Large Bias Research Articles

Cultural Bias in Large Language Models: A Comprehensive Analysis and Mitigation Strategies

Abstract This paper delves into the intricate relationship between Large Language Models (LLMs) and cultural bias. It underscores the significant impact LLMs can have on shaping a more equitable and culturally sensitive digital landscape, while also addressing the challenges that arise when integrating these powerful AI tools. The paper emphasizes the immense significance of LLMs in contemporary AI research and applications, underpinning many systems and algorithms. However, their potential role in perpetuating or mitigating cultural bias remains a pressing issue warranting extensive analysis. Cultural bias stems from various intertwined factors; the following analysis categorizes cultural bias shaping LLMs into three dimensions: data quality, algorithm design, and user interaction dynamics. Furthermore, the impacts of LLMs on cultural identity and linguistic diversity are scrutinized, highlighting the interplay between technology and culture. The paper advocates responsible AI development, outlining mitigation strategies such as ethical guidelines, diverse training data, user feedback mechanisms, and transparency measures. In conclusion, the paper emphasizes that cultural bias in LLMs is not solely a problem but also presents an opportunity. It can enhance our awareness and critical understanding of our own cultural biases while fostering curiosity and respect for diverse cultural perspectives.

Revisiting Thorpe scale analysis and diapycnal diffusivities in Drake Passage

Ship heave effects in the swell-prevalent Southern Ocean induce frequent false overturns in density profiles, and conventional Thorpe scale analysis leads to unrealistically large diapycnal diffusivities. Another critical factor causing large positive biases in Thorpe scale analysis concerns the Ozmidov to Thorpe scales ratio α often considered constant and equal to 0.8. A revised Thorpe scale analysis is proposed with an innovative segment-by-segment approach to circumvent heave-induced false overturns together with best-fitting α functions to Diapycnal and Isopycnal Mixing Experiment in the Southern Ocean (DIMES) data, which yields realistic diffusivities in the Drake Passage region. The revised Thorpe scale method applied to a finely-resolved CTD section across the Drake Passage yields typical diffusivities of O (10-5-10-3 m2 s-1), with the largest values being preferentially concentrated close to the bottom at major circumpolar fronts, consistent with nearby DIMES microstructure observations. In the southern Drake Passage the method highlights a middepth (500-1500 m) mixing maximum of 10-4-10-3 m2 s-1 in the Lower Circumpolar Deep Water (LCDW) layer due to large intrusions of Warm Deep Water (a cold/fresh variety of LCDW) of Weddell Gyre origin. The inferred middepth mixing maximum in intrusive regions, which is associated with isopycnal eddy stirring rather than caused by internal waves or double diffusive mixing, supports the short-circuiting paradigm of meridional overturning circulation in the southwestern Scotia Sea region and has important climatic implications in a warming climate.

Empirical Monte Carlo evidence on estimation of timing-of-events models

This article builds on the Empirical Monte Carlo simulation approach to study the estimation of Timing-of-Events (ToE) models. We exploit rich Swedish data of unemployed job seekers with information on participation in a training program to simulate placebo treatment durations. We first use these simulations to examine which covariates are key confounders to be included in dynamic selection models for training participation. The joint inclusion of specific short-term employment history indicators (notably, the share of time spent in employment), together with baseline socio-economic characteristics, regional and inflow timing information, is important to deal with selection bias. Next, we omit subsets of explanatory variables and estimate ToE models with discrete distributions for the ensuing systematic unobserved heterogeneity. In many cases, the ToE approach provides accurate effect estimates, especially if time-varying variation in the unemployment rate of the local labor market is taken into account. However, assuming too many or too few support points for unobserved heterogeneity may lead to large biases. Information criteria, in particular those penalizing parameter abundance, are useful to select the number of support points. A comparison with other duration models shows that a Stratified Cox model performs well with abundant multiple spells but less well when multiple spells are uncommon. The standard Cox regression model performs poorly in all configurations as it is unable to account for unobserved heterogeneity.

The efficiency of detecting seabird behaviour from movement patterns: the effect of sampling frequency on inferring movement metrics in Procellariiformes

BackgroundRecent technological advances have resulted in low-cost GPS loggers that are small enough to be used on a range of seabirds, producing accurate location estimates (± 5 m) at sampling intervals as low as 1 s. However, tradeoffs between battery life and sampling frequency result in studies using GPS loggers on flying seabirds yielding locational data at a wide range of sampling intervals. Metrics derived from these data are known to be scale-sensitive, but quantification of these errors is rarely available. Very frequent sampling, coupled with limited movement, can result in measurement error, overestimating movement, but a much more pervasive problem results from sampling at long intervals, which grossly underestimates path lengths.MethodsWe use fine-scale (1 Hz) GPS data from a range of albatrosses and petrels to study the effect of sampling interval on metrics derived from the data. The GPS paths were sub-sampled at increasing intervals to show the effect on path length (i.e. ground speed), turning angles, total distance travelled, as well as inferred behavioural states.ResultsWe show that distances (and per implication ground speeds) are overestimated (4% on average, but up to 20%) at the shortest sampling intervals (1–5 s) and underestimated at longer intervals. The latter bias is greater for more sinuous flights (underestimated by on average 40% when sampling > 1-min intervals) as opposed to straight flight (11%). Although sample sizes were modest, the effect of the bias seemingly varied with species, where species with more sinuous flight modes had larger bias. Sampling intervals also played a large role when inferring behavioural states from path length and turning angles.ConclusionsLocation estimates from low-cost GPS loggers are appropriate to study the large-scale movements of seabirds when using coarse sampling intervals, but actual flight distances are underestimated. When inferring behavioural states from path lengths and turning angles, moderate sampling intervals (10–30 min) may provide more stable models, but the accuracy of the inferred behavioural states will depend on the time period associated with specific behaviours. Sampling rates have to be considered when comparing behaviours derived using varying sampling intervals and the use of bias-informed analyses are encouraged.

A Comprehensive Evaluation of Mean and Extreme Climate for the Conformal Cubic Atmospheric Model (CCAM)

Abstract This study evaluates the performance of the Conformal Cubic Atmospheric Model (CCAM) in dynamically downscaling fifth major global reanalysis produced by European Centre for Medium-Range Weather Forecasts (ECMWF) (ERA5) reanalysis data from 1985 to 2014, following a 5-yr spinup period. It focuses on daily maximum and minimum temperatures and daily precipitation, comparing CCAM to ERA5 and the Australian Gridded Climate Data (AGCD). The CCAM effectively reduces warm biases in daily minimum temperatures but struggles with cold biases in daily maximum temperatures, particularly in northern Australia during the wet season, possibly due to high-level cloud overestimation. Precipitation tends to be overestimated, especially in extreme rainfall events, though offset by an underestimation of low rainfall. The study showcases improvements in the annual minimum of daily minimum temperatures across most of Australia, while identifying challenges in forecasting cooler extreme temperatures. It adds value to annual maximum daily maximum temperatures in southern Australia but less so in the north. The analysis of the 5% annual exceedance probability (AEP5%) yields mixed results influenced by location and potential ocean temperature changes. Some coastal areas exhibit lost value, possibly linked to ocean temperature shifts. Furthermore, CCAM’s representation of maximum annual daily and 5-day rainfall reveals lost value, particularly in eastern Australia due to an overestimate of extreme rainfall. Despite the challenges of comparing a dynamical downscaling model like CCAM to ERA5, this study highlights its benefits in reducing biases, especially in temperature representation. Given the larger biases in phase 6 of Coupled Model Intercomparison Project (CMIP6) global climate models, CCAM appears suitable for dynamic downscaling in climate projections, emphasizing the need for ongoing model enhancements, including addressing biases related to ephemeral water bodies and extreme rainfall. Significance Statement This study critically assesses the performance of the Conformal Cubic Atmospheric Model (CCAM) in dynamically downscaling ERA5 reanalysis data from 1985 to 2014, offering valuable insights into climate modeling. Focusing on temperature and precipitation, CCAM proves effective in mitigating warm biases in daily minimum temperatures but encounters challenges with cold biases in daily maximum temperatures, particularly in northern Australia. The analysis reveals the overestimation of precipitation, especially in extreme events, yet identifies improvements in annual minimum daily minimum temperatures across Australia. The study underscores CCAM’s potential in reducing biases compared to CMIP6 global climate models, making it a promising tool for dynamic downscaling in climate projections. It emphasizes the necessity for ongoing model enhancements, particularly addressing biases related to ephemeral water bodies and extreme rainfall.

Negative global metacognitive biases are associated with depressive and posttraumatic stress disorder symptoms and improve with targeted or game-based cognitive training.

Metacognition is disrupted in several clinical populations. One aspect of metacognition, global metacognitive bias (difference between objective and self-reported abilities), has shown to be particularly relevant to clinical functioning. However, previous studies of global metacognitive biases in populations with elevated depressive/posttraumatic stress disorder (PTSD) symptoms have not measured objective and self-reported abilities relative to normative samples, limiting the quantification of biases. Additionally, few studies have examined whether cognitive interventions can improve metacognitive biases or how this relates to depressive/PTSD symptom severity. A total of 84 participants with mild traumatic brain injury (77% veterans) performed PTSD and depression assessments along with self-reported and objective measures of global cognition. Age-adjusted norm-based z scores were used for self-reported and objective cognition, and bias was calculated by subtracting objective minus self-report scores. Participants then received 13 weeks of targeted cognitive training or entertainment games training (both providing performance feedback). Participants were assessed at baseline, immediately posttraining, and 3 months posttraining. We found large negative metacognitive biases in those with clinically significant severity of depressive symptoms (z score difference = -1.77), PTSD symptoms (-1.47), and depressive + PTSD symptoms (-2.29). Metacognitive biases improved after both targeted and entertainment training and was associated with reductions in depressive/PTSD symptom severity (r = -.41/-.42, respectively), led by the entertainment training group (r = -.54/-.46, respectively). These findings show that clinically significant severity of depressive/PTSD symptoms is associated with substantial negative global metacognitive biases and preliminarily suggests that cognitive training may improve these biases and depressive/PTSD symptom severity. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

The Use of Deep Learning in Open Learning: A Systematic Review (2019 to 2023)

No records of systematic reviews focused on deep learning in open learning have been found, although there has been some focus on other areas of machine learning. Through a systematic review, this study aimed to determine the trends, applied computational techniques, and areas of educational use of deep learning in open learning. The PRISMA protocol was used, and the Web of Science Core Collection (2019–2023) was searched. VOSviewer was used for networking and clustering, and in-depth analysis was employed to answer the research questions. Among the main results, it is worth noting that the scientific literature has focused on the following areas: (a) predicting student dropout, (b) automatic grading of short answers, and (c) recommending MOOC courses. It was concluded that pedagogical challenges have included the effective personalization of content for different learning styles and the need to address possible inherent biases in the datasets (e.g., socio-demographics, traces, competencies, learning objectives) used for training. Regarding deep learning, we observed an increase in the use of pre-trained models, the development of more efficient architectures, and the growing use of interpretability techniques. Technological challenges related to the use of large datasets, intensive computation, interpretability, knowledge transfer, ethics and bias, security, and cost of implementation were also evident.

Unconventionally Large Exchange Bias and Cluster Spin Glass State in Mn2(Ni,Sn) Heusler Alloys Melt Spun Ribbons

Unconventionally Large Exchange Bias and Cluster Spin Glass State in Mn₂(Ni,Sn) Heusler Alloys Melt Spun Ribbons

Design and implementation of an adaptive unscented Kalman filter with interval Type-3 fuzzy set for an attitude and heading reference system considering gyroscope bias

Low-cost microelectromechanical sensors used in attitude and heading reference systems (AHRS) suffer from unstable on–off biases, scaling errors, and nonlinearities that lead to an increase of navigation errors over time. To damp these errors and uncertainties from ultimate attitude and heading angles, the bias term of gyroscopes is augemented in under estimation state vector for high frequency feedback compensation. In this paper, an adaptive estimation algorithm to combine the strong tracking filter (STF) and interval type-3 fuzzy set (IT3FS) with an unscented Kalman filter (UKF) is proposed to deal with the large uncertainties in the microelectromechanical AHRS. The UKF, involving strong capability of dealing with nonlinearity, is limited by large instable bias uncertainty. By combining the STF to adaptively tuning of the UKF gain matrix and the IT3FS to determine the STF fading factor as well, the new IT3FS-STUKF algorithm is developed for high estimation performance against the significant uncertainties. To validate the proposed IT3FS-STUKF, real experiments are conducted with ground vehicles in urban areas, which show an improved accuracy compared to the extended Kalman filter (EKF) at reasonable computational burden.

An improved method for statistics estimation of out-of-plane load resistance of masonry walls using design-code models and mechanics-based finite element models

The susceptibility of masonry walls to out-of-plane failure under seismic loading presents a significant engineering concern. In adherence to contemporary limit state design philosophy and as mandated by performance-based design guidelines, it is crucial to accurately estimate statistics parameters such as the mean and variance of out-of-plane resistance. To achieve this, two deterministic models are often considered within the Monte Carlo simulation framework: design-code models, known for their computational efficiency but potential for significant model error, and mechanics-based finite element models, noted for their accuracy but computational intensity. Relying solely on either model for statistics estimation presents challenges due to their respective inherent limitations. This study introduces an improved strategy that synergizes the strengths of both models, resulting in enhanced statistics estimators for the out-of-plane resistance of masonry walls. The proposed methodology involves the use of the control variate method by integrating numerous design-code model evaluations to boost computational efficiency while incorporating a limited number of finite element model evaluations to ensure accuracy. The constructions of the proposed estimators follow rigorous optimization procedures to minimize associated errors and variances. Theoretical derivations are provided for essential elements in the estimator constructions, such as model evaluation numbers for finite element and design-code models and control variate coefficients. Moreover, the accuracy of the proposed estimators is compared with that of the crude Monte Carlo estimator. A key benefit of the proposed estimators is their unbiased nature, and their accuracy is not compromised by the discrepancies between the finite element model and the design-code model. To demonstrate the practicality of these estimators, two case studies are illustrated: one focusing on unreinforced masonry walls and the other on reinforced masonry walls. The results indicate that design-code model-based estimators exhibit large bias, and compared to the estimators that solely rely on the finite element model, the proposed estimators achieve higher accuracy given the same computational budget.

Evaluation of clear-sky surface downwelling shortwave fluxes computed by three atmospheric radiative transfer models

In this study the clear-sky total, direct, and diffuse shortwave (SW) fluxes at the surface, have been calculated by three radiation transfer models (RTMs) – MODTRAN6.0 (M6.0), Canadian Centre for Climate Modelling and Analysis (CCCma), and Langley-modified Fu-Liou (NASA CERES). These calculations have been evaluated by surface measurements collected from seven sites that represent different climatological regimes with various surface scene types including ocean, grassland/continental, desert, and snow/sea ice. For pristine atmospheric conditions, SW fluxes predicted by CCCma and M6.0 shows little variation, which lays a baseline for further analysis. Note that computing time required by CCCma is ∼1000 times smaller than M6.0. Based on all samples collected from seven sites, mean differences of total, direct, and diffuse fluxes between surface measurements and CCCma / M6.0 / Fu-Liou are [5.3 / 2.4 / 0.9], [-2.2 / -5.1 / -13.7], and [7.5 / 7.5 / 14.6] W m-2, respectively. Histograms of differences between the three RTM calculations and surface measurements show that CCCma computed direct and diffuse fluxes have the smallest biases with standard deviations similar to those for M6.0, while Fu-Liou values have the largest biases and standard deviations. While Fu-Liou outperforms for total flux, especially for desert conditions, it is hampered by large biases for direct and diffuse across all scene types. The three RTMs are consistent with showing the least error for total flux and the largest in diffuse based on bias, correlation, and root mean square error.

Deep residual fully connected network for GNSS-R wind speed retrieval and its interpretation

Global navigation satellite system reflectometry (GNSS-R) has emerged as a new technique to provide L-band bistatic measurements for ocean wind speed retrieval, in which traditional geophysical model functions (GMFs) or shallow neural networks (NNs) are normally used. However, it is still challenging to identify and consider all relevant parameters in the GMF. Meanwhile, NN models face limitations due to the degradation problem, which restricts their depth and consequently their performance. Furthermore, the interpretation of NN models for GNSS-R wind retrieval is another issue. To this end, we propose a residual fully connected network (RFCN) fusing auxiliary information such as geometry, receiver gain, significant wave height, and current speed with track-wise corrected σ0. Referred to the European Centre for Medium-Range Weather Forecast (ECMWF) ERA5 wind product, the root mean square error (RMSE) and bias of RFCN winds are 1.031 m/s and -0.0003 m/s, respectively, with a 6% improvement in RMSE compared to debiased NOAA Cyclone Global Navigation Satellite System (CYGNSS) Version 1.2 (V1.2) wind speed retrieval. Moreover, in an intertropical convergence zone (ITCZ) area with large current speeds, the RMSE and bias are 1.006 m/s and -0.022 m/s: an improvement of 11.6% and 87.9% compared to debiased NOAA CYGNSS V1.2 winds. The bias ‘strips’ in these areas are nearly eliminated. Daily averaged error analyses also demonstrate that RFCN winds are more robust and consistent with ECMWF winds. For wind speeds larger than 20 m/s, referred to Soil Moisture Active Passive (SMAP) Level 3 final wind products, the RMSE and bias of fine-tuning RFCN (FT_RFCN) winds are reduced by 25.7% and 91.5% compared to NOAA winds. Finally, the RMSE and bias of retrievals in tropical cyclones, measured by Stepped Frequency Microwave Radiometer (SFMR) during 2021-2022, reveal an improvement of 3.5% and 21.2% compared to NOAA winds. Through SHapley Additive exPlanations (SHAP) models developed for RFCN and FT_RFCN, the contribution of each feature is quantitatively evaluated, while providing insights into their interactions within the ‘black-box’ NN models with clear physical meanings.

Arctic September Sea Ice Concentration Biases in CMIP6 Models and Their Relationships with Other Model Variables

Abstract The models that participated in the Coupled Model Intercomparison Project (CMIP) exhibit large biases in Arctic sea ice climatology that seem related to biases in seasonal atmospheric and oceanic circulations. Using historical runs of 34 CMIP6 models from 1979 to 2014, we investigate the links between the climatological sea ice concentration (SIC) biases in September and atmospheric and oceanic model climatologies. The main intermodel spread of September SIC is well described by two leading EOFs, which together explain ∼65% of its variance. The first EOF represents an underestimation or overestimation of SIC in the whole Arctic, while the second EOF describes opposite SIC biases in the Atlantic and Pacific sectors. Regression analysis indicates that the two SIC modes are closely related to departures from the multimodel mean of Arctic surface heat fluxes during summer, primarily shortwave and longwave radiation, with incoming Atlantic Water playing a role in the Atlantic sector. Local and global links with summer cloud cover, low-level humidity, upper or lower troposphere temperature/circulation, and oceanic variables are also found. As illustrated for three climate models, the local relationships with the SIC biases are mostly similar in the Arctic across the models but show varying degrees of Atlantic inflow influence. On a global scale, a strong influence of the summer atmospheric circulation on September SIC is suggested for one of the three models, while the atmospheric influence is primarily via thermodynamics in the other two. Clear links to the North Atlantic oceanic circulation are seen in one of the models.

An end-to-end image-text matching approach considering semantic uncertainty

We propose a novel end-to-end image-text matching approach considering semantic uncertainty (SU-ITM), aiming to deal with the one-to-many semantic diversity involved in image-text matching in order to capture the associations between them more comprehensively and improve the robustness of the model. Traditional methods map images and texts as definite points in an embedding space to measure cross-modal similarity. However, the point-based embedding cannot capture the semantic uncertainty, leading to a large bias in the matching results. To address this problem, we model the one-to-many associations between image and text in a way that establishes a probability distribution, incorporating the uncertainty information into the final semantic representation of the text. In addition, we optimize the image-text matching loss so that the different text features approximate the image features in a distributed manner while maintaining the discriminative nature of the semantic representation, effectively reducing the matching uncertainty. Notably, our method achieves end-to-end training by not using pre-trained target detection branches throughout the training process. We fully demonstrate the excellent performance of our method in the image-text matching task through experimental validation on Flickr30k and MSCOCO. Excellent performance levels of 546.1 and 545.0 are achieved on the R@SUM metric for Flickr30k and MSCOCO 1k, respectively.

Observation of the multiple magnetic phases in double perovskite Pr1.8La0.2CoFeO6

The structural and magnetic properties of the double perovskite Pr1.8La0.2CoFeO6 have been investigated. The X-ray diffraction study shows that the system acquires room-temperature orthorhombic phase with the Pnma space group. The X-ray photoemission spectroscopy (XPS) measurement confirmed that the B-site ions are present in mixed valence states i.e. Co has been found in Co2+ and Co3+ valance states whereas Fe has been found in Fe3+/Fe4+. Magnetic measurements of the system confirm the existence of several fascinating magnetic behaviors, such as long-range canted antiferromagnetism withTN ∼ 271 K, giant exchange bias, and re-entrant cluster glass phase (TG ∼ 33 K). Field-dependent magnetization shows a large spontaneous exchange bias; HSEB∼ 1.8 kOe and giant conventional exchange bias, HCEB∼ 2.4 kOe at 5 K. Antiferromagnetic materials with large exchange bias can be utilized in high-density spintronic devices. Temperature-dependent Raman study demonstrates that the observed phonon mode exhibits anomalous behavior close to the magnetic transition temperature. Analysis of anomalous softening of phonon mode below TN, clarifies the presence of significant spin-phonon coupling while the magnetostriction effect does not play any significant role in the observed phonon anomaly.

Rapid Recovery of Degraded Perovskite Single-Crystal Radiation Detectors via Infrared Healing.

Radiation detectors based on metal halide perovskite (MHP) single crystals (SCs) have exhibited exceptional sensitivity, low detection limit, and remarkable energy resolution. However, the operational stability issue still dramatically impedes their commercialization due to degradation induced by high-energy irradiation and large bias. Here, we propose an innovative infrared healing strategy to restore the devices that have undergone severe damage from both long-term biasing and X-ray irradiation. Compared to the slow and inefficient intrinsic self-healing process of MHPs, the infrared healing method demonstrates the capacity to achieve rapid recovery of the detection performance of the degraded devices within just 1 h. We reveal that the healing mechanism is mainly related to the reduction of the ion-migration activation energy in MHP SCs under infrared illumination, which promotes the back diffusion of the displaced ions to their original lattice positions and remedies defects. Finally, the healing effect is further confirmed through the gamma-ray spectroscopy acquisition with degraded MHP SCs, whose energy resolution at 59.5 keV of 241Am source is improved from 36% to 12% following infrared illumination. These results present infrared healing as a simple and economic method to extend the service life of MHP SC-based detectors.

Analytical evaluation of the automated interleukin-6 assay on the Roche cobas e602 analyzer.

Interleukin-6 (IL-6) is a proinflammatory cytokine that is associated with many inflammatory diseases. This validation study evaluates the automated Roche Elecsys IL-6 electrochemiluminescent immunoassay that has been granted emergency use authorization by the US Food and Drug Administration. The Elecsys IL-6 assay was evaluated for precision, linearity, interference (by hemoglobin, bilirubin, triglycerides, and biotin) and clinical performance was compared to the V-PLEX Human IL-6 immunoassay (Meso Scale Discovery), performed by a reference laboratory. The Elecsys IL-6 assay is precise (intra-assay <3% coefficient of variation [CV], interassay <5% CV), exhibits an analytical measurable range of 1.5-4790 pg/mL, and is tolerant of significant interferences (H < 2522, I <62, L<2101, biotin <50ng/mL). Comparison with the V-PLEX assay revealed a 2.95 slope bias in patient samples evaluated for IL-6 concentration (n=43, range=1.5-1891 pg/mL, y=2.95x - 32.7, r2=0.84). Bland-Altman analysis revealed an absolute mean bias of 152 pg/mL (SD=254 pg/mL), or a mean percentage difference of 73%. The Roche IL-6 assay showed good analytical performance. The large systematic bias compared with another reference method precludes using multiple methods to monitor IL-6 response. The random-access nature of an automated IL-6 assay on the Roche platform makes the test available on demand.

High-k Dielectrics-Based Electric Field Engineering in β-Ga2O3 Power Diodes

We present β-Ga2O3 high-voltage Schottky barrier diodes (SBD) incorporating advanced electric field engineering approaches through the use of high-permittivity dielectrics. We present two approaches: 1) the lateral dielectric superjunction SBD, where the electric field within the drift layer is modified from triangular to rectangular profile, enhancing the breakdown voltage and enabling the use of higher drift layer doping to reduce on-resistance, and 2) the vertical high-k RESURF trench SBD, which leverages the reduced surface electric field effect (RESURF) to reduce the electric field at the metal-semiconductor junction and reverse leakage currents.A superjunction-like structure uses charge balance to push the VBR-Ron-sp trade-off beyond the material’s unipolar figure of merit. The lack of p-type doping in β-Ga2O3 makes the realization of such structures very difficult. However, field flattening can be achieved in β-Ga2O3 drift layers using a high permittivity material with appropriate choice of dielectric constant, device aspect ratio and semiconductor/dielectric width. An analytical model is discussed to highlight the device design strategy and experimental demonstration is reported in the lateral geometry. 500 nm of Si-doped Ga2O3 epitaxial layer with a sheet charge of 1.6×1013 cm-2 is grown using metal organic vapor phase epitaxy (MOVPE). Trenches with three different fin widths (WFin = 2, 5, 10 µm) were made and high-k dielectric BaTiO3 (BTO) with dielectric constant of 220 is deposited on the sample. A low specific on resistance of 1.65 mΩ-cm2 is observed for LAC = 5µm due to the high charge density epitaxial lateral Fins. For the SBD with LAC = 5µm and WFin = 5µm, a high average electric field of ~3 MV/cm (VBR/ LAC) is estimated across the entire length of the device. A very high Baliga's figure of merit (BFOM) of 1.35 GW/cm2 is observed for the RESURF SBD with LAC=5µm and WFin=2µm (VBR=1.49 kV and Ron-sp=1.65 mΩ-cm2), surpassing SiC unipolar FOM when considering just the conducting area.While superjunction-like structures effectively flatten the electric field inside the drift layer, high electric fields at the metal/semiconductor junction results in increased leakage current at large reverse bias, particularly in the case of vertical Schottky Barrier Diodes (SBDs). To circumvent this challenge, we present the approach of β-Ga2O3 Vertical Trench Schottky Barrier Diode (SBD) with a high-permittivity dielectric RESURF. The incorporation of a trench geometry, coupled with the high-permittivity dielectric RESURF, effectively reduces the surface electric field at the metal-semiconductor junction. This reduction facilitates the use of a lower work-function anode contact, further diminishing the turn-on voltage. The use of high-k dielectric also significantly reduces the dielectric leakage contribution to the overall reverse leakage of the diode. Large-area SBDs with areas of 1mm2 and 4mm2 exhibit pulsed forward current of 6A/20A and a catastrophic breakdown voltage of 1.74kV/1.4kV, surpassing other β-Ga2O3-based large-area power diodes, while small-area SBDs (200×200 µm2) do not exhibit catastrophic breakdown until 3kV. The large-area SBDs also demonstrate lower capacitance, stored charge, and stored energy compared to commercial SiC SBDs, indicating potential for superior switching performance. The combination of lower stored charge and a low forward voltage drop results in an excellent trade-off between conduction and switching power loss, yielding a QCVF figure of merit comparable to commercial bare die SiC SBDs. We demonstrate the potential of high permittivity dielectrics-based extreme field engineering to push the performance limits of β-Ga2O3 diodes for next-generation power electronics.We acknowledge funding from AFOSR MURI program (Dr. Ali Sayir) and Coherent/II-VI foundation.

Large model biases in the Pacific centre of the Northern Annular Mode due to exaggerated variability of the Aleutian Low

AbstractThe Northern Annular Mode (NAM) is traditionally defined as the leading empirical orthogonal function (EOF) of mean sea‐level pressure (MSLP) anomalies during winter. Previous studies have shown that the Pacific centre‐of‐action of the NAM is typically more amplified in models than in reanalysis. Here, we analyse the NAM in hindcasts from nine seasonal prediction models over 1993/1994–2016/2017. In all the models, the Pacific centre‐of‐action is much larger than in reanalysis over that period, during which the NAM and the North Atlantic Oscillation (NAO) are almost indistinguishable. As a result, the NAM in the models is correlated with Aleutian Low variability around four times more strongly than in reanalysis. We show that this discrepancy can be explained primarily by the amplitude of Aleutian Low variability, which is on average 17% higher in models than in reanalysis, with a secondary effect from a stronger correlation between the Aleutian Low and NAO. When the NAM is computed using zonally averaged MSLP, the Aleutian Low amplitude does not influence the pattern directly. Instead, the amplitude of the Pacific centre‐of‐action is governed primarily by the correlation between the Aleutian Low and NAO, reducing the apparent Pacific biases in models. While the two methods yield almost identical results in reanalysis, the large Aleutian Low biases result in differences when applied to model data. Modifying the MSLP statistically to alter the Aleutian Low amplitude reveals that the spatial pattern of the traditionally defined NAM is highly sensitive to Aleutian Low variability, even without modifying the correlation between the Aleutian Low and NAO. Hence, the NAM in models may not be as biased as the traditional method would suggest. We therefore conclude that the traditional EOF method is unsuitable for defining the NAM in the presence of highly amplified Aleutian Low variability, and encourage the use of the zonal‐mean method.

Analysis of anthropogenic CO2 emission uncertainty and influencing factors at city scale in Yangtze River Delta region: One of the world's largest emission hotspots

Cities occupy only 3% of the earth's land area, yet they contribute over 70% of anthropogenic CO2 emissions. Accurately quantifying the corresponding uncertainty of CO2 emissions at the city scale is the first step in simulating global/regional greenhouse gas concentration and implementing effective emission reduction policies. As the world's largest emitter, China has committed to reaching its CO2 emissions peak by 2030, and cities are facing substantial pressure to reduce CO2 emissions. The Yangtze River Delta (YRD) region is treated as the largest urban agglomeration in China and globally. Despite the availability of multiple emission inventories, comprehensive assessments of city-level CO2 emissions uncertainty within this region are still lacking. This study focused on the YRD region, compared six inventories and used nighttime light intensity, GDP, population, and satellite-based xCO2 concentrations as proxy data to identify potential sources of emission bias. The findings are as follows: (1) The city-level CO2 emissions in the YRD region ranged from tens of Mt to approximately 600 Mt. From 2010 to 2019, emissions in the region increased slowly. However, emission intensity (calculated by dividing the CO2 emissions by the Gross Domestic Product) showed a declining trend. The relatively low proportion of point source emissions in EDGAR inventory is attributed to its reliance on the point source CARMA, which overlooks smaller point sources. (2) The average uncertainties for low emission (0–50 Mt), medium emission (50–100 Mt) and high emission (>100 Mt) cities were 51.1%, 34.0%, and 45.0%, respectively. The absolute errors of emissions showed strong positive correlations with proxy data. There exists a logarithmic relationship between them and uncertainty, which can assist in estimating emission uncertainty in other cities in the future. (3) The ratios of biological CO2 flux to the ten-year average of CO2 emissions varied between-6.79% and −0.02%, which are much smaller than the uncertainty of anthropogenic emissions, the comparisons indicate that recent large biases in estimating anthropogenic CO2 emissions hinder the evaluation of carbon neutrality ability.