Electrical and optical interconnects for high-performance computing

Technology scaling has enabled drastic growth in the computational and storage capacity of integrated circuits (ICs). This constant growth drives an increasing demand for high-bandwidth communication between and within ICs. In this dissertation we focus on low-power solutions that address this demand. We divide communication links into three subcategories depending on the communication distance. Each category has a different set of challenges and requirements and is affected by CMOS technology scaling in a different manner. We start with short-range chip-to-chip links for board-level communication. Next we will discuss board-to-board links, which demand a longer communication range. Finally on-chip links with communication ranges of a few millimeters are discussed.

Electrical signaling is a natural choice for chip-to-chip communication due to efficient integration and low cost. IO data rates have increased to the point where electrical signaling is now limited by the channel bandwidth. In order to achieve multi-Gb/s data rates, complex designs that equalize the channel are necessary. In addition, a high level of parallelism is central to sustaining bandwidth growth. Decision feedback equalization (DFE) is one of the most commonly employed techniques to overcome the limited bandwidth problem of the electrical channels. A linear and low-power summer is the central block of a DFE. Conventional approaches employ current-mode techniques to implement the summer, which require high power consumption. In order to achieve low-power operation we propose performing the summation in the charge domain. This approach enables a low-power and compact realization of the DFE as well as crosstalk cancellation. A prototype receiver was fabricated in 45nm SOI CMOS to validate the functionality of the proposed technique and was tested over channels with different levels of loss and coupling. Measurement results show that the receiver can equalize channels with maximum 21dB loss while consuming about 7.5mW from a 1.2V supply. We also introduce a compact, low-power transmitter employing passive equalization. The efficacy of the proposed technique is demonstrated through implementation of a prototype in 65nm CMOS. The design achieves up to 20Gb/s data rate while consuming less than 10mW.

An alternative to electrical signaling is to employ optical signaling for chip-to-chip interconnections, which offers low channel loss and cross-talk while providing high communication bandwidth. In this work we demonstrate the possibility of building compact and low-power optical receivers. A novel RC front-end is proposed that combines dynamic offset modulation and double-sampling techniques to eliminate the need for a short time constant at the input of the receiver. Unlike conventional designs, this receiver does not require a high-gain stage that runs at the data rate, making it suitable for low-power implementations. In addition, it allows time-division multiplexing to support very high data rates. A prototype was implemented in 65nm CMOS and achieved up to 24Gb/s with less than 0.4pJ/b power efficiency per channel. As the proposed design mainly employs digital blocks, it benefits greatly from technology scaling in terms of power and area saving.

As the technology scales, the number of transistors on the chip grows. This necessitates a corresponding increase in the bandwidth of the on-chip wires. In this dissertation, we take a close look at wire scaling and investigate its effect on wire performance metrics. We explore a novel on-chip communication link based on a double-sampling architecture and dynamic offset modulation technique that enables low power consumption and high data rates while achieving high bandwidth density in 28nm CMOS technology. The functionality of the link is demonstrated using different length minimum-pitch on-chip wires. Measurement results show that the link achieves up to 20Gb/s of data rate (12.5Gb/s/$\mu$m) with better than 136fJ/b of power efficiency.

Robust near-threshold QDI circuit analysis and design

The two most important digital-system design goals today are to reduce power consumption and to increase reliability. Reductions in power consumption improve battery life in the mobile space and reductions in energy lower operating costs in the datacenter. Increased robustness and reliability shorten down time, improve yield, and are invaluable in the context of safety-critical systems. While optimizing towards these two goals is important at all design levels, optimizations at the circuit level have the furthest reaching effects; they apply to all digital systems. This dissertation presents a study of robust minimum-energy digital circuit design and analysis. It introduces new device models, metrics, and methods of calculation—all necessary first steps towards building better systems—and demonstrates how to apply these techniques. It analyzes a fabricated chip (a full-custom QDI microcontroller designed at Caltech and taped-out in 40-nm silicon) by calculating the minimum energy operating point and quantifying the chip’s robustness in the face of both timing and functional failures.

Cloud computing services for seismic networks

This thesis describes a compositional framework for developing situation awareness applications: applications that provide ongoing information about a user's changing environment. The thesis describes how the framework is used to develop a situation awareness application for earthquakes. The applications are implemented as Cloud computing services connected to sensors and actuators. The architecture and design of the Cloud services are described and measurements of performance metrics are provided. The thesis includes results of experiments on earthquake monitoring conducted over a year. The applications developed by the framework are (1) the CSN --- the Community Seismic Network --- which uses relatively low-cost sensors deployed by members of the community, and (2) SAF --- the Situation Awareness Framework --- which integrates data from multiple sources, including the CSN, CISN --- the California Integrated Seismic Network, a network consisting of high-quality seismometers deployed carefully by professionals in the CISN organization and spread across Southern California --- and prototypes of multi-sensor platforms that include carbon monoxide, methane, dust and radiation sensors.

Nanostructured tungsten trioxide photoanodes for solar energy conversion

Nanostructured tungsten trioxide (WO₃) photoelectrodes are potential candidates for the anodic portion of an integrated solar water-splitting device that generates hydrogen fuel and oxygen from water. These nanostructured materials can potentially offer improved performance in photooxidation reactions compared to unstructured materials because of enhancements in light scattering, increases in surface area, and their decoupling of the directions of light absorption and carrier collection. To evaluate the presence of these effects and their contributions toward energy conversion efficiency, a variety of nanostructured WO₃ photoanodes were synthesized by electrodeposition within nanoporous templates and by anodization of tungsten foils. A robust fabrication process was developed for the creation of oriented WO₃ nanorod arrays, which allows for control nanorod diameter and length. Films of nanostructured WO₃ platelets were grown via anodization, the morphology of the films was controlled by the anodization conditions, and the current-voltage performance and spectral response properties of these films were studied. The observed photocurrents were consistent with the apparent morphologies of the nanostructured arrays. Measurements of electrochemically active surface area and other physical characteristics were correlated with observed differences in absorbance, external quantum yield, and photocurrent density for the anodized arrays. The capability to quantify these characteristics and relate them to photoanode performance metrics can allow for selection of appropriate structural parameters when designing photoanodes for solar energy conversion.

Functional genomic studies of the structure and regulation of eukaryotic transcriptomes

The main focus of this thesis is the use of high-throughput sequencing technologies in functional genomics (in particular in the form of ChIP-seq, chromatin immunoprecipitation coupled with sequencing, and RNA-seq) and the study of the structure and regulation of transcriptomes. Some parts of it are of a more methodological nature while others describe the application of these functional genomic tools to address various biological problems. A significant part of the research presented here was conducted as part of the ENCODE (ENCyclopedia Of DNA Elements) Project.

The first part of the thesis focuses on the structure and diversity of the human transcriptome. Chapter 1 contains an analysis of the diversity of the human polyadenylated transcriptome based on RNA-seq data generated for the ENCODE Project. Chapter 2 presents a simulation-based examination of the performance of some of the most popular computational tools used to assemble and quantify transcriptomes. Chapter 3 includes a study of variation in gene expression, alternative splicing and allelic expression bias on the single-cell level and on a genome-wide scale in human lymphoblastoid cells; it also brings forward a number of critical to the practice of single-cell RNA-seq measurements methodological considerations.

The second part presents several studies applying functional genomic tools to the study of the regulatory biology of organellar genomes, primarily in mammals but also in plants. Chapter 5 contains an analysis of the occupancy of the human mitochondrial genome by TFAM, an important structural and regulatory protein in mitochondria, using ChIP-seq. In Chapter 6, the mitochondrial DNA occupancy of the TFB2M transcriptional regulator, the MTERF termination factor, and the mitochondrial RNA and DNA polymerases is characterized. Chapter 7 consists of an investigation into the curious phenomenon of the physical association of nuclear transcription factors with mitochondrial DNA, based on the diverse collections of transcription factor ChIP-seq datasets generated by the ENCODE, mouseENCODE and modENCODE consortia. In Chapter 8 this line of research is further extended to existing publicly available ChIP-seq datasets in plants and their mitochondrial and plastid genomes.

The third part is dedicated to the analytical and experimental practice of ChIP-seq. As part of the ENCODE Project, a set of metrics for assessing the quality of ChIP-seq experiments was developed, and the results of this activity are presented in Chapter 9. These metrics were later used to carry out a global analysis of ChIP-seq quality in the published literature (Chapter 10). In Chapter 11, the development and initial application of an automated robotic ChIP-seq (in which these metrics also played a major role) is presented.

The fourth part presents the results of some additional projects the author has been involved in, including the study of the role of the Piwi protein in the transcriptional regulation of transposon expression in Drosophila (Chapter 12), and the use of single-cell RNA-seq to characterize the heterogeneity of gene expression during cellular reprogramming (Chapter 13).

The last part of the thesis provides a review of the results of the ENCODE Project and the interpretation of the complexity of the biochemical activity exhibited by mammalian genomes that they have revealed (Chapters 15 and 16), an overview of the expected in the near future technical developments and their impact on the field of functional genomics (Chapter 14), and a discussion of some so far insufficiently explored research areas, the future study of which will, in the opinion of the author, provide deep insights into many fundamental but not yet completely answered questions about the transcriptional biology of eukaryotes and its regulation.

Towards understanding the mixing characteristics of turbulent buoyant flows

This work proposes a new simulation methodology in which variable density turbulent flows can be studied in the context of a mixing layer with or without the presence of gravity. Specifically, this methodology is developed to probe the nature of non-buoyantly-driven (i.e. isotropically-driven) or buoyantly-driven mixing deep inside a mixing layer. Numerical forcing methods are incorporated into both the velocity and scalar fields, which extends the length of time over which mixing physics can be studied. The simulation framework is designed to allow for independent variation of four non-dimensional parameters, including the Reynolds, Richardson, Atwood, and Schmidt numbers. Additionally, the governing equations are integrated in such a way to allow for the relative magnitude of buoyant energy production and non-buoyant energy production to be varied.

The computational requirements needed to implement the proposed configuration are presented. They are justified in terms of grid resolution, order of accuracy, and transport scheme. Canonical features of turbulent buoyant flows are reproduced as validation of the proposed methodology. These features include the recovery of isotropic Kolmogorov scales under buoyant and non-buoyant conditions, the recovery of anisotropic one-dimensional energy spectra under buoyant conditions, and the preservation of known statistical distributions in the scalar field, as found in other DNS studies.

This simulation methodology is used to perform a parametric study of turbulent buoyant flows to discern the effects of varying the Reynolds, Richardson, and Atwood numbers on the resulting state of mixing. The effects of the Reynolds and Atwood numbers are isolated by looking at two energy dissipation rate conditions under non-buoyant (variable density) and constant density conditions. The effects of Richardson number are isolated by varying the ratio of buoyant energy production to total energy production from zero (non-buoyant) to one (entirely buoyant) under constant Atwood number, Schmidt number, and energy dissipation rate conditions. It is found that the major differences between non-buoyant and buoyant turbulent flows are contained in the transfer spectrum and longitudinal structure functions, while all other metrics are largely similar (e.g. energy spectra, alignment characteristics of the strain-rate tensor). Also, despite the differences noted between fully buoyant and non-buoyant turbulent fields, the scalar field, in all cases, is unchanged by these. The mixing dynamics in the scalar field are found to be insensitive to the source of turbulent kinetic energy production (non-buoyant vs. buoyant).

Chemical and Electrochemical Behavior of Graphene-Covered Silicon Photoanodes

This dissertation describes efforts over the last five years to develop protective layers for semiconductor photoelectrodes based on monolayer or few-layer graphene sheets. Graphene is an attractive candidate for a protective layer because of its known chemical inertness, transparency, ease of deposition, and limited number of electronic states. Monolayer graphene was found to effectively inhibit loss of photocurrent over 1000 seconds at n-Si/aqueous electrolyte interfaces that exhibit total loss over photocurrent over 100 seconds. Further, the presence of graphene was found to effect only partial Fermi level pinning at the Si/graphene interface with respect to a range of nonaqueous electrolytes. Fluorination of graphene was found to extend the stability imparted on n-Si by the monolayer sheet in aqueous Fe(CN)₆^3-/4- electrolyte to over 100,000 seconds. It was demonstrated that the stability of the photocurrent of n-Si/fluorinated graphene/aqueous electrolyte interfaces relative to n-Si/aqueous electrolyte interfaces is likely attributable to the inhibition of oxidation of the silicon surface.

This dissertation also relates efforts to describe and define terminology relevant to the field of photoelectrochemistry and solar fuels production. Terminology describing varying interfaces employed in electrochemical solar fuels devices are defined, and the research challenges associated with each are discussed. Methods for determining the efficiency of varying photoelectrochemical and solar-fuel-producing cells from the current-voltage behavior of the individual components of such a device without requiring the device be constructed are described, and a range of commonly employed performance metrics are explored.