Hydrogen Sensing Properties of FET-Type Sensors with Pt-In₂O₃ at Room Temperature

Abstract

In this paper, a field effect transistor (FET)-type sensor with Pt-decorated In₂O₃ (Pt-In₂O₃) nanoparticles is fabricated for detecting H₂ gas at room temperature. A pulsed measurement method is adopted to continuously alternate between pre-biasing the gate and reading the drain current of the FET-type sensor. This method effectively reduces the drift in the sensing signal. It is also found that negative pre-bias voltages can dramatically shorten the recovery time of the sensor after sensing H₂, while positive pre-bias voltages have the opposite effect. The H₂ sensing performance of the sensor is characterized under the enhancement of a pulsed negative pre-bias. By calculating and comparing the root mean square, signal-to-noise ratio, and detection limit of the sensor under different operating regions, it is found that the sensor has the best sensing performance in the subthreshold region, which is suggested to be the optimum operating region for FET-type sensors. In addition, the presence of oxygen significantly consumes the hydrogen molecules and reduces the room-temperature H₂ sensitivity of the sensor. The proposed sensor presents promising H₂ sensing properties, and this research could be a guide for the use of FET-type sensors in more gas detection applications.

1. Introduction

Hydrogen is a promising energy carrier because it has a high calorific value and nonpolluting combustion products (water) [1,2,3]. It has been widely developed for the applications in vehicle energy and aviation fuel [4,5]. However, H₂ is also one of the most dangerous inflammable and explosive gases under normal temperature and pressure, and requires strict monitoring for its safe use and storage. Therefore, the manufacture of reliable hydrogen detectors and controllers that operate at room temperature is required, while ensuring safety and a low power consumption.

Up to now, a lot of effort has been put into developing novel nanoscale metal oxide semiconductors for H₂ detection at low temperature, as nanostructures have an ultrahigh surface-to-volume ratio that provides more active centers for surface reactions [6]. In addition, noble metals, such as Pt, Pd, and Au, have been utilized to decorate the nano metal oxides, since their catalytic effect can promote the dissociation of gas molecules and further enhance the gas-sensitive properties at room temperature [7]. Jieting Zhao et al. reported that Pt-SnO₂ prepared using a solution reduction method exhibited a response of up to 9700 (resistance ratio) for 1% H₂ at room temperature [8]. Riya Wadhwa et al. used Pt-modified MoS₂ flacks to achieve a significant resistivity change (23% at 50 ppm H₂) at low hydrogen concentrations (0.5–50 ppm) at room temperature, which the authors attributed to the Pt NP-induced H₂ spillover effect [9]. Among the metal oxide semiconductors, In₂O₃ has attracted attention because of its high electrical conductivity, large band gap, and easy synthesis [10,11,12]. Sensors based on nanostructured In₂O₃ with different morphologies have been reported for H₂ detection at relatively low temperatures [13,14]. These sensors are based on changes in the resistance of In₂O₃ with a variation in H₂ concentration, which might be limited by problems such as large sensing areas and a high energy consumption. Recently, there has been an increasing interest in field effect transistor (FET)-type sensors because they are compatible with standard CMOS technology [15,16]. They can be integrated with other functional circuits on the same chip for complex, high-precision gas species identification and concentration detection. Consequently, it is possible to scale down the size of the sensors, reduce the power consumption, and fabricate sensors in a reproducible way, meeting the requirements for miniaturized, integrated, and intelligent sensors [17,18].

In an FET-type gas sensor, the sensing material can be used as a channel or a gate. When the sensing material serves as the channel of the FET, which is known as a TFT sensor, the sensing mechanism is similar to that of a resistive device, but the gas adsorption and desorption generates significant noise [19]. When the sensing material is used as the gate of the FET, the change in gas concentration will regulate the work function of the sensing material and thus indirectly control the electron concentration in the channel, producing a sensing current signal. This sensing principle and sensor structure can effectively suppress noise and prevent contamination during the production of sensing materials [20]. In this case, the importance of the resistivity of the sensing material comes to the fore. This is because a high resistance means significant delays and energy consumption, which can be fatal for large sensing systems. As a result, semiconductor materials with a high conductivity, such as In₂O₃, are more suitable for the application of FET sensors. Although FET sensors offer so many advantages, there is still a need for more in-depth exploration of how to solve the signal drift that FETs are prone to, further reduce the temperature of H₂ sensing and efficiently improve the H₂ sensing characteristics of FET-type sensors.

In this work, nanoscale In₂O₃ decorated by Pt (Pt-In₂O₃) is used to serve as the sensitive gate. The sensing material is deposited on the FET platform in the final step of sensor fabrication by using an inkjet printing process that is highly compatible with CMOS processes. The operating temperature is reduced to room temperature by the Pt decoration. A pulse measurement method is proposed for the sensor to suppress common signal drift problem in FETs and improve the recovery rate of the sensor. Moreover, the effect of oxygen in room-temperature H₂ sensing is also discussed.

2. Materials and Methods

2.1. FET Platforms

Figure 1 shows the SEM images and the schematic diagrams of the FET-type sensor. The sensor has four electrodes, including a drain, body, source, and control gate (CG), and an n+ polysilicon floating gate (FG) passivated by a SiO₂/Si₃N₄/SiO₂ (ONO) insulator stack. One end of the FG is located above the channel for controlling the concentration of carriers in the channel, and the other end extends 10 μm away from the channel to be capacitively coupled with the interdigitated CG. The sensing material is deposited on top of the coupling region of the two gates, which is enclosed by the dashed line in Figure 1b. The channel length and width of the FET platform are 2 μm and 2.4 μm, respectively. Figure 1c,d present the schematic diagram of the cross-section along A-A′ and B-B′ in Figure 1b. Due to the unique FG structure, the channels and oxides are effectively insulated from contaminants and molecules from the deposition of sensing materials as well as from the air.

In this paper, a pMOSFET platform is adopted because it has lower 1/f noise than the nMOSFET [21]. The fabrication of FET substrates began with 6-inch n-type single-crystal silicon wafers. The active regions of the FET platforms were defined and separated by field oxides using the local oxidation of silicon process. Buried channels were formed using the ion implantation process to reduce the carrier scattering at the interface between the channel and the gate oxide, thereby reducing the noise of the FET. A gate oxide of 10 nm was grown via dry oxidation, followed by the deposition and patterning of 350 nm-thick in situ n⁺-doped polysilicon as the CG. After forming the source and drain regions using ion implantation, layers of SiO₂ of 10 nm, Si₃N₄ of 20 nm, and SiO₂ of 10 nm were sequentially deposited on the wafer, i.e., ONO passivation, as shown in Figure 1. The ONO passivation layer protects the active region and the FG of the FET, which further reduces noise and increases the lifetime of the sensor. After defining the contact holes via photolithography, metal stacks consisting of 30 nm-thick Cr and 50 nm-thick Au were deposited and patterned through the lift-off process to form the CG, drain, source, and body electrodes. At this point, the structure of the FET was essentially complete. However, to further protect the FET from contamination during the fabrication of sensing materials, a final SU-8 passivation layer was spin-coated on the wafer and patterned via the photolithographic process, exposing only the coupled FG-CG region, i.e., the opened SU-8 window in Figure 1a, as well as all contact pads.

2.2. Sensing Materials

The Pt-In₂O₃ sensing material was deposited on top of the SU-8 window via an inkjet printing process. The ink was formulated using In₂O₃ nanopowders (≤100 nm diameter) and 8 wt% H₂PtCl₆ (in H₂O), all of which were purchased from Sigma-Aldrich (USA). Firstly, In₂O₃ nanopowders were ultrasonically dispersed in ethanol (99%) at room temperature. The original H₂PtCl₆ solution was further diluted to the desired concentration with deionized water and then mixed with the prepared In₂O₃ suspension to serve as the ink. After printing (Omni Jet 100), a two-hour annealing process at 300 °C in air was conducted to fully evaporate the solvent from the Pt-In₂O₃ sensing layer. The weight concentration (wt.%) of Pt was set to be 2%, 5% and 10%.

2.3. Measurement Setups

Figure 2 presents the experimental setup. During characterization of the hydrogen sensing performance of the sensor, dry synthetic air (20 vol.% of O₂ and 80 vol.% of N₂) was used as the reference and carrier gas to simulate the detection of H₂ leaks in air environments. Hydrogen samples of various concentrations were prepared by mixing 1% H₂ diluted in N₂ (with an uncertainty of 2%) with synthetic air with MFC 3 and MFC 2; at the beginning, a reference gas of 400 sccm controlled by MFC 1 was blown onto the sensor, and when the tests of the gas concentration response were initiated, the reference gas was switched to a hydrogen sample of a certain concentration at 400 sccm. In addition, the effect of O₂ on the H₂ detection characteristics was also investigated in this paper. During this test, the reference and carrier gases were all replaced with N₂, and the H₂ sensing characteristic tests were repeated. All tests were conducted using an Agilent B1500A at room temperature.

3. Results and Discussion

3.1. Pt Concentration Impact

The basic electrical and sensing properties of the FET-type sensor are shown in Figure 3. Figure 3a shows the double-sweep pulsed IV (PIV) and DC IV curves measured at room temperature. Figure 3b illustrates the response of the sensor with different Pt concentrations (2 wt.%, 5 wt.% and 10 wt.%) to 0.1% H₂ when a DC bias was applied to the electrodes. The response (R) of the sensor is defined as the following (Equation (1)):

$$Response=\frac{I_{\mathrm{D}\_\mathrm{B}}-I_{\mathrm{D}\_\mathrm{G}}}{I_{\mathrm{D}\_\mathrm{B}}}\times 100\mathrm{\%}$$

(1)

where I_{D_B} and I_{D_G} denote the drain currents of the sensors when exposed to the reference gas and to the H₂ target gas mixture, respectively. It can be visualized from Figure 3b that the response of the sensor to the H₂ gas increases with an increasing Pt concentration in the sensing layer, and the response becomes negligible when the Pt concentration is set to 2 wt.%.

Moreover, all three sensors with different Pt concentrations presented current drift, which affected the H₂ sensing accuracy. This phenomenon can be attributed to charge traps in the In₂O₃ nanoparticles and their interfaces, corresponding to the hysteresis in the DC IV in Figure 3a.

3.2. Pulse Measurement

To suppress the current drift, a pulse biasing scheme that continuously alternated between pre-biasing the gate and reading the drain current of the FET-type sensor was designed, as shown in Figure 4. Figure 4a shows the pulse schemes applied to CG (upper) and D (lower) electrodes. The pulse cycle of the scheme was divided into a holding phase (t_h) and a reading phase (t_r). During t_h, a prebias voltage (V_pb) was applied to the CG, which equals zero in Figure 4 (V_CG = V_pb = 0 V), and the D was grounded (V_DS = 0 V). In this period, the drain current was not recorded, so unwanted power consumption was removed. During the t_r phase, the V_CG and V_D values were simultaneously set to V_gr and V_dr, respectively, and the drain current values were recorded to form a discontinuous sensing signal. Here, t_h and t_r were set to 1 s and 20 µs, respectively. In Figure 4b, transient responses of the FET-type sensors are plotted as a parameter of Pt wt.% at room temperature. The drift caused by charge trapping is eliminated, since t_r is applied for a shorter duration than the charge trapping time. The results illustrate that the sensor based on a 10 wt% Pt sensing layer has the highest hydrogen response.

Figure 4c demonstrates the effect of the polarity of V_pbs on the H₂ sensitivity characteristics of the sensor with 10 wt% Pt. The dynamic response of the sensor to 0.1% H₂ was tested at a V_pbs of −0.5 V and +0.3 V, separately, and compared with the result at a V_pb of 0 V. The response time is defined as |I_D|, reaching 90% of the maximum change under H₂ gas exposure, and the recovery time is defined as |I_D|, falling to 90% of the difference between the maximum current and the baseline. The recovery times of the sensor after 0.1% H₂ exposure at a V_pbs of −0.5 V, 0 V and +0.3 V are about 700 s, 2000 s and 5000 s, respectively. That is, a V_pb of −0.5 V significantly speeds up sensor recovery. Meanwhile, the response at a V_pb of −0.5 V decreases by 14% compared to that at a 0 V pre-bias, and the response at a V_pb of +0.3 V increased by 9.3% compared to that at a 0 V pre-bias. This suggests that the effect of pre-bias on the response is relatively small compared to the recovery time. The response time and recovery time of the sensor at a V_pbs of −0.5 V (results in Figure 4c) were also compared to those of some typical FET-type sensors reported recently [22,23,24,25], which are shown in Table S1 of the Supplementary Material. In consideration, negative pre-biases were chosen to enhance the sensing characteristics in the later tests.

It is worth mentioning that in our previous study, the negative bias voltage had a similar effect on SnO_x, which accelerated the recovery of SnO_x after sensing the reducing gas H₂S [16]. We attributed this to the fact that the negative bias modulates the energy band bending of the SnO_x surface, allowing oxygen molecules to chemisorb more quickly on the surface of the sensing material during the recovery process. However, unlike previous studies, the sensing material in this paper is In₂O₃ modified with Pt. Although both In₂O₃ and SnO_x are n-type semiconductor materials, the sensing process of the sensor in this paper involves complex catalysis under the modification of Pt. Therefore, the mechanism for the effect of the pre-bias voltage in this study needs to be studied further in continued research.

3.3. H₂ Concentration Sensing

Next, the H₂ sensing performance of the sensor was measured under the enhancement of pulsed negative pre-biases, which is shown in Figure 5. According to the characteristics of FETs, the FET-type sensors in this paper can operate in three different states when different bias voltages are applied to the control gate and drain, i.e., subthreshold, linear, and saturation regions. To find the most suitable working region for H₂ detection, controlled experiments were conducted for the three different working regions. The other conditions of the test were kept constant, and a pre-bias voltage of −0.5 V was applied during the measurement with t_h = 1 s and t_r = 20 µs.

Figure 5a–c show the transient response of the sensor to H₂ samples of various concentrations in the subthreshold, linear and saturation regions, respectively. In the initial stage of all the tests, 400 sccm of dry air was blown onto the sensor. From 50 s in Figure 5a–c, the air was replaced by a H₂ sample of a certain concentration and maintained for 300 s for the sensor to respond. After that, the gas flowing into the chamber was changed back to the reference gas, and the sensor entered the recovery period. The hydrogen concentration was set from 0.1% to 0.6% with a step of 0.1%. All responses to H₂ under the three working regions increase with an increase in H₂ concentration. Figure 5d shows the responses calculated using Equation (1) and the linear fittings of the relationship between the responses and the H₂ concentrations for the three operating conditions of the sensor. According to the results, in the subthreshold region, the sensor has the largest response–concentration curve slope, and the sensor recovered to about 85% within 500 s for all the H₂ concentrations. Meanwhile, the sensor in the linear region exhibits the lowest response values, as well as the flattest slope, which are even negligible compared to those of the other two regions.

Next, the sensor noise, signal-to-noise ratio (SNR) and detection limit (DL) of the sensor at different operating regions were calculated. The sensor noise was determined by the root-mean-square deviation (RMS) of the baseline, which was calculated from the fifth-order polynomial fit of 10 data points on the baseline before gas sensing using Equation (2) [26].

$$RMS=\sqrt{\frac{\Sigma {\left(x_i-x\right)}^2}{N}}$$

(2)

where x_i is the measured data, x is the corresponding fitted value and N is the number of data points (N is 10 in the present case). The SNR is calculated by Equation (3) [27].

$$SNR=\frac{R}{RMS}$$

(3)

The calculated SNR versus H₂ concentration was plotted and is shown in Figure 5e. As the H₂ concentration increased, the SNR of the sensor increased in all three operating regions, but all SNRs for the subthreshold region were higher than those for the other two regions. The DL of the sensor is calculated using Equation (4) [26].

$$DL=\frac{3RMS}{a}$$

(4)

where a is the slope of the linear calibration curve of response R vs. H₂ concentration in Figure 5d. The DL values of the sensor in linear, saturation and subthreshold regions are 308 ppm, 189 ppm and 91 ppm, respectively. Combining the above measurements and calculations, the sensor has the highest response and SNR and the lowest DL when it is in the subthreshold region. Therefore, the FET-type sensor in different operating regions can exhibit different hydrogen sensing characteristics, and the subthreshold region is suggested to be a relatively suitable choice.

3.4. Impact of O₂ on H₂ Sensing

In previous sections, all the results were measured with dry synthetic air as the carrier gas to obtain a certain concentration of H₂ in the target gas. Pt is a kind of powerful catalyst that can promote the reaction between H₂ and O₂ even at room temperature [28]. Owing to the air carrier gas, O₂ also exists in the target gas samples during the gas sensing measurements, so that H₂ is consumed by O₂ on the surface of Pt [28]. The catalytic effect of Pt may promote the dissociation of O₂ and reactions between oxygen and hydrogen species to produce water [9], [29,30]. Therefore, in the above measurements, the exact amount of H₂ molecules that results in a change in sensing signal is limited. In other words, ambient O₂ plays a non-negligible role in real H₂ detection.

In this section, the effect of O₂ on H₂ sensing was investigated by replacing the synthetic air with pure N₂ to serve as carrier and reference gases. The gas sensing experiments in Figure 5c were repeated with the difference that O₂ was removed from all gas samples this time. The dynamic response curve is shown in Figure 6.

Based on Equations (1) and (4), the R and SNR values were also calculated for the N₂ case and plotted together with the results of synthetic air case, as shown in Figure 7a,b for comparison. The R and SNR of the sensor in the H₂ concentration variation range of 0.1% to 0.4% were significantly improved when N₂ served as the carrier and reference gas, but the R of the sensor appeared to be saturated when the H₂ concentration reached 0.5%. This phenomenon can be explained by the fact that the number of hydrogen molecules is retained without the consumption of O₂, but at the same time, the presence of large quantities of hydrogen molecules causes the sensing material to become saturated very quickly. In other words, the number of oxygen ion species adsorbed on the surface of the sensing material in advance is not sufficient to detect more H₂. The DL of the sensor tested in this section was also calculated according to Equation 4. If the relationship between the response and concentration is fitted linearly over a range of hydrogen concentrations from 0.1 to 0.6%, the slope is 79 in Figure 7a and the DL is 104 ppm. However, if the response–gas concentration relationship is fitted linearly over the 0.1–0.4% concentration range, the slope is 108 and the DL is 77 ppm, considering that the response of the sensor saturates when the hydrogen concentration reaches 0.5%. It is suggested that the H₂ sensing performance of the sensor in N₂ will be enhanced if more oxygen ion species can be introduced onto the surface of the sensing material during device preparation.

4. Conclusions

In summary, FET-type H₂ sensors with Pt-In₂O₃ nanoparticles were prepared and investigated in this study. An increase in the Pt concentration in the sensing material can improve the H₂ sensitivity of the sensor at room temperature, which can be attributed to the catalytic effect of Pt on the dissociation of H₂ molecules into H atoms. The current drift of the FET-type sensor was significantly suppressed by using the pulse measurement method, which alternately pre-biases the gate and reads the drain current of the sensor. Furthermore, applying negative voltages during the pre-bias phase of the pulse test can accelerate the sensor’s recovery, while applying positive voltages achieves the opposite. The FET sensor exhibits different H₂ sensing characteristics when it is operated in the linear, subthreshold and saturation regions. Among them, the sensor in the subthreshold region has the highest H₂ sensitivity, signal-to-noise ratio and the lowest detection limit, so it can be used as the best operating state for FET sensors. In addition, the presence of O₂ severely weakens the sensitivity of the sensor. This is suggested to be due to the catalytic effect of Pt, which promotes reactions between O₂ and H₂. As a result, the amount of H₂ that is required to change the work function of the sensing material is reduced. Further studies on the influence of O₂ will be detailed in the future work.