Faculty: Prof. Martin Herbordt
Mon/Wed 2-4PM

This is an incredibly exciting time in computer engineering with processors getting very complex, but few programmers having the knowledge to use them effectively. Although GPU acceleration has received much attention, there is (surprisingly) an even more dire need with respect to efficiently using multicore and even single core.

Prof. Hatice Altug

An interdisciplinary team of researchers led by Department of Electrical & Computer Engineering Assistant Professor Hatice Altug has created a highly sensitive, infrared (IR) absorption spectroscopy technique that can identify specific proteins and other molecules using far less sample material than what conventional spectrometers require. Exploiting recent advances in nanophotonics, the technique constitutes a powerful new tool for biomolecular studies and drug discovery, and could considerably enhance biological and chemical sensor detection capabilities.

Infrared absorption spectroscopy uses infrared light to excite the bonds that connect atoms within molecules, causing them to vibrate at a specific resonant frequency. By examining what frequencies of light are absorbed by a material, scientists can determine what kind of bonds it contains, and thus identify the material.

Because absorption signals are often weak, conventional IR spectroscopy requires large samples of target molecules in many layers. To overcome this limitation, the research team used tiny gold nanoparticles as highly efficient “nanoplasmonic” antennas that greatly amplify the signal received from an individual protein molecule.

“Our technique enhances the signal by a factor of up to 100,000,” said Altug.  “Because our technique is ultra-sensitive, we don’t need a large number of molecules from which to obtain signals. In fact, we can obtain signals from even a single-molecule-layer thick protein film.”

Altug and her collaborators—Professor Shyamsunder Erramilli (BME, Physics); Research Professor Mi Hong (Physics); graduate student Ronen Adato and post-doctoral fellow Ahmet Ali Yanik in Altug’s lab; and Tufts University bioengineers David Kaplan, Fiorenzo Omenetto and Jason Amsden—report on this unprecedented achievement in this week’s online edition of Proceedings of the National Academy of Sciences.

Nanoantennas Dramatically Improve Detection Capability

To obtain the high sensitivity needed to detect vibrations from an extremely small sample of silk protein molecules, the team designed a 50-by-50 array of gold, rod-shaped nanoantennas and tuned their resonant frequency to match that of the bonds within the sampled molecules.

The 2,500 strategically configured antennas focus infrared light on nearly 145 silk protein molecules deployed at the tip of each nanoantenna. The light, in turn, excites the bonds within the molecules to vibrate at their signature 6.6 micron wavelength. After absorbing a significant fraction of the incoming IR light, the silk protein molecules reflect the rest back through the nanoantennas. Upon receipt of the reflected signal, the spectrometer deduces the vibrational signature of the silk protein molecules.

Combining theoretical calculations and advanced nanofabrication techniques, Yanik and Adato obtained up to a 100,000-fold enhancement of the molecules’ vibrational signatures and whittled the sample thickness down to a single layer of protein.

Drawing on seed funding from an ENG Dean’s Catalyst Award and ongoing support from the National Science Foundation, Massachusetts Life Science Center and Department of Defense, Altug and her co-investigators are now applying their novel IR spectroscopy technique to other kinds of molecules.

“Our plasmonic method is quite general and can be adapted to enhance the infrared fingerprints of other biomolecules, such as nucleic acids and lipids,” said Altug. “It therefore provides a general purpose toolkit for ultra-sensitive vibrational spectroscopy of biomolecular systems.”

Drug Discovery Implications

Because the technique requires only one-layer, two-nanometer-thick samples, it may ultimately enable scientists to obtain much more accurate and useful data.

“The sensitivity of our technique can be high enough to provide spectroscopy at the single-molecule scale,” said Altug, “and a single-molecule response can be very different from that of an ensemble of molecules.”

Studying protein molecules in one layer offers yet another advantage.

“Conventional IR spectroscopy requires a large number of proteins, usually 5,000 to 10,000 layers of them in one stack that resembles a baklava,” said Erramilli. “With our single-layer substrate we can capture proteins in their native environment.”

As a result, the new technique could be used to improve our understanding of how protein molecules interact and how external forces alter their shape and behavior — questions of fundamental importance in biochemistry and drug discovery.

The method may also help amplify biological and chemical sensing capabilities in defense and other applications.

“Chemical sensors detect the presence of specific molecules via molecular fingerprints, telltale vibrational frequencies of the molecules’ bonds,” Altug explained. “Our technique’s ultra-sensitivity enables us to pick up clear, identifiable response signals even from a trace amount of a chemical.”

Prof. Lev Levitin

Prof. Tommaso Toffoli

Constant advances in technology may make it seem like computing power has infinite potential, but two ECE professors have determined that one day processors will hit an ultimate speed limit.

Department of Electrical & Computer Engineering Distinguished Professor Lev Levitin and Research Professor Tommaso Toffoli recently published new research in Physical Review Letters which shows that while breakthroughs in hardware and quantum computing will continue to produce faster processing speeds for many years, we will one day hit a fundamental ceiling.

“No system can overcome that limit. It doesn’t depend on the physical nature of the system or how it’s implemented, what algorithm you use for computation—any choice of hardware and software,” Levitin said. “This bound poses an absolute law of nature, just like the speed of light.”

Levitin’s research had earlier identified a quantum elementary operation, the simplest possible task that could be performed by a quantum computer. Using this idea, Levitin and Toffoli derived an equation that represents the minimum amount of time it takes for such an operation to occur, for a given amount of energy invested. As quantum computing represents the theoretical limit of computing performance, this time interval illustrates the fastest possible operation rate that could ever be achieved by any computer.

For perspective, contemporary computers typically perform around 10^10 elementary operations per second, while an ideal computer would breeze through 10^20 operations using the same amount of time and energy—effectively making it 10 billion times faster.

While reaching this limit is hindered by current technological limitations, Levitin and Toffoli anticipate that we could hit this barrier within this century.

“If we believe in Moore’s law, it would take only 50 years to achieve this quantum limit.”

Prof. Ayse Coskun

Department of Electrical & Computer Engineering Assistant Professor Ayse Coskun recently received the “Best Paper Award” at the 17th IFIP/IEEE International Conference on Very Large Scale Integration (VLSI-SoC), 2009. The conference, which explores developments in the field of Very Large Scale Integration Systems and their designs, was held October 12-14 in Florianópolis, Brazil.

Coskun’s research was performed in collaboration with Jose L. Ayala (Compultense University of Madrid, Spain), David Atienza (EPFL, Switzerland), and Tajana Simunic Rosing (UC San Diego).

Paper Abstract

Modeling and Dynamic Management of 3D Multicore Systems with Liquid Cooling

Three-dimensional (3D) circuits reduce communication delay in multicore SoCs, and enable efficient integration of cores, memories, sensors, and RF devices. However, vertical integration of layers exacerbates the reliability and thermal problems, and cooling efficiency becomes a limiting factor. Liquid cooling is a solution to overcome the accelerated thermal problems imposed by multi-layer architectures. In this paper, we first provide a 3D thermal simulation model including liquid cooling, supporting both fixed and variable fluid injection rates. Our model has been integrated in HotSpot to study the impact on multicore SoCs. We design and evaluate several dynamic management policies that complement liquid cooling. Our results for 3D multicore SoCs, which are based on 3D versions of UltraSPARC T1, show that thermal management approaches that combine liquid cooling with proactive task allocation are extremely effective in preventing temperature problems. Our proactive management technique provides an additional 75% average reduction in hot spots in comparison to applying only liquid cooling. Furthermore, for systems capable of varying the coolant flow rate at runtime, our feedback controller increases the improvement to 95% on average.

Prof. Luca Dal Negro

Department of Electrical & Computer Engineering Assistant Professor Luca Dal Negro was recently awarded $479,997 in funding for his research proposal, “Deterministic Aperiodic Structures for On-chip Nanophotonics and Nanoplasmonics Device Applications.” The grant, which spans four years, was awarded by the US Air Force Office of Scientific Research.

Research Proposal:

Understanding the role of aperiodic order in optical systems provides unique opportunities for the engineering of light dispersion, density of states fluctuations, radiation patterns, frequency spectra and localized field states in complex non-periodic photonic-plasmonic nanostructures created by simple mathematical rules and amenable to predictive theories. The specific objective of this research project is to design, explore and engineer a novel class of Si-based, photonic-plasmonic planar structures capable of providing ultra broadband (0.4μm-1.8μm) enhancement of nanoscale-localized optical fields and nonlinear processes in metal-dielectric Deterministic Aperiodic Nano Structures (DANS). Differently from conventional crystal structures, which are limited by translational invariance symmetry, the proposed research approach focuses on the design and fabrication of on-chip localized field states by Fourier-space engineering in photonic-plasmonic nanostructures with controllable degree of aperiodic order and non-crystallographic point symmetries (i.e. high degree of rotational invariance, statistical symmetries, multi-fractal scaling).

In this project, we will engineer DANS in the form of metal-dielectric nano-particles, nano-pillars, and nano-holes arrays fabricated on Si substrates by Electron Beam Lithography (EBL). All the proposed structures are amenable to inexpensive pattern replication on the wafer scale by using nano-imprint lithographic techniques, thus reducing the costs associated to EBL nanofabrication. DANS structures will be fabricated with varying degree of structural complexity (ranging from quasi-periodicity to pseudo-randomness) combined with statistical and higher-order rotational symmetries, not yet explored in the field of photonics. These structures will be designed using a number of rigorous analytical and numerical techniques including multi-particle scattering theories (GMT, T-matrix theory, Discrete Dipole Approximations, FDTD, Finite Elements) and will be experimentally characterized using dark-field scattering, micro-Raman, time-resolved emission and non-linear optical spectroscopy.

In relation to novel device synthesis and characterization, this project will explore the science and technological opportunities of DANS in relation to: (a) broadband enhancement and control of far-fields and radiative processes; (b) modification of non-linear optical responses in quantum dots; (c) engineering of angle-insensitive, broadband scattering substrates for plasmon-enhanced solar cells; (d) engineering of strongly-coupled photonic-plasmonic optical sensors.

Prof. Paschalidis

Prof. Cassandras

Department of Electrical & Computer Engineering Professors Ioannis Paschalidis and Christos Cassandras were recently awarded $360,000 in funding for their research proposal, “Distributed Wireless Sensor Networks for Long-Term Deployments.” The grant, which spans two years, was awarded by the US Department of Energy and is supplemental funding to an existing grant as part of the 2009 stimulus act.

Research Proposal

The proposed research project aims at addressing a number of fundamental issues in Distributed Wireless Sensor Networks (DWSN), thus, contributing to the maturing of this technology so that it can be useful in long-term surveillance missions. First, recognizing the limited resources available to DWSN nodes, we will pursue a number of novel optimization-based control approaches aiming at (i) maximizing network throughput using joint routing, power control, and transmission scheduling policies, (ii) reducing network latency, (iii) maximizing network lifetime through dynamic control schemes, (iv) performing critical self-organization tasks, and (v) minimizing node energy consumption while guaranteeing real-time constraints. Second, we will develop localization and tracking capabilities in DWSNs where most nodes do not have access to a GPS system. In conjunction, we will develop strategies for coverage control, where (possibly mobile) DWSN nodes coordinate to position themselves in a given surveillance region so as to maximize the probability of detecting a variety of events. These main objectives will be aided by a number of supporting research tasks including (i) simulation-based on-line parameter adaptation methods enabling a network to automatically adjust crucial parameter settings in order to maintain a desirable level of performance; and (ii) statistical anomaly detection providing the capability to identify non-typical patterns in the set of environment variables the DWSN may monitor. Throughout our proposed research work, we will seek techniques that are scalable, implementable in distributed fashion, and computationally compatible with the limited processing capabilities of most network nodes we expect to be dealing with. The project will capitalize on the accumulated knowledge and resources of the Boston University Sensor Network Consortium (SNC) that the PIs have spearheaded and on a recently established small robot laboratory testbed with wireless sensing capabilities to implement and test several of the approaches to be developed. The proposing team commits to closely collaborate with Los Alamos National Laboratory (LANL) researchers working on a DOE funded project entitled Distributed Sensor Networks with Collective Computation.

Moresco’s paper, “Full-Band Monte Carlo Simulation of HgCdTe APDs [avalanche photodetectors],” aims to provide a theoretical model of the operation of both e-APD and h-APD that is as much as possible free from fitting parameters and can be used to predict the multiplication gain and noise factor of devices with arbitrary structures. Past theoretical studies have evaluated the performance of electron-initiated APD, but these studies relied on a significant number of assumptions and fitting parameters that require a substantial amount of experimental information.

APDs based on HgCdTe have been the focus of a significant research effort because of the potential use in many imaging applications in the short- and mid-infrared spectral region. The most attractive feature of these devices is the potential for single carrier ionization when electrons are used as the primary injection carrier.

This research was conducted under the supervision of ECE Associate Professor Enrico Bellotti and in collaboration with ECE research associate Francesco Bertazzi, ECE research assistant Michele Penna, and Professor Michele Goano of the Politecnico di Torino.

The workshop was held from October 6-8, 2009 in Chicago.

A printed version of the publication is available in the Department’s main office.

For more information, please see our employment webpage.

Prof. Luca Dal Negro

The National Science Foundation (NSF) has named ECE assistant professor Luca Dal Negro as the recipient of a 2009 Faculty Early Career Development (CAREER) Award. Spread over five years, the CAREER Award supports junior faculty who exemplify the role of teacher-scholars through outstanding research, excellent education and the integration of education and research within the context of their institution’s mission.

Dal Negro’s CAREER award supports his project, “Combined Light and Carrier Localization in High-refractive Index Silicon Nanocrystal Structures: a Novel Approach for Si-based Lasers,” which aims to advance nanophotonics by developing a new class of light-emitting devices that will enable unprecedented control and enhancement of certain light-emitting elementary particles. His research aims to enhance electromagnetic fields to the point where the radiation properties of the silicon nanostructures can be changed to efficiently and rapidly emit light for on-chip light emitters and lasers applications.

“We are doing this within fully silicon-based technology,” Dal Negro said. “It’s an exciting project for us, and allows us to take the ultimate advantages of field localization to better engineer the spontaneous and stimulated emission properties of silicon nanostructures.”