Graphene/AlGaN/GaN RF Switch

RF switches, which use a combination of graphene and two-dimensional high-density electron gas (2DEG) in the AlGaN/GaN system, were proposed and studied in the frequency band from 10 MHz to 114.5 GHz. The switches were integrated into the coplanar waveguide, which allows them to be used in any system without the use of, e.g., bonding, flip-chip and other technologies and avoiding the matching problems. The on-state insertion losses for the designed switches were measured to range from 7.4 to 19.4 dB, depending on the frequency and switch design. Although, at frequencies above 70 GHz, the switches were less effective, the switching effect was still evident with an approximately 4 dB on–off ratio. The best switches exhibited rise and fall switching times of ~25 ns and ~17 ns, respectively. The use of such a switch can provide up to 20 MHz of bandwidth in time-modulated systems, which is an outstanding result for such systems. The proposed equivalent circuit describes well the switching characteristics and can be used to design switches with required parameters.


Introduction
Radio frequency (RF) and terahertz dynamically reconfigurable multi-element devices often require electronic switches. Efficient switches are important components for the development of new communication, sensing, imaging, testing, and instrumentation systems. Applications of such switches include switched-beam reconfigurable antennas, polarization switching, multi-band receivers, transceivers, time division duplexing systems, and test circuits with multiple signal paths.
Graphene and transition metal dichalcogenides also attract attention for designing RF and terahertz switches and other devices [16,17]. A monolayer MoS 2 RF switch with a 0-50 GHz operating frequency range was reported in [18]. Several publications consider

Structure of Graphene/AlGaN/GaN Switch and DC Characteristics
The proposed hybrid structure includes a CPW with an embedded graphene/AlGaN/ GaN switch. A typical CPW consists of two ground conductors and a signal conductor line in the middle. In the proposed structure, the signal conductor line is made of Ti/Al/Ni/Au on the top of the AlGaN/GaN structure. The signal line is interrupted in the middle and the graphene gate device is placed in the gap between two parts of the signal line (see Figure 1). In the on state, the connection between two parts of the signal line is provided by 2DEG. The graphene layer in the middle of the device, above 2DEG, acts as a gate. The graphene gate is connected to the ground (GND) of the CPW; therefore, the device is controlled by the bias voltage applied to the CPW signal pads along with a high-frequency signal. As a result, the structure operates as a high-frequency single pole, single throw (SPST) switch. We used AlGaN/GaN epitaxial heterostructures grown by Metalorganic Vapor Phase Epitaxy (MOVPE) on a silicon carbide substrate. MOVPE growth was started from a 38nm-thick AlN nucleation layer on a commercially available 500-µm-thick semi-insulating SiC substrate. The next layer was 2.3 µm high-resistivity (HR) GaN buffer followed by a 0.7 µm unintentionally doped (UID) GaN layer. The AlGaN barrier consisted of: 1.2 nm AlxGa1−xN (x = 66%), 5 nm AlxGa1-xN UID (x = 28%), 10 nm AlGaN:Si (n~1.5·10 18 cm −3 ), and a 2 nm UID AlGaN layer. The whole heterostructure was covered by a 2 nm GaN cap As opposed to a conventional transistor-based design, where a metal gate is used, the properties of a graphene gate also change when the voltage is applied between the graphene gate and 2DEG. Therefore, the properties of both the 2DEG and graphene are controlled by the bias. In other words, the 2DEG also can be considered as a "gate" relative to graphene. Since CVD graphene is of p-type conductivity, a positive voltage on the 2DEG relative to graphene reduces the concentration of holes and increases the graphene resistivity [34]. Therefore, with properly designed graphene and two-dimensional electron gas at the AlGaN/GaN interface, a positive voltage on the 2DEG relative to graphene can completely remove the conductive layers in the central line gap of the CPW. This should improve the off-state characteristics.
We used AlGaN/GaN epitaxial heterostructures grown by Metalorganic Vapor Phase Epitaxy (MOVPE) on a silicon carbide substrate. MOVPE growth was started from a 38-nm-thick AlN nucleation layer on a commercially available 500-µm-thick semi-insulating SiC substrate. The next layer was 2.3 µm high-resistivity (HR) GaN buffer followed by a 0.7 µm unintentionally doped (UID) GaN layer. The AlGaN barrier consisted of: 1.2 nm Al x Ga 1−x N (x = 66%), 5 nm Al x Ga 1−x N UID (x = 28%), 10 nm AlGaN:Si (n~1.5 × 10 18 cm −3 ), and a 2 nm UID AlGaN layer. The whole heterostructure was covered by a 2 nm GaN cap layer. A schematic diagram of the fabricated heterostructures is shown in Figure 1. This is a typical structure of AlGaN/GaN high electron mobility transistors (HEMTs) for high-frequency applications [32].
The CPW processing was performed using a commercial laser writer system for lithography based on a 405 nm wavelength GaN laser source with a minimum 1 µm linewidth. The first step in the processing was 150 nm mesa etching provided by an Inductively Coupled Plasma-Reactive Ion Etching system. As a result of the etching, the 2DEG remained only under the signal line. Then, ohmic contacts were formed by thermal evaporation of Ti(15 nm)/Al(100 nm)/Ni(40 nm)/Au(50 nm) and rapid thermal annealing at 780 • C for 1 min under a N 2 atmosphere. Metallization for the ohmic contacts was deposited on the central line and interrupted in the middle. The ground lines were fabricated simultaneously. In order to be able to tune the width of the gap in the central line, the same Ti/Al/Ni/Au metal stack may be deposited at a later time in order to reduce the gap width. This metal stack was not annealed.
The last step was graphene transferring and its patterning. A chemical vapor deposition (CVD) graphene layer was deposited on the whole GaN-based wafer by the high-speed electrochemical delamination technique [35]. The detailed step-by-step procedure of the graphene delamination and transferring from Cu foil onto AlGaN/GaN can be found in [36]. Finally, graphene patterning was performed by oxygen plasma etching. Graphene remained in the central line gap and was extended to the ground pads as shown in Figure 1a. A cross-section of the structure is shown in Figure 1b.
During the CPW fabrication, the quality of the graphene layer was controlled with Raman spectroscopy. Figure 2 shows the Raman spectra of graphene transferred onto the AlGaN/GaN wafer, recorded with a Renishaw inVia micro-Raman system using a 532 nm frequency doubled Nd:YAG laser as an excitation source. The typical graphene peaks were observed: G mode at 1590 cm −1 and 2D mode at 2685 cm −1 , which are characteristic of the sp 2 hybridization of carbon. The full width at half maximum (FWHM) of the 2D band and the intensity ratio of the 2D/G peaks were used to determine the number of graphene layers and their quality. The FWHM of the 2D peak in Figure 2 was~34 cm −1 , which is typical for monolayer graphene. The ratio I 2D /I G was over 2, which is characteristic of a monolayer of graphene as well. The obtained results confirmed that the graphene sample was of high quality and defect-free. The spectrum in the range from 1050 cm −1 to 1950 cm −1 showed also features stemming from the SiC substrate. The optical microscope images of three of the fabricated devices are shown in Figure  3. Since graphene is barely seen in an optical microscope, its location is outlined with red dashed lines. The dark color corresponds to the annealed contact metallization, and the lighter color is the metallization deposited at the second step. The dimensions of the structures are summarized in Table 1.  The structures shown in Figure 3 represent the field effect transistors with the left and right sides of the central line acting as the source and drain. The graphene layer in the middle, which is located between the source and drain on the top of the AlGaN barrier layer, acts as a gate. It was shown previously [36] that graphene forms a high-quality Schottky barrier to AlGaN and graphene gate AlGaN/GaN transistors demonstrate very good characteristics. Figure 4 shows the transfer current-voltage characteristics of the transistor in the E4 structure. The transistors demonstrated around six orders of magnitude for the on-off ratio and the subthreshold slope n = 1.3-1.4. The subthreshold current, which is determined by the gate leakage current, was very small, even for the transistor with the highest gate area, as shown in Figure 4. The current-voltage characteristic shown in the linear scale in Figure 4 only slightly tends to saturate at high gate voltages, indicating the minor contribution of the contact resistance. The threshold voltage for the studied transistors determined from the linear extrapolation of the current-voltage characteristic at a small drain voltage was within the range of Vt = −3.0 to −3.2 V. This means that at Vg The optical microscope images of three of the fabricated devices are shown in Figure 3. Since graphene is barely seen in an optical microscope, its location is outlined with red dashed lines. The dark color corresponds to the annealed contact metallization, and the lighter color is the metallization deposited at the second step. The dimensions of the structures are summarized in Table 1. The optical microscope images of three of the fabricated devices are shown in Figure  3. Since graphene is barely seen in an optical microscope, its location is outlined with red dashed lines. The dark color corresponds to the annealed contact metallization, and the lighter color is the metallization deposited at the second step. The dimensions of the structures are summarized in Table 1.  The structures shown in Figure 3 represent the field effect transistors with the left and right sides of the central line acting as the source and drain. The graphene layer in the middle, which is located between the source and drain on the top of the AlGaN barrier layer, acts as a gate. It was shown previously [36] that graphene forms a high-quality Schottky barrier to AlGaN and graphene gate AlGaN/GaN transistors demonstrate very good characteristics. Figure 4 shows the transfer current-voltage characteristics of the transistor in the E4 structure. The transistors demonstrated around six orders of magnitude for the on-off ratio and the subthreshold slope n = 1.3-1.4. The subthreshold current, which is determined by the gate leakage current, was very small, even for the transistor with the highest gate area, as shown in Figure 4. The current-voltage characteristic shown in the linear scale in Figure 4 only slightly tends to saturate at high gate voltages, indicating the minor contribution of the contact resistance. The threshold voltage for the studied transistors determined from the linear extrapolation of the current-voltage characteristic at a small drain voltage was within the range of Vt = −3.0 to −3.2 V. This means that at Vg  The structures shown in Figure 3 represent the field effect transistors with the left and right sides of the central line acting as the source and drain. The graphene layer in the middle, which is located between the source and drain on the top of the AlGaN barrier layer, acts as a gate. It was shown previously [36] that graphene forms a high-quality Schottky barrier to AlGaN and graphene gate AlGaN/GaN transistors demonstrate very good characteristics. Figure 4 shows the transfer current-voltage characteristics of the transistor in the E4 structure. The transistors demonstrated around six orders of magnitude for the on-off ratio and the subthreshold slope n = 1.3-1.4. The subthreshold current, which is determined by the gate leakage current, was very small, even for the transistor with the highest gate area, as shown in Figure 4. The current-voltage characteristic shown in the linear scale in Figure 4 only slightly tends to saturate at high gate voltages, indicating the minor contribution of the contact resistance. The threshold voltage for the studied transistors determined from the linear extrapolation of the current-voltage characteristic at a small drain voltage was within the range of V t = −3.0 to −3.2 V. This means that at V g < −3.5 V, the channel is fully depleted and the transmission line is interrupted. It is connected only due to the highly resistive graphene layer and fringing capacitances. At zero gate voltage, the central line is connected by the highly conductive 2DEG. < −3.5 V, the channel is fully depleted and the transmission line is interrupted. It is connected only due to the highly resistive graphene layer and fringing capacitances. At zero gate voltage, the central line is connected by the highly conductive 2DEG.

RF Characteristics
On-chip S-parameter measurements of the graphene/AlGaN/GaN switches in the 70.5-114.5 GHz frequency range were carried out using a measurement setup configured as shown in Figure 5. The setup consisted of the Agilent N5245A PNA-X vector network analyzer (VNA) with VDI WR-10 waveguide frequency extenders and 100 µm pitch Cascade Microtech Infinity WR-10 waveguide GSG probes. The probes were positioned using the Cascade Microtech EPS200MMW probe station. For the measurements at frequencies below 50 GHz, the same setup was used but without frequency extenders and with coaxial GSG probes. Figure 6 shows an optical microscope image of the structure under testing with the probes attached.

RF Characteristics
On-chip S-parameter measurements of the graphene/AlGaN/GaN switches in the 70.5-114.5 GHz frequency range were carried out using a measurement setup configured as shown in Figure 5. The setup consisted of the Agilent N5245A PNA-X vector network analyzer (VNA) with VDI WR-10 waveguide frequency extenders and 100 µm pitch Cascade Microtech Infinity WR-10 waveguide GSG probes. The probes were positioned using the Cascade Microtech EPS200MMW probe station. For the measurements at frequencies below 50 GHz, the same setup was used but without frequency extenders and with coaxial GSG probes. Figure 6 shows an optical microscope image of the structure under testing with the probes attached. < −3.5 V, the channel is fully depleted and the transmission line is interrupted. It is connected only due to the highly resistive graphene layer and fringing capacitances. At zero gate voltage, the central line is connected by the highly conductive 2DEG.

RF Characteristics
On-chip S-parameter measurements of the graphene/AlGaN/GaN switches in the 70.5-114.5 GHz frequency range were carried out using a measurement setup configured as shown in Figure 5. The setup consisted of the Agilent N5245A PNA-X vector network analyzer (VNA) with VDI WR-10 waveguide frequency extenders and 100 µm pitch Cascade Microtech Infinity WR-10 waveguide GSG probes. The probes were positioned using the Cascade Microtech EPS200MMW probe station. For the measurements at frequencies below 50 GHz, the same setup was used but without frequency extenders and with coaxial GSG probes. Figure 6 shows an optical microscope image of the structure under testing with the probes attached.  For the calibration and setting of the measurement plane to the contact pads, the SOLR Cascade 101-190C and LRM Cascade 138-357 standard impedance substrates were used at low and high frequencies, respectively. The S-parameter measurements were performed with bias voltages applied between the central lines and ground. The DC connections were provided through the bias ports of the VNA and through the bias ports of the waveguide probes in the low-and high-frequency configurations, respectively. As the graphene layer was connected to the GND conductors (Figure 1a), a positive voltage applied to the signal pads corresponded to the negative bias of the graphene gate. Micromachines 2021, 12, x 6 of 12 For the calibration and setting of the measurement plane to the contact pads, the SOLR Cascade 101-190C and LRM Cascade 138-357 standard impedance substrates were used at low and high frequencies, respectively. The S-parameter measurements were performed with bias voltages applied between the central lines and ground. The DC connections were provided through the bias ports of the VNA and through the bias ports of the waveguide probes in the low-and high-frequency configurations, respectively. As the graphene layer was connected to the GND conductors (Figure 1a), a positive voltage applied to the signal pads corresponded to the negative bias of the graphene gate. Figure 7 shows the measured transmission coefficient S21 characteristics in the on and off state of three examined graphene/AlGaN/GaN switches (see Table 1 for details). In the studied configuration, the on state corresponds to zero gate voltage and the off state was studied at Vg = −5 V.   Figure 7 shows the measured transmission coefficient S21 characteristics in the on and off state of three examined graphene/AlGaN/GaN switches (see Table 1 for details). In the studied configuration, the on state corresponds to zero gate voltage and the off state was studied at V g = −5 V. For the calibration and setting of the measurement plane to the contact pads, the SOLR Cascade 101-190C and LRM Cascade 138-357 standard impedance substrates were used at low and high frequencies, respectively. The S-parameter measurements were performed with bias voltages applied between the central lines and ground. The DC connections were provided through the bias ports of the VNA and through the bias ports of the waveguide probes in the low-and high-frequency configurations, respectively. As the graphene layer was connected to the GND conductors (Figure 1a), a positive voltage applied to the signal pads corresponded to the negative bias of the graphene gate. Figure 7 shows the measured transmission coefficient S21 characteristics in the on and off state of three examined graphene/AlGaN/GaN switches (see Table 1 for details). In the studied configuration, the on state corresponds to zero gate voltage and the off state was studied at Vg = −5 V. The structure E4 with the large gap in the central line L = 2L gap + L g = 65 µm exhibits insertion loss between 12.5 and 14.9 dB in the on state across a very wide frequency range. On the other hand, the off-state characteristics show the strong frequency dependence of isolation, which is typical for parasitic capacitive coupling. Isolation values decrease from over 77 dB below 100 MHz to 17.5-19.4 dB in the 70.5-114.5 GHz frequency range. Despite poor isolation at frequencies above 70 GHz, the switching effect is still evident, with an on-off ratio of approximately 4 dB.
Narrowing the gap in the central line to L = 2L gap + L g = 20 µm in the E1 and G1 structures allowed us to achieve lower on-state insertion loss (7.4-14.1 dB) at frequencies above 200 MHz. A smaller gap between extended pads has an adverse impact on the isolation due to increased capacitive coupling between the input and output. As a result, at frequencies above~100 GHz, switching is not effective. E1 and G1 structures have different dimensions of extension pads, which causes their slightly different behavior at low frequencies.
In order to study the switching properties in the time domain, the DC power supplier was replaced with the function generator. Figure 8 shows the switching dynamics of a 1 GHz signal at the output. The waveforms presented in Figure 8 were obtained with 100 kHz, with a 5 V square wave signal applied to the structures.
Narrowing the gap in the central line to L = 2Lgap + Lg = 20 µm in the E1 and G1 structures allowed us to achieve lower on-state insertion loss (7.4-14.1 dB) at frequencies above 200 MHz. A smaller gap between extended pads has an adverse impact on the isolation due to increased capacitive coupling between the input and output. As a result, at frequencies above ~100 GHz, switching is not effective. E1 and G1 structures have different dimensions of extension pads, which causes their slightly different behavior at low frequencies.
In order to study the switching properties in the time domain, the DC power supplier was replaced with the function generator. Figure 8 shows the switching dynamics of a 1 GHz signal at the output. The waveforms presented in Figure 8 were obtained with 100 kHz, with a 5 V square wave signal applied to the structures. The measured switching times are listed in Table 2. The large differences in switching times between structures can be attributed to technological uncertainties, particularly to the graphene quality. The switch based on the G1 structure exhibits a very fast switching time. The rise and fall times are 25 ns and 17 ns, respectively. The use of such a switch can provide up to 20 MHz of bandwidth in time-modulated systems, which is an outstanding result for such systems [10,37]. The measured transmission characteristics shown in Figure 7 were compared with the characteristics of an equivalent small-signal circuit of the structures shown in Figure  9a. The circuit represents the physical design of the structure.
The graphene gate is represented by resistors R5-R4-R5 arranged according to the gate shape (Figure 9b). The graphene gate and 2DEG underneath constitute a capacitor with resistive plates, which is represented as an element R3-C3-R4. This element is actually a distributed capacitor and it is modeled as an infinite number of elementary stages dR3-dC3-dR4, connected as shown in Figure 9a. The values of R3-C3-R4 elements are the sums of their elemental counterparts. Two capacitors C2 represent the fringe capacitances of the graphene gate (mainly responsible for limited isolation in the off state). The measured switching times are listed in Table 2. The large differences in switching times between structures can be attributed to technological uncertainties, particularly to the graphene quality. The switch based on the G1 structure exhibits a very fast switching time. The rise and fall times are 25 ns and 17 ns, respectively. The use of such a switch can provide up to 20 MHz of bandwidth in time-modulated systems, which is an outstanding result for such systems [10,37]. The measured transmission characteristics shown in Figure 7 were compared with the characteristics of an equivalent small-signal circuit of the structures shown in Figure 9a. The circuit represents the physical design of the structure.
The graphene gate is represented by resistors R5-R4-R5 arranged according to the gate shape (Figure 9b). The graphene gate and 2DEG underneath constitute a capacitor with resistive plates, which is represented as an element R3-C3-R4. This element is actually a distributed capacitor and it is modeled as an infinite number of elementary stages dR3-dC3-dR4, connected as shown in Figure 9a. The values of R3-C3-R4 elements are the sums of their elemental counterparts. Two capacitors C2 represent the fringe capacitances of the graphene gate (mainly responsible for limited isolation in the off state).
Similarly, pad extension with 2DEG underneath constitutes a capacitor with a single resistive plate R1-C1. Values of circuit elements are given in Table 3. They were extracted from nominal structure parameters and DC current-voltage characteristics, except for C2, which was matched based on the measured data.
The comparison of measurements and simulations of the S11 and S21 parameters for the G1 structure is presented in Figure 10. For both reflectivity (S11) and transmission (S21), the simulation corresponds to the measurement in both operating states. Simulations of the E4 and E1 structures also showed good agreement with the measurements. Similarly, pad extension with 2DEG underneath constitutes a capacitor with a single resistive plate R1-C1. Values of circuit elements are given in Table 3. They were extracted from nominal structure parameters and DC current-voltage characteristics, except for C2, which was matched based on the measured data.
The comparison of measurements and simulations of the S11 and S21 parameters for the G1 structure is presented in Figure 10. For both reflectivity (S11) and transmission (S21), the simulation corresponds to the measurement in both operating states. Simulations of the E4 and E1 structures also showed good agreement with the measurements.    Similarly, pad extension with 2DEG underneath constitutes a capacitor with a single resistive plate R1-C1. Values of circuit elements are given in Table 3. They were extracted from nominal structure parameters and DC current-voltage characteristics, except for C2, which was matched based on the measured data.
The comparison of measurements and simulations of the S11 and S21 parameters for the G1 structure is presented in Figure 10. For both reflectivity (S11) and transmission (S21), the simulation corresponds to the measurement in both operating states. Simulations of the E4 and E1 structures also showed good agreement with the measurements.
(a) (b) Figure 10. S11 and S21 scattering parameters for the G1 structure calculated from equivalent small-signal circuit along with measurement results for (a) on and (b) off states. [dB] [dB] R5 R5 R4 Figure 10. S11 and S21 scattering parameters for the G1 structure calculated from equivalent small-signal circuit along with measurement results for (a) on and (b) off states.
The results of simulations shown in Figure 10 indicate that the proposed equivalent circuit represents well the behavior of the studied graphene switches and can be used for the designing of switches with required parameters. Table 4 compares the parameters of the studied switches with previously published results for graphene-based RF switches. As can be seen, the graphene/AlGaN/GaN switches demonstrate very good characteristics, in many ways better than other published simulated and experimental results. The proposed devices operate with a low driving voltage and very low power consumption, providing very fast switching. As seen in Table 4, this kind of graphene-based switch is one of the very few switches whose parameters have been studied experimentally. Although the switches in more conventional technologies demonstrate better highfrequency performance, this research is in the early stage and there is room for improvement. The advantage of the application of graphene as a gate in a millimeter-wave switch is that its parameters change with bias at the same time as the channel parameters. Therefore, in the off state, with positive bias on the 2DEG relative to the graphene gate, the conductive layers can be eliminated from the gap in the CPW, making the impact of parasitic connection through fringing capacitances and the gate less significant. The main directions of the future development of this type of structure are the optimization of graphene characteristics and switch geometry as well as the evaluation of shunt architecture to achieve lower insertion losses and a higher operating frequency band.

Conclusions
We have proposed and studied the design of a switch that uses a combination of graphene and two-dimensional high-density electron gas (2DEG) in an AlGaN/GaN system in order to provide effective switching at RF frequencies. The switch was integrated into the coplanar waveguide, which was chosen due to lower losses compared to the microstrip line, especially at high frequencies. The presented design is an on-chip solution fabricated in one technological process. The proposed equivalent circuit describes well the switching characteristics and can be used to design switches with required parameters. The switching times are sufficiently low to use this kind of switch in time-modulated systems.