Gallium nitride (GaN) power devices are redefining the limits of switching converters by combining wide bandgap physics with lateral HEMT structures optimized for fast, low-loss operation. This article describes GaN as the natural successor to silicon MOSFETs in the 100–650 V class, showing how material figures of merit directly translate into lower on-resistance, higher switching frequency, and much higher power density at competitive cost.
Silicon power MOSFETs have driven the evolution of switch-mode power conversion since the late 1970s, replacing bipolar transistors, thanks to majority-carrier operation, ruggedness, and ease of drive. For decades, continuous structural improvements—cell pitch, trench, and superjunction—pushed RDS(on) down while keeping breakdown capability and manufacturability. However, silicon is now essentially at its theoretical limit for unipolar devices in the 100–600 V range.
The bandgap of a semiconductor is related to the strength of the chemical bonds between the atoms in the lattice. Stronger bonds make it more difficult for electrons to transition between atomic sites. This leads to several important consequences, including lower intrinsic leakage currents and the ability to operate at higher temperatures. Based on scientific data, both GaN and silicon carbide (SiC) exhibit significantly wider bandgaps than silicon.
The theoretical specific on-resistance RDS(on) of a majority-carrier device is constrained by the material’s critical electric field, permittivity, and mobility. For a one square millimeter device area, the drift region controls the trade-off between breakdown voltage and conduction loss.
The approximate breakdown voltage can be written as:
VBR = ½ wdrift Ecrit
wdrift is the drift region thickness and Ecrit is the material’s critical electric field.
The number of electrons available in the drift region between two terminals is calibrated by a simplified, one-dimensional Poisson relation:
q ND = ε0 εr Ecrit wdrift
Where q is the electron charge, ND the doping concentration (or equivalent electron density), ε0 the vacuum permittivity, and εr the relative permittivity.
Combining this with the usual expression for resistance of the drift region (for area = 1 mm²) results in the below equation:
RDS(on) = w_drift / (q μn ND)
It yields the well-known relation between specific on-resistance and breakdown voltage:
RDS(on) = 4 VBR² / (ε0 εr μn Ecrit³)
This equation shows the dominant role of the critical field: RDS(on) scales as VBR² but inversely as Ecrit³. A material that can withstand a much higher electric field and maintain good mobility will deliver orders of magnitude lower specific resistance at the same breakdown voltage.
Figure 1 See theoretical on-resistance for a one square millimeter device versus blocking voltage capability for Si-, SiC-, and GaN-based power devices. Source: Efficient Power Conversion (EPC)
In Figure 1, silicon, 4H-SiC, and GaN theoretical limits diverge dramatically as breakdown voltage increases. At 600 V, GaN’s theoretical specific RDS(on) is roughly two orders of magnitude lower than silicon, and significantly better than SiC, highlighting why GaN is particularly attractive in the 100–650 V class.
Crystal structure and 2DEG formation
GaN’s crystal structure is a key enabler for these performance gains. Crystalline GaN adopts a wurtzite hexagonal structure, while 4H-SiC also has a hexagonal lattice but with different stacking. Both materials are mechanically robust, chemically stable, and tolerant of high operating temperatures, but GaN additionally exhibits strong piezoelectric effects due to the asymmetry of the wurtzite lattice. This effect brings GaN to achieve very high conductivity compared with either silicon or silicon carbide.
When a thin layer of AlGaN is grown on top of GaN, lattice mismatch and spontaneous polarization create strain at the interface. This strain, combined with the intrinsic polarization of the wurtzite structure, generates a strong internal electric field. To compensate this, a two-dimensional electron gas (2DEG) forms at the AlGaN/GaN interface with sheet carrier density on the order of 1013 cm-2 and electron mobility significantly higher than bulk GaN (up to 1500–2000 cm²/V·s versus ~1000 cm²/V·s). This ultra-thin, highly conductive channel is at the heart of the GaN HEMT.
Figure 2 Simplified cross section of a GaN/AlGaN heterostructure shows the formation of a 2DEG created due to the strain-induced polarization at the interface between the two materials. Source: Efficient Power Conversion (EPC)
From an electrical standpoint, this 2DEG behaves like a very low-resistance sheet: the product of carrier density and mobility (ns μn) is much higher than in a doped silicon drift region, while the conduction path is extremely short and lateral. This combination is what allows GaN devices to reach very low RDS(on) for a given chip area and breakdown rating. In addition, the wide bandgap (3.39 eV vs. 1.12 eV for silicon) yields much lower intrinsic leakage and supports higher operating temperatures.
GaN, SiC, and silicon: Material figures of merit
Let’s compare key material parameters for Si, GaN, and 4H-SiC: bandgap, critical field, electron mobility, permittivity, and thermal conductivity. Both SiC and GaN have wider bandgap and much higher critical fields than silicon. In addition to its wide bandgap, GaN exhibits significantly higher electron mobility than both silicon and silicon carbide, enabling faster carrier transport, higher current density, and superior high-frequency performance.
Moreover, GaN’s Ecrit is about 3.3 MV/cm, compared to 0.23 MV/cm for silicon, allowing a much thinner drift region for the same breakdown voltage. The previous RDS(on)–VBR equation directly shows that increasing Ecrit reduces the specific on-resistance by orders of magnitude.
Silicon carbide has even better thermal conductivity than GaN, which is an advantage for very high-power densities and high-voltage systems (>1 kV). However, in the mid-voltage range up to a few hundred volts, GaN’s combination of lateral HEMT structure, very high Ecrit, and 2DEG conduction gives it a superior theoretical figure of merit compared to both silicon and SiC. This positions GaN as the primary technology for replacing MOSFETs in most 40–650 V applications.
From depletion-mode to enhancement-mode GaN HEMTs
The native GaN HEMT is a depletion-mode device: at zero gate bias the 2DEG under the AlGaN barrier provides a low-resistance channel between source and drain, and a negative gate voltage is required to pinch it off. Source and drain contacts reach the 2DEG through the AlGaN layer, while the gate sits on top and modulates the channel by depleting or restoring that electron gas.
This normally-on behavior is acceptable in RF power amplifiers, but it’s problematic in switching converters, where a device that conducts at VGS = 0 V can cause shoot-through during startup or fault conditions.
For power conversion, enhancement-mode operation (normally-off) is therefore essential. With an enhancement-mode HEMT, the 2DEG is suppressed at zero gate bias and re-formed only when a positive gate voltage is applied, making its behavior similar to a power MOSFET.
Several device architectures implement this transition from depletion- to enhancement-mode:
- In recessed-gate structures, the AlGaN barrier is locally thinned beneath the gate. Reducing the barrier thickness lowers the internal polarization-induced field to the point where the 2DEG vanishes at VGS = 0 V. A positive gate voltage then recreates the channel and allows current to flow.
- Fluorine-implanted gates introduce negative charge into the AlGaN barrier by ion implantation. The fixed negative charge depletes the 2DEG under the gate at zero bias, shifting the threshold into the positive range. Applying a positive gate voltage compensates this charge and restores conduction.
- In p‑GaN gate HEMTs, a thin p-type GaN layer is grown on top of the AlGaN barrier. The positive charge in this p-GaN region creates a built‑in potential that overcomes the polarization field and depletes the 2DEG at zero gate bias. When a positive voltage is applied to the gate, electrons are again attracted to the interface and the 2DEG reforms, turning the device on.
- Hybrid solutions combine a low-voltage enhancement-mode silicon MOSFET with a depletion-mode GaN HEMT in series. In the cascode configuration, the MOSFET gate becomes the external control terminal. When the MOSFET turns on, the GaN gate is effectively driven to a voltage that enables the HEMT; when the MOSFET turns off, the GaN gate is driven negative, and the composite behaves as a normally‑off device.
All these approaches pursue the same goal: eliminate conduction at VGS = 0 V using an architecture that remains compatible with practical gate‑drive levels and offers stable threshold voltage. In practice, p‑GaN gate devices have become the most widely used in commercial power conversion, while cascode hybrids are attractive at higher voltages where the on‑resistance of the silicon MOSFET adds only a small penalty to the GaN device.
Figure 3 An enhancement-mode (e-mode) device depletes the 2DEG with zero volts on the gate (a). By applying a positive voltage to the gate, the electrons are attracted to the surface, re-establishing the 2DEG (b). Source: Efficient Power Conversion (EPC)
The second and final part of this article series about GaN technology fundamentals will explain hybrid structures and RDS(on) penalty as well as vertical GaNs and how to build a GaN HEMT transistor.
Maurizio Di Paolo Emilio is director of global marketing communications at Efficient Power Conversion (EPC), where he manages worldwide initiatives to showcase the company’s GaN innovations. He is a prolific technical author of books on GaN, SiC, energy harvesting and data acquisition and control systems, and has extensive experience as editor of technical publications for power electronics, wide bandgap semiconductors, and embedded systems.
Editor’s Note:
The content in this article uses references and technical data from the book GaN Power Devices for Efficient Power Conversion (Fourth Edition) authored by Alex Lidow, Michael de Rooij, John Glaser, Alejandro Pozo Arribas, Shengke Zhang, Marco Palma, David Reusch, Johan Strydom.
Related Content