Progress in performance development of room temperature direct terahertz detectors

In this paper issues, associated with the development of THz direct detectors and focal plane arrays in the last decade are discussed. After short description of general classification of THz detectors, more details concern Schottky barrier diodes, CMOS-based detectors, microbolometers, and field-effect transistor detectors, where links between THz devices and modern technologies such as micromachining are underlined. Special attention has been paid to the development of detectors made of two-dimensional materials. Their performance is comparable to that presented for classical terahertz detectors. It is shown that applications of nanoscale materials and devices, in particular, made of two-dimensional materials, open the door for further performance improvement of THz detectors operated at room temperature.


Introduction
The development of THz technology started in the 1980s mainly for laboratory applications (see Fig. 1). It was an expensive scientific device that could only be used by highly trained personnel. However, over the last two decades, efforts have been made to implement easy-to-use and cost-effective systems by developing more compact and reliable components. Now, THz techniques are entering the commercial market, and the first products for applications such as industrial process monitoring on the production line have already been sold.
The total size of the THz market appears to be difficult to determine. The market reports differ in their estimates for 2020 by more than a factor of 4 (Fig. 2). According to a French research firm, Tematys, the total market is about 100 USD million [ 3]. At the other end of the spectrum is Beijing-based QY Research, which believes that the current market size is $385 million [4]. Similarly, forecasts for the annual growth rate (CAGR) vary by a factor of two, with values ranging from 16 (Tematys) to nearly 32% (QY Research). From Fig. 2 results that a simple arithmetic average from the global market five reports, is approximately 230 million US$ in 2020. All reports agree that the THz technology market will continue to grow, and even the most pessimistic forecast CAGR of 16% still indicates significant growth.

General Classification of Terahertz Detectors
The terahertz (THz) detectors are generally classified in two broad categories: thermal detectors and photon detectors. Their operation principled is briefly described in Table 1.
THz detector should be characterized by high sensitivity, short response time, low-power consumption, and compatibility with the electronic input circuit, as well as operational reliability, high quantum efficiency, good thermal properties, low dark current, low noise, long lifetime, small size, and low manufacturing costs. In the following, we will define some important parameters that characterize terahertz detectors and their detector arrays. Responsivity is defined as the output signal (typically voltage or current) of the detector produced in response to a given incident radiant power falling on the detector.
Noise equivalent power (NEP) is the incident power on the detector that produces a signal-to-noise ratio of unity at the output of a given optical detector at a given data-signaling rate or modulation frequency, operating wavelength, and effective noise bandwidth. Some manufacturers and authors define NEP as the minimum detectable power per square root bandwidth [W/Hz 1/2 ].
In terms of responsivity, it can be written: where V n and I n are voltage and current noises and R v and R i are voltage and current responsivities. The unit of NEP is the watt. The NEP is also quoted for a fixed reference bandwidth which is often assumed to be 1 Hz. This "NEP per unit bandwidth" has a unit of watts per square root Hertz (W/Hz 1/2 ). The NEP depends on the optical wavelength, as well, since the detector responsivity is wavelength dependent.
The normalized detectivity D * (or D-star) is related to the unit detector area and the unit bandwidth. This means that both NEP and detectivity are functions of the electrical bandwidth, ∆f, and detector area, A d : It is the most important detector parameter which allows the comparison of the same type detectors but having different areas. D * is expressed in the unit of cmHz 1/2 /W which is more commonly called "Jones." Noise equivalent difference temperature (NEDT) is a fundamental figure of merit for focal plane arrays (FPAs) and characterizes the thermal sensitivity of a system. It can be defined as the amount of temperature difference required to produce a unity signal-to-noise (S/N) ratio. NETD is defined as: where V n is the noise and ∆V represents the signal measured for the temperature difference ∆T. NETD characterizes the detection system sensitivity for low spatial frequencies. A smaller NEDT value indicates a better thermal sensitivity. Table 2 compares the general performance of the state-of-the-art THz detectors in terms of their operating frequency, speed, responsivity, and noise equivalent power.

Uncooled Direct THz Detectors
The performance of different types of THz detectors operating at room temperature is summarised in Table 3 [5]. As we see, the spread in detector parameters is large. The sensitivity parameter only partially characterizes the detector. NEP is much more important from the detector system point of view-there is a lot of experimental data on NEP for rectifier detectors. In the last decade CMOS detectors, microbolometers and plasma detectors have been growing rapidly, indicating that these technologies are widely available. The data for pyroelectric detectors and Golay cells are the most sparse of all.
The performance results to date for different detector type are compared in the bar chart of Fig. 3. We chose the frequency range close to 650 GHz since it is the center frequency located in a well-known window useful for terrestrial sensing (at higher frequency the atmosphere is prohibitively lossy). As is shown, the Si-CMOS and SiGe-based devices are characterized by NEP values around 50 pW/Hz 1/2 . For pyroelectric detectors and Golay cells, the NEP values lie above 100 pW/Hz 1/2 and close to 1 nW/Hz 1/2 . Schottky-barrier detectors (SBDs) respond to the THz electric field and usually generate an output current or voltage through a quadratic term in their current-voltage characteristics. In general, the NEP of SBD and FET detectors is better than that of Golay cells and pyroelectric detectors around 650 GHz. Both the pyroelectric and the bolometer FPAs with detector response times in the millisecond range are not suitable for heterodyne operation. FET detectors are clearly capable of heterodyne detection with improved sensitivity. 2D materials like graphene and black phosphorus (bP) (FET and PTE detectors) are promising materials for the development of room temperature detectors operating across the FIR in spectral range above 100 μm (at frequency below 3 THz).

2D Materials in THz Detector Family
Since the discovery of graphene, intensive research has begun on 2D materials for their potential applications as photodetectors over a wide range of the electromagnetic spectrum. Most of the research has been devoted to photon detectors in the visible and NIR range. However, a more interesting range of potential applications for these materials is the terahertz range.
It is well-known that the performance of photodetectors mainly depends on the properties of the photodetector active regions, such as absorption coefficient, e-h pair lifetime, and charge mobility [6]. The high dark current of conventional   Table 4 summarizes the performance of representative detectors [7]. Figure 4 compares NEP values of graphene-based and bP detectors with existing standard THz ones. Most of the experimental data collected in the literature are for single graphene detectors operating above 100 μm wavelength (frequency range below 3 THz). In general, the performance of graphene FET detectors is lower compared to CMOS and plasmonic detectors made of III-V Schottky-barrier and silicon-based materials, SiGe and InGaAs. However, compared to VO x and amorphous silicon microbolometers, the performance of graphene detectors is close to the trend line estimated for microbolometers in the THz spectral region (see Fig. 5). The best quality VO x bolometer arrays are characterized by lower NEP value of about 1 pW/Hz 1/2 in the LWIR range (≈ 10 μm). However, it should be noted here that the microbolometer data are for above megapixels monolithic arrays, what is big advantage of this monolithic infrared detector technology in comparison with THz range.
As Fig. 4 shows, black phosphor has emerged as candidate of 2D material for terahertz detection applications. However, surface instability due to chemical

Uncooled THz Focal Plane Arrays in Active Imaging
There are two methods used in THz imaging: passive imaging and active imaging. The passive imaging uses the thermal radiation emitted by an object. Since the lowtemperature objects emit very low power radiation, the THz detector sensitivity is required to be very high (NEP value on the order of several fW/√Hz), which can only be achieved using cryogenic or heterodyne detector systems. The manufacturing and operating costs as well as the size of these systems limit them to applications such as defense, space, and security. To omit these technological barriers, the uncooled THz active systems are manufactured with considerable lower costs. However, these systems require the use of external sources of illuminations. The schematic arrangement of active THz imaging system is shown in Fig. 6. Most of the systems use monochromatic THz sources such as quantum cascade lasers (QCLs) or far infrared optically pumped lasers delivering mW-range powers. As is shown in Fig. 6, the reflected beam backlights an object with a maximum area, and the transmitted light is collected by a camera lens. The focal plane is positioned behind the camera lens, making the object plane in front of the lens. Also shown is  the modified reflection mode setup, where a specular reflection is collected by the repositioned lens and camera.
In the period 2010-2011, three different companies/organizations announced cameras optimized for the > 1-THz frequency range: NEC (Japan) [36], INO (Canada) [37], and Leti (France) [38]. Several vendors fabricated cameras with array size from 16 × 16 to 384 × 288 pixels which is listed in Table 5 [39]. The largest and most sensitive ones are provided by microbolometer technology-based sensors. Their minimum detectable power (MDP) is in the range of a few dozen pW (at frequency a few THz). For comparison, the MDPs of FET-based cameras are in nW range, e.g., 9 nW for Tic-Wave camera developed by the University of Wuppertal. FET-based cameras are smaller and, due to larger footprints, are more suitable for longer wavelength range, up to100 GHz. Sensitivity of NeTHIS and Ophir Photonics cameras is considerably lower, around 10 μW, but their spectral response is very broad from infrared to terahertz. The first of them is thermoconversion membrane-based camera. The second one contains hybrid array which consists of a LiTaO 3 pyroelectric crystal connected with indium bumps to a silicon readout multiplexer. Figure 7 presents development status of commercial THz cameras containing two-dimensional detector arrays. Their details characteristics are given in Table 5. Figure 8 presents trends in development of room-temperature 32 × 32 FET CMOS THz detectors. As can be seen, in the last 10 years, the NEP value has been improved by two orders of magnitude. The lowest reported NEP is in the range of 12-14 pW/√Hz at operating band around 650-1000 GHz in 65-nm CMOS [46]. It should be noted that although device scaling may result in lower thermal noise of the MOSFET channel, the detector parameters (sensitivity, NEP, operating bandwidth) reported so far have not been significantly improved by migration to nanometer CMOS technology nodes. It is expected that one of the main possible reasons for this is the very high (in the kΩ range) and frequency-dependent impedance of the MOSFET operating in moderate inversion, which makes efficient and broadband coupling to the on-chip antenna very difficult.

Conclusions
Over the past 20 years, there have been significant technological developments that have enabled real-time high-resolution THz imaging. Today, THz imaging research is gradually being transferred from laboratory instruments to commercial products. Intensive research is being conducted to develop THz cameras similar to those available in the infrared spectrum. The technical requirements for THz cameras are the possibility of mass production, compatibility with standard fabrication technologies (e.g., CMOS process), low weight, small size and low-power consumption, and high sensitivity when operating at room temperature (allowing THz imaging even without an active THz source).
To realize THz cameras that meet these criteria, mainly two main technology directions are being pursued: THz thermal cameras and FET-based THz cameras. In particular, microbolometers have achieved remarkable sensitivity at room  The discovery of graphene and alternative 2D materials with direct energy gaps has opened a new window for THz photodetector fabrication. The combination of scalability, prospects for integration into Si-platforms, and the ability to implement flexible devices makes graphene competitive for the next generation of sensor systems. However, at the present stage of technology, 2D films have a small area (typically 100 μm 2 ), which makes it expensive for industrial applications. Moreover, this technology is not mature, which has detrimental influence on uniformity of pixel arrays. The development of efficient 2D photodetector arrays integrated in the focal plane of the imaging system is particularly critical in industrialization. Most experimental methods for transferring 2D materials from their substrate to the desired electronic circuits are either incompatible with high-volume production or lead to significant degradation of the 2D material and its electronic properties.
The above difficulties in scaling wafer sizes of 2D materials may be overcome in the future. In particular, this is already noticeable in the adaptation of 2D materials for the production of 2D transistors. For example, IMEC introduces tungsten disulfide (WS 2 ) in the logic device scaling roadmap [50]. Recently, the European Commission launched a 20 million project to bridge the gap between lab-scale manufacturing and large volume production of electronic devices based on 2D materials [51].