A gate-all-around transistor (GAA transistor or GAAFET) is a metal–oxide–semiconductor field-effect transistor whose gate surrounds the cross-section of its conducting channel. An insulating dielectric separates the gate from the channel, while source and drain regions connect to the channel’s ends. By controlling the channel from all sides, the gate can suppress unwanted current more effectively than a planar gate and, for suitable dimensions, a three-sided gate. GAA structures are used in advanced integrated circuits to extend transistor scaling beyond conventional fin-based designs.
The term describes a device geometry rather than a particular material or manufacturing process. Channels may take the form of narrow wires, broad thin sheets, or other elongated cross-sections, and multiple channels may share one gate.
Structure and operating principle
Like other field-effect transistors, a GAA device uses an electric field from the gate to regulate current between source and drain. In a typical enhancement-mode device, a suitable gate voltage creates a conducting channel; removing that voltage suppresses conduction.
The channel is made from a semiconductor, most commonly silicon in mainstream logic processes. The surrounding gate stack consists of a thin insulating layer, generally incorporating a high-permittivity material, and conductive gate materials. The insulator prevents direct electrical contact between gate and channel while allowing electrostatic coupling.
“Gate-all-around” refers to the gate surrounding the channel in a cross-section perpendicular to current flow. It does not mean that the gate encloses the entire transistor or covers the source and drain contacts.
Electrostatic control
As a transistor’s gate length decreases, the drain’s electric field increasingly influences the channel. These short-channel effects can lower the barrier that normally prevents current from flowing when the transistor is off. Consequences include increased leakage and a threshold voltage that depends more strongly on drain voltage.
A surrounding gate gives the gate electrode greater control relative to the drain. Thin channel bodies further reduce regions that are weakly controlled by the gate. This permits shorter channel lengths or lower off-state leakage than would otherwise be practical.
One measure of switching behavior is the subthreshold swing:
where is gate voltage and is drain current. A smaller value means that less gate-voltage change is needed to change current by a factor of ten. Improved electrostatics can bring a conventional GAA MOSFET closer to the thermal limit,
approximately 60 millivolts per decade at room temperature. Here is the Boltzmann constant, is absolute temperature, and is the elementary charge. Surrounding the channel does not, by itself, overcome this limit for ordinary thermionic MOSFET operation.
Main geometries
Nanowire transistors
A nanowire transistor has a narrow channel with a roughly circular, square, or similar compact cross-section. Small channel dimensions provide strong electrostatic control, but a single wire offers limited current. Multiple wires can be connected in parallel to increase drive current.
Nanowires may be horizontal, with current flowing approximately parallel to the wafer surface, or vertical, with current flowing approximately perpendicular to it.
Nanosheet transistors
A nanosheet transistor uses a thin, relatively wide channel. Several horizontal sheets are commonly stacked vertically, with the gate surrounding each sheet.
The sheet thickness can remain small for strong gate control, while the width provides a larger conducting perimeter and greater current capacity. Designers can select different sheet widths within manufacturing and layout constraints, providing an additional way to balance performance, power consumption, and area.
For an ideal rectangular sheet of width and thickness , its gated perimeter is approximately
For similar sheets, the total gated perimeter is approximately . This is a useful geometric comparison, not a complete current model: mobility, surface orientation, series resistance, and electrostatic behavior also affect performance.
The boundary between “nanowire” and “nanosheet” is not universally standardized. Both belong to the broader GAA family.
Comparison with planar transistors and FinFETs
A planar MOSFET places its gate primarily above a channel near the semiconductor surface. A FinFET raises the channel into a fin and places the gate over its top and along its two sidewalls. A GAA transistor extends gate coverage around the full channel perimeter.
| Characteristic | Planar MOSFET | FinFET | Stacked nanosheet GAA |
|---|---|---|---|
| Typical channel geometry | Surface channel | Upright fin | Vertically stacked thin sheets |
| Gate coverage | Primarily one side | Typically three sides | Full perimeter of each sheet |
| Electrostatic control at small dimensions | More limited | Stronger than planar | Potentially stronger than FinFET |
| Common drive-strength adjustment | Channel width | Number and geometry of fins | Sheet width, number, and geometry |
| Fabrication demands | Relatively simpler geometry | Fin patterning and gate integration | Multilayer growth, channel release, and surrounding-gate formation |
These comparisons describe architectural tendencies rather than guaranteed performance rankings. A finished circuit depends on the complete process, device dimensions, interconnects, and design rules.
GAA technology also remains compatible with complementary metal–oxide–semiconductor logic: n-channel and p-channel transistors are combined to implement logic functions. The transition changes transistor structure rather than the underlying Boolean functions implemented by logic gates.
Manufacturing
A common horizontal nanosheet process begins with alternating layers of silicon and silicon–germanium grown on a semiconductor wafer. The silicon layers become channels, while intervening silicon–germanium layers serve as sacrificial material.
A simplified fabrication sequence includes:
- Stack formation and patterning. The multilayer stack is patterned into narrow structures using photolithography and etching.
- Temporary gate formation. A sacrificial gate defines the channel region while other structures are fabricated.
- Source and drain formation. Semiconductor regions are grown and processed to connect the channel ends. Doping and strain engineering help establish the required electrical properties.
- Spacer formation and channel release. Insulating inner spacers separate the eventual gate from source and drain regions. Selective etching removes sacrificial layers beneath the gate while preserving the channels.
- Replacement gate formation. Gate dielectric and metal layers are deposited around the released sheets, including within the narrow gaps between them.
- Contacts and interconnects. Conductive connections integrate the devices into circuits.
The exact sequence varies between processes. Highly selective etching and conformal deposition, including atomic layer deposition, are important because small material-removal or thickness errors can affect several channel surfaces simultaneously.
Development and adoption
Surrounding-gate transistor concepts were investigated well before their commercial use, with substantial research into silicon wires and related structures during the late twentieth century. Their industrial importance increased as planar MOSFET scaling became more difficult and FinFETs became established in advanced logic manufacturing.
In 2017, IBM and research partners announced a 5-nanometer-class research technology using stacked silicon nanosheet transistors. In 2022, Samsung announced initial production using a GAA architecture at its 3-nanometer process generation, employing its Multi-Bridge-Channel FET implementation.
Manufacturers use different names for related designs. Samsung’s MBCFET and Intel’s RibbonFET are implementations within the broader GAA family, not separate fundamental operating principles.
Process-node labels such as “3 nm” and “2 nm” denote technology generations. They do not directly specify gate length, channel thickness, or a single physical dimension, and they are not sufficient for comparing different manufacturers’ processes.
Applications and design implications
GAA transistors support continued increases in logic density and improvements in performance or energy efficiency associated with Moore’s law. Intended applications include central processing units, graphics processing units, mobile processors, and specialized accelerators.
Better electrostatic control can reduce leakage or permit operation at a lower supply voltage for a given performance target. However, the outcome depends on the complete circuit. Dynamic switching power is approximately
where is switching activity, is switched capacitance, is supply voltage, and is operating frequency. Lower voltage can substantially reduce this component of power, but transistor geometry also affects capacitance and achievable frequency.
Adoption requires new device models, layout rules, and circuit libraries within electronic design automation systems. Logic cells and static random-access memory must be optimized for the new architecture; smaller transistors do not automatically yield proportional reductions in every circuit block.
Limitations and further development
GAA structures introduce several engineering challenges:
- Process complexity and yield: releasing thin channels and depositing uniform gate stacks around them require tight control.
- Contact and access resistance: resistance outside the gated channel can limit gains in intrinsic transistor performance.
- Parasitic capacitance: coupling involving gates, source and drain regions, and nearby interconnects can offset switching-speed improvements.
- Self-heating: narrow, partially isolated channels can make heat removal more difficult.
- Variability and reliability: channel dimensions, interface quality, gate materials, and dielectric defects affect electrical behavior and operating lifetime.
- Scaling limits: direct source-to-drain tunneling and other quantum effects become increasingly important at very small dimensions.
Further developments include forksheet structures, which introduce dielectric separation between neighboring devices, and the complementary field-effect transistor or CFET, which stacks n-channel and p-channel devices vertically. These approaches address device spacing and circuit density as well as gate control, and require additional solutions for fabrication, contacts, and heat removal.