What is a High Electron Mobility Transistor (HEMT)?

A High Electron Mobility Transistor (HEMT) is a field-effect transistor incorporating a junction between two materials with different bandgaps – known as a heterojunction – as the channel. Traditional field-effect transistors rely on doped semiconductor regions where carrier conduction is inherently limited by impurity scattering. HEMTs bypass this limitation through modulation doping or spontaneous/piezoelectric polarization.

This induces a highly conductive, undoped spatial layer known as a Two-Dimensional Electron Gas (2DEG).

The core physics of the 2DEG

In Gallium Nitride (GaN) heterostructures, typically comprising an Aluminum Gallium Nitride (AlxGa1-xN) barrier layer grown on a GaN buffer layer, the 2DEG is formed due to macroscopic polarization fields.

Because GaN crystallizes in the non-centrosymmetric wurtzite structure along the c-axis, it exhibits strong spontaneous polarization (PSP).

Furthermore, growing a lattice-mismatched AlGaN layer pseudomorphically on top of the thicker GaN buffer induces tensile strain.

This strain generates piezoelectric polarization (PPE). This abrupt polarization discontinuity induces a massive fixed positive sheet charge at the interface.

To preserve charge neutrality, free electrons from the surface states and bulk donor states accumulate in the localized potential well formed at the conduction band.

GaN HEMT - Structure, physics & E-HEMTs

GaN HEMT structure 

The D-mode vs.E-HEMT paradigm shift

Standard GaN HEMT heterostructures naturally yield a highly conductive 2DEG channel at zero gate bias (VGS = 0 V). This behavior classifies them as Depletion-mode (D-mode) or "normally-on" devices.

In industrial power electronics, normally-off devices are preferred. Achieving Enhancement-mode (E-mode) or "normally-off" operation—where the channel is completely non-conductive at VGS = 0 V and requires a positive threshold voltage (VTH > 0 V) to initiate conduction—is mandatory for safe, reliable power distribution.

Structural architecture of GaN HEMT devices

To transform a native D-mode GaN heterostructure into an E-HEMT, the conduction band at the gate region must be modulated upward to lift the potential well above the Fermi level (EF), depleting the 2DEG under equilibrium conditions. Manufacturers deploy several specialized gate architectures to achieve this.

  • Mechanism: A thin layer of p-type GaN (doped heavily with Magnesium, Mg) is grown directly atop the AlGaN barrier layer beneath the gate metal contact. The p-type doping creates a built-in potential that raises the electrostatic potential of the AlGaN barrier. This shifts the conduction band minimum at the AlGaN/GaN interface above the Fermi level, entirely depleting the 2DEG zone directly beneath the gate electrode at VGS = 0 V.
  • Operation: Applying a positive gate voltage higher than the built-in potential (VGS > VTH, typically around +1.5 V to +2.0 V) lowers the conduction band band-edge, pulling the potential well back beneath the Fermi level and restoring the 2DEG channel for forward conduction.
  • Mechanism: In a recessed gate HEMT, the highly conductive AlGaN barrier layer is physically thinned or entirely etched away in the region directly beneath the gate electrode via highly controlled dry etching.
  • Physics: The density of the polarization-induced 2DEG is monotonically dependent on the thickness of the AlGaN barrier. By reducing the barrier thickness below a critical threshold (typically < 5 nm), the polarization field drops below the level required to form a potential well deeper than the Fermi energy level. Consequently, the 2DEG disappears locally under the gate.
  • Insulated variation: Often, a thin atomic-layer-deposited (ALD) dielectric layer (such as Al₂O₃ or SiO₂) is inserted between the recessed trench and the gate metal, forming a Gate Injection Transistor (GIT) or a metal-insulator-semiconductor HEMT (MIS-HEMT) to suppress gate leakage current.
  • Mechanism: Negative fluorine ions (F⁻) are precisely injected into the AlGaN barrier layer underneath the gate region via ion implantation, followed by a localized thermal anneal.
  • Physics: The strongly electronegative fluorine ions act as fixed, immobile negative sheet charges within the lattice. These negative charges effectively neutralize the positive polarization charges at the AlGaN/GaN interface. This electrostatically pinches off the local 2DEG, shifting the threshold voltage into positive territory.
  • Mechanism: A thin layer of p-type GaN (doped heavily with Magnesium, Mg) is grown directly atop the AlGaN barrier layer beneath the gate metal contact. The p-type doping creates a built-in potential that raises the electrostatic potential of the AlGaN barrier. This shifts the conduction band minimum at the AlGaN/GaN interface above the Fermi level, entirely depleting the 2DEG zone directly beneath the gate electrode at VGS = 0 V.
  • Operation: Applying a positive gate voltage higher than the built-in potential (VGS > VTH, typically around +1.5 V to +2.0 V) lowers the conduction band band-edge, pulling the potential well back beneath the Fermi level and restoring the 2DEG channel for forward conduction.
  • Mechanism: In a recessed gate HEMT, the highly conductive AlGaN barrier layer is physically thinned or entirely etched away in the region directly beneath the gate electrode via highly controlled dry etching.
  • Physics: The density of the polarization-induced 2DEG is monotonically dependent on the thickness of the AlGaN barrier. By reducing the barrier thickness below a critical threshold (typically < 5 nm), the polarization field drops below the level required to form a potential well deeper than the Fermi energy level. Consequently, the 2DEG disappears locally under the gate.
  • Insulated variation: Often, a thin atomic-layer-deposited (ALD) dielectric layer (such as Al₂O₃ or SiO₂) is inserted between the recessed trench and the gate metal, forming a Gate Injection Transistor (GIT) or a metal-insulator-semiconductor HEMT (MIS-HEMT) to suppress gate leakage current.
  • Mechanism: Negative fluorine ions (F⁻) are precisely injected into the AlGaN barrier layer underneath the gate region via ion implantation, followed by a localized thermal anneal.
  • Physics: The strongly electronegative fluorine ions act as fixed, immobile negative sheet charges within the lattice. These negative charges effectively neutralize the positive polarization charges at the AlGaN/GaN interface. This electrostatically pinches off the local 2DEG, shifting the threshold voltage into positive territory.

Structural architecture of E-HEMTs

To understand the positioning of E-HEMT technology, it must be evaluated alongside Silicon (Si) MOSFETs.

Material property analysis

GaN's high critical field translates directly into a high breakdown voltage, letting GaN HEMTs block hundreds of volts in a thin layer. On a Si substrate, they also offer solid thermal performance for power devices.

Beyond the raw material numbers, three device-physics differences set E-HEMTs apart:

  • Si MOSFETs: Inversion-mode devices. When a positive gate voltage is applied, electrons are drawn to the oxide-semiconductor interface to form an inversion layer. These electrons experience intense surface roughness scattering at the amorphous oxide interface, lowering effective mobility.
  • GaN E-HEMT: Heterojunction-field device. Conduction occurs via the 2DEG, which is physically isolated in a pure, undoped monocrystalline layer. This eliminates impurity and surface scattering, providing a low specific on-resistance (RDS(on),sp).

Two primary Figures of Merit dictate power efficiency in high-frequency applications:

  • RDS(on) x Qg: Dictates total gate drive power loss.
  • RDS(on) x QOSS: Dictates output capacitance losses during hard-switching transitions.

For the RDS(on) × Qg figure of merit, lower is better — and here GaN has a decisive edge.

Because GaN E-HEMTs use tiny gate structures and high carrier mobility, their Qg and QOSS are up to an order of magnitude lower than silicon devices of equal current rating.

That is what allows switching frequencies into the megahertz (MHz) range with minimal parasitic losses.

  • Si MOSFETs: Feature a parasitic p-n body diode between the source and drain. During inductive commutation, this body diode conducts, injecting minority carriers. Turning off the diode requires clearing these carriers, resulting in a large Reverse Recovery Charge (Qrr) and severe switching loss.
  • GaN E-HEMT: The lateral HEMT architecture lacks a physical p-n body diode. It can conduct in reverse (third-quadrant operation) by naturally turning on the 2DEG channel when the drain voltage drops below the gate voltage. Because conduction is governed exclusively by majority carriers, the reverse recovery charge is zero (Qrr = 0). This eliminates a principal cause of electromagnetic interference (EMI) and power loss in bridge topologies.
  • Si MOSFETs: Inversion-mode devices. When a positive gate voltage is applied, electrons are drawn to the oxide-semiconductor interface to form an inversion layer. These electrons experience intense surface roughness scattering at the amorphous oxide interface, lowering effective mobility.
  • GaN E-HEMT: Heterojunction-field device. Conduction occurs via the 2DEG, which is physically isolated in a pure, undoped monocrystalline layer. This eliminates impurity and surface scattering, providing a low specific on-resistance (RDS(on),sp).

Two primary Figures of Merit dictate power efficiency in high-frequency applications:

  • RDS(on) x Qg: Dictates total gate drive power loss.
  • RDS(on) x QOSS: Dictates output capacitance losses during hard-switching transitions.

For the RDS(on) × Qg figure of merit, lower is better — and here GaN has a decisive edge.

Because GaN E-HEMTs use tiny gate structures and high carrier mobility, their Qg and QOSS are up to an order of magnitude lower than silicon devices of equal current rating.

That is what allows switching frequencies into the megahertz (MHz) range with minimal parasitic losses.

  • Si MOSFETs: Feature a parasitic p-n body diode between the source and drain. During inductive commutation, this body diode conducts, injecting minority carriers. Turning off the diode requires clearing these carriers, resulting in a large Reverse Recovery Charge (Qrr) and severe switching loss.
  • GaN E-HEMT: The lateral HEMT architecture lacks a physical p-n body diode. It can conduct in reverse (third-quadrant operation) by naturally turning on the 2DEG channel when the drain voltage drops below the gate voltage. Because conduction is governed exclusively by majority carriers, the reverse recovery charge is zero (Qrr = 0). This eliminates a principal cause of electromagnetic interference (EMI) and power loss in bridge topologies.

Applications and real-world impact

The physical characteristics of E-HEMTs render them highly effective for applications requiring compact size, high power density, and high efficiency.
 
  • High-density power supplies and USB-C chargers
    The rapid commercialization of GaN E-HEMTs is highly visible in consumer electronics. By migrating from Si MOSFETs to GaN E-HEMTs running at switching frequencies above 500 kHz, the volume of passive magnetic components (transformers and inductors) and bulk capacitors can be cut by more than half. This transition facilitates ultra-compact 65  W - 240W chargers operating at high efficiency levels.
  • Server and data center power delivery
    Modern data centers running AI workloads consume megawatts of power. Regulations dictate strict efficiency standards (e.g., Titanium efficiency, requiring >96% efficiency). E-HEMTs deployed in the power factor correction (PFC) stage and LLC resonant DC-DC converters in power supplies, in DC-DC conversion in intermediate bus converters (IBC) and in the battery backup units (BBU) reduce switching and conduction losses, saving gigawatt-hours across large IT factories.
  • Automotive on-board chargers (OBC)
    Within Electric Vehicles (EVs), GaN E-HEMTs serve in 400 V system architectures for on-board chargers and DC-DC converters. Their low switching losses enable higher frequency operation, reducing weight and shrinking the footprint of the cooling systems and passive filters, which extends overall driving range.

Conclusion

Enhancement-mode High Electron Mobility Transistors (eHEMTs) represent a significant advance in power semiconductor evolution. By addressing the "normally-on" characteristics of native Gallium Nitride devices through advanced p-GaN and gate-recession architectures, E-HEMTs deliver the benefits of wide-bandgap physics directly to commercial power systems.

With zero reverse recovery charge, low specific on-resistance, and high switching frequencies, E-HEMTs offer increased performance in efficiency and power density for applications in consumer electronics, data center infrastructure, renewable energy systems and automotive power electronics.

Recommended documents

Explore more about GaN

Frequently asked questions about GaN HEMT structure

Enhancement-mode High Electron Mobility Transistor.

A standard HEMT is typically "normally-on" (depletion-mode), whereas an eHEMT is "normally-off" at zero gate bias.

It means no current flows through the device when the gate voltage is zero (VG = 0).

The Two-Dimensional Electron Gas, a thin, highly concentrated sheet of free-moving electrons trapped at the heterojunction interface.

Electrons are tightly confined vertically but can move freely along the horizontal plane of the interface.

Electrons travel through an undoped channel, meaning they do not slow down by crashing into ionized dopant impurities.

MOSFETs use a continuous doped channel, while eHEMTs separate the electron generation region from the undoped transport channel to prevent electron collisions.

They feature exceptionally low on-resistance (RDS(on)), significantly reducing power lost as waste heat during conduction.

Fast switching allows power converters to use much smaller passive components, like inductors and capacitors, downsizing the overall device footprint.