Improving terahertz-detection sensitivity of 8x8 FET arrays through liquid-nitrogen cooling in a compact low-noise cryostat

TL;DR

This work demonstrates a compact, liquid-nitrogen–cooled FET-based THz detector system built around an $8\times8$ patch-antenna–coupled Si-CMOS TeraFET array, optimized for $2.85$–$3.4$ THz. By cooling to $77\ \mathrm{K}$ and down to $20\ \mathrm{K}$, the authors observe continuous improvements in noise-equivalent power ($NEP$) driven by reduced thermal noise and enhanced nonlinear mixing, achieving up to $\sim$11–15× NEP improvement at $20\ \mathrm{K}$ and about $4$–$6\times$ at $77\ \mathrm{K}$ relative to room temperature. In a 2.85 THz, $2.1$ mW QCL-excited configuration, the 8×8 detector demonstrates a minimum optical NEP of $\approx 190$ pW/$\sqrt{\mathrm{Hz}}$ (Johnson-noise limited) and $\approx 420$ pW/$\sqrt{\mathrm{Hz}}$ experimentally, with a dynamic range exceeding 67 dB and a readout bandwidth near 5 MHz. The results show that a compact, broad-temperature THz detector system can approach performance levels of superconducting sensors under efficient optical coupling, while offering fast response and ruggedness suitable for space-borne spectroscopy and other constrained platforms.

Abstract

We show that the sensitivity of antenna-coupled field-effect transistors (FETs) to terahertz (THz) radiation improves continuously with decreasing temperature. The noise-equivalent power (NEP) of 540 GHz patch-antenna-coupled FETs decreases as temperature reduces to 20 K. We project NEP values approaching 1 to 2 pW/sqrt(Hz) under efficient power coupling conditions (e.g., using a superstrate Si-lens), which is comparable to superconducting niobium transition-edge sensors (TESs) at 4 K. Building on these findings, a compact, low-noise, liquid-nitrogen-cooled (77 K) FET-based direct (incoherent) THz-power sensing system} for spectroscopy applications was realized. Here, an 8x8 pixel-binned detector array fabricated in a commercial 65-nm Si-CMOS process, was optimized for operation in the 2.85 to 3.4 THz band. Characterization was performed in the focal plane of a 2.85-THz quantum-cascade laser delivering approx. 2~mW of THz power. A linear dynamic range exceeding 67 dB was achieved without saturation (for 1~Hz-detection bandwidth). The system provides a -3 dB readout bandwidth of 5 MHz, exceeding that of conventional thermal detectors (typically 1 kHz). Combined with its broad temperature operability 20 K to 300 K and compact design, the system is particularly well suited for space- and payload-constrained platforms such as balloon- and satellite-based missions, where deep cryogenic cooling is impractical.

Figures (9)

• Figure 1: a) Cross-sectional view of the patch-antenna implementation, representing the unit cell of the $8\times8$ TeraFET array (not to scale; graphic adapted from ludwig_modeling_2024). Relevant dimensions and permittivities are indicated. b) Top view of the transistor layout. The strongly doped $p^+$-well region, shown in pink, was introduced to pin the body potential to source ground (guard ring). The dimensions of the square-shaped ring are $d_{1,p\text{-}well} = { \color{black} \qty{6}{\micro\meter} }$ and $d_{2,p\text{-}well} = { \color{black} \qty{1}{\micro\meter} }$. Depth of the $p^+$-well is estimated as $\sim$60 and doping density around 2.5e19^-3 (compare ludwig_2-d_2025).
• Figure 2: Experimental and simulated THz spectral response of the patch-antenna coupled FET shown in \ref{['fig:TeraFETcrossSectionB6']} at $V_{GS}=\qty{0.55}{\volt}$. Stars: Normalized responsivity taken from measurements with $8\times8$ detector using THz-QCLs at four frequencies (2,5,2,85,3,4 and 3,6). Data at 2.85 and 3.4 was reproduced from holstein_88_2024. Red Line: Spectral response of a single FET-detector (unit cell of the $8\times8$-detector) coupled to a superstrate Si-lens as introduced in krysl_si_2024. Spectral response was determined using a Michelson interferometer with a thermal (Globar) source (vendor: Bruker Corp.). Blue: Simulated responsivity of the patch antenna element. The antenna structure's impedance and efficiency was simulated using CST Studio Suite and the transistor's terahertz response was modeled using an in-house hydrodynamic model implemented in Keysight's Advanced Design System (ADS-HDM) as presented in ludwig_circuit-based_2019 and ludwig_modeling_2024. All measurements presented within this graph were acquired, while the detector element was operated at room-temperature.
• Figure 3: Experimental noise spectral amplitude of both room-temperature and LN2-cooled $8\times 8$ FET array operation at $V_{GS}=\qty{0.55}{\volt}$measured with a Tektronix DPO4104 oscilloscope. (Inset) Schematic of the electronic amplifier circuitry ( Low Noise, High Input Impedance Composite Amplifer , adapted from texas_instruments_jfe2140_2023) consisting of a JFET-input stage and a low-noise amplifier followed by a second low-noise amplifier stage (total gain $G_{V}=100$). The complete amplifier chain is not actively cooled, while the detector is operated at 77. The amplifier chain is enclosed in a metal shielding box (see \ref{['fig:ln2cryo']}).
• Figure 4: Photograph of the LN2-cooled detector system and its internal components. The first inset shows the aluminum-based PCB mounted on the dewar’s cold finger, carrying the detector die. The second inset provides a zoomed-in view of the $8\times8$ pixel binned TeraFET detector array. The resulting effective active detector area is approximately 600x600 (introduced in holstein_88_2024). The housing provides access for liquid nitrogen cooling (top) and includes a vacuum port compatible with external pumping systems (left). Buffer and amplifier electronics (see \ref{['fig:ElectronicsAndNoisespectrum']}) are housed in a shielded metal enclosure to suppress electromagnetic interference from the laboratory environment. Power supply, on/off switch, and the SMA output connector are located on the rear side of the shielding box.
• Figure 5: Experimental setup used for FET detector characterization at cryogenic temperatures at 540. Terahertz radiation generated using an RPG ZTX750 multiplier ($\times$36) (vendor: Radiometer Physics GmbH, Germany) was guided through an optical system onto the detector (without Si-lens) inside a closed cycle cryostat. For this measurement, an OAP with RFL of 3inch was used as the final focusing element due to space constraints. The biasing/readout scheme was same for the 2.85 THz-QCL based setup, while the setup geometry can be seen in \ref{['fig:LeedsQCLSetup']}.