For GaN Schottky diode-type hydrogen sensors, it was reported that an oxidic intermediate layer between the catalytic Schottky contact and the GaN surface is the origin of the hydrogen sensitivity of Pd–GaN Schottky diodes, implying that the metal /semiconductor interfacial modification would lead to significant change in the interaction of hydrogen with devices [15]. Several authors have reported the critical role of the dielectric layer between the Schottky metal and the GaN surface in hydrogen sensitivity of devices [16–22]. In this report, firstly, hydrogen sensing characteristics of Pt–GaN Schottky diodes fabricated on intentionally defect-introduced GaN surfaces are investigated. Secondly, influence of dielectric layers, including SiO₂, Si_xN_y and AlGaN layers, between the metal and semiconductor interface on hydrogen detection was studied. Figure 1 shows the I–V characteristics for a Pt–(Si-doped) GaN Schottky diode in a N₂ ambient and the response to the exposure to 1% H₂ in N₂ at room temperature.

The diode exhibits a good rectifying behavior in both ambients. The I–V characteristics consist of the exponential region in the low-current region and the resistance-limited region in the high-current region. On the basis of the thermionic emission model for V > 3kT/q, the I–V characteristics of Schottky diodes are described using: (1) I F = SA * * T 2     exp ( − q Φ B kT ) { exp ( qV nkT ) − 1 }where I_F is the forward current, S is the diode area, A** is Richardson’s constant for n-GaN (24 Acm⁻² K⁻²), T is the absolute temperature, k is Boltzmann’s constant, q is the electron charge, and Φ_B is the Schottky barrier height [23]. The ideality factor n and Schottky barrier height Φ_B are extracted from Equation (1). Table 1 shows a summary of the extracted n and Φ_B in both N₂ and 1% H₂ in N₂ from the I–V characteristics shown in Figure 1.

The obtained n and Φ_B in N₂ are consistent with those previously reported [24]. Upon exposure of the device to 1% H₂ in N₂ at room temperature, n and Φ_B were found to be 1.22 and 0.662 eV, respectively. The reduction in the Schottky barrier height may be explained by the formation of a hydrogen-induced dipole layer between the metal and the semiconductor, which leads to the change in interfacial potential, as reported previously [6,7].

Next, hydrogen sensing characteristics of Pt–GaN Schottky diodes fabricated on as-grown and annealed GaN surfaces are studied in order to investigate the correlation between the surface defects of GaN and hydrogen sensing performance. The GaN samples were annealed before contact formation in the RTA system. Figure 2 shows the sensitivity of Pt–(Si-doped) GaN Schottky diodes for 1% H₂ in N₂ at room temperature as a function of the annealing temperature of the GaN surfaces. Here, the sensitivity is defined as a shift of voltage upon exposure to 1% H₂ in N₂ at a current of 0.1 mA. The error bars represent the range of the measured values.

As shown in Figure 2, the averaged value of the sensitivity of devices fabricated on as-grown surfaces, 900 °C and 1,000 °C annealed surfaces were 0.34, 0.35 and 0.34 eV, respectively. Significant differences of the sensitivity among the devices are not observed in Figure 2. In order to investigate the effects of carrier concentration of the films on hydrogen detection characteristics, undoped GaN films were employed. Figure 3 shows the sensitivity of Pt–(undoped) GaN Schottky diodes for 1% H₂ in N₂ at room temperature as a function of the annealing temperature of the GaN surfaces.

Here, the sensitivity is defined as a shift of voltage upon exposure to 1% H₂ in N₂ at a current of 0.1 nA. The error bars represent the range of the measured values. Although the devices fabricated on annealed surfaces displayed a slightly increased sensitivity, taking the error bars of the sensitivity into consideration, significant differences of the sensitivity among the devices are not observed, once again. On the other hand, the inset of Figure 3 shows the room temperature photoluminescence spectra of as-grown and 900 °C annealed GaN surfaces. The spectra have near-bandedge emission around 3.4 eV. The broad deep level emission around 2.2 eV is modulated by the Fabry-Perot interference fringes. As shown in the inset of Figure 3, 900 °C annealed GaN sample shows stronger deep level emission around 2.2 eV than as-grown sample does. This deep level emission is generally attributed to point defects and antisites, suggesting that high temperature annealing created these defects in the sample surface [25]. Though the PL spectra reveal the clear difference of the surface defects between as-grown and annealed surfaces as shown in the inset of Figure 3, the devices fabricated on annealed surfaces show the similar in hydrogen detection sensitivity to those on as-grown surfaces as shown in Figures 2 and 3. These results suggest that the surface defects of GaN do not play a critical role in hydrogen sensing characteristics of Pt–GaN Schottky diodes.

In order to investigate the influence of dielectric layers, including SiO₂, Si_xN_y and AlGaN layers, between the metal and semiconductor interface, on hydrogen detection sensitivity, hydrogen response of Pt–GaN diodes with dielectric layers was studied. Figure 4 shows the schematic cross sections of the MIS Pt–GaN diodes with both 10 nm SiO₂ and 10 nm Si_xN_y dielectrics, respectively.

Figures 5(a,b) show the I–V characteristics of MIS Pt–GaN diodes with both 10 nm SiO₂ and 10 nm Si_xN_y dielectrics in N₂ and response to exposure to 1% H₂ in N₂ at room temperature, respectively. As shown in Figure 5(a), hydrogen exposure of a MIS Pt–GaN diode with a SiO₂ dielectric causes marked changes in I–V characteristics as compared with that of the conventional Schottky diode, as shown in Figure 1, demonstrating that the MIS devices with a SiO₂ dielectric have a higher sensitivity to hydrogen than the conventional Schottky devices. Table 2 shows the sensitivities of the conventional Pt–GaN Schottky diode and MIS Pt–GaN diodes in 1% H₂ in N₂ extracted from Figures 1 and 5. Here, the sensitivities are defined as shifts in voltage upon exposure to 1% H₂ in N₂ at a current of 0.1 mA.

As shown in Table 2, the MIS Pt–GaN diode with a 10 nm SiO₂ dielectric shows a marked improvement in hydrogen detection sensitivity, which is twice higher than that of the conventional Pt–GaN Schottky diode. Upon changing the hydrogen-containing ambient into N₂, the I–V characteristics of the MIS Pt–GaN diode are found to revert to the initial values, although it takes a long time, that is, more than three hours at room temperature, just like the conventional Pt–GaN Schottky diode. In sharp contrast, a MIS Pt–GaN diode with a 10 nm Si_xN_y dielectric does not show any hydrogen response, as shown in Figure 5(b). Therefore, the hydrogen detection sensitivity of the device is found to be 0.00 V, as shown in Table 2. More detailed electrical properties of these MIS Pt–GaN diodes upon exposure to hydrogen are discussed in Section 3.2.

Hydrogen sensing characteristics of Pt–AlGaN/GaN diodes are generally reported like those of Pt–GaN diodes. In Pt–AlGaN/GaN diodes, the AlGaN/GaN hetero-interface structurally plays a major role in carrier transportation in an SBD. Moreover, in an AlGaN/GaN hetero-structure, the polarization-induced 2DEG concentration at the AlGaN/GaN interface is extremely sensitive to surface states. Any potential changes on the surface by adsorption of gas or liquid polar molecules, or pressure change would affect the surface potential and modulate the 2DEG density, leading to application in excellent sensors. Since the AlGaN layer is undoped and considered to be an insulator, the AlGaN/GaN structure would be regarded as one of MIS structures. Previously, only a few studies revealed the investigation of hydrogen detection sensitivity of Pt–GaN and Pt–AlGaN/GaN Schottky diodes [26], and little is known about the comparative performances. In this report, hydrogen detection sensitivity of Pt–GaN and Pt–AlGaN/GaN Schottky diodes is compared, and the results are shown in Figure 6.

As shown in Figure 6, the sensitivities of the Pt–GaN and Pt–AlGaN/GaN diodes are 0.37 V and 0.44 V, respectively. Here, the sensitivities are defined as shifts in voltage upon exposure to 1% H₂ in N₂ at a current of 0.05 A. The Pt–AlGaN/GaN diode shows slightly higher sensitivity as compared with that of the Pt–GaN device, consistent with the previous report [26]. Note that the Pt–GaN diode show relatively higher leakage current in H₂ ambient, as shown in Figure 6(a). In sharp contrast, Pt–AlGaN/GaN diode shows no leakage current in both ambients. Therefore, Pt–AlGaN/GaN devices would be suitable for application to highly-sensitive hydrogen sensors operated at high temperature. The detailed hydrogen sensing mechanism of Pt–AlGaN diodes are discussed in Section 3.3.

In the previous section, I compared the hydrogen response of MIS Pt–GaN diodes with both 10 nm SiO₂ and 10 nm Si_xN_y dielectrics. As a result, MIS Pt–GaN diodes with a SiO₂ dielectric showed a marked improvement in hydrogen detection sensitivity. In sharp contrast, MIS Pt–GaN diodes with a Si_xN_y dielectric did not show any hydrogen response. Since the changes in the I–V characteristics of devices are totally different for the two dielectric layers, detailed investigation for these phenomena would lead to the elucidation of the interaction mechanism of hydrogen with semiconductor devices. Here, I–V and C–V characteristics are performed for MIS Pt–GaN diodes with both SiO₂ and Si_xN_y dielectrics, and the changes in the electrical properties are discussed.

The conduction mechanism for MIS Pt–GaN diodes with both SiO₂ and Si_xN_y dielectrics is investigated as follows. In the oxide field range of 0.3–0.9 MV/cm, for MIS Pt–GaN diodes with a SiO₂ dielectric, the measured I–V data in a N₂ ambient were fitted using the Fowler–Nordheim (FN) tunneling model, as shown in Figure 7(a).

The FN model is given by: (2) J = AE 2   exp ( − B E ) (3) A = q 3 m 8 π hm dielec Φ B (4) B = 8 π 2 m dielec ( q Φ B ) 3 / 2 3 qhwhere m_dielec is the effective electron mass in the dielectric, m is the free electron mass, q is the magnitude of electron charge, Φ_B is the barrier height, and E is the dielectric electric field [27]. Here, the value of the effective mass used was equal to m_dielec = 0.5 m [28]. Since the plot of log (J/E²) versus 1/E is linear, as shown in Figure 7(a), the oxide conduction is FN conduction. The barrier height extracted from the FN plot as shown in Figure 7(a) is 2.9 eV, and this value is found to be consistent with the previously reported values for the SiO₂ /GaN interface [28,29].

On the other hand, for MIS Pt–GaN diodes with a SiO₂ dielectric, the measured I–V data in a H₂ ambient were fitted using the Pool–Frenkel emission model, as shown in Figure 7(b). The Pool–Frenkel emission model is given by: (5) J ∝ E   exp ( − q ( ϕ B − qE / π ε ) kT )where E is the dielectric electric field, ε is the dielectric permittivity, ϕ_B is the trap level below the conduction band, T is the temperature, k is the Boltzmann’s constant [30]. The plot of log (J/E) versus E^1/2 is linear, as shown in Figure 7(b), and, therefore, the oxide conduction is Pool–Frenkel emission. It is found that the hydrogen ambient drastically changes the conduction mechanism of MIS Pt–GaN diodes with a SiO₂ dielectric, as shown in Figures 7(a,b). For MIS Pt–GaN diodes with a Si_xN_y dielectric, the measured I–V data in a N₂ or H₂ ambient were fitted using the Pool–Frenkel emission model, as shown in Figure 7(c). Figure 7(c) shows a linear relation, suggesting that the dielectric conduction is Pool–Frenkel emission. The Si_xN_y dielectric is well known for showing Pool–Frenkel emission [28,31,32], consistent with the result shown in Figure 7(c). The Pool–Frenkel emission is due to emission of trapped electrons into the conduction band. The supply of electrons from the traps is through thermal excitation. For trap states with Coulomb potentials, the expression is similar to that of the Schottky emission. The barrier height, however, is the depth of the trap potential well. The barrier reduction is larger than in the case of Schottky emission by a factor of two, since the barrier lowering is twice as large due to the immobility of the positive charge [33]. The principal defect in the band gap of Si_xN_y was reported to be the silicon dangling bond [31]. The hydrogen ambient does not change the conduction mechanism for MIS Pt–GaN diodes with a Si_xN_y dielectric, as shown in Figure 5(b).

Figures 8(a,b) show room-temperature C–V characteristics at 10 kHz of the MIS Pt–GaN diodes with both SiO₂ and Si_xN_y dielectrics in a N₂ ambient and the response to the exposure to 1% H₂ in N₂, respectively. Bias voltage was swept from accumulation to depletion.

Note that the capacitances in the accumulation region for MIS Pt–GaN diodes with a Si_xN_y dielectric is about twice as large as those for MIS Pt–GaN diodes with a SiO₂ dielectric, consistent with their dielectric constants that are 3.8–4.2 for SiO₂ and 7.5 for Si_xN_y, respectively. As shown in Figure 8(a), for MIS Pt–GaN diodes with a SiO₂ dielectric, the C–V curve in H₂ significantly shifts toward negative bias values. The capacitance at each value of bias is just shifted by the difference of a constant voltage so that the shape of the C–V curve is unaltered. These C–V curves are usually explained by hydrogen-induced dipole layer formed in the metal–dielectric interface [12]. That is, the atomic hydrogen formed on the Pt surface diffuses through the metal film and is trapped at the metal-insulator interface where a polarized layer is formed. The hydrogen response ΔV can be expressed as: (6) Δ V = n i μ εwhere n_i is the concentration of hydrogen atoms at the interface, μ is the effective dipole moment per trapped hydrogen atom, ε is the permittivity [34]. The hydrogen-induced shift in the C–V curves, ΔV, for MIS Pt–GaN diodes with a SiO₂ dielectric is 1.0 V, as shown in Figure 8(a). This value is much larger than the C–V curve shift of 0.3 V for Pt–AlGaN/GaN Schottky barrier diodes at the similar experimental condition, [as shown in Figure 12(a)], reflecting the drastic I–V changes in MIS Pt–GaN diodes with a SiO₂ dielectric. Based on the Equation (6), the values of n_i and μ should depend on the properties of interfaces. In other words, hydrogen could be a probe in order to investigate the interface properties of the devices. The C–V curves at a measurement frequency of 1 MHz display the similar value of shift upon exposure to hydrogen.

In sharp contrast, for MIS Pt–GaN diodes with a Si_xN_y dielectric, the C–V curve does not show any shifts upon exposure to hydrogen. Note that the capacitances in the accumulation region (the voltages ranging from 1 to 2 V) in H₂ show slight increase as compared with those in N₂. This is probably due to the change in the dielectric constant of Si_xN_y in the different ambients. The results shown in Figure 8(b) reveal that the hydrogen-induced dipole layer is not formed in the metal–dielectric interface for MIS Pt–GaN diodes with a Si_xN_y dielectric. These C–V characteristics are quite anomalous and have not been reported yet. There is a key point that emerges from the results shown in Figure 8. Previously, it is reported that a work function decrease in the Schottky metals, such as Pd and Pt, on exposure to hydrogen is the origin of the hydrogen sensitivity, that is, changes in the characteristics of devices [13,14]. If the proposed mechanism is true, the results shown in Figure 8 is not consistent, because no work function change is observed in Figure 8(b), despite using the same Schottky metal in the both devices. Note that the work function change should be reflected in the flatband voltage shifts in the C–V curves. Therefore, it is obvious that the interface between the metal and the semiconductor plays a critical role in the interaction of hydrogen with semiconductor devices.

Figure 9 shows room-temperature C–V characteristics of the MIS Pt–GaN diodes with a Si_xN_y dielectric at measurement frequencies of 10 kHz, 100 kHz and 1 MHz in a N₂ ambient and the response to the exposure to 1% H₂ in N₂.

As shown in Figure 9, the C–V curves for MIS Pt–GaN diodes with a Si_xN_y dielectric show frequency dispersion at measurement frequencies ranging from 10 kHz to 1 MHz. Although the C–V curves for MIS Pt–GaN diodes with a Si_xN_y dielectric show frequency dispersion at measurement frequencies ranging from 10 kHz to 1 MHz, all the C–V curves do not show any shifts upon the exposure to hydrogen regardless of measurement frequencies. Note that these C–V characteristics of the MIS Pt–GaN diodes with both SiO₂ and Si_xN_y dielectrics in H₂ shown in Figure 8 are found to revert to the initial values upon changing the hydrogen-containing ambient into N₂, just like the I–V characteristics.

Although the detailed reaction mechanism of these MIS Pt–GaN diodes to hydrogen is unknown at present, the plausible mechanisms are described as follows. First, the hydrogen-induced changes in the electrical properties may be related to the number of adsorption sites for hydrogen at the metal/dielectric/semiconductor interface of the devices [35]. Second, hydrogen could be contained in both the SiO₂ and Si_xN_y dielectrics during the sputtering deposition, and some of the hydrogen molecules trapped at the metal/dielectric interface may affect the hydrogen detection sensitivity. Third, the Si_xN_y is well known to have a high density of trap states and hence potentially suffer from charge trapping instabilities [32]. This property of the Si_xN_y may relate to the acquired I–V and C–V data. Otherwise, these obtained results may be explained only by a novel reaction mechanism. Therefore, it would be important to understand the reaction mechanism underlying these phenomena, possibly leading to chemical selectivity to hydrogen. Especially, more detailed investigation of the interface states in a Schottky contact could help to interpret the reaction mechanisms at the interfaces of the MIS Pt–GaN diodes to hydrogen.

Furthermore, Si_xN_y passivation is known to be effective in improving the reliability of AlGaN/GaN high-electron mobility transistors (HEMTs), and the stability of MIS Pt–GaN diodes with a Si_xN_y dielectric exposed to hydrogen, as reported here, may be related to the excellent passivation property of Si_xN_y [36].

In the previous studies, high-frequency C–V characterization was typically employed in order to elucidate the reaction mechanism. As a result, a shift of the C–V curves toward negative bias values upon hydrogen exposure was observed, concluding that the formation of a dipole layer reduces the effective Schottky barrier height. The formation of the dipole layer should increase the capacitance; however, no increase of the capacitance has been reported yet. We presumed that very low frequencies are required in order to observe the proposed dipole layer and to investigate the reaction mechanism. Here, we demonstrate a clear change in the C–V characteristics of Pt–AlGaN/GaN SBDs exposed to hydrogen using a low frequency [37]. Typical I–V characteristics of the Pt–AlGaN/GaN SBD in N₂ and response to exposure to 1% H₂ in N₂ at room temperature are shown in Figure 10.

From I–V measurements, good rectifier characteristics of the n-type Schottky diode were confirmed for SBD samples in both environments. The introduction of hydrogen clearly enhances the current under a given bias both in forward and reverse I–V characteristics, as reported previously [38,39]. The forward I–V characteristics consist of the exponential region in the low-current region and the resistance-limited region in the high-current region. The ideality factor n and Schottky barrier height Ö_B were calculated by applying the thermoionic emission theory to the forward I–V curves. The calculated values of n and Ö_B were 2.3 eV, 0.88 eV in N₂ and 1.9 eV, 0.81 eV in 1% H₂, respectively. These values are consistent with those previously reported [39]. Upon changing the hydrogen-containing ambient into N₂, the I–V characteristics of the Pt–AlGaN/GaN SBD are found to revert to the initial values. According to the previous literature, the reduction in the Schottky barrier height is explained as a result of the formation of a hydrogen-induced dipole layer between the metal and the semiconductor [38,39].

In order to investigate the detailed mechanism of the interaction between hydrogen and Pt–AlGaN/GaN SBDs, we have conducted low frequency G–V and C–V characterization of Pt–AlGaN/GaN SBDs exposed to hydrogen. The G–V and C–V characterization was conducted with an ac modulation level of 100 mV and frequencies ranging from 1 Hz to 1 kHz. Figures 11 and 12 show room-temperature G–V and C–V characteristics at various frequencies of the Pt–AlGaN/GaN SBD in N₂ and response to exposure to 1% H₂ in N₂, respectively.

The G–V curves in Figure 11 suggest the following three issues: First, the conductance goes through a peak and approaches zero on either side. Second, the maximum values of the conductance in H₂ is much higher than those in N₂, which is probably related to the enhanced current under a given bias in I–V characteristics upon the introduction of hydrogen, as shown in Figure 10. Third, the G–V curves show frequency dependence. The peak magnitude of the conductance increases with increasing frequency in both ambients.

In Figure 12(a), the C–V characteristics in N₂ can be roughly classified into two diode bias (V_G) regions: (i) −1.5 V ≤ V_G, (ii) V_G < −1.5 V. In region (i), the capacitance is almost flat due to only small changes induced by applying V_G, where the depletion layer width is calculated to be 22.2–23.3 nm, in reasonable agreement with the AlGaN layer thickness. The almost flat portion of the capacitance is interpreted as the dc depletion of 2DEG by negative dc voltages. In region (ii), the capacitance rapidly decreases with decreasing V_G down to −2.8 V which is the pinch-off voltage, V_TH. This region corresponds to the depletion of GaN after almost complete pinch-off of 2DEG. At a frequency of 1 kHz, the C–V curve in H₂ shifts toward negative bias values, as reported previously [38]. As the frequency decreases from 1 kHz to 1 Hz, the capacitance in H₂ dramatically increases and the oscillation of the capacitance is observed, as shown in Figure 12. These C–V characteristics are quite anomalous and have not been reported so far. In sharp contrast, the C–V curves in N₂ do not change significantly over all the frequencies ranging from 1 kHz to 1 Hz. Note that these C–V characteristics of the Pt–AlGaN/GaN SBD in H₂ shown in Figure 12 are found to revert to the initial values upon changing the hydrogen containing ambient into N₂, just like the I–V characteristics. There are several key points that emerge from the results shown in Figure 12. First, at a frequency of 1 kHz, the C–V curve in H₂ does not reflect any polarization because the capacitance should increase when polarization is formed in the dielectric material [40]. In Figure 12(a), the capacitance at each value of bias is just shifted by the difference of their pinch-off voltages so that the shape of the C–V curve is unaltered. These C–V curves could be explained by positive fixed charge in the Pt–AlGaN interface [41]. Second, at a frequency of 1 Hz, the capacitance in H₂ dramatically increases with decreasing V_G down to V_TH. This anomalous C–V curve would not be attributed to mobile ionic charge because mobile ionic charge can be detected in C–V characteristics even at much higher frequency like 1 MHz [41]. Third, the capacitance in region (ii), e.g., at a bias voltage of −2 V, at a frequency of 1 Hz in N₂ is slightly higher than that at a frequency of 1 kHz. This is probably due to the tiny amount of remnant hydrogen in the devices. Here, no thermal treatment was performed in order to expel the remnant hydrogen from the devices prior to the experiment. Fourth, generally speaking, interfacial polarization, which occurs when mobile charge carriers are impeded by a physical barrier that inhibits charge migration, could produce such an increased capacitance in H₂ at a low frequency [40]. Although little is known about interfacial polarization in the case of hydrogen interaction with semiconductor devices, hydrogen-related dipoles may have the function of interfacial polarization. Adsorbed hydrogen on the AlGaN surface is positively charged and fixed on the surface, forming a dipole layer. Sufficiently low ac oscillations switch the polarity of dipoles, leading to the significant increase in the capacitance. Fifth, as for the C–V curves at low frequencies, the capacitance in H₂ dramatically increases with decreasing V_G down to V_TH as compared with those in N₂ although the capacitance at zero bias has nearly the same value both in N₂ and H₂, suggesting that the alignment of electric dipoles by applied biases may be related. A similar phenomenon is observed in ferroelectric materials in which the domains can switch from one direction of spontaneous alignment to another when an electric filed is applied, giving rise to large changes in the polarization and dielectric constant [42]. Furthermore, the oscillation of the capacitance observed in the C–V curve at 1 Hz in H₂ may be explained by the alignment of hydrogen-related dipoles when the ac oscillation is applied, i.e., the ac oscillation tries to switch the polarity of dipoles, but some of them cannot follow the oscillation simultaneously, resulting in the capacitance oscillation as a function of the applied bias voltages. In addition, surface charges on AlGaN may influence the dipole formation and alter C–V curves, implying that low-frequency C–V measurements could be useful for monitoring the surface states of AlGaN.

I–V and C–V characteristics of the Pt–AlGaN/GaN SBD for much lower concentration of hydrogen, which was 100 ppm H₂ in N₂, were performed at room temperature. Figure 13 shows I–V characteristics of the device in N₂ and response to exposure to 100 ppm H₂ in N₂ at room temperature. Comparing Figure 13 with Figure 10, the current enhancement under a given bias is different, i.e., 1% H₂ enhances more current under a given bias than 100 ppm H₂ does both in forward and reverse I–V characteristics.

Figure 14 shows C–V characteristics at 1 kHz (a), 100 Hz (b), 10 Hz (c) and 1 Hz (d) of the Pt–AlGaN/GaN SBD in N₂ and response to exposure to 100 ppm H₂ in N₂ at room temperature, and Figure 15 shows time evolution of C–V characteristics in 100 ppm H₂ at 1 Hz.

Note that the C–V characteristics with 100 ppm H₂ result in similar curves to those with 1% H₂, as shown in Figures 12 and 14, but the saturation time for 100 ppm H₂ was much longer than that for 1% H₂. As shown in Figure 15, upon H₂ exposure, the capacitances increase till ∼60 min and saturate after 60 min upon H₂ exposure.

By sharp contrast, at 1% H₂, capacitances saturate instantaneously. In low H₂ concentrations like 100 ppm, it takes more than 60 min to saturate the C–V characteristics (and also I–V characteristics), as shown in Figure 15, suggesting that hydrogen-induced dipole formation needs longer time in lower concentration of a hydrogen ambient. But, using low-frequency C–V characterization leads to clear detection of 100 ppm hydrogen even at room temperature where it is hard to detect hydrogen by using conventional I–V characterization, suggesting that low-frequency C–V method would be effective in detecting very low concentration of hydrogen.