Reliability of Buried InGaAs Channel n-MOSFETs With an InP Barrier Layer and Al2O3 Dielectric Under Positive Bias Temperature Instability Stress

The positive bias temperature instability (PBTI) reliability of buried InGaAs channel n-MOSFETs with an InP barrier layer and Al2O3 gate dielectric under medium field (2.7 MV/cm) and high field (5.0 MV/cm) are investigated in this paper. The Al2O3/InP interface of the insertion of an InP barrier layer has fewer interface and border traps compared to that of the Al2O3/InGaAs interface. The subthreshold slope, transconductance, and shift of Vg are studied by using the direct-current Id-Vg measurements under the PBTI stress. The experimental results show that the degradation of positive ΔVg under the medium field stress is mainly caused by the acceptor trap, while the donor trap under the high field stress become dominant in the subthreshold region, which leads to the negative shift in Vg. The medium field stress-induced acceptor traps are attributed by the InP barrier layer in the subthreshold region, resulting that the low leakage current can be achieved in the buried InGaAs channel n-MOSFETs with an InP barrier layer compared to the surface InGaAs channel n-MOSFETs.


INTRODUCTION
InGaAs was considered for use as the n-type high-mobility channel material because it has higher electron mobility and smaller electron effective mass than that of silicon [1][2][3]. The complementary metal oxide semiconductor (CMOS) structure can be realized by integrating III-V n-MOSFETs and Ge p-MOSFETs on a Si CMOS platform [4][5][6]. However, one of the most critical problems that must be solved to realize III-V MOSFETs is the formation of a stable MOS interface with low trap density [7]. Compared with the SiO 2 /Si system, the III-V native oxides negatively affect fermi-level pinning and current drift [8][9][10]. The atomic layer deposited (ALD) Al 2 O 3 dielectric in surface InGaAs channel MOSFETs can achieve a thermally stable interface and large band offsets, as confirmed by the previous research on the dielectric layer of InGaAs MOSFETs [11][12][13]. However, Al 2 O 3 /InGaAs interface traps and border traps in the dielectric layer remain high, which reduces the effective channel mobility and results in reliability instability in InGaAs MOSFETs [14][15][16]. Based on the poor interface quality of InGaAs and Al 2 O 3 , the introduction of a barrier layer between the Al 2 O 3 dielectric and InGaAs channel considerably improves channel electron mobility, transconductance, and drive current [17][18][19]. Although the InGaAs channel and Al 2 O 3 dielectric are separated by the barrier layer, high interface traps and border traps considerably affect device reliability under bias temperature instability (BTI) stress [20][21][22]. To reduce interface defect density, the interface passivation techniques of N passivation treatment [23][24][25] and sulfur passivation treatment [26][27][28] have been investigated to improve the interface properties and reliability.
Bias temperature instability stress directly leads to the degradation of threshold voltage, subthreshold slope, and onstate current. The interface trap and border trap induced by bias temperature instability stress are also considered to be the causes of the degradation of III-V MOSFET performance. Li et al. investigated the surface InGaAs channel n-MOSFETs under positive bias temperature instability (PBTI) stress and recovery [29]. They explained that high defect density exists at the InGaAs and Al 2 O 3 interface, which includes both interface traps and border traps, and the PBTI stress induces mainly border traps. However, few studies have reported the buried channel InGaAs MOSFETs with a barrier layer under PBTI stress.
In this paper, we experimentally studied the mechanisms of the buried InGaAs channel n-MOSFETs with an InP barrier layer under PBTI stress and recovery. The interface and border traps are estimated in the Al 2 O 3 /InP and Al 2 O 3 /InGaAs interfaces. The degradation of I d -V g during the PBTI tests shows a shift in V g under a medium field (2.7 MV/cm), which is the opposite of that observed under a high field (5.0 MV/cm) in buried InGaAs channel n-MOSFETs. The effects of PBTI stress in the buried InGaAs channel n-MOSFETs with an Al 2 O 3 /InP interface were investigated by performing the subthreshold slope, transconductance, and V g shift. The specific border traps are quantified to analyze the reliability of the device under the PBTI stress.

EXPERIMENTAL Fabrication Process
The main structure of Si-based buried In 0.25 Ga 0.75 As channel n-MOSFETs used in this paper is illustrated in Figure 1. The layer structure was grown on InP substrate by metal-organic chemical vapor deposition (MOCVD) and consisted of a 20 nm In 0.52 Al 0.48 As buffer layer, a 2 nm In 0.6 Al 0.4 As doping layer with Be doping concentration of 3 × 10 18 cm −3 , a 5 nm In 0.52 Al 0.48 As barrier layer, a 5 nm In 0.25 Ga 0.75 As channel layer, a 3 nm InP barrier layer, and a 40 nm In 0.53 Ga 0.47 As cap layer with N-type doping concentration of 2 × 10 19 cm −3 . During the device fabrication process, benzocyclobutene (BCB) is used to bond the InP wafer to the Si wafer, and the two-step surface cleaning process was carried out. First, a 10% w/t HCl solution was applied for 1 min to remove the native oxide layer, and 20% w/t NH 4 OH solution was applied for 6 min. Second, 20% (NH 4 ) 2 S solution was applied to passivate the interface of the InP barrier layer for 15 min at room temperature [26,27]. Then, 8 nm of Al 2 O 3 (i.e., gate dielectric) was deposited by Beneq TFS-200 atomic layer deposition (ALD) system at the substrate temperature of 300 • C. A postdeposition anneal (PDA) was carried out at 400 • C for 30 s in N 2 atmosphere. Ti/Au gate metal was evaporated by an

Measurement Methods
DC current-voltage (I-V) characterization was achieved with an Agilent B1500A semiconductor device analyzer. In the I-V measurements [30], the drain voltage (V d ) was set to 50 mV, and the source and substrate were grounded. During the PBTI stress phase, two different gate voltages were selected during the PBTI stress phase, and the gate field strengths were calculated to be E = 2.7 and E = 5.0 MV/cm, respectively, based on the simulation, while V s =V d =V b = 0 V. All DC I d -V g tests were carried out at room temperature (300 K). The PBTI test contains a 500 s stress phase and a 500 s recovery phase, as shown in Figure 2. During the stress phase, PBTI stress is set to 2.7 MV/cm and to 5 MV/cm for Al 2 O 3 for a duration of 500 s. Before applying the stress, we first measured the initial I d -V g curve by using a fresh sample (Iline). After a 500-s PBTI stress, the S line was measured; the R lines were the I d -V g curves obtained from the sample during the 500 s recovery.

Interface Characteristics
The distribution curves of interface trap density (D it ) are extracted from the multifrequency (1 MHz to 1 KHz) C-V   [29] can be described by the C-V hysteresis curve shown in Figure 3B. The N bt distribution of the Al 2 O 3 /InP interface is less than that of Al 2 O 3 /InGaAs, which indicates that low border traps are achieved in the Al 2 O 3 /InP interface.

Direct-Current I d -V g Measurements
Unlike the Si MOSFET under positive bias temperature instability stress, the oxide traps of the InP/Al 2 O 3 are generated during the stress phase. An uninterrupted cycle test on the same device can determine whether the test stress contributes to the I d -V g curve drift. The I d -V g curves of the buried InGaAs n-MOSFETs are shown in Figure 4. Compared with a fresh line, there is no distinct shift in I d -V g curves for either the subthreshold or the on-state region, and there is no clear change in current after the first cycle test. In the second cycle measurement curve, the I d -V g curves still do not show any shift. The subthreshold slope (SS) and on-state current remain the same compared with the fresh line, which indicates that neither negative nor positive charges were created under the measuring stresses.
According to the simulation results by Varghese et al. [33], the recoverable donor traps impact negative V g in the subthreshold region. Acceptor traps are essential for inducing a positive I-V curve shift in both the subthreshold and on-state regions. Figures 5A,B show the DC I d -V g curves measured for the fresh device before stress (I line) and under PBTI stress (S line) (E = 2.7 and E = 5.0 MV/cm) after 500 s as well as the recovery (R line) after 500 s. Compared with the I, S, and R lines in the medium and high fields (2.7 MV/cm and 5.0 MV/cm), there are two cases in the subthreshold region and in the on-state region. (1) In a medium field (E=2.7 MV/cm), the V g shift V g at a constant drain current is positive both in the subthreshold and on-state regions, which indicates that negative charges were created after the PBTI stress. The I d -V g curve of the R line still demonstrates a negative shift in the on-state region compared with that of the S line. It is clear that acceptor traps, which are induced by the medium field strength stress, are recoverable. The stressinduced recoverable acceptor traps exist in the on-state region. However, the drain current of the R line coincides with that of S line in the subthreshold region, indicating the medium field strength stress-induced recoverable traps are not shown in the subthreshold region. (2) In a high field (E=5.0 MV/cm), the V g shift V g is negative in the subthreshold region, indicating donor traps are created after the PBTI stress of 5.0 MV/cm. The V g shift V g is positive in the on-state region, demonstrating acceptor traps are created after the PBTI stress. The two crossing points, which mean a balance between acceptor trap and donor trap, are founded in 500 s of PBTI and 500 s of recovery with initial curve. The results show that donor traps induce negative V g shift of I d -V g curves, larger shift with lower I d current level. The acceptor traps induce positive shift of I d -V g curves, larger shift with higher I d current level. The high field stress induced donor traps have a large density in distribution of energy gap, and their distribution extends to the conduction band, just opposite to the distribution trend of acceptor traps in the energy gap, which are consistent with that of the surface channel InGaAs n-MOSFETs [29,33]. By comparing the S line with initial curve under the high field strength stress, the shift of crossing point is found to be negative in the R line with initial curve, indicating the donor traps are almost recoverable. Compared with the S line, the R line demonstrates a negative shift in the on-state region, indicating there are fewer recoverable donor traps than recoverable acceptor traps in the on-state region.
In surface channel InGaAs n-MOSFETs [29], stress-induced donor traps produce a negative shift in threshold voltage under the 2.7 MV/cm stress. However, stress-induced acceptor traps produce a positive shift in threshold voltage under the 2.7 MV/cm stress in buried channel InGaAs n-MOSFETs. Meanwhile, the I d -V g curves of the S and R lines do not have an offset in the subthreshold region. The results indicate that no donor traps are induced under medium field for the buried channel device. The Al 2 O 3 /InP interface has a lower interface trap density than that of the Al 2 O 3 /InGaAs interface in the distribution of energy gap, especially the downtrend of D it distribution of the Al 2 O 3 /InP interface near the mid-gap, which is opposite of the D it distribution of the Al 2 O 3 /InGaAs interface near the mid-gap shown in Figure 3A. The effect of low defect density is not serious in the recovery curve of the buried channel InGaAs n-MOSFETs in the subthreshold region under medium field strength, so buried channel InGaAs n-MOSFETs with Al 2 O 3 /InP interface show as a completely different trend than that of surface channel InGaAs n-MOSFETs in the recovery phase.
Time-dependence of V g in the on-state and subthreshold regions is shown in Figures 6A,B respectively. The value of V g shifts to positive direction in the on-state region under medium and high field strengths. From 500 to 1,000 s of recovery, V g of the two recovery curves continues the downward trend in the on-state region. The result shows recoverable acceptor traps have been proved to be recoverable in the on-state region under the field stress. The degradation of negative V g under high field stress is clearly smaller than the degradation of positive V g under medium field stress, indicating donor traps are induced under high field stress in the subthreshold region. Because stress-induced donor traps are fully recovered in the subthreshold region, V g is totally recovered in the subthreshold region. The degradation of the subthreshold slope (SS) and transconductance (G m ) are reflected by the stress-induced border traps under medium and high fields, as shown in Figures 7A,B. The downtrends of S and G m are clear during the recovery process from 500 to 1,000 s, revealing that stress-induced recoverable traps were released in the recovery process. When high field stress-induced recoverable donor traps become dominant in the subthreshold region, the degradation of S and G m are more pronounced compared with that of medium field stress. This finding illustrates that the degradation of the subthreshold slope and the transconductance are mainly caused by the donor traps under the high field stress.

Extractions of Trap Energy Densities
According to the similar explanation given by Li et al. [29], the distribution curves of border traps were investigated under the stress from V g among the I, S, and R lines. Specifically, (1) donor and acceptor traps are induced at the end of the 500 s stress. (2) Stress-induced donor traps fully recover, while acceptor traps are partially recoverable and partially permanent at the end of 500-s recovery. Although stress-induced acceptor traps are dominant, donor traps may also exist. To distinguish between donor traps and acceptor traps, N AP ox (I d ) represents the density  of negatively charged permanent acceptor traps, and N DR ox (I d ) represents recoverable donor traps. N AP is the total trap, which represents the density difference of negatively charged acceptor traps or positively charged donor traps. N AR ox (I d )-N DR ox (I d ) is the density of negatively charged recoverable acceptor traps. These parameters can be obtained from: where C ox is the gate oxide capacitor per unit area and q is the electron charge. We obtain N AP ox (V g ), N DR ox (I d ), N AR ox (I d ), and N AR ox (V g )+ N DR ox (V g ) as functions of gate bias V g , as shown in Figures 8A,B. For buried channel InGaAs MOSFETs, the magnitude of total traps is averagely 1.5 × 10 12 cm −2 and 1.6 × 10 12 cm −2 under medium and high field strengths, respectively, indicating more traps are induced by high field stress than the medium field stress. The medium field stress-induced permanent acceptor trap is calculated to be 1.1 × 10 12 cm −2 , which is larger than recoverable acceptor trap with the average density of 3.8 × 10 11 cm −2 . By comparison of the surface channel InGaAs MOSFETs, donor trap is not induced by the medium field stress in the buried InGaAs channel MOSFETs. However, the recoverable acceptor trap and recoverable donor trap are generated by the high field stress with a common density of 8.7 × 10 11 cm −2 , and the permanent acceptor trap is averagely 7.7 × 10 11 cm −2 . The results indicate that recoverable trap is easily induced by high field stress. Meanwhile, the medium field stress-induced permanent acceptor trap is larger than the high field stress-induced permanent acceptor trap in the subthreshold region, indicating the permanent acceptor trap and recoverable acceptor trap have been neutralized by recoverable donor trap in the high field stress.
Compared to the surface channel InGaAs n-MOSFETs by considering the experimental results of Li et al. [29], the impacts of PBTI stress on buried InGaAs channel n-MOSFETs are summarized as follows: (1) The D it and N bt distribution of the Al 2 O 3 /InP interface is smaller than that of Al 2 O 3 /InGaAs interface through the sulfur passivation treatment, indicating the interface reliability of buried InGaAs channel n-MOSFETs are improved by the Al 2 O 3 /InP interface. (2) In the surface channel InGaAs n-MOSFETs with the Al 2 O 3 /InGaAs interface donor traps become the dominant under the medium field stress. In contrast, the medium field stress-induced the permanent acceptor trap and recoverable acceptor trap are contributed by the degradations of V g and S in the buried InGaAs channel n-MOSFETs with the Al 2 O 3 /InP interface, which indicates that the generation of acceptor trap are attributed by the insertion of the InP barrier layer. (3) Compared to the surface InGaAs channel n-MOSFETs under the medium field stress, the acceptor trap become dominant in the subthreshold region for the buried channel one. The buried channel InGaAs MOSFETs is better to maintain the low off-state current developing III-V MOSFETs technology for low-power application.

CONCLUSIONS
In summary, the degradation of the buried InGaAs channel n-MOSFETs with an InP barrier layer under PBTI stress and recovery were investigated. The Al 2 O 3 /InP interface helps achieve low interface and border traps compared to the Al 2 O 3 /InGaAs interface through the sulfur passivation treatment. Contrary to the shift direction of V g under the medium field stress in the surface InGaAs channel n-MOSFETs, the permanent acceptor trap of 1.1 × 10 12 cm −2 and recoverable acceptor trap of 3.8 × 10 11 cm −2 become the dominant to produce a positive shift in V g in the buried InGaAs channel n-MOSFETs. The high field stress-induced recoverable donor trap of 8.7 × 10 11 cm −2 cause degradation of S and G m in the subthreshold region, whereas the degradation of I d -V g is contributed by the recoverable acceptor trap and permanent acceptor trap in the on-state region. Compared to the surface InGaAs channel n-MOSFETs under medium field stress, the low leakage current can be achieved in the buried InGaAs channel n-MOSFETs with an InP barrier layer.

DATA AVAILABILITY STATEMENT
All datasets generated for this study are included in the article/supplementary material.

AUTHOR CONTRIBUTIONS
HLi was the leader of the work and responsible for the main of experiment and paper writing. KQ, XG, YLi, YC, ZZ, and LM were responsible for single step of the fabrication process. FZ and XZ were responsible for device testing. TF, XL, YLiu, TS, and HLiu were mainly engaged in picture editing and related data processing. TS and HLiu contributed to the modification and suggestion in this paper.