Search

An equilibrium carrier statistics approach with a partial ionization model is proposed to determine the energy level of CN deep donors in p-type GaN with heavy Mg doping. Unlike usual compensating centers that are assumed to be fully ionized under equilibrium, partial ionization of the CN donor was taken into consideration. The energy level of the CN donor is determined to be EV + (0.20 ± 0.01) eV at elevated temperatures (∼350 K) using such a partial ionization model. The donor level for an isolated C center at a low temperature limit is further calculated considering the doping and temperature effects, which is EV + (0.32 ± 0.01) eV. Furthermore, the ionization ratio of CN is found to be dependent on the C concentration and can then be estimated to be in the range of 0.3–0.8. Such a partial ionization characteristic of CN may capture/emit free carriers during device operation and should be taken into account when analyzing device reliability. [ABSTRACT FROM AUTHOR]

The control of carbon (C) impurity plays a crucial role in achieving high mobility GaN drift layers for GaN vertical power devices. We investigate the effects of NH3 partial pressure on the C incorporation in GaN and find out that the NH3 partial pressure is the key parameter controlling the C concentration. It is found that the C concentration is inversely proportional to the square of NH3 partial pressure under a fixed growth rate. Further physical analysis indicates that higher NH3 partial pressure results in a lower surface N vacancy (VN) concentration and, thus, the lower C incorporation efficiency. Under such a guidance, the C concentration in GaN can be reduced to as low as ∼1.7 × 1015/cm3 at a growth rate of 1 μm/h. As a consequence, a record high electron mobility of 1227 cm2/V·s at the room temperature is achieved for a GaN drift layer grown on Si substrates. Our strategy is effective in C control and paves the way for realizing high performance kV-class GaN vertical power devices on Si substrates. [ABSTRACT FROM AUTHOR]

Heavy carbon (C) doping is of great significance for semi-insulating GaN in power electronics. However, the doping behaviors, especially the atomic configurations and related self-compensation mechanisms, are still under debate. Here, with the formation energy as the input parameter, the concentrations of C defects with different atomic configurations are calculated by taking the configurational entropy into account. The result shows that the concentrations of tri-carbon complexes (CNCiCN, where Ci refers to interstitial carbon) and dicarbon complexes (CNCGa) cannot be neglected under heavy doping conditions. The concentration of CNCiCN can even exceed that of CN at sufficiently high doping levels. Especially, we suggest that it is the tri-carbon complex CNCiCN, instead of the commonly expected CGa, that acts as the self-compensation centers in semi-insulating GaN under heavy C doping conditions. The results provide a fresh look on the long-standing problem about the self-compensation mechanisms in C doped GaN. [ABSTRACT FROM AUTHOR]

Identifying atomic configurations of impurities in semiconductors is of fundamental interest and practical importance in designing electronic and optoelectronic devices. C impurity acting as one of the most common impurities in GaN, it is believed for a long time that it substitutes at Ga site forming CGa with +1 charge-state in p-type GaN, while it substitutes at N site forming CN with -1 charge-state in n-type GaN. However, by combining x-ray absorption spectroscopy and first-principles simulations, we observed that C is mainly occupying the N site rather than the Ga one in p-GaN. We further reveal that this is due to an H-induced EF-tuning effect. During growth, the existing H can passivate Mg dopants and upshifts the EF to the upper region of bandgap, leading to the CN formation. After the p-type activation by annealing out H, although the EF is pushed back close to the valence band maximum, whereas the extremely large kinetic barrier can prevent the migration of C from the metastable CN site to ground-state CGa site, hence stabilizing the CN configuration. Additionally, the CN with neutral charge-state ( C N 0 ) in the p-GaN is further observed. Therefore, the real C-related hole-killer in p-type GaN could be CN rather than the commonly expected CGa. Our work not only offers the unambiguous evidence for the C defect formation in p-GaN but also contributes significantly to an in-depth understanding of the C-related hole-killers and their critical role on electrical and optoelectrical properties of p-GaN and even p-AlGaN. [ABSTRACT FROM AUTHOR]

It is empirically well acknowledged that C doping makes GaN high-resistive. However, the detailed doping type and high-resistivity mechanisms of C doped GaN, which are extremely essential for GaN power electronics, still remain unclear. In this work, we clarify the mutative (from downward to upward) shift of the Fermi level and the n -type conductivity in heavily C doped GaN grown by MOCVD for the C concentration increases over a critical value, by combining photo-assisted KPFM and Seebeck coefficient measurements. We also discuss the reverse transition of Fermi level and ultimately n -type conductivity should be attributed to the self-compensation and other donor-type compensation centers introduced along with C doping. [ABSTRACT FROM AUTHOR]

It has been established that the formation of point defects and their behaviors could be regulated by growth details such as growth techniques and growth conditions. In this work, we prove that C doping approaches have great influence on the charge state of C N , thus the interaction between H and C in GaN. For GaN with intrinsic C doping, which is realized by reducing the V/III ratio, C N mainly exists in the form of C N − charged from the higher concentration of V N and, thus, may attract H + by coulomb interaction. Whereas for the extrinsically C doped GaN with propane as the doping source, the concentration of V N is reduced, and C N mainly exists in neutral charge state and, thus, nearly does not attract H ions. Therefore, we demonstrate that the interplay between H and C atoms is weaker for the extrinsically C doped GaN compared to the intrinsically doped GaN, thus gives a clear picture about the different charge states of C N and the formation of C–H complexes in GaN with different C doping approaches. [ABSTRACT FROM AUTHOR]

In this work, we study the impacts of different types of strain relief layer (SRL) on dynamic on-resistance (R on) degradation of GaN power devices on Si by back-gate ramping and vertical leakage measurement. Our study reveals that the SRL has important effects on the dynamic R on. Compared with step-graded AlGaN SRL, the superlattice SRL possesses much more energy barriers, which can more effectively block the leakage of holes from GaN buffer and the injection of electrons from Si substrate. Enhancing the carrier blocking ability of SRL could contribute to the suppression of dynamic R on degradation. [ABSTRACT FROM AUTHOR]

The time and monetary costs of training sophisticated deep neural networks are exorbitant, which motivates resource-limited users to outsource the training process to the cloud. Concerning that an untrustworthy cloud service provider may inject backdoors to the returned model, the user can leverage state-of-the-art defense strategies to examine the model. In this paper, we aim to develop robust backdoor attacks (named RobNet) that can evade existing defense strategies from the standpoint of malicious cloud providers. The key rationale is to diversify the triggers and strengthen the model structure so that the backdoor is hard to be detected or removed. To attain this objective, we refine the trigger generation algorithm by selecting the neuron(s) with large weights and activations and then computing the triggers via gradient descent to maximize the value of the selected neuron(s). In stark contrast to existing works that fix the trigger location, we design a multi-location patching method to make the model less sensitive to mild displacement of triggers in real attacks. Furthermore, we extend the attack space by proposing multi-trigger backdoor attacks that can misclassify inputs with different triggers into the same or different target label(s). We evaluate the performance of RobNet on MNIST, GTSRB, and CIFAR-10 datasets, against four representative defense strategies Pruning, NeuralCleanse, Strip, and ABS. The comparison with two state-of-the-art baselines BadNets and Hidden Backdoors demonstrates that RobNet achieves higher attack success rate and is more resistant to potential defenses. [ABSTRACT FROM AUTHOR]

We have investigated the interaction between carbon impurities and threading dislocations and their impact on the transport properties of GaN grown on Si substrates. The incorporation of carbon impurity was found to be associated with dislocation density, with a linear dependence. It indicates that the carbon may accumulate around the dislocations. The temperature-dependent Hall-effect measurement further confirmed that those carbon-decorated dislocations can act as acceptor-like traps, existing at every c-lattice spacing along a threading dislocation. The acceptor-like traps are important scattering centers and, thus, cannot be neglected. By reducing the density of the carbon-decorated dislocation via introducing a thick dislocation filtering layer to reduce the dislocation-related acceptor-like trap scattering, a record room-temperature electron mobility of 1090 cm2/V s with a carrier concentration of ∼2 × 1016 cm−3 for n--GaN on Si was achieved. Our results provide an effective approach to obtain high-quality n−-GaN on Si for vertical GaN based devices. [ABSTRACT FROM AUTHOR]

Carrier compensation traps in n−-GaN drift layers grown on Si substrates were investigated using high-temperature deep-level transient spectroscopy (DLTS). The upper limit of the temperature range (700 K) allows for the study of deeper levels in the bandgap than those previously reported by conventional DLTS. Three trap states were revealed to be responsible for carrier compensation. Besides the residual carbon (C) acceptor, two deep electron traps detected in the DLTS high-temperature range, labeled E2 and E3 with energies EC of 0.98 and 1.38 eV, respectively, were also found to have contributions to the carrier compensation. A comprehensive investigation combining with positron annihilation spectroscopy measurements revealed that E2 and E3 are related to the (–/2–) and (0/–) acceptor levels of the VGa–ON complex, respectively. The relatively high concentrations of E2 and E3 imply that the VGa–ON complex is an essential carrier compensation source in the drift layer and plays a crucial role in developing kV-class vertical GaN power devices. [ABSTRACT FROM AUTHOR]

Deep neural networks have achieved tremendous success in various fields, especially in recognition and classification applications. However, faced with the difficulty of training millions of parameters of such networks, many users outsource the training procedure of a specific prediction work to the powerful cloud servers that own abundant computation and storage resources. Although such outsourced training can significantly simplify and expedite the development circles, it also introduces many security risks. In recent years, a new type of attack, the so-called backdoor attack, has attracted much attention, where the attacker's goal is to create a maliciously deep neural network to make misclassification on the special inputs with the backdoor trigger. For its concealment, such attacks can potentially cause disastrous consequences. Subsequently, many defense mechanisms against this attack are also appearing. In this article, we conduct a retrospective review on the existing schemes of the backdoor attacks and defenses in outsourced cloud environments. According to the resources the adversary has, and whether the detection time is during run-time or not, we classify the attack and defense approaches into multiple categories. We present a detailed overview of each category, and we provide a comparison of these approaches and evaluate part of the attack schemes by the experiments. We also highlight various future research directions in this field. These views shed light on possible avenues for future research. [ABSTRACT FROM AUTHOR]

It has been established that hydrogen (H) plays a key role in p-type doping of GaN and it must be removed by dissociation of the Mg–H complex in order to achieve p-type conductivity. However, in carbon (C)-doped semi-insulating GaN, which is the core component of power electronic devices, the role of H, especially the formation and dissociation process of C–H defects, has remained to date a mystery. In this work, we provide a direct evidence for the interaction between H and C in the form of the C N − H i complex in as-grown C-doped GaN. The complex can be dissociated into C N − and H + after post-growth annealing. The activation energy is estimated to be about 2.3–2.5 eV from the temperature-dependent annealing experiments. Our study reveals that the C N − H i complex plays an essential role in understanding the variation of optical and electronic properties of C-doped GaN. [ABSTRACT FROM AUTHOR]

It is commonly believed that a torn ACL or a damaged meniscus may be associated with altered knee joint movements. The purpose of this study was to measure the tibiofemoral kinematics of ACL deficiency with concomitant meniscus deficiency. Unilateral knees of 28 ACL deficient participants were studied while ascending stairs. Among these patients, 6 had isolated ACL injuries (group I), 8 had combined ACL and medial meniscus injuries (group II), 8 had combined ACL and lateral meniscus injuries (group III) and 6 had combined ACL and medial-lateral meniscus injuries (group IV). Both knees were then scanned during a stair climb activity using single fluoroscopic image system. Knee kinematics were measured at 0°, 5°, 10°, 15°, 30° and 60° of flexion during ascending stairs. At 0°, 15° and 30° flexion of the knee, the tibia rotated externally by 13.9 ± 6.1°,13.8 ± 9.5° and 15.9 ± 9.8° in Group I. Group II and III exhibited decreased external rotation from 60° to full extension. Statistical differences were found in 0°, 15°and 30° of flexion for the 2 groups compared with Group I. In general, the tibia showed anterior translation with respect to the femur during ascending stairs. It was further determined that Group III had larger anterior translation compared with Group IV at 0° and 5° of flexion (−6.9 ± 1.7 mm vs. 6.2 ± 11.3 mm,P = 0.041; −9.0 ± 1.8 mm vs. 8.1 ± 13.4 mm,P = 0.044). During ascending stairs the ACL deficient knee with different deficiencies in the meniscus will show significantly different kinematics compared with that of uninjured contralateral knee. Considering the varying effect of meniscus injuries on knee joint kinematics, future studies should concentrate on specific treatment of patients with combined ACL and meniscus injuries to protect the joint from abnormal kinematics and subsequent postoperative degeneration. [ABSTRACT FROM PUBLISHER]

Background: Numerous studies have investigated kinematic alterations in patients with an isolated anterior cruciate ligament (ACL) injury. However, a substantial proportion of patients with injured ACLs also have concomitant meniscal tears. Purpose: To evaluate the in vivo alteration of knee kinematics after an ACL tear, with or without a combined medial or lateral meniscal tear, during level walking activity. Study Design: Controlled laboratory study. Methods: Fifty-six patients with unilateral ACL-deficient (ACLD) knees were studied. Among these patients, 15 had isolated ACL injuries (group 1), 15 had combined ACL and medial meniscal injuries (group 2), 15 had combined ACL and lateral meniscal injuries (group 3), and 11 had combined ACL and medial/lateral meniscal injuries (group 4). The kinematics of each knee was determined using an optical tracking system during treadmill gait. Range of motion (ROM) and kinematic alterations were compared between the contralateral ACL-intact (ACLI) and ACLD knees. Results: All ACLD knees, with or without meniscal deficiency, had significantly less flexion than the ACLI knees (~3°-8°; P < .05). In groups 1, 3, and 4, the injured knees exhibited more femoral external rotation by approximately 1° to 2° (P < .05). Group 4 showed posterior femoral translation (anterior tibial translation) in the swing phase (~13 mm), while groups 2 and 3 showed increased anterior femoral translation (~2-3 mm). During medial-lateral translation, patients in group 2 had a more medial femoral shift (~4 mm) relative to the tibia initially, while those in group 4 exhibited a lateral femoral shift. Conclusion: The results indicate that meniscal injuries alter the kinematics of the ACLD knee when compared with knees with an isolated ACL injury. The location of the meniscal tear also affects knee kinematics. Clinical Relevance: Considering the varying effects of meniscal injuries on knee joint kinematics, these data provide insight into the pathological function of the ACL-injured knee joint during walking. [ABSTRACT FROM AUTHOR]

The rapid corrosion rate hinds the clinical application of magnesium-based materials. Therefore, microarc oxidation (MAO) coatings were fabricated on a magnesium alloy named AZ31B to improve its anti-corrosion ability, while the surface morphology and degradation behavior were studied. The adhesion, proliferation and osteogenic differentiation of bone marrow stromal cells (BMSCs) cultured with MAO–AZ31B were investigated. In contrast to uncoated one, MAO–AZ31B performed a long-term corrosion resistance for more than five weeks. Evidenced by cell test, MAO coatings could promote BMSCs adhesion, proliferation and induced osteogenic differentiation in comparison with uncoated samples. Thus MAO showed beneficial effects on the corrosion resistance of, and thus improved cell adhesion to the biodegradable AZ31B, and is a promising method for surface modification of biomedical magnesium alloys. [ABSTRACT FROM AUTHOR]

Normative tibiofemoral data of Chinese or Asian subjects during gait is rarely reported. This study is aimed at investigating the six-degree-of-freedom (6DOF) knee kinematics of adult Chinese during gait, based on a novel portable system. Twenty-eight healthy Chinese subjects (56 knees) were studied during their treadmill gaits. A set of optical marker clusters were attached to the thighs and shanks of each subject, who was tracked by an optical joint kinematics measurement system. Knee landmarks were initially digitized with respect to the marker cluster sets to determine the local coordinate systems for calculation of 6DOF knee joint kinematics. The range of motion (ROM) in 6DOF and 5 kinematic parameters were calculated and compared between bilateral knees and genders. We discovered that knee rotations, as well as motion in proximodistal and mediolateral translations, showed similar patterns in flexion and extension. However, the anteroposterior translations did not show a clear pattern. The results of ROM in 6DOF obtained in this study are comparable with those reported in existing literature. No statistical difference was found between left and right knees either in the ROMs or in the 5 kinematic parameters. However, the ROM in the mediolateral direction during gait was found to be higher in men than women ( P = 0.014). In addition, the femurs of female subjects rotated more internally than the femurs of male during the stance phase ( P = 0.011). We concluded that normal Chinese knees exhibited distinct gait patterns, except for anteroposterior motion. Women and men exhibit different axial rotations and mediolateral translation patterns during their treadmill gait. [ABSTRACT FROM AUTHOR]