High sensitivity HgTe room temperature terahertz photodetector

The terahertz (THz) wave, which refers to electromagnetic radiation loosely defined as a frequency range from 0.1 to 10 THz,¹ has raised burgeoning interest in both fundamental science and everyday life due to its application-oriented issues. Benefitting from its special position, the THz wave is endowed with unique advantages, such as the fact that it is penetrable to many objects and overlaps with the rotational and vibrational energy levels of biochemical molecular systems,² laying the foundation for applications in biomedicine,³ security imaging,⁴ environmental monitoring,⁵ and communications.⁶ However, efficient exploration of the terahertz technique still lags behind the infrared and microwave ones due to the drawbacks of traditionally light-induced principles that hinder the development of photodetectors with capabilities of high-sensitivity, fast response, as well as on-chip integration.

In the last few decades, much effort has been devoted to the innovation and development of efficient THz detection technology.^7–9 Nevertheless, since the frequency of a THz wave is much higher than the cut-off frequency of traditional electronic devices and the photon energy is much lower than thermal-agitation energy at room-temperature,¹⁰ it is hard to achieve a photovoltaic (PV) effect by making use of light-induced charge-separation via a built-in field in a reverse-biased semiconductor junction. The emergence of two-dimensional (2D) materials offer a fertile playground for exploring THz optoelectronics at the nanoscale.^11–13 Owing to the novel photoelectric properties of 2D materials such as high electron mobility,¹⁴ intriguing band structure,¹⁵ and a larger Seebeck coefficient,¹⁶ a variety of THz detectors based on 2D materials such as graphene,¹⁷ MXenes,^18,19 and TMDCs^20–22 have been reported. However, the performance of 2D material-based THz detectors is aggregated by their weak electromagnetic coupling efficiency,²³ weak absorption,²⁴ and air-stability,²⁵ which hinders their practical applications.

Considerable effort has been invested within the material-science and optoelectronic communities to search for alternative material platforms that enable better device performance and circumvent the above bottlenecks.^26–28 The advent of topological semi-metals (TSM) enables alternative platforms for exploring exotic quasi-particles with the topological character of bulk wave functions for THz detection due to their versatile operation of symmetry-breaking.^29–31 In the last few years, a few caveats of multiple excitations (e.g., Weyl fermions³² and Dirac fermions³³) have been reported in three-dimensional solid-state systems, which has prompted enormous interest in exploiting properties of nontrivial topology that are absent in other materials. By breaking either time reversal or inversion symmetry,³⁴ numerous peculiar quantum effects such as nonlinear Hall effect,³⁵ nonlinear optical properties,³⁶ and chiral anomalies³⁷ were found, offering ad-hoc properties to target the expected performance and functionalities.

HgTe is a semi-metallic material hosting a zinc blende (ZB) crystal structure^38,39 and band-inversion with a negative bandgap (−0.3 eV)^40,41 so that the selective wavelength dependence of the photoresponse imposed by the gap is absent. Benefiting from the large Seebeck coefficient (about−135 µV/K) and excellent thermoelectric characteristics, a HgTe-based PTE THz detector could possess better response time and high frequency characteristics than electronic detection based on the rectifier effect or nonlinear frequency mixing mechanism in theory. The research on HgTe mainly focuses on quantum wells, which is a typical two-dimension (2D) topological insulator and has shined in the field of infrared detection for decades.⁴² In recent years, HgCdTe has previously been utilized for bow-tie antenna terahertz detectors.^43,44 However, since HgCdTe belongs to the ternary metal, this results in a huge obstacle to epitaxial growth. Meanwhile, the expensive CdZnTe substrate used for growth also hindered its further development. HgTe, with its large area epitaxial growth, low cost, and high performance, becomes the best choice. The transformation of dimensions may bring new possibilities. The strain HgTe thin film is a three-dimensional (3D) topological insulator, which could evolve into a semi-metallic state under the regulation of stress and other factors.⁴⁵ Traditionally, semimetals have not been considered as candidates for photodetection due to the large dark current that occurs when bias-voltage traverses across the channel. However, the successful demonstration of a graphene-based field-effect-transistor (FET) photodetector offers the feasibility of using semi-metallic materials for high-speed and broadband photodetection down to the far-infrared regime in terms of the photothermoelectric or carrier multiplication effect.⁴⁶ A large Seebeck coefficient and the ultra-high electron mobility (4.5 × 10⁴ cm²/V s) in combination with excellent ambient-stability⁴⁷ render HgTe an ideal candidate for the development of semimetal-based photodetectors beyond their graphene-based counterparts.

In this work, a large-area HgTe film with high crystallinity and mobility on a CdTe substrate is successfully grown via the molecular-beam epitaxial (MBE) method, and the light-resonator has been exploited in terms of a bow-tie antenna to achieve strong electromagnetic coupling of THz radiation, which converts the incident electromagnetic waves into a direct current by triggering the photothermoelectric (PTE) effect in the HgTe channel. The excellent broadband responsivity (0.36 A W⁻¹), fast response, high-resolution THz imaging, and blackbody response are all demonstrated in our bow-tie antenna-based HgTe detector. Our results provide a guarantee for sensitive, ultra-fast, and large-area application based on controllable HgTe-film growth.

It is well known that there exist a series of difficulties in the growth and device fabrication of Hg-based materials. The adhesion coefficient of Hg on the surface of the crystal material is very low, and resulting in a high vapor pressure at room temperature. Therefore, the growth temperature of the Hg-based material in the MBE system should be controlled precisely. When the growth is completed, the wafer should be cooled below 80 °C in a Hg atmosphere. Otherwise, large quantities of Hg desorb in during cooling, resulting in serious surface degradation. Meanwhile, the selection of the buffer layer and crystal phase is also significant. Both HgTe and CdTe belong to the zinc blende (ZB) structure, and the lattice mismatch is only 0.3%. In theory, CdTe is an ideal lattice matching buffer layer. However, the lattice mismatche between the buffer layer and the substrate is as high as 14.7% with the lattice constants of 6.48 Å for CdTe and 5.65 Å for GaAs. The large lattice mismatch results in the growth of CdTe along different crystalline orientations on the GaAs surface. Improper crystalline orientation matching will produce a large number of stacking faults and twin crystals at the CdTe-GaAs interface, which then extend to the CdTe buffer layer and affect the quality of the HgTe crystal. In the epitaxy of HgTe wafers, our substrate and buffer layer are both [211]-crystal orientation. From the perspective of epitaxial relations, the orientation of CdTe [211]-crystalline is parallel to that of GaAs [211] in the epitaxial plane, which, therefore, greatly reduces the dislocation defects caused by stress release. On the other hand, the bond energy of Hg–Te is only 1/3 that of the Cd–Te bond. HgTe crystal may be damaged in the process of micro/nano fabrication. Compared with traditional dry etching, inductively coupled plasma etching (ICP) is used for device preparation, which is based on the chemical reaction of plasma gas molecules with the material, and less damage is caused to the material. All of the above strategies ensure the success of epitaxial growth and device preparation. In ordinary scientific research, we focus on the exploration and study of the physical mechanisms in individual devices. However, all scientific research will eventually move out of the laboratory and serve practical applications. On the one hand, the market demands large-scale production of large wafers for integrated chips. On the other hand, a single device is only suitable for the exploration of physical mechanisms. In order to further pursue excellent performance, linear array and focal plane array (FPA) detector occupy the majority of practical applications. Therefore, the large size and high quality HgTe wafers we grow lay the foundation for the preparation of FPA devices in practical applications.

Figure 1 displays a high-quality single crystal of HgTe films with a diameter of 3 in., which were grown via the MBE method on a CdTe (211)/GaAs substrate (see more details in Sec. IV for Materials Preparation). The crystallographic structure of HgTe/CdTe is presented in Fig. 1(a) (the green, yellow, and blue balls denote Hg, Te, and Cd atoms), and the small lattice mismatch between HgTe and CdTe substrate (0.3%) enables the successful growth of single crystal HgTe film on CdTe/GaAs. A scanning electron microscopy (SEM) image of the HgTe/CdTe interface and corresponding energy-dispersive spectroscopy (EDS) elemental mapping images are provided in Figs. 1(b) and 1(c), which reveal a perfect HgTe/CdTe heterointerface accompanied by the uniformity of Hg and Te components across the whole sample. The high-resolution X-ray Diffraction (HRXRD) spectrum of a HgTe film with a thickness of 600 nm is shown in Fig. 1(d). There are two sharp diffraction peaks of (211)-oriented CdTe and HgTe at 70.97° and 71.49°, respectively, corresponding to the 422 crystal plane,⁴⁸ validating the excellent crystalline quality of the as-grown HgTe film. The Raman spectra [see Fig. 1(e)] of HgTe films with the varied thicknesses (240, 400, and 600 nm) show two characteristic Raman peaks at 118 and 137 cm⁻¹, corresponding to the TO and LO phonon mode,⁴⁹ and the full width half maximum (FWHM) of the two modes are 12.5 and 8 cm⁻¹, respectively. The microstructure of the HgTe/CdTe interface is revealed by the high-resolution transmission electron microscope (HRTEM) image in Fig. 1(f). No dislocations, stacking or twin faults, at the HgTe/CdTe interface are found, indicating a very smooth transition interface plane on the atomic scale. Correspondingly, selected area electron diffraction (SAED) in Fig. 1(g) further validates the excellent interface (see more details in Sec. IV for materials characterization).

The electron mobility of HgTe film with a thickness of 600 nm is 2.7 × 10⁴ cm²/V s (300 K) and 4.5 × 10⁴ cm²/V s (77 K) by using the Van der Pauw Hall measurement (see more details in the Hall Test section, supplementary material), where the better crystal quality and electrical properties (such as lower carrier concentration and higher mobility) are revealed compared to the other two thicknesses of wafers and thus performed in THz device preparation. In addition, to achieve efficient conversion of electromagnetic wave into electrical signal, a bow tie antenna structure is engineered numerically by using the finite element software-COMSOL with a particular design for both strong THz-focusing at the channel and readable electrical-output. Figure 2(a) displays schematically the HgTe-based bow-tie antenna THz detector with two arms of antenna connected electrically for measurements (see more details in Sec. IV for device fabrication). The optical image of the whole bow tie antenna detector with a 6 µm channel-length and simulated electric-field profile under THz radiation in Fig. 2(b) clearly verifies the efficiency of THz coupling, showing strong polarization-dependence in Fig. 2(c). Meanwhile, the far-field enhancement with varied channel-lengths at different frequencies shown in Fig. 2(d), substantiating evidently stronger light–matter interaction, can be achieved by reducing the antenna-gap.

The photoelectronic process of bow-tie antennas-based HgTe THz detector is portrayed in Figs. 2(e) and 2(f) schematically. Due to the extremely low energy of THz waves, it is hard to excite the electrons effectively in traditional semiconductor materials, which could be solved by the inherent negative bandgap characteristics of HgTe topological semi-metals. There exists a strong coupling between THz waves and interband electrons, benefiting from the innate advantages of Dirac fermions.⁵⁰ When the external bias voltage is absent, the electrons are in thermodynamic equilibrium, filling the Dirac cone below the Fermi level. Meanwhile, the electrons are driven into non-equilibrium states under the THz radiation, making it feasible to produce a photocurrent. As the external bias voltage is traversing across the channel, the interface barrier will become asymmetric to facilitate the carrier transport process, and the nonequilibrium carriers will be driven unilaterally to form the bias-dependent photocurrent.

In order to better elucidate the underlying mechanism of a bow-tie antenna-based THz detector, the band diagram at the junction region is depicted in Fig. 2(f). Due to the different work functions between electrodes (Au) and HgTe film, the band of HgTe will bend at the metal-material interface. Benefiting from the large Seebeck coefficient and excellent thermoelectric characteristics, the HgTe-based PTE THz detector possesses better response time and high frequency characteristics than electronic detection based on the rectifier effect or nonlinear mixing mechanism. When the electrons/holes are heated up by strong optical-field coupling, depending on the bias conditions, there are two regimes of detector operation that can be distinguished. While the bias voltage is absent, a weak net photocurrent is observable under the THz radiation, which is probably caused by the residual symmetry-breaking artifacts or imperfections. As the external bias voltage is scanned across the channel, the interface barrier of Au–HgTe will become asymmetric, leading to the Seebeck coefficient difference and the non-zero photocurrent production that is changeable by the scanning bias voltage.⁵¹ Alternatively, a large photoresponse at zero bias is achievable by engineering an asymmetric short channel detector with tailored distribution of non-equilibrium carriers.

Following the above understandings, we proceed with our experiment by quantifying the performance of the prepared HgTe-based THz photodetector with different channel lengths and symmetry. The Seebeck coefficients of HgTe films with a thickness of 600 nm are measured to be around −135 µV/K at 300 K by the Seebeck coefficient measurement system (see more details in the Seebeck coefficient measurement section, supplementary material). A large Seebeck coefficient is essential to ensure the efficiency of photothermal detection, aside from the strong interband excitation. The schematic magnification of the bow-tie antenna-coupled symmetric long channel detector structure is illustrated in Fig. 3(a), where HgTe serves as the channel connecting the electrodes. The THz wave is coupled to the bow-tie antenna to complete the detection. To validate the underlying mechanism of the bow-tie antenna-based HgTe THz detector, an electromagnetic wave source (Agilent E8257D) at 30 GHz is used to excite the non-equilibrium carriers, and the results are shown in Figs. 3(b) and 3(c). Linear power-dependence of photoresponse at different bias voltages is observed, which is in accordance with the above analysis that the direction of photocurrent is reversible.

Furthermore, the fast-pulsed shape of photocurrent is well preserved in Fig. 3(c) at 50 or 100 mV, indicating the excellent performance and good signal-to-noise ratio of the detector. In order to accurately determine the response time, an ultra-fast light-modulator connected with a 100 GHz IMPATT diode is utilized [Fig. 3(d)]. The pulsed-photocurrent at different bias-voltages with the extension of the rising/falling edges is shown in Figs. 3(e) and 3(f), respectively. The rising and recovery time-scales are around 3 and 4 µs, which are defined as the time required to reach 90% of the maximum photocurrent and the time needed to drop to 10% of the maximum photocurrent.⁵² 

Actually, the sensitive and stable THz response could be retained across a wide-frequency range from 0.02 to 0.3 THz in our HgTe-based detector (see Figs. S2 and S3 in the supplementary material). The photocurrent responsivity (R_I) and noise equivalent power (NEP) vs the bias voltage at 0.03, 0.10, and 0.30 THz are recorded in Figs. 3(g)–3(i). The responsivity, which is derived from the formula R_I = I_ph/P_inS_d, is one of the significant norms to evaluate the capacity of a detector. I_ph is the detector photocurrent, and P_in and S_d refer to the irradiated power densities of the THz wave and the irradiated device area (140 × 420 µm²),⁵³ respectively. Under zero bias voltage, the responsivity reaches 554.3, 4.8, and 0.65 mA W⁻¹ at 0.03, 0.1, and 0.3 THz, respectively. It is worth noting that the corresponding responsivity could be further improved to 25.6, 0.48, and 0.038 A W⁻¹ when the voltage is fixed at 100 mV. The NEP, which is defined as the minimum incident power required when the signal-to-noise ratio reaches unity within the 1 Hz bandwidth, has been extracted from the ratio ν_n/R_I.⁵⁴  ν_n is the root-mean-square of noise rooted from thermal Johnson–Nyquist noise (ν_th) related to the non-zero resistance detector and shot noise due to bias current (ν_b), in the form of ν_n = (ν_th + ν_b)^1/2 = (4k_BT/r + 2qI_d)^1/2, in which k_B is the Boltzmann constant, T is room temperature, r is detector resistance (r = 107 Ω is obtained from the measurement), q is the elementary charge, and I_d is a direct current. The system-specific flicker noise (ν₁/f) is dominated at low-frequency (less than 150 Hz), and it has been ignored. When the bias voltage is zero, the value of NEP is large since the configuration of antennas is geometrically symmetric and the photocurrent depends weakly on the excitation of the THz wave. Nevertheless, when the bias voltage increases, the non-equilibrium carriers are transported under the action of an external direct current electric field, and the required threshold power of the THz wave decreases rapidly. Therefore, a smaller NEP value is obtained. Under 0.03, 0.10, and 0.30 THz-irradiation, the measured NEPs are 0.008, 2.73, and 28.6 nW/Hz^1/2 at 0 mV bias and 0.8, 44.8, and 560 pW/Hz^1/2 at 100 mV bias voltage, respectively. Benefitting from the nature of the negative bandgap and oxidation resistance in HgTe, the symmetric bow-tie antenna-assisted HgTe-based THz detector exhibits excellent detection stability, as could also be inferred from the photoresponse at 3.3 THz from the QCL laser after two months [see Fig. S2(b) in the supplementary material].

It is worth mentioning that, in terms of the optical coupling with the magnetic dipole (MD) mode, the symmetric cubic resonator across the 6 µm HgTe channel in the aforementioned detector supports all magnetic dipole modes. Unfortunately, the MD mode Mz does not couple to a plane wave with the k-vector normal to the xy-plane; only the Mx or My mode could be directly excited by this plane wave. Therefore, perfect absorption using only MD modes in symmetric cubic resonators is impossible.⁵⁵ In order to further improve the sensitivity of the HgTe-based THz detector, short channel bow-tie antennas formed by incorporating an asymmetric cubic resonator and a 2 µm gap-channel are also implemented.

Figure 4(a) illustrates the position-dependence of the Seebeck coefficient for scanning a THz spot. The fact that the S(x) achieves its maximum with opposite signs at two junctions, combined with a value close to zero with the THz spot focused at the middle of the channel. The simulated localized electric-field enhancement (for the situation of a THz spot focused at the middle of the channel) is shown in Fig. 4(b) (see simulation method in supplementary material).⁵⁶ Here, we simultaneously excite the MD modes, in particular Mx and Mz, and thus realize perfect absorption in the HgTe channel through breaking the resonator cubic symmetry, i.e., a bow-tie antenna-based asymmetric cubic resonator. Compared with the previous symmetric long channel bow-tie antenna, the electromagnetic coupling gain in the HgTe channel increases by ∼11%, and the far-field enhancement reaches 140 at 0.03 THz. Figure 4(c) shows the physical process of hot electron transport excited by a potential gradient ∇V(x) with asymmetric THz-irradiation, in which a maximum value of S(x) could be completed. Corresponding profiles of electron temperature T(x), Seebeck coefficient S(x), and potential gradient ∇V(x) along the channel are all shown in Fig. 4(d). Hence, the detector equipped with an asymmetric short channel cubic resonator leads to a high temperature T(x) along the channel. The incident photons will heat the carriers in HgTe via strong electron–electron interactions. The hot carriers and the lattice finally reach thermal equilibrium via the scattering process between carriers and phonons. By taking advantage of the large Seebeck coefficient in HgTe, a potential gradient ∇V(x) = –S × ∇T(x) opens up the feasibility of unbiased operation, obliterating the excess thermal-noise and power-consumption.

Figures 4(e)–4(i) summarize the detector performance of a bow-tie antenna-based HgTe THz detector with an asymmetric 2 µm channel length. The comparison of photocurrent between the symmetric long channel and asymmetric short channel detectors in a single time-period of 0.1 THz with a 100 mV bias voltage is also displayed in Fig. 4(e), substantiating the improved performance and speed. To evaluate the detector performance at higher frequency operation, the transient photocurrent is also presented at 0.3 THz in Fig. 4(f), where the pulsed shape is well preserved with a better signal-to-noise ratio than that in Fig. 3(f).

In addition, the responsivity and NEP vs the bias voltage at 0.3 THz are summarized in Figs. 4(g) and 4(h), respectively. In Fig. 4(g), the responsivity shows a linear growth by varying bias voltage, which could be regarded as strong evidence that non-equilibrium carriers are driven by the external bias for THz detection. Notably, the asymmetric nanostructure near the HgTe channel manipulates the localized intensity-distribution of THz waves, whereas it is immune to the intrinsic symmetry of bow-tie antennas, as can be inferred from the far-field characters in a 2 µm channel [see Fig. S4(b) in the supplementary material]. Therefore, the detector shows excellent prospects for improving the ability to detect polarization. Figure 4(i) depicts both the responsivity and the NEP of the detector at different frequencies with a bias voltage of 100 mV for a symmetric long channel detector and an asymmetric short channel detector, respectively. Here, the photodetector acquires a higher responsivity (0.36 A W⁻¹) and a lower NEP (88.5 pW Hz^−1/2) at 0.3 THz. Compared with the previous detector, the sensitivity of the asymmetric short channel detector at 0.1 and 0.3 THz is improved by an order of magnitude.

The stronger zero-bias photocurrent mediated by the unilateral hot-carrier flow validates the success of the implementation of the asymmetric structure that strengthens the efficiency of charge-separation with a higher signal-to-noise ratio and lower power-consumption [see Fig. S5(b) in the supplementary material]. Even though it is only a preliminary attempt at HgTe-based THz detectors and far from an optimal detector design method. However, the obtained R_I and NEP have shown comparable properties to other existing materials. To prove the superiority of the HgTe-based photodetector, the room-temperature responsivity, NEP, and bandwidth of the detector are compared with those of the commercial photodetector and the two-dimensional based photodetectors in Table I. The 3 dB bandwidth corresponds to the frequency when the normalized responsivity decreases to 0.707 of its original value, and a value of 30 kHz is obtained for the HgTe photodetector as shown in Fig. S5(c) (see more details in the supplementary material). The responsivity and bandwidth of the HgTe photodetector are comparable to those of the most advanced commercial photodetectors, and the NEP is lower than that of photodetectors based on 2D materials and commercial photodetectors in the 0.3–3.3 THz region. In addition, it is worth mentioning that the photoresponse is also valid at higher frequency (e.g., 3.3 THz from a QCL laser) despite the size of the antenna being incommensurate with the incident photons, which is leveraged on the semi-metallic nature and ultrahigh carrier mobility of HgTe film.

To provide a concrete exploitation of HgTe for performing system application, for the first time, transmission imaging of an asymmetric short channel HgTe-based THz detector with a 2 µm channel is demonstrated via a THz quasi-optical system [Fig. 5(a)]. Here, 0.3 THz radiation is delivered onto the metallic letters “MBE” (the thickness is only 0.025 mm and is hand-cut from aluminum foil), which are concealed in a 3 mm thick cardboard and used as the target object with transmitted power recorded by the detector in unbiased mode. The 80 × 70 pixels image is obtained via scanning the metallic letters (20 ms integration time for each pixel), and the size of the metallic letters “MBE” is around 40 × 35 mm. Therefore, the resolution of the HgTe detector imaging for metallic letters is 0.5 mm⁻¹. Figures 5(b) and 5(c) show the optical picture of an object covered with cardboard and the image acquired by THz transmission imaging. Here, the words “MBE” can be clearly revealed by the detector, even though they are covered by a 3 mm thick cardboard and invisible to the naked eye [also a metal key transmission imaging was performed in Fig. S6(b), supplementary material]. To further validate the sensitivity of HgTe photodetectors at THz band, blackbody-response measurements are performed at 30 and 300 K for asymmetric short channel detectors. All responses show linear growth with varying bias voltages. The results suggest that HgTe is one of the competitive candidates for sensitive room temperature THz photodetectors (see more details in the Blackbody Measurement section, supplementary material). Therefore, the controllable synthesis of high-quality HgTe films provides feasibilities to achieve high-sensitive, fast response photodetectors working at THz band for large area imaging. Our work lays a solid foundation for the preparation of FPA devices for practical application.

In this work, large area HgTe-based TSM thin-films with high crystallinity, high carrier mobility, as well as good ambient-stability were successfully grown via the MBE method. Here, for the first time, bow-tie antenna-based photodetectors have been well-designed by taking advantage of the gapless nature and superior carrier-transport of HgTe-film, allowing for broadband photodetection down to the far-infrared/terahertz regime. Even with its first implementation, the HgTe-detector already exhibits superior performance with capabilities of fast response, low-noise, as well as high-sensitivity across the THz band. Our results reveal the efficiency of charge-separation mediated by the photothermoelectric process and polarization sensitive THz detection in the HgTe detector. By breaking the symmetry of the bow-tie antenna-based cubic resonator, responsivity (0.36 A W⁻¹) and fast response at 0.3 THz can be obtained in low-power mode, enabling well-resolved, high-quality imaging for macroscopic objects. Our work opens up a new avenue to explore the fascinating properties of TSM systems led by the strong light–matter interaction and bulk-wave functions in HgTe, with the target of large-area and fast imaging applications, promising superb impacts on THz photonics.

HgTe films (240, 400, and 600 nm) were grown on the GaAs(211)B substrate with a CdTe buffer layer via a DCA 450 MBE system. Firstly, the GaAs substrate was deoxidized in an As atmosphere, and then a CdTe buffer layer was grown on the GaAs(211)B substrate. The epitaxy temperature of HgTe film was 173 °C, and the growth rate was around 40 nm/min. After the completion of the growth, the wafer was annealed in situ at 130 °C for 30 min in the Hg atmosphere.

The morphology of as-grown HgTe samples was analyzed via SEM (ZEISS sigma 500) operated at 10 kV and a Leica DM2700M optical microscope. The EDS Mapping (ULTIM MAX) was utilized to investigate the element composition of HgTe films. Raman spectra were obtained from a LabRAM HR Evolution Raman spectrometer with a 532 nm wavelength. The HRXRD (Bruker D8 Discover) pattern was measured to characterize the phase purity and crystal structure of the samples. The atomic structure and crystallization of the samples were evaluated by HRTEM and SAED. Herein, the cross-section sample was fabricated by dual beam FIB (Thermo Fisher Helios G4). The electron mobility of the HgTe films was obtained from the Ecopia HMS 3000 Hall Measurement System (see more details in the Hall Test Section, supplementary material).

In the process of device preparation, the device was prepared via ultraviolet lithography and inductively coupled plasma (ICP) to form HgTe micro-channels. A double-layer photoresist lift-off technology was applied for bow-tie antenna patterning, where magnetron sputtering evaporation technology was exploited in metal electrode fabrication (Ti/Au = 20/300 nm) deposition.

We performed photoelectric measurements on a 600 nm HgTe-based THz detector at room temperature in ambient conditions, where the Keithley 4200 semiconductor parameter analyzer is employed in the electrical data collection. The recorded signals were passed to the oscilloscope for display and then amplified by a lock-in amplifier (LIA) and preamplifier. The value of I_ph can be obtained from the photo signal on the LIA via the relation I_ph = 2.2 LIA/Gn, where Gn is the gain factor and 2.2 accounts for the sine wave Fourier component. A tunable signal-generator (Agilent E8257D) with a frequency range of 0.02–0.04 THz is used for parameters calibration. The higher frequency (0.1 and 0.3 THz) THz sources are generated from an IMPATT diode connected with a VDI multiplier (WR 2.8). The power densities of 2.5 mW cm⁻² (0.03 THz), 1.05 mW cm⁻² (0.1 THz), and 1.05 mW cm⁻² (0.3 THz) are calibrated by a Golay cell. The quantum cascade laser (QCL) system possesses a power density of 2 mW cm⁻² at 3.3 THz and is calibrated by a THz power meter (TK100). The responsivity is derived from the formula R_I = I_ph/P_inS_d, where I_ph is the device photocurrent and P_in and S_d refer to the irradiated power density of the THz wave and the irradiated device area, respectively. The value of NEP is extracted from the ratio ν_n/R_I. ν_n is the root-mean-square of noise arises from the thermal Johnson–Nyquist noise (ν_th) related to the non-zero resistance device and shot noise due to bias current (ν_b). In the response time measurement section, an ultra-fast light modulator connected with a 100 GHz IMPATT-diode is utilized to achieve the rapid modulation of the THz source.

The supplementary material is available free from the author.

This work was supported by the National Key R&D Program of China (Grant No. 2021YFB2800702), the National Natural Science Foundation of China (Grant Nos. 12027805, 61974166, 61875217, and 11991060), Zhejiang Lab (Grant No. 2021MB0AB01), the Shanghai Municipal Science and Technology Major Project (Grant No. 2019SHZDZX01), and the Hunan Provincial Natural Science Foundation of China (Grant No. 2021JJ20080).

The authors have no conflicts to disclose.

Xinrong Zuo: Conceptualization (lead); Data curation (lead); Formal analysis (lead); Investigation (lead); Software (lead); Validation (lead); Visualization (lead); Writing – original draft (lead). Chenwei Zhu: Investigation (equal); Methodology (equal); Resources (equal). Chenyu Yao: Formal analysis (equal); Investigation (equal); Methodology (equal). Zhen Hu: Conceptualization (equal); Data curation (equal); Formal analysis (equal). Yan Wu: Conceptualization (equal); Investigation (equal); Resources (equal). Liuyan Fan: Conceptualization (equal); Resources (equal); Visualization (equal). Zhifeng Li: Resources (equal); Validation (equal). Jun He: Resources (equal); Supervision (equal). Xiaoshuang Chen: Supervision (equal). Pingping Chen: Funding acquisition (lead); Resources (lead); Supervision (lead); Writing – review & editing (lead). Xiaoming Yuan: Funding acquisition (equal); Resources (equal); Supervision (equal); Writing – review & editing (equal). Lin Wang: Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Methodology (equal); Resources (equal); Writing – review & editing (equal). Wei Lu: Resources (equal); Supervision (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.