First-generation gallium nitride (GaN) technology refers primarily to the earliest commercial and near-commercial high-electron-mobility transistors (HEMTs) and related devices that emerged in the mid-2000s, building on foundational material and device breakthroughs from the 1960s–1990s. These early devices leveraged GaN’s wide-bandgap properties for high-power, high-frequency RF applications and later power switching, typically in depletion-mode (normally-on) configurations grown on mismatched substrates such as silicon carbide (SiC) or silicon.
Historical Context and Foundations
GaN research began long before practical devices. Polycrystalline GaN was synthesized in the 1930s. Single-crystal GaN films on sapphire via hydride vapor phase epitaxy appeared in 1969 (Maruska and Tietjen). Early metal-insulator-semiconductor LEDs followed in the early 1970s, but progress stalled due to poor crystal quality, high defect densities (often >10⁹ cm⁻²), and the inability to achieve reliable p-type doping.
Breakthroughs in the late 1980s and early 1990s—low-temperature buffer layers (Amano/Akasaki), thermal annealing for p-type activation (Nakamura), and high-quality InGaN growth—enabled efficient blue LEDs (Nobel Prize in Physics 2014). These advances in metal-organic chemical vapor deposition (MOCVD/OMVPE) epitaxial growth transferred directly to electronic devices.
Key electronic milestones:
- AlGaN/GaN heterojunction reported around 1991.
- First AlGaN/GaN HEMTs demonstrated in 1993, showing high channel electron mobility from polarization-induced two-dimensional electron gas (2DEG).
- Commercial RF GaN HEMTs introduced by Eudyna Device Inc. (Japan) in 2004 as depletion-mode devices, primarily for RF power amplifiers.
- Nitronex followed in 2005 with GaN-on-silicon depletion-mode RF HEMTs.
- Early power-focused work included high-voltage demonstrations on SiC, with enhancement-mode devices emerging later (e.g., EPC in 2009).
First-generation technology is thus best understood as the initial commercial RF HEMT platform of the mid-to-late 2000s, often with optically defined gates around 0.5–0.9 µm and SiN passivation, optimized for high power density on SiC or early silicon substrates.
Material Properties Enabling First-Generation Devices
GaN is a III-V wide-bandgap semiconductor (direct bandgap ~3.4 eV) with a wurtzite crystal structure. Compared with silicon:
- Critical electric field is roughly 10× higher (~3 MV/cm vs. ~0.3 MV/cm), supporting higher breakdown voltages.
- Electron saturation velocity is higher, enabling faster switching and higher-frequency operation.
- Thermal conductivity is good (though substrate-dependent), and the material tolerates high temperatures.
- Polarization (spontaneous + piezoelectric) at AlGaN/GaN interfaces creates a high-density 2DEG without intentional doping of the channel.
These properties yield a Baliga figure of merit orders of magnitude better than silicon, translating into higher power density, efficiency, and frequency capability. Early devices still suffered from high defect densities due to lattice and thermal mismatch with foreign substrates (sapphire, SiC, Si), which limited reliability and yield until buffer-layer engineering matured.
Device Structure and Physics of First-Generation GaN HEMTs
First-generation devices are lateral AlGaN/GaN HEMTs. A typical epitaxial stack (grown by MOCVD on SiC or Si) includes:
- Substrate (semi-insulating SiC preferred for RF to minimize losses; silicon for cost).
- Nucleation/buffer layers to manage mismatch and isolate the channel.
- GaN channel/buffer.
- AlGaN barrier (typically 15–30 nm, Al fraction ~20–30%).
- Optional thin GaN cap.
- Surface passivation (SiN) to reduce current collapse and trapping.
At the AlGaN/GaN interface, polarization discontinuity induces a high-density 2DEG (sheet densities often >10¹³ cm⁻²) with high mobility. Current flows laterally between source and drain through this 2DEG. The gate (Schottky metal in early designs) modulates the 2DEG density.
Early commercial devices were predominantly depletion-mode (normally-on): a negative gate-source voltage is required to deplete the 2DEG and turn the device off. Threshold voltages were typically around –3 to –5 V. Field plates (extensions of the gate metallization) were introduced relatively early to manage peak electric fields in the gate-drain region, improving breakdown voltage and reducing trapping effects.
Fabrication often used gold-based metallization for ohmic contacts and gates, with optical lithography defining gates in the 0.5–0.9 µm range for early RF processes.
Performance Characteristics and Early Applications
First-generation RF GaN HEMTs delivered power densities far exceeding GaAs or silicon LDMOS (often several W/mm), high operating voltages (28–50 V or more), and good efficiency. Continuous-wave power amplifiers reaching tens of watts at microwave frequencies were demonstrated early. Applications centered on wireless infrastructure (base-station power amplifiers), radar, and satellite communications, where high power density allowed smaller, more efficient systems.
In power electronics, early lateral devices showed promise for high-voltage switching, but commercial enhancement-mode (normally-off) devices suitable as silicon MOSFET replacements arrived later. Cascode configurations (depletion-mode GaN paired with a low-voltage silicon MOSFET) were sometimes used to achieve normally-off behavior while retaining GaN’s switching advantages.
Limitations of First-Generation Technology
Several constraints defined the generation and motivated subsequent improvements:
- Normally-on operation complicated circuit design and safety (fail-to-on behavior).
- Current collapse / dynamic on-resistance from surface and buffer trapping, mitigated but not eliminated by passivation and field plates.
- Substrate mismatch leading to high dislocation densities, impacting reliability, leakage, and long-term performance.
- Cost and manufacturability — SiC substrates were expensive; early silicon processes faced quality challenges; gold-based metallization was not CMOS-compatible.
- Voltage and frequency scaling limited by lateral architecture and gate length; thermal management also constrained power handling.
- Limited integration and packaging maturity compared with silicon.
These issues drove the shift toward enhancement-mode designs (p-GaN gates, recessed gates, etc.), CMOS-compatible gold-free processes, improved epitaxial quality, advanced field-plate and packaging techniques, and eventually vertical GaN architectures for higher voltages.
Evolution Beyond the First Generation
Subsequent generations introduced enhancement-mode devices optimized for power conversion (chargers, data centers, automotive), higher-voltage ratings (650 V and above becoming standard for many applications), better reliability, monolithic integration (GaN power ICs), and larger-wafer manufacturing (200 mm and now 300 mm efforts). Vertical GaN-on-GaN structures represent a more fundamental architectural shift, allowing current to flow through the bulk rather than laterally for potentially higher voltage and current density.
In optoelectronics, “first-generation” nitride LEDs referred to early double-heterostructure and single-quantum-well InGaN devices of the mid-1990s that established blue/green emission before multi-quantum-well and advanced light-extraction structures.
Summary Table: Key Attributes of First-Generation GaN Technology
| Aspect | Typical First-Generation Characteristics |
|---|---|
| Device type | Lateral AlGaN/GaN HEMT (depletion-mode dominant) |
| Substrate | SiC (RF preferred) or early Si |
| Gate | Schottky, optically defined (~0.5–0.9 µm) |
| Operation | Normally-on; negative Vgs to turn off |
| Key enablers | Polarization 2DEG, SiN passivation, early field plates |
| Strengths | High power density, high frequency, high voltage capability |
| Main limitations | Normally-on, trapping, defect density, cost |
| Primary applications | RF power amplifiers (base stations, radar) |
| Approximate timeline | Commercial RF introduction ~2004–2005 onward |
First-generation GaN technology proved that wide-bandgap nitride semiconductors could deliver transformative performance in high-power RF electronics. It established the epitaxial, process, and device-physics foundations that later generations refined into the efficient, compact power and RF solutions widely deployed today in consumer chargers, data-center power supplies, 5G infrastructure, and emerging electric-vehicle and renewable-energy systems. The progression from imperfect crystals considered “useless” to a mature commercial platform remains one of the clearer illustrations of how materials science breakthroughs, once unlocked, cascade into system-level advantages across multiple industries.
1) Historical Context and Foundations
Early Material Synthesis and the Long Struggle for Crystal Quality
Gallium itself was predicted by Dmitri Mendeleev in 1871 and isolated in 1875. The first synthesis of polycrystalline GaN occurred in the early 1930s by reacting gallium metal with ammonia at high temperatures (approximately 900–1000 °C). Early efforts produced only powder or polycrystalline material of limited usefulness for electronic devices.
A decisive step came in 1969 when H. P. Maruska and J. J. Tietjen at RCA Laboratories grew the first single-crystal GaN films on sapphire substrates using hydride vapor phase epitaxy (HVPE). This demonstrated that epitaxial GaN was possible, yet the films remained highly defective. Lattice mismatch between GaN and sapphire (roughly 16 %) combined with large differences in thermal expansion coefficients produced dislocation densities often exceeding 10⁹–10¹⁰ cm⁻². Such high defect densities made controlled doping extremely difficult and rendered the material nearly unusable for reliable p–n junctions or transistors.
Throughout the 1970s, researchers explored metal-insulator-semiconductor (MIS) structures and achieved rudimentary blue-violet electroluminescence, but efficiency remained too low for commercial application. GaN was widely regarded as an intractable material—interesting in principle because of its wide direct bandgap (~3.4 eV) but practically unworkable.
Breakthroughs in Epitaxial Growth and p-Type Doping
The transformation began in the mid-to-late 1980s in Japan. Hiroshi Amano and Isamu Akasaki introduced low-temperature-deposited buffer layers (initially AlN, later GaN). Growing a thin, disordered buffer at reduced temperature before raising the temperature for the main GaN layer dramatically improved surface morphology and reduced dislocation density. This technique, refined by Shuji Nakamura at Nichia Chemical, became a cornerstone of modern GaN epitaxy.
Equally critical was the achievement of p-type conductivity. Magnesium doping had been attempted for years, yet the material remained highly resistive. In 1989, Amano and Akasaki demonstrated local p-type conversion using low-energy electron-beam irradiation (LEEBI). Nakamura later clarified the underlying mechanism: hydrogen atoms from the ammonia ambient form Mg–H complexes that passivate the acceptors. Thermal annealing in a hydrogen-free ambient above approximately 700 °C releases the hydrogen and activates the magnesium acceptors. This thermal-annealing process proved scalable and became the industrial standard.
With both high-quality epitaxial layers and reliable p-type doping available, efficient blue light-emitting diodes based on InGaN active regions followed rapidly in the early 1990s. The 1993–1995 demonstrations of bright blue and multi-color quantum-well LEDs by Nakamura and colleagues completed the primary-color set and launched the solid-state lighting revolution. These optoelectronic advances simultaneously matured the metal-organic chemical vapor deposition (MOCVD, also called OMVPE) growth technology that would later be essential for electronic devices.
Emergence of the AlGaN/GaN Heterostructure and the HEMT Concept
While LED research dominated attention, electronic device work proceeded in parallel. The high-electron-mobility transistor (HEMT) concept itself originated with Takashi Mimura’s 1979 invention of the AlGaAs/GaAs HEMT. The idea of a polarization-induced two-dimensional electron gas (2DEG) at a heterojunction without intentional channel doping proved especially powerful when applied to the nitride system.
In 1991, researchers reported the AlGaN/GaN heterojunction. By 1993 the first AlGaN/GaN HEMTs had been demonstrated. Unlike conventional doped-channel field-effect transistors, the strong spontaneous and piezoelectric polarization discontinuity at the AlGaN/GaN interface spontaneously generates a high-density 2DEG (typically >10¹³ cm⁻²) with high electron mobility. This polarization doping is unique to the III-nitrides and became a defining advantage of GaN HEMTs.
Early devices were grown primarily on sapphire or silicon carbide. SiC offered better thermal conductivity and closer lattice matching, making it preferred for high-power RF work. Silicon substrates, although more defective, promised lower cost and compatibility with larger-diameter wafers.
From Laboratory Demonstrations to First Commercial Devices
By the early 2000s, epitaxial quality, surface passivation (especially silicon nitride), and basic field-plate structures had advanced sufficiently for practical RF power transistors. In 2004, Eudyna Device Inc. in Japan introduced the first commercial depletion-mode GaN HEMTs intended for radio-frequency applications. These early devices typically featured optically defined gates in the 0.5–0.9 µm range, operated at supply voltages of 28–50 V, and delivered power densities several times higher than contemporary GaAs or silicon LDMOS transistors. Nitronex followed in 2005 with GaN-on-silicon RF HEMTs, further lowering the barrier to entry.
These mid-2000s commercial products constitute the practical beginning of “first-generation” GaN technology in the electronics sense. They still relied on depletion-mode (normally-on) operation, gold-based metallization, and relatively immature packaging and reliability qualification. Nevertheless, they validated the material system for high-power, high-frequency amplification and established the process baseline upon which later enhancement-mode power devices, CMOS-compatible flows, and higher-voltage architectures were built.
Why These Foundations Matter
The historical path of GaN illustrates a classic pattern in semiconductor technology: a theoretically attractive material is held back for decades by crystal-growth and doping problems; once those problems are solved—often driven by a high-value application such as blue LEDs—the same epitaxial and process knowledge rapidly transfers to electronic devices. The polarization-induced 2DEG, wide bandgap, high critical electric field, and high saturation velocity that make modern GaN power and RF devices superior to silicon were already present in the material system of the 1990s. What first-generation technology demonstrated was that these intrinsic advantages could be realized in manufacturable transistors with useful reliability and performance.
Subsequent generations refined threshold-voltage control, reduced trapping, improved substrate quality, eliminated gold metallization, and moved toward larger wafers and vertical architectures. Yet every later advance rests on the crystal-growth techniques, doping methods, and heterostructure understanding established between 1969 and the mid-2000s. Understanding this historical and scientific foundation remains essential for evaluating both the capabilities and the remaining limitations of GaN technology today.
2) Material Properties Enabling First-Generation Devices
First-generation devices succeeded because GaN combined a set of properties that simultaneously supported high voltage, high frequency, high power density, and efficient carrier transport in a heterostructure format that did not require intentional channel doping.
Wide Bandgap and High Critical Electric Field
GaN possesses a direct bandgap of approximately 3.4 eV at room temperature. This is substantially larger than silicon’s 1.12 eV and comparable to or slightly higher than 4H-SiC (~3.2–3.3 eV). The wide bandgap produces two immediate benefits:
- Extremely low intrinsic carrier concentration, which allows operation at elevated temperatures with reduced leakage.
- A high critical (breakdown) electric field, typically reported in the range of 3.0–3.5 MV/cm (roughly 10–12 times higher than silicon’s ~0.3 MV/cm).
The elevated critical field means that a thinner drift or access region can support a given blocking voltage. In first-generation lateral HEMTs this translated into the ability to operate at drain voltages of 28–50 V (and higher in research devices) while still achieving useful current densities. Higher voltage capability directly improves power density and efficiency in RF amplifiers.
Electron Transport Properties
Bulk GaN electron mobility is typically 1000–1400 cm²/V·s under low doping, competitive with silicon. More important for HEMTs is the two-dimensional electron gas (2DEG) that forms at the AlGaN/GaN interface. In the 2DEG, room-temperature mobility commonly exceeds 1500–2000 cm²/V·s in well-grown structures of the first-generation era, with sheet densities routinely above 1 × 10¹³ cm⁻².
Electron saturation velocity in GaN reaches approximately 2.0–2.5 × 10⁷ cm/s, higher than both silicon (~1.0 × 10⁷ cm/s) and 4H-SiC (~2.0 × 10⁷ cm/s). High saturation velocity, combined with the high sheet density of the 2DEG, enables high current density and fast switching, which are essential for microwave power amplification.
Polarization Effects: The Defining Advantage
GaN crystallizes in the wurtzite structure, which lacks inversion symmetry along the c-axis. This produces strong spontaneous polarization. When a thin AlGaN barrier is grown on GaN, the lattice mismatch introduces additional piezoelectric polarization. The net polarization discontinuity at the interface creates a fixed positive sheet charge that induces a high-density 2DEG in the GaN channel without any intentional doping of the channel itself.
This polarization doping is unique among the commercially mature semiconductors of the early 2000s. It yields:
- High channel conductivity with low scattering (because the electrons are spatially separated from ionized impurities).
- Natural normally-on (depletion-mode) behavior, which matched the early commercial RF target applications.
- Simplified epitaxial structures relative to doped-channel devices.
First-generation HEMTs exploited this polarization-induced 2DEG as the primary conduction path between source and drain. Gate control was achieved by a Schottky metal contact that modulated the 2DEG density under the gate.
Thermal and Dielectric Properties
GaN’s thermal conductivity is typically cited in the range of 1.3–2.5 W/cm·K (bulk values; actual device performance is strongly substrate-dependent). While lower than 4H-SiC (~3.7–4.9 W/cm·K), it is still superior to many III-V materials and adequate when the device is grown on high-thermal-conductivity SiC substrates—the preferred platform for early high-power RF transistors.
The relative dielectric constant of GaN is approximately 8.9–9.5, lower than silicon’s ~11.8. Lower permittivity reduces parasitic capacitances, contributing to higher operating frequencies and lower switching losses.
Figures of Merit Quantifying the Advantage
Two widely used figures of merit capture the system-level impact of these properties:
- Baliga figure of merit (BFOM) – relevant for power-switching performance: proportional to εμE_c³. GaN’s BFOM is hundreds to roughly 1000 times higher than silicon’s, reflecting the combination of high critical field and respectable mobility.
- Johnson figure of merit (JFOM) – relevant for high-frequency power: proportional to (v_sat E_c)². GaN again substantially outperforms silicon and is competitive with or superior to SiC for many RF metrics.
These theoretical advantages were already evident in the mid-1990s laboratory devices and became commercially relevant once epitaxial quality and surface passivation reached levels that allowed reliable RF operation.
Comparative Summary of Key Material Properties
| Property | Silicon | 4H-SiC | GaN (typical) |
|---|---|---|---|
| Bandgap (eV) | 1.12 | 3.2–3.3 | 3.4 |
| Critical electric field (MV/cm) | ~0.3 | 2.2–2.5 | 3.0–3.5 |
| Electron mobility (cm²/V·s) | ~1400 | ~800–1000 | ~1000–1400 (bulk); higher in 2DEG |
| Saturation velocity (×10⁷ cm/s) | ~1.0 | ~2.0 | ~2.0–2.5 |
| Thermal conductivity (W/cm·K) | ~1.5 | ~3.7–4.9 | ~1.3–2.5 |
| Relative permittivity | ~11.8 | ~9.7 | ~8.9–9.5 |
| Polarization-induced 2DEG | No | Limited | Strong (key enabler) |
Values are approximate and vary with doping, temperature, crystal quality, and measurement method; they represent consensus ranges used in device literature of the first-generation period.
How These Properties Shaped First-Generation Devices
In the commercial RF HEMTs introduced around 2004–2005 (primarily depletion-mode devices on SiC or early silicon substrates), the material advantages manifested as:
- Power densities several times higher than contemporary GaAs or silicon LDMOS transistors.
- Ability to operate at higher supply voltages with acceptable efficiency.
- High gain and efficiency at microwave frequencies relevant to cellular base stations and radar.
- Compact die size for a given output power, reducing parasitics and enabling smaller system form factors.
Limitations of the first generation (normally-on operation, trapping effects, substrate cost, and defect-related reliability issues) were engineering and process challenges rather than fundamental material shortcomings. The intrinsic properties listed above remained the foundation upon which later enhancement-mode power devices, improved passivation, field-plate designs, and eventual vertical architectures were built.
In short, first-generation GaN technology succeeded because the material itself offered a rare combination of high breakdown strength, excellent high-field transport, and a natural high-density channel formed by polarization—properties that silicon could not match and that silicon carbide could only partially equal in the RF domain. These characteristics turned an experimentally difficult semiconductor into a practical high-performance technology once crystal growth and device processing matured.
3) Device Structure and Physics of First-Generation GaN HEMTs
Overall Device Architecture
First-generation GaN HEMTs are lateral devices. Current flows parallel to the surface of the epitaxial layers between source and drain electrodes, modulated by a gate electrode placed between them. This contrasts with later vertical GaN structures in which current flows perpendicular to the surface through a bulk drift region.
A typical early commercial device consisted of an epitaxial heterostructure grown by metal-organic chemical vapor deposition (MOCVD) on a foreign substrate, followed by device isolation, ohmic contacts, a Schottky gate, surface passivation, and interconnect metallization. Gate lengths in the initial commercial processes were commonly in the 0.5–0.9 µm range, defined optically rather than by electron-beam lithography.
Epitaxial Layer Stack
The active device is formed in a carefully engineered epitaxial stack. A representative first-generation structure (grown on semi-insulating SiC for RF or on silicon for cost-sensitive applications) included:
- Substrate — Semi-insulating 4H-SiC was preferred for high-power RF transistors because of its high thermal conductivity and relatively good lattice match. Silicon (111) substrates were used in early commercial GaN-on-Si processes (e.g., Nitronex) to reduce cost, despite greater lattice and thermal mismatch.
- Nucleation / transition layers — A thin low-temperature AlN or GaN nucleation layer, often followed by graded or superlattice (Al,Ga)N transition layers, managed strain and prevented cracking or excessive defect propagation.
- GaN buffer / channel layer — A thick (typically 1–5 µm), highly resistive GaN buffer provided electrical isolation from the substrate and served as the channel region. Intentional carbon doping or iron compensation was sometimes used to raise resistivity and suppress buffer leakage or parallel conduction. An undoped GaN channel region sat above the compensated buffer.
- AlGaN barrier — A thin (typically 15–30 nm) AlₓGa₁₋ₓN layer with aluminum mole fraction usually between 0.20 and 0.30. This barrier is the source of the polarization-induced two-dimensional electron gas (2DEG).
- Optional thin GaN cap — A few-nanometer GaN or n⁺-GaN cap was sometimes added to protect the barrier surface, improve ohmic contacts, or modify surface potential.
No intentional doping of the channel was required. The high-density conduction path formed spontaneously from polarization effects.
Physics of the Polarization-Induced 2DEG
The defining physical mechanism of GaN HEMTs is the formation of a two-dimensional electron gas at the AlGaN/GaN heterointerface without chemical doping of the channel.
GaN and AlGaN crystallize in the non-centrosymmetric wurtzite structure. Along the c-axis this produces a strong spontaneous polarization. When a thin, pseudomorphic AlGaN barrier is grown on GaN, the lattice mismatch generates additional piezoelectric polarization. The net polarization discontinuity creates a fixed positive sheet charge at the AlGaN/GaN interface (on the order of 10¹³ cm⁻² for typical aluminum fractions).
This fixed charge induces free electrons that accumulate in a narrow potential well on the GaN side of the interface, forming the 2DEG. Typical sheet densities in first-generation structures exceeded 1 × 10¹³ cm⁻², with room-temperature mobilities often in the 1200–2000 cm²/V·s range. Because the electrons reside in the undoped GaN and are spatially separated from ionized impurities, scattering is reduced and mobility remains high.
Surface donor-like states on the AlGaN (or GaN cap) surface supply the electrons that populate the 2DEG. The surface potential is approximately pinned near the Schottky barrier height, allowing the polarization charge to be largely transferred into the mobile 2DEG.
Device Terminals and Operation
- Source and drain contacts — Ohmic contacts were formed by depositing multilayer metal stacks (commonly Ti/Al-based, sometimes with Au or other overlays) followed by high-temperature annealing. These contacts reach the 2DEG either by alloying through the barrier or by recessing it.
- Gate electrode — A Schottky metal (often Ni/Au or Pt-based stacks) formed a rectifying contact on the AlGaN barrier. Applying a negative gate-source voltage widens the depletion region under the gate, reducing the 2DEG density and eventually pinching off the channel. Threshold voltages in early depletion-mode devices were typically –2 V to –5 V.
- Surface passivation — Silicon nitride (SiNₓ) deposited by plasma-enhanced chemical vapor deposition was critical. It reduced surface-state-related current collapse (dispersion) and improved reliability by stabilizing the surface potential.
When a positive drain-source voltage is applied, electrons in the 2DEG drift from source to drain under the lateral electric field. The gate modulates the available charge and therefore the current. At high drain bias the electric field peaks near the drain-side edge of the gate; early field-plate extensions of the gate metallization were introduced to redistribute this peak field, raise breakdown voltage, and further suppress trapping.
Key Physical Behaviors and Early Limitations
The lateral geometry concentrates the highest electric fields near the surface and the gate edge. This made devices sensitive to surface trapping, which manifested as current collapse or dynamic on-resistance increase under RF stress. Passivation and field plates mitigated but did not eliminate these effects in the first generation.
Because the 2DEG exists at zero gate bias, the devices were normally-on. Circuit designers therefore needed negative gate bias supplies or cascode arrangements with silicon MOSFETs to achieve fail-safe normally-off behavior.
Buffer leakage, dislocation-related trapping, and thermal management (especially on silicon substrates) also constrained performance and reliability. Nevertheless, the combination of high 2DEG density, high mobility, high critical field, and high saturation velocity delivered power densities several times higher than contemporary GaAs or silicon LDMOS transistors, establishing the commercial viability of the technology.
Summary of Structural and Physical Features
| Element | Typical First-Generation Implementation | Primary Role |
|---|---|---|
| Substrate | Semi-insulating SiC or Si(111) | Mechanical support, thermal path, isolation |
| Buffer / channel | 1–5 µm resistive GaN | Isolation + 2DEG host |
| Barrier | 15–30 nm AlₓGa₁₋ₓN (x ≈ 0.20–0.30) | Polarization charge source |
| Cap (optional) | Few nm GaN | Surface protection, contact improvement |
| Conduction path | Polarization-induced 2DEG | High-density, high-mobility channel |
| Gate | Schottky metal, optically defined (~0.5–0.9 µm) | Channel modulation |
| Passivation | SiNₓ | Suppress surface trapping |
| Operation mode | Depletion-mode (normally-on) | Natural 2DEG at Vgs = 0 |
First-generation GaN HEMTs succeeded because their structure directly translated the unique polarization physics of the III-nitrides into a practical high-power, high-frequency transistor. The same fundamental heterostructure and 2DEG mechanism remain the foundation of nearly all subsequent lateral GaN power and RF devices, even as later generations added enhancement-mode gates, advanced field plates, improved buffers, and CMOS-compatible processing.
4) Performance Characteristics and Early Applications
Key Performance Metrics of First-Generation Devices
First-generation commercial and near-commercial GaN HEMTs (introduced by Eudyna/Sumitomo around 2004–2006, Nitronex in 2005, and subsequent entries from Cree, RFMD, and others) exhibited the following characteristic performance attributes:
Power density
Laboratory and early commercial devices routinely achieved continuous-wave (CW) power densities of 5–10 W/mm or higher at microwave frequencies. Notable early results included:
- Approximately 9–12 W/mm at 4–10 GHz in research structures with field plates.
- Packaged commercial transistors delivering tens to hundreds of watts total output while occupying significantly smaller die area than equivalent GaAs or LDMOS devices.
This represented roughly a 5–10× improvement over contemporary GaAs FETs (typically ~1 W/mm) and a clear advantage over silicon LDMOS at higher frequencies and voltages.
Operating voltage and current density
Devices operated at drain voltages of 28–50 V (some research and early commercial parts pushed higher). Peak drain current densities commonly reached 800–1000 mA/mm. The combination of high voltage and high current density enabled high absolute power levels from compact transistors.
Efficiency
Power-added efficiency (PAE) or drain efficiency in the 50–65% range was frequently demonstrated under CW or pulsed conditions at L-, S-, and C-band frequencies. High efficiency reduced thermal load and system cooling requirements—an important system-level benefit.
Frequency capability and gain
Early processes with optically defined gates (~0.5–0.9 µm) supported useful performance from UHF through X-band and into higher frequencies in research devices. Typical power gain values were in the 10–15 dB range at relevant operating frequencies. Cut-off frequencies (f_T and f_max) were sufficient for cellular base-station bands (around 2 GHz) and radar applications.
Other electrical characteristics
- Threshold voltages typically –2 V to –5 V (depletion-mode).
- Breakdown voltages often exceeding 100 V in well-designed structures.
- On-resistance and sheet resistance of the 2DEG supported low conduction losses relative to the high power levels achieved.
These metrics were realized despite the challenges of surface trapping, current collapse, and thermal management that still limited reliability and consistency in the earliest products. Field-plate structures, silicon-nitride passivation, and improved epitaxial buffers progressively mitigated these issues.
Comparison with Competing Technologies
Relative to the dominant RF power technologies of the early 2000s:
- Versus GaAs FETs/HEMTs: GaN offered substantially higher power density, higher operating voltage, and better thermal performance on SiC substrates.
- Versus silicon LDMOS: GaN provided higher frequency capability at high power, higher power density, and improved efficiency, particularly above 2 GHz. LDMOS remained competitive at lower frequencies and for very high total power levels where cost and packaging maturity favored silicon.
The net system impact was the ability to deliver more RF power from a smaller, more efficient amplifier module, reducing size, weight, and cooling demands.
Early Commercial Applications
First-generation GaN HEMTs found their initial markets in applications that valued high power density, efficiency, and high-frequency operation:
Wireless infrastructure (base stations)
The primary commercial driver. GaN transistors began appearing in 3G (W-CDMA/UMTS) base-station power amplifiers around 2006–2007. High power density allowed smaller remote radio heads and improved efficiency under complex digital modulation with digital pre-distortion. Early products included 50–180 W class devices operating at 28–50 V for cellular bands. Broadband GaN-on-Si devices also targeted WiMAX and multi-band operation.
Radar systems
Military and civil radar benefited from GaN’s combination of high power, efficiency, and bandwidth. Solid-state power amplifiers (SSPAs) using GaN began displacing traveling-wave tubes in L-, S-, and X-band radar. Advantages included higher reliability, faster response, and the ability to support active electronically scanned arrays (AESA) with many transmit/receive modules. Early demonstrations reached hundreds of watts pulsed output with good efficiency.
Satellite communications and other microwave systems
High-power, high-efficiency amplifiers for satellite uplinks and broadband microwave links exploited GaN’s voltage and frequency capability. Compact size and reduced thermal load were particularly valuable in space-constrained or thermally limited environments.
Defense and specialized RF systems
Beyond radar, GaN entered electronic warfare, jamming, and other high-power RF applications where power density and ruggedness mattered.
System-Level Impact and Limitations in Early Deployment
Early adopters gained measurable reductions in amplifier size and weight, improved DC-to-RF efficiency, and the ability to support wider instantaneous bandwidths. However, first-generation devices carried practical constraints:
- Depletion-mode (normally-on) operation required negative gate bias circuitry or cascode configurations.
- Higher cost relative to LDMOS or GaAs, driven by SiC substrates and process maturity.
- Ongoing reliability qualification, particularly under long-term RF stress and high-temperature operation.
- Thermal packaging challenges when pushing high power densities.
Despite these limitations, the demonstrated performance was sufficient to secure design wins in infrastructure and defense programs, creating the volume and learning-curve momentum that later enabled lower-cost GaN-on-silicon power devices and enhancement-mode transistors.
Summary of Typical Early Performance
| Parameter | Typical First-Generation Range | Significance |
|---|---|---|
| Power density (CW) | 5–12 W/mm (research); multi-W/mm packaged | Compact high-power amplifiers |
| Drain voltage | 28–50 V | Higher voltage than GaAs; competitive with LDMOS |
| Peak current density | 800–1000 mA/mm | High absolute current from small periphery |
| Efficiency (PAE/drain) | 50–65% | Reduced thermal and power-supply burden |
| Frequency coverage | UHF to X-band (and higher in research) | Cellular, radar, satellite |
| Gate length | ~0.5–0.9 µm | Optical definition; adequate for target bands |
| Mode | Depletion-mode | Natural 2DEG; required bias adaptation |
First-generation GaN HEMTs proved that the material’s intrinsic advantages could be translated into practical RF power devices with clear system benefits. Their early success in base stations and radar laid the foundation for the broader adoption of GaN across both RF and power-conversion markets in subsequent generations.
5) Limitations that constrained first-generation GaN HEMT technology
Normally-On (Depletion-Mode) Operation
The polarization-induced 2DEG exists at zero gate bias, making the transistors naturally conducting (normally-on). Threshold voltages typically ranged from –2 V to –5 V. This required negative gate-source bias to turn the device off.
In power-amplifier circuits this complicated bias sequencing, increased the risk of catastrophic failure during power-up or control-signal loss, and necessitated additional circuitry (negative voltage supplies or cascode pairing with a silicon MOSFET). Circuit designers accustomed to normally-off silicon LDMOS or GaAs pHEMTs faced a steeper integration barrier. The inconvenience of depletion-mode operation was repeatedly cited as a factor restricting acceptance outside specialized RF infrastructure and defense applications.
Current Collapse and Dynamic On-Resistance
Surface states and bulk traps in the buffer or barrier caused significant dispersion. Under high drain bias or RF drive, electrons could be trapped at surface states in the gate-drain access region or in deep levels within the GaN buffer. This depleted the 2DEG, raising the effective on-resistance (dynamic R_ON) and reducing available current—a phenomenon widely known as current collapse or knee-voltage walk-out.
The result was degraded large-signal RF performance, lower efficiency under modulation, and potential thermal runaway in extreme cases. Early silicon-nitride passivation and field-plate structures mitigated the problem but did not eliminate it. Residual dispersion remained a reliability and performance concern, particularly under pulsed or high-voltage stress conditions typical of radar and base-station use.
Material Defects and Substrate Mismatch
Because bulk GaN substrates were not commercially available at scale, devices were grown on foreign substrates (primarily SiC or silicon). Lattice and thermal-expansion mismatches produced high densities of threading dislocations (often still in the 10⁸–10⁹ cm⁻² range in early material). These defects acted as trapping centers, leakage paths, and sources of long-term degradation.
Buffer leakage, parallel conduction, and reduced breakdown voltage were common consequences. On silicon substrates the mismatch was more severe, requiring sophisticated nucleation and strain-management layers that added process complexity and cost while still limiting ultimate material quality relative to later native-substrate approaches.
Reliability and Long-Term Stability
Early devices exhibited a range of reliability challenges under high-temperature, high-voltage, and high-RF-stress conditions:
- Hot-carrier effects and trap generation near the gate edge.
- Time-dependent dielectric or barrier degradation.
- Threshold-voltage shifts and gate leakage increases.
- Limited high-temperature operating-life (HTOL) data at the highest voltages.
Statistical reliability data were sparse compared with mature silicon or GaAs technologies. Qualification for infrastructure and defense programs therefore required extensive additional testing, slowing adoption and increasing development cost.
Cost and Manufacturing Maturity
SiC substrates were expensive. Gold-based metallization schemes (common for ohmic contacts and gates) were incompatible with standard silicon CMOS lines, limiting foundry options and volume scaling. Process control, yield, and packaging technology lagged behind silicon LDMOS. Early GaN-on-silicon efforts reduced substrate cost but introduced additional defect and thermal-management trade-offs.
These factors kept device prices high relative to competing technologies, confining first-generation products largely to performance-critical rather than cost-sensitive markets.
Thermal Management and Packaging Constraints
Although GaN itself has respectable thermal conductivity, heat extraction depended heavily on the substrate and package. High power densities concentrated heat in small die areas. On silicon the thermal path was poorer than on SiC. Packaging solutions were still evolving; early air-cavity and plastic packages did not always provide optimal thermal resistance or RF performance, limiting the practical power levels that could be extracted reliably.
Voltage and Frequency Scaling Limits of the Lateral Architecture
The lateral geometry concentrates the peak electric field near the surface at the drain-side edge of the gate. This made further increases in breakdown voltage difficult without longer gate-drain spacing (raising on-resistance) or more elaborate field-plate structures. Frequency scaling was also constrained by the relatively thick barriers needed for adequate 2DEG density and by short-channel effects once gate lengths were reduced.
Summary of Primary Limitations
| Limitation Category | Manifestation | Impact on Adoption |
|---|---|---|
| Device polarity | Normally-on (depletion-mode) | Circuit complexity, safety concerns |
| Trapping / dispersion | Current collapse, dynamic R_ON | RF performance degradation, efficiency loss |
| Material quality | High dislocation density from heteroepitaxy | Leakage, trapping, reliability |
| Reliability data | Limited HTOL and failure-mode understanding | Slow qualification for critical systems |
| Cost structure | Expensive SiC, Au metallization, low volume | Restricted to high-value RF niches |
| Thermal / packaging | High power density vs. heat extraction | Practical power limits, system design constraints |
| Architectural scaling | Lateral field crowding | Barriers to higher voltage and frequency |
These limitations did not prevent first-generation GaN HEMTs from achieving commercial success in wireless infrastructure and radar. They did, however, define the technical agenda for the next phase of development: enhancement-mode (normally-off) gate structures, improved surface passivation and field-plate designs, higher-quality buffers and substrates, gold-free CMOS-compatible processes, better packaging, and eventually vertical device architectures. Addressing these constraints transformed GaN from a specialized high-performance RF technology into a broader platform for both RF and power-conversion applications.