Fully Integrated THz Schottky Detectors Using Metallic Nanowires as Bridge Contacts

This paper investigates fully integrated Terahertz (THz) Schottky detectors using silver (Ag) metallic nanowires (NWs) with 120 nm diameter as bridge contacts for zero-bias operating THz detectors based on highly doped Gallium Arsenide (GaAs) and Indium Gallium Arsenide (InGaAs) layers. The combination of InGaAs and metallic NWs shows improved performance at zero-bias than a GaAs based detector with a simulated capacitance of 0.5 fF and a series resistance of 29.7<inline-formula> <tex-math notation="LaTeX">$\Omega $ </tex-math></inline-formula>. Thus, the calculated maximum cut-off frequency of 2.6 THz was obtained for a NW contacted vertical InGaAs THz detector. Initial THz measurements were carried out using a common THz setup for frequencies up to 1.2 THz. A responsivity of 0.81 A/W and a low noise-equivalent power (NEP) value of 7 pW/<inline-formula> <tex-math notation="LaTeX">$ {~\sqrt {Hz}}$ </tex-math></inline-formula> at 1 THz were estimated using the measured IV-characteristics of the zero-bias NW-InGaAs based THz Schottky detector.


I. INTRODUCTION
THz (100 GHz -10 THz) nanotechnologies represent a new approach to material science and engineering, as well as for design of new devices and processes for the fabrication of compact high-frequency devices. THz detectors, such as Schottky diodes, were intensively investigated in the last 30 years [1]- [3]. Further good and reliable THz detector concepts include complementary metal-oxide-semiconductor (CMOS) [4] and bolometers with some drawbacks. The most sensitive bolometers require a low temperature environment, which increases the cost and the fabrication complexity. CMOS detectors are frequently implemented with resonant antennas in order to tune out the device reactance, limiting the bandwidth [5]. THz Schottky detectors need reliable submicron anode contacts with a low capacitance. For future generations of THz systems, an improvement of the detection sensitivity and system resolution is a must in many applications. In particular, the THz imaging systems suffer from the considerably low power of THz sources as compared to other spectral domains requiring very sensitive detectors. Further potential applications in non-destructive testing of small cell The associate editor coordinating the review of this manuscript and approving it for publication was Jenny Mahoney. structures in biomedical research, skin cancer detection or in safety engineering can profit from improved detectors. Over the last two decades, tremendous advances have been made in devices ranging from the introduction of new operation concepts to new material choices and improvement in fabrication processes [6]- [12]. Examples of them are High-Electron Mobility Transistors (HEMTs) which use the rectification effect by the Dyakonov-Shur instability. Even HEMTs with much longer gate lengths and cut-off frequencies in the GHz range enable detection of THz waves [13].
To date, Schottky diodes are still the most sensitive detector concept for direct detection of THz radiation at room temperature. This kind of detector combines a high responsivity with very short response time and compact dimensions. It shows excellent performance in the sub-mm wavelength range. The output signal is proportional to the incoming power over a wide range of powers. AC-coupled Schottky receivers can measure the field intensity of individual terahertz pulses [14], [15]. Nagatsuma et al. used a Schottky receiver to build a wireless communication system towards 100 Gbit/s at 100 GHz [16]. Typically, the best performance at room-temperature conditions can be achieved for waveguidecoupled Schottky diode detectors with responsivities ranging from 4000 V/W at 100 GHz to about 400 V/W at 900 GHz and the minimum NEP of about 1 pW/

√
Hz at 100 GHz [2]. However, the frequency bandwidth of waveguide-based detectors is limited by the physical size, whereas for many applications such as THz spectroscopy, it is highly desirable to achieve wider operating bandwidths going to more broadband radiation coupling schemes [17]. In combination with fast response time this contributes to the emergence of new application areas [26]. The low NEP of Schottky detectors can allow the passive detection of low power down to black body radiation. It was also shown that alternative rectifier methods based on InP/InGaAs heterostructures [18] with Fermi-level managed barriers and heterojunction low barrier diode based on AlGaInAs [19] can be used for low noise, lowlevel radiometric detection with a responsivity of 1110 V/W at 300 GHz and 1700 V/W at 220 GHz, respectively. Ito and Nadar achieved a NEP of ∼ 1.3 pW/

√
Hz at 300 GHz, respectively. Other publications have shown a modified Schottky barrier height on semiconductors due to reduced Fermi-Level pinning with nanoparticles from few nanometer to > 20 nm [3]. Similar to this approach, our metallic nanowires reduce the Schottky barrier height. This property and the low capacitance lead to efficient zero-bias operation and low noise THz detection capability.
In this contribution, the conventional Schottky contact was replaced with a NW to achieve a new Schottky diode for THz detection. The use and further investigation of these emerging and extremely demanding concepts with metallic NWs on semiconducting mesas (SC-Mesa) constitute the significant novelty of this work over earlier efforts in the field with standard contacts [5]. The new devices have benefits besides the reduction of the fabrication complexity (no need for high resolution lithography for small contact in nanometer range, like electron-beam lithography), enhanced electric field at the Schottky contact with reduced Schottky barrier height [20]- [24]. Furthermore, NWs can be aligned using several methods: for example by nanostructure grooves [25], with nano-tweezers [26] or using the dielectrophoresis (DEP) technique [27]. In this work, the metallic NWs were aligned using DEP. These THz Schottky detectors based on GaAs and InGaAs were simulated, fabricated and characterised using Ag-NWs with 120 nm diameter as bridge contacts and were compared to standard detectors (SD) fabricated with evaporated finger contacts.

II. SIMULATIONS
The device capacitances were simulated for different configurations using CST EM Studio. In the CST simulation, a NW with a diameter of 120 nm and a contact length of 2 µm was used for the capacitance calculation. The mesa has a width of 7 µm and a length of 4 µm and contents of two layers. The bottom layer has a height of [1]µm with n ++ = 5 · 10 18 cm −3 and the top layer has the height of [0.1]µm with n + = 1 · 10 18 cm −3 , respectively. The thickness of the antenna arms (Ant 1 and Ant 2, Fig. 1) and the ohmic contact was 120 nm. The parameters were set using CST EM Studio environment which can simulate the FIGURE 1. CST simulation of the potential distribution of the GaAs Schottky diode using silver NW as air-bridge contact. Ant1 and Ant2 are the inner parts of both antenna arms that were connected to the Schottky diode. The same configuration was also used for the InGaAs diode. The total simulated capacitance of both devices was 0.5 fF.
potential distribution of the device. The electrostatic solver was used which can provide the capacitance of the module. The capacitance of the NW and of the SD based diodes was simulated as 0.5 fF and 2 fF, respectively, which can be used to determine the maximum cut-off frequencies. Figure 1 shows the CST simulation results of the potential distribution, from which the capacitance was calculated. The simulated capacitance of the InGaAs diode was almost equal to the simulated capacitance of the GaAs diode due to the same geometry. Furthermore, numerical modelling of the device was performed using COMSOL Multiphysics. For the COMSOL simulation, the semiconductor module was used as a setting environment at a temperature of 300 K. The model under study (inset of Fig. 2) consists of two sandwiched n-doped layers of GaAs or InGaAs with the same thicknesses and doping concentrations used in the CST simulation. The impurity concentration of the semiconductor material was set in the doping module for the Schottky contact. A Ag-NW was placed on the top of the semiconductor structure as Schottky contact with an estimated width and length of 30 nm (only the bottom curvature of the NW was used) and 2 µm, respectively. Then, a voltage between 0 V and 1 V was applied. This sweep voltage was set in the auxiliary sweep under study/stationary. Figure 2 represents the simulated IV curve obtained with our model. VOLUME 9, 2021

III. FABRICATION AND RESULTS
In order to investigate the room temperature THz detectors, four detector types were fabricated: Two detectors were based on standard contacts and the other two were based on Ag-NW contacts. A small NW diameter of 120 nm was used to have better contact in comparison to a thicker NW because of its mechanical flexibility on the active layer. The finger contact was evaporated using Ti/Au (20 nm/120 nm) with a finger width of 1 µm and a length of 2 µm. All detectors were fabricated with highly doped epitaxial layers with the same material configurations mentioned in the simulation part. The semi-insulating (s.i.) substrates have a thickness of 300 µm (s.i.-GaAs for n-GaAs, s.i.-InP for n-InGaAs). Figure 3 (a) illustrates the diode RF equivalent circuit used to calculate the power lost due to the impedance mismatch of the device and the antenna and (b) the 3D close-up of device structures. The ohmic contact of the diodes was fabricated with Ni/AuGe/Ni and annealed at 420 • C. A silicon dioxide (SiO 2 ) passivation layer under the antenna is used to avoid any leakage current. The antenna metallization is made of Ti/Au with thickness of 20 nm and 120 nm, respectively. Afterwards, an additional passivation layer was deposited on top to prevent any short circuit of the antenna with the aligned NWs. Only the part between the mesa and one antenna arm was opened to contact the Schottky diode with the NW. DEP was then used for the selective alignment. The DEP uses an AC electric field to manipulate neutrally charged particles in a solution and offers a promising method to attract NWs onto predefined electrodes. The process parameters depend on the volume, length and material of the NWs. A potential of 2 Vp-p with a square-wave and a frequency of 30 kHz was applied to align the NW. The NW diameter and the controlled alignment determine the junction and parasitic capacitance as well as the series resistance. Figure 4 shows the SEM images of the Schottky contacts using DEP technique. Figure 4 (a) illustrates a single Ag-NW. In Fig. 4 (b), a NW was contacted with a part of an antenna. A comparison between aligned single NW Fig. 4 (c) and multi-NWs as a multi-finger Schottky contact Fig. 4 (d) are also shown for selective alignments using DEP. Figure 5 shows a SEM image of the fabricated detector with an aligned NW in the middle of the antenna structure. All other NWs have no influence due to surface passivation. These NWs are leftovers of the process due to the use of the DEP. The alignment of the DEP technique is accurate to contact many NWs in a small area. On one hand however, the NWs are in suspension and this medium contains tens of NWs per device area which makes the alignment of a single NW difficult, therefore few NWs are spread together on the whole structure. On the other hand, the DEP force depends on the fluid flow and the interaction between the NW and the surfaces around which impaired the accuracy of this method. An alignment accuracy of 500 nm was achieved in the way that several NWs lie side by side without causing a short circuit.

A. IV-CHARACTERISTICS OF THE DIODES
First of all, both SD-GaAs and NW-GaAs based detectors were characterized. Figure 6 shows the IV-characteristics of these diodes. In addition, we changed the active material to InGaAs because of its higher electron mobility, smaller band gap and smaller Schottky barrier height. The SD-InGaAs  and NW-InGaAs based detectors were characterized as well (Fig. 7). Figure 8 shows a comparison of the NW-GaAs and the NW-InGaAs based detectors. The NW-based devices were characterised with one single NW with a diameter of 120 nm.
Similar to Casini et al. [3], the metallic NW increases the electrical field at the Schottky contact to achieve higher current at lower voltage than the standard Schottky diode with evaporated contact. A parasitic thin oxide layer under the Schottky contact is less critical for the NW due to the field enhancement. The modification at the forward bias with relatively high current density creates an effective channel, unlike for the evaporated contact. This shows that a semiconductor surface treatment is not as critical as for the evaporated Schottky contact. Unlike the simulated structure which has a contact length of 2 µm, the fabricated diodes have an estimated contact length of < 1 µm due to the bending of the NW and the passivation layer (to protect a short circuit between the NW and the ohmic contact), so the lowermost part of the NW curvature does not contact the active material well. Therefore, the forward current is smaller for the fabricated devices. Generally, the simulations show an ideal scenario, therefore the results look better. There is probably still 0.5 fF of parasitic capacitance in the system, which is not captured by the simulation. In the future, the NW will be aligned closer to the designed structure as the simulation shows to achieve a longer contact on the mesa for a higher current.

B. EXTRACTION OF DIODE PARAMETERS
The resulting DC IV-characteristics show typical diode shapes of the thermionic emission model given by with the magnitude of electron charge q, the Boltzmann constant k B and absolute temperature T of 300 K. A * indicates the Richardson constant of [8.6] cm −2 K −2 A. The series resistances R s , the barrier height φ, and the ideality factor n of both detectors were extracted using the Cheung method [28]. This method enables the determination of the Schottky diode parameters from a single forward bias IV-characteristic at a certain temperature. From equation (2) we get equations (3) and (4) where β = q/k B T . The plot between d(V ) d( lnI ) and the current I gives a straight line (y = ax + b), where R s can be indicated from the slope and the y-intercept of this plot VOLUME 9, 2021 provides the n value. The intercept of the linear fit of the plot between H (I ) and I gives φ. We have extracted different diodes and found out that n was about 1.2 -1.49. These different values are attributable to the varying alignments of the Schottky contacts. In comparison to the extracted barrier height using the thermionic emission model found in the literature as 0.56 eV [29] and 0.24 eV [30] for the SD-GaAs and SD-InGaAs, respectively, the lowest determined effective barrier height according to equation (5) for our NW-GaAs and NW-InGaAs based diodes was 0.4 eV and 0.21 eV with the saturation current I s . The saturation current was about 1 nA and 138 nA for NW-GaAs and NW-InGaAs, respectively. The enhanced electrical fields for the NW based diodes can decrease the Schottky barrier height related to the image-force barrier lowering [31]. The effective area A eff was defined as an approximation of 6 · 10 −10 cm 2 because only the lowermost part of the NW curvature [width 30 nm times length 2 µm] under the NW was considered as contact area. From Sze et al. [32], the Schottky junction capacitance was calculated as C j = 0.48 fF, which fits with our simulated one of 0.5 fF. Using this value of the junction capacitance and the series resistance R s of 29,7 (with an error of ±30%) in f max cut−off = 1/(2πC j R s ), we estimate a maximum cut-off frequency of 10.7 THz for the vertical contacted NW-InGaAs diode. However, the junction resistance R j and antenna resistance R a influence the actual cut-off frequency. Using equation (6) [33] the calculated cut-off frequency for the vertical contacted NW-InGaAs diode at zero-bias operation is obtained as 1.7 THz with C j = 0.5 fF, R j = 420 , R a = 70 and R s = 29.7 , where R a is the average radiation resistance of the logarithmic-periodic antenna according to simulations. For the same diode at near zero-bias operation at 0.1 V, a cut-off frequency of 2.6 THz is expected with C j = 0.5 fF, R j = 214 , R a = 70 and R s = 29.7 . This is due to the decreased R j with increased bias voltage. At frequencies greater than 1 THz, the frequencydependent skin effect should be taken into account [34] due to the increase of the series resistance which in turn reduces the cut-off frequency. In the Schottky diode, the skin effect δ occurs in the metallic NW and in the semiconductor buffer layer. The skin depth can be described as follows [34], [35] where ω is the radial frequency, µ the permeability and σ the conductivity of the NW. For Ag-NWs at a frequency of 1 THz, the skin depth is 63.42 nm. Therefore a NW diameter of 120 nm was used to avoid the skin effect. This means at frequencies higher than 1 THz, a part of the NW is free from the AC currents.

C. THz RESPONSIVITY AND NEP
In order to highlight the peculiarity of these devices, a comparison of the quasi-DC responsivity (the numerical second derivative of the IV curve) was performed for all detectors (Fig. 9), which is a direct indicator for the responsivity. This method was also used in [36] and showed an accurate agreement with the measured responsivity. Thus, it is useful to look at the quasi-DC responsivity as a function of the DC bias voltage hence as long the detected current I DC THz is in the small signal limit proportional to the second derivative of the IV curve. Consequently, the theoretical responsivity can be estimated by where η tot contains all losses, including the impedance mismatch of the antenna and the device η imp calculated using the equivalent circuit shown in Fig. 3 (a), as well as the optical losses η opt at the interface of the silicon lens and air. The relative comparison between our SDs with an evaporated contact ( Fig. 9 (a)) shows that the SD-InGaAs based Schottky diodes have lightly higher responsivity at zero-bias operation than the SD-GaAs based Schottky diodes and a higher quasi-DC responsivity with a factor of about 2 at 0.45 V (operation points 3 and 4). Moreover, the detector based on NW-InGaAs has significantly higher responsivity under zero-bias operation than the NW-GaAs with a factor of about 15 ( Fig. 9 (b) point 1 and 2). Therefore the NW-InGaAs was used to reduce the noise floor and the complexity of the setup for zero-bias THz detection. At this point of operation, the NW-InGaAs indicates a maximum possible responsivity of = 15 mA/V 2 · R a = 1.08 A/W. Using equation (8) with the impedance losses η imp of approximately 25%, we expect a responsivity of 0.81 A/W and a low NEP value of 7 pW/ √ Hz at 1 THz using the measured IV-characteristic of the NW-InGaAs based THz Schottky detector. The NEP value was calculated by Note that the losses of approximately 25% were calculated without taking the optical losses η opt into account.

D. THz MEASUREMENTS
A common THz setup with a LTG-GaAs lateral photomixer with conventional interdigital fingers was used [37], [38]. The THz beam was generated by using two tunable distributed feedback laser (DFB) with a wavelength of about 850 nm which were coupled by a 50:50 combiner. An optical laser power of the beat signal of 20 mW was focused onto the active region of the photomixer employing a polarizationmaintaining optical fiber. To accelerate the generated carrier, a DC potential of up to 12 V was applied. The photomixer was mounted on a hyper-hemispherical Si lens to couple and direct the THz radiation into two parabolic mirrors which in turn direct the THz beam to the Schottky detector. The dark current and the photocurrent were measured as 200 nA and 1.0 mA at 12 V, respectively. Commercial transimpedance amplifier and Lock-in technique for noise suppression were used to measure the relative THz signal. A Golay cell as a reference for the THz Schottky detector measurements was used. The initial THz measurement was carried out using the zero-biased NW-InGaAs based Schottky detector (Fig. 10). The THz detector was measured at 100 GHz, 300 GHz, 500 GHz, 700 GHz, 1.0 THz and 1.2 THz and showed an output voltage of 0.15-1.69 mV. The used photomixer was limited to about 1.2 THz. These measurements are still initial and need optimizations on one hand on the source side and on the other hand on the detector side as well.
In Table 1, a comparison of the cut-off frequency and NEP value of the NW-InGaAs based Schottky detector compared to the state-of-the-art results is given. The Schottky diode with a curvature contact presented in [3] operates at 550 GHz with a NEP of 0.5 nW/

√
Hz and needs an electron-beam lithography. With the Fermi-level managed barrier Schottky diode [18] and a heterojunction low barrier AlGaInAs based Schottky diode [19], a NEP of 3 pW/ Hz at 0.6 GHz presented in [33] uses the CNT as an active material but it includes only the diode without antenna, so the manufacture constraints for THz frequencies were only modelled. Another device based on phosphorus implanted Schottky diode on graphene/SiC presented in [36] shows an estimated NEP of 5 pW/

√
Hz at ≈ 90 GHz (limited by the RC circuitry). A Schottky diode with a standard evaporated contact presented in [39] reaches a frequency of about 0.8 THz with a NEP of 10 pW/ √ Hz. Further low Barrier Schottky diode with standard contact found in [40] has a NEP of 0.39 pW/

√
Hz at ≈ 90 GHz. The vertical NW-InGaAs based Schottky detector presented in this work reaches frequencies up to 1.2 THz (limited by the used source) with an estimated NEP of 7 pW/

IV. CONCLUSION
In this paper, a metallic NW configuration was used for THz Schottky detectors. The vertical NW-InGaAs Schottky detector showed much higher zero bias responsivity in comparison to the detector based on NW-GaAs and thus to the SDs. Both Schottky detectors were fabricated on n+GaAs or on n+InGaAs structure using the NW as bridge-contact on the semiconductor mesa. The alignment with dielectrophoresis with an alignment accuracy of 500 nm was applied to contact the NW based Schottky diodes. The simulated capacitance of the device was 0.5 fF which was used to calculate the cut-off frequency of about 2.6 THz. Several initial THz measurements up to 1.2 THz were carried out using a common VOLUME 9, 2021 THz setup. The detectors work at room temperature and achieved zero-bias operation. A responsivity of 0.81 A/W and a low NEP value of 7 pW/ √ Hz at 1 THz were estimated using the measured IV-characteristics of the NW-InGaAs based THz Schottky detector. THz detector devices based on dielectrophoretically aligned metallic NWs on low permittivity substrates can result in low cost, room temperature detector operation with high cut-off frequency. Thus, this nano-micro integrated THz detector is promising to fabricate highly sensitive, high frequency imaging and spectroscopy systems for material inspection, security and biomedical applications. where compound semiconductor materials and technologies were used to develop sensors and electric devices on the basis of GaAs, GaN, and ZnO as well as carbon nanotubes and graphene for applications in mechanical engineering and at high-frequency. He has 20 years of expertise in microelectronic devices and components. This includes the design, fabrication, and characterization of 3-D magnetic field sensors, pressure sensors, gas sensors, acceleration sensors, high-frequency diodes, and nanostructured field emission devices for high frequency generation. His work includes the experimental use of 1-D or 2-D materials, as well as the theoretical evaluation of related quantum size effects. He is currently working in the field of THz photomixers as well as CNT-and graphene-based field-emission devices for sensing, electron source and microwave applications. These resulted in several patents and numerous publications in internationally highly regarded magazines.