Charge Trapping Mechanisms in Nitridated SiO2/ 4H-SiC MOSFET Interfaces: Threshold Voltage Instability and Interface Chemistry

Silicon dioxide (SiO2) layers deposited on 4H-SiC and subjected to different post deposition annealing (PDA) in NO and N2O were studied to identify the key factors influencing the channel mobility and threshold voltage stability in lateral implanted 4H-SiC MOSFETs. Cyclic gate bias stress measurements allowed to separate the contributions of interface states (Nit) and near interface oxide traps (NIOTs) in the two oxides. The reduction of these traps in the NO annealed sample is due to the lower amounts of sub-stoichiometric silicon oxide (~1nm) and carbon-related defects (<1nm) at the interface, as could be demonstrated by Electron Energy Loss Spectroscopy. The experimental results indicate that limiting the SiC re-oxidation during post-deposition annealing in MOSFET technology is a key factor to improve the mobility and threshold voltage stability.


Introduction
The performances of 4H-SiC MOSFETs are strongly influenced by the processing of the SiO2/4H-SiC interface [1,2]. In particular, threshold voltage (Vth) instability phenomena [3,4] and poor field effect channel mobility (µFE) [5,6] often occur in these devices and can be partially mitigated by post oxidation annealings (POA) or post oxide deposition annealings (PDAs) [5]. Hence, a great attention of the scientific community is devoted to optimize the SiO2/4H-SiC interface by POAs or PDAs. However, while the introduction of Nitrogen or others atomic species can produce some benefits in the interfacial carrier transport, they can introduce additional trapping states, which are detrimental for the Vth stability [7]. Hence, the introduction of such undesired traps at the SiO2/4H-SiC interface should be limited.
On the other hand, it is mandatory to develop reliable, fast and accurate Vth determination methods, able to minimize the amount of undetected traps during the investigation [8].
In this paper, SiO2/4H-SiC interfaces on implanted p-type 4H-SiC, formed with deposited SiO2 subjected to different PDAs in NO or N2O, were investigated by cyclic gate stress measurements to determine the amount of trapped charges. Furthermore, nanoscale structural/chemical analyses allowed to identify the key factors influencing the channel mobility and Vth stability in lateral MOSFETs. In particular, a correlation between the chemical disorder at the deposited SiO2/4H-SiC interface after the PDAs and the electrical properties of the lateral MOSFETs is presented.

Experimental
In this paper, different lateral MOSFETs were fabricated on 4•-off-axis n-type (0001) 4H-SiC epitaxial layers (1 × 10 16 cm -3 ), with a p-type Al-implanted body region (NA ~ 10 17 cm -3 ). The gate oxide was a 40 nm thick deposited SiO2 layer [9]. Different PDAs were performed in a horizontal furnace in a sub-atmospheric pressure regime at 1150 °C in NO or N2O [10], in order to investigate the electrical and chemical impact on the SiO2/4H-SiC interface.
The MOSFETs were characterized by means of current voltage (ID-VG) transfer characteristics measured in a CASCADE Microtech probe station, using a Keysight B1505A parameter analyzer. The MOSFETs chemistry interface was investigated by Scanning Transmission Electron Microscopy (STEM) analyses (JEOL ARM200CF at a primary beam energy of 200 keV) and electron energy loss spectroscopy (EELS).

Results and Discussion
The standard procedure to determine the threshold voltage Vth in lateral MOSFETs uses the linear fit of ID 0.5 vs VG. However, this method requires the gate bias sweep and can be not sensitive to an eventual variation of the charge state in the insulator during the measurement. Hence, it is important to minimize the system perturbation induced by the Vth measurement itself. Recently, we have presented a method to measure the Vth variation by means of a single point ID measurement, avoiding relaxing effects on the majority of trapped charge [8]. The whole semiconductor bandgap trapping states were probed by varying the gate bias stress from inversion to accumulation and backward followed by a single point Vth measurement at VG = Vread = +8 V. Fig. 1a describes schematically the cyclic gate bias sequence used to probe the whole 4H-SiC bandgap and to probe the Vth variations. Fig. 1b shows how the ID measured at Vread = +8 V varied upon the cyclic gate bias stress is applied. The ID variation can be correlated to the Vth variation.
The Vth variation occurring from the minimum and maximum VG stress values can be associated to the total amount of electrons (Nit) that can be trapped at the interface traps [8]. On the other hand, closing the cyclic gate bias sequence it is possible to notice a residual Vth variation at VG= 0 V. This can be associated to the remaining amount of charge at the near interface oxide traps (NIOTs) [8].

Residual charged NIOTs
By repeating the experimental procedure on the samples under investigation, it is possible to estimate the total amount of traps at the interface Nit and in the oxide NIOTs. In particular, the amount of detected Nit decreases with the increasing temperature (from 6×10 11 cm -2 ), while the detected NIOTs are nearly constant in the investigated temperature range (about 1×10 11 cm -2 ) [8,11]. This behavior can be associated to a tunneling mechanism from the semiconductor into a trap located into the insulator [11][12][13]. On the other hand, it has been demonstrated that the Vth main variation is due to the Nit and their charge retention capability has an activation energy of EA=0.1 eV, which in turn can be associated to the intrinsic interface state defects of the SiO2/4H-SiC system [13].
The PDAs performed in NO or N2O resulted into a different MOSFET electrical behavior. For example, the maximum field effect mobility value after annealing in NO (55 cm 2 V -1 s -1 ) was more than twice larger than the value measured in the sample annealed in N2O (20 cm 2 V -1 s -1 ). Moreover, the different PDAs performed in NO or N2O produced different Vth variation under the cyclic bias stress. As described previously, from the cyclic bias stress it is possible to extract both the Nit and the NIOTs. Hence, the relative amount of the Nit and NIOTs can be varied by a fine tuning of the PDA. As an example a PDA in NO reduced both trapping mechanisms reducing the Vth instability (and increasing the field effect mobility). In particular, Fig. 2a shows the comparison between the Vth variation during the cyclic stress obtained on the samples after PDAs in NO and N2O respectively. As can be seen in the highlighted region of Fig. 2a, the PDA in NO demonstrates a reduction of both the Nit and NIOTs with respect of the PDA in N2O [14]. Evidently, the two SiO2/4H-SiC interfaces produced with the two different PDAs are different. Such difference has to be pursued at the intimate chemical nature of the interface. Fig. 2b shows the comparison of the EELS profile collected on both PDA SiO2/SiC interfaces, demonstrating that the improved Nit and NIOTs in the NO sample can be associated to a reduced amount of SiOx and C-based defects [14].   2b shows the carbon and SiOx EELS profiles collected on both N2O and NO samples. The PDAs signals are compared to those collected on the as deposited reference sample. The as deposited interface shows carbon and SiOx profiles that represent the steepest transition between the SiC substrate and the SiO2 insulator. As can be noticed, both PDAs produced a broader interface element transition compared to the as deposited sample. This progressive change in the SiOx and carbon profiles across the SiO2/4H-SiC interface are due to the slight re-oxidation induced by the PDA. These results allowed to draw the correlation between the electrical properties and the chemical order at the SiO2/4H-SiC interface. In particular, it is worth noticing that the smaller amount of interface traps in the sample annealed in NO can be associated to a thinner and steeper sub-stoichiometric SiOx layer with respect to the N2O case. Furthermore, both samples show similar carbon. In particular, the carbon profiles show a decreasing tail within the oxide. The carbon tails extend from the SiC interface into the oxide for about 1.2 nm and 0.9 nm for the sample annealed in N2O in NO, respectively. It has to be emphasized that in Fig. 2b only the carbon profile of the sample annealed in N2O is shown for the sake of clarity. These carbon-related defects within the insulator can be associated to the residual NIOTs detected by the cyclic gate bias stress procedure.

Conclusion
In this paper, a combined electrical/chemical study of deposited SiO2/4H-SiC interfaces on implanted p-type 4H-SiC, subjected to different PDAs in NO or N2O is presented. The lateral MOSFETs were investigated by cyclic gate stress measurements and nanoscale structural/chemical analyses. This basic study allowed to identify the key factors influencing the channel mobility and Vth stability in lateral MOSFETs. The results demonstrated that the threshold voltage instability of 4H-SiC MOSFETs, associated to different trapping mechanisms, is directly related to the SiO2/SiC interface chemistry and can be mitigated by an accurate control of the nitridation conditions of deposited oxides. In particular, interface traps can be associated to a sub-stoichiometric SiOx layer and near interface oxide traps can be associated to carbon tails extending from the SiC interface into the oxide due to the interface re-oxidation occurred during the PDAs.