HEMT Notes (Chapter 20)

20.1 Introduction

  • The High-Electron Mobility Transistor (HEMT) was first demonstrated by Mimura and colleagues at Fujitsu Labs in 1980 [1].
  • HEMT represents the culmination of bandgap engineering and molecular beam epitaxy (MBE) advances that also enabled devices such as the heterostructure laser, the heterojunction bipolar transistor, and heterostructure avalanche photodiodes.
  • The HEMT is based on modulation doping, first demonstrated by Dingle and collaborators at Bell Labs in 1978 [2].
  • Modulation-doped structures create a two-dimensional electron gas (2DEG) at the interface between two semiconductors with different bandgaps. The spatial separation between dopants and electrons yields mobility that exceeds the bulk value, even at relatively high carrier concentrations (Fig. 20.1).
  • Initial modulation-doped structures and HEMTs used the AlGaAs/GaAs system, made possible by atomic-layer-precision growth via MBE, which enables near-monolayer interface abruptness, coherent interfaces, and precise dopant placement.
  • Early mobility enhancements were modest at room temperature but became significant at low temperatures. Over time, low-temperature mobility improved dramatically, enabling fundamental solid-state physics discoveries [4].
  • This chapter traces the evolution of the HEMT from inception to its current state across various III–V systems (arsenides and phosphides focus) and discusses how high-crystalline-quality AlGaAs/GaAs heterostructures spurred new physics.
  • It closes with a look at future potential for HEMT-like structures in multiple domains.

20.2 HEMT Electronics

  • From the outset, the combination of high mobility and strong 2DEG confinement suggested using modulation doping for high-speed FETs with excellent short-channel effects and great scaling potential.
  • Rapid progress in high-speed logic followed. Within one year, Fujitsu demonstrated the first HEMT IC featuring an enhancement/depletion-mode logic ring oscillator with a switching delay of 17.1 ps (17.1ps), the lowest among semiconductor logic technologies reported at that time [5].
  • HEMTs showed clear speed–power advantages over contemporary GaAs MESFETs and Si ECL, driving a race to demonstrate high-speed HEMT circuits.
  • Fujitsu achieved SRAMs of increasing density: 1 kbit static RAMs in 1984, 4 kbit in 1987, and 64 kbit in 1991 [3].
  • Despite impressive demonstrations in supercomputers and other systems, this early HEMT IC path did not outcompete Si CMOS in density, power, and economies of scale.
  • The first mass-market HEMT application emerged in 1983 as a four-stage HEMT amplifier operating at 20 GHz, with gain improving as temperature decreased due to mobility enhancements at cryogenic conditions. This cryogenic performance outperformed GaAs MESFETs of the era and contributed to the Nobeyama Radio Observatory’s adoption for their 45 m radio telescope, enabling the discovery of new interstellar molecules in Taurus Molecular Cloud (1985–1986) [3].
  • The stability of HEMTs vs. parametric amplifiers made them valuable for long observations in astronomy and radio astronomy instrumentation [7].
  • The first mass-market AlGaAs/GaAs HEMT applications surfaced with high-frequency noise advantages over MESFETs, due to superior current-gain cut-off frequency $f_T$ and high channel aspect ratio.
  • Direct Broadcasting Satellite receivers with low-noise HEMT amplifiers went commercial in 1987, enabling antenna-size reductions by about half. By 1988, worldwide HEMT receiver production reached roughly 20 million devices/year [8].
  • Mid-1980s innovations expanded GaAs-based high-frequency capabilities:
    • 1985: Invention of the Pseudomorphic HEMT (PHEMT), an AlGaAs/InGaAs/GaAs quantum-well structure where InGaAs provides enhanced electron transport and tight confinement with a large conduction-band discontinuity.
    • Planar (delta) doping of the AlGaAs barrier improved transconductance, channel aspect ratio, and device scalability; removing dopants from under the gate improved breakdown voltage.
    • PHEMTs enabled 0.1 μm gate-length devices and achieved record noise and power performance at very high frequencies (Fig. 20.2).
  • PHEMTs and related delta-doped HEMTs spurred co-integration of n-channel and p-channel devices, enabling complementary circuits in AlGaAs/GaAs [32] and InAlAs/InGaAs systems [33]. This led to highly integrated front-end modules and more complex single-chip systems (Fig. 20.2 cross-section).
  • InAlAs/InGaAs HEMTs on InP (InP HEMTs) excelled at ultra-low noise and very high frequency performance. They achieved record $fT$ and $f{ ext{max}}$ values, with $fT$ and $f{ ext{max}}$ exceeding $640$ GHz in InAlAs/InGaAs pseudomorphic HEMTs on InP, and the record $f_T$ reaching $738$ GHz. Notably, these results occurred at low operating voltage of $0.5$ V [35–36].
  • Other remarkable performance figures in InP HEMTs include $f_{ ext{max}} > 1$ THz and a noise figure as low as 0.71 dB at 95 GHz [37–39], with ultralow-noise performance reported at cryogenic temperatures [40].
  • Figure references:
    • Fig. 20.3 shows the state-of-the-art InAlAs/InGaAs HEMT on InP (“InP HEMT”) with the gate stack and delta doping from the era.
    • Fig. 20.4 tracks the evolution of the current-gain cut-off frequency $f_T$ for GaAs MESFETs, GaAs PHEMTs, and InP HEMTs over time, illustrating approaching and surpassing sub-THz regimes.
  • Metamorphic HEMTs (MHEMTs) on GaAs substrates emerged to improve manufacturability (larger wafers, better reproducibility, easier handling) with wafer sizes up to 6'' GaAs and compatible processing. InAlAs/InGaAs MHEMT processes have been available in foundry mode since 2003 [47]. A wide variety of circuits with performance rivaling the best InP HEMT technologies demonstrated sub-mm wave capabilities, including amplifiers operating up to 460 GHz [48].
  • The “delta-doped pseudomorphic HEMT” (S di/d gating) schematic (Fig. 20.2) illustrates a typical cross-section from ca. 1987 with features such as a high-dopant density region beneath the spacer and a thin AlGaAs barrier to improve breakdown and transconductance.
  • The state of the art includes InGaAs channels with higher In content to increase transport properties and confinement, enabling mm/sub-mm wave operation and extremely high $f_T$ values.

20.2 HEMT Electronics (continued)

  • The GaAs-based HEMTs and InP-based HEMTs show extremely high frequency performance, but the scaling of conventional planar HEMTs is approaching physical limits due to gate leakage as the wide-bandgap barrier is thinned.
  • Proposed future directions to push performance beyond the planar HEMT bottleneck include:
    • Introducing a dielectric in the gate stack and scaling gate length further via FinFET configurations or lateral/vertical nanowire geometries.
  • A major excitement in this search for higher performance is the AlGaN/GaN HEMT:
    • Modulation-doped GaN-based HEMTs were demonstrated in 1992–1993 by Khan and collaborators at APA Optics; today, GaN HEMTs offer a high breakdown voltage (wide bandgap), high electron saturation velocity, and high current concentration at the AlGaN/GaN interface, leading to unique power amplification attributes up to millimeter-wave frequencies.
    • GaN HEMTs are expected to advance high-power, high-frequency radar, communications, and power-management applications. MISFET/MOS-HEMT structures and p-GaN gate HEMTs are among the discussed device approaches for low-leakage, enhancement-mode devices. Detailed treatment of GaN HEMTs is beyond the scope of this chapter.

20.3 Modulation-Doped Structures in Physics

  • The AlGaAs/GaAs heterostructure is a nearly perfect crystalline interface with minimal dangling bonds or interface traps, enabling the formation of a pristine 2DEG and enabling the observation of quantum phenomena.
  • The pursuit of a near-perfect AlGaAs/GaAs interface also aimed to realize precise quantization of Hall resistance (IQHE), driving much of the research into reducing scattering and maximizing mobility at the interface [53].
  • Continuous MBE improvements increased the low-temperature mobility from about
    ext{mobility}_{ ext{low-T}}
    ightarrow 5.7 imes 10^{7} rac{ ext{cm}^2}{ ext{V·s}}
    (the highest mobility reported in any semiconductor structure), enabling the observation of a rich physics treasure-trove previously obscured by disorder [4].
  • Mobility increases correlated with the emergence of new phenomena:
    • Fractional Quantum Hall Effect (FQHE) was observed in samples with mobility around 9imes104extcm2extV⋅s9 imes 10^{4} \frac{ ext{cm}^2}{ ext{V·s}} [55].
    • Microwave-induced zero-resistance states appeared in samples with mobility around 3imes106extcm2extV⋅s3 imes 10^{6} \frac{ ext{cm}^2}{ ext{V·s}} [4].
    • Interactions between composite fermions were postulated in samples with mobility around 1imes107extcm2extV⋅s1 imes 10^{7} \frac{ ext{cm}^2}{ ext{V·s}} [4].
    • A fragile even-denominator 5/2 quantum Hall state was observed with mobility around 3.1imes107extcm2extV⋅s3.1 imes 10^{7} \frac{ ext{cm}^2}{ ext{V·s}}, speculated to hint at non-abelian quasi-particles relevant to quantum computation [4].
  • Gating of AlGaAs/GaAs heterostructures allows carving 1DEG (quantum wires), 0DEG (quantum dots), and other emergent quantum structures where electrons are coherent over long distances.
  • Notable quantum phenomena include quantized conductance in quantum wires (1DEG) and single-electron tunneling in quantum dots (single-electron transistor) [56–58].
  • The IQHE’s Hall resistance plateaus are tied to fundamental constants (Planck’s constant and electron charge); the Ohm is defined through measurements of the Hall resistance on the first plateau in arrays of quantum Hall bars in AlGaAs/GaAs heterostructures, achieving precisions of a few parts in 10^9 [59].
  • Sensor-density advances follow: ultrahigh mobilities enable high-sensitivity micro-Hall sensors for scanning Hall probe microscopy and bio-recognition [60], three-axis Hall sensors for integrated magnetic field measurements [61], and Terahertz (THz) detection/mixing using 2DEG structures [62–63].
  • Functionalized GaN-based 2DEG structures offer robust chemical and biological sensing in gases and liquids, with potential for multi-parameter sensing when combined with MEMS resonators [64–65].

20.4 Exciting Prospects

  • The future is bright for 2D carrier confinement devices across electronics, communications, physics, and sensing:
    • GaAs, InP, and GaN HEMTs will continue toward greater integration, higher frequency, higher power and efficiency, lower noise, and reduced cost.
    • GaN promises access to high-power, high-frequency regimes approaching vacuum-tube capabilities, with lighter, more efficient, and more reliable radar and communications systems on the ground and in space.
    • Anticipated device directions include MOS-HEMT or MISFET structures with enhanced gate control, p-GaN gate HEMTs for improved device performance.
    • Other material systems (e.g., antimonides) may emerge to enable ultralow-power, high-speed systems.
  • History suggests that continued improvements in AlGaAs/GaAs quality will yield new physical insights into quantized electron systems, and similar progress in other heterostructures will yield surprises as well.
  • InGaAs-based 2DEG devices enable THz sensing, mixing, and frequency multiplication [62], while GaN and related materials offer chemically robust platforms for sensing in harsh environments [64–65].

20.5 Conclusions

  • The first 40 years of the HEMT have been rich and exciting; the best is yet to come.
  • Kroemer’s Lemma of New Technology: “The principal applications of any sufficiently new and innovative technology have always been – and will continue to be – applications created by that technology” [66].
  • We are just beginning to discover what 2DEG semiconductor heterostructures can do, with potential impacts from cosmic to personal scales.
  • The next 40 years are expected to bring new surprises and satisfactions across devices, physics, sensing, communications, and power electronics.

Acknowledgments

  • The author thanks colleagues for valuable suggestions: Ray Ashoori, Brian Bennett, Bobby Brar, P. C. Chao, Takatomo Enoki, Augusto Guttierez-Aitken, Eric Higham, Jose Jimenez, Marc Kastner, Richard Lai, Takashi Mimura, Tomas Palacios, Loren Pfeiffer, Philip Smith, Tetsuya Suemitsu, Ling Xia.

References (selected)

  • [1] Mimura, T. et al. (1980). A new field-effect transistor with selectively doped GaAs/n-AlxGa1−xAs heterojunctions. Jap. J. Appl. Phys. 19: L225.
  • [2] Dingle, R. et al. (1978). Electron mobilities in modulation-doped semiconductor heterojunction superlattices. Appl. Phys. Lett. 33: 665.
  • [3] Mimura, T. (2005). Development of high electron mobility transistor. Japn. J. Appl. Phys. 44: 8263.
  • [4] Pfeiffer, L. and West, K.W. (2003). The role of MBE in recent quantum Hall effect physics discoveries. Phys. E. 20: 57.
  • [5] Joshin, T.M.K. et al. (1981). High electron mobility transistor logic. Japn. J. Appl. Phys. 20: L598.
  • [6] Niori, M. et al. (1983). A 20 GHz high electron mobility transistor amplifier for satellite communications. IEEE I.S.C.C. Conf. p. 198.
  • [7] Weber, J.C. and Pospieszalski, M.W. (2002). Microwave instrumentation for radio astronomy. IEEE Trans. Microw. Theory Techn. 50: 986.
  • [8] Mimura, T. (1990). HEMT technology: impact on computer and communications systems. Surf. Sci. 228: 504.
  • [9] Rosenberg, J.J. et al. (1985). In0.15Ga0.85As/GaAs pseudomorphic single quantum well HEMT. IEEE Electron Dev. Lett. 6: 491.
  • [10] Ketterson, A. et al. (1985). High transconductance InGaAs/AlGaAs pseudo-morphic modulation-doped FETs. IEEE Electron Dev. Lett. 6: 628.
  • [11] Chao, P.C. et al. (1987). High performance 0.1 μm gate-length planar-doped HEMTs. IEEE/Cornell Conf. Adv. Concepts… p. 410.
  • [12] Henderson, T. et al. (2007). High-performance BiHEMT HBT / E-D pHEMT integration. CS MANTECH, p. 247.
  • [13] Cheng, K.Y. et al. (1982). Electron mobilities in modulation doped Ga0.47In0.53As/Al0.48In0.52As heterojunctions. Appl. Phys. Lett. 40: 147.
  • [34] Brown, R.B. et al. (1998). Overview of complementary GaAs technology for high-speed VLSI circuits. IEEE Trans. VLSI Syst. 6: 47.
  • [35] Kim, D.-H. and del Alamo, J.A. (2010). 30 nm InAs PHEMTs with $fT=644$ GHz and $f{ ext{max}}=681$ GHz. IEEE Electron Dev. Lett. 31: 806.
  • [36] Jo, H.-B. et al. (2020). Lg = 19 nm In0.8Ga0.2As composite-channel HEMTs with $fT=738$ GHz and $f{ ext{max}}=492$ GHz. IEDM 2020.
  • [37] Lai, R. et al. (2007). Sub-50 nm InP HEMT Device with $f_{ ext{max}} > 1$ THz. IEDM.
  • [38] Kim, D.-H. et al. (2010). 50-nm E-mode In0.7Ga0.3As PHEMT on 100-mm InP substrate with $f_{ ext{max}}>1$ THz. IEDM 2010.
  • [39] Takahashi, T. et al. (2008). Improvement in high-frequency and noise characteristics of InP-based HEMTs by reducing parasitic capacitance. IC/DM 2008.
  • [40] Cha, E. et al. (2020). InP HEMTs for Sub-mW cryogenic low-noise amplifiers. EDL 41: 1005.
  • [41–44] Deal, W.R. et al. (various) on InP HEMT demonstrations operating beyond 1 THz. IEDM/IEEE publications.
  • [53] Schlom and Pfeiffer (2010). Upward mobility rocks! Nat. Mater. 9: 881.
  • [54] Chung, Y.J. et al. (2022). Understanding mobility limits in ultrahigh-mobility GaAs 2D systems. Phys. Rev. B 106: 075134.
  • [55] Tsui, D.C. et al. (1982). Two-dimensional magnetotransport in the extreme quantum limit. Phys. Rev. Lett. 48: 1559.
  • [59] Poirier and Schopfer (2009). Resistance metrology based on the quantum Hall effect. Eur. Phys. J. Spec. Top. 172: 207.
  • [60] Bando, M. et al. (2009). High sensitivity micro-Hall sensors in InAlSb/InAsSb/InAlSb heterostructures. J. Appl. Phys. 105: 07E909.
  • [61] Todaro, M.T. et al. (2010). Fully integrated GaAs-based three-axis Hall sensor. J. Microelectromech. Microeng. 20: 105013.
  • [62] Dyakonov, M. and Shur, M. (1996). Terahertz detection via 2D electron gas. IEEE Trans. Electron Dev. 43: 380.
  • [63] Kawano, Y. and Ishibashi, K. (2010). On-chip near-field terahertz detection with 2DEG. Phys. E. 42: 1188.
  • [64] Pearton, S.J. et al. (2004). GaN-based diodes and transistors for chemical, gas, biological and pressure sensing. J. Phys. Condens. Matter 16: R961.
  • [65] Niebelschutz, F. et al. (2008). AlGaN/GaN-based MEMS with 2DEG for novel sensor applications. Phys. Stat. Sol. C 5: 1914.
  • [66] Kroemer, H. (2000). Nobel lecture: quasielectric field and band offsets. Rev. Mod. Phys. 73: 783.

Notes on LaTeX usage in this summary:

  • Inline and display math is used where numerical quantities and frequencies are cited. Examples include:
    • $fT$ and $f{ ext{max}}$ values, e.g., f<em>T>640  GHzf<em>T > 640\ \,\mathrm{GHz} and f{ ext{max}} > 1\ \,\mathrm{THz}.
    • Record values: fT=738  GHzf_T = 738\ \,\mathrm{GHz}.
    • Gate length and feature sizes: 0.1 μm0.1\ \mu\mathrm{m} gate length.
    • Materials and mobility figures: 5.7×107 cm2 V−1 s−15.7 \times 10^{7}\ \mathrm{cm^{2}\,V^{-1}\,s^{-1}} (low-temperature mobility).
    • Operating voltage for InP HEMTs: 0.5 V0.5\ \mathrm{V}.
    • Metamorphic wafer sizes: $6''$ GaAs wafers.
    • THz and sub-THz demonstrations: >1\ \mathrm{THz} and >460\ \mathrm{GHz} demonstrations, etc.