Leakage Current Conduction Mechanism of Au-Pt-Ti / HfO 2 -Al 2 O 3 / n-InAlAs Metal-Oxide-Semiconductor Capacitor under Reverse-Biased Condition

: Au-Pt-Ti / high- k / n-InAlAs metal-oxide-semiconductor (MOS) capacitors with HfO 2 -Al 2 O 3 laminated dielectric were fabricated. We found that a Schottky emission leakage mechanism dominates the low bias conditions and Fowler–Nordheim tunneling became the main leakage mechanism at high ﬁelds with reverse biased condition. The sample with HfO 2 (4 m) / Al 2 O 3 (8 nm) laminated dielectric shows a high barrier height φ B of 1.66 eV at 30 ◦ C which was extracted from the Schottky emission mechanism, and this can be explained by fewer In–O and As–O states on the interface, as detected by the X-ray photoelectron spectroscopy test. These e ﬀ ects result in HfO 2 (4 m) / Al 2 O 3 (8 nm) / n -InAlAs MOS-capacitors presenting a low leakage current density of below 1.8 × 10 − 7 A / cm 2 from − 3 to 0 V at 30 ◦ C. It is demonstrated that the HfO 2 / Al 2 O 3 laminated dielectric with a thicker Al 2 O 3 ﬁlm of 8 nm is an optimized design to be the high- k dielectric used in Au-Pt-Ti


Introduction
According to the requirements of high speed, low power dissipation, and low noise application for RF devices used in telecommunication and other modern integrated circuits, increasing interest is focused on new III-V compound devices of InAs/AlSb and InAlAs/InGaAs HEMTs (high-electron mobility transistors), as these devices possess high electron mobility and peak velocity in the channel [1][2][3][4][5]. However, due to the narrow band gap of channels of these HEMTs, the devices suffer from serious current leakage which is considered as the biggest issue of InAs/AlSb and InAlAs/InGaAs HEMTs [6,7]. In this kind of device, InAlAs is most frequently used as the protective layer on the above barrier and as the gate contact semiconductor [1,3], and some reports proposed to deposit a high-k dielectric film on InAlAs, together with the gate electrode, to become a metal-oxide-semiconductor (MOS) capacitor isolated gate structure, in order to effectively suppress the leakage current. HfO 2 that presents a high dielectric constant is a popular candidate as the high-k dielectric [8][9][10], however it does not match well with InAlAs and the poor lattice match would degrade its performance [9]; Al 2 O 3 is used frequently as the high-k dielectric as well [10,11], however its dielectric constant is not high enough, and that will lead to a lower EOT (effective oxide thickness) which is not beneficial for reducing device size. For improvement, the HfO 2 -Al 2 O 3 laminated dielectric layer is proposed, and in this new device structure, a compromised dielectric constant can be achieved and the leakage current can be effectively suppressed [12][13][14]. In our previous paper [13], the physical and electrical performance of the new Au-Pt-Ti/high-k/n-InAlAs MOS capacitors with HfO 2 -Al 2 O 3 laminated dielectric were studied in detail. However, its leakage current mechanism was not mentioned, and this scheme has been covered in few other papers. To better understand the generation reason of the leakage current of the Au-Pt-Ti/HfO 2 -Al 2 O 3 /n-InAlAs MOS capacitor, we study the leakage current mechanism of the new devices at different bias condition ranges in detail. As InAs/AlSb HEMTs and InAlAs/InGaAs HEMTs work under negative gate voltage bias conditions, we studied the reverse-bias leakage current mechanism in this paper as in the case of a real application.

Experiment
The device structure from bottom to top is a 350 µm semi-insulating InP substrate, a 200 nm InP buffer layer, a 500 nm Si-doped In 0.5 Al 0.5 As semiconductor layer with a doping concentration of 1 × 10 17 cm −3 [15], a 12 nm oxide layer with Al 2 O 3 -HfO 2 dielectrics, and a metal with structure of Ti (20 nm)/Pt (20 nm)/Au (200 nm). The detailed schematic layer structures of the prepared sample can be found in our previous published paper [14]. In order to identify the impact of the thickness of the Al 2 O 3 inserting layer, we manufactured two kinds of samples with an oxide layer of HfO 2 (4 nm)/Al 2 O 3 (8 nm) laminated dielectrics (marked as Sample #1), and HfO 2 (8 nm)/Al 2 O 3 (4 nm) laminated dielectrics (marked as Sample #2), respectively. The detailed fabrication process is listed in Table 1 [16][17][18][19][20]. Table 1.

Process Step
Process Description

Measurement and Discussion
The leakage current measurement of the Au-Pt-Ti/HfO 2 -Al 2 O 3 /InAlAs MOS-capacitors under reversed bias condition is shown in Figure 1. It is found that the leakage current density J achieves a significant low value below 10 −6 A/cm 2 in the bias voltage V g ranges from −3 to 0 V at a temperature of 30 • C for the both Sample #1 and #2. In particular, Sample #1 shows a much lower leakage current, below 1.8 × 10 −7 A/cm 2 from −3 to 0 V at 30 • C which is three times less than Sample #2 with the leakage current value below 6.5 × 10 −7 A/cm 2 . This demonstrates that the thickness ratio of Al 2 O 3 and HfO 2 films clearly impact the leakage current. We note that the electric field E of Al 2 O 3 should be higher than that of HfO 2 , since the dielectric constant of HfO 2 is higher than that of Al 2 O 3 . However, we simply formulate E by applying V/T ox (T ox is oxide thickness) in order to present an intuitive view of the leakage current changing trend. In addition, it is found that the leakage current density J increases as the temperature increases. This may be explained by the violent electron movements at high temperatures. To better understand the generation reason of the leakage current of the Au-Pt-Ti/HfO 2 -Al 2 O 3 /InAlAs MOS capacitor and the reason for the lower leakage current of the sample with the higher thickness ratio of Al 2 O 3 and HfO 2 films, we will study the leakage current mechanism of the devices at different bias condition ranges and different temperatures in detail.
Coatings 2019, 9, x FOR PEER REVIEW 3 of 8 of the leakage current changing trend. In addition, it is found that the leakage current density J increases as the temperature increases. This may be explained by the violent electron movements at high temperatures. To better understand the generation reason of the leakage current of the Au-Pt-Ti/HfO2-Al2O3/InAlAs MOS capacitor and the reason for the lower leakage current of the sample with the higher thickness ratio of Al2O3 and HfO2 films, we will study the leakage current mechanism of the devices at different bias condition ranges and different temperatures in detail. Schottky emissions often happen in low electrical fields. The energy band diagram for Schottky emissions under the reversed bias condition is shown in Figure 2 [21][22][23][24]. Electrons surmount the metal-HfO2 surface barrier first, and then surmount the barrier at HfO2-Al2O3, and finally fall into the conduction band of InAlAs to form the leakage current. It can be verified by linear fitting the curve of ln J versus Ei 1/2 with a straight line, where J is the leakage current density and Ei is the electric field intensity under reversed bias voltage [25,26]. Schottky emission is temperature dependent, and the slope of linear approximation on the curve of lnJ vs Ei 1/2 needs to be consistent at different temperatures, which is the typical feature of Schottky emission [11,25,26]. Therefore, we make the line fit on the measurement curve under different temperatures of 30/50/70 °C in order to identify it correctly by the same scope of the fitting straight line, as shown in Figure 3. According to the analysis, the Schottky emission occurs in the bias range of −1.1-0 V and −1.7-0 V for Sample #1 and Sample #2, respectively. As we presented before, Sample #1, with a higher thickness ratio of Al2O3 and HfO2 films, shows a reduced leakage current in the Schottky region. This can be explained by its higher barrier height ϕB. The barrier height ϕB,n can be extracted from the equation ϕB,n = kTln(A*T 2 /J0)/q according to the Schottky emission mechanism, where k is the Boltzmann constant, T is the temperature, q is the electron charge, A* is the effective Richardson constant as the value of 105600A/K 2 m 2 , and J0 is the saturation current density [15]. The extracted ϕB,n values are listed in Table 2. The ϕB,n we extracted was illustrated as in Figure 2. It presents the barrier at HfO2/Al2O3 where is the top of the barrier. It is worth noting that the ϕB,n we obtained from the Schottky mechanism should be lower than the ideal barrier height by a value of △ϕB as shown in Figure 2. This gap can be explained by the Schottky barrier lowering effect [15,16]. It was found that Sample #1 presents a higher ϕB,n value than Sample #2, indicating that increasing the thickness of the Al2O3 film can suppress the accessibility of carriers to climb over the barrier to form a leakage current in the case of a fixed total thickness of laminated dielectric. In addition, it is noted that ϕB increases as Schottky emissions often happen in low electrical fields. The energy band diagram for Schottky emissions under the reversed bias condition is shown in Figure 2 [21][22][23][24]. Electrons surmount the metal-HfO 2 surface barrier first, and then surmount the barrier at HfO 2 -Al 2 O 3 , and finally fall into the conduction band of InAlAs to form the leakage current. It can be verified by linear fitting the curve of ln J versus E i 1/2 with a straight line, where J is the leakage current density and E i is the electric field intensity under reversed bias voltage [25,26]. Schottky emission is temperature dependent, and the slope of linear approximation on the curve of lnJ vs E i 1/2 needs to be consistent at different temperatures, which is the typical feature of Schottky emission [11,25,26]. Therefore, we make the line fit on the measurement curve under different temperatures of 30/50/70 • C in order to identify it correctly by the same scope of the fitting straight line, as shown in Figure 3. According to the analysis, the Schottky emission occurs in the bias range of −1.1-0 V and −1.7-0 V for Sample #1 and Sample #2, respectively. As we presented before, Sample #1, with a higher thickness ratio of Al 2 O 3 and HfO 2 films, shows a reduced leakage current in the Schottky region. This can be explained by its higher barrier height φ B . The barrier height φ B,n can be extracted from the equation φ B,n = kTln(A*T 2 /J 0 )/q according to the Schottky emission mechanism, where k is the Boltzmann constant, T is the temperature, q is the electron charge, A* is the effective Richardson constant as the value of 105600A/K 2 m 2 , and J 0 is the saturation current density [15]. The extracted φ B,n values are listed in Table 2. The φ B,n we extracted was illustrated as in Figure 2. It presents the barrier at HfO 2 /Al 2 O 3 where is the top of the barrier. It is worth noting that the φ B,n we obtained from the Schottky mechanism should be lower than the ideal barrier height by a value of φ B as shown in Figure 2. This gap can be explained by the Schottky barrier lowering effect [15,16]. It was found that Sample #1 presents a higher φ B,n value than Sample #2, indicating that increasing the thickness of the Al 2 O 3 film can suppress the accessibility of carriers to climb over the barrier to form a leakage current in the case of a fixed total thickness of laminated dielectric. In addition, it is noted that φ B increases as the temperature is increased, while the leakage current is also increased as the temperature is increased (Shown in Figure 1). This is because electrons move more violently at high temperatures, making it easier for the electrons to cross through the even higher barrier between the oxide layer and the gate electrode, to form an increased gate leakage current. As a result, the samples show a clear, increased leakage current by one order of magnitude when the temperature was increased from 30 • C to 70 • C.
Coatings 2019, 9, x FOR PEER REVIEW 4 of 8 the temperature is increased, while the leakage current is also increased as the temperature is increased (Shown in Figure 1). This is because electrons move more violently at high temperatures, making it easier for the electrons to cross through the even higher barrier between the oxide layer and the gate electrode, to form an increased gate leakage current. As a result, the samples show a clear, increased leakage current by one order of magnitude when the temperature was increased from 30 °C to 70 °C.  As Schottky emission is strongly related to dielectric-semiconductor interfacial conditions, a 60 s etching process was applied to the dielectric surface in order to make the interface suitable for X-ray photoelectron spectroscopy (XPS). An Al sputtering ion beam was applied in the XPS setup with a beam current density of 1.067 × 10 -5 A/cm 2 under 12,000 V. The XPS spectra were calibrated against the C-C peak at 284.8 eV. Advantage software was used to analyze the XPS data, the binding  the temperature is increased, while the leakage current is also increased as the temperature is increased (Shown in Figure 1). This is because electrons move more violently at high temperatures, making it easier for the electrons to cross through the even higher barrier between the oxide layer and the gate electrode, to form an increased gate leakage current. As a result, the samples show a clear, increased leakage current by one order of magnitude when the temperature was increased from 30 °C to 70 °C.  As Schottky emission is strongly related to dielectric-semiconductor interfacial conditions, a 60 s etching process was applied to the dielectric surface in order to make the interface suitable for X-ray photoelectron spectroscopy (XPS). An Al sputtering ion beam was applied in the XPS setup with a beam current density of 1.067 × 10 -5 A/cm 2 under 12,000 V. The XPS spectra were calibrated against the C-C peak at 284.8 eV. Advantage software was used to analyze the XPS data, the binding  As Schottky emission is strongly related to dielectric-semiconductor interfacial conditions, a 60 s etching process was applied to the dielectric surface in order to make the interface suitable for X-ray photoelectron spectroscopy (XPS). An Al sputtering ion beam was applied in the XPS setup with a beam current density of 1.067 × 10 −5 A/cm 2 under 12,000 V. The XPS spectra were calibrated against the C-C peak at 284.8 eV. Advantage software was used to analyze the XPS data, the binding energy peak positions indicate the chemical environment of the elements and the peak areas show the elemental composition. The measurements of the O 1s peak are shown in Figure 4. Figure 4a presents the XPS measurement of the O 1s peak of the dielectric-semiconductor interface, and Figure 4b presents the comparison of the ratio of various oxide contents on the interface. It is found that a larger amount of Al 2 O 3 with O 1s peak at 532.25 eV is presented in Sample #1, compared with Sample #2, which leads to enhanced matching between the dielectric and semiconductor and benefits the interfacial quality. Sample #2 presents Al-O and In-O on the interface, which forms interface states to degrade the interface quality. In addition, Sample #2 shows higher HfAlO components on the interface as well. The HfAl-O chemical bond is not as stable as the Al-O chemical bond because of its lower affinity with the oxide atom, make it easy to generate oxygen vacancies and dangling bonds in the presence of impurities from the epitaxial process to form interface states, i.e., As-O and In-O states, leading to a higher concentration of interfacial states. Thus, the leakage current that was contributed by interfacial traps is higher for Sample #2 with thinner Al 3 O 2 at 4 nm. Therefore, increasing the ratio of Al 2 O 3 to HfO 2 is helpful to increase the interface quality, and to suppress the Schottky leakage current.  Figure 4. Figure 4a presents the XPS measurement of the O 1s peak of the dielectric-semiconductor interface, and Figure  4b presents the comparison of the ratio of various oxide contents on the interface. It is found that a larger amount of Al2O3 with O 1s peak at 532.25 eV is presented in Sample #1, compared with Sample #2, which leads to enhanced matching between the dielectric and semiconductor and benefits the interfacial quality. Sample #2 presents Al-O and In-O on the interface, which forms interface states to degrade the interface quality. In addition, Sample #2 shows higher HfAlO components on the interface as well. The HfAl-O chemical bond is not as stable as the Al-O chemical bond because of its lower affinity with the oxide atom, make it easy to generate oxygen vacancies and dangling bonds in the presence of impurities from the epitaxial process to form interface states, i.e., As-O and In-O states, leading to a higher concentration of interfacial states.
Thus, the leakage current that was contributed by interfacial traps is higher for Sample #2 with thinner Al3O2 at 4 nm. Therefore, increasing the ratio of Al2O3 to HfO2 is helpful to increase the interface quality, and to suppress the Schottky leakage current. When Vg is continually biased to the negative direction, the Fowler-Nordheim (F-N) tunneling, which can be verified by fitting the curves of ln(J/Ei 2 ) versus 1/Ei with a straight line [27][28][29][30][31], becomes the main leakage mechanism. According to the extraction in Figure 5, the F-N emission occurs in the range −3-−1.1 V and −3-−1.7 V, for Sample #1 and Sample #2, respectively. It indicates that when the electrical field intensity is large enough, the electrons obtain enough energy to tunnel the potential barrier to make F-N tunneling occur. It is noted that the slope of the F-N fitting straight line under different temperatures is not the same. This is because the F-N tunneling is not proportional to temperature [32], which is different from the Schottky emission mechanism. The effective barrier height can be extracted from the F-N fitting curve, and values are shown in Table 3. It is found that the barrier extracted for Sample #1 is higher than that of Sample #2 at 30 °C and 70 °C, resulting in Sample #1 showing a reduced leakage current in the F-N region at room temperature. In contrast, Sample #1 shows a lower barrier height, of 0.18 eV, than Sample #2, showing a barrier height of 0.26 eV with the temperature of 50 °C. This should make Sample #1 present a higher leakage current under the F-N region, which is consistent with the leakage current test result in Figure 1. When V g is continually biased to the negative direction, the Fowler-Nordheim (F-N) tunneling, which can be verified by fitting the curves of ln(J/E i 2 ) versus 1/E i with a straight line [27][28][29][30][31], becomes the main leakage mechanism. According to the extraction in Figure 5, the F-N emission occurs in the range −3-−1.1 V and −3-−1.7 V, for Sample #1 and Sample #2, respectively. It indicates that when the electrical field intensity is large enough, the electrons obtain enough energy to tunnel the potential barrier to make F-N tunneling occur. It is noted that the slope of the F-N fitting straight line under different temperatures is not the same. This is because the F-N tunneling is not proportional to temperature [32], which is different from the Schottky emission mechanism. The effective barrier height can be extracted from the F-N fitting curve, and values are shown in Table 3. It is found that the barrier extracted for Sample #1 is higher than that of Sample #2 at 30 • C and 70 • C, resulting in Sample #1 showing a reduced leakage current in the F-N region at room temperature. In contrast, Sample #1 shows a lower barrier height, of 0.18 eV, than Sample #2, showing a barrier height of 0.26 eV with the temperature of 50 • C. This should make Sample #1 present a higher leakage current under the F-N region, which is consistent with the leakage current test result in Figure 1.

Discussion
The leakage current conduction mechanism of an Au-Pt-Ti/HfO2-Al2O3/InAlAs MOS-capacitor is dependent on the bias voltage. The Schottky emission occurs at a very low negative bias, and F-N tunneling becomes a dominant conduction mechanism when the electrical field intensity is increased continually. Due to the larger barrier height and lower interface states, the sample with HfO2 (4 m)/Al2O3 (8 nm) laminated dielectric shows a very low leakage current density of below 1.8 × 10 -7 A/cm 2 from −3 to 0 V at 30 °C. In conclusion, the HfO2 (4 m)/Al2O3 (8 nm) laminated dielectric structure is an optimized design to be the high-k dielectric used in Au-Pt-Ti/HfO2-Al2O3/InAlAs MOS capacitors to suppress the leakage current.

Conflicts of Interest:
The authors declare no conflict of interest.

Discussion
The leakage current conduction mechanism of an Au-Pt-Ti/HfO 2 -Al 2 O 3 /InAlAs MOS-capacitor is dependent on the bias voltage. The Schottky emission occurs at a very low negative bias, and F-N tunneling becomes a dominant conduction mechanism when the electrical field intensity is increased continually. Due to the larger barrier height and lower interface states, the sample with HfO 2 (4 m)/Al 2 O 3 (8 nm) laminated dielectric shows a very low leakage current density of below 1.8 × 10 −7 A/cm 2 from −3 to 0 V at 30 • C. In conclusion, the HfO 2 (4 m)/Al 2 O 3 (8 nm) laminated dielectric structure is an optimized design to be the high-k dielectric used in Au-Pt-Ti/HfO 2 -Al 2 O 3 /InAlAs MOS capacitors to suppress the leakage current.