aiwiki.page
English
Technology / Materials / epitaxy

Epitaxy

Epitaxy is the oriented growth of a crystalline material on a crystalline substrate, enabling precise control of semiconductor layers, interfaces, and functional materials.

26 keywords9 linked from10 not yet writtenWritten by AI
CrystalSemiconductorCrystal Structur…ThermodynamicsMass TransferLithographyLaserIon ImplantationEpitaxy

Epitaxy is the growth of a crystalline material on another crystal in a definite orientational relationship. The underlying crystal, called the substrate, provides an ordered surface that influences the orientation of the growing material. An epitaxial layer need not have the same composition, lattice spacing, or crystal symmetry as its substrate, and it need not be free of defects. Its defining feature is crystallographic alignment rather than chemical identity or perfection.

Epitaxy is central to semiconductor manufacturing because it permits the construction of layers whose composition, thickness, strain, and electrical properties differ deliberately from those of the supporting wafer. Such layers form the active regions of transistors, light-emitting diodes, semiconductor lasers, and high-efficiency photovoltaic devices. Beyond electronics, epitaxial growth is used to investigate magnetism, superconductivity, oxide interfaces, and materials only a few atomic layers thick.

The term describes an ordered growth relationship, not a single deposition method. Epitaxy can proceed from vapor, liquid, or solid phases. Its outcome depends on the interaction of crystallography, surface chemistry, atomic transport, and mechanical strain, making the interface between two materials as important as either material considered separately.

Crystallographic relationships and historical development

An epitaxial relationship specifies how crystallographic planes and directions in a film align with those in a substrate. For example, a cube-on-cube relationship may be written as (001)film∥(001)substrate(001)_{\mathrm{film}}\parallel(001)_{\mathrm{substrate}} together with [100]film∥[100]substrate[100]_{\mathrm{film}}\parallel[100]_{\mathrm{substrate}}. The plane relationship alone is insufficient: a film can have its surface normal aligned while remaining randomly rotated in the plane. Such a film is textured, but it does not necessarily possess the full orientational order normally implied by epitaxy. These distinctions follow from the symmetry and atomic arrangement of the materials’ crystal structures.

In homoepitaxy, the film and substrate are the same material, as in silicon grown on silicon. The layer may nevertheless differ substantially in dopant concentration, isotopic composition, or defect content. In heteroepitaxy, the materials differ: examples include silicon–germanium on silicon and aluminum gallium arsenide on gallium arsenide. Heteroepitaxy expands the range of attainable properties, but introduces possible mismatches in lattice spacing, thermal expansion, bonding, and polarity. A single substrate can also support several symmetry-equivalent orientations, producing domains rather than a single uniformly oriented crystal.

Epitaxial relationships were observed in minerals before they became tools of microelectronics. The French mineralogist Louis Royer introduced the term épitaxie in 1928 in work on oriented crystal overgrowth. Early explanations emphasized similarities between the atomic spacings exposed at crystal surfaces. Modern descriptions retain the importance of structural matching but also account for surface termination, chemical bonding, reconstruction, and growth kinetics. Simple numerical agreement between bulk lattice constants is therefore neither a universal requirement nor a sufficient predictor of successful growth.

Semiconductor epitaxy developed rapidly during the mid-twentieth century. Vapor-grown silicon layers became important in transistor production around 1960 because they allowed a relatively lightly doped active region to be formed on a heavily doped supporting wafer. Herbert Nelson demonstrated liquid-phase epitaxy for semiconductor junction structures in 1963. Work by John Arthur, Alfred Y. Cho, and others during the late 1960s established molecular-beam epitaxy as a method of growing controlled crystalline layers under ultrahigh vacuum. These developments made it practical to treat a crystal as an engineered stack rather than a chemically uniform block.

Epitaxy should also be distinguished from several neighboring concepts. A single-crystalline film may be produced without a conventional crystalline substrate, while an epitaxial film may contain twins, dislocations, or multiple domains. Conversely, an amorphous or polycrystalline coating deposited by the same equipment is not epitaxial merely because its composition and thickness are well controlled.

Surface energetics, nucleation, and growth modes

Epitaxial growth requires atoms or molecules to arrive at a surface, move or react there, and become incorporated into crystallographic sites. The balance between these processes determines whether the surface remains smooth, develops islands, or becomes rough. Thermodynamics identifies favorable configurations, but the structure actually produced also depends on how quickly the system can approach them. A low-energy arrangement may not form if arriving species become trapped before they can reach appropriate sites.

A useful starting point compares the substrate surface energy γs\gamma_s, film surface energy γf\gamma_f, and film–substrate interface energy γi\gamma_i. In a simplified, unstrained equilibrium picture, complete wetting is favored when

γs≥γf+γi.\gamma_s \geq \gamma_f+\gamma_i.

This condition expresses whether replacing the exposed substrate with an interface and a film surface lowers the energy. Actual growth also involves strain, chemical reactions, surface reconstructions, and nonequilibrium conditions. Consequently, the inequality is a guide to tendencies rather than a universal recipe for obtaining a flat film.

Three classical growth modes describe common morphologies. In Frank–van der Merwe growth, successive layers develop in an approximately layer-by-layer sequence. In Volmer–Weber growth, three-dimensional islands form directly because film atoms favor bonding to one another over complete wetting of the substrate. In Stranski–Krastanov growth, an initial wetting layer is followed by island formation. The latter often occurs in strained heteroepitaxy, where growing islands can relieve part of the elastic energy accumulated in a flat film. Indium arsenide deposited on gallium arsenide is a widely studied example used to form nanoscale islands.

Surface diffusion strongly affects which morphology appears. A rough estimate of the diffusion length is

LD∼Dsτ,L_D \sim \sqrt{D_s\tau},

where DsD_s is the surface diffusion coefficient and τ\tau is the time available before incorporation, desorption, or another terminating event. Increasing substrate temperature commonly increases diffusion, allowing deposited species to reach steps or existing islands. Excessive temperature, however, can cause desorption, decomposition, intermixing, or loss of volatile constituents. Growth rate and temperature must therefore be considered together.

A slightly misoriented, or vicinal, wafer presents terraces separated by atomic steps. If arriving species travel far enough to reach those steps, step-flow growth can occur without repeated nucleation of islands on each terrace. Barriers to crossing step edges can instead promote mound formation, while unequal step velocities can cause step bunching. Substrate cleaning is equally decisive: an oxide or carbon-rich contamination layer may interrupt crystallographic templating even when temperature and flux are otherwise suitable. Epitaxy is therefore controlled as much by the state of the first atomic layers as by the supply of material afterward.

Lattice mismatch, elastic strain, and crystal defects

In heteroepitaxy, different equilibrium lattice spacings create mechanical constraints at the interface. For a simple cubic comparison, the lattice mismatch is commonly defined as

f=af−asas,f=\frac{a_f-a_s}{a_s},

where afa_f and asa_s are the relaxed lattice parameters of the film and substrate. Other sign conventions are also used. A positive value under this convention means that the free-standing film would have a larger lattice parameter than the substrate. If the film adopts the substrate’s in-plane spacing, it is compressed in the plane and usually expands perpendicular to it through its elastic response.

A sufficiently thin film can remain coherent, maintaining lattice continuity across the interface without mismatch-relieving dislocations. Such a film is often described as pseudomorphic. Its elastic energy per unit area increases approximately in proportion to thickness when strain remains uniform. Beyond a certain thickness, introducing defects or changing morphology may lower the total energy. The resulting critical thickness is not an absolute material constant: it depends on mismatch, elastic properties, dislocation geometry, temperature, growth history, and the availability of dislocation sources.

The Matthews–Blakeslee force-balance model and energy-based approaches associated with People and Bean are widely used to estimate critical thickness under different assumptions. They do not describe identical physical limits. A film can remain metastably strained beyond an equilibrium estimate because dislocation nucleation or motion is kinetically inhibited. Conversely, defects inherited from the substrate may initiate relaxation earlier than an idealized model predicts. Reported critical thicknesses must therefore be interpreted with their experimental conditions.

Dislocations are line defects characterized by a Burgers vector describing a lattice displacement. Misfit dislocations running along an interface can accommodate the difference between film and substrate spacings. Threading segments extend through the growing film and may intersect the active device region. They can provide nonradiative recombination paths in optoelectronic materials, increase leakage, and alter mechanical behavior. Other important defects include stacking faults, twins, point defects, and antiphase boundaries, the last being especially relevant when a compound semiconductor with distinct sublattices grows on a nonpolar substrate.

Mismatch is not solely a growth-temperature problem. Different thermal expansion coefficients generate additional stress during cooling, potentially producing cracks or wafer curvature even when the film was comparatively well matched during deposition. Buffer layers, compositionally graded layers, patterned substrates, and lateral overgrowth can reduce the propagation of damaging defects into active regions. In silicon–germanium technology, relaxed graded buffers can provide a lattice spacing different from that of silicon, enabling deliberately strained layers above them. These approaches manage defects and strain rather than eliminating their underlying causes.

Vapor-phase epitaxy and chemical reactor processes

Many industrial epitaxial layers are grown by chemical vapor deposition (CVD), in which gaseous precursors react or decompose at a heated substrate. The process is epitaxial only when the deposited material adopts the required crystalline orientation. The same precursor chemistry can instead produce polycrystalline or amorphous material if the surface, temperature, or substrate is unsuitable. Reactor design controls precursor delivery, gas residence time, temperature uniformity, and removal of reaction products across the wafer.

Silicon epitaxy commonly uses silane or chlorinated silicon precursors such as dichlorosilane and trichlorosilane, often with hydrogen as a carrier gas. Depending on precursor, pressure, and application, growth temperatures range from several hundred degrees Celsius to above 1,000 ∘C1{,}000\,^\circ\mathrm{C}. Chlorine-containing chemistry can influence both deposition and etching, helping suppress unwanted nucleation in some processes. Dopant gases such as phosphine or diborane allow the electrical character of the growing layer to be adjusted independently of the starting wafer.

Two limiting regimes help explain reactor behavior. At relatively low temperatures, surface reaction rates may control deposition, producing strong temperature dependence. At higher temperatures, delivery of reactive species through the gas boundary layer may become limiting. This mass-transfer regime makes flow geometry and precursor depletion especially important. Real reactors can operate between these limits, and different reactions within the same process may respond differently to temperature and pressure. Wafer rotation and carefully designed gas distribution improve uniformity but do not replace control of surface chemistry.

Metalorganic vapor-phase epitaxy, also called metalorganic chemical vapor deposition, is especially important for compound semiconductors. Gallium arsenide can be grown using a gallium metalorganic precursor and arsine; gallium nitride commonly uses a gallium metalorganic precursor and ammonia. The ratio of group-V to group-III precursor supply strongly affects surface conditions and impurity incorporation, although inlet flow ratios do not directly equal the ratios of reactive species at the surface. These processes support production of light-emitting diodes, lasers, and high-frequency transistors.

Hydride vapor-phase epitaxy provides another route, particularly for thick compound-semiconductor layers. In one gallium nitride process, hydrogen chloride reacts with gallium to form volatile gallium chloride, which then reacts with ammonia in the growth region. High growth rates make the method useful for thick templates and substrates. Selective-area epitaxy, by contrast, uses patterned masks—often defined by lithography—to restrict growth to exposed crystalline regions. It enables localized device structures, but feature spacing and exposed area can alter precursor consumption and growth rate, a phenomenon known as loading.

Molecular beams and other physical deposition methods

Molecular-beam epitaxy directs beams of atoms or molecules toward a heated substrate in an ultrahigh-vacuum environment. Solid source materials are often evaporated from temperature-controlled effusion cells, while shutters start and stop their delivery. Some systems use gaseous sources or plasma-generated species. The low background pressure reduces contamination and allows particles to travel largely without collisions between source and substrate. Growth rates are usually lower than those of high-throughput vapor-phase processes, which favors precise layer sequencing over rapid production of thick material.

The method is particularly effective for making abrupt compositional changes. Opening and closing shutters allows one compound or alloy composition to follow another without replacing the entire reactor atmosphere. Nevertheless, an abrupt shutter action does not guarantee an atomically abrupt interface. Surface segregation, exchange reactions, residual flux, and interdiffusion can distribute atoms across several layers. Control therefore requires knowledge of both beam delivery and the residence and incorporation behavior of each element.

In conventional gallium arsenide growth, gallium and arsenic are supplied separately, with an arsenic overpressure helping maintain suitable surface conditions. The growth rate can be governed mainly by the gallium arrival rate within an appropriate temperature and flux window. Different compounds require different strategies: active nitrogen from a plasma source can support nitride growth, while oxide epitaxy may employ oxygen, ozone, or activated oxygen. The useful window is bounded by insufficient surface mobility at one end and desorption, decomposition, or unwanted reactions at the other.

An important feature is compatibility with reflection high-energy electron diffraction, or RHEED. An electron beam strikes the surface at a grazing angle, producing a diffraction pattern sensitive to near-surface order. Changes between streaked and spotty patterns can indicate changes in morphology. Under suitable layer-by-layer growth conditions, oscillations in diffraction intensity can be used to calibrate the deposition time for a known atomic-layer repeat. Such oscillations are condition-dependent and may disappear as growth shifts toward step flow or becomes rough.

Other physical deposition techniques can also produce epitaxy. In pulsed laser deposition, a high-energy laser pulse removes material from a target, generating a plume that reaches the substrate. This method is widely used for complex oxides because multicomponent targets can provide a convenient source of several elements, although film stoichiometry need not exactly reproduce target stoichiometry. Sputtering can likewise produce epitaxial metals, nitrides, and oxides when substrate preparation, energetic bombardment, and temperature are appropriately controlled. None of these techniques is intrinsically epitaxial; crystalline alignment remains an outcome to be established experimentally.

Liquid-phase growth and solid-phase recrystallization

In liquid-phase epitaxy, a crystalline substrate contacts a solution containing the material to be grown. Supersaturation causes solute to precipitate onto the substrate in an aligned crystalline form. Supersaturation can be generated by cooling, by establishing a temperature gradient, or by changing solution composition. Because atoms move readily through a liquid and growth can occur near equilibrium, the method can produce high-quality layers without the same highly reactive vapor precursors used in some gas-phase processes.

Compound-semiconductor liquid-phase epitaxy often employs a metallic solvent. Gallium arsenide, for example, can be grown from a gallium-rich solution containing dissolved arsenic. The substrate and melt are brought into contact under conditions chosen to avoid excessive substrate dissolution before growth begins. Boat and slider arrangements have historically allowed a substrate to contact successive melts, permitting multilayer structures. Temperature control and knowledge of the relevant phase diagram determine the amount and composition of material deposited.

Liquid-phase epitaxy was important in early semiconductor light emitters and remains relevant to selected materials and thick crystalline films. Epitaxial garnet films grown from high-temperature solutions are used in magnetic and magneto-optical applications. The method can provide comparatively high growth rates and good crystalline quality, but controlling extremely thin layers, abrupt composition changes, and large-area uniformity is generally more difficult than with beam-based or advanced vapor-phase techniques. Residual solvent incorporation and melt–substrate reactions can also matter.

Solid-phase epitaxy proceeds without a liquid growth medium. An amorphous layer in contact with a crystal can recrystallize during heating, with the crystalline region acting as the template. A prominent example is regrowth of silicon that has been amorphized by ion implantation. The crystalline–amorphous interface advances as atoms rearrange into the underlying lattice. The regrowth velocity is strongly temperature dependent and also responds to impurities, dopant concentration, stress, and crystallographic orientation.

This process is important because implantation can introduce dopants while also damaging the lattice. Annealing can restore crystalline order and place some implanted dopants on electrically active sites, although regrowth, dopant activation, and defect removal are distinct processes. Defects near the original end of the implanted-ion range may survive even after the amorphous layer has recrystallized. Laser-induced melting followed by epitaxial resolidification is another approach to repairing or restructuring a surface layer, but it is properly distinguished from solid-phase epitaxy because a transient liquid is involved. Heating and cooling histories determine whether regrowth follows the substrate or competes with unwanted nucleation.

Measuring orientation, interfaces, and material quality

No single measurement establishes all the characteristics of an epitaxial film. Orientation, strain, composition, thickness, surface roughness, and defect density are related but distinct quantities. A narrow diffraction peak does not by itself prove that an interface is chemically abrupt, just as a smooth surface does not prove that the underlying film is free of dislocations. Characterization therefore combines methods that probe different length scales and physical properties.

High-resolution X-ray diffraction is widely used to determine lattice spacings, orientation, and strain. Symmetric reflections primarily probe planes parallel to the surface, whereas asymmetric reflections contain information about both in-plane and out-of-plane lattice dimensions. Reciprocal-space maps can distinguish a film that remains coherently strained from one that has partly or largely relaxed. Rocking-curve widths are sensitive to orientational spread, but converting a width into a dislocation density requires assumptions about defect types and their statistical distribution.

Azimuthal scans and pole figures test in-plane alignment and identify rotational variants. This is important because an ordinary out-of-plane diffraction scan can show a strong preferred orientation even for a film lacking full epitaxial registration. Thickness fringes around film reflections can reveal coherent interference from smooth, reasonably uniform layers. Periodic multilayers generate satellite peaks whose spacing provides information about the repeat period, while peak intensities and shapes depend on composition, interface roughness, and structural coherence.

Transmission electron microscopy resolves interfaces, lattice defects, and local crystallographic relationships. Cross-sectional specimens reveal layer sequences and threading defects; plan-view specimens help assess lateral domain structures and dislocation arrangements. Scanning transmission electron microscopy, combined with suitable detectors and analytical spectroscopy, can distinguish atomic columns and map elemental distributions. Interpretation must account for specimen thickness, projection, preparation damage, and the limited region sampled. An exceptionally clean microscopic image may not represent an entire wafer.

Atomic force microscopy measures surface terraces, step bunching, pits, and islands, while secondary-ion mass spectrometry provides depth profiles of dopants and impurities. X-ray reflectivity is useful for thickness, density, and interface roughness in suitable films. Electrical measurements such as Hall measurements and capacitance–voltage profiling assess carrier concentration and mobility; optical measurements probe recombination and electronic transitions. In situ reflectometry, pyrometry, diffraction, and wafer-curvature monitoring can track growth as it occurs. Reliable interpretation depends on calibration and on separating instrument response from the physical property being inferred.

Semiconductor structures and device functions

Epitaxy enables independent control of a device’s supporting substrate and active layers. In silicon electronics, a lightly doped epitaxial layer can be grown on a more heavily doped wafer, providing the electrical properties needed near the surface while retaining a low-resistance mechanical support. Epitaxial drift regions are particularly important in many power devices because their thickness and doping help determine the trade-off between blocking voltage and conduction resistance. Comparable control is essential in silicon carbide devices used in power electronics.

Selective epitaxy allows material to be added only where a transistor architecture requires it. Silicon–germanium source and drain regions can impose compressive strain on neighboring silicon channels, while other material and stress arrangements are used to modify electron transport. The benefit comes from strain-induced changes in electronic band structure, not simply from making the crystal more ordered. Device integration must also manage contact resistance, unwanted defects, and the thermal exposure of structures already fabricated on the wafer.

Epitaxial heterojunctions join semiconductors with different electronic band structures. A thin, narrower-band-gap layer between wider-band-gap barriers can confine carriers to form a quantum well. Repeating layers produces a superlattice, in which sufficiently strong coupling between wells can generate minibands. Layer thicknesses of a few nanometers make quantum confinement important, so fluctuations of only one or two atomic planes can measurably shift optical transition energies.

These structures underpin many light-emitting diodes and semiconductor lasers. Gallium arsenide/aluminum gallium arsenide and indium phosphide-based heterostructures support a wide range of infrared devices. Indium gallium nitride quantum wells in gallium nitride structures provide the active regions of many blue and green light emitters. Crystal defects, polarization fields, alloy composition, and interface roughness all influence efficiency and emission wavelength. Epitaxy supplies the structural platform, but a complete device also requires contacts, patterning, heat removal, and optical design.

High-electron-mobility transistors exploit heterointerfaces that confine mobile electrons near an interface while controlling scattering and carrier supply. Aluminum gallium nitride/gallium nitride devices can use polarization-related charge to help create a high-density electron channel. In photovoltaic technology, epitaxial III–V junctions with different band gaps are stacked to absorb different parts of the solar spectrum, forming a multijunction solar cell. Lattice-matched designs favor crystalline compatibility, whereas metamorphic designs deliberately accommodate lattice changes through buffer structures. Across these applications, epitaxial quality is judged by the defects and interfaces that matter to the device, rather than by a single universal quality metric.

Complex oxides, two-dimensional materials, and integration limits

Epitaxy also provides a way to alter the behavior of materials whose properties are strongly coupled to atomic geometry. In complex oxides, changes in bond lengths, bond angles, oxygen content, and crystal symmetry can modify ferroelectricity, magnetism, electrical conduction, and superconductivity. A substrate is therefore more than a support: its lattice dimensions and surface termination can select structures or domain patterns that differ from those of a relaxed bulk crystal. Such effects are especially pronounced when layers are only a few unit cells thick.

Strontium titanate is a widely used substrate and model material for oxide epitaxy. The interface between lanthanum aluminate and strontium titanate became a major research subject after conducting behavior was reported in 2004 by Akira Ohtomo and Harold Hwang. The properties depend on growth conditions and processing, with oxygen vacancies, interfacial charge redistribution, chemical intermixing, and electrostatic boundary conditions playing roles in different samples. This illustrates why crystallographic alignment alone cannot specify an interface’s electronic state.

Van der Waals epitaxy extends the concept to systems in which weak interlayer interactions replace some of the strong bonding constraints typical of conventional semiconductor interfaces. Layered materials can sometimes grow with preferred orientations despite large differences in lattice constants. Graphene, hexagonal boron nitride, and transition-metal dichalcogenides are important examples in this broader research area. Weak bonding does not remove all structural problems: rotational domains, grain boundaries, substrate steps, and contamination can still control film continuity and electronic properties.

Heterogeneous integration seeks to combine materials that are difficult to grow together directly. III–V semiconductors on silicon are attractive for joining efficient light emitters with established silicon circuitry, but lattice mismatch, thermal expansion differences, and polar–nonpolar interfaces create substantial challenges. Direct epitaxy, patterned growth, defect-filtering buffers, and lateral overgrowth address parts of this problem. Wafer bonding and layer transfer provide alternatives; these can incorporate separately grown epitaxial materials without requiring an epitaxial relationship across the final bonded interface.

The practical limits of epitaxy are also economic and statistical. A structure demonstrated on a small substrate must often be reproduced over a much larger area with tight control of thickness, composition, bow, and defect distribution. Precursor purity, reactor deposits, substrate reuse, maintenance, and throughput affect manufacturing viability. An isolated defect may be unimportant in one device but cause failure in another, especially when large active areas or high electric fields are involved. Consequently, continuing development concerns not only new crystal combinations, but also reproducible process windows, nondestructive inspection, and integration with the temperature and contamination constraints of complete manufacturing sequences.