Studies on nanostructured transition and post-transition metal compound electrodes for photoelectrochemical cells

Jin Soo Kang 서울대학교 대학원 2015 국내박사

Photoelectrochemical cells are electrochemical cells that generally comprise semiconductor photoelectrode, electrocatalytic counter electrode, and electrolyte. There are two types of photoelectrochemical cells according to the function of the electrolyte; (i) photovoltaic cell with regenerative redox electrolyte and (ii) photosynthetic cell with sacrificial electrolye. Mesoscopic sensitized solar cells and photoelectrochemical water splitting cells are the typical examples of each kind, respectively. In this thesis, nanostructured electrodes composed of transition and post-transition metal compound for those cells were investigated. The main issues in the development of photoelectrodes, which are mainly metal oxide films, are improvements of light harvesting and charge collection. In contrast, replacement of high-cost and rare noble metal, typically platinum, with an economic and earth-abundant material has been the main objective of the research works on counter electrodes. Therefore, in this study, nanostructured metal oxides with intrinsically fast charge transport property and state-of-the-art nanomaterials with enlarged surface area were utilized as photoelectrodes. Also, nanostructured low-cost carbide and nitride materials with facile preparation methods were employed as electrocatalytic counter electrodes. Besides the discussions on quatitative improvements, in-depth physicochemical analyses for characterization of materials and photoelectrochemical cells were performed. After brief explanations of history, backgrounds, and previous research works on photoelectrochemical cells in chapter 1, in chapter 2 and 3, nanostructured metal oxide photoelectrodes with high carrier mobility are discussed. Mesoporous ZnO nanowire arrays with large surface area were synthesized by electrochemical anodization in a mild condition, and highly uniform SnO2 nanochannels were prepared by ultrasonic-assisted anodic oxidation. These nanostructured electrodes were employed as photoanode in quasi-solid dye-sensitized solar cells (DSCs), and fair energy conversion efficiencies were obtained, with further improvements by atomic layer deposition of TiO2 shells. In the next two chapters, strategies for efficient light harvesting by increased surface area were discussed. In chpater 4, TiO2 coated wrinkled silica nanoparticles were utilized as scattering centers in photoanode of DSCs. Superior performances to the DSCs with conventional sphere-shaped TiO2 scatterers were achieved due to the enlarged surface area, and futher observations on the relationship between spectral scattering properties and interwrinkle distances were made. In chapter 5, electrochemically anodized titanium and iron foams prepared by freeze-casting were used as photoanodes of DSCs and photoelectrochemical water splitting cells, respectively. By the anodic oxidation processes, one-dimensional TiO2 nanoatube arrays were formed on the titanium foam surface, and vertically aligned two-dimensional iron oxide nanoflakse were generated on the surface of the iron foam. These multidimensional structures had following three advantages; (i) large surface area for light harvesting, (ii) low-dimensionally confined semiconductor structure for enhanced charge transport, and (iii) three-dimensionally extended current collector with very low resistnace. Based on these properties, large photocurrent densities were obtained in both DSCs and iron oxide based photoelectrochemical water splitting cells. In chapter 6, nanostructured tungsten carbide synthesized by electrochemical anodization followed by heat treatment in carbon monoxide atmosphere was used as a counter electrode for DSCs employing cobalt bipyridyl redox electrolyte. The transformation of oxide into carbide was successfully and completely done due to the amorphous nature of anodic oxide materials. Moreover, superior electrocatalytic activity and photovoltaic performances were oberseved, which were attributed to the well known platinum-like electronic structure and nanoporous morphology. In chapter 7 and 8, nanostructured nickel nitride and cobalt nitride were fabricated by reactive sputtering of nickel and cobalt in nitrogen atmosphere, respectively. By physicochemical charaterizations based on electron microscopy and X-ray analyses, the films were addressed to Ni2N and CoN with cauliflower-like morphologies. When these electrodes were used as counter electrodes of quantum dot-sensitized solar cells (QDSCs), both electrodes displayed superior performances to Pt. In addition, the energy conversion efficiency of DSC employing cobalt nitride counter electrode was comparable to that with platinum. Furthermore, in QDSCs, CoN exhibited higher stability than the state-of-the-art Cu2S counter. Photocurrent densities of QDSC with CoN counter electrode exceeded that with Cu2S within 20 min, though the initial performance was higher in QDSC employing Cu2S.

Analysis of the current status of pharma CRO industry and strategies for enhancing global competitiveness in Korea

Kang, Soojin Sungkyunkwan university 2017 국내석사

Development of the Korean health literacy instrument for adults

강수진 Graduate School, Yonsei University 2011 국내박사

Deep collaborative filtering methods equipped with zero-injection

강진수 한양대학교 대학원 2019 국내석사

In the field of collaborative filtering (CF), Zero-Injection has been known to very effective in improving the accuracy of existing CF models universally. There are two main reasons for the huge improvement. First, owing to Zero-Injection, CF models can figure out some items that each user may dislike even if the items are recommended. Therefore, Zero-Injection prevents CF models from recommending such uninteresting items to a target user. Second, the rating matrix, which is generally very sparse, becomes more abundant when Zero-Injection is applied. The densified rating matrix eventually helps CF models to well understand users’ latent preferences on items more accurately. However, we noticed that Zero-Injection is very promising but has been only applied to some linear CF models, such as SVD, SVD++, and PureSVD. These linear CF models would discover only linear relationships among users and items. Motivated by the aforementioned research limitation, this thesis proposes to apply Zero-Injection to recently proposed, deep learning based CF Models, such as Collaborative Denoising Autoencoder (CDAE), AutoRec, and Collaborative Filtering framework based on Generative Adversarial Networks (CFGAN). We believe that we can achieve more accurate recommendation results if those deep learning based CF models meet Zero-Injection, since they can understand more complex relationships between users and items. Through our extensive experiments, we validate that our approach is really effective and thus providing greatly improved recommendation accuracy under not only basic top-N recommendation, but also difficult situations too, such as long-tail items recommendation, and recommendation to cold-start users. 협업 필터링 분야에서, 제로인젝션 방법은 기존의 협업 필터링 모델들의 정확도를 크게 향상시킬 수 있는 것으로 알려져 있다. 그 이유는 크게 두 가지가 있는데, 첫째는 제로인젝션을 사용함으로써 협업 필터링 모델들은 각 사용자들에게 상품을 추천해주기 전에 각 사용자들이 싫어하는 상품들에 대해 파악하고, 그 상품들이 추천되는 것을 막을 수 있다. 둘째로는 제로인젝션을 사용함으로써 기존의 매우 희소하던 평점 행렬에 풍부한 데이터가 추가로 생기는 효과를 볼 수 있다. 이렇게 풍부해진 평점 행렬은 협업 필터링 모델들이 상품들에 대한 사용자들의 상대적인 선호도를 더욱 정확하게 이해하는 데 도움이 된다. 하지만, 제로인젝션 방법은 위와 같이 그 효과와 가능성이 풍부함에도 불구하고 SVD, SVD++, PureSVD 등과 같이 단순한 선형 협업 필터링 모델들에만 적용되었다. 그러나 이러한 선형 모델들은 사용자와 상품 간의 단순한 관계밖에 알 수 없는 것으로 알려져 있다. 따라서 본 학위논문에서는, 제로인젝션을 최근에 제안된, 최신의 딥러닝 기법을 이용한 협업 필터링 모델들인 CDAE, AutoRec, CFGAN등에 적용해서 추천 정확도를 극대화시키는 연구에 대해 다루고자 한다. 제로인젝션과 딥러닝 모델들을 함께 이용한다면 딥러닝을 통해 사용자와 상품 간의 더 복잡한 관계를 유추할 수 있고, 이는 기존의 협업 필터링이 제공하던 추천의 정확도를 크게 향상시킬 수 있을 것으로 기대한다. 이를 증명하기 위해 본 논문에서는 다양한 실험을 진행하였으며, 실험을 통해서 제안하는 방법이 기존의 협업 필터링 모델들보다 높은 정확도를 제공하는 것을 보였다. 또한, 실험을 통해 단순한 top-N 추천 뿐 아니라, long-tail 상품 추천과 cold-start 사용자 추천 등 다른 어려운 상황에서도 좋은 정확도를 보이는 것을 확인하였다.

외국어로서의 한국어 반의어 교육 연구

강수진 부산대학교 2011 국내박사

<치다>류 움직씨의 다의관계 연구

강수진 부산대학교 2001 국내석사

컨테이너 터미널의 적하계획을 위한 탐색 알고리즘 개발

강진수 부산대학교 2001 국내석사

텍스트 난이도에 따른 고등학교 학생들의 읽기 전략 사용에 관한 연구

강수진 한국외국어대학교 교육대학원 2015 국내석사

A Study of High School Students' Reading Strategy Use Based on Text Difficulty Along with the importance and necessity of reading skill in language education, especially in EFL or ESL circumstances, a lot of research related reading strategies has been carried out and it has been proved that to teach and learn reading strategies is one of the most effective ways to improve reading proficiency. However, it is true that so far, in reading strategy instruction, most of research has been focused on the effects of learners’ language proficiency level on reading strategy use, without regard to text difficulty. Having this in mind, this study investigated the reading strategy use of Korean high school students based on text difficulty as well as reading proficiency level. For the research, 175 high school students were sampled as subjects and they were required to answer reading strategy questionnaires consisting of 29 reading strategies. They completed the questionnaire right after reading easy and difficult passages respectively, each followed by multiple choice comprehension questions. After that, a few of them were interviewed. The data from the questionnaires and the interview were analyzed in terms of text difficulty (easy and difficulty) and the students' level of reading proficiency (high, middle, low). The result of the study indicated that (1) considering the text difficulty, the frequency of reading strategy use differed among reading proficiency groups; (2) students used more bottom-down reading strategies under the easy text condition while they used more top-down ones under difficult text condition; (3) students used different reading strategies within the same classification according to the text condition; (4) the higher students’ level of reading proficiency was, the more reading strategies were used; (5) finally, high level students had more positive attitude to use and learn reading strategy, compared with middle and low level students. Based on the findings of this study, an important pedagogical implication or practical recommendation is suggested. It is text difficulty that could be another important independent variable which affects the students’ reading strategy use, except for learner variables. Therefore, it is necessary for English teachers to take into consideration reading strategies suitable for the text difficulty as well as learners’ level of language proficiency. Given that one of the most important goals of teaching reading is to help students become strategic and independent readers, it is hoped that these results can be applied to reading strategy instruction in the classroom. Furthermore, it is expected that the results of this study can be utilized as a suitable resource for further one.