Multichannel, Portable and Low-Cost Source and Measure Unit Platform for the Continuous Characterization of Organic Electrochemical Transistor Biosensors

1. Introduction

In the field of biosensing technologies, a novel category of biosensors is represented by Organic Electrochemical Transistors (OECTs). These devices are based on gate, drain and source electrodes and an ion-permeable organic semiconducting (OSC) channel [1]. OECTs operate in contact to an electrolyte, and gate-source voltage variations (input) modulate the drain-source current (output) through ions injection into the channel [2]. To date, OECTs have been exploited for detecting a wide range of analytes, including enzymes, metabolites, antibodies and nucleic acids [3]. Thanks to OSC properties, they have also emerged for biological models monitoring. Indeed, the most common channel material, namely poly3,4-ethylenedioxythiophene-polystyrene-sulfonate (PEDOT:PSS), exhibits electrochemical stability, biocompatibility and enables low voltages operation (< 1 V), endowing OECTs with a relatively high signal amplification [1,2,4].

OECTs are typically characterized by commercial Source and Measure Units (SMUs), instruments which simultaneously provide and measure voltage and current on the same channel [5]. In OECT biosensors characterization, generally the drain-source current (I_ds) is measured while supplying a variable gate-source voltage (V_gs) and a constant drain-source voltage (V_ds). Through this approach, biochemical processes are examined by the obtained transfer characteristic and transient response [6]. Common SMUs used in research are the B2900 Series by Keysight Technologies and the 2400 Standard Series by Keithley Instruments [7,8,9]. As representative model, the Keysight B2912A features full 4-quadrant operation, a maximum range of ± 210 V / ± 3 A, a maximum resolution of 100 nV / 10 fA, a noise below 100 nV / 10 fA, a maximum sampling rate of 10 μs and a memory buffer of 100,000 data points. It allows the user to configure versatile triggers, current and voltage ranges, and generate customized waveforms through GUI SW and SCPI programming. It supports USB, LAN and GPIB communication with PC and triaxial wiring via BNC connectors [10]. Even though commercial SMUs are state-of-the-art equipment, they are not intended for in-vitro experimentation. Standard models provide only 1 or 2 channels, which are insufficient for biological replicates. Modular and high-density channel models are significantly more expensive [11,12], constituting a considerable barrier for research institutions. Another constraint is the bulkiness, which complicates transportation and set-up in biological laboratories [7,13]. Concerning SW aspects, the limited memory of standard models is inadequate for long-term measurements, thus necessitating a PC for data logging [10,14]. The specialized syntax of the SCPI (Standard Commands for Programmable Instruments) language [15] requires a dedicated study for proper coding.

Hence, there is a need for a compact, affordable and user-friendly SMU that is compatible with in-vitro biosensing. Many low-cost prototypes, although not specifically designed for biosensors, have been developed [7,8,16,17,18]. They are based on Arduino microcontroller units (MCUs), digital-to-analog converters (DACs), analog-to-digital converters (ADCs) and additional HW and SW resources for communication and data management. With regard to biosensing, three recently proposed prototypes are noteworthy. The first, fabricated by Galanti and coworkers [9], integrates a commercial power supply, a general-purpose CPU and an analog circuit for the characterization of graphene field-effect transistors (FETs). It features full 4-quadrants operation, a high signal-to-noise ratio (SNR) and is provided with an electromagnetic interference (EMI) shielding box. The second, constructed by Bryantono et al. [19], is a STM32F1-based circuitry for the I – V tracing of FET biosensors. It presents an optimized architecture to minimize parasitic elements, an enclosing box for heat dissipation and a LabVIEW GUI for data analysis. The third, carried out by Park and colleagues [13], is a WeMos D1-controlled web-based system for measuring the transcharacteristic of ionic liquid-gated FETs. It is capable of automatically storing real-time measurements in a cloud server via wireless communication, generating large datasets for long-term experiments. Despite the achievements, these three systems lack a biosensor integration setup (thereby necessitating external probes), support only single device testing, are limited to I – V characterization (no transient or temporal measurements are reported) and require further HW resources for data analysis.

Based on the aforementioned limitations, in this work we have developed a multichannel, low-cost SMU platform for the I – V and transient response characterization of OECT biosensors. The platform controls a six-OECT set-up integrated in a standard culture plate, housed in a cell incubator during operation. The sensorized plate communicates with the breadboard via flat cables and re-attachable connectors, supporting in-vitro operations. The circuitry is based on Arduino Nano MCU and easy-to-use components, enabling automatic measurement and data logging. The platform features a maximum resolution of 1.22 mV / 78 nA (adequate for OECTs), a maximum per-channel current of ≈ 10 mA (≈ Ids,max) and a low power consumption (≈ 1 W), essential for long-term experiments. The breadboard-based circuitry facilitates prototyping and introduces an acceptable offset (≈ 30 µA) and negligible noise (≈ 150 nA) for this application. The user-friendly Arduino IDE is used for programs development and uploading to the MCU. This operator and PC-independent system is capable to continuously monitor the six OECT biosensors for long-term periods, achieving a large dataset. In this paper, we present the developed SMU platform and validate its OECT characterization performance through comparative measurements executed with the Keysight B2912A SMU. Results demonstrate that this prototype represents a valid alternative to standard SMUs in the context of OECT biosensors characterization. We believe that the proposed solution can support biological research by providing a portable, cost-effective and in-vitro compatible biosensing platform.

2. Materials and Methods

2.1. Hardware Design

2.1.1. SMU Circuit

The proposed SMU circuit, whose block diagram is shown in Figure 1, comprises five functional subsystems: digital, voltage-driving, voltage-sensing, current-sensing and power supply. The digital subsystem includes an MCU, a real-time clock (RTC) for timestamping, a memory card for data and timestamp storage, a red light-emitting diode (LED) for acquisition phase signalling and two multiplexers (MUXs) for converters selection. The MCU communicates with the RTC and MUXs via Inter-Integrated Circuit (I2C) and with the memory card via Serial Peripheral Interface (SPI) protocols. Although not essential for operations, a PC can be connected to the MCU through USB for real-time check. The voltage-driving subsystem consists of six DACs, three generating the gate voltages (V_g) and three the source voltages (V_s) for the OECTs. These voltages feed six operation amplifiers (op-amps) configured as voltage followers (VFs), three driving the Vg and three the Vs signals respectively. VFs source the required currents to the six OECTs while maintaining stable and load-independent driving voltages. OECTs drain voltages (V_d) are fixed to GND (0V). The voltage-sensing subsystem is composed of two ADCs which measure the three gate-source voltages (V_gs) and the 5V power supply, as explained below. The current-sensing subsystem is based on six high-side shunt resistors (R_s) in series with the corresponding OECTs, six instrumentation amplifiers (INAs) and three ADCs. OECTs drain-source currents (I_ds) are acquired with the shunt-resistance method: I_ds, driven by V_ds and flowing on R_s, produces by Ohm’s law a proportional voltage drop (V_Rs = I_ds ∙ R_s), which is amplified by the INA (V_out = G_INA ∙ V_Rs, where G_INA is the INA gain) and measured by the ADC. I_ds is computed as: I_ds = - V_out / (G_INA ∙ R_s) (1). The power supply subsystem includes a power supply unit (PSU), providing common GND, single + 8V and bipolar ± 8V lines, a voltage regulator (VR) generating single 5V supply (5V_VR), and the low-dropout regulator (LDO) integrated in the MCU, providing single 5V supply (5V_MCU). The MCU and VR are powered with + 8V, the op-amps with ± 8V, the converters with 5V_VR, and the digital subsystem with 5V_MCU. The two 5V sources are used to distribute the power load and prevent interference between the analog and digital circuits. These supply sources are respectively inspected by the MCU and the voltage-sensing ADC to detect eventual power anomalies.

The following sections describe the circuit components and the subsystems schematics, designed with KiCad (v9.0.5). Recurring components are omitted to avoid redundancy.

2.1.2. Digital Subsystem

The digital subsystem (Figure 2) implements Arduino Nano (Arduino) as MCU, which is a compact, low-power and user-friendly development board based on ATmega328 microcontroller (Microchip Technology) [20]. The other devices are based on breakout boards (hereafter referred to as modules), which are modular printed circuit boards (PCBs) containing the core integrated circuit (IC) and auxiliary components such as pull-up resistors, bypass capacitors and LEDs. Breakout boards allow ICs to be easily managed through their breadboard-friendly pins [21]. The RTC consists in the DS3231 module (Maxim Integrated, Adafruit Industries), a battery-integrated and highly accurate clock able to maintain date and time information for long periods [22]. MUXs are based on the TCA9548A module (Texas Instruments, Adafruit Industries), a low-standby current I2C switch able to select among up to eight devices [23]. The high-speed 8 GB USD500S microSDHC (Transcend) is used as memory card, integrated in the microSD card adapter module (Adafruit Industries).

2.1.3. Voltage-Driving Subsystem

The voltage-driving subsystem (Figure 3) exploits as DACs the rail-to-rail, low-power 12-bit MCP4725 modules (Microchip Technology, Adafruit Industries) [24]. The I2C pins of each module are connected to the corresponding I2C channel pins of the DAC-selecting MUX. The modules output signal is processed by a low-pass RC filter, constituted by a 1 KΩ resistor (CFR25 type) and the parallel of a 1 µF ceramic and a 100 nF film capacitors. This two-capacitors filter, with a cutoff frequency of ≈ 145 Hz, attenuates a broad range of noise frequencies while preserving the working signals. VFs are implemented with the LM358 IC (Texas Instruments), which is a cost-effective, low-offset and unity-gain stable dual op-amp [25].

2.1.4. Voltage and Current-Sensing Subsystems

The ADCs of the sensing subsystem (Figure 4) are based on the 4-channel, low-power 16-bit ADS1115 module (Texas Instruments, DFRobot), configured in differential mode for common noise reduction. By incorporating a low-drift voltage reference and a programmable gain amplifier (PGA), ADS1115 supports six input full-scale ranges (FSR) from ± 0.256 V to ± 6.144 V, with corresponding resolutions from 7.812 to 187.5 µV. It offers eight programmable data rate settings, from 8 samples per second (SPS) up to 860 SPS, clocked by an integrated oscillator [26]. The I2C pins are connected to the corresponding I2C channel pins of the ADC-selecting MUX. Shunt resistances consist in a precision 1 Ω resistor (E96 series), whose ends are connected to the positive and negative input pins of the INA121 instrumentation amplifier (Texas Instruments). This high-accuracy INA exhibits high common-mode rejection, low voltage offset and drift, and low bias and quiescent currents. By placing an external resistor RG of 511 Ω (E96 series) between its two R_g pins, the INA gain (G_INA = V_out / (V_in+ - V_in-)) was set to approximately 100, according to: G_INA = 1 + 50 kΩ / R_G [27]. Vout is subsequently processed by a low-pass filter identical to the voltage-driving subsystem one.

2.1.5. Power Supply Subsystem

The series connection of two PeakTech 6080A linear-regulated power supplies (PeakTech) is exploited as bipolar PSU, as shown in the subsystem schematic of Figure 5. A capacitor-input filter placed on the ± 8V lines, constituted by the parallel of a 100 nF film and a 10 µF tantalum capacitors, suppresses high-frequency noise on the power rail. The LM317 VR (Texas Instruments) is downstream of the PSU to generate 5VVR from +8V. This adjustable three-pin regulator supplies an output voltage equal to: V_out = V_ref ∙ (1 + R_adj / R_ref) + I_adj ∙ R_adj, where V_ref = 1.25 V, I_adj ≈ 50 µA, R_ref set to 220 Ω [28] and R_adj set to 660 Ω to obtain V_out = 5V. The PSU output was set to 8V to guarantee a minimum headroom of 3 V on the VR, and a 1 KΩ load resistor was placed at VO pin to achieve a minimum output current of 5 mA [29]. Moreover, LM317 is supported by the following external circuitry: a 100 nF bypass film capacitor at VI pin (input voltage stability), a 10 µF tantalum capacitor at ADJ pin (ripple rejection improvement), a 1 µF ceramic capacitor at VO pin (transient response improvement), and two protective 14001G diodes respectively between VO-VI pins and ADJ-VO pins (prevention of capacitors discharging across V_out).

2.1.6. OECT and Wiring Set-Up

The OECT set-up (Figure 6) was customized for integration in a 12-wells culture plate and interfacing with the breadboard. A 25 µm-thick flexible Kapton substrate (DuPont) was selected to allow sensors bending in the wells. Gold was employed for OECT electrodes, interconnections and contact pads, while PEDOT:PSS for the channel [30]. It was designed with a length of 200 µm and a width of 4 mm, while the gate as a 5 x 5 mm gold electrode. A polydimethylsiloxane (PDMS) layer was used for gold passivation. All materials were chosen for their biocompatibility and reliable operation in a biological environment. Following the approach of Fallager et al. [31], a guide-patterned PDMS mold was exploited as interconnector between pads and the stripped wires of a 10 cm-long ribbon cable. Silver paste was used to bond pads and wires, encapsulated by epoxy resin for protection towards the cell incubator humidity (95 % RH). This interconnection block was conceived to prevent short circuits between the bonded contacts and provide them mechanical stability. All aforementioned structures were designed by Rhinoceros CAD software (Robert McNeel). A wiring set-up was engineered for signal transmission between the OECT set-up and the breadboard, consisting of insulation-displacement contact (IDC) connectors linked by a 1 m-long ribbon cable (refer to Figure 9). Ferrite rings, enclosing the cable ends, were used for EMI attenuation. This wiring set-up was tailored for communication through a cell incubator and to let operators disconnecting the sensorized plate from the platform during biological procedures.

2.2. Hardware Fabrication

The designed circuit was assembled on a 230 x 175 mm breadboard (Digilent), powered by the PSU via screw terminals (Figure 7a). ADCs and connectors were integrated in the breadboard through 3D printed holders pasted on the breadboard sides. OECTs gold paths were fabricated according to the following cleanroom-based process (Figure 7b). First, a 10-nm thick Ti adhesion layer and a 100 nm-thick Au layer were sequentially deposited on Kapton by e-beam evaporation (FC-2000 evaporator, Ferrotec Temescal) [30,32]. Then, paths layout was transferred to the substrate by photolithography: the AZ1518 positive photoresist (MicroChemicals GmbH) was spin-coated (≈ 1.5 µm-thick at 4000 rpm, 30 s), soft-baked (110 °C, 60 s), exposed to the UV laser beam pattern at 468 mJ/cm2 (LW405 laser writer, Microtech) and developed for 3 minutes in a solution of AZ400K developer (MicroChemicals GmbH), diluted in distilled H2O (1:4 ratio). Finally, the exposed Ti-Au regions were removed by wet etching, respectively using TechniEtch ACI2 etchant (MicroChemicals GmbH) for Au (4 min treatment) and a solution of NH3 diluted in H2O2 for Ti (1:3 ratio, 2 min treatment). The OECT channel was fabricated according to the following protocol [32]. Pristine PEDOT:PSS (Clevios PH1000, Heraeus) was sonicated for 30 min, filtered and mixed with ethylene glycole (5% v/v) and (3-glycidoxypropyl)trimethoxysilane (GOPS) (1% v/v) for conductivity and stability improvement respectively. The obtained solution was diluted with distilled water (100% v/v) to decrease its viscosity at a proper range (1 – 10 cP) for Aerosol Jet Printing deposition (Aerosol Jet 200 Series printer, Optomec). After treating the substrate with 10 min of ozone for wettability enhancement, the OECT channel was printed by setting a fly velocity of 2 mm/s, a sheath gas pressure of 35 SCCM, a carrier gas pressure of 25 SCCM, a plate temperature of 60 °C and an ultrasonic atomizer current of 0.5 mA, following annealing at 120 °C for 30 min. The passivating PDMS layer was deposited by spin-coating (≈ 100 µm-thick at 1000 rpm, 30 s), masking the OECT sensing region and pads with dicing tapes, which were successively removed. Sensors layout was patterned by CO₂ laser marking (Laser Slider, Microla Optoelectronics), by setting a power of 12% (6 W), a frequency of 1500 Hz and a scan velocity of 50 mm/s. Pads and ribbon cable wires were bonded by aligning them on the interconnector guides and by applying silver paste, following curing at 60 °C for 1 h. The bonded region was encapsulated by mixing epoxy resin with hardener (1:1 ratio), waiting 1 h for partial curing, pouring the obtained gel on the bonded contacts and letting the gel completely curing overnight. Figure 7c reports the developed OECT set-up.

2.3. Software Development

Programs defining SMU operations (Arduino Sketches) were written in the open-source Arduino Programming Language (C/C++ dialect). The user-friendly Arduino IDE was used for library installation, Sketch writing and compilation, firmware upload and serial communication management between Arduino Nano and PC. The execution flow of the Sketches is visualized in Figure 8 and summarized as follows. It starts with the declaration and initialization of SW libraries, modules, variables and functions. The Arduino setup function initializes the communication protocols, the SD card data files (in CSV format) and the modules. The I2C clock frequency was set to 400 kHz (fast mode) and the UART speed to 115200 baud, these options being a compromise between data rate and communication stability. Within the Arduino loop function, after power supply and I2C checks, the developed source_measure function is sequentially called for each OECT. This function applies the V_ds and V_gs signals, records the I_ds, V_gs and time values in a data buffer and log them in the data files. These can be accessed after the measurement by transferring the SD card from the adapter module to PC. The used Sketches, documented through in-code comments, are open-source and can be found in the Supplementary Materials.

2.4. Keysight B2912A SMU Configuration

The proposed SMU was validated by comparing its characterization performance with respect to the commercial B2912A SMU (Keysight Technologies), presented in the Introduction. The instrument was configured with the same settings of the developed SMU. It was set in the 2-wired mode, with a cutoff frequency of 145 Hz and, at each test, with the same time, voltage and current settings of the current SMU. The Quick IV Measurement Software (Keysight Technologies) was used for B2912A configuration and measurements.

3. Results and Discussion

3.1. SMU Platform Development

Figure 9 shows the developed SMU platform operating in a biological laboratory. The equipment is portable, fully integrated and optimized for in-vitro experimentation. The system supports multi-sample, continuous and automated biosensing based on OECTs. Components are low-power consuming and the analog circuitry working range is ≈ 1 V, consuming a total average power of ≈ 1 W. Voltage and maximum current resolutions are respectively 1.22 mV and 78 nA, which are adequate for OECT characterization. By working with Ids max ≈ 10 mA, the 1 Ω shunt resistor ensures a maximum voltage drop of ≈ 10 mV on Vds signals, resulting in negligible perturbations of the driving voltages. The SD card enables PC-independent data collection without memory constraints, thereby overcoming the limited 2 KB SRAM of Arduino Nano. The three Arduino Nano serial communication interfaces (I2C, UART/USB, SPI) facilitate multi-components and multi-sensors integration. The total material cost of the platform is approximately € 400, resulting in a cost-effective instrument. The complete bill of materials and the components costs are provided in Table S1 of Supplementary Materials.

3.2. Calibration

By working in the converters linear region (≈ 20 % – 80 % of FSR), the driving and sensing subsystems output-input relationship can be modelled with a linear transfer function: Output = Gain ∙ Input + Offset, where Gain (ideally = 1) and Offset (ideally = 0) are the slope and the intercept respectively. Due to tolerances, Gain and Offset errors arise and can be compensated by SW calibration [33]. The subsystems were calibrated based on converters linear regions and OECT working ranges. For example, for source voltage drivers Offsets were measured with a 47 Ω test resistor (R_t) on the output pin to simulate the OECT channel resistance. The zero-scale (ZS) and full-scale (FS) voltages (V_ZS, ideal = 1 V, V_FS, ideal = 4 V) were respectively measured with R_t = 100 Ω and 470 Ω to simulate I_ds ≈ 10 mA [32]. The 1 mV-resolution Fluke 113 digital multimeter (Fluke) was used for voltages acquisition and the 10 mV-resolution PeakTech 6080A tuneable power supply (PeakTech) for voltages application. Output data are reported in Table 1a and 1b. Overall, Offset and Gain errors are respectively within ≈ 1 mV and 1 % for the voltage subsystem and within ≈ 0.2 mA and 5 % for the current subsystem. These errors can be mainly related to R_s contribution and to GINA tolerance, as well as reference variations [33].

Table 2. Offset and Gain errors for each subcircuit of the current sensing subsystem. Test conditions of each subcircuit are based on converters linear regions and OECT current ranges.

Currentsensing	∆Offset (mA)	∆Gain (%)	Conditions
Sensing Ch_1	0.005	4.3	FSR = 20.48 mA Rt = 47 Ω ZS: IZS 1.85 mA FS: IFS 12.5 mA
Sensing Ch_2	- 0.190	3.7
Sensing Ch_3	0.005	4.3
Sensing Ch_4	- 0.019	1.2
Sensing Ch_5	0.000	2.1
Sensing Ch_6	- 0.130	2.2

Table 2. Offset and Gain errors for each subcircuit of the current sensing subsystem. Test conditions of each subcircuit are based on converters linear regions and OECT current ranges.

Currentsensing	∆Offset (mA)	∆Gain (%)	Conditions
Sensing Ch_1	0.005	4.3	FSR = 20.48 mA Rt = 47 Ω ZS: IZS 1.85 mA FS: IFS 12.5 mA
Sensing Ch_2	- 0.190	3.7
Sensing Ch_3	0.005	4.3
Sensing Ch_4	- 0.019	1.2
Sensing Ch_5	0.000	2.1
Sensing Ch_6	- 0.130	2.2

3.3. Preliminary Validation

3.3.1 Residual Offset and Noise

After calibration, the residual offset and noise were assessed with R_t = 47 Ω (channel resistance simulation), FSR = 20.48 mA (OECT current range), null input voltages (V_Rt = 0 V), SPS = 128 (default), acquired samples = 70. The offset was calculated as the average of all acquired current samples, computed according to (1). The noise was defined and calculated as root mean square error (RMSE) of the samples with respect to the offset [9]. Outcomes (Figure 10 and Table S2) show that, despite of calibration, residual offsets between -30 – -45 µA and noises between 155 – 210 nA are still present. They can be attributed to ground loops of the breadboard [34], tolerances and temperature variations of components [19]. However, these errors can be considered acceptable with respect to the tolerances of the fabricated OECTs. Moreover, even B2912A is affected by an offset of 4 µA and a noise of 134 nA, this last slightly lower than the one of the prototype.

3.3.2. Test Resistor Characterization

The system was firstly evaluated by characterizing the already used 47 Ω test resistor, having a well-known behaviour. Test settings were based on OECT working ranges and were likewise applied on each channel and on B2912A. Currents were computed according to (1). The I – V characterization was executed by sweeping V_Rt between 0 and - 0.6 V (OECT Vds range [32]), applying FSR = 20.48 mA, SPS = 128, settle time = 200 ms (sufficient for OECT stabilization), and step size = 10 mV (compromise between resolution and slope fluctuations [18]). The acquired I_Rt – V_Rt curves, visualized in Figure 11a, agree with the expected trend, represented by the B2912A characteristic. As reported in Table S3, slopes are lower for approximately 1% with respect to B2912A ones, with a maximum reduction of ≈ 4%. Discrepancies among channels can be attributed to tolerances of test resistors and components, while deviations with respect to B2912A can be ascribed to the parasitic resistance of the breadboard tracks [7]. For the transient response characterization, VRt was stepped from 0 to -0.6 V with SPS = 860 (highest configurable value), resulting in a sampling rate of ≈ 2.5 ms, likewise set on B2912A. The obtained responses, plotted in Figure 11b, are coherent and follow the expected behaviour: currents are ≈ 0 mA at VRt = 0 V and stabilize at ≈ - 12.3 mA after a steep transient. Steady state channels currents differ for approximately 200 µA with respect to the B2912A one. This variation can be related to LM358, which simultaneously drives two loads while having a maximum output current of 20 – 30 mA [25]. Rise times of the obtained responses were calculated as the time required for current to rise from 0 % (last sample before step) to 100 % (first stable sample). As shown in Table S3, they are equal (∆tr = 0 %) or higher for 1 – 2 ms (10 – 20 %) than those obtained with B2912A (10 ms). Such time delays are supposed to arise from using ADS1115 in single-shot mode, which operates in low-power mode between acquisitions, but requires per-sample conversions [26].

3.4. OECT Characterization

The OECT characterization performance of the system was validated by comparing the acquired transfer characteristics (I_ds vs V_gs sweep) and transient responses (I_ds(t) vs V_gs step) with those obtained with B2912A. Test conditions were the same of resistor characterization (section 3.2 of Results), with the introduction of OECT-related settings: V_ds was fixed at - 0.3 V and V_gs was swept between 0.6 and 1.2 V for the transfer characteristic and stepped from 0 to 1.2 V for the transient response. These higher-values voltage ranges were applied to better appreciate current variations of the fabricated OECTs. The six-OECT set-up (Figure 6b) was tested in an electronic laboratory by placing in each well 2 mL of Phosphate-buffered saline (PBS) solution 1X (as done in [35]). In the prototype measurements, the OECT set-up was wired to the breadboard as shown in Figure 9 and characterized as explained in Section 4 of Materials and Methods. In the case of B2912A, for each measurement the SMU cables were manually wired and switched to the connector pins of the tested OECT. Results are reported in Figure 12 and Figure 13. First, Figure 12a and Figure 13a demonstrate that the developed SMU is able to correctly perform transfer characteristic and transient response characterizations on OECTs. Corresponding OECT curves are qualitatively coherent between the two instruments, allowing the different trends to be clearly characterized. Evidently, discrepancies with respect to B2912A are visible: as previously stated, these deviations can be ascribed to residual offsets, parasitic effects, components tolerances, temperature and reference voltages variations of the breadboard. The slower rise times of the spikes acquired with the prototype can be associated to the digital time delays, as also noticed with R_t. Nevertheless, it must be pointed out that the fabricated OECTs contribute considerably to the remarked variabilities. As shown in Figure S1, curves obtained by B2912A on three fabricated OECTs at two different rounds present appreciable deviations. Indeed, repeated measurements influence OECTs behaviour through ions channel doping and electrical stress, impacting on device stability [35]. Moreover, temperature variations of the environment can influence the PBS samples and OECTs conditions, slightly changing the device behaviour.

3.5. OECT Long-Term Continuous Monitoring

The actual aim of this work was to develop an operator-independent, long-term and continuous monitoring SMU for OECT biosensors. To validate these features, the platform was run continuously for 5 days while executing an automatic transient response characterization on OECTs every 30 minutes. The experimental set-up was as in previous section, except for setting V_ds = - 0.1 V and V_gs step = 0.3 V to simulate OECT-based cells observation [36]. Figure 14 shows the three first and last acquired transient responses for one representative OECT of the set-up. Figure 15 reports the extracted I_{ds on} values (I_ds values before step application) recorded every 30 minutes. The presented data confirm that the system is able to perform continuous characterization for periods comparable to ones of biological experiments. Thus, the platform can be used to monitor temporal changes on biological samples by analyzing time-varying figures of merit related to OECTs, such as the time constant (τ) of the transient response. For each OECT, the Ids on values exhibit an approximately stable trend, coherent with the absence of applied stimuli to the samples. To simulate the culture medium change, the 2 mL PBS sample of each well was replaced on the third day. Clearly, the replacement strongly affected the Ids on of OECTs 2, 3, 5 while minimally influencing OECTs 1, 4, 6. This effect can be attributed to the different stability of the PEDOT:PSS films, which responded distinctly when interfacing to the new solution. Thus, this effect must be taken into account in continuous biological experiments for proper data interpretation and analysis. Other I_{ds on} variations over time are attributable to OECTs repeated use, as mentioned in previous section.

4. Conclusions

In this work, we have carried out a low-cost SMU platform for OECT biosensors characterization. The system integrates open-source, fully documented and user-friendly HW-SW components, supporting prototyping and easy customization. The platform can characterize the vast majority of OECTs, exhibiting an operating range of ≈ 5 V / 10 mA, a resolution of 1.22 mV / 78 nA, an average noise lower than 200 nA and a maximum sampling rate of ≈ 2.5 ms. We have assessed the characterization performance of the system through I – V and transient response measurements on a commercial resistor and on a six-OECT in-vitro set-up, comparing the outcomes with ones obtained by Keysight B2912A SMU. The observed coherencies suggest that, for OECT-based in-vitro biosensing, the proposed system is more appropriate than the expensive, low-density channel and bulky conventional SMUs. More importantly, the platform durability and automatic operativity have been validated through a continuous 5-day test, with data acquisition every 30 minutes. It is also noteworthy that the platform can be applied to other classes of organic transistor biosensors, such as Electrolyte-Gated Organic Field Effect Transistors (EGOFETs) and Organic Charge Modulated Field Effect Transistors (OCMFETs). Their low operating voltages (< 2 V) and similarity in I – V behaviour to OECTs enable their characterization by the developed SMU. In contrast, the system is incompatible with Organic Field-Effect Transistors (OFETs) as they typically operate at higher voltages (~ 10 V) and frequencies (~ MHz) [1,2,4]. As this work concerns the initial version of the platform, we are aware of its limitations and we are working on its improvement. As unequivocal first upgrade, the breadboard is being transitioned to a PCB. This will strongly minimize offset and gain inaccuracies and will strengthen the stability of reference voltages. Secondly, the single channel MCP4725 DACs are being replaced by a single octal channels DAC7578 module (Texas Instruments, Adafruit Industries) for space and I2C channels saving, while the 860 SPS ADS1115 is being replaced by the 3300 SPS ADS1015 module (Texas Instruments, Adafruit Industries), increasing the maximum sampling rate. Finally, we are developing the PCB enclosure for environmental agents protection and a customized power supply to comply with biological laboratory regulations. Still, limitations that can be overcome by major design changes are the 1^st quadrant and the voltage-controlled only operation. Overall, we have paved the way to the realization of SMU platforms for the long-term and continuous in-vitro biosensing based on OECTs, combining portability, multi-sample monitoring and operator-independency. Next work will concern the application of the platform for the continuous observation of in-vitro cell models.