A field-effect transistor (FET) is a transistor in which an electric field controls the flow of charge through a conducting channel. The field is established principally by a voltage applied to a gate electrode, while current flows between two other terminals, the source and drain. Unlike a bipolar junction transistor, whose operation depends on coupled electron and hole injection, a conventional FET conducts mainly through one carrier type. This distinction gives rise to the traditional description unipolar transistor, although some field-effect devices support ambipolar conduction.
FETs are central components of modern electronics. The metal–oxide–semiconductor field-effect transistor (MOSFET), their most widely used form, is the principal switching element in integrated circuits, including processors and semiconductor memories. Other forms provide low-noise amplification, microwave power generation, high-voltage switching, and the electronic backplanes of displays. Their importance follows from a combination of electrostatic control, low steady-state gate current, small physical size, and compatibility with large-scale fabrication.
The term describes a family rather than a single structure. Junction gates, insulated gates, metal–semiconductor contacts, and semiconductor heterojunctions can all produce field-effect control. Understanding their differences requires distinguishing the common electrostatic principle from the particular materials, geometries, and operating conditions of each device.
Electrostatic control and channel formation
A FET contains a channel whose electrical conductivity changes in response to the gate potential. In a semiconductor, conductivity depends on the concentration and mobility of mobile charge carriers: electrons in a conduction band or holes associated with a valence band. Semiconductor doping establishes background carrier concentrations, while the gate field changes the local carrier population. Depending on the structure and bias, the field can remove carriers from an existing channel, accumulate additional carriers, or create a conducting layer of the opposite carrier type.
The source supplies carriers to the channel, and the drain collects them under an applied drain–source voltage. These names indicate electrical function rather than necessarily identifying permanently different physical structures. In many integrated MOSFETs, the source and drain are approximately symmetric, and their operational roles depend on the applied potentials. In discrete power devices, structural asymmetry, the body connection, and an intrinsic diode make the distinction more consequential.
An n-channel enhancement-mode MOSFET illustrates channel formation. Its simplest form has two heavily doped n-type regions in a p-type body, separated by a surface region beneath an insulated gate. A sufficiently positive gate voltage first repels holes from that surface, producing depletion, and then attracts enough electrons to create an inversion layer. This layer connects source and drain. The gate voltage associated with a specified onset of appreciable channel conduction is called the threshold voltage. It is not an abrupt physical boundary: current varies continuously through weak, moderate, and strong inversion.
A p-channel MOSFET operates with reversed voltage polarities and holes as its channel carriers. In an enhancement-mode device, the conducting channel is absent or weak at zero gate–source voltage under specified conditions. A depletion-mode device has a conducting channel at zero gate bias and requires a gate voltage of the appropriate polarity to suppress conduction. These classifications are distinct from the choice between n-channel and p-channel operation.
The gate does not ordinarily supply the carriers that constitute the drain current. Instead, it changes their electrostatic environment. This accounts for the high input resistance of insulated-gate devices, but not for zero input power under all conditions. Gate capacitances must be charged and discharged whenever the voltage changes, and real dielectrics exhibit leakage. Many MOSFETs also have a fourth terminal, the body or bulk. Its voltage affects depletion charge and threshold voltage, producing the body effect. Thus, the familiar three-terminal representation often conceals an electrically significant fourth region.
Principal device families
The junction field-effect transistor (JFET) controls a pre-existing semiconductor channel with a reverse-biased p–n junction. Increasing the reverse bias widens the junction’s depletion region, reducing the channel’s conducting cross-section. Conventional JFETs are normally on at zero gate–source voltage and are therefore depletion-mode devices. Their gate current is usually small when the junction remains reverse biased, although junction leakage increases with temperature. JFETs are used in high-input-impedance amplifiers, analog switches, current regulators, and low-noise input stages.
In an insulated-gate FET, a dielectric separates the gate conductor from the semiconductor. The MOSFET is the dominant example, although modern devices need not literally contain a metal gate above silicon dioxide: historical and contemporary processes have used different gate materials and dielectric stacks. Insulated gates permit both enhancement- and depletion-mode implementations. They also allow the gate to control the channel without the forward-bias limitation of a junction gate, subject instead to dielectric leakage and breakdown limits.
A metal–semiconductor field-effect transistor (MESFET) uses a rectifying metal–semiconductor contact, commonly called a Schottky gate, to deplete a conducting channel. Gallium arsenide MESFETs became important in microwave electronics because the material supports high electron mobility and favorable high-frequency operation. The gate is normally operated without substantial forward conduction. Unlike a MOSFET, the conventional MESFET has no insulating gate dielectric.
A high-electron-mobility transistor (HEMT) uses a heterojunction between dissimilar semiconductors to confine carriers near an interface. In many structures, this produces a two-dimensional electron gas with high mobility. Aluminum gallium arsenide/gallium arsenide devices can separate mobile carriers from ionized dopants that would otherwise scatter them. Aluminum gallium nitride/gallium nitride devices exploit polarization effects to generate high sheet carrier densities. HEMTs are especially important in low-noise microwave receivers, radio-frequency power amplifiers, and power conversion.
A thin-film transistor (TFT) is distinguished chiefly by its construction in a thin semiconductor film on a supporting substrate, which may be glass or plastic rather than a single-crystal semiconductor wafer. Amorphous silicon, polycrystalline silicon, metal oxides, and organic semiconductors support different compromises among mobility, uniformity, processing temperature, and cost. TFTs switch and drive individual elements in liquid-crystal displays and organic light-emitting diode displays. These categories can overlap: “thin-film,” “insulated-gate,” and “heterojunction” identify different aspects of a device rather than mutually exclusive alternatives.
Invention and industrial development
The field-effect principle predates the first successful transistor. Julius Edgar Lilienfeld filed patent applications in the 1920s describing semiconductor devices controlled by an electric field; an important United States application followed a Canadian filing in 1925. Oskar Heil proposed another insulated-gate arrangement in a patent filed in 1934. These early proposals did not establish a practical manufacturing technology. Semiconductor purity, surface preparation, reliable contacts, and control of interface states remained major obstacles.
Research at Bell Laboratories during the 1940s initially included efforts to obtain field-effect amplification. Surface states could trap charge and screen the applied field, preventing the expected modulation of the semiconductor interior. Investigations of these effects contributed to the work that produced the point-contact transistor in 1947, associated with John Bardeen and Walter Brattain. William Shockley subsequently developed the junction-transistor concept and published a major theoretical treatment of a unipolar field-effect transistor in 1952. Experimental junction FET work in the early 1950s established practical field-effect devices before MOS technology became commercially dominant.
The decisive step toward MOS devices was control of the silicon surface. Mohamed Atalla’s work on silicon surface passivation with thermally grown silicon dioxide helped make the silicon–oxide interface usable for electronic devices. Atalla and Dawon Kahng fabricated a working MOSFET at Bell Laboratories in 1959 and publicly demonstrated the technology in 1960. The oxide served both as a gate insulator and as a means of stabilizing the semiconductor surface, linking a physical operating principle to a reproducible fabrication route.
MOS technology developed rapidly during the 1960s. Frank Wanlass introduced the complementary MOS circuit concept in 1963, combining n-channel and p-channel devices to reduce static power consumption. Early commercial MOS circuits often used p-channel devices because of manufacturing considerations, while later n-channel processes offered higher speed. Self-aligned silicon-gate processes, developed in the late 1960s, reduced alignment-related parasitic capacitance and improved manufacturability. These changes supported increasingly dense memories and the first commercial microprocessors.
During the following decades, complementary metal–oxide–semiconductor technology became dominant in digital integrated circuits. Its expansion depended on improvements in lithography, oxidation, implantation, interconnection, and circuit design rather than on a single invention. The scaling framework published by Robert Dennard and colleagues in 1974 explained how reducing device dimensions and voltages together could preserve electric fields while improving density and speed. Later developments—including strained channels, high-permittivity gate dielectrics, metal gates, and three-dimensional channel geometries—addressed limits that emerged as straightforward geometric scaling became insufficient.
Current–voltage behavior and device models
FET behavior is described by relationships among terminal voltages, currents, charge, and temperature. For an ideal long-channel n-channel enhancement MOSFET with its body tied to its source, a useful strong-inversion approximation begins with the gate overdrive . Here is gate–source voltage and is threshold voltage. In the gradual-channel approximation, with constant mobility and negligible parasitic resistance, the drain current in the linear or triode region is
for . The quantities , , , and are electron mobility, gate-oxide capacitance per unit area, channel width, and channel length.
When reaches approximately , inversion charge near the drain becomes small. This condition is called pinch-off. It does not stop conduction: carriers arriving from the channel are swept through the high-field drain-end region. The idealized saturation current is
Actual current commonly increases with drain voltage because the effective channel length changes and other nonideal effects occur. A simple correction multiplies the ideal saturation current by , where approximates channel-length modulation over a limited operating range. These square-law expressions are instructional models, not universal descriptions of nanoscale devices.
Below threshold, current is not identically zero. In weak inversion, diffusion contributes strongly, giving an approximately exponential dependence on gate voltage. The subthreshold swing, the gate-voltage change needed for a tenfold current change, is ideally
under a simplified electrostatic model. Here is Boltzmann’s constant, absolute temperature, elementary charge, and depletion capacitance per unit area. For a conventional thermionic MOSFET at about 300 K, the ideal limiting value is approximately 60 millivolts per decade; interface traps and imperfect gate control generally increase it.
For small signals around a bias point, transconductance is , while output conductance is , with other relevant conditions held fixed. Together with gate–source, gate–drain, and junction capacitances, these quantities determine gain and bandwidth. Short-channel transistors require models incorporating mobility degradation, velocity saturation, leakage, geometry-dependent electrostatics, and parasitic resistance. Circuit simulators therefore use calibrated compact models, such as the BSIM family, rather than treating the elementary square law as a complete device description.
Materials and fabrication
Most integrated logic FETs are fabricated on silicon because it combines suitable electronic properties with a highly developed manufacturing infrastructure and a useful native oxide. A typical process begins with a semiconductor wafer, followed by repeated steps that define isolated active regions, form gates, introduce dopants, create contacts, and build interconnecting metal layers. Photolithography transfers geometric patterns into photoresist; etching, deposition, implantation, and thermal processing translate those patterns into electrically functional structures.
In a traditional self-aligned MOS process, the gate helps define where source and drain regions are formed. This reduces the dependence of channel geometry and overlap capacitance on the alignment between separate masks. Modern processes add refinements such as sidewall spacers, shallow extensions, raised source/drain regions, and selective epitaxial growth. These control resistance, electric fields, strain, and junction placement. Isolation structures prevent neighboring devices from conducting unintentionally through the substrate, while wells allow n-channel and p-channel devices to coexist.
The gate stack presents a central materials trade-off. A thinner dielectric increases gate capacitance and strengthens electrostatic control, but very thin silicon dioxide permits appreciable electron tunneling. High-permittivity gate dielectrics, notably hafnium-based materials, provide a larger capacitance than silicon dioxide at the same physical thickness. Their performance is often expressed through equivalent oxide thickness: the thickness of silicon dioxide that would provide the same capacitance per unit area. Metal gates help establish suitable work functions and avoid limitations associated with heavily doped polysilicon gates.
Silicon is not optimal for every application. Silicon carbide and gallium nitride have wide bandgaps and high critical electric fields, making them valuable for high-voltage and high-temperature electronics. Gallium arsenide and indium-based compounds support high-frequency and optoelectronic applications. Material selection also depends on thermal conductivity, defect density, wafer availability, contact resistance, dielectric quality, and cost. A high bulk carrier mobility alone does not guarantee a superior finished transistor.
Fabrication must control variations across both individual devices and entire wafers. Local differences in gate dimensions, interface charge, grain structure, doping, or strain can change threshold voltage and current. Such variation affects digital timing and memory yield, as well as analog matching. Electrical characterization measures transfer curves, output curves, capacitances, leakage, and breakdown behavior. Process integration must also accommodate interconnect resistance and capacitance: a fast isolated transistor does not necessarily produce a fast circuit when its wiring and neighboring devices are included.
Scaling and three-dimensional geometries
Reducing transistor dimensions allows more devices to occupy a given chip area, a trend associated with Moore’s law. Classical constant-field scaling reduces device dimensions, voltages, and certain material dimensions together so that internal electric fields remain approximately unchanged. Under its ideal assumptions, smaller devices switch faster and use less energy per transition without increasing power density. These assumptions became progressively harder to maintain as threshold voltages, leakage currents, atomic dimensions, and practical supply voltages constrained further reductions.
A principal difficulty is maintaining gate control over a short channel. In a long-channel device, the gate largely determines the electrostatic barrier that regulates carrier injection from the source. As the source and drain approach each other, their fields influence that barrier more strongly. Drain-induced barrier lowering allows increasing drain voltage to raise off-state current. Related short-channel effects include threshold-voltage shifts and degraded subthreshold behavior. Very short channels may also exhibit substantial source-to-drain tunneling, which cannot be eliminated merely by improving classical gate control.
One response is to make the active semiconductor body thin enough for the gate to control it more completely. Fully depleted silicon-on-insulator devices place a thin silicon layer above an insulating layer, reducing unwanted substrate coupling. A FinFET forms the channel in a narrow vertical fin, with the gate controlling multiple sides. This geometry improves electrostatic control relative to a comparable planar device. Commercial high-performance logic adopted FinFET technology during the 2010s, while planar structures remained useful in many analog, power, and specialized processes.
A gate-all-around transistor extends the principle by surrounding the channel more completely. Nanowire and nanosheet implementations place one or more narrow semiconductor channels inside the gate structure. Stacked nanosheets can provide substantial effective channel width within a compact footprint, while sheet dimensions offer a means of adjusting drive strength. These structures require complex fabrication, including selective removal of sacrificial layers and controlled formation of gate material around the remaining channels.
Scaling is also limited by resistance, heat, and variability outside the ideal channel. Source/drain contacts can contribute a large fraction of total resistance; closely spaced interconnects increase capacitive coupling; and dense switching activity produces difficult thermal conditions. Contemporary process-node labels therefore do not specify a universal gate length or a single physical dimension. Progress involves coordinated changes in transistor geometry, wiring, cell layout, electronic design automation, and packaging. Performance comparisons require actual electrical and dimensional information rather than node names alone.
Digital, analog, and memory circuits
The fundamental digital application of the FET is the controlled switch. In a CMOS inverter, a p-channel transistor connects the output toward the positive supply, while an n-channel transistor connects it toward the lower supply. Their gates share an input. A low input turns the p-channel device on and the n-channel device off; a high input reverses those states. Ideally, either stable logic state has no direct conducting path between supply rails. During transitions, capacitances charge and discharge, and both devices may conduct simultaneously for part of the transition.
The approximate dynamic power associated with capacitive switching is
where represents switching activity under the chosen convention, the effective switched capacitance, supply voltage, and operating frequency. Leakage and short-circuit current add to this component. The quadratic voltage dependence explains the importance of supply-voltage reduction, while the need to preserve speed and noise margins limits how far voltage can fall. Networks of complementary transistors implement logic gates, registers, and the functional units of a central processing unit.
FETs also serve as continuously controlled analog elements. A common-source amplifier converts a gate-voltage variation into a drain-current variation, which a load converts into an output voltage. Source followers provide buffering, while differential pairs compare two input voltages and underpin many amplifiers and comparators. Current mirrors reproduce or scale bias currents. Important design quantities include transconductance, output resistance, voltage headroom, device matching, noise, and parasitic capacitance. A transistor optimized for digital switching is not automatically optimal for precision analog use.
Memory circuits exploit several distinct FET functions. A typical six-transistor static random-access memory cell uses two cross-coupled CMOS inverters and two access transistors. A conventional dynamic random-access memory cell combines an access transistor with a storage capacitor and requires periodic refresh. In flash memory, charge stored in a floating gate or charge-trapping dielectric shifts a transistor’s threshold voltage. Reading detects that threshold-dependent conduction, while programming and erasing alter the stored charge through mechanisms such as tunneling.
Field-effect control also integrates sensing with readout. In a CMOS image sensor, transistors reset and select pixels and buffer electrical signals associated with collected photogenerated charge. An ion-sensitive FET replaces the usual gate environment with a chemically responsive interface, allowing surface potential changes to modulate channel current. Such sensing depends on interface chemistry, reference potentials, drift, and calibration in addition to transistor characteristics. Across these applications, the same electrostatic mechanism is adapted to switching, amplification, charge storage, or measurement rather than representing a single fixed circuit role.
Power switching and radio-frequency operation
Power FETs are designed to control substantial current and voltage while limiting conduction and switching losses. Their structures differ markedly from the small lateral transistors used in digital logic. A vertical power MOSFET conducts mainly through the thickness of the die, allowing many parallel cells to share a voltage-supporting drift region. Current capability scales strongly with active area, while blocking capability depends on the drift region’s thickness, doping, and material properties.
The principal conduction loss in a fully enhanced MOSFET is approximately
where is on-state drain–source resistance under specified gate-voltage and temperature conditions. Increasing voltage capability generally requires a more resistive drift region in conventional silicon structures. Superjunction devices reduce this trade-off through alternating, charge-balanced regions that reshape the electric field. Silicon carbide MOSFETs exploit the material’s higher critical field to achieve favorable resistance at high blocking voltages.
Switching performance depends on gate charge and nonlinear internal capacitances, not simply on a nominal input resistance. During part of a switching transition, gate charge changes the drain voltage through the gate–drain capacitance, producing the familiar Miller plateau in the gate-voltage waveform. Gate-drive resistance, package inductance, and circuit layout affect switching speed, overshoot, and electromagnetic emissions. Consequently, a device with lower on-resistance may not yield lower total loss if its gate charge or output-capacitance losses are substantially greater.
A silicon power MOSFET usually includes an intrinsic body diode. Its conduction and reverse-recovery behavior matter in bridges and other circuits where current reverses direction. Gallium nitride HEMTs can conduct in the reverse direction without the same intrinsic p–n body-diode structure, although their losses still depend on gate bias, stored capacitive energy, and circuit conditions. Power FETs share applications with the insulated-gate bipolar transistor, which uses an insulated gate but relies on bipolar conduction and is therefore not simply a conventional unipolar FET.
At radio frequencies, gain, noise, output power, efficiency, and stability become central. The current-gain cutoff frequency describes where short-circuit current gain falls to unity under specified conditions, while the maximum oscillation frequency reflects power-gain limitations. Gate resistance, transit time, feedback capacitance, and access resistance all matter. Gallium arsenide and indium-based HEMTs are used in low-noise microwave amplification, while gallium nitride HEMTs support high-power radio-frequency applications. Thermal design and impedance matching remain essential even when the intrinsic transistor has excellent high-frequency characteristics.
Reliability, physical limits, and emerging devices
FET reliability depends on how electrical and thermal stress change materials and interfaces over time. Gate dielectrics may accumulate defects until leakage rises or a conductive breakdown path forms. Time-dependent dielectric breakdown therefore differs from an immediate failure caused by a single excessive voltage. Insulated gates are also vulnerable to electrostatic discharge, which can create localized damage. Protection structures reduce this vulnerability but introduce capacitance and other electrical trade-offs.
Carrier trapping can gradually alter threshold voltage and drive current. Bias-temperature instability depends on gate bias, temperature, dielectric composition, and stress history. Some trapped charge can be released after the stress changes, so measured degradation depends partly on the measurement procedure. Hot-carrier effects arise when energetic carriers damage interfaces or become trapped in dielectric regions. These mechanisms are not interchangeable: each responds differently to voltage, temperature, geometry, and operating duty cycle.
Self-heating changes carrier mobility, resistance, leakage, and aging rates. It is especially important in power devices and in structures whose channels are thermally isolated by surrounding dielectrics. Parallel operation does not automatically ensure uniform current distribution; thermal feedback, device mismatch, and circuit parasitics can concentrate stress. High-field operation may also produce avalanche breakdown. A device’s safe operating area consequently includes constraints involving voltage, current, pulse duration, temperature, and heat removal, rather than a single maximum-current rating.
At small dimensions, individual defects can produce measurable random telegraph fluctuations in current. Low-frequency noise, often approximated by a spectrum over a limited range, can involve trapping and carrier-number or mobility fluctuations. Quantum confinement changes the available carrier states in thin channels, while short transit distances can make transport partly ballistic. These effects require treatments beyond a bulk, constant-mobility drift model, even though compact circuit models often represent them through calibrated effective parameters.
Research devices explore alternatives to conventional materials and injection mechanisms. Tunnel FETs seek to control band-to-band tunneling and can, under particular conditions, avoid the conventional thermionic subthreshold-swing limit; achieving high useful current and reproducible fabrication remains difficult. Carbon nanotube FETs offer thin channels and favorable transport, but placement, contacts, and material uniformity complicate integration. Atomically thin semiconductors such as molybdenum disulfide provide strong geometric gate control, while contact resistance and dielectric interfaces remain important challenges. Ferroelectric gate stacks support nonvolatile threshold states in ferroelectric FETs, and related negative-capacitance concepts investigate ways of modifying gate electrostatics. These approaches target different requirements and cannot be compared adequately through channel dimensions or a single performance metric alone.