramic families.\n\nParts made by additive manufacturing present additional reasons for post-build heat treatment. Metal powder-bed processes can generate steep thermal gradients, residual stress, cellular solidification structures, and metastable phases. A stress-relief anneal may reduce distortion before removal from the build plate, while a higher-temperature treatment may homogenize composition or transform the microstructure. The appropriate sequence depends on alloy, manufacturing route, and intended service.\n\nFor polymer prints, annealing can improve heat resistance through crystallization or relaxation, but dimensional changes may be substantial and direction-dependent. In both metals and polymers, annealing does not automatically eliminate pores or lack-of-fusion defects. Pressure-assisted densification is a separate process. The distinction matters because a part can become less stressed or more crystalline while retaining defects that still control fatigue life or fracture."
    },
    {
      "heading": "Industrial equipment, atmosphere control, and verification",
      "text": "Industrial annealing equipment ranges from small laboratory furnaces to continuous lines processing strip, wire, glass, or semiconductor wafers. Batch furnaces treat a stationary load through a complete cycle; continuous furnaces move products through successive heating, soaking, and cooling zones. Bell furnaces are widely used for metal coils, while roller-hearth, mesh-belt, and other conveyor systems serve products that can pass continuously through a controlled environment.\n\nThe furnace temperature is not necessarily the workpiece temperature. Thick sections, tightly packed loads, and coiled strip can develop substantial temperature differences. Heat reaches the load through radiation, convection, and sometimes contact conduction, while the load’s internal temperature field evolves through conduction. Holding time must therefore distinguish heating through the section from time actually spent at the intended treatment temperature. Excessive heating rates can produce distortion even before the desired annealing reaction begins.\n\nAtmosphere control limits chemical changes at the surface. Air can cause oxidation and, in steels, decarburization. Protective or reducing gases and vacuum are selected according to alloy and treatment objective. Hydrogen-containing atmospheres can support bright annealing, but their suitability depends on material chemistry and equipment design. Oxygen-bearing copper, for example, can suffer internal damage when hydrogen reacts with copper oxide to form steam. Gas purity, dew point, oxygen content, leakage, and contamination can therefore be as important as nominal furnace temperature.\n\nModern installations use sensors, recorded temperature histories, and process control to maintain reproducibility. Thermocouples measure furnace or load temperatures, while optical pyrometers support noncontact measurement where suitable. Pyrometry requires attention to emissivity and intervening windows or atmospheres. Furnace temperature uniformity and instrument accuracy must be established over the working volume rather than assumed from a single control reading.\n\nVerification depends on the intended outcome. Hardness measurements can reveal softening but may not distinguish all relevant structures. Metallography measures grain size and phase distribution; an electron microscope can reveal finer features, and diffraction-based orientation mapping can quantify texture and recrystallization. Residual stresses may be assessed by diffraction or mechanical methods. Glass commonly requires optical stress examination, while semiconductor processes use electrical and structural measurements suited to the device.\n\nAn apparently acceptable average property can conceal local failures. Excessive grain growth, surface decarburization, incomplete recrystallization, sensitization, or nonuniform dopant activation may occur in restricted regions. Nondestructive testing can reveal some defects, but structural confirmation may require representative destructive sampling. Qualification consequently connects the recorded thermal cycle with measured material properties, rather than treating the furnace program alone as proof of successful annealing."
    },
    {
      "heading": "Historical development and the computational analogy",
      "text": "Practical annealing predates scientific explanations of crystal structure. Metalworkers discovered that copper and other worked metals could recover formability after heating, enabling repeated cycles of hammering and shaping. Glassmakers likewise learned that cooling conditions affected cracking and durability. These practices developed empirically: the useful relationship between heating and subsequent behavior was recognized long before defects, diffusion, or glass relaxation could be measured.\n\nNineteenth-century metallography made heat treatment a subject of systematic structural investigation. Henry Clifton Sorby’s work in the 1860s helped establish microscopic examination of metals, revealing that apparently uniform metallic objects contained distinct structural constituents. Subsequent development of phase diagrams, thermal analysis, and controlled furnaces allowed annealing conditions to be related to transformations rather than judged only by appearance, furnace color, or craft experience.\n\nDuring the twentieth century, physical metallurgy connected annealing with dislocations, vacancies, boundary migration, and diffusion. Gustav Tammann’s investigations of metals contributed to early scientific understanding of recrystallization. William A. Johnson and Robert F. Mehl’s 1939 transformation model and Melvin Avrami’s papers of 1939–1941 provided influential mathematical descriptions of nucleation and growth. A commonly used idealized expression is\n\n\\[X(t)=1-\\exp(-Kt^n),\\]\n\nwhere \\(X\\) is transformed fraction, \\(K\\) is a temperature-dependent rate parameter, and \\(n\\) reflects assumptions about nucleation and growth. Real recrystallizing materials may depart from these assumptions because deformation is heterogeneous, nucleation sites are limited, or growth rates change during transformation.\n\nAnnealing also supplied a metaphor for simulated annealing, a method in mathematical optimization. The method explores candidate solutions using a temperature-like control parameter. A move that worsens the objective by \\(\\Delta E>0\\) may be accepted with probability\n\n\\[p=\\exp\\left(-\\frac{\\Delta E}{T_a}\\right),\\]\n\nwhere \\(T_a\\) is an algorithmic parameter with units scaled to the objective. At high values, unfavorable moves are accepted relatively often, helping the search escape local minima; lowering the parameter progressively favors better solutions.\n\nThe acceptance rule draws on the Monte Carlo method published by Nicholas Metropolis and collaborators in 1953. Scott Kirkpatrick, C. Daniel Gelatt Jr., and Mario P. Vecchi presented the influential optimization formulation in 1983. Computational temperature need not correspond to any physical temperature, and a finite cooling schedule does not automatically guarantee a globally optimal result. The analogy rests on exploration followed by increasing selectivity, not on a literal reproduction of grain growth or metallurgical transformations. Related computational uses of “annealing” therefore require their own definitions, separate from the heat-treatment processes that inspired the name."
    }
  ],
  "links": [
    "metallurgy",
    "diffusion",
    "temperature",
    "thermodynamics",
    "recrystallization",
    "martensite",
    "brass",
    "electrical-transformer",
    "electric-motor",
    "birefringence",
    "glass-transition",
    "semiconductor",
    "ion-implantation",
    "laser",
    "solar-cell",
    "polymer",
    "thermosetting-polymer",
    "3d-printing",
    "sensor",
    "process-control",
    "electron-microscope",
    "simulated-annealing",
    "mathematical-optimization"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}