Atomic-layer-deposited Al2O3 and HfO2 on InAlAs: A comparative study of interfacial and electrical characteristics
Wu Li-Fan1, 2, Zhang Yu-Ming1, Lv Hong-Liang1, †, , Zhang Yi-Men1
School of Microelectronics, Xidian University, Key Laboratory of Wide Band-Gap Semiconductor Materials and Devices of China, Xi’an 710071, China
School of Electronic Engineering, Xi’an University of Posts and Telecommunications, Xi’an 710121, China

 

† Corresponding author. E-mail: hllv@mail.xidian.edu.cn

Project supported by the National Basic Research Program of China (Grant No. 2010CB327505), the Advanced Research Foundation of China (Grant No. 914xxx803-051xxx111), the National Defense Advance Research Project, China (Grant No. 513xxxxx306), the National Natural Science Foundation of China (Grant No. 51302215), the Scientific Research Program Funded by Shaanxi Provincial Education Department, China (Grant No. 14JK1656), and the Science and Technology Project of Shaanxi Province, China (Grant No. 2016KRM029).

Abstract
Abstract

Al2O3 and HfO2 thin films are separately deposited on n-type InAlAs epitaxial layers by using atomic layer deposition (ALD). The interfacial properties are revealed by angle-resolved x-ray photoelectron spectroscopy (AR-XPS). It is demonstrated that the Al2O3 layer can reduce interfacial oxidation and trap charge formation. The gate leakage current densities are 1.37 × 10−6 A/cm2 and 3.22 × 10−6 A/cm2 at +1 V for the Al2O3/InAlAs and HfO2/InAlAs MOS capacitors respectively. Compared with the HfO2/InAlAs metal–oxide–semiconductor (MOS) capacitor, the Al2O3/InAlAs MOS capacitor exhibits good electrical properties in reducing gate leakage current, narrowing down the hysteresis loop, shrinking stretch-out of the CV characteristics, and significantly reducing the oxide trapped charge (Qot) value and the interface state density (Dit).

1. Introduction

The InGaAs-based high electron mobility transistor (HEMT) has been widely studied for the applications in high-speed, low-power and high-frequency circuits due to its low effective electron mass and high electron mobility.[1,2] Its current flow is controlled by a Schottky metal gate. However, the Schottky gate brings a high leakage current. Recently, the InGaAs-based metal–oxide–semiconductor field-effect transistors (MOSFETs) have received much attention because of their relatively low gate leakage currents and large voltage swings compared with those of the Schottky-gate devices.[3] Each of these MOSFET and HEMT architectures typically has a higher band gap barrier layer, such as InAlAs, as a surface material, on which high-κ dielectrics will be deposited. Compared with the InP barrier layer, InAlAs provides not only good lattice matching with InGaAs, but also a large conduction band offset with InGaAs, thereby improving the confinement of electrons in the channel. Using in-situ epitaxial growth, a high-quality interface between InGaAs and InAlAs can be obtained. However, the interface between the barrier layer InAlAs and high-k oxide still plays an important role in device operation, affecting the subthreshold swing and the threshold voltage.[4] So, it is necessary to investigate interfacial and electrical characteristics between the barrier layer InAlAs and high-k oxide.

SiO2 has been replaced by the high-k dielectric because of its high leakage current when scaling down the device.[5] Research focuses on high-k dielectrics, including Al2O3,[6] HfO2,[7] and so on. Among these high-k dielectrics, the HfO2 gate dielectric has a high dielectric constant (18–20). However, HfO2 has a smaller bandgap of 5.45 eV than other high-k dielectrics.[8] In comparison, Al2O3 has a larger band gap of 9 eV, stronger adhesion to many surfaces of material, and serves as a protection layer due to its stabilities against moisture. In recent studies, an HfO2/InAlAs gate stack has been reported in relation to the surface passivation chemistries of InAlAs.[913] However, interfacial and electrical characteristics of HfO2/InAlAs MOS have been rarely reported.

In this work, we investigate the interfacial and electrical properties of the ALD-Al2O3 and ALD-HfO2 gate dielectrics on the InAlAs epitaxial layer. Compared with that of the HfO2/InAlAs capacitor, the capacitance–voltage (CV) characteristic of the Al2O3/InAlAs capacitor is significantly improved, and in particular, the stretch-out of CV curve for the Al2O3/InAlAs capacitor is considerably shrunk. The interface study is done by angle-resolved x-ray photoelectron spectroscopy (AR-XPS) and the electrical characterizations based on the CV hysteresis, current density–voltage (JV), and charge-trapping behavior of Al2O3 gate stacks on InAlAs are performed to evaluate its potential as an alternative gate dielectric for the InAlAs MOS capacitor.

2. Experiments

In this work, a 1.5-μm n-type In0.5Al0.5As layer (Si doped to ∼ 1 × 1017 cm−3) was epitaxially grown on a semi-insulating n-type GaAs substrate. Prior to deposition, the samples were first rinsed by using acetone and deionized (DI) water, then soaked in HCl (37%) for 1 min for native oxide removal, and finally dipped in (NH4)2S (10%) for 15 min to remove elemental As and In, cleaned in deionized water for 3 min and dried in N2. Two types of ALD dielectrics, i.e., (I) Al2O3 (10 nm) and (II) HfO2 (10 nm) film (sample I: Al2O3/InAlAs MOS capacitor, and sample II: HfO2/InAlAs MOS capacitor), were used for fabricating the high-k/InAlAs MOS capacitors. The ALD-Al2O3 film was deposited at 245 °C by using tri(methyl)aluminum (TMA-Al) and H2O as the precursors. The ALD-HfO2 film was deposited at 245 °C including an alternating pulse of tetrakis-(ethyl-methyl-amino)-hafnium (TEMA-Hf) and H2O used as the precursors. Post deposition annealing (PDA) was carried out at 380 °C for 1 min under N2 ambient by rapid thermal annealing (RTA). Finally, an MOS capacitor structure was produced by sputtering Ti/Pt/Au (20 nm/20 nm/200 nm) to form the gate electrode with an area of 1 × 10−4 cm2.

All the CV hysteresis measurements were carried out at a high frequency of 1 MHz at room temperature (295 K). AR-XPS was performed by using a monochromated Al–Kα x-ray source (1486.6 eV). The C 1 s line with a bonding energy of 284.6 eV was used as a reference to eliminate the charge effect during the analysis. AR-XPS is a useful tool for profiling depth in a substrate or a hetero-structure. If necessary, the photoelectron take-off angle, θ (defined as the elevation angle according to the sample surface), was changed by tilting the sample to change the photoelectron escape depth, t, in accordance with

where λ is the inelastic mean free path of photoelectrons.

3. Results and discussion

To quantify the density of interfacial trapped charge in high-k oxides on III–V surfaces, the CV hysteresis measurements are executed at a high frequency (1 MHz) and room temperature (295 K), and the results are shown in Fig. 1.

Fig. 1. CV characteristics of samples (I) Al2O3/n-InAlAs and (II) HfO2/n-InAlAs, measured at a frequency of 1 MHz.

The upwards line represents the CV curve measured by the gate voltage swept from −2.0 V to 2.0 V for the Al2O3/InAlAs capacitor and from −2.0 V to 2.0 V for HfO2/InAlAs capacitor respectively, and the downwards line shows the CV curve contained by the gate voltage swept in the opposite direction. There are obvious accumulation and depletion regions for both MOS capacitors. The hysteresis is likely to be caused by charges trapped at the high-κ/InAlAs interface as well as in the high-κ layer. The CV curves for sample I show reducing stretch-out compared with those for sample II, which implies that the interfacial trapped charge densities for sample I are lower than those for sample II. In addition, as can be seen from Fig. 1, the difference in the flat-band voltage (ΔV) of CV hysteresis for sample I is 0.02 V, which is smaller than the ΔV of 0.16 V for sample II. The value of Qot (cm−2) can be calculated from the following formula through using the CV measurements:

where Qot is the oxide trapped charge density, ΔV is the difference in the flat-band voltage of CV hysteresis, q is the elementary charge, and Cox is the oxide capacitance. According to ΔV estimated from the CV hysteresis sweep, the oxide trapped charge densities are 8.7648×1011 cm−2 for sample II and 7.1980×1010 cm−2 for sample I, respectively, using an oxide capacitance (Cox) Cox = 0.3599 μF·cm−2 for the Al2O3 layer and Cox = 0.5478 μF/cm2 for the single HfO2 layer. Thus, sample I shows much smaller Qot. This is attributed to the good interface between the InAlAs material and Al2O3, which prevents interfacial oxide and trap charge from forming. The electrical characterizations of the MOS capacitors are summarized in Table 1 for the Al2O3 and HfO2 layers as the dielectrics, respectively.

Table 1.

Electrical properties of MOS capacitors with an area of 100 μm×100 μm for the Al2O3 and HfO2 layers.

.

In order to further study the interface quality of the two samples, the interface state density Dit is analyzed. Dit can be extracted from the CV curves by the Terman method [14] as follows:

where Cox is the insulator capacitance, Vs the surface potential of InAlAs, Vg the gate voltage, Cs the substrate capacitance, εs the relative dielectric constant of the substrate, Et the energy level of the interface trap state, Ei the intrinsic Fermi level, qVB the doping-induced Fermi level in the bulk InAlAs, q the unit charge, and N the doping concentration of the substrate. According to the experimental CV curves, we can extract the values of Cs from Eq. (3). Then the relationship between Vs and Vg can be extracted from Eq. (4). Fitting the experimental data, an approximate functional relationship between Vs and Vg can be developed. According to Eqs. (5)–(7), we can obtain the curves of Dit versus EtEi in Fig. 2.

Fig. 2. Curves of Dit versus EtEi for sample I: Al2O3/InAlAs and sample II: HfO2/InAlAs.

The midgap Dit for sample I is 2.5×1011 cm−2·eV−1, while the value of Dit for sample II is 3×1012 cm−2·eV−1. It is found that the interface trap density for sample I is lower than that for sample II, which means that the Al2O3 layer can effectively reduce the interface trap density, and thus improve the interfacial characteristics on the InAlAs layer. Moreover, Al2O3 has better lattice matching with the InAlAs epitaxial layer, which may contribute to the low dangling bonds, such as In–, As–, O–. The Al2O3 is known for its high-temperature stability against recrystallization. But HfO2 forms polycrystalline after annealing 600 °C from the amorphous film. Thus HfO2 is much weaker especially in the temperature stability. This indicates that a thermally stable and reliable Al2O3/InAlAs capacitor has been obtained.

To investigate the interfacial properties of the HfO2 and Al2O3 stack films on InAlAs, the chemical states are analyzed by AR-XPS, and the spectra are fitted to Gaussian–Lorentzian functions. Figure 3(a) shows the Al 2p XPS photoelectron spectra obtained from 380 °C annealing for sample I at different take-off angle θ. For sample I, the Al 2p spectrum shows the peak at 74.2 eV, which indicates the formation of Al2O3 at θ = 90°. It is shown that there is no obvious change in the peak of Al 2p at θ = 10° and 30°, which suggests that there is no detectable chemical reaction in the Al2O3 layer. This is also confirmed by the O1s XPS photoelectron spectra shown in Fig. 5, which shows that no other aluminum compounds are formed in the Al2O3 layer. The XPS probing on the Hf 4f of sample II is shown in Fig. 3(b). For sample II, the Hf 4f spectrum shows two components 4f5/2 at 18.18 eV and 4f7/2 at 16.48 eV with a spin–orbital splitting of 1.86 eV at θ = 10°, indicating the formation of HfO2.[15] Figure 3(b) shows that the peaks shift to higher binding energies of 18.41 eV and 16.76 eV in the case of θ=30° and 90° uniformly. The binding energies of Hf 4f7/2 and 4f5/2 in the HfAlO film are 16.76 eV and 18.41 eV respectively,[16] therefore it is concluded that a thin Hf–O–Al mixture layer is formed at the interface of samples II. This is nicely consistent with the changes in the O 1s spectra in Fig. 5.

Fig. 3. XPS spectra of (a) Al 2p in sample I and (b) Hf 4f in sample II at different take-off angle θ (10°, 30°, and 90°).

To further investigate the interfaces of sample I and sample II, XPS analyses are performed on elements O and As in the gate dielectric structures. Figure 4 gives the information about the As 3d core level spectra for HfO2 and Al2O3 on the n-InAlAs at the different values of take-off angle θ.

Fig. 4. As 3d photoelectron spectra of sample I and sample II at the different values of take-off angle θ (10°, 30°, and 90°).

Figure 5 shows the O 1s core level spectra for sample I and sample II at θ = 90°. In Fig. 4, neither As– nor As–O is detected at θ = 10°, and only a little As–O is detected at θ = 30° for sample I, which reveals that Al2O3 can be effective in lowering the diffusion of As for sample I. Meanwhile for sample II, a great amount of As2O3 is seen at θ = 10° and 30°, which suggests that As has great ability to diffuse in the HfO2 gate dielectric. Figure 5 reveals that a certain number of O dangling bonds are shown in the HfO2 sample, and it is the reason for forming interface trap density.[17,18]

Fig. 5. O1s photoelectron spectra at the take-off angle θ = 90° for sample (I) Al2O3/InAlAs capacitor, and sample (II) HfO2/InAlAs capacitor.

The gate leakage currents of sample I and sample II can be seen in Fig. 6. At a gate voltage of +1 V, the gate leakages of the Al2O3 and HfO2 samples are 1.37×10−6 A/cm2 and 3.22 × 10−6 A/cm2, respectively. The result indicates that sample I shows an appreciably lower gate leakage current than sample II. The gate leakage line is well fitted by the Fowler–Nordheim (FN) tunneling model (data not shown), thus the FN tunneling is the dominant current-conduction mechanism for sample I. The FN tunneling conduction demands a low density of oxide traps and a big band offset, which are required by the large bandgap and the high-quality oxide film of Al2O3. The decrease of leakage current density of sample I may be attributed to high-thermodynamic stability, high quality, robustness and large conduction band offset between Al2O3 and InAlAs. The Al2O3 is known for its high temperature stability against recrystallization and its large measured conduction band offset with many semiconductor materials; while the higher gate leakage current of sample II may be explained by the smaller band-gap of HfO2 and the change from amorphous HfO2 films to polycrystalline after 600 °C annealing. HfO2 is much weaker especially in the temperature stability. Moreover, the poor lattice-matching and interfacial quality between HfO2 and InAlAs assist the carriers in tunneling the gate dielectric to form the high leakage current.

Fig. 6. JV characteristics for sample I: Al2O3/InAlAs and sample II: HfO2/InAlAs.
4. Conclusions

In this work, the characteristics of the ALD-Al2O3/InAlAs and ALD-HfO2/InAlAs MOS capacitors are investigated by AR-XPS. It is found that the Al2O3 layer can prevent the As 3d from diffusing, and the interfacial oxidation and trap charge from forming. Moreover, the CV hysteresis measurements are carried out at a high frequency (1 MHz) and room temperature (295 K). Compared with the HfO2/InAlAs MOS capacitor, the Al2O3/InAlAs MOS capacitor exhibits good electrical properties in narrowing down the hysteresis loop, shrinking stretch-out of CV characteristics, and significantly reducing the oxide trapped charge density (Qot) and the interface state density (Dit). Additionally, the gate leakage current densities of 1.37 × 10−6 A/cm2 and 3.22 × 10−6 A/cm2 at +1 V for the Al2O3/InAlAs and HfO2/InAlAs capacitors respectively also prove that there are better electrical properties for the Al2O3/InAlAs capacitor. Therefore, the Al2O3/InAlAs MOS capacitor can make the device achieve good performances.

Reference
1Xuan YWu Y QYe P D 2008 IEEE Electron Dev. Lett. 29 294
2Schleeh JAlestig GHalonen JMalmros ANilsson BNilsson P AStarski J PWadefalk NZirath HGrahn J 2012 IEEE Electron Dev. Lett. 33 664
3Yanning SKiewra E WKoester S JRuiz NCallegari AFogel K ESadana D KFompeyrine JWebb D JLocquet J PSousa MGermann RShiu K TForrest S R 2007 IEEE Electron Dev. Lett. 28 473
4Zhao L FTan ZWang JXu J 2015 Chin. Phys. 24 018501
5Sang H PYu S KJimin CHyo J KMann CDae H KYoung C BHyoungsub KSang W CChung Y KJung H S 2013 Appl. Surf. Sci. 283 375
6Jun LYuri Y GScott MIan M PKarim CEamon CMaire PPaul K H 2013 J. Appl. Phys. 114 144105
7Cheng YZou J JWan MWang W LPeng X CFeng LDeng W JZhu Z F 2015 Chin. Phys. Lett. 32 58102
8Wang L KLiu R J YYang H YLi G XZhang Y TZhang B L 2015 Chin. Phys. 24 018102
9Kobayashi MChen P TSun YGoel NMajhi PGarner MTsai WPianetta PNishi Y 2008 Appl. Phys. Lett. 93 182103
10Liu ALiu G XZhu H HByoungchul SElvira FRodrigo MFukai S 2015 RSC Adv. 5 86606
11Zhang FLiu G XLiu AByoungchul SShan F K 2015 Ceram. Int. 41 13218
12Masaharu KGaurav TYun SNiti GMike GWilman TPiero PYoshio N 2010 Appl. Phys. Lett. 96 142906
13Brennan BGalatage R VThomas KPelucchi EHurley P KKim JHinkle C LVogel E MWallace R M 2013 J. Appl. Phys. 114 104103
14Terman L M 1962 Solid State Electron. 5 285
15Yamamoto KHayashi SKubota MNiwa M 2002 Appl. Phys. Lett. 81 2053
16Lan X XGong C JOu XCao Y QSun CChen YXu BXia Y DLi A DYin JLiu Z G 2015 Microelectron. Eng. 133 88
17Wei H HHe GChen X SCui J BZhang MChen H SSun Z Q 2014 J. Alloys Compd. 591 240
18Aguirre F SMilojevic MChoi K JKim H CHinkle C LVogel E MKim JYang TXuan YYe P DWallace R M 2008 Appl. Phys. Lett. 93 061907