Solid State Phenomena Vol. 359

Abstract: Free standing wafers of the cubic polytype of silicon carbide (3C-SiC) grown on micromachined silicon substrates can be a platform for new power electronic devices, provided that suitable device fabrication processes are understood and optimized. In this frame, p-type doping is still an open issue, as results on the electrical activation of ion implanted Al in 3C-SiC are limited. This work analyses high level p-type doping with post-implantation annealing carried out at temperatures in the range 1650-1850 °C with different durations. A coherent picture emerges, showing that the resulting resistivity in 3C-SiC Al-implanted layers is higher than the one obtained in 4H-SiC implanted layers, the result being ascribed to low carrier mobility and possibly presence of compensation centers, rather than to poor Al electrical activation. The reported results highlight the importance of working on material and processing optimization.

Abstract: Within this work, the effect of high dose Al ion implantation on 4H-SiC epitaxial layer is displayed. Through TEM investigation it is demonstrated that the implanted surface is suitable as seed for subsequent epitaxial regrowth generating a crystal free of extended defects. In order to assess the defects within the projected range of the ion implanted area, High Angle Annular Dark Field STEM (HAADF-STEM) analyses were performed demonstrating the atomic arrangement of the lattice in correspondence of the dislocation loop and the deviation of the crystallographic planes of 4H-SiC, driven by stress relaxation, that determine the staircase configuration of the implant pattern. Further emphasis is given to the detailed analysis of the precipitates atomic structure, whose preferential localization is ascertained. Using Energy-Dispersive X-ray spectroscopy (EDS) analysis, the precipitate is finally established as Al crystal with an FCC structure.

Abstract: The problem of crystal damage recovery and of impurity substitution in implanted semiconductors is considered from a statistical mechanical viewpoint. This is done by resorting to a thermodynamic pseudo-potential originally developed for cooperative structural rearrangements in disordered systems close to their glass transition. The dependence of the substitutional fraction φ on the post-implantation annealing temperature T_ann in Al/4H-SiC systems is discussed in the light of these ideas. After completion of the annealing process, an Arrhenius plot of φ(T_ann) shows a slope in the order of 1 eV or less, depending on the amount of lattice damage initially produced by the implantation. Slopes ∼4 eV are found after incomplete annealing, indicating that substitution occurs mainly in damaged crystal cells. These concepts are suggested to be used for optimization of the doping procedure by ion implantation.

Abstract: This paper demonstrates for the first time a new annealing scheme to form p-type junctions in SiC by high temperature ion implantation followed by laser annealing without the use of a protective carbon capping layer. This novel approach leverages higher substrate temperatures during implant to minimize implant-induced defects during ion implantation, which enables the use of reduced thermal budget laser annealing for dopant activation. Laser annealing enables higher surface temperatures in the implanted layer than conventional annealing using a high temperature furnace. The shorter thermal budget results in higher dopant activation while minimizing, the formation of extended defects observed during high thermal budget furnace annealing, which can lead to undesirable degradation in device performance. By using laser annealing with no carbon capping layer, the sheet resistance of the implanted layers is reduced up to 6 times with respect to the conventional process (using a furnace anneal and carbon capping layers).

Abstract: We investigated the electrical and structural effects of silicon (Si), yttrium (Y) and lanthanum (La) doping in 10-45 nm thick hafnium dioxide (HfO₂) films on silicon carbide (SiC) and Si substrates. We show that the introduction of Si dopants leads to a significant enhancement of the electric breakdown field and a reduction of the leakage current density by elevating the crystallization temperature. This effect becomes stronger with higher Si content. In contrast, Y and La doping does not raise T_C but increases the tetragonal and orthorhombic phase portion within the crystalline films and therefore enhances the dielectric constant k. Furthermore, we show that larger grains in crystalline films are associated with a higher leakage current density.

Abstract: Bulk mobility and dopant activation of implanted species into 4H-SiC plays a crucial role in the carrier conduction, blocking behavior, and channel properties of a 4H-SiC vertical power MOSFET. Nitrogen and phosphorus ion implantation became the norm as n-type dopants for 4H-SiC. Therefore, the doping and temperature behavior of both species in 4H-SiC needs to be well characterized. In this study, we report a comparison in electrical characteristics between nitrogen and phosphorus implanted 4H-SiC as a function of temperature for various doping levels. For this purpose, 4-point van der Pauw samples are prepared, resistivity and Hall measurements are conducted. We found that resistivities drop as temperature increases from 140 K with phosphorus having higher resistivities at all implanted doping concentrations. The carrier concentrations increase with increase of temperature, indicating incomplete ionization of dopants. Mobilities drop at low temperature due to increased impurity scattering, reaches a peak near 300 K and drops at higher temperature due to increased phonon scattering. From the obtained data, using a two-level charge neutrality equation, the activation percentage and ionization energies of dopants in hexagonal and cubic sites for both species are extracted and compared.

Abstract: Semiconductor devices rely on the incorporation of donor and acceptor atoms into the crystal lattice to form locally doped regions. For dopant atoms incorporated into SiC by ion implantation, a high-temperature annealing step is required to achieve electrical activation. This annealing step is accompanied by redistribution of the implanted atoms. The influence of the annealing parameters on dopant redistribution is crucial when aiming for ever smaller device dimensions. In this work, we present a consistent analysis of the diffusion of Al implanted in 4H-SiC after high-temperature annealing at 1650 °C and 1800 °C for different annealing times. We identify the equilibrium diffusion coefficient at long annealing times from Al profiles obtained by SIMS analyses for both annealing temperatures. The temperature dependence is determined using an Arrhenius representation. This allows to quantify the equilibrium diffusion lengths for the actual temperature profiles, including heating and cooling rates. We find that the measured diffusion lengths for short annealing times are larger than expected from equilibrium diffusion and attribute the excess length to transient enhanced diffusion. Comparing the transient diffusion lengths of room-temperature and 500 °C-implanted samples, we conclude that the transient behavior is likely related to residual crystal damage induced during the implantation process.

Abstract: Ion implantation, as a way of doping the 4H-SiC crystal, is one of the key components of modern power device fabrication. Aluminum is used to form p-type wells for the body of n-MOSFETs and low resistance p-type contacts using heavy doping. Therefore, the ion implantation process needs to be controlled over a wide range of process conditions including implant energies and doses. The fact that Al in 4H-SiC exhibits very low diffusion puts additional burden on the accuracy and predictability of any ion implantation engineering. In device design, these requirements can be addressed by applying computer simulations to predict doping profiles ahead of the actual implant step performed in a manufacturing facility. The accepted way to predict doping profiles is based on the binary-collision approximation (BCA), numerically implemented as a statistical Monte-Carlo (MC) method [1]-[3]. Nowadays, one can refer to simulation packages available from commercial vendors [4] for studying ion implantation using BCA-MC algorithms. However, while the physical accuracy of BCA models implemented in these packages has shown to be quite remarkable, predictable simulations for a complex material system as 4H-SiC requires calibration from data including secondary ion mass spectrometry (SIMS) and scanning electron microscopy (SEM).

Abstract: In this study, machine learning (ML) was employed to predict the electrical properties of finished devices, specifically focusing on the state of the contacts at the electrodes. The predictions are based on optical microscope images of the surface conditions, which were captured immediately following the laser doping of nitrogen atoms into 4H-SiC. The laser doping process involved varying the laser fluence from 0.4 to 4.0 J/cm² and using number of laser irradiation to 5, 10, 20, and 100 shots. The ML prediction was carried out in two steps. In STEP1, we classified the contact status into three types.: 1) Schottky junctions (insufficient doping), 2) Ohmic contact (good contact), and 3) Not ohmic (damage caused by laser irradiation). In STEP2, contact resistance prediction (numerical regression) was performed using the dataset predicted as an ohmic contact. As a result, we found that the three classifications in STEP1 could be predicted with a high accuracy of over 88%. Furthermore, the contact resistance prediction in STEP2 could be made with an accuracy (RMSPE: root mean square percent error) of 27.2%. Visualizing the prediction basis of numerical regression using modulus-reweighted grad-regression activation mapping (MoRAM) revealed that the ML model focused on the inside of the laser-irradiated area in the optical microscope image. The results of the scanning electron microscopy observation of the laser-irradiated area showed that ablation and residuals were generated during laser doping in that area. Consequently, it was concluded that our ML model predicted the contact resistance of the finished device taking into consideration these surface conditions. Even highly-skilled laser doping technicians have difficulty predicting the resistance values arising from the ablation and residue conditions. Based on above results, we conclude that our ML model is capable of predicting the electrical characteristics of a finished device, a task that is often considered challenging for humans.