Graphene Solution‐Gated Field‐Effect Transistor for Ultrasound‐Based Wireless and Battery‐Free Biosensing

The development of wireless and battery‐free sensors for biomedical applications is a fast growing research and industrial field. It promises to greatly improve the patient's comfort during the diagnosis phase, but also in the treatment of chronic diseases. While the standard technologies are based so far on electromagnetic waves, ultrasonic powering and communication is offering perspectives to further reduce the size of the sensor in order to develop minimally invasive electronic implants. Wireless and battery‐free ultrasound‐based devices for healthcare monitoring comprise a piezoelectric element for the powering and communication, and a variable shunt load that varies according to a physiological parameter of interest. The changes in the load modify the acoustic reflectivity of the piezoelectric element, which can be detected using a pulse‐echo protocol. In the present study, the use of graphene solution‐gated field‐effect transistor is introduced as a new type of shunt load. Using an mm‐sized device, it is shown that the amplitude of an ultrasonic wave that reflects on the piezoelectric component is modulated by the voltage applied on the gate of the transistor both in physiological medium and biological tissue. This study sets the basis toward a new type of ultrasound‐based wireless and battery‐free biosensors.


Introduction
The past decade has seen the rising of wireless and wearable devices for healthcare monitoring that greatly improved the patient comfort during diagnosis and rehabilitation phases. [1,2] One of the striking examples that illustrate this major contribution is the multisensing experiments required to monitor physiological parameters during sleep. [3] Indeed, the wired-based solutions are great burden because they reduce the patient mobility and strongly alter their sleep state while wireless solutions can be worn without discomfort. More recently, the emerging concepts of fully wireless and battery-free technology represent a new step toward the improvement of the patient comfort especially in the case of implantable medical devices since they facilitate the insertion surgery and avoid the cumbersome use of wires and batteries. [4] This technology directly uses the energy delivered by an external probe to power and interrogate a sensor. So that the implant is only active when interrogated by the external probe as in the case of Radio Frequency Identification technologies (RFID). So far, most of the standard wireless and wearable technologies have been based on the electromagnetic (EM) communication that are very efficient in the air. However, they face very strong limitations when dealing with implants that aim at collecting biosignals from the inner organs. Indeed, electromagnetic waves are attenuated by the tissue and thus require high power to be transmitted, [5] leading to detrimental heating effects on tissues in the surrounding of the implant. [6,7] Besides, the dimension of an EM antenna is rather difficult to shrink down to the ideal implant size of hundred micrometers because the attenuation of EM scales with the frequency. [5] To circumvent such drawbacks, some EM-based implantable strategies consist in deporting the antenna and the digitization unit under the skin [8] or at the skin surface using wires. [9] More recently Nurmiko et al. reported a remarkable network of wireless and battery-free epicortical implants. [10] They managed to record and stimulate from several implants on the rat cortex using an RF powered integrated circuit. Nevertheless, this technology might remain only usable for epicortical recordings due to overheating issues when attempting to power intracortical implants. Table 1. Comparison between the state or the art Ultrasound based wireless and Battery-Free implantable sensors. It can be noticed that all of them are using a rigid integrated circuit. Most of them are encompassing a power block to supply DC voltage to the sensor.
Ref. [14] Ref. [13] Ref. [15] Ref. [12] Implant size [mm 3 ] 0 . 3  Ultrasounds offer a very interesting alternative to communicate with implants. The attenuation of the ultrasonic waves in the tissue is rather weak compared to that of EM waves and the size of the piezoelectric component for communication has been evaluated to be able to reach dimensions as low as 10 μm when using ultrasound frequencies ≈10 MHz. [5] The underlying physical mechanism that allows the ultrasonic-based wireless and battery-free sensing consists mainly in modulating the acoustic reflective coefficient of a piezoelectric component using a variable shunt load [11] connected at its two terminals. The implant can thus be interrogated wirelessly and does not contain any battery. The external ultrasound source is used to power the sensor that modulate the echo according to a physiological parameter that changes the equivalent impedance of the sensor. Using this concept several implantable millimeter and submillimeter-scale devices have already been reported for temperature monitoring, oxygen monitoring and also to elicit and measure neural activity [3,[12][13][14][15] ( Table 1). The implant for temperature monitoring consists in a bulk lead zirconate titanate (PZT) ceramic connected to a complementary metal oxide semiconductor (CMOS) integrated circuit (IC) composed of power management block to supply a direct current (DC) voltage to a temperature sensor. [14] The temperature sensor modulates the input impedance of the PZT to reflect an echo containing the temperature information. The implant for oxygen monitoring is based on a similar architecture with an oxygen sensor used as a modulator. [13] In addition, the IC developed in this study is comprising a module that conditions the echo to contain a binary information of the oxygen concentration. This binary coding of the information of interest in the echo signal is also used in a stimulation implant, [15] where the echo contains the information of the stimulation parameters used by the IC. Finally, the sensor developed by Carmena et al. [12] simply uses a CMOS transistor where the gate is connected to an electrode in contact with the biological tissue to measure the sciatic nerve activity. In this case no power block is used thus lowering the complexity of the system. The transistor impedance modulations induced by the electrical neural activity directly modifies the acoustic reflectivity of the piezo, eliminating the use of an integrated circuit for power management.
Consequently, while the size of the implant, as well as the mechanical flexibility are key parameters to provide all the expectations from this technology, it can be noticed that all the ultrasonic implants that have been developed so far have been constituted of a bulk piezoelectric ceramic coupled to a rigid CMOS-based integrated circuit used as a shunt load that limits their flexibility and their easy fabrication. In particular, most of them must encompass a power block in the IC to DC power their sensor while a sensor working in alternating current (AC) configuration could be sufficient. So new strategies remained to be found to fully exploit the potential of this technology.
Solution-gated Field-Effect transistors (SGFET) that have been developed for biosensing, are very interesting electronics component that can certainly meet this demand. Actually the channel of a SGFET is directly in contact with the medium and do not require the use of an additional electrode to be connected to the gate terminal of the transistor as in the case of CMOS-based technology. It thus reduces the fabrication and integration complexity inherent to the use of a CMOS circuit. Many inorganic, organic and carbon-based SGFET have been reported [16][17][18][19] to be very efficient biosensors. Among them H-terminated diamond, poly(3,4-ethylenedioxythiophene) doped with polystyrene sulfonate (PEDOT-PSS), and more recently graphene, were proved to be biocompatible and very well adapted for biosensing. [20,21] While diamond can be integrated in the flexible substrate, graphene and PEDOT-PSS SGFETs inherently present high mechanical flexibility that makes them ideal for implantable technologies. [22,23] Furthermore, graphene offers the ultimate atomically flat dimension, ideal for low invasive sensor. Thanks to its high mobility, graphene has also unique properties to be used in high frequency regimes and thus in the MHz range, which is the main target for ultrasound-based implantable technologies. Additionally, graphene SGFET (gSGFET) have been successfully used for a wide range of healthcare applications, such as epi [18] an intracortical [24] neural recording, as well as a very large number of biosensors. [25] In particular, a previous study reported a graphene sensor for in situ wireless and battery-free monitoring of dental bacteria development using a radio-frequency antenna. [26] In the present study, we propose gSGFET as an alternative to the CMOS technology to further reduce the size of ultrasoundbased implants and to eventually improve their mechanical flexibility. We combine a gSGFET to a bulk piezoelectric PZT ceramic to show that it can be used as a modulator of the acoustic reflective coefficient of a piezoelectric component. The standard pulse-echo technique is used to power and interrogate the state a millimeter-sized gSGFET without the use of a CMOS IC, in order to established the transfer curve of the sensor that correlates the echo amplitude to the gate to source voltage (V gs ) applied on the transistor (Echo vs. V gs ), as well as the noise of the echo measurement. We thus show that the ultrasonic echo can be used as a mean to interrogate the state of the transistor which load. c) Transfer curve that links the amplitude of the Echo signal to the value of the resistive load. d) Schematic of a gSGFET. e) Typical transfer curve that links the resistance of the graphene channel (R ds ) to the gate to source voltage (V gs ). The transfer curve has a lambda shape and the position V gs that corresponds to the maximum value of resistivity, also called the charge neutrality point (CNP), depends on the charge dopant at the surface of the channel. f) Principle of the biodetection based on the CNP shift of a gSGFET illustrated with the example of the antigen recognition. When an antigen binds to the antibody the electric charge change at the surface of the channel thus inducing a shift of the CNP. g) Principle of potential monitoring in the case of an electrogenic cell. h) Schematic of piezoelectric shunted by a gSGFET. i) Transfer curve that links the Echo amplitude to V gs . This curve is the combination of the Echo versus R transfer curve and the R ds versus V gs transfer curve. j) Principle of the wireless and battery-free measurement of the CNP shift of a gSGFET for biosensing. k) Principle of the wireless and battery-free potential monitoring.
constitute the first step toward real-time monitoring of physiological parameters.

Concept of the Ultrasound-Based Interrogation of the gSGFET
The wireless ultrasound-based interrogation consists in probing the state of a variable shunt load connected to the two terminals of a piezoelectric component by measuring the amplitude of the ultrasonic wave that reflected by its surface (the echo wave). Indeed, the acoustic reflectivity of the piezoelectric component depends on the resistance of the load connected between its two terminals. [11] Consequently, the amplitude of the echo wave also depends on the resistivity of the shunt load (Figure 1b). Hence, it is possible to establish a transfer curve that links the amplitude of the echo signal to the resistance (R) of the shunt load ( Figure 1c).
In the case of gSGFET, the modulation of the resistivity is directly linked to the variations of the electrical environment of the graphene channel. This change can be induced by pH variation, neural activity or any other cellular or molecular binding on the graphene channel. Measuring the transfer curve that links the graphene channel resistance (R ds ) to the gate to source voltage is a way to access these modulations. The typical R ds versus V gs transfer curve of a gSGFET has a lambda shape (Figure 1e) that directly reflects the band energy diagram of the graphene. [27] The maximum value of the resistivity corresponds to the region with the less available energy states accessible to the charge carriers and is called the charge neutrality point (CNP). The transition from the transmission to the reception is ensured by a RF switch. b) Picture shows the US transducer above the gSGFET sensor in a PBS tank during a pulse-echo experiment. c) Protocol to extract V ds versus V gs and Echo versus V gs . V gs is swept within the potential window of graphene in PBS. For each value of V gs , the amplitude of the Echo and V ds are measured. The maximum peak to peak amplitude of the Echo and V ds are extracted for each V gs to reconstruct the transfer curves. d) Protocol to evaluate the noise of the echo measurement. The pulse-echo is repeated at a fixed time interval. The maximum of the peak to peak amplitude of the Echo is extracted for each pulse-echo and then plotted versus time.
From this transfer curve, two main working principles of gS-GFET for biosensing applications can be distinguished. One, relies on the shift of the CNP position induced by the charge doping at the surface of the graphene layer. [28,29] An example of this working principle is the antigenic recognition using a gSGFET where the CNP is shifted when the antibodies grafted at the surface recognize their associated antigen [30] (Figure 1f). The second working principle relies on the modulation of V gs with an additional voltage induced by electrical activity in the vicinity of the gSGFET as in the case of electrogenic cells recording [18] (Figure 1g).
Consequently, by combining a gSGFET and a piezoelectric component, it is possible to establish a novel transfer curve that links the amplitude of the echo signal to V gs (Figure 1i). This curve is the combination of the Echo versus R transfer curve and the R ds versus V gs transfer curve. As for gSGFET, this transfer curve can either be used to extract the position of the CNP for wireless and battery-free biosensing (Figure 1j) or as a calibration curve for wireless and battery-free bioactivity recording (Figure 1k).

Measurement Protocols of the Ultrasound-Based Interrogation of the gSGFET
A pulse-echo home-made setup was developed to measure the echo signal reflected by a piezoelectric ceramic disc shunted either by a resistive load or a gSGFET (Figure 2a,b). It is composed of a focused immersion ultrasonic transducer that operates at 2.22 MHz and can be connected to either a high-speed acquisition system or to a function generator with a power amplifier to generate ultrasonic wave thanks to a RF switch. The acquisition card, the RF switch and the waveform generator are connected www.advancedsciencenews.com www.advmattechnol.de to a personal computer (PC) and are interfaced using a custom python code. The setup allows the collection of the voltage across the source and drain terminals (V ds ) of the sensor. The device to be interrogated is placed in a transparent tank coated with PolyDimethylSiloxane(PDMS) to minimize the acoustic reflection from the walls and filled with a Phosphate Buffered Saline solution (PBS). An Ag/AgCl electrode is used as a gate electrode and the V gs is controlled by the python code.
The pulse-echo protocol is composed of a transmission and a reception phase whose duration can be tuned by the user using the python code. During the transmission phase (pulse), the ultrasonic transducer is connected to the waveform generator coupled to the power amplifier. During the reception phase (echo) the ultrasonic transducer is connected to the acquisition system to record the echo and V ds signals. The transition from the transmission to the reception phase is ensured by the RF switch.
In order to obtain the Echo versus V gs transfer curve, the gate to source voltage (V gs ) is swept within the potential window of the graphene electrode in PBS. For each V gs , a pulse-echo protocol is performed to obtain the associated Echo and V ds . The maximum of the peak to peak amplitude of both signals are then collected and plotted versus V gs (Figure 2c).
For a time course representation, V gs is fixed and the pulseecho protocol is repeated at fixed time intervals until the buffer of the acquisition system is filled. The maximum of the peak to peak amplitude of each Echo is extracted. The output values are then used to reconstruct a temporal signal (Figure 2d).

Determination of the Echo versus R Transfer Curve
Prior to the fabrication of the sensor that combines the piezoelectric ceramic disc to the gSGFET, the modulation of the acoustic reflectivity of this piezoelectric disc is assessed using various resistive loads. The goal of this step is to establish the Echo versus R transfer curve. To that aim, the piezoelectric disc is placed on a PolyEthylene Terephthalate (PET) substrate and its two terminals are connected to external wires using a silver paste. The device is manually encapsulated with a thin epoxy layer to avoid any contact with the liquid. Loads presenting values from 50 Ω to 10 kΩ are plugged to the connector that is placed outside of the PBS bath. For each R value, the pulse-echo protocol is applied using a 30 Vpp, 2.22 MHz sinusoidal signal. The transmission phase was set to 4 μs. The acquisition phase starts 4 μs later and last 90 μs. The peak to peak maximum of Echo is extracted and then plotted versus R. On Figure 3 we can observe that the echo amplitude is modulated over a range of 1.2 mV for load values between 50 Ω and 5 kΩ. Most of the modulation can be observed in the 50 Ω -2 kΩ. Ideally, the transistor should thus exhibit R values modulated between 0 to 2 kΩ with a maximum variation in the 0-750 Ω region.

Sensor Design
In order to obtain a rough estimation of the gSGFET geometry that match the resistance range imposed by the piezoelectric disc, we simulated with Python the transfer curve R ds versus V gs using a simplified compact model [31] defined by the following equation (see Supporting Information for the derivation of this equation).
Where 'μ' is the charge carrier mobility, 'e' is the electronic charge, 'L' and 'W' are the length and the width of the graphene channel in contact with the liquid, 'n 0 ' is the impurity charge density of the channel, C dl is the normalized double layer capacitance, and R c is the contact resistance that also includes the access resistance. For the simulations, we used a mobility of 2000 cm 2 ⋅(Vs) −1 , which is standard for gSGFET. The double layer capacitance was fixed to 2 μF⋅cm −2 , which is a value commonly reported for graphene/PBS interface. [32] R c was fixed to 400 Ω. This large value is inherent to a connection using conductive ink and to the manual insulation of the channel using epoxy (see Sensors Fabrication Section). Finally, n 0 was fixed to 10 12 cm −2 . Considering these parameters, W/L values between 5 and 7 appear to be a good compromise to reach both a low channel resistance and a high modulation (Figure 4). For these aspect ratios, the minimum resistance is ≈450 Ω and varies between 450 and 1000 Ω that would correspond to a variation of the echo signal ≈300 μV.

Sensor Fabrication
In order to validate the concept of the ultrasound-based wireless interrogation of gSGFETs, we used cheap and up-scalable fabrication techniques. Graphene on parylene developed by Grapheal has been chosen because it simultaneously offers high crystalline quality ( Figures S1 and S2, Supporting Information), easy handling and high flexibility. Indeed, this type of graphene has already proved very good efficiency when used for highly demanding fundamental physics [33] and presents good material properties. Mm-sized pieces of graphene on parylene can be easily cut with a sharp blade and directly placed on any substrates. Sheets of flexible PET are used as substrate and insulator. The fabrication of the sensor is made as depicted in Figure 5a. First, the window for the exposition of the graphene to the liquid are opened in a polyethylene terephthalate (PET) sheet by laser etching (Figure 5a.2). Then the electric contacts are patterned on the same PET sheet using a conductive ink printer (Figure 5a.3). A graphene sheet deposited on parylene is manually cut using a sharp blade and manually placed in contact with the silver ink (Figure 5a.4). A second PET sheet is glued to the first sheet to encapsulate the metallic contacts. The pads are then opened on the first PET sheet using the laser etching machine (Figure 5a.5). A commercial piezoelectric PZT ceramic disc (2 mm diameter and 1 mm thick) and the wires to measure the voltage at the two terminals of the sensor are soldered manually using a silver paste (Figure 5a.6,a.7). The device is finally sealed and encapsulated in an epoxy layer (Figure 5a.8). The exploded view and the picture of the final device are presented in Figure 5b,c, respectively.

Electrochemical and DC Characterization of the Sensors
After the fabrication, the device is characterized by cyclic voltammetry (CV) in PBS. This approach allows to determine the V gs range where the interface between the graphene sheet and the PBS is only capacitive. This defines the potential range where the transistor can be operated properly without current leakage from the graphene channel to the PBS. Additionally, the CV is performed to verify that there is no leakage in the encapsulation layer. A leakage would impair the sensitivity of the system since it would alter the effective resistance channel. For this characterization, the two terminals of the sensor are shortcut to form a graphene electrode. Being very far from its resonant frequency, the piezoelectric component can be considered as an open circuit and does not interfere with the measurement. An example of the cyclic voltammetry curves is presented in Figure 6a. The curve shows the CV of a graphene electrode has a typical potential window of ≈0.9 V. No current leakage from the contact can be observed, showing that they are properly encapsulated by the epoxy.
The sensors are then characterized in DC configuration to obtain the standard transistor transfer curve (R ds vs. V gs ) (see Experimental Section). This step gives the load modulation at the two terminals of the piezoelectric component within the limit of the potential window. Figure 6b shows the typical R ds versus V gs transfer curve of a gSGFET. The CNP is found ≈−0.15 V indicating a slight in doping of the graphene. The resistance of the channel varies between 700 and 1600 Ω. According to the Echo versus R transfer curve the modulation in the echo response should be ≈180 μV.

Ultrasound-Based Interrogation of the gSGFET State in Phosphate Buffered Saline
In order to obtain the Echo versus V gs transfer curve, the gate to source voltage (V gs ) is swept within the potential window of the graphene electrode in PBS between −0.5 and 0.3 V with a constant step of 25 mV. For each V gs value, the ultrasonic transducer is excited by a 2.22 MHz, 30 Vpp sinusoidal signal during 4μs. Another 4 μs after the end of the pulse, the ultrasonic transducer is connected to the acquisition system for 90 μs. Figure 7a shows the V ds time response for various V gs values. It can be observed that the voltage generation starts ≈8 μs which corresponds to the distance of ≈1.5 cm between the transducer and the surface of the sensor. The modulations of the V ds amplitude with V gs can clearly be seen. Figure 7b presents the Echo response for different V gs . One can notice that the echo signal starts at ≈16 μs, twice the time at which the generation of V ds is observed. This signals thus corresponds to the ultrasonic wave reflected from the piezoelectric disc.
When plotting V ds versus V gs (Figure 7c blue curve), the standard modulation of the graphene transistor can be observed proving that the sensor is changing its resistivity with V gs as it is expected for a gSGFET. A variation of 30 mV (from 105 to 75 mV) is observed on the 500 mV V gs range. The CNP is found ≈−0.16 V which is similar to the value obtained with the DC characterization.
Similarly, when plotting Echo versus V gs (Figure 7c green), a modulation of the echo clearly appears. As expected, the shape of the modulation of the Echo is reversed with respect to that of V ds and the position of the minimum of the Echo amplitude is very close to that of the CNP obtained in the V ds versus V gs transfer curve. This proves that the modulations of the echo amplitude are indeed induced by the change of the graphene channel resistivity. Additionally, a variation of 200 μV is observed in the echo signal across the whole V gs range. This value is close to the 180 μV that was expected from the Echo versus R curve.
In order to extract the sensitivity defined as the variation in the echo amplitude as a function of the variation of the gate voltage, the Echo versus V gs was fitted by a polynomial curve (Figure 7d). The maximum of the slope was found to be ≈840 μV V −1 .
We further assessed the repeatability of the measurements. Twelve voltage sweeps were performed and compared. We observed that the total amplitude of the signal is slightly shifted (few tens of microvolts) from one measurement to another ( Figure S3, Supporting Information). This shift can be attributed to a change www.advancedsciencenews.com www.advmattechnol.de of the position of the sensor which is not firmly fixed to the PDMS backing in the tank. By subtracting the minimum value of each single sweep, we observe that the signal is actually repeatable (Figure 7e) since the amplitude of the modulation is the same from one sweep to another. Moreover, the shape and the position of the CNP are the same for all the recordings. Interestingly, the position of the CNP of the average curve perfectly matches that measured with the V ds versus V gs transfer curve. The maximum slope is ≈820 μV V −1 (Figure 7f) which is close to the best single curve obtained for a single sweep. This indicates that repeatable signals can be measured properly.
Finally, in order to evaluate the accuracy of the measurement, a time reconstruction of the Echo amplitude is performed. V gs is fixed to 0 V and the pulse-echo patterns are repeated every 100 μs (thus mimicking a 10 kHz sampling frequency), >10 ms. Figure 5g (black curve) shows the evolution of the echo amplitude with time. This reconstructed recording exhibits a noise ≈80 μVpp. Considering the maximum of the sensitivity is ≈820 μV V −1 , the minimum gate variation that can currently be measured is ≈50 mV. This recording has been repeated 11 times. When averaging 11 recordings, the noise decreased down to 25 μVpp (Figure 5g, red curve).

Ultrasound-Based Interrogation of the gSGFET State in Chicken Muscle
In order to validate the concept with biological tissue, the sensor has been placed <2 cm of fresh chicken breast. Several pieces were superimposed in order to achieve a 2 cm thick tissue. Similar features compared to the experiment in PBS can be observed (Figure 8). V ds amplitude is modulated by V gs and the maximum voltage generated by the sensor is ≈90 mV (Figure 8a blue curve) thus reduced compared to that measured in PBS. This is consistent with the fact that the attenuation coefficient that is higher in the chicken tissue than in PBS. [34][35][36] Considering the echo signal (Figure 8a green curve), the amplitude varies by 80 μV over the 500 mV range applied between the gate and the source. The maximum sensitivity is ≈350 μV V −1 which is lower than in PBS (Figure 8b). By averaging over six sweeps, the position of the CNP can be determined with a minimal accuracy of 25 mV (Figure 8c). The maximum sensitivity of the averaged transfer curve over six sweeps is ≈350 μV V −1 (Figure 8d). This experiment thus shows the possibility to sense change of the gSGFET state at 2 cm deep inside a body tissue.

Discussion
We demonstrated for the first time an ultrasonically-powered sensor that coupled a gSGFET and a piezoelectric component without any embedded batteries nor capacitor to store electric energy. The experiment demonstrates the possibility to wirelessly interrogate the state of a gSGFET using an ultrasonic wave. In particular, by measuring the transfer curve Echo versus V gs , we showed that the resistivity modulations induced by the change in V gs , directly modulate the amplitude of the echo signal that is  reflected by the piezoelectric component. The variations of V gs induced by local changes of the electronic environment of graphene are the core of the biosensing using gSGFET. Consequently, this transfer curve sets the first crucial step for a wide range of wireless ultrasound-based gSGFET biosensors.
The stability of the CNP is a major issue for biosensors. [37][38][39] In PBS and in the chicken tissue, the position of the CNP point was found to be stable over the measurements of the Echo versus V gs transfer curve. Since this parameter is directly linked to the electric environment of the graphene channel, it is already possible to consider this type of device for applications, such as wireless pH sensing, ionic force sensing or any biosensing where the position of the CNP is modified.
By repeating the pulse-echo pattern every 100 μs, we also derived a time recording of the echo amplitude sampled at 10 kHz for 10 ms. Provided further software developments, the acquisition time can be extended to several hours to perform real-time potential monitoring. The time recording presented in Figure 7g showed that the noise of the echo measurement is ≈80 μVpp. Considering that the maximum sensitivity is ≈820 μV V −1 , 1 kHz signals with an amplitude of 50 mV can already be recorded thanks to this sensor. Moreover, by averaging the signal  Figure 8. a) Best echo and V ds versus V gs curves obtained for a single sweep (Echo is in green and V ds is in blue). b) Derivative of the polynomial fit of the echo versus V gs . c) Six sweeps with DC subtraction and then averaged (red curve). d) Derivative of echo amplitude with respect to V gs . This curve gives the sensitivity of the echo signal with respect to V gs . e) Picture of the ultrasound interrogation when the sensor is placed <2 cm of chicken tissue. multiple times it would be possible to further decrease the noise. For instance, by averaging 100 identical signals, the noise level can be decreased down to 8 μVpp and thus offering the possibility to measure 5 mV signals. In the same way, by sampling at 100 Hz with each sample being the mean of 100 pulse-echo, it would be possible to measure in real-time a 50 Hz signal with an amplitude of at least 5 mV. Hence, wireless electromyography can already be considered using our sensors.
The possibility to measure the Echo versus V gs transfer curve of the sensor when it is placed inside 2 cm of chicken tissue indicates that ultrasound-based wireless and battery-free biosensing using subcutaneous implants can already be envisioned. Additionally, devices placed on the surface of the skin can also already be considered. Actually, the lower acoustic frequency can be used to cross the air thus presenting an ideal strategy for wearable and battery-free devices.
The sensitivity of the present device is still lower than the tethered sensors [40] or even to the competitive ultrasound-based recording. [12] The main reason is that the full systems have not been optimized. Nevertheless, there are many ways to improve it. Working with piezoelectric devices that can further modulate their acoustic impedance according to the load value at their two terminal would be key to further improve the sensitivity of the devices and measure action potential of ≈1 mV and below as reported by Seo et al. [12] Additionally, the gSGFET sensors used here can be further improved to exactly match the region where the modulation of the acoustic reflection of the piezoelectric is maximum. Actually, graphene SGFETs are quite easy to tune and can be adapted to operate in the region where the maximum variation of the echo versus the load is expected. By controlling the graphene crystalline quality, the number of layers and using clean room microfabrication technics, it is possible to tune the contact resistance, as well as the transfer curve. The miniaturization and flexibility of the implant can also be greatly improved by using piezoelectric micromachined ultrasonic transducers and organic polymer as piezoelectric components.

Conclusion
In this study, we demonstrated a new type of ultrasound-based wireless and battery-free devices for biosensing. The cm-sized sensors made with graphene on parylene-C connected to a bulk piezoelectric were fabricated and fully characterized in PBS prior to the determination of the transfer curve of the system. We proved that it is possible to obtain the Echo-V gs transfer curve which is the first step toward biosensing and real-time monitoring. Beyond this first proof a concept, this study is showing that SGFETs are the interesting component to be used as a shunt load for ultrasound-based implantable technologies. Given, the large number of studies and opportunities reported on SGFETs for biosensing, we expect that many implantable biosensors can be designed and optimized following the same concept.

Experimental Section
Device fabrication: Laser etching was performed using a CO 2 Laser (Laserbox, Makeblock). The etching parameters were set to 45% of the power and a displacement speed of 15 mm⋅s −1 The printing of the metallic contacts was performed using a printing circuit board printer (Voltera V-one). The silver conductive ink was Conductor 2 from Voltera. Graphene on parylene was produced by Grapheal using a process reported previously. [33] The piezoelectric PZT ceramics (2 mm diameter and 1 mm thickness) were purchased from PI-Ceramic (PIC 181). The silver paste used for the soldering and the epoxy used for the encapsulation were purchased from RS (ref 186-3600 and 850-956, respectively).
Pulse-Echo Set Up: The system was composed of a manual homemade 3-axes translational stage offering a 10 μm precision to hold the ultrasonic transducer. An additional rotation stage was fixed on the z-axis to allow the rotation of the ultrasonic transducer. A customized transparent tank (dimension 80×35×50 mm) was placed on a 2-axes platform. The transducer was a 6 mm diameter immersion focused ultrasonic transducer (Olympus, USA Model: v323) with nominal frequency at 2.25MHz and a focal between 11 and 15 mm.
In order to perform both transmission and reception with the same transducer, a RF single pole double throw switch (Mini circuit, Model: ZX80-DR230-5+) was used to connect the ultrasonic transducer either to a function generator or to the acquisition system. In transmission mode, the ultrasonic transducer was powered by a function generator (Rigol Model: DG1062) followed by a power amplifier (Mini circuit, ZFL 22+). In reception mode, the transducer was connected to a high speed (1GS/s) data acquisition card (Teledyne SP devices, Linköping, Sweden, Model: ADQ14DC-2A). A buffer amplifier was placed at the input of the acquisition card to impose a high input impedance. A PCIe multichannel voltage source (NI 6738) was used to control V gs and the switch terminals. V ds signal was acquired by the same data acquisition card with a buffer amplifier at the input.
Cyclic Voltammetry: The cyclic voltammetry was performed in the three electrodes configuration using an Autolab PGSTAT 128N. The reference electrode was an Ag/AgCl electrode and the counter electrode was a curled platinum wire. Cyclic voltammetry was performed at 0.2 V⋅s −1 .
Measurement of the R versus V gs Transfer Curve: The solution-gated graphene field effect transistors were characterized with a custom made software to acquire the current across the two terminals of the gSGFET. V ds and V gs were applied using a NI PCIe 6738. A current to voltage amplifier (Femto, Model: DHPCA-100) was used to measure the current and the signal were acquired by the ADQ14 card. The measurement were performed at a fixed V ds of 100 mV.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.