Processing math: 100%
J. Semicond. > 2016, Volume 37 > Issue 2 > 026001

SEMICONDUCTOR TECHNOLOGY

Characterization of the effects of nitrogen and hydrogen passivation on SiO2/4H-SiC interface by low temperature conductance measurements

Yiyu Wang1, 2, Zhaoyang Peng1, Huajun Shen1, , Chengzhan Li2, Jia Wu2, Yachao Tang1, Yanli Zhao2, Ximing Chen2, Kean Liu2 and Xinyu Liu1

+ Author Affiliations

 Corresponding author: Huajun Shen, shenhuajun@ime.ac.cn

DOI: 10.1088/1674-4926/37/2/026001

PDF

Abstract: We investigate the effects of NO annealing and forming gas (FG) annealing on the electrical properties of a SiO2/SiC interface by low-temperature conductance measurements. With nitrogen passivation, the density of interface states (DIT) is significantly reduced in the entire energy range, and the shift of flatband voltage, Δ VFB, is effectively suppressed to less than 0.4 V. However, very fast states are observed after NO annealing and the response frequencies are higher than 1 MHz at room temperature. After additional FG annealing, the DIT and Δ VFB are further reduced. The values of the DIT decrease to less than 1011 cm-2 eV-1 for the energy range of EC - ET ≥0.4 eV. It is suggested that the fast states in shallow energy levels originated from the N atoms accumulating at the interface by NO annealing. Though FG annealing has a limited effect on these shallow traps, hydrogen can terminate the residual Si and C dangling bonds corresponding to traps at deep energy levels and improve the interface quality further. It is indicated that NO annealing in conjunction with FG annealing will be a better post-oxidation process method for high performance SiC MOSFETs.

Key words: SiO2/SiC interfaceNO annealingforming gas annealingdensity of interface states

Silicon carbide (SiC),which exhibits a wider band gap as well as a superior breakdown field and thermal conductivity over conventional Si,has gained considerable attention for a new generation of power devices[1, 2, 3]. However,the metal oxide semiconductor field effect transistors (MOSFETs) fabricated on 4H-SiC present low inversion channel mobility and have a gate oxide reliability problem due to the high density of interface traps at the SiO2/SiC interface[4, 5, 6]. The nitridation process that nitrogen is incorporated into the SiO2/SiC interface,is considered to play a major role in defect passivation[6, 7, 8]. More recently,new routes to improve channel performance have also been proposed,including alkali and alkaline earth elements passivation[9]. Although other processes such as wet oxidation[10],sodium-contaminated oxidation[11, 12],or POCl3 annealing[13, 14] also improve the mobility,nitridation is the most common process,taking reliability and stability into consideration. It lowers the DIT and increases the field effect mobility from 5 to 30 cm2/(Vs),which makes the commercialization of SiC MOSFETs a reality. However,the peak electron inversion-layer mobility at moderate doping levels is expected to be 200 cm2/(Vs) or higher,considering the SiC bulk electron mobility of 954 cm2/(Vs)[15, 16] }.Asthedevelopmentofthecharacterizationtechnique,theproblemsofthenitridationprocessaregraduallyrevealed.LargefrequencydispersionisobservedbylowtemperatureCV$ measurements[17]. It is indicated that very fast states still exist at the SiO2/SiC interface even after NO annealing. The traditional high-low method with a maximum frequency of 1 MHz will seriously underestimate the DIT[18]. These conclusions are further confirmed by Kimoto's group,and they find that only a part of the fast states can be detected by conductance measurements up to 100 MHz at room temperature[19]. They also propose a C-ψS method to accurately evaluate the DIT at the SiO2/SiC interface[20]. Recently,a low temperature conductance method,which is supposed to be more suitable for characterization of very fast interface states at the nitrided interface,has been used to study the effects of the fast interface states[21]. Yoshioka reports that the capture cross section of these fast states is uniquely larger than that of conventional interface states and the density of these fast states also increases with NO temperature.

Carbon related defects have been found at the SiO2/SiC interface[8, 22, 23]. Physical defects that are attributed to the high DIT are carbon clusters,C dangling bonds and a nonstoichiometric sub-oxide near the interface[24, 25, 26]. It is suggested that NO annealing can dissolve interfacial carbon clusters,which leads to a reduction of the size of these clusters and possibly generates carbon dangling bonds at the interface[8, 27, 28]. In addition,since the nitridation process is usually performed at a temperature as high as 1350 \du,sub-oxide generation may become non-negligible during the high temperature process[29, 30]. Hydrogen passivation treatment (forming gas annealing: FGA) is routinely applied for Si-based MOS devices. Theoretical studies show that H2 can passivate isolated C dangling bonds as well as Si related defects[31]. It has been shown that nitrogen plasma exposure promotes hydrogen passivation of electrical defects at the SiO2/SiC interface[32].

In this article,we propose a treatment that combines NO annealing and FGA treatment to improve the electrical properties of the SiO2/SiC interface further. The role of nitrogen and hydrogen passivation on the SiO2/4H-SiC interface is evaluated by low-temperature conductance measurements.

MOS capacitors were used to characterize the effects of the passivation treatments on the interface of SiO2/SiC. Figure 1 represents the structure of the SiC MOS capacitor and its fabrication process flow. The starting substrate was an n-type 4H-SiC (0001) wafer with a 12 μm epilayer (ND 4.4 × 1015 cm3). After standard RCA cleaning,45-nm-thick oxides were formed by dry oxidation at 1300 \du,followed by post-oxidation annealing at 1300 \du in N2. Film thickness was determined by UVISEL spectroscopic ellipsometry[33]. Nitridation was carried out in NO (sample label: NO) at 1300 \du for 120 min. During the nitridation,an extra oxide growth took place and the thickness of SiO2 was about 61 nm. FGA annealing was performed in forming gas (5% H2 + 95% N2) ambient (sample label: FGA). For the combined treatment (sample label: NO & FGA),the NO & FGA sample first underwent nitridation and was then annealed in forming gas. A control sample without NO annealing or FG annealing was also prepared (sample label: as-oxidized). Circular Al electrodes (gate) with a diameter of 200 μm were deposited on the sample surface by a lift-off process.

Figure  图 1.  4H-SiC MOS capacitor and its fabrication process flow. The post-oxidation treatments for each sample are also shown.
Fig.1.

The room temperature conductance measurements were measured at probe frequencies from 1 kHz to 1 MHz using a precision impedance analyzer (4284A,Agilent Technologies). For the NO sample and the NO & FGA sample,low temperature conductance measurements were also carried out at temperatures ranging from 150 to 293 K with a step of 30 K to characterize the fast states. Figure 2 shows the typical parallel-model capacitance (CP) and conductance (GP) for the MOS capacitors.

Figure  Fig2.  (Color online) (a) Typical capacitance–voltage and (b) conductance–voltage characteristics of the MOS capacitors measured at different frequencies (parallel mode).

Based on an equivalent circuit shown in Figure 3,the interface state conductance (GPIT) was extracted from the measured impedance. Before applying the conductance method,the values of the Z(ω) were determined for each frequency from the impedance and Cox measured in a strong accumulation region. Then,series resistance was removed from the measured impedance[34]. The interface state density is linked to GPIT by[35]

Figure  Fig3.  Equivalent circuit for a MOS capacitor, where Cox, CD, CIT, and GPIT are the oxide capacitance, the semiconductor capacitance, the interface-state capacitance, and the interface-state conductance, respectively.

where ω is the angular frequency,S is the area of the gate electrode,and e (1.602 × 1019 C) is the elementary electric charge. The interface state density (DIT),the time constant of the interface states (τ),and the standard deviation (σ) are determined using a simple method provided by Nicollian and Brews[35]. The energy level (Ec - ET) of interface states is calculated from

by taking into account the Fermi level of the SiC epilayer (Ec - 0.24 eV) at room temperature. Here ψs is the surface potential determined from a comparison of an ideal C-V curve and the measured high frequency C-V curve. For the low temperature measurements,the incomplete dopant ionization effect in 4H-SiC is also taken into consideration when calculating the Fermi level[36]. The stability of the flatband voltage was also evaluated by C-V measurements.

The C-V characteristics measured with various frequencies from 1 kHz to 1 MHz for the MOS capacitors are shown in Figure 4. Both the as-deposited sample and the FGA sample show large frequency dispersion,which indicates the formation of an interface with a high density of interface defects. As for the NO sample and the NO & FGA sample,the C-V curves agree well with each other. Figure 5 shows interface-state conductances (GPIT) at different gate voltages for these samples. A broad single peak is observed for the as-oxidized sample and FGA sample,which is consistent with C-V results. From the peaks,DIT values of the as-oxidized sample and FGA sample are determined at each gate voltage and shown in Figure 6(a) as a function of the energy level with reference to the conduction band edge of SiC. H2 has only a limited effect on the DIT,which is accorded with earlier reports[37]. Two reasons may account for this result. Firstly,H2 can passivate Si dangling bonds and isolated C dangling bonds,but these kinds of defects are not the dominant defects in the as-oxidized SiO2/SiC interface. Secondly,H2 has a large diffusion barrier (> 2.5 eV) in SiO2 and needs a relatively high annealing temperature[38]. Although FGA treatment has a small effect on the DIT,the values of time constant (τ) show a significant decrease compared with the as-oxidized sample,especially at a trap energy level close to the conduction band edge of SiC.

Figure  Fig4.  The parallel-mode capacitance measured at different frequencies for SiC-MOS capacitors underwent various fabrication conditions. The measurements are performed at room temperature.
Figure  Fig5.  Interface-state conductance against frequency for SiC-MOS capacitors under various fabrication conditions. The measurements are performed at room temperature.
Figure  Fig6.  (a) The densities of interface states and (b) time constant for the as-oxidized sample and FGA sample. The values were extracted from conductance measurements at room temperature.

As shown in Figure 5,the NO sample and NO & FGA sample do not show any peaks throughout the measurable frequency range because these peaks take a maximum frequency higher than 1 MHz. However,it does not mean that there is a low density of interface states in these samples. It has been reported that the density of conventional states existing at an as-oxidized interface decreases significantly after NO annealing and new defects with a very fast time constant are observed[19]. These fast interface states can respond to frequencies higher than 100 MHz at room temperature (RT),so it needs an ultra-high frequency measurement or other measurement approach. Since the response frequency of fast states is lowered with decreasing temperature,the DIT is evaluated by low temperature conductance measurements for the NO sample and NO & FGA sample.

Here the impedance-frequency characteristics are measured in a temperature range of T= 150 - 298 K and the peaks of GPIT are only found in a range of T 210 K. Figure 7 shows the parallel-mode capacitance measured at low temperatures (210 K/180 K/150 K) for the NO & FGA sample. The frequency dispersion increases obviously with reducing the measurement temperature. Since the response frequency of interface states is lowered with decreasing temperature,for T < 210 K,1 MHz is almost high enough for the GPIT peak to be clearly observed. Figure 8 shows the GPIT at different gate voltages for the NO & FGA sample. Distinct peaks are observed for both the samples. The DIT of all the samples is then summarized in Figure 9. The detectable energy level is as shallow as 0.1 eV by low temperature measurements. Even though the DIT decreases obviously after NO annealing,the interface states density is still more than 1012 cm2 eV1 at shallower than 0.2 eV. The values of DIT are reasonable compared with earlier reports[21]. As for the NO & FGA sample,the DIT is further reduced after additional FG annealing. Though FG annealing has a limited effect on the shallow traps,deeper in the energy bandgap (EC - ET 0.4 eV),the DIT is reduced to less than 1011 cm2 eV1.

Figure  Fig7.  The parallel-mode capacitance for the NO & FGA sample measured at low temperatures.
Figure  Fig8.  Interface-state conductance against frequency for the NO & FGA sample measured at low temperatures.
Figure  Fig9.  The density of interface-states for SiC-MOS capacitors under various fabrication conditions.

As shown in Figure 9,FGA treatment at 450 \du alone does not have any impact improvement of the electrical properties. It is well known that Si-O-C bonds and C-C clusters,as well as C dangling bonds are supposed to be the possible origins of interface states in SiC-MOS devices. However,the Nitridation process can dissolve carbon clusters and associated complex Si-O-C bonds,which leads to the formation of simple Si and C dangling bonds that can be readily terminated by H2. Thus,the density of interface states (EC - ET 0.2 eV) decreases dramatically after a treatment combining NO and forming gas annealing. This result is consistent with the theoretical study that carbon-related defects can explain most of the observed interface states but not the high density near the conduction band of 4H-SiC[39]. Figure 10 shows the interface-state conductance against temperature for the NO & FGA sample measured at 100 kHz. Multiple lines show the results measured at different gate voltages. The conventional interface-state conductance peak,which usually exists at the as-oxidized SiO2/SiC interface and responds at room temperature,is effectively suppressed. The fast interface-state conductance peak is larger than the conventional interface-state conductance peak and is located at lower temperature in T-sweep measurements at 100 kHz. As for these fast interface states,we can rule out the possibility that simple Si dangling bonds and C dangling bonds,which were easy to be terminated with H2,are the dominant component of these traps. Yoshioka reports that the density of fast states increases almost in proportion to the N concentration at the SiO2/SiC interface[21]. Distinct N peaks are also found for both the NO sample and the NO & FGA sample in XPS measurements as shown in Figure 11. N atoms accumulated at the interface by the process are considered as a possible origin of these shallow traps (EC - ET 0.2 eV).

Figure  Fig10.  Interface-state conductance against temperature for the NO & FGA sample measured at 100 kHz. Multiple lines show the results measured at different gate voltages.
Figure  Fig11.  XPS spectra of N 1s for the SiO2/SiC structures under various fabrication conditions.

The stability of VFB against the accumulation voltage is investigated at room temperature. All the C-V sweeps (1 MHz) are started from -5 V,and the maximum accumulation voltage (Vacc) is varied from 10 to 24 V with a 2-V step. The flatband voltage shifts corresponding to these C-V curves are summarized in Figure 12. Significant positive VFB shifts are observed for the as-deposited sample and FGA sample. Positive shifts occur when the acceptor-like deep interface states are filled with electrons under a strong accumulation region. Most of them are difficult to detrap and act like negative fixed charges. Thus,a positive shift of the C-V curves is observed. The VFB shift almost disappears for the NO & FGA sample. This is because the DIT of the NO & FGA sample decreases by over an order of magnitude compared to that of the as-oxidized sample and especially at the deep energy level. It is indicated that the combined treatment of NO annealing and forming gas annealing is quite effective to suppress the VFB shift for the SiC-MOS devices.

Figure  Fig12.  Flatband voltage shift extracted from C–V characteristics as a function of the maximum accumulation voltage of C–V sweep.

We investigate the role of nitrogen and hydrogen incorporation into the SiO2/SiC interface by conductance measurements. Even though the DIT is significantly reduced in the entire energy range after NO annealing,very fast states are observed at the nitride SiO2/SiC interface and respond to frequencies higher than 1 MHz at room temperature. Through low-temperature conductance measurements,the conductance peak of the NO-annealed capacitor is clearly observed. Forming gas annealing alone is not effective for reducing interface defects,but hydrogen can terminate the residual Si and C dangling bonds generated after NO annealing and the DIT is further reduced. The combined treatment of NO and forming gas annealing reduces the density of interface states in the deep energy trap level to less than 1011 cm2 eV1. The stability of VFB is also remarkably improved due to the reduction of DIT. It is indicated that combined treatment using the nitridation process in conjunction with hydrogen passivation is a promising method for improving the performance of SiC-based MOS devices.



[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
图 1.  4H-SiC MOS capacitor and its fabrication process flow. The post-oxidation treatments for each sample are also shown.

Fig.1.


Fig2.  (Color online) (a) Typical capacitance–voltage and (b) conductance–voltage characteristics of the MOS capacitors measured at different frequencies (parallel mode).

Fig3.  Equivalent circuit for a MOS capacitor, where Cox, CD, CIT, and GPIT are the oxide capacitance, the semiconductor capacitance, the interface-state capacitance, and the interface-state conductance, respectively.

Fig4.  The parallel-mode capacitance measured at different frequencies for SiC-MOS capacitors underwent various fabrication conditions. The measurements are performed at room temperature.

Fig5.  Interface-state conductance against frequency for SiC-MOS capacitors under various fabrication conditions. The measurements are performed at room temperature.

Fig6.  (a) The densities of interface states and (b) time constant for the as-oxidized sample and FGA sample. The values were extracted from conductance measurements at room temperature.

Fig7.  The parallel-mode capacitance for the NO & FGA sample measured at low temperatures.

Fig8.  Interface-state conductance against frequency for the NO & FGA sample measured at low temperatures.

Fig9.  The density of interface-states for SiC-MOS capacitors under various fabrication conditions.

Fig10.  Interface-state conductance against temperature for the NO & FGA sample measured at 100 kHz. Multiple lines show the results measured at different gate voltages.

Fig11.  XPS spectra of N 1s for the SiO2/SiC structures under various fabrication conditions.

Fig12.  Flatband voltage shift extracted from C–V characteristics as a function of the maximum accumulation voltage of C–V sweep.

[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
  • Search

    Advanced Search >>

    GET CITATION

    shu

    Export: BibTex EndNote

    Article Metrics

    Article views: 2914 Times PDF downloads: 40 Times Cited by: 0 Times

    History

    Received: 22 May 2015 Revised: Online: Published: 01 February 2016

    Catalog

      Email This Article

      User name:
      Email:*请输入正确邮箱
      Code:*验证码错误
      Yiyu Wang, Zhaoyang Peng, Huajun Shen, Chengzhan Li, Jia Wu, Yachao Tang, Yanli Zhao, Ximing Chen, Kean Liu, Xinyu Liu. Characterization of the effects of nitrogen and hydrogen passivation on SiO2/4H-SiC interface by low temperature conductance measurements[J]. Journal of Semiconductors, 2016, 37(2): 026001. doi: 10.1088/1674-4926/37/2/026001 ****Y Y Wang, Z Y Peng, H J Shen, C Z Li, J Wu, Y C Tang, Y L Zhao, X M Chen, K A Liu, X Y Liu. Characterization of the effects of nitrogen and hydrogen passivation on SiO2/4H-SiC interface by low temperature conductance measurements[J]. J. Semicond., 2016, 37(2): 026001. doi: 10.1088/1674-4926/37/2/026001.
      Citation:
      Yiyu Wang, Zhaoyang Peng, Huajun Shen, Chengzhan Li, Jia Wu, Yachao Tang, Yanli Zhao, Ximing Chen, Kean Liu, Xinyu Liu. Characterization of the effects of nitrogen and hydrogen passivation on SiO2/4H-SiC interface by low temperature conductance measurements[J]. Journal of Semiconductors, 2016, 37(2): 026001. doi: 10.1088/1674-4926/37/2/026001 ****
      Y Y Wang, Z Y Peng, H J Shen, C Z Li, J Wu, Y C Tang, Y L Zhao, X M Chen, K A Liu, X Y Liu. Characterization of the effects of nitrogen and hydrogen passivation on SiO2/4H-SiC interface by low temperature conductance measurements[J]. J. Semicond., 2016, 37(2): 026001. doi: 10.1088/1674-4926/37/2/026001.

      Characterization of the effects of nitrogen and hydrogen passivation on SiO2/4H-SiC interface by low temperature conductance measurements

      DOI: 10.1088/1674-4926/37/2/026001
      Funds:

      Project supported by the National Natural Science Foundation of China (Nos. 61106080, 61275042) and the National Science and Technology Major Project of the Ministry of Science and Technology of China (No. 2013ZX02305).

      More Information
      • Corresponding author: shenhuajun@ime.ac.cn
      • Received Date: 2015-05-22
      • Accepted Date: 2015-09-01
      • Published Date: 2016-01-25

      Catalog

        /

        DownLoad:  Full-Size Img  PowerPoint
        Return
        Return