Highly Sensitive Detection of Urea Using Si Electrolyte-Gated Transistor with Low Power Consumption

We experimentally demonstrate Si-based electrolyte-gated transistors (EGTs) for detecting urea. The top-down-fabricated device exhibited excellent intrinsic characteristics, including a low subthreshold swing (SS) (~80 mV/dec) and a high on/off current ratio (~107). The sensitivity, which varied depending on the operation regime, was analyzed with the urea concentrations ranging from 0.1 to 316 mM. The current-related response could be enhanced by reducing the SS of the devices, whereas the voltage-related response remained relatively constant. The urea sensitivity in the subthreshold regime was as high as 1.9 dec/pUrea, four times higher than the reported value. The extracted power consumption of 0.3 nW was extremely low compared to other FET-type sensors.


Introduction
Urea is a crucial biomarker for diagnosing various malfunctions in the human body. High urea levels in the blood can indicate conditions such as indigestion, kidney malfunction, renal failure, urinary tract obstruction, and gastrointestinal bleeding. In contrast, its low levels can indicate hepatic failure, nephritic syndrome, and cachexia [1]. The urea concentration (pUrea = − log 10 [Urea]) in human blood ranges from 2.1 to 2.6 (3.5 mM to 7.5 mM).
Electrochemical biosensors have advantages such as fast response time, cost effectiveness, portability, and so on [8,9]. In order to further enhance their efficacy, improving key sensing parameters such as sensitivity, selectivity, and response time is of utmost importance. Recently, various types of transistors including ion-sensitive field-effect transistors (ISFETs) and biologically active FETs (BioFETs) have been developed to detect urea [10][11][12][13]. Nanostructure FET sensors have high sensitivity and can provide real-time and label-free detection [14,15]. However, the small sensing area of these sensors can limit the receptor density, resulting in insufficient output signals and significant device-to-device variations. Extended-gate FETs (EGFETs) are another type of ISFET consisting of a conventional FET and a separated sensing membrane connected to the gate [16,17]. However, the inherent interface between the gate and membrane generates additional parasitic capacitance and resistance, which worsens the sensitivity and reproducibility. More recently, electrolytegated FETs (EGTs) that use a functionalized gate electrode as the sensing surface have been developed [18][19][20][21][22]. The larger gate area, typically one order of magnitude larger than the channel area, is beneficial to achieve higher receptor density, thus enhancing output signals and reducing performance variations, which is crucial for the commercialization of BioFETs.
Herein, we investigated the electrical responses of Si-based EGTs for detecting urea. The device was fabricated using microfabrication technology. The Ag gate was functionalized with urease, and the current-voltage characteristics were experimentally measured at different pUrea values. The sensitivity and the limit of detection were analyzed in the subthreshold regime. Additionally, interference tests using typical biomolecules found in human blood were performed to evaluate the selectivity of the EGTs for detecting urea.
Prior to the experiments, a urea solution was prepared by dissolving urea powder in a 1 × PBS solution of pH 6. To test the selectivity of the device, other biomolecules such as glucose, AA, and KCl were also dissolved in the 1 × PBS solution with a pH of 6. The electrical characteristics of the device were measured using a semiconductor parameter analyzer (Keithley 4200, Keithley, Solon, OH, USA). The gate voltage (V G ) was applied in increments of 50 mV through a buffer solution, while the drain current (I D ) was measured with a fixed drain voltage (V D ) of 0.1 V. The source and body voltages (V S and V B ) were set to 0 V. I D was limited to 10 −7 A to prevent the degradation of the device. The I D −V G characterizations were performed after exposing the target solution of 20 µL for 10 min.

Fabrication of EGTs
The EGTs were fabricated using a top-down method ( Figure 1a) on a silicon-oninsulator wafer (p-type, 10 Ω·cm, (100)) with a 140 nm-thick top-Si layer and 400 nm-thick buried oxide layer as the substrate material. The top Si layer was thinned to 100 nm using thermal oxidation to ensure the uniform doping of deep Si during ion implantation. The active region, consisting of the source, drain, and channel, was formed using an I-line stepper and an inductively coupled plasma reactive-ion etching (ICP-RIE) process. Using electron-beam lithography and ICP-RIE etching, the channel region was then patterned into nanowires with a width of 50 nm, 80 nm, and 110 nm, respectively. Arsenic ions (5 × 10 15 /cm 2 , 60 keV) were implanted into the source and drain regions, followed by rapid thermal annealing (RTA) at 1000 • C for 20 s. A 5 nm-thick SiO 2 gate insulator was then thermally grown in a furnace at 800 • C for 5 min. Contact electrodes and transmission lines were formed using Ag/Ti (500 nm/50 nm) layers deposited via an e-beam evaporator and lift-off process. Finally, a 2 µm-thick SU-8 layer was passivated on the surface for electrical isolation, excluding the channel, gate electrode, and contact pads (Figure 1b).

Functionalization of EGTs
As a urea receptor, the urease was immobilized on the gate area. The gate electrode was first treated with UV/ozone for 90 s under a light intensity of 200 µW/cm 2 to generate hydroxyl groups (OH − ). The surface was then exposed to vaporized APTES at 55 °C for 1 min, rinsed with anhydrous ethanol to remove unbound APTES molecules, and dried using N2 blowing. The devices were then immersed in a glutaraldehyde solution (2.5 %, 1 × PBS, pH 7.4) for 90 min, washed with 1 × PBS and DIW, and dried with N2 blowing. Finally, the devices were exposed to a urease solution (10 mg/mL, 1 × PBS, pH 7.4) for 18 h in a humid environment at 4 °C, followed by rinsing with 1 × PBS and DIW and drying with N2 blowing.
The urea functionalization on the Ag gate was verified using atomic force microscopy (AFM, VEECO, New York, NY, USA), as shown in Figure 2. The average roughness values were determined to be 0.7 nm for the bare Ag surface, 0.13 nm after APTES/GA treatment, and 4.2 nm following the immobilization of urease, respectively. The reduction in roughness observed after APTES/GA treatment can be attributed to the effective filling of APTES molecules within the Ag grain boundaries [23].  Figure 3 shows the intrinsic transfer curve (ID vs. VG) and gate leakage current (IG) of the EGT device. It exhibits excellent n-type characteristics including a low subthreshold swing (SS) of ~80 mV/dec, high on/off current ratio (ION/IOFF) of ~10 7 , and low threshold

Functionalization of EGTs
As a urea receptor, the urease was immobilized on the gate area. The gate electrode was first treated with UV/ozone for 90 s under a light intensity of 200 µW/cm 2 to generate hydroxyl groups (OH − ). The surface was then exposed to vaporized APTES at 55 • C for 1 min, rinsed with anhydrous ethanol to remove unbound APTES molecules, and dried using N 2 blowing. The devices were then immersed in a glutaraldehyde solution (2.5 %, 1 × PBS, pH 7.4) for 90 min, washed with 1 × PBS and DIW, and dried with N 2 blowing. Finally, the devices were exposed to a urease solution (10 mg/mL, 1 × PBS, pH 7.4) for 18 h in a humid environment at 4 • C, followed by rinsing with 1 × PBS and DIW and drying with N 2 blowing.
The urea functionalization on the Ag gate was verified using atomic force microscopy (AFM, VEECO, New York, NY, USA), as shown in Figure 2. The average roughness values were determined to be 0.7 nm for the bare Ag surface, 0.13 nm after APTES/GA treatment, and 4.2 nm following the immobilization of urease, respectively. The reduction in roughness observed after APTES/GA treatment can be attributed to the effective filling of APTES molecules within the Ag grain boundaries [23].

Functionalization of EGTs
As a urea receptor, the urease was immobilized on the gate area. The gate electrode was first treated with UV/ozone for 90 s under a light intensity of 200 µW/cm 2 to generate hydroxyl groups (OH − ). The surface was then exposed to vaporized APTES at 55 °C for 1 min, rinsed with anhydrous ethanol to remove unbound APTES molecules, and dried using N2 blowing. The devices were then immersed in a glutaraldehyde solution (2.5 %, 1 × PBS, pH 7.4) for 90 min, washed with 1 × PBS and DIW, and dried with N2 blowing. Finally, the devices were exposed to a urease solution (10 mg/mL, 1 × PBS, pH 7.4) for 18 h in a humid environment at 4 °C, followed by rinsing with 1 × PBS and DIW and drying with N2 blowing.
The urea functionalization on the Ag gate was verified using atomic force microscopy (AFM, VEECO, New York, NY, USA), as shown in Figure 2. The average roughness values were determined to be 0.7 nm for the bare Ag surface, 0.13 nm after APTES/GA treatment, and 4.2 nm following the immobilization of urease, respectively. The reduction in roughness observed after APTES/GA treatment can be attributed to the effective filling of APTES molecules within the Ag grain boundaries [23].  Figure 3 shows the intrinsic transfer curve (ID vs. VG) and gate leakage current (IG) of the EGT device. It exhibits excellent n-type characteristics including a low subthreshold swing (SS) of ~80 mV/dec, high on/off current ratio (ION/IOFF) of ~10 7 , and low threshold  Figure 3 shows the intrinsic transfer curve (I D vs. V G ) and gate leakage current (I G ) of the EGT device. It exhibits excellent n-type characteristics including a low subthreshold swing (SS) of~80 mV/dec, high on/off current ratio (I ON /I OFF ) of~10 7 , and low threshold voltage (V TH ) of~0.65 V. The low leakage current (<10 pA) and negligible hysteresis (inset of Figure 3) guarantee a reliable and reproducible operation during sensing responses.

Intrinsic Electrical Characteristics
voltage (VTH) of ~0.65 V. The low leakage current (<10 pA) and negligible hysteresis (inset of Figure 3) guarantee a reliable and reproducible operation during sensing responses.  Figure 4a shows the current monitoring result for 1 × PBS with and without urea (pUrea 0.5) at a fixed VG of 0.3 V. Five devices were used to obtain each data point, and the average value and 1σ of those measurements are plotted. Over time, ID continuously decreased for the urea solution, whereas it remained constant for 1 × PBS. Since the response for the urea saturated within the first 10 min of exposure, 10 min exposure time was used for all experiments. The urea in a solution reacts with the urease on the Ag surface to produce the OHions and to increase the pH value. Figure 4b shows the change in the transfer curve as the device is exposed to different pUrea values. The initial state denotes the ID−VG curve without urea. An increase in the urea concentration or a decrease in the pUrea value caused the curve to shift toward a positive VG direction.

Sensing Characteristics
The current-related response (RI) is defined as follows [24,25]: where ID0 and ID1 represent drain currents at a fixed VG0 before and after the reaction, respectively. VG0 of 0.3 V was selected to calculate RI from the data presented in Figure 4b. The voltage-related response (RV) is defined as follows [26]: where VG0 and VG1 represent gate voltages at a fixed ID0 before and after the reaction, respectively. The ID0 of 3 nA and VG of 0.3 V were chosen because the current was significantly higher than the noise level (~1 pA), and it ensured the device was operated in the subthreshold regime below the VTH of 0.65 V.     Figure 4b shows the change in the transfer curve as the device is exposed to different pUrea values. The initial state denotes the I D −V G curve without urea. An increase in the urea concentration or a decrease in the pUrea value caused the curve to shift toward a positive V G direction.

Sensing Characteristics
The current-related response (R I ) is defined as follows [24,25]: where I D0 and I D1 represent drain currents at a fixed V G0 before and after the reaction, respectively. V G0 of 0.3 V was selected to calculate R I from the data presented in Figure 4b. The voltage-related response (R V ) is defined as follows [26]: where V G0 and V G1 represent gate voltages at a fixed I D0 before and after the reaction, respectively. The I D0 of 3 nA and V G of 0.3 V were chosen because the current was significantly higher than the noise level (~1 pA), and it ensured the device was operated in the subthreshold regime below the V TH of 0.65 V.
To achieve a high sensitivity, FET-based biosensors should be operated in the subthreshold regime [27,28], where I D and SS are defined as follows [29]: where µ n is the electron mobility; C ox is the oxide capacitance; C it is the interface state capacitance; W is the channel width; L is the channel length; k is the Boltzmann constant; T is the temperature; q is the electron charge; and C d is the depletion capacitance in the channel. R I at a fixed V D can also be expressed as follows: where V TH0 and V TH1 represent threshold voltages before and after the reactions, respectively. Therefore, R I can exponentially increase as R V increases. Figure 5 illustrates the dependence of R I and R V with respect to the SS value at a pUrea of 0.5. The extracted R V was approximately 120 mV, displaying a consistent behavior across different SS values. In contrast, R I was inversely proportional to SS values and decreased as SS increased. The exponential calibration curve of R I and R V was obtained as R I = 100 × (e ln(10) × 122/SS − 1) and R V = 61.2 × e −SS/36.0 + 112. Figure 6 shows the relationship between the R I and pUrea for different SS values. The EGTs with low SS values (75 < SS < 85) exhibit a saturated R I of 3.3 × 10 3 (%) at a pUrea of 1.0. Conversely, EGTs with higher SS values (95 < SS < 105) exhibit a lower saturated R I of 1.3 × 10 3 (%) at the same pUrea value. As determined by the slope of the logistic fitted line of R I , the consistent urea sensitivity of 1.9 dec/pUrea is achieved across all SS values, which is more than four times higher than the previous results ( Table 1). The dynamic range, defined as the difference between 10% and 90% of the maximum sensitivity, is observed to be between pUrea 2.0 and pUrea 3.4 regardless of SS values, which fully encompasses the clinical range of human urea. The limit of detection (LOD) of R I , determined using the 3-σ method from the logarithmic trend line [30,31], is as low as pUrea 3.22 for 75 < SS < 85, pUrea 3.04 for 85 < SS < 95, and pUrea 2.99 for 95 < SS < 105. Figure 7 shows the relationship between R V and pUrea over the whole range of SS (75 < SS < 105). Each point represents the average of five different devices. A dynamic pUrea range of 1.8-2.9 was obtained. The urea sensitivity extracted from the R V curve was 120 mV/pUrea, with a LOD of pUrea 3.14.    Figure 7 shows the relationship between RV and pUrea over the whole range of SS (75 < SS < 105). Each point represents the average of five different devices. A dynamic pUrea range of 1.8-2.9 was obtained. The urea sensitivity extracted from the RV curve was 120 mV/pUrea, with a LOD of pUrea 3.14.
Power consumption is a crucial factor for portable biosensing applications. The calculated power consumption with VD = 0.1 V and ID = 3 nA is significantly lower than that of other FET-type biosensors due to the operation in the subthreshold regime. Table 1 compares the sensing performance of the EGT with that of previously reported FET-type sensors.    Figure 7 shows the relationship between RV and pUrea over the whole range of SS (75 < SS < 105). Each point represents the average of five different devices. A dynamic pUrea range of 1.8-2.9 was obtained. The urea sensitivity extracted from the RV curve was 120 mV/pUrea, with a LOD of pUrea 3.14.
Power consumption is a crucial factor for portable biosensing applications. The calculated power consumption with VD = 0.1 V and ID = 3 nA is significantly lower than that of other FET-type biosensors due to the operation in the subthreshold regime. Table 1 compares the sensing performance of the EGT with that of previously reported FET-type sensors.  Power consumption is a crucial factor for portable biosensing applications. The calculated power consumption with V D = 0.1 V and I D = 3 nA is significantly lower than that of other FET-type biosensors due to the operation in the subthreshold regime. Table 1 compares the sensing performance of the EGT with that of previously reported FET-type sensors.  Figure 8 shows the RI of various common interferents found in human blood including glucose (100 mM, 1 × PBS), AA (100 µM, 1 × PBS) and KCl (10 mM, 1 × PBS), and RI of urea (100 mM) with unmodified EGT (without urease) to demonstrate the lack of nonspecific binding of the device. All devices except the unmodified EGT were functionalized using the same method described in Section 2.3. Each data point corresponds to the average measurement obtained from five devices. The interference response for individual ingredients was found to be minimal, with less than a 10% change compared to the signal observed with urea at a concentration of 3.16 mM. A negligible RI of the unmodified sample indicates that there is no nonspecific binding between the urea and a Ag surface. Although the RI of the urea/mixture sample was reduced due to the opposite signal direction of the interferents compared to urea, it was still detectable at a sufficient level. This suggests the stability of the EGT sensing performance and the minimal impact of interfering ions on its urea response.   Figure 8 shows the R I of various common interferents found in human blood including glucose (100 mM, 1 × PBS), AA (100 µM, 1 × PBS) and KCl (10 mM, 1 × PBS), and R I of urea (100 mM) with unmodified EGT (without urease) to demonstrate the lack of nonspecific binding of the device. All devices except the unmodified EGT were functionalized using the same method described in Section 2.3. Each data point corresponds to the average measurement obtained from five devices. The interference response for individual ingredients was found to be minimal, with less than a 10% change compared to the signal observed with urea at a concentration of 3.16 mM. A negligible R I of the unmodified sample indicates that there is no nonspecific binding between the urea and a Ag surface. Although the R I of the urea/mixture sample was reduced due to the opposite signal direction of the interferents compared to urea, it was still detectable at a sufficient level. This suggests the stability of the EGT sensing performance and the minimal impact of interfering ions on its urea response.  Figure 8 shows the RI of various common interferents found in human blood including glucose (100 mM, 1 × PBS), AA (100 µM, 1 × PBS) and KCl (10 mM, 1 × PBS), and RI of urea (100 mM) with unmodified EGT (without urease) to demonstrate the lack of nonspecific binding of the device. All devices except the unmodified EGT were functionalized using the same method described in Section 2.3. Each data point corresponds to the average measurement obtained from five devices. The interference response for individual ingredients was found to be minimal, with less than a 10% change compared to the signal observed with urea at a concentration of 3.16 mM. A negligible RI of the unmodified sample indicates that there is no nonspecific binding between the urea and a Ag surface. Although the RI of the urea/mixture sample was reduced due to the opposite signal direction of the interferents compared to urea, it was still detectable at a sufficient level. This suggests the stability of the EGT sensing performance and the minimal impact of interfering ions on its urea response.

Conclusions
We investigated the label-free sensing response of urea using Si-based EGTs. The device was fabricated using a top-down microfabrication technique and operated in the subthreshold regime to enhance the sensitivity. The EGTs with a low SS could further increase the current-related responses. The urea sensitivities determined from R I and R V were as high as 1.9 dec/pUrea and 120 mV/pUrea, respectively. The calculated power consumption was as low as 0.3 nW and three orders of magnitude lower compared to previously reported results. In addition, the extracted dynamic range fully covered the human clinical range of urea. These results suggest that Si-based EGTs have significant potential for clinically diagnosing urea-related diseases.