Browsing by Author "Cai, Yunjiao"
Now showing 1 - 3 of 3
Results Per Page
Sort Options
Item Comparison of Statistical Learning and Predictive Models on Breast Cancer Data and King County Housing Data(Department of Applied Economics and Statistics, University of Delaware, Newark, DE., 2017-09) Cai, Yunjiao; Fu, Zhuolun; Zhao, Yuzhe; Hu, Yilin; Ding, ShanshanIn this study, we evaluate the predictive performance of popular statistical learning methods, such as discriminant analysis, random forests, support vector machines, and neural networks via real data analysis. Two datasets, Breast Cancer Diagnosis in Wisconsin and House Sales in King County, are analyzed respectively to obtain the best models for prediction. Linear and Quadratic Discriminant Analysis are used in WDBC data set. Linear Regression and Elastic Net are used in KC house data set. Random Forest, Gradient Boosting Method, Support Vector Machines, and Neural Network are used in both datasets. Individual models and stacking of models are trained based on accuracy or R-squared from repeated cross-validation of training sets. The final models are evaluated by using test sets.Item Quantitative Analysis of Spin Relaxation and Spin Transfer in Mesoscopic Nonlocal Spin ValvesCai, YunjiaoSpintronics is a new, emerging, and advancing academic research area in physics. Understanding the spin relaxation mechanism and spin transfer effect is crucial in spintronics. Accordingly, new methods are developed in this dissertation to determine the spin relaxation lengths accurately, and to explore the spin relaxation mechanism in mesoscopic Cu channels. In addition, novel nonlocal structures are designed and fabricated to generate efficient spin transfer switching with pure spin currents. A large number of mesoscopic nonlocal spin valve (NLSV) devices have been used to determine the Cu spin relaxation length and analyze the spin relaxation mechanism. Two different but related methods are used. In the first method, many NLSVs are fabricated on the same substrate under identical processing conditions, and the average Cu spin relaxation lengths at 10 K and 295 K are accurately determined. This method relies on Cu resistivity values determined directly from the NLSV devices. An iterative approach is used to take into account the dependence of injection/detection spin polarization on the size of the ferromagnetic electrodes. The probabilities of spin-flip for bulk defects and phonons in the mesoscopic Cu channels are shown to be $\mathrm{<}$ 5 $\times$ ${10}^{-4}$.. However, experimental data suggest that spin relaxation length and resistivity could both vary even for NLSVs fabricated under identical conditions. Therefore, a second method is developed to extract a distinct value of spin relaxation length from each individual NLSV. A dependence of Cu spin relaxation length on the Cu resistivity can then be established from a large number ($\mathrm{>}$100) of NLSVs. Such a dependence is very important because it reflects the underlying mechanism of spin relaxation. By changing the dimensions of the Cu channels and measurement temperatures, the Cu resistivity is tuned by more than one order of magnitude. By analyzing the relationship between Cu spin relaxation length and Cu resistivity, we conclude that the spin relaxation can be described by the Elliott-Yafet model. However, the spin-flip probabilities at surfaces are substantially higher than those in the bulk. Large values of spin relaxation lengths ($\mathrm{\sim}$ 2.0 $\mu$m at 10 K and $\mathrm{\sim}$ 700 nm at 295 K) can be achieved in Cu channels with lower resistivity. This is encouraging for the prospect of using mesoscopic Cu wires as spin transport channels. Another theme of this dissertation is nonlocal spin transfer switching with pure spin currents. Spin transfer effects in nonlocal lateral structures are pivotal in realizing spintronic devices such as all-spin logic with built-in memory. Efficient nonlocal spin transfer switching is achieved by using separately tailored polarizer and free-layer interfaces. A low-resistance oxide interface with larger area is used between the ferromagnetic polarizer and the Cu channel to achieve substantial spin polarization with low injection charge current density. An ohmic interface with smaller area is used between the ferromagnetic free-layer and the Cu channel to facilitate the absorption of spin current with high areal density. A feasibility study is first conducted to demonstrate that the spin polarization provided by a low-resistance oxide interface with relatively large area (330 nm $\times$ 170 nm) can be sufficiently high. Subsequently, nonlocal spin transfer devices are fabricated and characterized. By designing the polarizer and free-layer interfaces separately, we achieve reversible and bistable spin transfer switching with a modest charge current density of $\mathrm{\sim}$ 6 $\times$ 10${}^{6}$ A$\cdot$cm${}^{-2}$ between 100 K and 150 K. With potentials for improvements, nonlocal spin transfer structures can be as efficient and robust as the nanopillar spin transfer structures.Item Tuning of spin relaxation and the Kondo effect in copper thin films by ionic gating(Physical Review B, 2022-08-11) Shen, Xingyu; Cai, Yunjiao; Wu, Yizheng; Ji, YiSpin relaxation length is a fundamental material parameter that influences all aspects of spin dependent transport. The ability to tune the spin relaxation length leads to novel spintronic phenomena and functionalities. We explore the tunability of the spin relaxation length in the mesoscopic Cu channels of nonlocal spin valves by using the ionic gating technique via a Li+ containing solid polymer electrolyte. At 5 K, the Cu spin relaxation length λCu is tuned reversibly between 670 and 410 nm and the Cu resistivity ρCu is tuned by 9%. The strength of the Kondo effect due to the Fe impurities in Cu is tuned by one order of magnitude. At 295 K, λCu is tuned between 380 and 300 nm and ρCu is tuned by 7%. A gradual amplification of the tuning ranges by repeated gate cycling is observed and clearly suggests an electrochemical origin. Tunable spin relaxation in simple metals enriches functionalities in metal-based spintronics and shines light on fundamental spin relaxation mechanisms.