Semiconductor and Device Technology III-V Semiconductors Informational

What is the advantage of InP over GaAs for millimeter wave applications above 100 GHz?

InP (indium phosphide) has fundamental material advantages over GaAs for frequencies above 100 GHz: (1) Higher electron velocity: InP has higher peak and saturated electron drift velocity: InP: v_sat = 2.5 × 10^7 cm/s, peak velocity = 4.0 × 10^7 cm/s. GaAs: v_sat = 1.0 × 10^7 cm/s, peak velocity = 2.0 × 10^7 cm/s. The higher velocity directly translates to higher transistor speed (fT and fmax). (2) Higher fT and fmax: InP HEMT (InAlAs/InGaAs on InP substrate) with 25-50 nm gate length: fT > 500 GHz. fmax > 1 THz (demonstrated in research). GaAs pHEMT with 100 nm gate length: fT ≈ 150 GHz. fmax ≈ 300 GHz. InP provides usable gain at frequencies where GaAs has no gain. Above 100 GHz: InP HEMT provides 10-15 dB gain per stage. GaAs pHEMT provides < 5 dB gain (marginal). Above 200 GHz: only InP HEMT (and InP HBT) provide useful gain. (3) Lower noise: the noise figure of a transistor is fundamentally related to the noise temperature and the transit frequency: NF_min ≈ 1 + k×f/fT (simplified Pospieszalski model). Since InP has higher fT: NF_min is lower at a given frequency. At 100 GHz: InP HEMT NF ≈ 1.5-2.5 dB. GaAs pHEMT NF ≈ 3-5 dB. At 200 GHz: InP HEMT NF ≈ 3-5 dB. GaAs: no usable gain (cannot function as a LNA). (4) Higher output power at W-band: InP HBT (heterojunction bipolar transistor, InP/InGaAs): provides usable output power at 94-220 GHz. P_sat at 94 GHz: 50-200 mW per device. PAE at 94 GHz: 10-20%. GaAs at 94 GHz: limited to < 50 mW per device (insufficient fmax for efficient PA design). (5) Substrate quality: InP substrates are higher quality (lower defect density) than GaAs for these advanced epitaxial structures. The InAlAs/InGaAs channel on InP has excellent lattice matching, enabling very thin, high-quality layers for sub-50 nm gate lengths.

Category: Semiconductor and Device Technology

Updated: April 2026

Product Tie-In: Transistors, MMICs, Evaluation Boards

InP for Sub-THz Applications

InP technology is the enabling semiconductor for the sub-THz frequency range (100-1000 GHz), which is increasingly important for 6G communications, imaging, and spectroscopy.

InP Device Types

(1) InP HEMT (High Electron Mobility Transistor): channel: InGaAs (lattice-matched or pseudomorphic on InP). Barrier: InAlAs. Gate length: 25-100 nm. fT: 200-600 GHz (scaling with gate length). fmax: 500-1500 GHz. Noise figure: excellent (the lowest NF of any transistor technology above 100 GHz). Used for: LNAs, mixers, and low-noise receivers at 100-300+ GHz. Application: radio astronomy receivers (cryogenic InP HEMTs at 4 K achieve NF < 0.5 dB at 100 GHz), 5G/6G receivers at 100+ GHz, and sub-THz imaging for security screening. (2) InP HBT (Heterojunction Bipolar Transistor): vertical transport device (collector-base-emitter stack on InP substrate). Emitter: InP or InGaP. Base: thin InGaAs (< 25 nm for speed). Collector: InP (high breakdown voltage for output power). fT: 300-500 GHz. fmax: 500-1000 GHz. Higher output power than InP HEMT at the same frequency (the HBT can handle higher current and has higher breakdown voltage). Used for: PAs, oscillators, and power DACs at 100-300 GHz. The InP HBT is the preferred device for transmitter IC design above 100 GHz. (3) Comparison: InP HEMT: best for LNA (lowest noise). InP HBT: best for PA and oscillator (higher power, higher breakdown). Many sub-THz transceiver ICs use both: HEMT for the receive chain and HBT for the transmit chain (this requires a BiHEMT process, combining both device types on the same wafer).

Applications Above 100 GHz

(1) 6G communications: the 100-300 GHz band (D-band: 110-170 GHz, sub-THz) is being explored for future 6G wireless links: data rates: 100+ Gbps per link (using 10-50 GHz bandwidth). Range: 10-100 m (very short range due to high path loss). InP ICs are the only technology capable of providing the necessary gain and output power at these frequencies. (2) Sub-THz imaging: at 200-500 GHz: materials have unique absorption signatures. Security screening: detect concealed weapons under clothing (passive mmWave/THz imaging). Medical: imaging of skin tissue for cancer detection (the sub-THz wavelength penetrates 0.1-0.5 mm into tissue). InP receiver ICs provide the low-noise front end for these imagers. (3) Radio astronomy: detection of faint cosmic signals at 100-1000 GHz. Cryogenic InP HEMT LNAs achieve noise temperatures of 5-50 K (the lowest achievable with any technology at these frequencies). Used in telescopes: ALMA (Atacama Large Millimeter Array), NOEMA, and Herschel Space Observatory.

InP vs GaAs Parameters

InP HEMT: fT > 500 GHz, fmax > 1 THz
GaAs pHEMT: fT ≈ 150 GHz, fmax ≈ 300 GHz
InP NF @100GHz: 1.5-2.5 dB
GaAs NF @100GHz: 3-5 dB (marginal gain)
v_sat: InP 2.5×10⁷ vs GaAs 1.0×10⁷ cm/s

Common Questions

Frequently Asked Questions

Why not use InP for everything?

InP has significant disadvantages at lower frequencies: (1) Cost: InP wafers are 3-5× more expensive than GaAs, and 10-30× more than Si. The InP substrate is fragile and limited to 3-4 inch diameters (vs 6 inch for GaAs and 12 inch for Si). (2) Mechanical fragility: InP is softer and more brittle than GaAs. Processing and handling are more challenging. Yield is lower. (3) Low breakdown voltage: standard InP HEMT has V_BR ≈ 4-7 V (much lower than GaN at 60+ V or GaAs at 15-25 V). This limits the output power per device. For PA applications below 100 GHz: GaN provides 10-100× more output power per device. (4) Limited market: the commercial InP market is small (radio astronomy, military, specialized research). The process is not available at many foundries. This means: fewer process options, longer lead times, and higher NRE. At frequencies below 40 GHz: GaN (for power) and GaAs (for low noise and moderate cost) are superior choices. InP comes into play only when the frequency exceeds the capabilities of GaN and GaAs.

What foundries offer InP processes?

Commercial InP MMIC foundries: (1) Northrop Grumman (US): InP HEMT and HBT processes. Available to US government programs and selected commercial customers. (2) Teledyne Scientific (US, now part of FLIR): high-performance InP HEMT (25 nm gate, fmax > 1 THz). (3) III-V Lab (France): InP HEMT and HBT for research and commercial. (4) Fraunhofer IAF (Germany): InP HEMT (35-50 nm gate, fmax > 800 GHz). Available through the EU foundry access programs. (5) WIN Semiconductors (Taiwan): InP HBT (fT > 300 GHz). Commercial foundry with production capability. (6) GlobalFoundries / Tower Semiconductor: SiGe BiCMOS is an alternative to InP for some applications at 100-200 GHz (lower cost but also lower performance). The InP foundry landscape is much smaller than GaAs or CMOS. Most InP designs are done in close collaboration with the foundry (custom process modifications are common).

Is SiGe a viable alternative to InP above 100 GHz?

SiGe BiCMOS (e.g., GlobalFoundries 8XP, IHP SG13S): fT = 300-400 GHz, fmax = 400-700 GHz. This provides usable gain up to approximately 200 GHz. At 100-200 GHz: SiGe provides: lower cost (processed on standard 8-inch Si wafers), higher integration (millions of transistors per die, enabling complex digital beamforming and signal processing), and adequate gain (5-10 dB per stage at 140 GHz). But: higher noise figure than InP HEMT (SiGe NF at 100 GHz: 5-8 dB vs 1.5-2.5 dB for InP). Lower output power (SiGe PA at 140 GHz: 5-15 mW per amplifier). For applications where noise and power are critical: InP is superior. For applications where cost and integration are critical: SiGe is preferred. The 6G vision: SiGe for the digital transceiver and baseband, InP for the highest-performance RF front end (PA and LNA MMICs).

Keep Reading