Photovoltaic (PV), which converts sunlight into electricity, is a promising solution to the energy and environmental crisis we are facing right now. In this dissertation, we are focusing on next generation semiconductors as the photoactive materials, i.e. organic and orgainic/inorganic hybrid perovskite semiconductors, to achieve cost effective and energy efficient solar cell technology.

Organic semiconductor always shows narrow absorption compared with the conventional inorganic semiconductors, which is one of the limiting factors of the organic photovoltaic (OPV). In the second chapter, we solve this problem by employing the multi-absorber bulk heterojunction (BHJ) device architecture, in which different polymer absorbers are blended together to cover panchromatic absorption. A comparison study reveals the working mechanism of this novel device, and the material selection rule is summarized. Eventually, the 8.7% power conversion efficiency is realized in ternary BHJ solar cell. Besides the narrow absorption, the organic BHJ layer is usually very thin (~100 nm) due to its limited carrier mobility, hence the absorption is insufficient in the photoactive layer. In the third chapter, we utilize the plasmonic effect of the metal nanoparticle to enhance the absorption of the OPVs. The addition of spectrally tuned SiO2-Au nanorods leads to improved photocurrent and device efficiency. On the other hand, the narrow absorption of organic materials also creates new device possibility that traditional semiconductor can't deliver. One of the good examples is the visibly semitransparent solar cells. In the fourth chapter, we develop this idea by engineering transparent top electrode and incorporate photonic distributed bragg reflector to further improve the device efficiency without compromising the visible transparency.

Organic/inorganic hybrid perovskite semiconductor attracts incredible attentions in the past few years. It originates from the excitonic semiconductor family but delivers superior photovoltaic device performance than the typical excitonic solar cell e.g. dye sensitized solar cells and organic solar cells. However, the photophysics of this type of material is still mysterious in many aspects. In the fifth chapter, we try to understand its photophysical property by some basic characterization methods. We discover that the photo excitations in this materials varies with different crystal size. Besides, we observe the photoluminescence lifetime and intensity improve dramatically after exposing to the moisture. The improved photophysical property eventually results in greatly enhanced photovoltaic device performance. In addition to the solar cell, we successfully demonstrates the ultral high photodetectivity based on the perovskite materials, the results are included in sixth chapter.

The barrier to utilize solar generated electricity mainly comes from their higher cost relative to fossil fuels. However, innovations with new materials and processing techniques can potentially make cost effective photovoltaics. One such strategy is to develop solution processed photovoltaics which avoid the expensive vacuum processing required by traditional solar cells. The dissertation is mainly focused on two absorber material system for thin film solar cells: chalcopyrite CuIn(S,Se)2 (CISS) and kesterite Cu2ZnSn(S,Se)4 organized in chronological order.

Chalcopyrite CISS is a very promising material. It has been demonstrated to achieve the highest efficiency among thin film solar cells. Scaled-up industry production at present has reached the giga-watt per year level. The process however mainly relies on vacuum systems which account for a significant percentage of the manufacturing cost. In the first section of this dissertation, hydrazine based solution processed CISS has been explored. The focus of the research involves the procedures to fabricate devices from solution. The topics covered in Chapter 2 include: precursor solution synthesis with a focus on understanding the solution chemistry, CISS absorber formation from precursor, properties modification toward favorable device performance, and device structure innovation toward tandem device.

For photovoltaics to have a significant impact toward meeting energy demands, the annual production capability needs to be on TW-level. On such a level, raw materials supply of rare elements (indium for CIS or tellurium for CdTe) will be the bottleneck limiting the scalability. Replacing indium with zinc and tin, earth abundant kesterite CZTS exhibits great potential to reach the goal of TW-level with no limitations on raw material availability. Chapter 3 shows pioneering work towards solution processing of CZTS film at low temperature. The solution processed devices show performances which rival vacuum-based techniques and is partially attributed to the ease in controlling composition and CZTS phase through this technique. Based on this platform, comprehensive characterization on CZTS devices is carried out including solar cells and transistors. Especially defects properties are exploited in Chapter 4 targeting to identify the limiting factors for further improvement on CZTS solar cells efficiency. Finally, molecular structures and precursor solution stability have been explored, potentially to provide a universal approach to process multinary compounds.

Pre-2018 CSE ID: CS2002-0721

Metal halide perovskite material had emerged as a rising star in the photovoltaic application due to its excellent optoelectronic properties. Perovskite solar cells with a power conversion efficiency of over 25% have been demonstrated within a few years. However, the stability issues of this material still prohibit the commercialization of perovskite solar cells because of the low ion-migration activation energy, phase instability at room temperature, and sensitivity to the external environments. Attacked by the external environments through the defects in the perovskite has been considered one of the most critical issues of the instability of perovskite solar cells. Since the defects are preferred formed at the surface and grain boundaries of the perovskite, developing facile and effective strategies to passivate the defects at surface and grain boundaries is essential to enhance the stability and efficiency of perovskite solar cells.

In Chapter two, 1,3,7-Trimethylxanthine, a commodity chemical with two conjugated carboxyl groups better known by its common name caffeine, improves the performance and thermal stability of perovskite solar cells based on both MAPbI3 and CsFAMAPbI3 active layers. The strong interaction between caffeine and Pb2+ ions serves as a "molecular lock" that increases the activation energy during film crystallization, delivering a perovskite film with a preferred orientation, improved electronic properties, reduced ion migration, and greatly enhanced thermal stability. A champion stabilized efficiency of 19.8% and retain over 85% of their efficiency under continuous annealing at 85�C in nitrogen. Later on, in Chapter three, CuBr2 was introduced into the all-inorganic perovskites to control the crystallization and passivate the grain boundary defects, therefore a power conversion efficiency of over 16% was realized.

Apart from grain passivation by modulating the crystallization process of perovskites, a facile surface-induced secondary grain growth by utilizing the surface anisotropic was developed to enlarge the grain size to reduce the grain boundaries. As a result, grain size as large as 4 microns was realized through the oleylammonium treatment, the power conversion efficiency of 16.58% was achieved with 4,000-hour shelf stability.

In Chapter 5, the chemical environment of a functional group that is activated for defect passivation was systematically investigated with theophylline, caffeine and theobromine. When N-H and C=O were in an optimal configuration within the molecule, hydrogen-bond formation between N-H and I assisted the primary C=O binding with the antisite Pb defect to maximize surface-defect binding. A stabilized power conversion efficiency of 22.6% of photovoltaic device was demonstrated with theophylline treatment.

Graphene and its derivatives have shown potential to replace indium tin oxide in electronic applications because of their high theoretical conductivity and high optical transmittance. While certain factors prevent the immediate incorporation of graphene as a general transparent electrode material, particularly in solution-processible organic electronics, advances have been made in bringing this material closer to commercial applications. This work explores graphene derived from two popular synthesis methods as an electrode material in organic electronic devices, namely solution processible reduced graphene oxide and chemical vapor deposition (CVD) on copper. Chemical modification with thionyl chloride greatly enhanced the conductivity of chemically reduced graphene. When examined in detail, it was found that thionyl chloride adsorbed on the surface induces charge transfer from the graphitic base to the adsorbed molecule. Although the conductivity of this material is now almost comparable with conventional transparent electrode materials, the process requires blending with carbon nanotubes to achieve consistency in production of continuous large area films.

Chemical vapor deposition is an effective method to deposit large areas of graphene, and after optimization of the growth and transfer conditions, continuous single layer graphene with minimal defects was synthesized on copper foils. When used as a transparent anode at the bottom of an organic bulk heterojunction device, the surface roughness and hydrophobicity hinder its performance. To improve adhesion and conductivity of the electrode, a layer of the conducting polymer poly(3,4-ethylenedioxythiophene) was electropolymerized directly and conformally on the graphene surface. The surface morphology and thickness of the polymer material was tracked with the growth time, and was also found to be heavily dependent on the quality of the graphene and contact with the potentiostat. With further optimization, this process could be an important advance toward creating multilayer polymer devices on graphene electrodes.

With the rise of the Internet, social computing and numerous mobile applications have brought about a large amount of unstructured entity data, including places, events, and things. On one hand, the entity data, populated by a variety of data sources, is growing at an ever increasing speed. On the other hand, the service suppliers expect to render the entities pertinent to the things present in the information need of the users, rather than the data that merely matches the request strings. To meet these challenges, knowledge graph has been widely adopted in practice and become a fundamental building block for many commercial products from better recommendation systems to enhanced search engines.

As the knowledge graph subsumes humongous valuable content, there is an emerging need for the querying techniques that can extract and deliver the information from the data to the users. However, querying the real-life knowledge graph is not a trivial task. It has to deal with complex, unstructured graph data that can't fit a specific data model. Although there are many research efforts that aim in this direction, they typically have not addressed the real challenges: (1) In contrast to the complex and tedious graph data, the query from the end users tends to be ambiguous and underspecified. (2) It requires the semantic query understanding since the rigid string matching is often not desirable. (3) The querying system must scale to large heterogeneous graph data. This suggests the techniques that combine effective ranking and searching strategies.

In this thesis, we propose to tackle the challenges by primarily introducing a novel framework, Schemaless Graph Querying (SLQ), which supplies a flexible mechanism to querying large knowledge graphs. In essence, the query engine is built upon a set of transformation functions that automatically map keywords and linkages from a query to their matches in a graph. The end users are therefore not required to describe their queries precisely as that by most querying systems. To return the most relevant results in a fast way, SLQ also incorporates a learned ranking model that shall not rely on manually labeled data especially when the training data is difficult to obtain. This in turn give rise to the efficient top-K search techniques.

SLQ is designed inherently to better understand the user query intent. Rather than merely performing syntax based search, SLQ introduces the ontology base matching technique that can as well produce semantically related results. By leveraging an ontology index, an effective filtering mechanism is proposed to fast extract the top-K results based on the similarity quality. Moreover, a result summarization technique is invented to help the user inspect the excessive number of results: The users can enlarge and get detailed information on a summarized result view based on their search intents.

Another key problem in the context of graph querying is that this process needs to run under severe time constraints. The efficient search techniques are introduced from both the algorithm and the system aspects. In terms of the algorithm, we propose STAR, a framework for fast top-K searching for star queries. The technique is further extended to tackle general graph queries. From the system aspect, we investigate the distributed processing of large graphs in a cluster and propose the graph partitioning approaches. The technique is able to adapt in real time to changes in the query workload.

In summary, my research is introduced based on the hypotheses that as the knowledge graph becomes more available, there is a great chance of serving user queries in a better way. My contributions devote to realize this vision mainly for the applications of graph querying.