Currently, about one in seven people worldwide lack access to clean water, and meeting the needs of a growing population will only further stress our existing resources (Gleick 2002). It is for these reasons that sustainable, cost effective, efficient water disinfection systems are needed. Therefore, the subject of this research is to explore a possible method of enhancing the Solar Water Disinfection (SODIS) method by using a Fresnel lens to concentrate solar energy, with the goal of attaining pasteurization temperatures to reduce holding times for disinfection. This method will be referred to as "PULSE" - Pasteurization Using a Lens and Solar Energy. The hypotheses are that pasteurization temperatures can be reached using a Fresnel lens and, secondly, that the Fresnel lens will accelerate the rate of disinfection by causing a rapid temperature rise during the solar disinfection process.

To test the first hypothesis, a Fresnel lens was used under varying weather conditions to determine its heating capabilities and conditions under which it is able to reach pasteurization temperatures. In testing the second hypothesis, disinfection rate as a function of temperature increase was established using a hot plate. Following demonstration that disinfection increases with temperature increase, an experiment utilizing PET bottles (typically employed in the SODIS method) (SODIS: How does it work? 2011) filled with water spiked with E. coli, was carried out to determine bacterial inactivation rates. Each condition (Dark, Hotplate, SODIS, and PULSE) was tested, but SODIS and PULSE were done over the summer months, when weather conditions would be similar, to ensure reproducibility of results. Temperature and E. coli numbers were monitored as a function of time. The rates of bacterial inactivation in each of the PET bottles were compared to confirm how temperature affects disinfection rates and to determine whether the Fresnel lens will speed up the solar disinfection process. The goal was to demonstrate that by achieving pasteurization temperatures, a Fresnel lens could significantly increase disinfection rates, thereby providing clean water in a much shorter amount of time.

In exploring this technology approach, this research sought to lay a foundation in which the mechanism, application, and relevance of lens-enhanced solar disinfection can be more closely studied. Better designs, better materials, and a better understanding of the fundamental principles governing disinfection can be brought to the forefront with the examination of existing designs, materials, and current data that this research looks to elucidate. Ultimately, it is towards this end in which this research is directed - to help mitigate human suffering due to something well within our capability.

Coenzyme Q (also termed ubiquinone or Q) is an electron carrier in the mitochondrial respiratory chain that functions as an essential component in energy metabolism processes. Q transfers electrons from NADH and succinate to cytochrome c, a heme protein, and also serves as a vital lipid soluble antioxidant via its redox activity, whereby it is able to undergo a radical intermediate. Thus, sufficient de novo Q biosynthesis is crucial for proper health maintenance in humans. Patients with deficiencies in Q suffer from a wide spectrum of health disorders, including kidney disease, neurodegenerative diseases, ataxia, and cardiovascular complications. Additionally, decreased Q levels have been linked to aging.

Saccharomyces cerevisiae (baker’s yeast) serves as an excellent model for studies on Q because of its powerful molecular genetics and its close homology to human Q biosynthesis and function. Q biosynthesis in S. cerevisiae (which makes Q₆ with six isoprene units, versus humans whose Q₁₀ has ten isoprene units) takes place in the mitochondria. Fourteen known mitochondrial proteins are responsible for facilitating this process—Coq1-Coq11 (Coq11 being recently discovered and, renamed from YLR290C, by Allan and Awad), Yah1 (ferredoxin), Arh1 (ferredoxin reductase), and Hfd1 (aldehyde dehydrogenase). Null coq1-coq9 mutants lack Q₆, are respiratory impaired, and sensitive to lipid autoxidation stress by polyunsaturated fatty acids. Also, many of the proteins necessary for the biosynthesis of Q₆ are associated in a high molecular weight complex that localizes in the inner mitochondrial membrane, as part of what has been termed the ‘CoQ-Synthome’.

Chapter 2 investigates potential additional protein partners of the CoQ-Synthome. Via tandem affinity purification and proteomic and mass spectrometry analysis—facilitated by dual-tagged Coq polypeptides—YLR290C was identified as a CoQ-Synthome binding partner. This protein of unknown function co-purified with Coq4, Coq5, and Coq7 and was verified via Immunoblotting and proteomic analysis. Additionally, yeast null mutants for ylr290c showed impaired ¹²C-Q₆ and ¹³C₆-Q₆ levels, indicating that this novel binding partner was needed for efficient Q biosynthesis. Altogether, the entirety of the data supported YLR290C as a novel Coq polypeptide, and hence it was renamed “Coq11”. Additionally, previous work elucidated that the stability of the CoQ-Synthome relied on the presence of the Coq8 protein, the putative kinase of the system, which was not previously shown to be a member of the complex. Moreover, several Coq polypeptides—namely Coq3, Coq5, and Coq7—are phosphorylated in a Coq8-dependent manner. Chapter 2 demonstrates the novel finding that Coq8 was captured in association with Coq6, a known member of the high molecular-weight complex, hence asserting a physical association between Coq8 and Coq6. Thus, it is now believed that Coq8 is a novel member of the CoQ-Synthome.

Given the findings contained in Chapter 2 related to the association of Coq6 with the putative kinase of the system, Coq8, the question now remains whether Coq6 is a phosphorylated Coq polypeptide, similar to Coq3, Coq5, and Coq7. Chapter 3 further investigates this potential for phosphorylation as it related to Coq6. In fact, 2-Dimensional Isoelectric Focusing Assays conducted on dual-tagged Coq6 showed that indeed it is a phosphorylated polypeptide. With this novel information regarding Coq6, purified dual-tagged Coq6 was subjected to phosphoproteomic analysis, the results of which proved inconclusive. Hence, predictive software and algorithms were employed to assist in the elucidation of potential sites of phosphorylation on Coq6. Additionally, a putative model of Coq6 was generated using the predictive software PHYRE2. Using this information, site-directed mutagenesis of Coq6 was conducted in order to investigate the phenotypes of particular sites that were predicted to be “high potential for phosphorylation sites” based on the predictive algorithms and the Coq6 model.

Chapter 4 details the extensive story of alternative splicing in yeast, and in particular, as it relates to the Coenzyme Q biosynthetic pathway. Previous findings from another collaborating laboratory showed that Coq7 contained several phosphorylated residues that were regulatory in terms of Q biosynthesis. Namely, when three particular residues were mutated to be constitutively nonphosphorylatable, Q levels were dramatically higher compared to the WT. In addition, the phosphatase responsible for the dephosphorylation of Coq7 was identified as Ptc7, a known mitochondrial phosphatase. Chapter 4 highlights the alternative splicing of PTC7, which results in two distinct and functional isoforms of this gene: Ptc7_s (spliced) and Ptc7_ns (nonspliced). Particularly, it is the spliced form which was previously cited as the mitochondrial phosphatase; the nonspliced form is, indeed, a novel protein of unknown function. Chapter 5 will further delve into the potential function of Ptc7_ns. In addition to highlighting a rare case of alternative splicing in yeast, and particularly that of a gene (PTC7) involved in Q biosynthesis, Chapter 4 also uncovers the gene that is responsible for this alternative splicing event. In particular, SNF2, which stands for sucrose-nonfermentable, is a nutrient sensing gene that modulates the splicing of PTC7 and is, thereby, able to control Q-biosynthesis in yeast. Moreover, Chapter 4 highlights how SNF2 is necessary for the metabolic shift from fermentation to respiration in yeast, and how nonconsensus splicing is initiated when SNF2 becomes absent from the cell.

Chapter 5 provides insight and perspectives on the current projects at hand and suggests future directions for each portion of the thesis projects. Additional experimental techniques and approaches are highlighted and ideas regarding how to advance each project further are proposed.

Finally, the Appendix contains the recent publication that elucidates kaempherol as a novel coenzyme Q biosynthetic precursor. When tested in comparison to various polyphenolic compounds, kaempherol supplementation showed the greatest effect on increasing Q content in mammalian cell lines. The data, overall, suggests that kaempherol is able to be metabolized into a novel ring precursor for coenzyme Q biosynthesis.

This dissertation comes as an application of receive and transmit phased array beamformers on 22 nm fully depleted silicon-on-insulator technology. They both use the concept of passive vector modulation for higher linearity. The advantage of the smaller feature size switches was used to implement the design at the Ka-band with minimized losses. The receiver IP-1dB is higher than -20 dBm with a power consumption of 60 mW at typical biasing. The biasing was varied for gain from 11 dB to 18 dB, 5 GHz bandwidth (24-29 GHz), and power consumed from 35 mW to 180 mW. The noise figure is optimized to be at 3 dB while minimizing the power consumption. The phase error standard deviation is less than 4.5 degrees for 6-bit phase shifters. The gain error standard deviation is less than 0.25 dB for the attenuators. The Ka-band transmit channel, beamformer, has an OP-1dB is 15.5 dBm with 12.5 % efficiency of the entire channel. For the QAM-64-100 MHz modulated signal, the efficiency is 7.9 % at EVM of 3.7 % and output power of 13.1 dBm. The gain is 27 dB, and the 3 dB bandwidth extends from 22.5 GHz to 26.8 GHz. The channel 6-bit phase shifters have 5.5-degree phase error standard deviations at the center frequency and 2 degrees at 26.8 GHz. The attenuator's gain error standard deviation is less than 0.4 dB within the transmitter band.

Building efficient concurrent data structures that scale to the level of GPU parallelism is a challenging problem. Solutions designed for the CPU do not scale to the thousands of cores that modern GPUs offer. In this dissertation, we show how to efficiently build a lock-based B-Tree on the GPU; then, we show how to extend the B-Tree to support snapshots and linearizable multipoint queries. Finally, we show how to compose a special-purpose data structure from general-purpose ones (e.g., hash tables) and discuss the design decisions that make composing data structures easy.

Our B-Tree supports concurrent queries (point, range, and successor) and updates (insertions and deletions). The B-Tree design use fine-grain locks to synchronize between concurrent updates, yet with clever design designs that reduce contention, the tree provide high update throughput. We show how our cache-aware B-Tree design take advantage of the cache-hierarchy of the GPU, achieving lookup throughput that exceeds the DRAM bandwidth of the GPU.

We address the critical question of understanding and providing semantics for concurrent updates and multipoint queries (e.g., range query). Using linearizability, we offer an intuitive understanding of concurrent operations. We show how to build a linearizable GPU B-Tree that uses snapshots to ensure linearizable multipoint queries. To support our snapshot B-Tree, we design a GPU epoch-based-reclamation scheme.

Finally, we show how to compose a graph data structure from general-purpose hash tables resulting in a hash-based graph data structure that excels at updates and point queries. Our graph data structure outperforms sorted-array-based solutions. The design of the data structure is a general one such that we can replace the hash tables with a B-Tree to address graph problems that require sorted adjacency lists.

Modern digital communication systems target satisfying multiple standards and different operating scenarios. Applications include read channels of data storage systems, PCIe links, FPGA I/Os, and multi-standard radios. This stimulates the research on reconfigurable analog-to-digital converters (ADCs) to serve as a key building block at the front-end of such systems. Conventional reconfigurable designs suffer from poor figure-of-merit (FoM) scaling with different resolutions, which reduces their flexibility. The limited efficiency of these techniques is attributed to the fact that they fix the ADC architecture for all configurations, whereas the optimum architecture depends on the target resolution. This dissertation introduces an architecture reconfigurable ADC that efficiently covers a wide range of resolutions by configuring the ADC to the proper architecture for each resolution. This leads to a reconfigurable ADC nearly as efficient as dedicated designs in both area and power.

Device matching is the heart of precision analog design. Basically, well matched devices come at the expense of larger die, parasitic capacitance and power consumption. Instead of only sizing the devices to achieve the desired accuracy, the intrinsic accuracy of an area efficient converter is designed worse than its resolution. Before or during chip usage, self-calibration automatically detects and corrects for elements mismatches and leads to reduced silicon area and improved yield. This dissertation investigates the efficiency of using body voltage trimming calibration for data converters. The tradeoffs of this technique are studied in details. Suggested methods have been presented to extend the use of bulk voltage trimming beyond technology limitations with minimal area and power overhead and no special technology requirements. The study shows that best results are achieved when mixing bulk trimming with other calibration techniques.

Two prototype chips are implemented in 65-nm CMOS to verify the results of this study. The first chip is a 2.5-10GS/s reconfigurable flash ADC. The ADC can be configured to work as a 3-bit, a 4-bit, or a 5-bit ADC with worst case integral nonlinearity (INL) and differential nonlinearity (DNL) of 0.48LSB and 0.35LSB respectively. The ADC achieves a figure-of-merit of 0.46pJ/conv-step and the active area is 0.13mm². The second chip is a 1.5-4GS/s "architecture" reconfigurable ADC. The ADC covers resolution range from 3b to 7b, and achieves a figure-of-merit of 0.46pJ/conv-step at 7-bit and the active area is 0.15mm².