1. Introduction
Frequency multiplication is a common operation performed by nonlinear electronic circuits for a wide plethora of applications. Shifting the frequency of a reference signal is generally easier than generating a signal at an arbitrary frequency, and allows relaxing the complexity of other parts of the electronics for the signal manipulation, allowing the design of more performing circuits under less demanding requirements. The frequency up conversion therefore enables the development of electronic circuits capable of enhancing either data transmission capacity and rate, or spatial resolution, miniaturization, power consumption, based on the considered application. In general, frequency multipliers are relevant for any improvement related to the increase of the system’s operating frequency. Hence, research in novel devices and architectures for frequency multiplication is powered by the interest in opening new possibilities for applications in the sub-THz and THz frequency range [
1].
The key application of electronic circuits performing frequency multiplication has historically been that of signal modulation and mixing for wireless communication systems. However, frequency multipliers are also a basic component of test and measurement equipment, such as spectrum analyzers, and medical imaging equipment, such as magnetic resonance and ultrasound devices.
Traditional Silicon-based frequency multiplier circuits include diode-based multipliers (varistor or varactor type), known for wideband and low-noise operation, and active transistor-based multipliers, with higher conversion gain but increased noise. In these two cases, the generation of higher harmonics is obtained either by clipping the output waveform or by exploiting the nonlinearity of the input-output relation. Concerning nonlinear analog circuits for high-frequency RF applications, graphene has shown great potential among the emerging technologies [
2]. The peculiar capability of tuning the polarity of charge carriers across the channel of a graphene field-effect transistor (GFET) by simply acting on the gate bias inspires the development of single-device ambipolar circuits that are usually made of several transistors, with the consequent area, mass, and power saving, which are crucial for several critical applications [
3]. Circuits that typically exploit the ambipolar conduction and the nonlinear relation between the drain voltage and the gate voltage of GFETs are mixers, phase shifters, frequency multipliers.
The first experimental demonstration of a GFET-based frequency doubler dates back to 2009 [
4], when a 10 kHz sine waveform was multiplied by a single back-gated GFET obtaining >90% RF power delivered to the second harmonic. Shortly after, an improved GFET with a top-gate design allowed better performance providing frequency multiplication from 100 kHz to 200 kHz [
5], and then from 700 MHz to 1.4 GHz, achieving a spectral purity >90% [
6] against the ~35% obtained from traditional unipolar field-effect devices. Following, the progress in the RF-GFET transistor fabrication allowed achieving the first above-cut-off 3-GHz frequency doubler [
7], and the first single-transistor frequency doubler operating in the Ku-band, up to 16 GHz [
8]. From that moment, inspired by the impressive results obtained for the frequency doublers, GFETs were proposed also for the design of frequency triplers [
9] and quadruplers [
10] by exploiting a W-shaped drain current-gate voltage transfer curve, achieved by using two GFETs in series and a double-gate GFET configuration, respectively. Overall, the GFET-based frequency multipliers have outperformed other transistor technologies [
11]. However, technological limitations are still hindering the widespread diffusion of graphene-based electronic devices, due to the issues related to graphene deposition and transfer process and non-negligible sensitivity to manufacturing-related uncertainties [
12]. These issues hold graphene-based components from achieving the high-frequency performance of state-of-the-art traditional silicon electronics and III-V compound semiconductor technology. Both technologies can deliver gain up to frequencies above 100 GHz, making them readily applicable to cutting-edge high-frequency RF wireless communications. Nevertheless, the interest towards graphene electronic devices lies in their versatility, allowing multiple applications to be integrated into a single component, rather than on achieving top-level performance.
Besides the interesting opportunities in RF electronics, GFETs have gained great attention mostly as high-sensitivity chemical sensors and biosensors [
13,
14,
15]. Thanks to the successful demonstration of single-molecule high-sensitivity sensing property, and being the simplest device capable of delivering high-purity frequency multiplication (especially at double-frequency harmonic component), the GFET was considered for the realization of nanomaterial-based harmonic transponders [
16,
17]. In this application, the GFET acts as a chemically-sensitive frequency multiplier: the chemical gating effect causes a shift of the transistor’s Dirac point, thus modulating the second-harmonic output signal and affecting the nonlinear conversion gain. Based on the fact that graphene can also be employed for energy harvesting applications and for tunable high-frequency antennas, it is possible to envisage all-graphene systems for combined sensing, modulation, and transmission, with the added value of being light-weight, small-footprint, low-power, in line with the ongoing Internet of Things and forthcoming Internet of Everything scenario. This envisaged application takes advantage of the unique characteristics of GFETs, offering the potential to integrate the functionality of sensing and modulation within a single component. Due to the frequency limitation and the fabrication-related performance impairment, at the present stage of technological implementations, graphene technology can be explored to develop integrated systems working below the sub-THz frequency range. The advantage related to the shift towards this novel technology remains in the development of lightweight, small, thermally-efficient, power-saving components. On the other hand, as nanotechnology is still in the first stages, the absence of established design rules poses a significant challenge for designers seeking to define optimal operating conditions and selection criteria during the system parameter setup. To address this challenge, this work proposes an original analysis of the performance of a GFET-based frequency doubler model, providing choice criteria for making informed decisions during the design of operating conditions of GFET-based signal modulators. Building on our previous activity [
3,
12,
18] and exploiting a GFET Verilog-A model of proven reliability, the proposed work explores the application of a recently presented high-performing GFET [
19] to the frequency doubling operation by carrying out circuit simulations. The analysis is approached by comprehensively computing relevant figures of merit under varying conditions, such as different bias points and operating frequencies. In this way, the proposed study targets the definition of criteria not yet available in the literature for the design of the intrinsic GFET operating conditions. Therefore, it supports the simulation-based design of novel components bypassing the need for expensive prototypes and time-consuming measurements.
In this paper, the operational principle of a single-device frequency doubler based on an intrinsic (i.e., without any external component) graphene field-effect transistor device is described in
Section 2, discussing the relevance of the bias point for the frequency multiplication. In
Section 3, the simulated GFET device is described, and the model validity is verified by comparing the simulation results with available measurement data. The frequency doubler’s performance in the frequency domain and its dependence on the input power level is explored in
Section 4, by evaluating relevant figures of merit such as the power conversion gain and the output power delivered at doubled frequency. The presented results are discussed in
Section 5, drawing some considerations on the application of GFETs in combined sensing and modulation applications. Finally, the conclusions are reported in
Section 6.
2. Operational Principle of a Single-Device GFET-Based Frequency Doubler
The ambipolar conduction of charge carriers across the graphene channel is a consequence of the V-shaped transfer curve of GFETs, which enables changing the polarity of the current, I
D, by varying the gate-source voltage, V
GS. The polarity of the drain-source current (I
DS) switches from p-type to n-type when the gate-source bias V
GS crosses the minimum conduction point (Dirac point), generally referred to as the Dirac voltage, V
dirac.
Figure 1a reports the I
DS-V
GS transfer curve of a typical GFET simulated by a compact model for DC operation. In
Figure 1a, the transfer curve is symmetrical because the model adopted for the circuit simulation assumes that, as in [
20], the mobility of both holes and electrons is the same, and the saturation velocity is independent on the carrier density. Under these assumptions, the value of V
dirac is roughly given by V
dirac = V
G0 + (V
D + V
S)/2 [
21], where V
G0 is the gate voltage offset due to charges trapped in the graphene channel, and V
D and V
S are the voltage values at the drain and source terminal, respectively.
The static gate and drain bias point (V
GS, V
DS) is fundamental for the correct operation of the frequency doubler: it is important that, at the chosen operating point, the small-signal variation of the drain current depends on the square of the gate voltage while being linearly dependent upon the drain voltage [
22]. When the transistor is biased at the Dirac point and a drain voltage is applied, both types of carriers are injected in the channel and recombine. This condition is represented in
Figure 1b by the region in proximity of V
DS = 0.5 V. For this reason, even by increasing the V
DS, the I
DS does not increase, and the current curve shows the typical inflection point related to the carrier polarity inversion [
23,
24]. If biased in this flat region, in which the odd symmetry property is lost, odd harmonics of the input signal can appear in the output waveform, impairing the output voltage spectral purity and reducing the power transfer to the harmonic of interest [
22], that, in the case of the frequency doubler, is the first even harmonic component.
The GFET frequency doubler circuit configuration is shown in
Figure 2a and the operational principle at the basis of the frequency doubling operation is explained in
Figure 2b. In this common-source configuration, the source voltage is V
S = 0, and therefore V
GS = V
G. When the gate is biased at the minimum conduction point V
G = V
G0 + (V
D + V
S)/2, and an alternating signal at frequency f
in, v
in = V
in sin(2πf
int), is applied to the gate so that v
G = V
dirac + v
in, the conduction is hole-based for one half of the period T = 1/f
in, and electron-based for the other half period. This concept is described in
Figure 2b, where the red curve representing the drain current I
D has a minimum in V
G − V
dirac = 0 and the left and right branches of I
D are related to holes (h+) and electron (e−) current, respectively. When the ac component of the gate voltage v
g oscillates between +V
in and −V
in, as indicated by the green dotted lines in
Figure 2b, the operating point shifts alternatively from the hole-based to the electron-based branch of I
D, and, due to the square-law dependence of the drain voltage on the gate voltage (see the blue curve in
Figure 2b), the signal v
D = V
D + v
d at the drain terminal oscillates at twice the input frequency, 2f
in (v
d ∝ sin
2 (2πf
0t)). The load is R
load = 50 Ω, i.e., no output impedance matching network is applied to maximize the power transfer, and therefore the circuit analysis of the schematic in
Figure 2a describes only the intrinsic device behaviour.
3. The Simulated GFET Device
The circuit simulation of the GFET-based frequency doubler was performed by using the RF design commercial software suite Advanced Design System−ADS
® (Keysight Technologies, Inc., Santa Rosa, CA, USA) and a GFET compact model [
25]. The adopted model was presented in [
26] and validated in [
27]. The model reproduced successfully the nonlinear behaviour of GFET-based ambipolar components for RF applications [
28,
29,
30], proving to be a valuable tool for the purpose of predicting the GFET frequency doubler’s performance. The device model considered in our simulations is the top-gated graphene FET described in [
19]. Employing an Al
2O
3 substrate (relative dielectric permittivity ε
ox = 7.5), with higher optical phonon energy compared to traditional SiO
2 substrates, the device exhibited enhanced high-frequency performance. This improvement was achieved thanks to the reduced impact on charge carrier saturation velocity of impurities injected into the graphene channel from the substrate. The geometrical parameters and the graphene electronic properties of the simulated top-gate GFET device were taken from [
19], which reports mobility up to μ = 2000 cm
2 V
−1 s
−1 and saturation velocity up to v
sat = 2 × 10
7 cm s
−1 for a transistor with channel length L = 0.5 μm and width W = 2 × 15 μm. The gate voltage offset V
G0 was determined from V
G0 = V
dirac − (V
D + V
S)/2 with V
dirac, V
D, V
S taken from the drain resistance curve shown in [
19]. The surface potential inhomogeneity Δ was fixed considering a fluctuation of ±50 mV [
31], corresponding to an energy variation of 0.1 eV. A sketch of the considered device is shown in
Figure 3. The complete set of the model input parameters used in the simulations and their description are reported in
Table 1.
Due to the lack of experimental data concerning the use of the considered GFET in frequency doubling operation, the model performance was compared to the available data reported in [
19]. The comparison between the simulated and measured short-circuit current gain, h
21, and its transition frequency, f
T, computed for increasing values of the drain bias, V
D, is shown in
Figure 4a and
Figure 4b, respectively. The simulations match the experiments quite well, however the model fails to catch the high-frequency behaviour in high-field condition, as can be noticed by the deviation of f
T from the measured values when V
D > 1 V. This issue is related to the emergence of phenomena linked to the high-field condition which is not accounted for by the GFET compact model. For instance, the effects related to carrier saturation velocity necessitate a more refined model. Moreover, a high drain field induces a non-negligible temperature increase, intensifying the scattering of charge carriers due to their increased interactions with lattice vibrations, impurities, and defects. Additionally, the increase in power consumption amplifies the self-heating of the device, further enhancing the temperature rise and its related effects. These circumstances ultimately lead to a significant impairment of channel mobility, with a consequent impact on the output current. The multifaceted nature of high field condition, therefore, demands more complex models to achieve accurate experimental fitting. To avoid this issue, the following simulations are performed at V
D < 0.5 V, in low-field condition.
It is worth noting that, since the considered device was never characterized as frequency multiplier, the proposed work relies on the consistency of the simulation framework. However, because the adopted model underwent validation for the specific device under consideration (see
Figure 4), and for the frequency doubling operation in the time domain [
26], we infer that the model can be assumed to describe accurately the combination of this device in frequency doubling operation.
Moreover, the input and output impedance matching networks are intentionally excluded in the performed analysis. By investigating the intrinsic GFET behaviour in frequency doubling operation, the conditions where the device performs more effectively without any other external component are identified. In this way, by exploiting the device functionality at its best, the complexity of matching networks that must be later designed can be reduced, and the design constraints relaxed. This approach allows preserving the simplicity of the circuit, aiming at minimizing the number of components to safeguard footprint, weight, energy consumption, and cost. Furthermore, optimizing the operating conditions to enhance performance of the intrinsic device mitigates the impact of unavoidable parasitic elements (e.g., pad capacitance, contact resistance, interconnect parasitics), which typically impair overall circuit efficiency.
5. Discussion
The proposed work supports the design of GFET-based frequency doublers for application to signal modulation systems in the MHz-range, such as harmonic transponders and integrated sensing and modulation systems. Based on the findings presented in
Section 4, concerning the performance of a recent GFET device in frequency doubling operation, informed criteria for the selection of the optimal operating conditions can be obtained. In particular, the results of this study restrict the range of frequencies for selecting the operating frequency of the system to those in which the device demonstrates higher efficiency, and provides information that can guide the shaping the input signal interrogating the sensor.
Indeed, the design of nonlinear circuits for frequency multiplication starts with the optimization of the bias levels. To operate as frequency multiplier, the GFET must be biased at the Dirac point, posing a first constraint on the gate bias. Following the analysis reported in
Figure 7b, it becomes evident that increasing the drain voltage is advantageous when the specific application demands a shift towards higher frequencies. This adjustment allows enhancing the rejection of the fundamental tone by design. Once the static operating condition is fixed, the input signal can be tailored.
Both the amplitude and frequency of the input signal can be set aiming at maximizing the device performance. To improve the power conversion efficiency of the intrinsic device, the most suitable operating frequency can be fixed based on the P
21 bandwidth (see
Figure 7b). It is important to bear in mind, during this phase, that increasing the frequency unavoidably worsens the power conversion gain, regardless of the input signal power level, as can be seen in
Figure 6. After fixing the signal frequency, appropriate measures for output impedance matching can be implemented, so that undesired harmonic components can be effectively driven through low-impedance paths (ideally, short-circuited) and the power transfer from the input to the double-frequency harmonic component can be enhanced. Concerning the input signal level,
Figure 7c,d show that higher amplitude of the input signal supports higher rejection of the fundamental tone, and improves the conversion gain through the greater input power, as proven by
Figure 8. However, it is worth noting that high amplitude of the input signal may lead to greater harmonic distortion due to the nonlinear behaviour in large-signal operation of the GFET.
Figure 8 shows that this may lead to saturation of the conversion gain. Therefore, the input signal amplitude should be set to achieve an input power level, based on the chosen operating frequency, that prevents saturation of the conversion gain.
Finally, the doubler should be cascaded with high-pass filters to remove the DC offset and enhance the rejection of the fundamental tone, and eventually with selective filters to relax the requirements on the impedance matching networks.
Although the operating conditions can be designed to maximize the efficiency of the GFET for its specific use as frequency doubler, it is worth discussing also the open issues related to the use of GFETs, and more broadly of graphene components, for the proposed application. Currently, the performance of GFETs in terms of amplification properties and maximum operating frequency still lags behind state-of-the-art components, mainly due to the poor conversion gain and impurity-related impairments. Moreover, shifting from traditional electronics to all-graphene electronics will require technological improvements in the manufacturing and transfer of graphene, to enhance the reliability and repeatability of graphene components. Nevertheless, simulation works suggest new possibilities and applications not affordable by traditional technology, inspiring the exploration of innovative solutions to overcome current technological limitations and drive graphene electronics towards its full potential.
6. Conclusions
This work investigated the performance of a single-device GFET-based frequency doubler by means of circuit simulations. The observed performance indicators include the ratio between the output power at double-frequency and that at fundamental frequency, and the power conversion gain at double frequency. The performed analysis offers guidelines for the design of operating conditions that maximize the device efficiency, thereby relaxing constraints on signal conditioning circuits and impedance matching networks. By providing the knowledge required to perform the tuning of the GFET frequency doubler design, the findings of this study can support the development of all-graphene sensing, modulation, and transmission systems. This study, combined with the envisioned future research directions, aims to contribute to the ongoing technological transition towards lightweight, small, thermally-efficient, power-saving carbon-based components for the pervasive Internet-of-Everything scenario. Future research will focus on the analysis of the frequency doubler’s robustness against fabrication-related uncertainties, addressing the concern about the process-induced non-uniformity of the device performance. Further, the development of criteria for designing input and output impedance matching network will be investigated, based on the latest advancements in graphene passive components. Finally, the GFET frequency doubler’s integration into highly-efficient and compact all-graphene systems will be explored.