Title
stringlengths 16
413
| text
stringlengths 602
4.94k
|
|---|---|
A Peristaltic Oxygenating Mixer for Miniature Integrated Bioreactors | We have developed a mixer and corresponding fabrication process to address problems involved in the development of a miniaturized parallel integrated bioreactor array system, whose functional objectives include: (1) the ability to support cell growth of aerobic micro-organisms without oxygen limitation, (2) scalability to a large number of reactors, (3) online sensing of culture parameters, and (4) individual control over pH. In order to achieve these design objectives, we have developed a flat form factor, all PDMS (silicone elastomer), peristaltic oxygenating mixer (Figure 1), using a fabrication process that allows integrating multiple scale (100µm-1cm) and multiple depth (100µm-2mm) structures in a simple molding process. The flat form factor ensures a high surface area to volume ratio for high oxygen transfer rates, and the peristaltic action achieves in-plane homogeneous mixing within 5-20 seconds, depending on the depth of the well and actuation parameters, which is three orders of magnitude faster than lateral mixing from diffusion alone. The peristaltic action also contributes to mixing in the vertical direction, which further improves the oxygen transfer rate. The volumetric oxygen transfer coefficient (kLa) was measured by a gassing-in method [1], using an integrated platinum-octaethylporphyrine based dissolved oxygen sensor [2]. Calibrated measurements of the oxygen transfer coefficient (Figure 2) in devices of various well depths agree with theoretically expected oxygen transfer coefficients for unmixed devices. For devices mixed with various actuation frequencies, the measured oxygen transfer coefficient falls short of the theoretical values due to non-instantaneous vertical mixing. Even with non-optimized devices, preliminary results from eight simultaneous bacteria growth experiments, using four different medium compositions with online measured optical density and dissolved oxygen concentration, indicate that the oxygen transport is sufficient to maintain a greater than 55% dissolved oxygen concentration for the duration of the bioreaction. |
Microfluidic Platform for High-Density Multiplexed Biological Assays | We have developed a microfluidics-based technology that will support the ongoing need to reduce the cost and increase the capabilities of genetic testing in areas such as: population studies for the identification of inherited disease genes, more effective evaluation of drug candidates, and rapid determination of gene expression in tissues for disease management. This technology will also reduce the cost of the clinical testing of novel genetic targets related to disease risk and drug response.Specific improvements promised by this technology are the following:Provides a flexible microfluidic enabling platform for genomic, proteomic and cellular array-based assays;Can be used with current diagnostic protocols and instrumentation;Tests many samples in parallel on the same microarray; Reduces the time it takes to perform genetic tests on microarrays from hours to minutes.The elastomeric microfluidic device can print high-density DNA microarrays with dimensions as small as 10 µm. The device (Figure 1), which hermetically seals to a glass slide, patterns hundreds of DNA targets in parallel as lines on the glass surface. DNA samples are introduced into the sample entry ports and drawn along the channels, where they are exposed to and bind to the slide. After patterning, subsequent probe-target hybridization is simply achieved by running fluorescently labeled samples orthogonally over the target DNA-patterned glass slide, using a second microfluidic chip. Hybridization is achieved in less than 5 minutes; orders of magnitude faster than conventional DNA microarrays that require 16 hours for the same process. Using 10 µm wide microchannels, the hybridization spot density can be increased to over 400,000 assays per cm2. |
Polymer-Based Microbioreactors for High Throughput Bioprocessing | This project aims to develop high-throughput platforms for bioprocess discovery and developments, specifically automated microbioreactors; each with integrated bioanalytical devices, and operating in parallel. By microfabrication and precision machining of polymer material such as poly(dimethylsiloxane) (PDMS) [1] and poly(methylmethacrylate) (PMMA) [2, 3], we realize microliter (5~150 µl) microbioreactors (Figure 1) with integrated active magnetic mixing and dissolved oxygen, optical density, and pH optical measurements (Figure. 2) for monitoring nutrients and products. Reproducible batch and fed-batch [2] fermentation of Escherichia coli and Saccharomyces cerevisiae have been demonstrated in the microbioreactor. With the integration of local temperature control, cell-resistance surface modification, and pressure-driven flow at ~µL/min rates, the microbioreactor was also proven to be capable for chemostat continuous cell culture [3], which is a unique and powerful tool for biological and physiological research. As examples of bioanalysis, HPLC [1] and gene expression analysis [4] using microbioreactors have demonstrated potential applications in bioprocess developments. Parallel microbial fermentations were undertaken in a multiplexed system demonstrating the utility of microbioreactors in high-throughput experimentation [5]. A key issue for high-throughput bioprocessing is to have inexpansive and disposable microbioreactors to save operation time and labor. Current works include the integration of plug-n-pump microfluidic connections [6] in the microbioreactor system, as well as, incorporation of fabricated polymer micro-optical lenses and connectors for biological measurements to produce “cassettes” of microbioreactors. |
Cell stimulation, Lysis, and separation in Microdevices | Quantitative data on the dynamics of cell signaling induced by different stimuli requires large sets of self-consistent and dynamic measures of protein activities, concentrations, and states of modification. A typical process flow in these experiments starts with the addition of stimuli to cells (cytokines or growth factors) under controlled conditions of concentration, time, and temperature, followed at various intervals by cell lysis and the preparation of extracts (Figure 1). Microfluidic systems offer the potential to do these experiments in a reproducible and automated fashion. Figure 1 shows a schematic of a microfluidic device for rapid stimulus and lysis of cells. The fluidic systems with stimulus and lysis zones are defined using soft lithography in a poly(dimethylsiloxane) (PDMS) layer, which is then bonded to a glass slide. Temperature regulation for the two zones is achieved by using a thermo electric (TE) heater at 37oC to mimic physiological conditions during stimulation and a TE cooler at 4oC to inhibit further stimulus during lysis. Mixing in the device is enhanced by the use of segmented gas-liquid flow.To extract meaningful data from cellular preparations, current biological assays require labor-intensive sample purification to be effective. Micro-electrophoretic separators have several important advantages over their conventional counterparts, including shorter separation times, enhanced heat transfer, and the potential to be integrated into other devices on-chip. A PDMS isoelectric focusing device has been developed to perform rapid separations by using electric fields orthogonal to fluid flow (Figure 2). This device has been shown to separate low molecular weight dyes, proteins, and organelles [1]. |
Microfluidic Devices for Biological cell capture | Over the past century, cellular biology and biomechanical engineering blazed ahead in areas, such as: genome sequencing, optical probes, and high-throughput biochemical testing. For example, an increasing variety of optical imaging probes now are available for chemical and biological analyses of molecular events, physiological processes, and pathologic conditions. In contrast, cell culture techniques have remained virtually stagnant [1]. Advances in MEMS, including microfluidics and soft lithography, are providing a toolset from which to develop biological MEMS devices. In addition to miniaturizing macro biological analysis tools, techniques, and assays, microfluidic devices can utilize microscale phenomena and systems to probe single- and multi-cellular levels yielding complimentary static and dynamic data sets [1,2]. Combining these advances with more traditional microtechnology provides groundwork for developing a new generation of cell culture and analysis. Assay protocols can be run in parallel, and dynamic single-cell event information can be collected on a small or large population of cells. Cells can be probed rapidly and inexpensively in large or small quantities with small sample sizes in custom, portable microenvironments developed to more physiologically resemble in vivo conditions [2]. Modular microfluidic devices are expanding possibilities, enabling snap-in modifications for different or second-pass assays. Biological cell capture and analysis devices are shown in Figure 1 and Figure 2. Designed to capture and maintain a specific number of cells in predetermined locations, the devices yield a mechanism by which to study isolated cells or cell-to-cell interaction. Once captured, the cells can be probed and static and dynamic data extracted on the single- and multi-cellular levels. |
Manipulating solid Particles In Microfluidic systems | Microfluidic systems offer a unique toolset for the separation of microparticles and for the study of the growth kinetics of crystal systems because of laminar flow profiles and good optical access for measurements. Conventional separation techniques for particles, such as sieving, are limited to sizes larger than ~ 50 microns with large dispersion. Sorting microparticles (e.g. small crystals, single cells), requires different techniques. Dielectrophoresis is particularly attractive for microfluidic systems because large electric field gradients that drive the force are easily generated at low voltages using microfabricated electrode structures, and fixed charges are not required as in electrophoresis. It is possible to continuously separate particles of 1-10 microns with ~ 1 micron resolution (Figure 1) using dielectrophoresis with asymmetric electric fields and laminar flow (Figure 2).Microfluidic devices can also be used to study crystallization and extract kinetic parameters of nucleation and growth, and to study different polymorphs of a system. Crystallization has been achieved in some batch processes that do not have uniform process conditions or mixing of the reactants, resulting in polydisperse crystal size distributions (CSD) and impure polymorphs. Microsystems allow for better control over the process parameters, such as the temperature and the contact mode of the reactants, creating uniform process conditions. Thus, they have the potential to produce crystals with a single morphology and uniform size distribution. |
BioMEMs for control of the stem cell Microenvironment | The stem cell microenvironment is influenced by several factors, including: cell-media, cell-cell, and cell-matrix interactions. Although conventional cell-culture techniques have been successful, they offer poor control of the cellular microenvironment. To enhance traditional techniques, we have designed a microscale system to perform massively parallel cell culture on a chip. To control cell-matrix and cell-cell interactions, we use dielectrophoresis (DEP), which uses non-uniform AC electric fields to position cells on or between electrodes [1]. We present a novel microfabricated DEP trap, designed to pattern large arrays of single cells. We have experimentally validated the trap using polystyrene beads, and have shown excellent agreement with our model predictions without the use of fitting parameters (Figure1A) [2]. In addition, we have demonstrated trapping with cells by using our traps to position murine fibroblasts in a 3x3 array (Figure 1B).To control cell-media interactions, a 4x4 microfluidic parallel cell culture array has been designed and fabricated (Figure 2A). Each of the 16 culture chambers has microfluidic inlets and outlets that geometrically control the flow rate and type of media in each cell culture chamber. Reagent concentration is varied along one axis of the array, while the flow rates are varied along the other axis. The system is fabricated out of multilayer polydimethylsiloxane (PDMS) on glass and includes an on-chip diluter to generate a range of concentrations. We have cultured murine fibroblasts in a similar PDMS-on-glass environment at comparable flow rates (Figure 2B). This microfabricated system will serve as an enabling technology that can be used to control the cellular microenvironment in precise and unique ways, allowing us to do novel cell biology experiments at the microscale. |
Development of Microfluidic channels for Endothelial cell chemotaxis | Many cells have the ability to sense the direction of external chemical signals and respond by polarizing and migrating toward chemoattractants or away from chemorepellants. This phenomenon, called chemotaxis, has been shown to play an important role in embryogenesis, neuronal growth and regeneration, immune system response, angiogenesis, and other biological phenomena.[1] In addition, cell migration is also important for emerging technologies, such as tissue engineering and biochemical implants.[2] This simple behavior is apparently mediated by complex underlying diffusion and migration mechanisms that have been the focus of many studies and models. These mechanisms may be studied by various chemotactic assays. There have been several chemotaxis assay chambers developed in the past. The most widely used is the Dunn chamber.[3] The drawback of this chamber is that the cells that are squeezed between the cover glass and the chamber walls might release toxic enzymes and organells, and their effect, if any, on the viable neighborhood cells can not be easily quantified. Additionally, the linear gradient of chemokines lasts only 1 to 2 hours. The Whitesides group at Harvard designed a chemotaxis assay chamber using soft lithography.[4] They incorporated several serial mixers to generate multi-profile chemical gradient. This chamber can generate gradient with a simple linear or complex profile without limit in time, but it needs continuous flow to maintain gradient in the direction normal to gradient direction, which is not physiological. A novel chemotaxis chamber using diffusion characteristics to develop a chemotactic gradient has been developed.[5] This chamber generates a stable and linear gradient along a narrow channel without limitation in time and unnecessary physical stresses. The chamber has 2 inlet ports for 2 kinds of solutions and 1 outlet. One of the input solutions is mixed with a growth factor, and the other solution is mixed with a fluorescent dye or microspheres to verify that there is no bypass flow through the cross channel that supports diffusion. There are two main channels through which the input solutions flow and one narrow cross channel that connects the two main channels, into which a growth factor diffuses from one main channel by diffusion. |
A Microfabricated sorting cytometer | This research involves the development of a microfabricated sorting cytometer for genetic screening of complex phenotypes in biological cells (Figure 1). Our technology combines the ability to observe and isolate individual mutant cells from a population under study. The cytometer merges the benefits of both microscopy and flow-assisted cell sorting (FACS) to offer unique capabilities on a single technology platform. Biologists will be able to use this platform to isolate cells based upon dynamic and/or intracellular responses, enabling new generations of genetic screens. We are implementing this technology by developing an array of switchable traps that rely upon the phenomena known as dielectrophoresis (DEP) [1,2]. The DEP-enabled traps allow for capturing and holding cells in defined spatial locations, and subsequently, releasing (through row-column addressing) a desired subpopulation for further study. Using DEP, non-uniform electric fields induce dipoles in cells that, in turn, enable cellular manipulations. At present, no scalable DEP-based trap configuration exists that can robustly capture single cells and is also amenable to high-throughput microscopy. Such a platform requires performance characteristics that can only be met through quantitative modeling. We have undertaken much of the front-end work necessary for such a system and are continuing our efforts to realize this desired functionality.To date, we have developed second-generation trap geometries implemented in 4x4 trap arrays (Figure 2) to compare our front-end simulation-based modeling with the performance of actual devices. We have designed, fabricated, and tested both n-DEP (cells held at electric-field minima) and p-DEP (cells positioned at electric-field maxima) based configurations [3]. Our first design and test iteration demonstrated partial functionality and first-order proof of concept, while offering insight for future design improvements. We are also investigating the effects of DEP trapping on cell health and the impact that it may have on our ability to assess specific complex phenotypic behaviors. |
Vacuum sealing Technologies for Microchemical Reactors | Current portable power sources may soon fail to meet the increasing demand for larger and larger power densities. To address this concern, our group has been developing MEMS power generation schemes that are focused around fuel cells and thermophotovoltaics. At the core of these systems is a suspended tube microreactor that has been designed to process chemical fuels [1]. Proper thermal management is critical for high reactor efficiency, but substantial heat loss is attributed to conduction and convection through air, as shown in Figure 1. A straightforward solution is to eliminate the heat loss pathways associated with air by utilizing a vacuum package. We are exploring a glass frit bonding method for vacuum sealing.The leading cause of failure for a glass frit hermetic seal is large voids that are formed in the frit while bonding [2]. Progress has been made toward the optimization of presintering and bonding parameters to reduce or eliminate void formation. A vacuum package of 150 mTorr was obtained after optimization, but became leaky shortly after. An alternate packaging method using a two-step bond process, inspired by [3], was devised and developed. Recent experiments of the process, depicted in Figure 2, show that the initial box bond is capable of producing a hermetic seal. Enhancements through the incorporation of non-evaporable getters will be assessed once a vacuum package is achieved. |
Scaled-out Multilayer Microreactors with Integrated Velocimetry sensors | Microreactors are a new class of continuous reactors, with feature sizes in the submillimeter range, which have emerged over the last decade and, for a number of applications, present capabilities exceeding those of their macroscale counterparts. Unlike conventional reactors, the throughput of microreactors is increased by “scale-out,” i.e., operating a large number of identical reaction channels in parallel under equal reaction conditions. We have developed a scaled-out gas-liquid microreactor, built by silicon processing, which consists of three vertically stacked reaction layers, each containing twenty reaction channels. The reaction channels are operated in parallel from single gas and liquid feeds with a liquid volumetric throughput of 80 mL/h. Gas and liquid are introduced to the device through single inlet ports, flow vertically to each reaction layer, and are distributed horizontally to the reaction channels via individual auxiliary channels that provide a significantly larger pressure drop than that across a single reaction channel. These auxiliary channels eliminate cross-talk between reaction channels and ensure uniform flow distribution. The product mixture flows out of the device through a single outlet port. The design rationale of the scaled-out microreactor is illustrated in Figure 1. It is based on flow visualization studies and pressure drop measurements, obtained in a single channel, with the same channel geometry as the reaction channels of the scaled-out device (triangular cross section, channel width = 435 µm, channel length ~ 20 mm) [1]. A photograph of the scaled-out unit is presented in Figure 2. Flow visualization by pulsed-fluorescence microscopy across the top reaction layer reveals that the same flow regime is present in all channels. To further validate the reactor design and monitor flows during continuous operation, pairs of integrated multiphase flow regime sensors are integrated into the device [2]. Comparable slug velocities are measured across the reaction layers. |
Multiphase Transport Phenomena in Microfluidic systems | Microscale multiphase flows (gas-liquid and liquid-liquid) possess a number of unique properties and have applications ranging from use in microchemical synthesis systems to heat exchangers for IC chips and miniature fuel cells. Our work is focused on gas-liquid flows in microfabricated channels of rectangular or triangular cross section. We characterize the phase distribution and pressure drop of such flows and apply such information to a systematic design of gas-liquid microchemical reactors. The inherently transient nature of such multiphase flows provides a rich variety of flow regimes and dynamic flow properties. Characterization is done using pulsed-laser fluorescence microscopy and confocal microscopy (spinning disk and scanning), as well as by integrated flow regime sensors. Superficial gas and liquid velocities were varied between 0.01-100 m/s and 0.001-10 m/s, respectively.Particular attention is given to segmented (slug or bubbly) flows in hydrophilic channels. Figure 1a illustrates the distribution of gas and liquid in the channel. Gas bubbles are surrounded by thin liquid films (thickness ~ 1µm) at channel walls and liquid menisci in the corners. Such flows create a recirculation in the liquid segments (Figure 1b) and can, therefore, be used to efficiently mix two miscible liquids on the microscale within a length of only a few tens of the microchannel width [1,2]. We demonstrate that the transient nature of gas-liquid flows can be used to significantly improve mixing of miscible liquids compared with existing methods. After mixing is accomplished–Figure 2 (bottom) provides an illustration for mixing of two differently colored streams–the gas can be removed from the mixed liquid phase in a capillary phase separator for arbitrary velocities and flow patterns [1]. In addition to providing mixing enhancement, segmented flows narrow the distribution of residence times of fluid elements in the liquid phase, as compared to single-phase flows [1]. A narrower residence time distribution is particularly essential for particle synthesis on a chip. |
Integrated Microreactor system | Individual microreactors have been fabricated for many different chemical reactions, but the development of microreaction technology will require combining separation with microreactors to enable multi-step synthesis. The realization of integrated microchemical systems will revolutionize chemical research by providing flexible tools for rapid screening of reaction pathways, catalysts, and materials synthesis procedures, as well as, faster routes to new products and optimal operating conditions. Moreover, such microsystems for chemical production will require less space, use fewer resources, produce less waste, and offer safety advantages. The need for synthesizing sufficient quantities of chemicals for subsequent evaluation dictates that microchemical systems are operated as continuous systems. Such systems require fluid controls for adjusting reagent volumes and isolating defective units. The integration of sensors enables optimization of reaction conditions, as well as, the extraction of mechanistic and kinetic information.We are developing integrated microchemical systems that have reactors, sensors, and detectors with optical fibers integrated on one platform. New approaches for connecting modular microfluidic components into flexible fluidic networks are being explored. Real-time control of reaction parameters, using online sensing of flowrate, temperature, and concentration, allows for precise attainment of reaction conditions and optimization over a wide range of temperatures and flow-rates. The multiple microreactors on the system can be used together to give higher throughputs or they can be used independently to carry out different reactions at the same time. Figure 1 shows a schematic of an integrated microreactor platform along with an early stage microreactor “circuit board” [1]. |
Micro Gas Analyzer | The US Department of Defense is currently interested in developing the technology to sense, in real time, deployable agents used in chemical warfare. The Micro Gas Analyzer Project (MGA) is the result of this interest, and aims to develop a portable sensor of wide rage and robustness. Current state-of-the-art technology involves bulky equipment (not portable), high power consumption due to the use of thermionic sources and impact ionization mechanisms, high voltage (in the kilovolt range), and long processing times. Thus, the project has a number of key technological challenges, such as the enhancement of the state-of-the-art sensitivity and specificity capabilities, power consumption reduction, and portability, while keeping the processing time below two seconds. The MGA is composed of an ionizer (a CNT field ionization array / CNT field emission array), a mass filter (a micro quadrupole mass spectrometer -µQMS), an ion counter/multiplier, an electrometer/mass detector, and a pumping system (passive – absorption pump/active – piezoelectric pump). A schematic of the MGA system is shown in Figure 1. The goal is to make low vacuum (in the millitor range), ionize the species inside the gas using the CNT arrays, filter them with the quadrupole, and then, sense them with the electrometer. The project team is composed of MIT (Ionizer, µQMS, µPump, Valves), University of Texas (Ionization, µPump), Cambridge University (Ion Counter), and Raytheon/CET (System Integration). |
Micro Quadrupole Mass spectrometer | One of the subsystems of the Micro Gas Analyzer Project is a mass filter. The purpose of this filter is to select the kind of species that will be sensed downstream by the electrometer. A microfabricated quadrupole mass filter array is being developed for this purpose where a confining potential sorts the unwanted species (Figure 1). Both high sensitivity and high resolution are needed over a wide range of ion mass-to-charge ratios, from 20 to 200 atomic mass units, to achieve the versatility and resolution that are intended for the program. In order to achieve the high resolution and sensitivity, multiple micro-fabricated quadrupoles, each with specific geometrical parameters, are operated in conjunction with each other. From a theoretical point of view, the Mathieu equations describe the dynamics of a particle inside the quadrupole. These equations predict a series of stability regions (Figure 2). Each stability region has its own strengths, such as: less power consumption, less operational voltage, or more sensitivity. For example, lower stability zones are used to improve ion transmission, whereas, higher stability zones are used to improve the selectivity of the filter. Therefore, we plan to explore the stability regions of the Mathieu equations to optimize our design. Two sets of variable voltage sources are needed for the mass filter to operate properly, with voltages ranging between 20 and 200 V, at frequencies of 250 and 500 MHz. We plan to try three different approaches to build the device: LIGA (a german acronym for the process that generates high aspect ratio metallic structures), rods assembled using micro-fabricated deflection springs [1], and rod mounts made with KOH [2]. The device has a cross-sectional area of 20 mm2. The aperture of the individual quadrupoles ranges from 10 to 100 microns. |
Design Tools for Bio-Micromachined Device Design | Using micromachining for biological applications requires complicated structures such as mixers, separators, preconcentrators, filters, and pumps; and these elements are used to process biomolecules or biological cells. To accelerate the design of these complicated devices, new tools are needed that can efficiently simulate mixing and particle or cell motion in complicated three-dimensional flows. In addition, for microfluidic devices intended for use in molecular separation, the length scales are such that noncontinuum fluid effects must be considered, and therefore, hybrid approaches that combine molecular and continuum models must be developed. Finally, the wide variety of structures being developed implies that generating models for system-level simulation will require efficient simulation combined with automated model extraction [3]. Our recent work in addressing these problems includes: the development of efficient time integration techniques for cells in flow [1], techniques for accurately extracting diffusion constants from measurements [2], and efficient techniques for extracting models from detailed simulations [4]. |
Microfabricated solid-Oxide Fuel cell systems | Solid-Oxide Fuel Cells (SOFCs), employing ceramic electrolytes, are a promising alternative to low-temperature PEM (proton-exchange membrane) fuel cells for portable power applications. The use of an oxygen-ion conducting electrolyte, operating at high temperatures, offers the potential for internal reforming of a variety of fuels, with improved tolerance to competitively adsorbing species at the anode (e.g. CO); thus, removing the need for pretreatment stages for conversion of hydrocarbon fuel to high-purity hydrogen. However, the appropriate thermal management of this high-temperature fuel cell system is required to achieve an energy-efficient device. A chip-scale micromembrane architecture has been developed for thermally efficient thin-film applications1 and has been successfully demonstrated for hydrogen separation via ultra-thin palladium films. Resistive heaters placed directly upon a thermally-isolated membrane allow for rapid heating and cooling of the supported thin film at a minimum expenditure of energy. In addition, the mechanical strength provided by the micromembrane support allows the use of sub-micron films for significant improvement in ion permeability. For these reasons, the micromembrane architecture has been investigated for SOFC development. The extension of this technology is achieved, utilizing a silicon-nitride girder-grid support system to mechanically reinforce the solid-oxide thin films (Figures 1 and 2).Efforts include: the determination of optimal free-standing fuel cell stack dimensions, integration of individual stacks into a reinforced membrane structure, design of current collectors, and electrical performance tests of fabricated devices. Stability tests of free-standing membranes of varying length scales and aspect ratios are performed for a variety of fuel cell stacks and individual stack layers, with results compared to mechanical models of layered free-standing films. The resulting information is incorporated into the design of a silicon-nitride reinforced free-standing membrane architecture. Lastly, microdevice testing stations allow for performance studies of prototype microdevices. |
Catalytic Micromembrane Devices for Portable High-Purity Hydrogen Generation | The development of portable-power systems employing hydrogen-driven fuel cells continues to garner significant interest in the scientific community, with applications ranging from the automotive industry to personal electronics. While progress has been made in the development of efficient hydrogen-storage devices, it is still preferable for portable-power systems to operate from a liquid fuel with a high energy density (e.g., methanol, ammonia). This necessitates the integration of a hydrogen generator capable of converting stored fuels to hydrogen to drive the fuel cell. Previous research has focused upon the development of novel catalysts and autothermal microreactor designs for efficient conversion of liquid fuels (e.g. methanol, ammonia) into hydrogen for use by a polymer-electrolyte fuel cell [1]. Additionally, micromembrane devices (Figure 1) have been developed for purification of the resulting hydrogen stream to remove impurities (e.g. CO) that adversely affect fuel cell performance [2]. Our current research aims to integrate (i) catalyst design, (ii) autothermal microreformer design, and (iii) micromembrane technology to realize microscale chemical systems capable of producing high-purity hydrogen for fuel cell operation. By combining microfabrication techniques for generation of micromembrane devices with wet-chemical deposition methods for a variety of catalysts, multiple membrane reactor applications for hydrogen generation can be realized, taking full advantage of superior mass transport and film permeabilities achievable at the microscale. Results obtained for LaNi0.95Co0.05O3 perovskite catalysts integrated with 23 wt% Ag-Pd membranes (Figure 2) demonstrate promising high-purity hydrogen yields at low methanol feed compositions, and demonstrate the applicability of catalytic membrane reactors effected at the microscale for efficient production of high-purity hydrogen. Resulting microdevices are directly applicable as part of an integrated portable-power system. |
Thermal Management in Devices for Portable Hydrogen Generation | As power requirements of portable electronic devices continue to increase, the development of an efficient portable power generation scheme has remained an active research area. Specifically, hydrogen-driven fuel cells have received significant attention. This work focuses on microreaction technology for the conversion of fuel to electrical power. Emphasis has been placed on developing microreactors for high-purity hydrogen production. Critical issues in realizing high-efficiency devices capable of operating at high temperatures have been addressed: specifically, thermal management, the integration of materials with different thermophysical properties, and the development of improved packaging and fabrication techniques.A microfabricated suspended-tube reactor (Figures 1, 2) has been developed for efficient combustion and reforming of chemical fuels.[1] The reactor, designed specifically to thermally isolate the high-temperature reaction zone from the ambient, consists of thin-walled U-shaped silicon nitride tubes formed by deep reactive ion etching (DRIE) and subsequent nitride deposition via chemical vapor deposition (CVD). Thin-film platinum resistors are integrated into the reactor for heating and temperature sensing. Detailed thermal characterization demonstrates reactor operation up to 900ºC and quantifies heat losses. Additionally, this high-temperature microcombustor is applicable for thermophotovoltaic generation. A new fabrication scheme for the suspended-tube reactor incorporates wet potassium hydroxide (KOH) etching, an economical and time-saving alternative to DRIE]. In this design, pre-fabricated thin-walled glass tubes replace the silicon nitride tubing to provide inlet and outlet conduits. The thermal conductivity of the resulting tubes is 50% lower than that of silicon nitride. Hence, this technique allows for the incorporation of robust tubing, while maintaining thermal efficiency. |
Materials and structures for a MEMs solid Oxide Fuel cell | Microfabricated solid oxide fuel cells are currently being investigated for portable power applications requiring high energy densities [1, 2]. Reducing the thickness of fuel cell stack materials improves the electrochemical performance versus traditional devices. This motivation for thinner structures, combined with significant temperature excursions during processing and operation (~600 – 1000 °C), presents the thermomechanical stability of such membranes as a major challenge. A buckled electrolyte/SiN thin film is shown in Figure 1. The prediction and management of structural stability (buckling) and failure require accurate knowledge of many parameters including: thermomechanical properties, residual stress, and fracture strength.Our group has characterized the residual stress and microstructure of the electrolyte layer of the fuel cell stack. Residual stress in sputter-deposited yttria stabilized zirconia (YSZ) thin films (5nm – 1000nm thickness), as a function of deposition pressure and substrate temperature, has been completed [3]. The results indicate variations in intrinsic stress from ~0.5GPa compressive to mildly tensile (~50 MPa) (Figure 2). Changes in microstructure are subsequently characterized using X-ray diffraction of as-deposited and annealed films and correlated with relevant mechanisms/models of residual stress evolution. Frameworks for using such residual stress data to design mechanically stable membranes for µSOFC devices have also been developed.Current research areas include: continued microstructural and residual stress characterization under thermal cycling, elastic/fracture properties characterization, design and fabrication of thermomechanically stable fuel cell stacks, exploration of proton conducting solid oxide thin films for lower-temperature operation, investigation of the mechanical properties of anode and cathode materials, and nonlinear modeling of film postbuckling and failure. |
Microfabricated Proton-conducting solid Oxide Fuel cell system | Owing to their high efficiency and energy density, miniaturized fuel cells are an attractive alternative to batteries in the mW-W power generation market for portable consumer and military electronic devices [cf. 1-3]. Hydrogen is being actively considered as a fuel for power generation. It can be supplied either by storage devices or its in-situ generation using reformers. However, safety and reliability issues persist with current storage choices, such as zeolites and carbon nanotubes [4]. For these reasons, fuel cells based on direct fuel reforming are advantageous. The processes typically involve either high temperature reforming of fuel to hydrogen combined with a low temperature Proton exchange membrane (PEM) fuel cell, which implies significant thermal loss. Alternatively, fuel reforming can be combined with solid oxide fuel cells capable of operating at high temperatures. Typical components of a solid oxide fuel cell include electrodes and an electrolyte. Typically ZrO2, CeO2, and LaGaO3, which are oxide ion conductors are used as separator materials [5]. However, one of the disadvantages of these materials is the need for operation at high temperatures (~700oC). These operating temperatures, in turn, lead to associated problems of materials compatibility and low tolerance with respect to variations in operating conditions. As an alternative, proton conducting solid oxide membranes, typically alkaline earth metal substituted perovskites, such as BaCeO3, SrCeO3, and BaZrO3, exhibit high protonic conductivity even at 400oC [6]. In the current research, we explore the possibility of fabricating a fuel cell using these low temperature electrolytes. Previous work on Pd-based membranes on MEMS-supported membranes indicates that hydrogen yields up to 93% can be achieved for methanol using LaNiCoO3 anode catalyst at 475oC. We plan to extend this concept further to prepare a complete fuel cell assembly and test its performance. |
Thermophotovoltaic (TPV) MEMs Power Generators | Batteries have, for a number of years, not kept up with the fast development of microelectronic devices. The low energy densities of even the most advanced batteries are a major hindrance to lengthy use of portable consumer electronics, such as laptops, and of military equipment that most soldiers carry with them today. Furthermore, disposing of batteries constitutes an environmental problem. Hydrocarbon fuels exhibit very high energy densities in comparison, and micro-generators converting the stored chemical energy into electrical power at even modest levels, are, therefore, interesting alternatives in many applications. This project focuses on building thermophotovoltaic (TPV) micro-generators, in which photocells convert radiation from a combustion-heated emitter, into electrical power. TPV is an indirect conversion scheme that goes through the thermal domain and therefore, does not exhibit very high efficiencies (10-15% max). However, because of its simple structure and because the combustor and photocell fabrication processes do not need to be integrated, the system is simpler to micro-fabricate than other generator types (e.g. thermoelectric systems and fuel cells). It is also a mechanically passive device that is virtually noiseless and less subject to wear than engines and turbines. In this TPV generator, a catalytic combustor, the suspended micro-reactor (Figure 1) is heated by combustion of propane and air, and the radiation emitted is converted into electrical energy by low-bandgap (GaSb) photocells (Figure 2). Net power production of up to 1 mW has been achieved [1], constituting a promising proof of concept. Work is underway to build a new micro-reactor more suited for the needs of TPV than the original design. |
Thermoelectric Energy conversion: Materials and Devices | Thermoelectric devices based on Peltier effect and Seebeck effect use electrons as a working fluid for energy conversion. These solid-state energy conversion devices have important applications in refrigeration and electrical power generation. Our work follows two directions: nanostructured materials and microdevices. The efficiency of thermoelectric devices is characterized by the nondimensional thermoelectric figure of merit 2ZT S T / k= σ, where S is the Seebeck coefficient, σ the electrical conductivity, and k the thermal conductivity of their constituent materials, and T is the average device temperature. Identifying materials with a large ZT has been challenging because of the interdependency of those three properties. With both quantum size effects on electrons and classical size effects on phonons, nanostructures provide an alternative way to engineer thermoelectric properties.1,2 Our current effort is focused on designing, synthesizing, and characterizing nanostructures in bulk form that can be produced for mass applications. Figure 1 illustrates ballistic phonon transport in a unit cell of a nanocomposite, which leads to low thermal conductivity.3 We are also working on fabricating micro thermoelectric devices, first using thin film devices such as SiGe alloy and Si-Ge superlattices,4 and more recently on thick films to reduce parasitic heat losses.5 In addition, we are also exploring novel microdevice configurations that can improve energy conversion efficiency, by utilizing the hot electron concepts.6,7 |
Far-Field spectral control and Near-Field Enhancement of Thermal Radiation Transfer for Energy conversion Applications | The performance of thermophotovoltaic (TPV) energy conversion systems is greatly affected by the radiation characteristics of the thermal emitter. Ideally, one would want a selective emitter with high emissivity above the band gap and low emissivity below the band gap. Various approaches have been proposed to fabricate effective selective emitters with 2D or 3D photonic crystals, which involve considerable intricate microfabrication. Instead, we have proposed a simpler-to-fabricate 1D structure that exhibits many of the features of its 2D and 3D counterparts [1]. The key has been to use ultra thin metallic films arranged as a periodic multilayer stack with a suitable non-absorbing dielectric material in-between. Figure 1 shows the numerical computation of the total hemispherical emissivity of two such structures as a function of wavelength. In addition to improving the selective emission of thermal radiators, we are also exploring near field effects to improve the energy density and efficiency of thermal-to-electric energy conversion devices. Electromagnetic surface waves, like surface phonon polaritons or surface plasmon polaritons, can increase the energy transfer by two or three orders of magnitude compared to the near-field enhancement between materials that do not support such surface waves. Our work has shown that such enhancements in thermal radiative transfer can not only increase the power density and efficiency of TPV devices [2] but can also contribute to the improvement of thermoelectric devices [3]. We are also exploring a new TPV device structure involving interdigitized hot-and-cold fingers with increased surface area, built-in photon recycling, and potentially built-in spectral control [4]. Experimental work involving microfabrication and device testing is in progress. |
Development of a High Power Density Microscale Turbocharger | A microscale turbocharger has been fabricated as part of a program to develop a microfabricated gas turbine generator to serve as a battery replacement with seven times the energy density of today’s best batteries. The turbocharger will evolve into the gas turbine generator with minimal fabrication process changes. The turbocharger lacks an electric generator, and its turbine and compressor flow paths are independent; otherwise, the two devices are virtually identical. The turbocharger is a test vehicle for developing fabrication processes and turbomachinery/bearing technology. The turbocharger is formed by fusion bonding six silicon wafers. The hatched structure in Figure 1 is the rotor, which is free to spin within the device on hydrostatic gas bearings. The turbocharger has a design rotation rate of 1.2 million rpm and a design compressor pressure ratio of 2.2.Journal bearing dimensional control is a key challenge: 15 +/- 0.75 µm in width and 330 +/- 5 µm in depth. The bearing width tolerance, which is half that of previous devices in this program, is achieved through refinements in the etch recipe as well as modifications to the masking material profile. The masking material must be carefully controlled because of its finite etch rate and the effects of sidewall-passivation-layer erosion from ions deflected by the resist slope. The journal bearing specification is met on device wafers with a yield of more than 60%. Another challenge for this device is obtaining a rotor blade height uniformity of about 1%, which is critical for low levels of imbalance in the rotor.A turbocharger has been operated to a rotation rate of 480,000 rpm, which is equivalent to a tip speed of 200 m/s (450 miles per hour). Figure 2 shows the measured compressor pressure ratio for two runs of the same device with different throttle settings. The compressor achieved a pressure ratio of 1.21 with a flow rate of 0.14 g/s at its top speed. The measured pressure and flow characteristics are consistent with the design models for this device. |
A MEMS Electroquasistatic Induction Turbine-Generator | Presented here is a microfabricated electroquasistatic (EQS) induction turbine-generator that has generated net electric power. A maximum power output of 192 µW was achieved under driven excitation. We believe that this is the first report of electric power generation by an EQS induction machine of any scale in the open literature. This work forms part of a program at MIT to fabricate a MEMS-scale gas turbine-generator system. Such a system converts the enthalpy of combustion of a hydrocarbon fuel into electric power. For even modest efficiency levels of the gas turbine engine cycle (10-15%), a small gas turbine would be a portable energy source with higher energy density than the best batteries available [1]. In MIT's device, this small engine provides the shaft power needed to drive a small electric generator. Although magnetic machines are preferred at large scales, EQS machines become attractive at small scales, primarily because very small airgaps between the rotor and stator allow higher breakdown electric fields of approximately 108 V/m. The generator comprises five silicon layers (Figure 1) fusion bonded together at 700oC. The stator is a platinum electrode structure formed on a thick 20 µm recessed oxide island. The rotor is a thin film of lightly doped polysilicon also residing on an oxide island, which is 10 µm thick. We also present a generalized state-space model for an EQS induction machine that takes into account the machine and its external electronics and parasitics. This model correlates well with measured performance, and was used to find the optimal drive conditions for all driven experiments. Figure 2 shows the results of an experiment under driven excitation. In this particular experiment, 108 µW was generated at 245krpm. Good correlation with the models is observed. In other experiments, self-excited operation was attained. In this case, the generator self-resonates and generates power without the use of any external drive electronics [3]. |
Multi-Watt Electric Power from a Microfabricated Permanent-Magnet Generator | Presented here are the design, fabrication, and characterization of three-phase permanent magnet (PM) machines that convert 2.3 W of mechanical power and deliver 1.1 W of DC electrical power to a resistive load at a rotational speed of 120,000 rpm. Such microgenerators are an important system-level component of compact MEMS-based power sources, such as combustion-driven or air-driven microengines [1].The generators are three-phase, eight-pole, synchronous machines, each consisting of a surface-wound stator (Figure 1) and a multi-poled PM rotor (Figure 2(a)). The stator uses three Cu windings that are dielectrically isolated from a 1-mm thick NiFeMo (Supermalloy) substrate by a 3 µm spin-on-glass layer and/or 5 µm polyimide layer. The coils were fabricated using a two-layer electroplating process [2]. They were measured to be 80-120 µm thick and 50-550 µm in width. The microfabricated coils, with their small inter-conductor gaps and variable width geometry, are the key for enabling high power output. The rotor contains an annular SmCo PM and a ferromagnetic FeCoV (Hiperco50) backiron, each 9.525 mm OD, 3.175 mm ID, and 500 µm thick. The SmCo PM and FeCoV backirons were, then, assembled and glued into a pre-formed PMMA cup, which was fit onto a 1.6 mm shaft (Figure 2(b)).For characterization, a high-speed spinning rotor test stand, incorporating an air-turbine driven spindle, was constructed. The stator was positioned under the rotor using an xyz-micropositioner, which permitted precise (± 5 µm) adjustment of the air gap. A three-phase step-up transformer (1:6 turn ratio) and Schottky diode bridge were used to rectify the output voltage for DC power generation across a load resistor. The power data for the 2-turn/pole machine shows a quadratic dependence on speed for a fixed load (Figure 2(c)) and typical power transfer dependence for varying loads (Figure 2(d)), with a maximum demonstrated power of 1.1 W (2.9 MW/m3 power density). |
High-speed Micro-scale Gas Bearings for Power MEMs | The high-speed micro hydrostatic gas journal bearings used in the high-power density MIT micro-engines are of very low aspect ratio, with a bearing length-to-diameter ratio of less than 0.1, and are running at surface speeds of order 500 m/s. These ultra-short high-speed bearings exhibit whirl instability limits and dynamic behavior very different from conventional hydrostatic gas bearings. The design space for stable high-speed operation is confined to a narrow region and involves singular behavior [1]. The narrow design space together with the limits on achievable fabrication tolerance that can be achieved in the silicon chip manufacturing technology severely affects journal bearing operability and limits the maximum achievable speed of micro turbomachinery. The hydrostatic gas thrust bearings are located near the center of the rotor, and play a vital role in providing axial support for the rotor. The thrust bearing geometry is designed to provide the required axial and tilting stiffness, and ensures stable thrust bearing operation at high-speed [2].Our technical approach involves the combination of numerical simulations, experiment, and simple, first principles based on modeling of the gas journal and gas thrust bearing flow fields and the rotordynamics. A novel variation of the axial-flow hydrostatic micro-gas journal bearing concept is introduced that yields anisotropy in bearing stiffness [3]. By departing from axial symmetry and introducing biaxial symmetry in hydrostatic stiffness (Figure 1), the bearing's top speed is increased and fabrication tolerance requirements are substantially relieved, making more feasible extended stable high-speed bearing operation. An existing analytical hydrostatic gas journal bearing model [4] is extended and modified to guide the journal bearing design with stiffness anisotropy. In addition, a novel micro gas thrust bearing model is established. High-speed experimental spin tests were conducted in several micro-bearing test devices, and all 11 test devices were spun to high-speed, achieving an average rotor speed of 720,000 rpm. Figure 2 depicts a typical test run, and shows good agreement between the newly established bearing theory and the measurements. |
Piezoelectric Micro Power Generator (PMPG): A MEMS-based Portable Power Device | A thin-film lead zirconate titanate Pb(Zr,Ti)O3 (PZT), MEMS energy-harvesting device is developed to enable autonomous sensors for in-service integrity monitoring of large scale infrastructures. It is designed to resonate at specific frequencies from external vibrational energy sources, thereby creating electrical energy via the piezoelectric effect. The corresponding energy density of the 1st prototype is 0.74 mW-h/cm2, which compares favorably to lithium ion batteries. [1] Current efforts are focused on improving the harvest efficiency of the device. A geometric optimization of the cantilever design is made to suppress damping contributions from air and structural dissipation. Additionally, a serpentine cantilever has been designed to achieve a low resonant frequency structure. The dominant contributors to low Q factor at the MEMS scale are air damping and internal structure damping. For 2nd generation PMPG [3], we have optimized the cantilever shape to minimize the damping effect. Analytical modeling of PMPG predicts a 77% decrease of the damping coefficient of a new PMPG device.[4] This reduced damping coefficient enables 4.3 times larger resonance amplitude of the cantilever structure and 10.2 times larger maximum strain of the PZT layer. As a result, power density increases up to 1850% of the old PMPG device at the same footprint. We also designed a serpentine cantilever to achieve a low resonant frequency structure, as well as, a low damping effect, when it resonates. (Figure 2)PMPG has been integrated with a commercial wireless sensor, Telos, to simulate a self-powered RF temperature monitoring system. Such devices will play an important role in remote sensing network applications. Telos on average consumes 350µJ for 38 ms per measurement. Since PMPG offers limited power, a storage capacitor and a power management module are implemented to power the node at discrete time intervals. |
MEMs Piezoelectric Ambient Vibration Energy Harvesting for Wireless sensors | Recently, numerous investigations have focused on the development of distributed wireless sensor node networks. Power for such devices can be supplied through harvesting ambient environmental energy, available as: mechanical vibrations, fluid motion, radiation, or temperature gradients [1]. Envisioned applications include: building climate control and warehouse inventory control, identification and personalization (RFID tags), structural health monitoring (aerospace and automotive sectors), agricultural automation, and homeland security.Advances in “low-power” DSP’s (Digital Signal Processors) and trends in VLSI (Very Large Scale Integration) system design have reduced power requirements to 10’s-100’s of µW. These power levels are obtainable through piezoelectric harvesting of ambient vibration energy. Current work focuses on harvesting this energy with MEMS resonant structures. Coupled electromechanical models have been developed to predict the electrical and mechanical performance obtainable from known low-level ambient vibration sources. These models have been validated by comparison to prior published results [2] and tests on a MEMS device. A non-optimized, uni-morph beam prototype (Figure 1) has been designed and modeled to produce 30 µW/cm3 [3]. A MEMS fabrication process for a prototype device is presented based on past work at MIT [4]. Dual optimal frequencies with equal peak powers and unequal voltages and currents are characteristic of the response of such coupled devices when operated at optimal load resistances (Figure 2).Future work will explore active sources, such as: aircraft skin for harvestable power, fabrication and testing of the uni-morph prototype beam, and optimization of device configurations for aerospace structural health monitoring applications. System integration and development, including modeling the power electronics, will be included. |
Micro chemical Oxygen Iodine Lasers (MicrocOIL) | Conventional Chemical Oxygen Iodine Lasers (COIL) offer several important advantages for materials processing, including short wavelength (1.3 µm) and high power. However, COIL lasers typically employ large hardware and use reactants relatively inefficiently. This project is creating an alternative approach called microCOIL. In microCOIL, most conventional components are replaced by a set of silicon MEMS devices that offer smaller hardware and improved performance. A complete microCOIL system includes: microchemical reactors, microscale supersonic nozzles, and micropumps. System models incorporating all of these elements predict significant performance advantages in the microCOIL approach [1]. Initial work is focused on the design, microfabrication, and demonstration of a chip-scale Singlet Oxygen Generator (SOG): a microchemical reactor that generates singlet delta oxygen gas to power the laser. Given the extensive experience with microchemical reactors over the last decade [2-4], it is not surprising that a microSOG would offer a significant performance gain over large scale systems. The gain stems from basic physical scaling; surface to volume ratio increases as the size scale is reduced, which enables improved mixing and heat transfer. The SOG chip being demonstrated in this project employs an array of microstructured packed-bed reaction channels interspersed with microscale cooling channels for efficient heat removal. Figure 1 shows a schematic top view of the microSOG chip, including inlets and outlets for the reactant and product flows, and packed-bed reaction channels. Figure 2 shows a schematic diagram of stacked microSOG chips, micronozzles, and micropumps forming a complete microCOIL system. |
Linear Array of Electrospray Micro Thrusters | Electrospray thrusters are electrostatic accelerators of charged particles that use the electrohydrodynamic effect known as Taylor cone as propulsive effect [1]. These particles could be charged droplets, solvated ions, or a mix of the two. Since the new advances in electrospray technology that occurred in the late 1980s [2], the field of electrospray propulsion has experienced a renaissance, specifically aiming to provide efficient high-tunable precision low-thrust engines for micro-satellites and high accuracy astrophysics missions [3]. The MIT Space Propulsion Laboratory and the Microsystems Technology Laboratories are currently pursuing the development of a micro-fabricated electrospray emitter array for space propulsion. The project is developing in parallel two radically different concepts, a pressure-fed engine, and a surface tension-fed engine. This abstract reports the design, fabrication, and experimental characterization of a micro-fabricated, internally-fed linear array of electrospray emitters (Figure 1). This work demonstrates the feasibility of high clustering of electrospray emitters. The linear array is composed of 1 plenum, 12 manifolds, and 240 emitters. The emitters are sharpened to reduce the startup voltage. The electrodes are micro-fabricated with conductive paths made of tungsten and electrical insulation provided by vacuum gaps 350 µm wide and 10 µm thick PECVD silicon oxide. The electrodes are hand-assembled to the engine using a novel technique that relies on clusters of micro-fabricated springs [4]. This assembly scheme allows us to have two independent process flows for the electrodes and the engine hydraulics. The emitter-to-emitter separation is 130 µm, and the hydraulic diameter is 12 µm. The length of each channel is 15 mm. The engine uses highly doped formamide as propellant, with electrical conductivity in the 0.3 – 3.0 S/m range. The electrospray array operates in the single Taylor cone droplet emission regime, and it requires about 2000 V to become activated. The engine implements the concept of hydraulic and electrodynamic flow rate matching to achieve electrical control. Current versus flowrate characteristics of the engine are in agreement with a well-established reduced order model (Figure 2). Experimental data, demonstrating the low divergence of electrospray emitter arrays operated in the single Taylor cone, is in qualitative agreement with a reduced order mode that assumes the absence of a thermalized tail in the plume. |
Planar Array of Electrospray Micro Thrusters | Electrospray thrusters are electrostatic accelerators of charged particles using the electrohydrodynamic effect known as Taylor cone to generate thrust [1]. These particles could be charged droplets, solvated ions, or a mix of the two. Since the new advances in electrospray technology that occurred in the late 1980s [2], the field of electrospray propulsion has experienced a renaissance, specifically aiming to provide efficient high-tunable precision low-thrust engines for micro-satellites and high accuracy astrophysics missions [3]. The MIT’s Space Propulsion Laboratory and the Microsystems Technology Laboratories are currently pursuing the development of a micro-fabricated electrospray emitter array for space propulsion applications. The project is developing, in parallel, two radically different concepts, a pressure-fed engine and a surface tension-fed engine. This abstract reports the design, fabrication, and experimental characterization of a hybrid macro-fabricated/micro-fabricated, externally fed planar array of micro-fabricated electrospray emitters with macro-fabricated electrodes (Figure 1). An externally-fed engine has a number of advantages compared to the other implementations reported in the literature. For example, the engine lacks a static pressure difference between the plenum and the emitters; therefore, there cannot be propellant emission unless it is electrically activated. In this sense, the planar array is less vulnerable to unplanned propellant emission compared to pressure fed schemes. Additionally, clogging is not an issue in this engine because the propellant is not doped, and the flow channels are open. The planar array uses the ionic liquid EMI-BF4 as a propellant. The ionic liquid EMI-BF4 has a very low vapor pressure, making it suitable to be used in an open architecture engine. The array is composed of a set of spikes, i.e., emitters, coming out from a propellant pool. There are two configurations for the emitters: fully sharpened slender emitters, i.e., pencils, and truncated pyramidal emitters, i.e., volcanoes. The arrays have between 4 and 1024 emitters in an active area of 0.64 cm2. The surface of the engine (tank and emitters) is covered with “black silicon” that acts as wicking material. The hydraulic system has been experimentally characterized, including: start-up tests (Figure 2), wettability tests, current-per-emitter versus voltage characteristics, imprints of the exit stream on a collector, and a thrust test in agreement with the current-per-emitter versus voltage characteristics and the time-of-flight measurements that we have independently obtained at the Space Propulsion Laboratory. Preliminary results demonstrating the feasibility of obtaining substantially larger emission currents at the same extraction voltage by controlling the temperature have also been obtained. The emission from the array seems to be described by a Schottky emission mechanism. |
Numerical Techniques for Integral Equations | Finding computationally efficient numerical techniques for simulation of three-dimensional structures has been an important research topic in almost every engineering domain. Surprisingly, the most numerically intractable problem across these various disciplines can be reduced to the problem of solving a three-dimensional potential problem with a problem-specific Greens function. Application examples include: electrostatic analysis of sensors and actuators, electromagnetic analyses of integrated circuit interconnect and packaging, detailed analysis of frequency response and loss in photonic devices, drag force analysis of micromachined structures, and potential flow based aircraft analysis. Over the last fifteen years, we have been developing fast methods for solving these problems, and have developed widely used programs such as FastCap (capacitance), FastHenry (magnetoquasistatics), FastLap (general potential problems), FastImp (full wave impedence extraction),and FastStokes (fast fluid analysis). Our most recent work is in developing higher order methods[1], methods that efficiently discretize curved geometries[2], methods that are more efficient for substrate problems [3], and methods for analyzing rough surfaces [4]. |
Characterization and Modeling of Nonuniformities in DRIE | We contribute a quantitative and systematic model to capture etch nonuniformity in the deep reactive ion etching (DRIE) of microelectromechanical systems (MEMS) devices [1]. DRIE is commonly used in MEMS fabrication where high-aspect ratio features are to be produced in silicon. It is typical for many devices, of diameters on the order of 10 mm, to be etched simultaneously into a silicon wafer of diameter 150 mm. Devices containing a range of feature diameters exhibit aspect ratio-dependent etching rates, a phenomenon that is well understood [3]. In addition, equivalent features within supposedly identical devices are observed to etch at varying rates. These spatial variations have been explained in terms of uneven distributions of SxFy ions and fluorine neutrals at the wafer scale, and of competition for those species at the device, or die, level. An ion–neutral synergism model [7] is constructed from data obtained by etching several layouts of differing pattern opening densities (Figure 2). Such a model is used to predict wafer-level variation with an r.m.s. error below 3% (Figure 1). This model is combined with a die-level model, which we have reported previously [2,8], on a MEMS layout. The two-level model is shown to enable prediction of both within-die and wafer-scale etch rate variation for arbitrary wafer loadings. |
Measuring the Mechanical Properties of Thin Films Using MEMS structures | Simple micromechanical devices are being developed to measure the mechanical properties of thin films in localized areas after processing. The simplest devices to fabricate are cantilevers overhanging a pit formed using an anisotropic etch. Cantilevers formed from a material of interest can be used to measure the through-thickness stress-gradient and the elastic modulus of that material. Measuring the elastic modulus requires applying a known force to the tip of the cantilever and measuring the subsequent deflection or curvature. We have developed a technique for high accuracy modulus measurement by application of a force with a beam having known properties, with deflection measurements made in an optical profilometer.Membrane devices, as shown in Figure 1, can be used to measure the stress in a thin film without further processing. The membranes are fabricated using an SOI wafer as the starting material. An anisotropic etch from the backside is used to form the membrane, which consists of two layers: buried silicon dioxide under the device single crystal silicon. The membrane buckles because the buried silicon dioxide is under compressive stress relative to the silicon. The amount of buckling is determined by the mechanical properties and the geometry of the membrane, and is measured using optical profilometry. Depositing a film on either side of the membrane changes the buckling, and therefore, the stress of the new material can be determined. Films deposited on both sides of the membrane contribute to the change in deflection; consequently, the stress in CVD films can be measured. Buckling of doubly-supported beams can be used to charac-terize compressive stresses. To characterize tensile stresses, we have recently developed a new type of device, a V-shaped beam, as shown in Figure 2(a). The V-beam is made from a material of interest. A tensile stress causes out-of-plane bending that can be measured using an optical profilometer. The measured deflections are then compared to finite element analyses. Two modes of bending have been seen in V-beams produced from silicon nitride thin films. Finite element models of the 2 modes showing vertical deflection contours can be seen in Figures 2(b) and 2(c). Mode 1 bending is symmetric and produces very large deflections that are often too large to measure in an optical profilometer. Most beams tend to bend into Mode 2, which is asymmetric, but easily measured us-ing an optical profilometer. Mode 2 deflections also have the advantage that the through-thickness stress gradient does not change the deflection. Because all the devices described above are small, they can be placed in many locations on the wafer. |
Scanning Probe Microscopy with Inherent Disturbance Suppression Using Micromechanical Devices | Scanning probe microscopes are notoriously susceptible to disturbances, or mechanical noise, from the surrounding environment that couple to the probe–sample interaction. These disturbances include vibrations of mechanical components, piezo drift, and thermal expansion. Disturbance effects can be substantially reduced by designing a rigid microscope, incorporating effective vibration isolation, and selecting an appropriate measurement bandwidth and image filter. However, it is not always possible to satisfy these requirements sufficiently, and as a result, critical features in an image can be obscured. The cause of this problem is that the actuator (control) signal is used both to readout topography and correct for disturbances. We have introduced a general approach for inherently suppressing out-of-plane disturbances in scanning probe microscopy [1]. In this approach, two distinct, coherent sensors simultaneously measure the probe-sample separation. One sensor measures a spatial average distributed over a large sample area, while the other responds locally to topography underneath the nanometer-scale probe. When the localized sensor is used to control the probe-sample separation in feedback, the distributed sensor signal reveals only topography. This configuration suppresses disturbances normal to the sample. We have applied this approach to scanning tunneling microscopy (STM) with a microcantilever that integrates a tunneling tip and an interferometer (Figure 1) and have shown that it enables Angstrom resolution imaging of nanometer-sized gold grains in a noisy environment (Figure 2). For disturbances applied normal to the sample, we measured disturbance suppression of -50 dB at 1 Hz, compared to 0 dB with conventional imaging. |
In-Plane AFM Probe with Tunable stiffness | We developed an in-plane Atomic Force Microscope (AFM) probe that is specifically tailored to the needs of biological applications. It features a variable stiffness, which makes the stiffness of the probe adjustable to the surface hardness of the sample [1]. The inherent capability of the in-plane AFM probe for building a massively parallel array is also an important feature that greatly affects the speed of the AFM scanning process. Concept and FunctionalityThe switchable stiffness probe allows the scanning of biological samples with varying surface hardness without changing probes during scanning and therefore, prevents a loss of positional information, as is unavoidable with conventional devices. For the integration of the components into a MEMS device, the conventional cantilever-type design of AFM probes has been abandoned in favor of an in-plane design. The new design has an advantage in that it facilitates a high-density array of AFM probes and allows for easy surface micromachining of the integrated device. It also enables the integration of micro-fluidic channels for reagent delivery and nanopipetting. For scanning nano-scale trenches and grooves, a multi-walled carbon nanotube, embedded in a nanopellet [2], is mechanically assembled to the AFM probe as a high-aspect-ratio tip. Design and FabricationThe variable stiffness is accomplished in a mechanical way by engaging or disengaging auxiliary beams to the compliant beam structure by the means of electrostatically actuated clutches (Figure 1). Figure 2 shows the integrated AFM probe system. For actuation, an electrostatic combdrive is considered to move the probe tip up and down. The vertical displacement of the tip can be measured by a capacitive sensor, which can easily be integrated into the system. |
Direct Patterning of Organic Materials and Metals Using a Micromachined Printhead | Organic optoelectronic devices are promising for many commercial applications, if methods for fabricating them on large area low-cost substrates become available. Our project investigates the use of MEMS in the direct patterning of materials needed for such devices.In our first demonstration, we used an electrostatically actuated micromachined shutter integrated with an x-y-z manipulator to modulate the flux of evaporated organic semiconductors and metals and to generate patterns of the deposited materials. The micromachined printhead consists of a free-standing silicon microshutter actuated over a 25 micron square aperture by a comb-drive actuator. Figure 1 shows the microshutter and aperture. The device is fabricated, starting with a SOI (silicon on insulator) wafer, and using deep reactive ion etching to pattern both the through-wafer aperture and the free-standing structure and actuation mechanism. An operating voltage of 30 V is needed to obstruct the aperture with the microshutter. The simulated first mechanical resonant frequency of the device is 6 kHz.We tested the printing method in a vacuum chamber by depositing an organic semiconductor, Alq3 (tris (8-hydroxyqunolinato) aluminum), and silver on glass substrates. We also printed arrays of organic light emitting devices (OLED). Figure 2 shows patterns obtained using this method: photoluminescence image of 40 micron pixels of Alq3, optical microscope image of 30 microns wide line patterns of silver, and electroluminescence of 30 micron pixels arrays of TPD:10%DCM/Alq3/TAZ at 20 V (with blue filter), and of TPD/Alq3/TAZ at 10V (no filter). The results show that this printing technique is capable of patterning small molecule organic light emitting devices at high resolution (800 dpi in our case).The next stage of this project will involve investigating the use of a microporous layer with integrated heaters for local evaporation of the materials. |
Nanometer-Level Positioning in MEMs without Feedback control | Traditional macro-scale nanopositioners rely on sensors and feedback control to achieve nanometer-level accuracy and repeatability. The need for low-cost, high-speed precision positioning devices has led to a trend in miniaturization of these machines. Miniaturization of precision positioning devices is problematic as precision positioners require feedback control, and feedback control is not readily adapted to small-scale machines. The difficulty in adaptation is due mainly to the challenges encountered during the integration of small-scale sensors, mechanisms, and actuators. In this work, we are designing multi-axis MEMS that are capable of nanometer-level positioning without sensing/feedback control. The approach has grown from binary actuation technologies used in macro-scale robotics [1,2].In our approach, Digital Nanoactuation Technology (DNAT), a positioner is equipped with actuator-flexure building blocks. The blocks consist of a pair of binary actuators that work together to generate discrete, repeatable positions. The actuators are attached to a positioning stage via flexures such that the actuator-flexure sets are diametrically opposed. An actuator set is shown on the left side of Figure 1. The opposed flexures differ in stiffness, one compliant, KC, and one stiff, KS. When both actuators are activated (four possible on-off combinations), four repeatable positions may be obtained. DNAT building blocks may be superimposed to provide many position states. For example, the 64 states shown on the right side of Figure 1 are obtained by superimposing the output of three blocks. The number of states scales with the number of actuator pairs, N, as 22N. A positioner with N = 6 is capable of over 4000 discrete positions. If these points are encompassed within a space of a few microns, simple on and off actuator commands may be used to obtain nanometer-level repeatability without sensing/feedback. A macro-scale analogy of a small-scale device has been constructed and tested [3] to demonstrate that nanometer-level positioning is possible. The small-scale prototype shown in Figure 2 is being tested to characterize a 64 state prototype before we progress to a smaller, 4000 state device. |
An Electrostatic, circular Zipping Actuator for the Application of a Tunable capacitor | A tunable capacitor is devised using a circular zipping actuator, based on its ability to potentially control a gap between two large surfaces with nanometer resolution [1]. The device consists of three wafers; a SOI (Silicon-On-Insulator) wafer sandwiched by two Pyrex glass wafers that are anodically bonded together, as shown in Figure 1. In the center of the device is a circular membrane that is supported by tethers that are connected to the outer walls. A cylindrical fulcrum, fabricated by the deep reactive ion etching technique, acts as the pivot for the membrane and divides the membrane into the outer actuator region and the center capacitor region. The top of the fulcrum is bonded to the top glass wafer for structural rigidity. The SOI layer is used as the membrane-actuator because of its uniform thickness and the low stress of single-crystal silicon. Thermally grown silicon dioxide is used as dielectric insulation. The bottom wafer contains the bottom electrodes for the actuator and the capacitor. The actuator electrode is etched into the glass to form the gap of the actuator. Gold is deposited on top of the glass wafer as both actuator and capacitor electrodes. Voltage is applied between the top and the bottom actuator electrodes. At a certain threshold, the outer membrane snaps down. With increasing actuation voltages, the membrane zips along the radial direction, as shown in Figure 2, and results in the separation of the two capacitor surfaces. Because of the poor adhesion of gold to oxide, the membrane will not be bonded to the gold surface, although the two are in close contact during operation. Thus, the design makes it possible to have two initially closed-contacted surfaces that can be pried apart. By changing the gap between the two plates of the capacitor, the capacitance can be tuned.The device is modeled using both numerical methods with Matlab and FEM with ANSYS. Tests are done using a laser interferometer to measure the center displacement and a network analyzer to measure the capacitance change. |
A Low contact Resistance MEMs Relay | An electrostaticaly driven, bulk micromachined, low contact resistance MEMS cross bar relay has been designed, and is currently under fabrication. This relay will be used to study and optimize the behavior of micro-scale contacts for power applications.Many MEMS relays have been reported in the literature [1,2,3]; most, however, are not suited for practical power applications due to their high contact resistance. A contact resistance of 50 mΩ [4] has been achieved by our group using a bulk micromachined, externally actuated structure as a proof of concept for this design [4].The electrostatic “zipper” actuators [4,5] are designed for low pull-in voltage (~100 V) and large contact travel (~40 µm) to prevent arcing as the load circuit (up to 600V) is switched on and off. Figure 1 shows the MEMS relay. Figure 2 shows a detailed view of the actuator. The two arms of the parallelogram flexure are used as the traveling electrodes of the electrostatic actuators. Each traveling electrode, or arm of the parallelogram flexure, is adjacent to a pair of stationary electrodes: an engaging and a disengaging stationary electrode. The relay is engaged by electrostatic attraction between the traveling electrodes and the engaging stationary electrodes. Similarly, the MEMS relay is disengaged through electrostatic attraction between the traveling electrodes and the disengaging stationary electrodes. Each stationary electrode is comprised of a stiff component and a compliant, cantilevered component. The cantilevered component reduces the pull-in voltage by reducing the distance between the electrodes. As the actuator is energized, the compliant end of the stationary electrode, having the lower stiffness, is attracted by and deflected toward the moving electrode, making initial contact at the loose end of the cantilever. As the actuation voltage is increased, the contact point between the electrodes is displaced along the stationary electrode over the stiff component of the electrode in a “zipping” motion. Our group continues to develop these MEMS relays for power applications. |
A Variable capacitor Made from single crystal silicon Fracture surfaces | A process for the fracture fabrication of single crystal silicon surface pairs with nanoscale roughness has been developed, and a prototype variable capacitor, featuring fracture surfaces as the moveable parallel plates, has been fabricated. The surfaces are fabricated by notching a portion of a compliant structure with either potassium hydroxide (KOH) or Focused Ion Beam (FIB) milling to produce a stress concentration. The device is fractured by pulling on the compliant structure with a probe. Post-fracture, the compliant structure acts as a bearing so the two surfaces can be brought back into intimate contact without misalignment. Proper alignment ensures that nanometer scale gaps can be maintained with surfaces that are perfectly smooth or complementary. Complementary surfaces have been closed to gaps less than 20 nm. For a successful fracture, the notch must be very sharp and properly aligned to the crystal structure, and the compliant structure (typically etched into the device layer of a Silicon On Insulator (SOI) wafer) must attenuate stray forces and moments and withstand the trauma of fracture. Experiments with different specimens have shown 10 µm to be the optimal thickness (Figure 1).An updated version of the device used for the surface fabrication experiments has been fabricated, assembled, and sealed (Figure 2). This device includes an integrated zipper actuator [1] for controlling the separation of the surfaces, as well as, provision for wirebonding the device into its hermetically sealed package. Testing has confirmed that the actuator functions properly and that the specimens survived the fabrication process. The device also validated the electrical model used to design the capacitance measurement circuitry. Unfortunately, fracturing of these new devices has been problematic: growing the actuator’s thermal oxide has likely blunted the notches. The fabrication process has been debugged, and a new round of fabrication (with an improved design) is nearing fruition. |
A High-Q Widely Tunable Gigahertz Electromagnetic cavity Resonator | RF systems need high-frequency widely tunable high-Q bandpass filters for channel selection filters and local oscillators. Our work describes the design, fabrication, and testing of an electromagnetic cavity resonator designed for such applications. Alternative technologies provide wide tuning or high Q, but not both, and are generally not tunable. This resonator is distinguished by its simultaneous high Q near 200 and its wide high-frequency tuning range of 2.5 GHz to 4.0 GHz, which have been experimentally demonstrated. The resonator is fabricated using standard MEMS technologies and consists of a gold-lined capacitor and toroidal inductor cavity formed by etching silicon in potassium hydroxide (Figure 1). Frequency tuning is performed by compressing the cavity to close the capacitor gap. Testing was done with a piezoelectric actuator for this task. The match between the modeled and measured impedance is extremely good up to and beyond 5 GHz, with less than a 1% error in magnitude and phase. |
Lateral, Direct contact RF MEMs switch with PZT Actuation | A novel direct contact MEMS switch is developed with compliant lateral metal contacts to address the need for low contact resistance and long life cycles. The device is unique in its self-alignment of the contact surfaces, self-cleaning of particles generated at each contact cycle, and mechanical anchoring method of the contact metal to the side of the Su-8 beam structures. The fabricated device maintains less than 0.1Ω contact resistance for up to 10 billions of cycles of contact. A fabricated device is shown in Figure 1 (a). Each switching member consists of two parallel beams with angled contact surfaces. One side of the contacting surfaces is undulated with micro grooves, as shown in Figure 1 (b). When the movable member is actuated to meet the fixed one, the gold on each side of the contact creates a short circuit. When the movable member is on the other side, enough gap is maintained to open the circuit with high isolation. The angled contact orientation makes the undulated surface slide over the static surface, which pushes entrapped particles or generated micro-weldments into the micro-grooves. By cleaning the surface at every cycle of switching, the micro-undulated surface ensures a low contact resistance over long cycles of switching operation. The grooved contact surfaces show successfully that the self-cleaning concept works and that a low contact resistance below 0.1Ω has been maintained over 10 billion cycles. (Figure 2) Applications of the self-cleaning MEMS switch, such as tunable antennas, are being investigated to assess the commercial potential of our switch. |
Design and Fabrication of Nano-Tweezers | Since the invention of atomic force microscopes (AFM) that provided researchers with a convenient tool to observe objects at nanoscale, manipulation tools at nanoscale have been in high demand. There have been several attempts to create nanomanipulation devices, such as nano-tweezers, to address this challenge. Most such attempts have amounted to single proofs of concepts rather than a practical, readily producible manipulation tool. The goal of this project was to further the current state of nanomanipulators, by producing nano-tweezers that are consistently producible, using batch microfabrication processes. In addition, given the regularity and practicality of the AFM as a nano-scale research tool, the nano-tweezers were intended to also serve as a scanning probe for the AFM. This way, the same tool can to be used to both image and manipulate samples, and the utility of the devices is increased.A two-fold approach was used to tackle the problem. First, using complete batch fabrication methods, a process was created to generate nano-scale tweezer tips separated by a nano-scale gap. This process uses standard micron scale batch lithography to define pyramidal walls in silicon. It then produces an extremely thin cut that self-aligns to the apex of the pyramid. Thus far, tip separations of 358nm and tip widths of 50nm have been repeatably produced. The alignment of the process is within 35nm and is much smaller than that of the lithography tool. The second phase was to create free standing, protruding structures that can serve as the tweezing arms and move with nano-scale resolution. Cantilevered flexural members, coupled with electro-static actuation, were successfully fabricated. These slender cantilevered flexural components measure only 1-2 um in width. A novel process was developed that overcomes problems due to surface tension, and protects the released devices all the way through die separation.The devices have shown actuation behavior that is consistent with theory and design intent. Resolution of motion of 40nm has been verified using SEM through the entire working range of the device. Resolution of less than 10nm is expected based on data but has not been verified due to the limits of this SEM. |
Induced-charge Electro-Osmotic Pumps and Mixers for Portable or Implantable Microfluidics | Microfluidic technology offers great promise in diverse fields such as bioinformatics, drug delivery, and analytical chemistry. In spite of involving microchannels, however, current lab-on-chip technologies are mostly limited to bench-top analysis due to various bulky external elements. For example, peristaltic pumping in soft-polymer channels requires complicated tubing and flow meters, and capillary electro-osmosis requires a high-voltage power supply. Miniaturizing and integrating the power source is a crucial next step toward portable or implantable devices for medical diagnostics, localized drug delivery, artificial organs, or pressure control to treat diseases such as glaucoma.We are developing new kinds of pumps and mixers exploiting “induced-charge electro-osmosis” (ICEO) [1], as a potential platform for portable microfluidics. ICEO refers to the slip of a liquid electrolyte at a polarizable (metal or dielectric) solid surface, driven by an electric field acting on its own induced surface (double-layer) charge. Unlike classical (fixed-charge) electro-osmosis, which requires large DC voltages (>100V) applied down a channel, ICEO can be driven locally by small AC voltages (<10V). It is sensitive to the geometry, ionic strength, and driving frequency and scales with the square of the applied voltage. The effect generalizes “AC electro-osmosis” at planar electrode arrays [2] and offers some more flexibility. We originally demonstrated ICEO flow in dilute KCl around a platinum wire by comparing flow profiles from micro-particle-image velocimetry (µPIV) to our theory [3]. We have also fabricated many devices involving electroplated gold structures on glass in PDMS microchannels, which exhibit mm/sec flow rates in 100 V/cm fields at kHz AC, and further optimization is underway. As a first application, we are developing a portable ICEO-powered biochip to detect blood exposure to toxic warfare agents by lysing cells and amplifying and detecting target genes. |
Resonant Body Transistor with MIT Virtual Source (RBT-MVS) Compact Model | High-Q mechanical resonators are crucial components for filters and oscillators that are essential for RF and ana-log circuits. It is highly desirable for resonators to scale to GHz-frequencies and beyond to meet today’s challenging requirements in terms of speed and data rates. Further-more, aggressive scaling requirements call for monolith-ic integration with CMOS circuits to allow for a smaller footprint and reduced parasitics and power consumption. Micro-electromechanical (MEM) resonators represent a potential solution for frequency and footprint scaling, along with monolithic integration in CMOS.A resonant body transistor (RBT) is a MEM resonator with a field-effect transistor (FET) incorporated into the resonator structure. The FET is intended for active sensing of the mechanical vibrations through piezoresistive modulation of the channel mobility. RBTs also rely on electrostatic internal dielectric transduction for actuation, by means of MOS capacitors (MOSCAPs). Such sensing and actuation enable these devices to easily scale to multi-GHz frequencies, while being compatible with CMOS manufacturing technologies.Compact modeling for these devices is essential to gain a deeper insight into the tightly coupled physics of the RBT while emphasizing the effect of the different parameters on the device performance. It also grants circuit designers and system architects the ability to quickly assess the performance of prospective RBTs, while minimizing the need for computationally intensive coupled-multi-physics finite element method (FEM) simulations.The RBT compact model is developed as a set of modules, each representing a physical phenomenon. Mechanical resonance, FET sensing, MOSCAP driving, and thermal modules are the most notable. The modules are interconnected through a set of nodes (namely, mechanical nodes and a thermal node) to represent the coupling between the different physics. This modular approach enables the seamless expansion of the RBT model either by incorporating new physics, adding driving or thermal sources, or mechanically coupling multiple RBTs together. A modified version of the MIT Virtual Source (MVS) model is used to implement both the electrostatic driving (as a MOSCAP) as well as the piezoresistive active FET sensing. The full model is developed in Verilog-A and available on nanohub.org. |
Piezoelectric Micro-Machined Ultrasonic Transducer Array for Medical Imaging | Diagnostic medical ultrasound imaging is becoming in-creasingly widespread because it is relatively inexpen-sive, portable, compact, and non-invasive compared to other diagnostic scanning techniques. However, com-mercial realization of advanced imaging trends will re-quire cost-effective, large-scale arrays of miniaturized elements, which are expensive to fabricate with the current bulk piezoelectric transducers. At high volume, micro-fabricated transducers based on micro-electro-mechanical (MEMS) technology are an array-compati-ble and low-cost option. The piezoelectric micro-machined ultrasonic transducer (pMUT) is a promising alternative to previously proposed capacitive MUT devices since it does not suffer from electrostatic transduction limitations, including potentially unsafe high bias voltage, and non-linearity. With more effective transformation via the piezoelectric effect, pMUTs have already demonstrated viability for deep penetration imaging via high acoustic pressure output. However, insufficient modeling has produced pMUT devices that often fall short of predictions, resulting in low electromechanical coupling and reduced bandwidth. With an improved modelling framework and optimization, pMUT based arrays have the potential for efficient, low-power, and high-pressure operation necessary for wearable applications.Based on a high force-output figure of merit, a 31-mode, lead zirconate titanate (PZT)-based pMUT plate cell design is selected. Our previous work developed and validated an analytical, electro-acoustic model of the single cell through experiment and finite element simulation. By leveraging and building on the validated single-cell model, we further optimized parallelized multi-cell elements to achieve high acoustic power and power efficiency. These elements are incorporated into 1D arrays (Figure 1) to demonstrate basic beamforming and image collection capabilities of a pMUT-based ultrasound system.Current work focuses on fabrication of the pMUT arrays (Figure 2) using common micro-fabrication techniques including a PZT sol-gel deposition process. Beyond fabrication, the project aims to generate proof-of-concept images to demonstrate the commercial viability of pMUT-based array systems. |
Development of a Tabletop Deep Reactive-Ion Etching System for MEMS Development and Production | A general rule of thumb for new semiconductor fabrica-tion facilities (fabs) is that revenues from the first year of production must match the capital cost of building the fab itself. With modern fabs routinely exceeding $1 billion to build, this rule serves as a significant barrier to entry for research and development and groups seeking to commercialize new semiconductor devices aimed at smaller market segments that require a dedicated pro-cess. To eliminate this cost barrier, we are working to create a suite of tools that will process small (~1”) sub-strates and cost less than $1 million. This suite of tools, known colloquially as the 1” Fab, offers many advan-tages over traditional fabs. By shrinking the size of the substrate, we trade off high throughputs for significant capital cost savings while incurring substantial savings in material usage and energy consumption. This sub-stantial reduction in the capital cost will drastically increase the availability of semiconductor fabrication technology and enable experimentation, prototyping, and small-scale production to occur locally and econom-ically. To implement this suite of 1” Fab tools, our cur-rent research has been focused primarily on developing a deep reactive-ion etching (DRIE) system. DRIE tools are used to create highly anisotropic, high aspect-ratio trenches in silicon—a crucial element in many MEMS processes that will benefit from a 1” Fab platform. A la-beled image of the 1” Fab DRIE system is shown in Fig-ure 1. The load lock and wafer lift assembly allow up to 2” wafers and pieces to be easily loaded and processed, and the modularized design of the processing chamber means that the (currently DRIE) system can be easily adapted to produce other plasma-based etching and deposition tools (such as PECVD and RIE). Using the switched-mode Bosch Process, the 1” Fab DRIE system currently can achieve silicon etch rates up to 6 µm/min with vertical sidewall profiles, an estimated photoresist selectivity greater than 75:1, and etch depth non-unifor-mity to less than 2% across the substrate. Several exam-ples of anisotropic etches performed with our system are included in Figure 2. Presently, we are working to refine the thermal design of the system and optimize recipes for high-aspect ratio etching. |
MEMS Energy Harvesting from Low-Frequency and Low-Amplitude Vibrations | Vibration energy harvesting at the microelectrome-chanical system (MEMS) scale will promisingly advance exciting applications such as wireless sensor networks and the Internet of Things by eliminating troublesome battery-changing or power wiring. On-site energy gen-eration could be an ideal solution to powering a large number of distributed devices usually employed in these systems. To enable the envisioned battery-less systems, a fully assembled energy harvester at the size of a quar-ter-dollar coin should generate robustly 101~102 µW of continuous power from ambient vibrations (mostly less than 100 Hz and 0.5 g acceleration) with wide bandwidth. We are inching close to this goal in terms of power densi-ty and bandwidth, but not in terms of low-frequency and low-amplitude operations.Most reported vibration energy harvesters use a linear cantilever resonator to amplify the energy absorption from weak ambient vibrations. While such structures are easy to model, design, and build, they typically have unusably narrow bandwidths. In contrast, nonlinear resonators have a different dynamic response and greatly increase the bandwidth by hardening or softening the resonance characteristic. Our previous research with nonlinear resonating bridge-structure-based energy harvesters achieved 2.0 mW/mm3 power density with >20% power bandwidth. However, they were operated with input vibrations of >1 kHz at 4 g, which practically limits the use of this technology for harvesting energy from real environmentally available vibrations. Many believed this is an inherent limitation imposed on the MEMS-scale structures.We approached this problem with a buckled-beam-based bi-stable nonlinear oscillator. Compared to mono-stable nonlinear oscillations, we found bi-stable oscillations could bring more dynamics phenomena to help reduce the operating frequency. An electromechanical lumped model has been built to simulate the dynamics of the buckled clamped-clamped beam-based piezoelectric energy harvesters. The two oscillation modes, intra-well and inter-well with respect to the double-well energy potential of the bi-stable system, have been predicted. The characteristic spring softening and spring stiffening responses corresponding to the simulations were observed by testing a meso-scale prototype. The testing results also verify the theoretical prediction on low-frequency operation, showing a shifted response of bi-stable configuration, which generates more power than the mono-stable configuration at lower frequencies. A MEMS mechanical bi-stable oscillator has also been fabricated to verify the operating frequency and amplitude of the new design (Figure 1). The multi-layer bridge structure has employed compressive residual stress in the micro-fabricated thin films to induce buckling and lower the operation frequencies. The dynamic responses were measured by a laser Doppler vibrometer (Figure 2). The wide-band nonlinear response shows a one-order-of-magnitude lower frequency range at low g’s. The fully functional piezoelectric devices are under fabrication. |
Close-Packed Silicon Microelectrodes for Scalable Spatially Oversampled Neural Recording | The extracellular recording of brain activity in the mammalian brain provides an important tool for un-derstanding neural codes and brain dynamics. Extra-cellular electrodes with recording sites that are closely packed can enable spatial oversampling of neural activ-ity, which facilitates data analysis; such oversampling becomes important when we aim to scale up the num-ber of neurons used for recording.We designed and implemented close-packed silicon microelectrodes (Figure 1) to enable the spatially oversampled recording of neural activity (Figure 2) in a scalable fashion, using a tight continuum of recording sites along the length of the recording shank, rather than discrete arrangements of tetrode-style pads or widely spaced sites. This arrangement, thus, enables spatial oversampling continuously running down the shank, so that sorting of spikes recorded by the densely packed electrodes can be facilitated for all the sites of the probe simultaneously.We use MEMS microfabrication techniques to create thin recording shanks, and a hybrid lithography process that allows a dense array of recording sites, which we connect with submicron dimension wiring. We have performed neural recordings with our probes in the live mammalian brain. Figure 2 illustrates the spatial oversampling potential of close packed electrode sites. |
Wearable Energy Harvesters Based on Aligned Mats of Electrospun Piezoelectric Nanofibers | Battery recharging and replacement are still chal-lenging after several decades of developing energy sources for portable and wireless devices. For this reason, new power sources have become essential for current and future stand-alone devices. Energy harvesters are an attractive alternative for supplying power in these systems. We are developing wearable energy harvesters based on electrospun piezoelectric nanofibers as transducing elements. The proposed harvesting device consists of a set of flexible interdigitated electrodes on a flexible substrate; the electrodes are coated with aligned piezoelectric nanofibers. Each time the substrate is stretched or bent, the piezoelectric nanofibers produce voltage and charge that can be used to feed low-power devices. Our energy harvesters could be integrated into garments, allowing people to carry less weight and volume in batteries, which is particularly advantageous on long journeys and when located far from the electrical grid. The piezoelectric nanofibers of our energy harvester are made of poly(vinylidene difluoride), i.e., PVDF, using the electrospinning technique. In electrospinning, a solution rich in long-chain polymers that is subject to a high electrostatic field ejects a jet that is thinned to a submicron diameter due to the interaction of the electric field and surface tension effects on the fiber (Figure 1). Highly aligned fiber deposition on the interdigitated electrodes of the energy harvester is necessary to achieve high efficiency. With this goal in mind, we developed a custom rotating collector system that allows control of the alignment and diameter of the deposited nanofibers. The collected fibers tend to be more aligned and exhibit smaller fiber diameters when the collector drum rotates at thousands of revolutions per minute (Figure 2). Current work focuses on controlling the morphology of the PVDF fibers and nanofiber mats, as well as on testing nanofiber harvester prototypes using a custom apparatus and benchtop electronics. |
Prediction and Characterization of Dry-Out Heat Flux in Micropillar Wick Structures for Thermal Management Applications | Thin-film evaporation in wick structures for cooling high-performance electronic devices is attractive be-cause it harnesses the latent heat of vaporization and does not require external pumping. However, optimiz-ing the wick structures to increase the dry-out heat flux is challenging due to the complexities in modeling the liquid–vapor interface and the flow through the wick structures. In this work, we developed a model for thin-film evaporation from micropillar array wick structures (Figure 1) and validated the model with ex-periments. The model numerically simulates liquid velocity, pressure, and meniscus curvature along the wicking direction by conservation of mass, momen-tum, and energy based on a finite volume approach. Specifically, the three-dimensional meniscus shape, which varies along the wicking direction with the lo-cal liquid pressure, is accurately captured by a force balance using the Young–Laplace equation. The dry-out condition is determined when the minimum con-tact angle on the pillar surface reaches the receding contact angle as the applied heat flux increases. With this model, we predict the dry-out heat flux on various micropillar structure geometries (diameter, pitch, and height) in the length scale range of 1–100 μm and discuss the optimal geometries to maximize the dry-out heat flux (seen in Figure 2). We also performed detailed experiments to validate the model predictions, which all show good agreement. This work provides many insights into the role of surface structures in thin-film evaporation and also offers important design guidelines for enhanced thermal management of high-performance electronic devices. |
Fabrication of Core-Shell Microparticles Using 3-D Printed Microfluidics | Coaxial electrospraying is an electrohydrodynamic process that creates core-shell microparticles by at-omization of a coaxial electrified jet composed of two immiscible liquids. Coaxial electrospraying has several advantages over other microencapsulation technol-ogies including higher encapsulation efficiency and more uniform size distribution. Coaxial electrosprayed compound microparticles can be used in exciting ap-plications such as feedstock microencapsulation, con-trolled drug release, and self-healing composites.Unlike traditional, i.e., uniaxial, electrospraying that has been investigated for over 100 years and of which many MEMS implementations exist, coaxial electrospraying was first described in 2002 and no microfabricated coaxial electrospray source had been reported due to the inherent three-dimensionality and complexity of its hydraulic system.Stereolithography (SLA) is a layer-by-layer additive manufacturing process that creates solid objects via photopolymerization of a resin using ultraviolet light. Additive manufacturing started as a visualization tool for mesoscaled objects, but recent developments in the resolution and capabilities of 3-D printing suggest that these manufacturing processes could address the complexity, three-dimensionality, and material requirements of many microsystems. In particular, high-resolution SLA can be used to manufacture freeform microfluidics at a small fraction of the cost per device, infrastructure cost, and fabrication time of a typical silicon-based microfluidic system.We developed SLA 3-D printed coaxial electrospray sources with one or two emitters that are fed by two helical channels (Figure 1). Each emitter spout is designed to produce a coaxial flow and to enhance the electric field on the liquid meniscus. Using these devices, we produced uniform core-shell microparticles using deionized water as the inner liquid and sesame oil as the outer liquid (Figure 2). The size of the droplets can be modulated by controlling the flow rates fed to the emitters. Electrical characterization of the devices demonstrates that the emitters operate uniformly. Current research efforts focus on demonstrating massively multiplexed sources with uniform array operation. |
Thin, Flexible, and Stretchable Tactile Sensor Based on a Deformable Microwave Transmission Line | Over the past decade, there have been numerous pub-lications on tactile sensors and skins aimed at repli-cating the human sense of touch in applications such as robotics, healthcare, and prosthetics. A variety of technologies are used, with the dominant ones being piezoresistive and capacitive, but both of these tech-nologies have limitations due to mechanical fragility, complex fabrication, and the need for large numbers of connections to external electronics. We have devel-oped another sensing technology that is mechanically robust, simple to fabricate, and requires only one con-nection to external electronics.The new sensor, shown in Figure 1, consists of a flexible and stretchable 1.6-mm-thick microstrip transmission line with conductors made of stretchable silver-based conductive cloth and a dielectric made of soft silicone rubber (PDMS). When pressure is applied to the line, the dielectric is deformed, causing a local impedance discontinuity in the line. We have developed an algorithm that can reconstruct the deformation of the line as a function of position, based on the measured impedance of the line across a wide frequency range (30 MHz to 6 GHz).To characterize the sensor and algorithm, the sensor was precisely deformed using a custom-designed jig based on a micrometer head while its impedance was measured in real time with a vector network analyzer. The analyzer was connected to a computer. in which the output was processed to display a plot of the reconstructed deformation, also in real time. To correct for imperfections in fabrication, any deformation present with the sensor at rest was subtracted from the responses with pressure applied. Three different pressures were applied at each of three locations, and the responses were combined to create Figure 2. Note that the reconstruction algorithm is derived entirely from physical theory and was calibrated to the measured velocity factor of the line but was not otherwise tuned to match the individual device. |
Extreme Heat Flux Thermal Management via Thin-film Evaporation | Thermal management is a primary design concern for nu-merous power-dense equipment such as power amplifiers, solar energy convertors, and advanced military avionics. During operation, these devices generate large amounts of waste heat (>1 kW/cm2) from sub-millimeter areas. These concentrated heat loads are spatially and tempo-rally non-uniform and cause hotspots which are localized regions with extreme heat flux and exceedingly high tem-perature that can adversely impact device performance and reliability. In this study, we demonstrate an extreme heat flux thermal management solution targeted towards cooling hotspots. Our test devices utilize well-defined silicon micropillar arrays which were fabricated via contact photolithography and deep-reactive-ion-etching for passive fluidic transport (i.e., capillary-wicking). Resistive thin-film heaters were integrated on the back side of our test device via electron-beam evaporation and acetone lift-off to emulate the heat generated by actual electronic chips during operation. The heaters which were used to measure temperature in addition to providing heating were calibrated prior to experiments in a convection oven. The hotspots (640×620 μm2) were spatially distributed over the microstructured surface (1×1 cm2). Uniform background heating was provided by heating the entire microstructured surface using a 1×1 cm2 thin-film heater. Experiments were conducted in a temperature-controlled stainless steel environmental chamber which was maintained at saturated temperature and the corresponding pressure. We dissipated ≈6 kW/cm2 from a single hotspot without background heating before the microstructured surface dried out (Figure 1a). Dryout occurs due to liquid starvation when the viscous losses exceed the capillary pressure generated owing to the meniscus shape. We activated concurrent hotspots on our test devices over the 1×1 cm2 microstructured surface and examined the hotspot dryout heat flux. Our experiments show that this hotspot dryout heat flux decreased monotonically when concurrent hotspots were present on the microstructured surface (left ordinate, Figure 1b). The dryout heat flux, which was ≈6 kW/cm2 when a single hotspot (H2) was present, decreased to ≈4 kW/cm2 per heater when two hotspots (H1/H3) were present (left ordinate, Figure 1b). This dryout heat flux decreased further to ≈3 kW/cm2 per heater when three hotspots (H1/H2/H3) were present (left ordinate, Figure 1b). When a 10 W/cm2 and 20 W/cm2 uniform background heating was superposed with a hotspot, the hotspot dryout heat flux, which was ≈6 kW/cm2 without background heating, decreased to ≈4 kW/cm2 and ≈3 kW/cm2, respectively (left ordinate, Figure 1c). Despite the decrease in the hotspot dryout heat flux, the total heating power increased when concurrent hotspots were created (right ordinate, Figure 1b) or when uniform background heating was superposed with a hotspot (right ordinate, Figure 1c). Our experiments show that thin-film evaporation is a promising thermal management solution for the next generation of power amplifiers and radio-frequency devices which generate extreme heat fluxes in excess of 1 kW/cm2. The insights gained from this study can be used to improve the design of wicking structures which are commonly used in phase-change-based thermal management devices such as heat pipes and vapor chambers. |
Enhanced Water Desalination in Electrochemical System | Currently, reverse osmosis (RO) is considered the lead-ing technology for desalination, and the operational efficiency of RO has been significantly improved over the last two decades with a thorough energy analysis. On the other hand, electrical desalination can be more advantageous in certain applications due to the diver-sity of allowed feed conditions, operational flexibility, and the relative low capital cost needed (the size of a system is generally small). Yet electrical desalination techniques such as electrodialysis (ED) have not been modeled in full detail, partially due to scientific chal-lenges involving the multiphysics nature of the pro-cess. In addition, while current ED relies on bipolar ion conduction, removing one pair of a cation and an anion simultaneously, one final but most important point is that desalination achieved by means of an an-ion exchange membrane (AEM) and a cation exchange membrane (CEM) should be considered separately and independently (Figure 1a). Based on the intrinsically different ion transport near AEM and CEM, our group previously presented a novel process of ion concentra-tion polarization (ICP) desalination (Figure 1b), which can basically enhance the amount of salt reduction, by examining unipolar ion conduction through both experiments and numerical modeling (Figure 1b). Since our experimental works are done in a model system for scalable electrochemical systems, the microfluidic device (Figure 1c) enables more scientific knowledge about ion transport phenomena through visualization. Meanwhile, the high-throughput module (stacked layer system, Figure 1d) enables us to realize a practical op-eration and evaluate the system’s performance. Along with the ICP desalination, we also employed an ED system as a model to investigate the mass transport effects of embedded microstructures between the ion exchange membranes. In this work, therefore, we aim to perform a high-level analysis of ion transport near IEMs in order to enhance water desalination in electro-chemical system. |
3-D Printed Massively Multiplexed Electrospray Sources | Electrospray is a electrohydrodynamic phenomenon that produces from a meniscus a stream of micro/nanoparticles that, depending on the properties of the liquid and the process conditions, can be droplets, ions, or fibers. The low spread in size and specific charge of the emitted particles makes the use of electrospray at-tractive in applications such as combustors, maskless micro/nanomanufacturing, and nanosatellite propul-sion. However, the throughput of an electrospray emit-ter is very low, limiting the applicability of single-emitter electrospray sources to a few practical cases, e.g., mass spectrometry of biomolecules. An approach to increase the throughput of an electrospray source without increasing the size variation of the emission is implementing arrays of electrospray emitters that operate in parallel. Miniaturization of the electrospray emitters results in less power consumption and lower onset voltage; in addition, using micro-fabrication, monolithic arrays of miniaturized emitters with large array size and emitter density can be made. Researchers have demonstrated a variety of MEMS multiplexed electrospray sources that operate uniformly. Although these devices work satisfactorily, they present a number of issues: (i) the device archi-tecture is often a compromise between what should be made based on the modeling and what can be made given the limitations of traditional microfabrication, sacrificing device performance; (ii) a change in any of the in-plane features of the design requires the redesign and fabrication of one or more lithography masks while causing added costs and time delays; (iii) these devices are fairly expensive because they are made in a multi-million semiconductor-grade cleanroom with advanced tools that are operated by highly trained staff, which restricts their application to high-end applications and research.We recently demonstrated the first 3-D printed multiplexed electrospray sources in the literature (Figure 1). The devices were fabricated with stereolithography and have associated two orders of magnitude less fabrication cost per device, fabrication time, and manufacturing infrastructure cost compared to a silicon MEMS multiplexed electrospray source. The 3-D printed devices include features not easily attainable with other microfabrication methods, e.g., tapered channels and threaded holes. Through the optimization of the fabrication process, arrays with as many as 236 internally fed electrospray emitters (236 emitters in 1 cm2) were made, i.e., a twofold increase in emitter density and a sixfold increase in array size compared with the best reported values from multiplexed, internally fed, electrospray sources made of polymer. The characterization of devices with a different array size suggests a uniform emitter operation (Figure 2). |
Optimization of Capillary Flow through Open Microstructured Arrays | Liquid propagation through porous microstructures has received significant attention due to the importance of precisely controlling flow in microfluidic systems. Peri-odic surface structures, e.g., arrays of open micropillars or open microchannels, sometimes can be used to con-trol the flow in a microsystem, introducing benefits such as direct access to the porous structure, device reusabil-ity, and resilience against clogging. In an open fluidic structure, the liquid is not actively pumped, e.g., using an upstream pressure signal; instead, the microstructured surface passively drives the liquid via capillary action. However, the same surfaces driving the flow via sur-face tension’s pull simultaneously impede it by way of viscous resistance. Therefore, optimization of the geom-etry of the microstructured surface is required to maxi-mize the flow rate it transports.We developed semi-analytical models that describe the dynamics of capillary flow against gravity in (i) vertical arrays of open microchannels with rectangular cross-section and (ii) arrays of open micropillars with square packing and square cross section. We also extended our analysis to capture the shear-thinning behavior typical of many non-Newtonian fluids. Our models indicate the existence of multiple flow rate maxima with respect to pore size. One maximum, which occurs only in micropillar arrays, arises from the trade-off between capillary pressure and viscous resistance. The two other maxima, which occur for both micropillar and microchannel arrays, are related to meniscus and gravitational effects and only appear at low aspect-ratio (i.e., in channels/gaps between adjacent pillars that are about as wide as they are deep) and high Bond number, respectively. Experimental capillary rise data demonstrate that incorporating first-order gravitational effects and the impact of meniscus curvature improved flow rate predictions relative to models that neglect these factors (Figures 1 and 2; in both figures the working liquid is 1% PEO in 40/60 ethanol/water). Experimental capillary rise data also confirm the existence and location of a flow maximum with respect to the width of an open-microchannel; operating at any of the maxima decreases the sensitivity of flow rate to geometric variation, allowing for more robust microfluidic systems. Finally, we demonstrated electrospray emission from the edge of a microstructured surface as an example of an application of the porosity geometries we in-vestigated in this study; the supply-limited regime of the current-voltage characteristics of these devices are in agreement with the literature on electrospray droplet emission, opening the possibility to implement arrays of externally-fed electrohydrodynamic jetting emitters that can operate continuously while producing droplets or nanofibers using suitable working liquids. |
Chip-Scale Electrostatic Vacuum Ion Pump with Nanostructured Field Emission Electron Source | Cold-atom interferometry of alkali atoms can be used in a variety of high-precision sensors and timing devices such as atomic clocks, gyroscopes, accelerometers, magnetom-eters, and gravimeters. These devices require ultra-high vacuum (UHV, pressure < 10-9 Torr) to operate; therefore, chip-scale versions require miniaturized UHV pumps re-silient to alkali metal vapors that consume power at lev-els compatible with device portability. In a macro-sized chamber, UHV-level vacuum can be maintained using a conventional magnetic ion pump, where electrons that swirl around the magnetic lines of a magnet create ions by impact ionization of neutral molecules, which in turn sputter a Ti getter. While scaled-down versions of mag-netic ion pumps have been reported, these are incompat-ible with miniaturized cold-atom interferometry systems because (i) a reduction in the pump size increases the required threshold magnetic field for electron trapping, and (ii) the larger magnetic field associated with a minia-turized ion pump can interfere with the operation of the cold-atom sensor, yielding flawed readings. Non-evapo-rable getter (NEG) pumps are used in some cold-atom in-terferometry systems, e.g., commercial chip-scale atomic clocks; however, NEG pumps are unable to pump noble gases such as He and N2 that are present in the chamber, and they inefficiently pump alkali vapors.We are developing vacuum ion pumps compatible with chip-scale cold-atom interferometry devices. The proposed field emitter array (FEA)-based magnet-free ion pump architecture is shown in Figure 1. In this pump design, a helical electron collector pulls the electrons toward itself, forcing them to first travel beyond the height of the electron collector, to then get pushed back due to the electrostatic mirror effect of the annular-shaped ion collector. Therefore, the trajectory of the electrons is significantly increased compared to a pump design with a parallel-capacitor electrode configuration, augmenting the probability of impact ionization. The FEA consists of arrays nano-sharp silicon tips, each surrounded by a self-aligned gate electrode; we have shown that these FEAs do not degrade in the presence of Rb vapor. Figure 2 shows the semi-log plot of the minus time derivative of the pressure versus time during pump-down, with the horizontal axis denoting the time since the beginning of each pump-down cycle; in these experiments, the pressure inside the chamber reached values as low as ~7×10-7 Torr. Each data point in the plot represents an average of the minus time derivative of the pressure considering all pump-down cycles. The R2 of the linear fit of the data evidences that our reduced-order model accurately explains the dynamics of the pump. The slope of the linear fit of the data estimates the experimental pumping time constant at about 161 seconds. |
Hardware Trojan Detection using Unsupervised Deep Learning on High Spatial Resolution Magnetic Field Measurements | One major vulnerability of integrated circuits (ICs) is the difficulty of ensuring that an IC fabricated in a third-party foundry is not a maliciously modified ver-sion of the original design. Such modifications by at-tackers, called hardware Trojans (example shown in Figure 1a), can leak private data from an IC, change its functionality, or have other effects. Attackers can de-sign Trojans so that their effects are not visible during simple functional tests, making detection difficult. However, side channel methods (Figure 1b) can mea-sure differences in circuit activity resulting from the modified logic to detect Trojans prior to the presence of functional changes.In this work, we achieve a method of detecting small footprint hardware Trojans in a field-programmable gate array by performing high spatial resolution and wide field-of-view imaging of the circuit magnetic fields using a quantum diamond microscope. These images are then separated into Trojan-free and Trojan-inserted measurements in an automated framework by using an unsupervised convolutional neural network and clustering. With this framework, we show detection ability comparable to previous literature without requiring any knowledge of the Trojan at test time. |
A Low-Power BLS12-381 Elliptic Curve Pairing Crypto-Processor | Pairing-based cryptography (PBC), a variant of elliptic curve cryptography (ECC), uses bilinear maps between elliptic curves and finite fields to enable novel applica-tions beyond traditional key exchange and signatures, e.g., signature aggregation and functional encryption. These protocols require special pairing-friendly ellip-tic curves; recent cryptanalysis has compromised the security of commonly used 254b BN curves. There-fore, the new BLS12-381 curve, based on a 381b prime field, is being standardized for PBC applications. How-ever, with strong security, the new curve has higher computational complexity, making implementating low-power embedded devices challenging. To address this challenge, we present the first BLS12-381 elliptic curve pairing crypto-processor, which enables two or-ders-of-magnitude energy savings through efficient hardware acceleration, implements countermeasures against timing and power side-channel attacks, and provides the flexibility to implement ECC and PBC pro-tocols for securing Internet of Things applications.Figure 1 shows the architecture of a pairing cryptoprocessor with the chip micrograph. Our test chip was fabricated in TSMC 40-nm low-power complementary metal-oxide-semiconductor process and supports voltage scaling from 1.1V down to 0.66V. The cryptographic core occupies a 0.2-mm2 area consisting of 112k logic gates and 16 KB SRAM. Programming with custom instructions for modular arithmetic, elliptic curve point and line arithmetic, pairing operations, control, and branching is possible. Key building blocks are constant-time and secure against timing and simple power analysis side-channel attacks. For high-security use, our chip can be configured to protect against stronger differential power analysis side-channel attacks at the cost of increased energy consumption. We have evaluated pairing-based public key cryptography protocols on our chip, including signature aggregation, identity-based signatures, identity-based encryption, inner product functional encryption, and multi-party key exchange. Our hardware-accelerated implementations are 130-140× more energy-efficient than software. The programmability of our pairing crypto-processor allows new protocols, algorithm optimizations, and side-channel countermeasures to be easily implemented using one chip. |
Direct Hybrid Encoding for Signed Expressions SAR ADC for Analog Neural Networks | Artificial intelligence (AI) has proven itself to be one of the most powerful techniques for computer vision, nat-ural language processing, and the automobile industry. Current AI algorithms that are based on deep neural networks (DNNs) are facing a crucial challenge from ef-ficient computing. State-of-the-art DNNs need millions of weights and plenty of computation. The huge ener-gy consumption is neither environmentally friendly nor practical in the battery-constraint edge devices. Conventional DNN hardware is based on fully digital implementation, where data movement is becoming the bottleneck. Data movement typically takes orders of magnitude more energy than the actual computa-tion. Analog neural networks (ANNs) are a promising solution for energy-efficient AI inference. The ANNs perform the in-memory-computing to reduce the ener-gy of data movement. Thus, the analog/digital interface circuits are a critical part of the ANNs and are often the key bottleneck of the performance, power consump-tion, and area of the resulting system.The hybrid encoding for signed expression (HESE) scheme is based on the booth encoding but with additional rules to provide the minimum-length signed-digit-representations (SDR) for efficient encoding for both DNNs including ANNs. This work focuses on a successive approximation register (SAR) analog-to-digital converter (ADC) that produces HESE encoded output on the fly. This ADC has two thresholds for 2-bits look ahead (LA). The proposed SAR ADC can directly encode the analog input to HESE instead of binary encoding. Preliminary results show that in a typical DNN, over 95% of weights can be represented by 5 terms of HESE for a 12-bits resolution. |
Efficient Computation of Map-scale Continuous Mutual Information on Chip in Real Time | Exploration tasks are essential to many emerging ro-botics applications, ranging from search and rescue to space exploration. The planning problem for explora-tion requires determining the best locations for future measurements that will enhance the fidelity of the map, for example, by reducing its total entropy. A wide-ly studied technique involves computing the mutual information (MI) between the current map and future measurements and utilizing this MI metric to decide on the locations for future measurements.However, computing MI for reasonably sized maps is slow and power-hungry, which has been the bottleneck in fast and efficient robotic exploration. In this paper, we introduce a new hardware accelerator architecture for MI computation that features 16 high-efficiency MI compute cores and an optimized memory subsystem that provides sufficient bandwidth to keep the cores fully utilized. Each core employs interleaving to counter the recursive algorithm and workload balancing and numerical approximations to reduce latency and energy consumption.We demonstrate an optimized architecture on a field-programmable gate array (FPGA) implementation, which can compute MI for all cells in an entire 201-by-201 grid (e.g., representing a 20.1-m-by-20.1-m map at 0.1-m resolution) in 1.55 ms while consuming 1.7 mJ of energy, thus finally rendering MI computation for the whole map in real time and at a fraction of the energy cost of traditional compute platforms. For comparison, this particular FPGA implementation running on the Xilinx Zynq-7000 platform is two orders of magnitude faster and consumes three orders of magnitude less energy per MI map compute than a baseline GPU implementation running on an NVIDIA GeForce GTX 980 platform. The improvements are more pronounced when compared to CPU implementations of equivalent algorithms. |
Simulation and Analysis of GaN CMOS Logic | There is an increasing demand for electronics that can operate in high-temperature conditions, such as space-craft applications and sensors for industrial environ-ments. Electronics based on wide-bandgap materials offer a promising solution, among which gallium ni-tride (GaN) stands out as a strong candidate due to its excellent material properties and potential for mono-lithic integration. Most current demonstrations of GaN logic are based on nanometal-oxide-semiconductor (nMOS) technology, which has a high static power con-sumption. Therefore, we are developing GaN comple-mentary metal-oxide-semiconductor (CMOS) technolo-gy, which has lower static power consumption.This work studies the effect of a p-channel transistor and circuit parameters on the performance of CMOS digital logic circuits. We used the MIT Virtual Source GaN-field-effect transistor (MVSGFET) model to accurately model the behavior of the n-channel and p-channel transistors, which were fabricated on the developed GaN complementary circuit platform. We simulated and studied several building blocks for digital logic, namely, the logic inverter, multi-stage ring oscillator, and static random-access memory (SRAM) cell, using the developed computer-aided design (CAD) framework. We conducted device-circuit co-design to optimize circuit performance, using a variety of design parameters, including transistor sizing and supply voltage scaling. We projected the high-temperature performance of the circuits through simulations based on experimentally observed device behaviors. The results indicate that GaN CMOS technology based on our monolithically integrated platform has potential for a variety of use cases, including harsh-environment digital computation. We will apply this technique for more complex combinational and sequential logic building blocks, with the eventual goal of realizing a GaN CMOS microprocessor. |
A 0.31 THz CMOS Uniform Circular Antenna Array Enabling Generation/Detection of Waves with Orbital-Angular Momentum | Multiplexing of electromagnetic (EM) waves with dif-ferent frequencies, polarizations, and coding has been extensively exploited in wireless systems. Recently, an-other dimension of EM waves-the orbital angular mo-mentum (OAM), is attracting increasing attention. An OAM-based wave possesses a wavefront with a helical phase distribution around the central axis of the beam. Different OAM modes, determined by the handedness and the total phase change () of the wavefront twist, are orthogonal. Wireless communication uses multi-OAM mode transmission to enhance spectral efficiency and physical-layer security. Conventional OAM-generation approaches incorporate dielectric spiral-phase plates, passive uniform circular antenna arrays, or metasur-faces in conjunction with separate signal drivers. These discrete solutions, however, lead to very bulky and cost-ly systems.In this project, we demonstrate the first chip-based (at any frequency) CMOS front-end that generates and receives electromagnetic waves with OAM, shown in Figure 1. The chip, based on a uniform circularly placed patch antenna array at 0.31THz, transmits reconfigurable OAM modes, which are digitally switched among the (plane wave), (left-handed), (right-handed), and superposition states. The chip is also reconfigurable into a receiver mode that identifies different OAM modes with >10dB rejection of unintended modes. The array, driven by only one active path, has a measured EIRP of -4.8dBm and consumes 154mW of DC power in the OAM source mode. In the receiver mode, it has a measured conversion loss of 30dB and consumes 166mW of DC power. The OAM chip output mapped from a repeated Keccak-generated data sequence was verified, and the time-domain outputs of the Rx with different SPP configurations are shown in Figure 2, which shows good correlation with matched modes, partial correlation of multiplexed mode, and rejection of unmatched modes. |
Stability Improvement of CMOS Molecular Clocks Using an Auxiliary Loop Based on High-Order Detection and Digital Integration | Recently, chip-scaled molecular clocks (CSMC) have achieved high-frequency stability with low power and compact size by using a rotational-mode transition of carbonyl sulfide (OCS) centered around 231.061 GHz as a frequency reference (f0). In the molecular clock, the probing signal generated from the transmitter is fre-quency-modulated at fm around the center frequency (fc). Since fc is locked to f0 in a feedback loop, the output frequency inherits the excellent stability of the OCS transition frequency.Due to its fully-electronic implementation, CSMC has reduced the cost of high-stability miniaturized clocks. However, the frequency stability is still limited by a finite loop gain of the frequency locked loop and detection non-idealities, which are susceptible to environmental disturbance even though an invariant physical constant is used as the frequency reference. In this work, we propose a new dual-loop CSMC architecture based on both fundamental and high-order transition probing as well as digital integration.In order to achieve a high long-term stability without compromising the signal-to-noise ratio, the fundamental harmonic detection forms the main loop while the higher-order probing is used in an auxiliary loop. The loop fine-tunes the phase-locked loop’s frequency multiplication ratio according to the sign of the high-order detection output. With a proper selection of gain and bandwidth in each loop, the main loop enables the fast correction of frequency, and the auxiliary loop responds against long-term frequency variation. Also, the frequency offset between the clock output and the OCS reference can be eliminated when the clock is locked because the auxiliary loop includes a digital integrator to obtain an infinite DC gain. As a result, the proposed CSMC implemented in 65-nm CMOS process achieved Allan deviation of 5.4×10-10 and 2×10-11 at 1 s and 104 s averaging times, respectively, with 71 mW power consumption. |
A Sampling Jitter Tolerant Continuous-Time Pipelined ADC in 16-nm FinFET | Almost all real-world signals are analog. Yet most of the data is stored and processed digitally due to advances in the integrated circuit technology. Therefore, ana-log-to-digital converters (ADCs) are an essential part of any electronic system. The advances in modern com-munication systems including 5G mobile networks and baseband processors require the ADCs to have large dy-namic range and bandwidth. Although there have been steady improvements in the performance of ADCs, the improvements in conversion speed have been less sig-nificant because the speed-resolution product is limit-ed by the sampling clock jitter (Figure 1). The effect of sampling clock jitter has been considered fundamen-tal. However, continuous-time delta-sigma modulators may reduce the effect of sampling jitter. But since del-ta-sigma modulators rely on relatively high oversam-pling, they are unsuitable for high-frequency applica-tions. Therefore, ADCs with low oversampling ratio are desirable for high-speed data conversion. In conventional Nyquist-rate ADCs, the input is sampled upfront (Figure 2). Any jitter in the sampling clock directly affects the sampled input and degrades the signal-to-noise ratio (SNR). It is well known that for a given root mean square (rms) sampling jitter σt the maximum achievable SNR is limited to 1/(2πfinσt), where fin is the input signal frequency. In an SoC environment, it is difficult to reduce the rms jitter below 100 fs. This limits the maximum SNR to just 44 dB for a 10 GHz input signal. Therefore, unless the effect of sampling jitter is reduced, the performance of an ADC would be greatly limited for high-frequency input signals.In this project, we propose a continuous-time pipelined ADC having reduced sensitivity to sampling jitter. We are designing this ADC in 16-nm FinFET technology to give a proof-of-concept for improved sensitivity to the sampling clock jitter. |
Bandgap-Less Temperature Sensors for High Untrimmed Accuracy | Temperature sensors are extensively used in measure-ment, instrumentation, and control systems. A sensor that integrates the sensing element, analog-to-digital converter, and other interface electronics on the same chip is referred to as a smart sensor. Complementa-ry metal-oxide-semiconductor- (CMOS) based smart temperature sensors offer the benefits of low cost and direct digital outputs over conventional sensors. How-ever, they are limited in their absolute accuracy due to the non-ideal behavior of the devices used to design them. Therefore, these sensors require either calibra-tion or gain/offset adjustments in the analog domain to achieve desired accuracies (Figure 1). The latter pro-cess, also called trimming, needs additional expensive test equipment and valuable production time and is a major contributor to the cost of the sensors. In order to enable high volume production of CMOS-based tem-perature sensors at low cost, achieving high accuracies without trimming is imperative.This work proposes the design of a CMOS temperature sensor that uses fundamental physical quantities resilient to process variations, package stress, and manufacturing tolerances to achieve high accuracies without trimming. Simulation results prove that 3σ inaccuracy of less than 1o C can be obtained with the proposed method. |
High Angular Resolution THz Beam Steering Antenna Arrays in 22-nm FinFET Technology | THz phased arrays are a promising emerging technolo-gy for many applications, including THz imaging, radar, communications, and other sensing applications. This is largely a result of the smaller wavelength at THz fre-quencies and accordingly smaller array size and weight. However, challenges exist in their design, particularly the design of THz phase shifters, which are often lossy, power hungry, and physically large, precluding their use in dense arrays. These losses often arise from the high-resolution nature of the phase shifters. In ad-dition, lossy on-chip transmission lines significantly degrade system performance. In this work, we apply phased array principles to yield dense THz antenna ar-rays with only one bit of phase resolution, yielding per-formance benefits in terms of DC power, THz loss, size, bandwidth, and simplicity. In addition, by distributing radio frequency (RF) power spatially, we mitigate many of the losses with RF signal distribution. This approach is termed reflectarray (reflector array). We demonstrate our approach on complementary metal-oxide semicon-ductor silicon in the form of a 4x4 mm2 chip containing 7x7 antenna elements, operating at 260 GHz. The chip is designed in Intel 22-nm FinFET process so that mul-tiple chips can be tiled to create large arrays that can be scaled in size based on performance requirements. The use of one-bit phase shifters comes at a cost in sys-tem-level performance by introducing sidelobes in the radiation pattern. Our work introduces a number of ap-proaches to mitigate this, allowing the one-bit phased array design to approach the performance of a phased array with a continuous, analog phase shifter. While still in progress, this work pushes towards practical large-scale THz phased arrays. |
DC-DC Converter Implementations Based on Piezoelectric Transformers | Power converters play major roles in many applica-tions ranging from power generation and distribution in electric grids to everyday devices such as mobile phones and computers. As many applications require small form factors, there has been a significant demand to miniaturize power converters while maintaining high performance. Typical converters rely on magnetic energy storage components, but the achievable power densities of magnetics fundamentally decrease at low volumes and therefore limit converter miniaturization. Piezoelectrics, which have more favorable power density scaling properties than magnetics, are a promising energy storage alternative to meet the demands for low-volume power electronics. Furthermore, multi-port piezoelectric transformers (PTs) offer the additional benefits of galvanic isolation and inherent voltage conversion ratios. Despite their potential, PTs have seen little use in converters without magnetics, and such design attempts have unreported or limited efficiencies.In this work, we systematically enumerate isolated and non-isolated converter switching sequences and topologies that best utilize PTs as their only energy storage components. We constrain this search for (1) high-efficiency behaviors such as zero voltage switching (ZVS) and all-positive instantaneous power transfer and (2) practical characteristics such as voltage regulation capability and control simplicity. To evaluate the selected switching sequences, we also develop a model for estimating the PT’s efficiency.Initial experimental results of these converter designs demonstrate promising high-efficiency behaviors and peak whole-converter efficiencies higher than reported for most magnetic-less PT-based converters in the literature. The prototype displayed in Figure 1 is based on a commercially available PT and achieves a peak efficiency of 89.3%, which is close to our estimation model’s predictions. These results suggest that PT-based converters can offer high efficiencies in addition to the low-volume scaling benefits of piezoelectrics. Such characteristics can be advantageous to high-voltage, low-power applications such as portable electronics and biomedical devices, particularly those requiring galvanic isolation. |
Closed Loop Control for a Piezoelectric-Resonator-Based DC-DC Power Converter | Electronics such as computers, mobile phones, house-hold appliances, and even electric vehicles can vary greatly in terms of supply requirements; power elec-tronics are necessary to power these devices from stan-dard sources. Reducing the sizes of power converters allows them to be more cost-effective and useful to a wider range of applications. Traditional DC-DC power converters make use of magnetics for energy storage, but these are less efficient and power-dense when scaled down to small sizes. Our prior work has explored the use of piezoelectric resonators (PRs) as alternative energy storage mechanisms for DC-DC converters, and we successfully demonstrated a magnetics-less PR-based converter with >99% efficiency. However, our initial prototypes depended on open-loop switching times that were manually tuned, meaning the convert-er could not dynamically handle transients or adjust operation when the load or temperature changed. This work presents a closed-loop control scheme for the PR-based DC-DC converter. For high efficiency, the converter is designed to cycle through a specific 6-stage “switching sequence” during each PR resonant cycle. In this sequence, the PR is switched between fixed-voltage energy transfer stages and resonant transition stages (shown in Figure 1), which is challenging to implement in a simple manner. The converter is controlled by two active switches, as shown in Figure 2. Both switches are triggered to turn on purely by voltage measurements of the PR node voltages. Switch 1’s on-time is modulated to control power output, and switch 2’s on-time is modulated to reach the specific high-efficiency point. Simulation results have shown that this control scheme is effective, and we are currently validating it on hardware. The successful implementation of this closed-loop control scheme will allow the PR-based converter to operate on its own, paving the way for use of these small and efficient DC-DC converters in commercial applications. |
Leveraging Multi-Phase and Fractional-Turn Planar Transformers for Power Supply Miniaturization in Data Centers | Data centers are the backbone of the Internet. Their servers represent an important and growing electrical load, and there is strong interest in miniaturizing the supplies that power them. Miniaturization is challeng-ing as it requires both a reduction in volume and an in-crease in efficiency and is bottlenecked in this applica-tion by the need for a high-current transformer. A common approach toward improving the current carrying capability of the transformer is to increase its phase count by employing multiple identical transformers in parallel. Every phase that is added proportionally decreases the “copper loss” (ohmic loss) of the transformer while proportionally increasing its core loss (i.e., loss in the magnetic material). We call this “linear rebalancing.”In this work, we fundamentally re-think the nature of the transformer to maximally leverage the connecting electronics. In particular, by careful placement of the active switching devices required in a converter around and the passive copper and magnetic material comprising the transformer, we can create a “fractional turn” transformer. Employing a half-turn fractional transformer reduces copper loss by a factor of four while increasing core loss by a factor of 2β, where β is between 2 or 3 depending on the core material. Thus, fractional turn transformers yield an “exponential rebalancing” of core and copper loss.We show that the fractional turn concept can also be combined with the common approach of adding transformer phases, enabling multi-phase fractional-turn transformers. For example, a split-phase half-turn transformer (SPHTT) combines the linear and exponential rebalancing of each of those transformers and allows a designer to get closer to the true optimum loss trade-off for a given application. We show that a SPHTT is optimal for a data center application, yielding 3.1x lower loss than a single-phase transformer in the same volume and demonstrating its clear miniaturization benefit. |
Soft-Actuated Micro Aerial Vehicles with High Agility | Developing agile and robust micro-aerial-vehicles (MAVs) that can demonstrate insect-like flight capa-bilities poses significant scientific and engineering challenges. Previously, we chose dielectric elastomer actuators (DEAs) to substitute for rigid actuators and achieved the first take-off and controlled flight of a soft-actuated MAV. In this work, we substantially im-prove the robot’s flight capability through redesigning the actuator, robot wings, and transmission. The new MAV weighs approximately 665 mg and can complete a somersault within 0.16 s. Furthermore, its vertical as-cending speed exceeds 70 cm/s, which makes it among the fastest soft mobile robots, and it outperforms rig-id-powered subgram MAVs. A major contribution to this excellent performance is that we switch to a less viscoelastic elastomer, Elastosil P7670. Compared to our previously used elastomer (5:4 mixture of Ecoflex 0030 and Sylgard 184), this new elastomer has a higher resonance peak, which implies a larger displacement at the resonant frequency. In addition, it has higher a di-electric strength and a shorter pot time. Based on our measurement, the new MAV achieves a high lift-to-weight ratio of 2.2:1, which is 83% better than our previous work. The large lift force enables us to demonstrate hovering flight, ascending flight, in-flight collision recovery, and--more impressively—a somersault. As shown in Figure 2, the MAV takes off and hovers, accelerates upward, flips along its body pitch axis, recovers attitude, and finally returns to hover. The somersault is completed in 0.16 s; during the body flip, the motion capture system loses tracking for approximately 0.1 s. This loss results in the MAV’s hitting the ground before recovering its attitude. Despite experiencing disturbance caused by the collision, the MAV quickly stabilizes its attitude and returns to hover. This is the first time that a soft-driven MAV performs agile tasks that rigid-driven MAVs have not yet demonstrated. |
Adjusting for Autocorrelated Errors in Neural Networks for Time Series Regression | Time series data are ubiquitous. Researchers in many fields, including the social sciences, operations re-search, and engineering often collect time series data to create models for systems without prior or precise knowledge of the model structure and, in turn, provide insight for such systems. During this process of collec-tion and creation, errors inevitably occur. Usually, the assumption is that the errors are uncorrelated at dif-ferent time steps. However, in practice, errors can be autocorrelated when (1) the function space of the mod-el and the true underlying system do not intersect, (2) some key explanatory variables are not collected, or (3) a measurement error at a current time step carries over to future time steps. To solve this issue, previous literature, such as the Cochrane–Orcutt estimation, focuses only on cases where the model is linear or contains only predefined nonlinearity. This focus greatly limits usage, as many systems today (such as in semiconductor manufacturing) are almost certainly nonlinear while the underlying nonlinearity is unknown.Here, we propose to use neural networks (NNs) to approximate the unknown nonlinearity and treat the autocorrelation coefficient ρ as a trainable parameter. The input to our model is a vector of features (i.e., regressors) at time t, and the output is the target scalar (i.e., regressand) also at time t. During training, we jointly optimize model parameters with the autocorrelation coefficient to adjust for the autocorrelated errors. This optimization enables us to train a NN that can fit the nonlinearity and adjust correspondingly to its autocorrelated noise. Compared to previous methods, this one has the advantages of (1) fitting unknown nonlinearity with autocorrelated noise and (2) better optimization via joint training of model parameters and autocorrelation coefficient. Our experimental results show that we obtain a better estimate of the autocorrelation coefficient and improve the model performance especially when the autocorrelated errors are substantial. |
Terahertz Wireless Link for Quantum Computing in 22-nm FinFET | Quantum computing can provide exponential speed-up in solving many of today’s intractable problems such as quantum chemistry, RSA encryption, DNA analysis, etc. In order to implement an error-protected quantum computer (QC), we will require approximately a million or thousands of qubits. State-of-the-art QCs have only around 100 qubits but still demand large-form-factor room- temperature electronics with many radio-fre-quency (RF) cables to realize the control and readout of quantum processors. These RF cables routed from room temperature to cryogenic temperature consume a non-negligible power due to the heat load, limiting the scalability and practical implementations of QCs.We propose a terahertz (THz) wireless link to efficiently deliver the control signals to the cryogenic environment, reducing the heat loss due to the physical conductive links (Figure 1). We implement a cryogenic THz receiver to send multi-Gb/s control signals modulated on a THz carrier (e.g., 260-GHz). The THz operation allows for a small antenna aperture size, high data rate, and minimal interference with the operation of the qubits, working around a few GHz. For the de-modulation of the sub-THz downlink control signal, a THz square-law detector, operating with zero drain bias, is used first to rectify the input to baseband, and then a low-power transimpedance amplifier followed by a post-amplifier are used to boost the baseband signal so that the subsequent digital circuits can operate reliably. Figure 2 shows the chip photo of this prototype. This system opens the door for scalable and practical realization of cryogenic quantum systems. |
Energy-Efficient System Design for Video Understanding on the Edge | With the rise of various applications including auton-omous driving, object tracking for unmanned aerial vehicles, etc., the need increases for accurate and ener-gy-efficient video understanding on the edge. Although plenty of deep learning chips designed for images ex-ist, little work has been done for videos. Video under-standing on the edge has three major challenges. First, video understanding requires temporal modeling. For example, it identifies the difference between opening and closing a box, which is distinguishable only with temporal information. Second, many applications are delay-critical, such as self-driving cars. Third, high en-ergy efficiency matters for edge devices with a tight power budget. Due to temporal continuity, consecu-tive frames might share much information, providing a potential to improve processing efficiency. However, an image-based processing system, which processes frames individually, cannot utilize that. In this project, we co-designed algorithms and hardware for energy-efficient video processing on delay-critical applications (Figure 1). We applied temporal shift module (TSM) on the backbone built on 2D convolutional neural network (Figure 2). To the best of our knowledge, our work is the first chip with temporal modeling support. Moreover, we propose a Real-Time DiffFrame method to reduce on-chip energy and DRAM traffic. It is based on the linearity of convolution, which has Conv(ft) = Conv(ft - ft-1) + Conv(ft-1), where ft and ft-1 are the successive frames. Due to temporal continuity, ft - ft-1 is usually sparse. Instead of the ordinary sparsity-aware convolution in previous work, our method utilizes SparseConv, which does not dilate the input pattern and further improves energy efficiency. The load and store of Conv(ft-1) are the overhead of the DiffFrame method. We propose a scheme to reduce memory traffic for real-time processing. The preliminary results show that our method achieves 1.6x reduction in DRAM traffic over previous work and 1.8x estimated reduction in computation and memory access over the baseline. |
Sparseloop: An Analytical, Energy-Focused Design Space Exploration Methodology for Sparse Tensor Accelerators | Many popular applications (e.g., deep neural networks) involve tensor computations (e.g., cross products) whose operand and result tensors can have high spar-sity. Due to the nature of multiplication, zero multipli-cands always result in zero products. Such computa-tions (which are called ineffectual) can be exploited by hardware sparse optimization features to improve ener-gy efficiency and throughput. We classify these sparse optimization features into three categories: zero-gat-ing, zero-skipping, and zero-compression. Zero-gating improves energy efficiency by keeping the associated hardware components idle for ineffectual computa-tions. Zero-skipping further improves throughput by skipping cycles where ineffectual computations would have taken place. Zero-compression reduces required storage by storing only nonzero values. In recent years, a variety of sparse tensor accelerators have been proposed. Based on the designer’s intuitions, each design applies variations of the aforementioned sparse optimization features differently to the storage and compute levels of the architecture. However, these specific designs are just points in a large and diverse space of sparse tensor accelerators. A fast, flexible, and accurate modeling framework would enable architects to perform early design space exploration in the complete space instead of picking specific points based on intuition.Existing tensor accelerator models are either very detailed and design-specific, leading to slow and limited design space exploration, or fast and flexible but unable to systematically evaluate the impact of sparse optimization features, resulting in inaccurate modeling. In this work, we propose Sparseloop, an analytical modeling infrastructure for performing fast design space exploration of sparse accelerators that vary in both (1) properties associated with sparsity (e.g., compression formats, ineffectual operations’ gating/skipping, and workload attributes) and (2) architecture properties (e.g., organization of the storage hierarchy). To the authors’ knowledge, Sparseloop is the first analytical model that allows systematic evaluation of sparse tensor accelerators. |
Multi-Inverter Discrete-Backoff: A High-Efficiency, Very-Wide-Range RF Power Generation Architecture | Radio-frequency (RF) power amplifiers (PAs) for indus-trial applications, e.g., plasma generation for semicon-ductor processing equipment, operate into variable load impedances at high frequency (e.g., tens of MHz) and power levels (e.g., peak power in kWs), and often with wide overall power ranges and high peak-to-aver-age-power ratios. To meet the evolving needs for semi-conductor processing, goals for RF PAs in these applica-tions include (1) operation over a wide load impedance (as determined by the plasma load); (2) operation across a very wide range of output power (e.g., 100:1 or 20 dB); (3) very fast dynamic response to output commands (e.g., at μs scale); and (4) high peak and average effi-ciency (to reduce cooling requirements and electricity costs). Unfortunately, meeting all these goals has not been possible to date, and efficiency is often sacrificed in order to meet the other performance metrics.This work introduces a scalable power amplifier architecture and control approach suitable for such applications. The architecture consists of modular PAs organized in groups and employs (1) a technique which we call Multi-Inverter Discrete Backoff (MIDB), which losslessly combines the outputs of parallel-grouped switched-mode PAs and modulates the number of active PAs within the same group to provide discrete steps in RF output voltage, and (2) outphasing among the voltage outputs of PA-groups, for fast-response and continuous output power control over a wide range. To further expand the high-efficiency output power range of the system, discrete drain modulation may be optionally employed. In doing so, the MIDB-based architecture can maintain high efficiency and fast RF power control across a very wide range of power backoff. |
Programming a Quantum Computer with Quantum Instructions | The use of quantum bits to construct quantum com-puters opens the door to dramatic computational speedups for certain problems. The maturity of mod-ern quantum computers has moved the field from be-ing predominantly a quantum-device-focused research area to also include practical quantum-computing-ap-plication-focused research. Our research explores a new experimental result on a foundational aspect of how to program quantum computers. A central prin-ciple of classical computer programming is the equiv-alence between data and instructions about what to do with that data. In quantum computers, this equiv-alence is broken: classical hardware is used to generate the sequence of operations to be executed on the quan-tum data stored in the quantum computer. Our experi-ment shows for the first time how the instruction-data symmetry can be restored to quantum computers. We use superconducting qubits as a platform to imple-ment high-fidelity quantum operations enabling the so-called density matrix exponentiation algorithm to generate these quantum instructions. This algorithm provides large quantum speedups for a family of other quantum algorithms. |
Silicate-Based Composite as Heterogeneous Integration Packaging Material for Extreme Environments | Electronic microsystems are foundational to today’s computational, sensing, communication, and informa-tion processing capabilities, therefore impacting indus-tries such as microelectronics, aerospace, healthcare, and many more. Cell phones are an example of what is possible when a variety of systems can be tightly integrated into a highly portable and capable system. However, as we aim to improve our ability to interact and operate (e.g., sense, communicate, record, compute, move, etc.) in extreme environments (such as outer space or the human body), new methods and materials must be developed to manufacture such integrated sys-tems that will endure post-processing, environmental, and operational challenges.Typical organic-based packaging materials (e.g., polymer adhesives, coatings, and molding materials) often suffer from outgassing and leaching that can lead to system contamination, as well as coefficient of thermal expansion (CTE) mismatches that can lead to warpage and breakage with fluctuations in system temperature during operation. This work demonstrates an alternative, by using a silicate-based inorganic glass composite as an electronics packaging material for stability in extreme environments. Combining liquid alkali sodium silicate (water glass) and nanoparticle fillers, composites can be synthesized and cured at low temperatures into chemically, mechanically, and thermally (up to 400oC) stable structures using high-throughput processing methods such as spin and spray coating. Further, this material can be processed into thick layers (10s to 100s of microns), fill high aspect ratio gaps (13:1), withstand common microfabrication processes, and have its CTE tailored to match various subs |
Rethinking Plant-Based Materials Production: Selective Growth of Tunable Materials via Cell Culture | Current systems for plant-based materials production are inefficient and place unsustainable demands on en-vironmental resources. Traditionally cultivated crops present low yields of industrially useful components and require extensive post-harvest processing to re-move extraneous portions of the plants. Large-scale monoculture remains the unchallenged standard for biomass production despite the negative impacts of the practice to the surrounding biome as well as a suscepti-bility to season, climate, and local resource availability. This work proposes a novel solution to these shortcom-ings based on the selective cultivation of useful, tun-able plant tissues using scalable, land-free techniques. By limiting biomass cultivation to only desirable plant tissues, ex planta farming promises to improve yields while reducing plant waste and competition for arable land. Employing a Zinnia elegans model system, we provide the first proof-of-concept demonstration of isolated, tissue-like plant material production by way of gel-mediated cell culture. Parameters governing cell development and morphology including hormone concentrations, medium pH, and initial cell density are optimized and implemented to demonstrate the tunability of cultured biomaterials at cellular and macroscopic scales. Targeted deposition of cell-doped, nutrient-rich gel scaffolds via injection molding and 3D bioprinting enable biomaterial growth in near-final form (Figure 1), reducing downstream processing requirements. These investigations demonstrate the implementation of plant cell culture in a new application space, propose novel methods for quantification and evaluation of cell development, and characterize morphological developments in response to critical culture parameters—illustrating the feasibility and potential of the proposed techniques.The proposed concept of selectively grown, tunable plant materials via gel-mediated cell culture is believed to be the first of its kind. This work uniquely quantifies and modulates cell development of cultured primary plant products to optimize and direct growth of plant materials. |
Absolute Blood Pressure Waveform Monitoring using Philips Ultrasound Probe | In an Intensive Care Unit (ICU), physicians use an inva-sive radial catheter to measure blood pressure (BP) to track the hemodynamic status of the subject, and these measurements are neither easy nor feasible to perform outside an ICU environment. In such non-ICU settings as a step-down clinical ward or an outpatient clinic, clinicians prefer to use a non-invasive arm-cuff device to measure BP. Even though these measurements are convenient, these devices cannot record the absolute BP (ABP) waveform. Hence, strong interest exists in developing a non-invasive device to monitor the ABP waveform as a quantitative option to perform rapid he-modynamic profiling of patients who cannot undergo invasive BP measurements. This project uses a Philips ultrasound-based transducer (XL-143) to measure BP from superficial arteries (carotid and brachial) proximal to the heart. We measure the arterial diameter and blood flow velocity waveforms from these arteries; an algorithm computes BP from this data in three stages, as illustrated in Figure 1. The algorithm uses the arterial area (A) calculated from arterial diameter and the blood flow velocity (F) waveforms to estimate the height of the ABP waveform, known as pulse pressure (PP), via standard fluid dynamics principles. Further, the algorithm uses a transmission line model of the human vasculature to estimate the mean arterial pressure (MAP). |
Electrochemical Neuromodulation Using Electrodes Coated with Ion-Selective Membranes | Developing precise and effective means of modulating the nervous system is a major challenge in neural pros-theses. While modalities such as deep brain stimula-tion (DBS), vagus nerve stimulation, and electric acous-tic stimulation (EAS) for cochlear implants are finally being realized on the clinical level, there still remains work to be done with respect to our ultimate goal. In the Micro/Nanofluidic BioMEMS research group, we are developing a type of electrode modified with an ion-selective material that can change the concentra-tion of chemicals around a nerve, which will enhance the level of control compared to traditional electrical stimulation.A type of material called the ion-selective membrane (ISM) has been used in the field of analytical chemistry for decades to measure ion concentrations. These membranes are composed of a polymer matrix modified with a chemical called an ionophore, which makes them selective to a particular ion species. In work published by our group, the functionality of these electrodes was inverted, using them for electrochemical stimulation in ex vivo studies of a frog sciatic nerve (see Figure 1 from Song et al.). As a continuation of this work, we are: (1) developing computational models that describe and predict physical behavior of chemical transport from galvanostatic operation of polymeric neutral-carrier based ion-selective membrane electrodes, (2) fabricating and characterizing practical devicesfor implementing ISM-based neuromodulation(see Figure 2 from Flavin et al.), and (3) employingprototype devices in in vitro and in vivo animal models. A successful implementation of this work will pavethe way for more advanced operations such as centralnervous system (CNS) intervention. |
Ultrasound-Based Cerebral Arterial Blood Flow Measurement | Ultrasound-based cerebral blood flow (CBF) monitor-ing is vital in the diagnosis and treatment of a variety of acute neurologic conditions. While flow velocity can be measured using Doppler ultrasound, accurate CBF measurement is difficult as vessel diameters cannot be determined reliably due to acoustic aberrations in-troduced by the skull and because cranial attenuation necessitates low frequency (1-2 MHz) insonation with poor spatial resolution.We have developed a CBF estimation technique that achieves the spatial resolution required for CBF determination by estimating the point spread function of the imaging system. The received data are then deconvolved to increase spatial resolution, and a correction is applied to account for cranial aberrations. Doppler data were collected from phantom blood vessels with diameters between 2 and 6 mm over a 150-mL/min range using a clinical ultrasound device.Our method achieved an RMSE of 26 mL/min, withinacceptable range for cerebral perfusion monitoring atthe bedside. |
Force-Coupled Ultrasound for Noninvasive Venous Pressure Assessment | Congestive heart failure is a clinical syndrome that affects about 6 million people and accounts for about 1 in 9 deaths in the United States. In this condition, the pumping ability of the heart decreases, causing a buildup of blood volume and pressure in the venous system as it returns blood to the heart. This buildup further decreases the pumping ability of the heart by over-stretching its ventricles. Additionally, increased venous pressure can lead to fluid migrating from the veins to the interstitial space, which is called edema. Left unchecked, edema can lead to death. Proper ad-ministration of diuretic drugs can allow venous pres-sure to drop back down by lowering intravascular vol-ume, which will improve a patient’s condition. However, thus far, only invasive catheterization can produce an accurate and reliable venous pressure measurement. Our goal is to produce an accurate, noninvasive means of assessing venous pressure by means of force-coupled ultrasound. By positioning our force-coupled ultrasound probe at the base of the neck, we can observe the compression of the internal jugular vein, which returns blood from the cerebral vasculature to the heart. Unlike in the case of an artery, we can safely observe compression from zero force all the way to complete occlusion of the vein. We can also observe compression of the internal jugular vein while increasing its pressure with the Valsalva maneuver, exhalation against a closed airway, and while decreasing its pressure by elevating it above the supine position. We expect these observations to give us excellent insight for our computational models to accurately assess venous pressure. |
An Electrokinetic-Based Concentrator for Ultra-Low Abundant Target Detection | The recent COVID-19 outbreak has sparked urgent in-terest in rapid and reliable viral identification. In fact, this is a recurring challenge in many other pathogen detection and diagnostics, where only a few target vi-ruses or bacterial cells are present in milliliters or even liters of volume, necessitating that a large volume of the sample must be concentrated for the targets to be introduced into the downstream detection system. Un-fortunately, due to the size of the virus or biomolecule, concentrating or retrieving the virus or biomolecule with a filter, ultracentrifuge, or any kind of method is extremely hard. Figure 1 (a) shows the purpose of this work and the overall concept. This technology is based on microfluidic devices that couple microchannel and cation exchange membrane (CEM) to play an electrophoretic force off a hydraulic drag force to enable charge-based concentration, without any physical filter. Under the electric field, the virus experiences the electrophoretic force and hydraulic drag force at the same time. The electrophoretic force is driven by the intensive electric field focused near the CEM while drag force is driven simply by the hydraulic flow. Efforts are being made to build electrokinetic concentrators using materials and processes that are more robust and scalable than those of traditional microfluidics. Instead of using polydimethylsiloxane (PDMS) that is patterned using photolithography, one can laser etch channels and ports into acrylic polymethyl methacrylate (PMMA). Thin adhesive films can have custom patterns cut into them using a digital die cutter and then be used to bond PMMA layers and seal channels. Designing the device in manner seen in Figure 1b also allows the use of ion- exchange membranes that are commonly used in electrodialysis and fuel cell systems, meaning these materials are robust and relatively inexpensive. |
Micro/Nanofluidic Technologies for Next-Generation Biomanufacturing | Biomanufacturing of therapeutic proteins and vaccines is crucial for modern medicine. Recently, the biophar-maceutical industry started to focus more on process intensification through continuous biomanufacturing. New therapeutic modalities such as cell and gene thera-pies are rapidly emerging as well. Accordingly, it has be-come increasingly important for biomanufacturers to improve manufacturing efficiency, quality, and safety. Compared to conventional biomanufacturing technol-ogies, micro/nanofluidic technologies can contribute to the improvement with their unique advantages. Here, we introduce our new micro/nanofluidic technologies for efficient, high-quality, and safe biomanufacturing. First, we developed spiral microfluidic devices for reliable and efficient perfusion culture and adventitious agent (AA) clearance. The devices enable size-based cell sorting without any physical barriers, so that mammalian cells can be continuously separated from cell culture. Using this feature, the spiral device was used for 1) cell retention for perfusion culture and 2) rapid AA clearance (Figure 1). This microfluidic technology could overcome the limitations (biofouling, cell damage) of conventional cell separation techniques (e.g., membrane-based filtration, centrifugation).Second, we introduce a new nanofluidic device for monitoring critical quality attributes (purity, binding affinity, glycosylation, etc.) of antibody therapeutics during biomanufacturing. The device has a nanofilter array and enables continuous-flow size or charge-based protein separation. Using this device, we demonstrated a fully automated continuous online protein-size monitoring during continuous perfusion culture. We are currently expanding the capability of the nanofluidic device to monitor binding affinity and glycosylation of antibodies at real-time speed (Figure 2). The technology could complement conventional protein-quality-monitoring equipment while producing a large amount of information about biologics quality. |
Measuring Eye Movement Features using Mobile Devices to Track Neurodegenerative Diseases | Current clinical assessment of neurodegenerative dis-eases (e.g., Alzheimer’s disease) requires trained special-ists, is mostly qualitative, and is commonly done only intermittently. Therefore, these assessments are affect-ed by an individual physician’s clinical acumen and by a host of confounding factors, such a patient’s level of attention. Quantitative, objective, and more frequent measurements are needed to mitigate the influence of these factors. A promising candidate for a quantitative and accessible diseases progression monitor is eye movement. In the clinical literature, an eye movement is often measured through a pro/anti-saccade task, where a subject is asked to look towards/away from a visual stimulus. Two features are observed to differ significantly between healthy subjects and patients: reaction time (time difference between a stimulus presentation and the initiation of the corresponding eye movement) and error rate (the proportion of eye movements towards the wrong direction). However, these features are commonly measured with high-speed, IR-illuminated cameras, which limits accessibility. A portable measurement system is required to track them longitudinally. Previously, we enabled ubiquitous tracking of eye-movement features by enabling app-based measurements of visual reaction time and error rates. In this work, we further show how we learn potential trends in these eye-movement features using Gaussian process modeling. Such modeling has allowed us to discover subjects’ task-performing strategies such as trading off between speed and accuracy. We hope that once we have collected data from patients, we can use the model to a) compare the trends of the features with the clinical assessments, b) distinguish the effect of strategies from the effect of disease progression, and c) evaluate the potential to use our system to track disease progression more frequently and widely than previously possible. |
A Comparison of Microfluidic Methods for High-Throughput Cell Deformability Measurements | The mechanical phenotype of a cell is an inherent bio-physical marker of its state and function, with many applications in basic and applied biological research. Microfluidics-based methods have enabled single-cell mechanophenotyping at throughputs comparable to those of flow cytometry. As shown in Figure 1, we present a standardized cross-laboratory study com-paring three microfluidics-based approaches for mea-suring cell mechanical phenotype: constriction-based deformability cytometry (cDC), shear flow deforma-bility cytometry (sDC), and extensional flow deform-ability cytometry (xDC). All three methods detect cell deformability changes induced by exposure to altered osmolarity. However, a dose-dependent deformability increase upon latrunculin B-induced actin disassembly was detected only with cDC and sDC, which suggests that when cells are exposed to the higher strain rate imposed by xDC, cellular components other than the actin cytoskeleton dominate the response. The direct comparison presented here furthers our understand-ing of the applicability of the different deformability cytometry methods and provides context for the inter-pretation of deformability measurements performed using different platforms. |
Electronics for Transparent, Long-Lasting Respirators | The use of personal protective equipment (PPE), includ-ing the N95 respirators and surgical masks, is essential in reducing airborne disease transmission, particularly during the COVID-19 pandemic. Unfortunately, there has been a shortage of PPE since the beginning of the pandemic. Also, the available N95 masks have major limitations, including masking facial features, waste, and lack of integrity after decontamination, forcing re-searchers to find alternatives.This work presents a transparent, elastomeric, adaptable, long-lasting respirator with an integrated biometric interface. The mask is made mostly of silicon rubber and comes with two replaceable filter cartridges. The electronic interface uses one of the filter insert locations to measure temperature, humidity, pressure, and air quality. The system uses Bluetooth Low Energy and sends real-time sensor data to a phone or a computer. The data can be used to inform the user regarding mask fit, fatigue, mask condition, and potential diagnostic information. |
Self-Editing or “Lamarckian” Genomes Using the Bio/Nano TERCOM Approach | Gene editing has been an area of active investigation for many decades. Some approaches introduce per-manent edits; others modify expression. In this work, conceptually, cells or cell-free reactions estimate their location by correlating the evolution of their sensed fluid environment (e.g., temp., salinity, sugar, pH, ion concentration, etc.) against an embodied map and then self-edit the content of their genomes in a way that depends on said estimate; editing the genome shifts the expressed phenotype and the heritable genotype. This approach is related to terrain contour matching (TERCOM), a technique used in air navigation. Current efforts focus on a reaction mixture containing a plas-mid that experiences path-dependent self-edits while en route to a target site. As envisioned (see Figure 1), a read-only so-called “junk DNA” segment of a plasmid transcribes into mRNA strands having coding heads and consumable tails; the tails are attacked by an ex-onuclease, the activity of which depends jointly on re-moved monomer species and local ion concentration (or another environmental variable), causing the tails to function as path-sensitive fuses and the mix of sur-viving mRNA to depend on the path. The surviving mRNA is reverse-transcribed into DNA and integrated as expressible genes in a read-write portion of the plas-mid; concurrent random erasures keep overall length roughly constant. In this process, the genetic composi-tion of the read-write region evolves with the changing environmental path. A related heritage effort explores drug delivery using particles that exhibit path-depen-dent doses or conformation. The current and heritage efforts build on prior study by the PI and his group of nanoparticles that record the trajectory of their environment. An experimental apparatus has been designed to test these various TERCOM-like reaction mixtures. Progress on the present effort may allow the engineering of organisms that exhibit Lamarckian evo-lution or gene therapies that confer this ability. |
Balancing Actuation Energy and Computing Energy in Motion Planning | Inspired by emerging low-power robotic vehicles such as insect-size flyers, high-endurance autonomous blimps, and chip-size satellites, we identify a new class of motion-planning problems in which the energy consumed by the computer while planning a path can be as large as the energy consumed by the actuators during the execution of the path. Figure 1 shows how the energy to move one meter on various low-powered robotic platforms is of a similar magnitude to the ener-gy to compute one second on various embedded com-puters. As a result, minimizing energy requires mini-mizing both actuation energy and computing energy since computing energy is no longer negligible. Figure 2 shows average actuation energy and computing energy curves for a selected robotic platform and a computing platform. Here, minimizing only actuation energy, as is conventionally done, does not minimize total ener-gy. Instead, stopping computing earlier and accepting a higher actuation energy cost for a lower computing energy cost minimizes total energy.We propose the first algorithm to address this new class of motion planning problems, called Computing Energy Included Motion Planning (CEIMP). CEIMP operates similarly to other anytime planning algorithms, except that it stops when it estimates that while further computing may save actuation energy by finding a shorter path, the additional computing energy spent to find that path will negate those savings. We evaluate CEIMP on realistic computational experiments involving 10 MIT building floor plans, and CEIMP outperforms the average baseline of using maximum computing resources. In one representative experiment on an embedded CPU (ARM Cortex A-15), for a simulated vehicle that uses one Watt to travel one meter per second, CEIMP saves 2.1-8.9x of the total energy on average across the 10 floor plans over the baseline, which translates to missions that can last equivalently longer on the same battery. |
Absolute Blood Pressure Measurement using Machine Learning Algorithms on Ultrasound-based Signals | Hypertension, or high blood pressure (BP), is a major cardiovascular risk factor. Therefore, measuring BP is of significant clinical value. At present, there are a few disadvantages for devices that measure a patient’s BP. For instance, in an Intensive Care Unit (ICU), physicians use an invasive radial catheter to measure BP, which is not feasible outside an ICU. In non-ICU settings, clini-cians use a non-invasive arm-cuff device to measure BP. This is convenient but can provide only a systolic and a diastolic pressure value and does not output the abso-lute BP (ABP) waveform. These devices also neglect the dynamic nature of the arterial system as they do not measure the morphology of the BP waveform, which may contain information on the underlying patho-physiology.In this work, we propose a non-invasive way to get BP waveform with blood flow velocity and arterial area obtained from non-invasive ultrasound signals. One key drawback of the ultrasound-based device is that the output BP waveform has an arbitrary reference, so we have to estimate the mean arterial pressure (MAP). We propose to use a machine learning model containing 1D convolution and Transformer encoder layers to regress the MAP accurately. The input features are arterial area, flow velocity, and several other scalar features such as pulse wave velocity and pulse pressure. They are first embedded into a 512-dimension vector. Then, the convolution layers perform feature extractions, and a transformer models the relationship between time steps. We perform the training on the Pulse Wave Database (PWDB) synthetic dataset and test on seven real patients. The model provides accurate results, with mean absolute error 2.6 mmHg and std 2.1 mmHg. This algorithm has large potential to make affordable BP waveform measurements accessible to everyone. |
Analytical and Numerical Modeling of an Intracochlear Hydrophone for Fully Implantable Assistive Hearing Devices | Cochlear implants with fully implantable microphones would allow directional and focused hearing by taking advantage of ear mechanics. They would be usable in almost all environmental conditions throughout the day and night. Current implantable microphones suf-fer from unstable mechanics, poor signal-to-noise ratio (SNR), and low bandwidth.In this work, we used analytical modeling, a finite element model, and experiments to design a polyvinylidene (PVDF) intracochlear hydrophone for high-bandwidth sensitivity, surgical viability, and improved SNR by electrical shielding and circuit design. Our analysis shows that the copolymer PVDF-TrFE should be used due to its higher hydrostatic sensitivity, the area of the sensor should be maximized to maximize gain, and the length should not exceed a maximal value determined by the bandwidth requirement. A short-circuit topology charge amplifier maximizes the SNR of the sensor by minimizing noise and attenuating electromagnetic interference by shielding. These advances in sensor performance bring fully implantable systems closer to reality. |
Fluorescent Janus Droplet and Its Application in Biosensing of Listeria Monocytogenes | Dynamic complex droplets afford versatile platforms for biosensing. The biosensing methods based on drop-lets enable a combination of advantages including speed, cost-effectiveness, and portability. This research explores a sensing method based on the agglutination of Janus emulsions for Listeria monocytogenes, which is a gram-positive bacterium and is responsible for a potentially lethal foodborne bacterial illness. We create a bio-recognition interface between the Janus emul-sions that comprises equal volumes of hydrocarbon and fluorocarbon oils in Janus morphology by attach-ing antibodies to a functional surfactant polymer with a tetrazine/trans-cyclooctene (TCO) click reaction. The Listeria antibodies would be on the surface of the hy-drocarbon hemisphere since the surfactant will stay at the interface of the hydrocarbon and water phase. Ag-glutinations of Janus droplets are formed when Liste-ria is added because of the strong binding between Listeria and the Listeria antibody located at the hydro-carbon surface of the emulsions. By incorporating one emissive dye in the fluorocarbon phase and a blocking dye in the hydrocarbon phase of Janus droplets, we conduct a two-dye assay, which enables the rapid detec-tion of trace Listeria in two hours via an emissive sig-nal produced in response to Listeria binding. To clarify, the Janus structures are tilted from their equilibrium position as a result of the formation of agglutinations and produce emissions that would ordinarily be ob-scured by a blocking dye. Overall, this method not only provides rapid and inexpensive Listeria detection with high sensitivity but also can be used to create a new class of biosensors by connecting with other related recognition elements. |
Dance-Inspired Investigation of Human Movement | This research focuses on efforts to formalize a dancer’s approach to movement. The overarching hypothesis is that dancers stabilize their joints through stretches – which are observed during common activities such as walking and running. However, most untrained indi-viduals are able to apply this form of stabilization only during activities such as walking that seemingly “just happen,” much as we “see.” In contrast, the best dancers and athletes are able to generalize this stretch-based joint stabilization beyond walking to their art form. To understand how dancers organize movement through stretches, the researchers use motion tracking and elec-tromyography. This work will potentially benefit sever-al fields, including soft robotics, neuroscience, and AI. |
Nanoscale Insights into the Mechanisms of Cellular Growth and Proliferation | The growth and proliferation of human cells are con-trolled by the large molecular machine called mTORC1 that acts as a molecular equivalent of an AND logic gate. mTORC1 integrates multiple environmental sig-nals, such as nutrients and growth factors, and orders the cell to either grow and divide in times of plenty or stand by and recycle when nutrients are scarce. Using electron cryomicroscopy, we revealed how mTORC1 recognizes nutrient signals, which provided a na-noscale-precision blueprint for the design of therapies aimed at deregulated mTORC1 in diseases of cellular growth, such as cancer. |
A Polarization-Encoded Photon-to-Spin Interface | The central goal of quantum communication is to de-liver quantum information in a way that is resilient against eavesdropping. One notable approach is the measurement-device-independent quantum key dis-tribution (MDI-QKD) protocol, in which a secret key is shared between two parties connected by quantum and classical channels. Essential to this architecture, however, is the ability to faithfully transfer quantum states between two distant qubits. Here, we propose an integrated photonics device for mapping qubits encod-ed in the polarization of a photon onto the spin state of a cavity-coupled artificial atom: a “polarization-encod-ed photon-to-spin interface” (PEPSI). We perform theo-retical analysis of the state fidelity’s dependence on the device’s polarization extinction ratio and atom-cavity cooperativity. Furthermore, we explore the rate-fidelity trade-off through analytical and numerical models. In simulation, we show that our design enables efficient, high-fidelity photon-to-spin mapping. |