Distinguished Lectures Series
The ECE Distinguished Lectures Series brings world-class researchers to the University to share their research and discoveries.
Knowledge is Power but Ignorance is Bliss: Theory and Methods for Asynchronous (QDI) Logic
Alain J. Martin
The use of a global clock to enforce sequencing in a VLSI implementation of a digital computation relies on a knowledge of timing (gate and wire delays) that constitutes both an advantage and a drawback. Its advantage is simplicity. But as continuing circuit miniaturization (“scaling”) makes this knowledge less and less precise, the drawbacks are starting to outweigh the advantages.
In this talk, Professor Martin will present an overview of asynchronous logic, a design approach for digital VLSI that does not use clock. He will show that delay insensitivity—-i.e. the complete ignorance of delays — is impossible. Instead, he will introduce the concept of “quasi-delay-insensitivity” (QDI), in which a minimal timing assumption, the “isochronic fork” is sufficient for implementing any computation. Martin will briefly describe the several asynchronous microprocessors designed at Caltech and will give evidence of the robustness and efficiency of QDI designs in the presence of large parameter variations.
System Implications of Integrated Photonics
Norman P. Jouppi
Micron-scale photonic devices integrated with standard CMOS processes have the potential to dramatically increase system bandwidths, performance, and configuration flexibility while reducing system power. Small devices have many advantages: reduced power, increased density, and increased speed. By integrating many thousands of these devices on a chip, photonics could potentially be used for most high-speed off-chip and global on-chip communication. Integrated photonics has many advantages at the board and rack scale as well. Recent high-speed board-level electrical signaling (>2.5GHz) precludes the use of multi-drop busses or communication over long distances on ordinary inexpensive PC board materials. By using photonics, high fan-out and high-fan-in bus structures can be built. Due to the low loss of optical signals versus distance, these structures can even be distributed over rack-scale distances. This dramatically increases system flexibility while reducing interconnect power.
As an example of the potential impact of photonics, I describe a system architecture for the 2017 time frame we call Corona. Corona is a 3D many-core architecture that uses nanophotonic communication for both inter-core communication and off-stack communication to memory or I/O devices. Dense wavelength division multiplexed optically connected memory modules provide 10 terabyte per second memory bandwidth. A photonic crossbar fully interconnects its 256 low-power multithreaded cores at 20 terabyte per second bandwidth. We believe that in comparison with an electrically-connected many-core alternative, Corona could provide 2 to 6 times more performance on many memory intensive workloads, while simultaneously significantly reducing power.
Devices for Optical Interconnects to Chips
David A.B. Miller
This talk will summarize recent work in device requirements for optical interconnects, in novel device approaches using germanium quantum wells and nanophotonic and nanometallic structures, and in fundamental limits to optical components.
Power has become a dominant constraint in information processing, and will continue to be for the foreseeable future and beyond. Electrical wires operate at energies of ~ 1 pJ/bit or higher for off-chip interconnects and ~ a few 100 fJ/bit to ~ 1 pJ/bit for global on-chip interconnects. Optics can in principle solve such energy problems while simultaneously allowing higher densities of interconnects, especially for longer off-chip distances. For sufficient benefit, however, total system energy of < 100 fJ/bit will be required for optical interconnects, which implies optical output devices (lasers or modulators) that are efficient at ~ 10 fJ/bit transmitter energies, and optical receivers (photodetectors and associated circuitry) that can operate with < 1 fJ/bit of received energy. These are aggressive numbers for optoelectronics, but not unphysical. Quantum-confined Stark effect (QCSE) quantum well modulators are one possible transmitter approach that could scale to such low numbers. Recent work at Stanford on such modulators using germanium quantum wells on silicon substrates represent an interesting recent development for low energy devices compatible with silicon technology. We have also recently combined Ge with nanometallic antenna structures to make enhanced photodetectors in deeply subwavelength detector elements and to consider other nanometallic structures such as “two-conductor” optical nanometallic waveguides for photodetectors and light concentration. Nanometallic structures are very promising for concentrating light into deeply sub-wavelength optoelectronic devices with high speeds and low capacitances.
For interconnects it will also be important to make optical devices such as wavelength splitters very small. We have been able to model, design, and demonstrate superprism wavelength splitters, which potentially could enable dense wavelength splitters for optical interconnects and other applications. We have also examined fundamental limits to dispersive optical device performance devising a limit that has now also been applied to bound the performance of “slow light” devices. In summary, these and other approaches to novel optical and optoelectronic devices exploiting nanophotonics and quantum confinement are very promising for future applications such as optical interconnects.
Metal Optics, Optical Antennas, and Spontaneous Hyper-Emission
For 50 years, stimulated emission has been stronger and far more important than spontaneous emission. Indeed spontaneous emission has been looked down upon, as a weak effect. Now a new science of enhanced spontaneous emission is emerging, that will make spontaneous emission stronger and faster than any possible stimulated emission. This new science depends upon the use of nanoscale metallic optical elements, as antennas for spontaneous emission.
The overall increase in spontaneous emission rate can be roughly 8 orders of magnitude! Under favorable circumstances the spontaneous emission rate can be comparable to the optical frequency itself, which is unprecedented.
Among the applications will be: (1) Direct modulation of LED’s will extend above 1THz, far faster than the direct modulation speed of any laser. This may define the future of short distance data-communications technology. (2) Materials which do not fluoresce or luminesce, owing to strong non-radiative losses (i.e. most molecules), will now become spectroscopically accessible since their spontaneous emission will now compete favorably with non-radiative losses. This is expected to have revolutionary implications in basic biological research, since a local probe can be inserted into a cell to optically interrogate the molecules at the tip.
The lecture will provide the basic background in metal optics, and in optical frequency antennas required to understand the photo-physics of this new form of light emission.
Solution-Processed Chalcogenide Semiconductors: Opportunities for Electronics Technologies
The effort to identify low-cost, high-throughput solution-based deposition techniques for thin-film field-effect transistors (TFTs), solar cells and related electronic devices has generated substantial interest in recent years, as a result of new applications potentially enabled by the alternative technologies (e.g., flexible displays, electronic newspapers, smart cards/fabric, large-area solar cells).
While much work in this field has focused on organic semiconductors, this talk will address a solution-based approach for processing inorganic metal chalcogenide semiconductors that relies on the low-temperature decomposition of highly soluble hydrazine-based metal chalcogenide precursors. The resulting metal chalcogenide films may be only a few unit cells thick, with large field-effect mobilities (up to 15 cm2/V-s) for use in both n- and p-channel TFTs. Phase change materials, useful for rewritable optical media, have also been demonstrated, while thicker metal chalcogenide films have been solution deposited as solar cell absorber layers, yielding devices with up to 13% power conversion efficiency.
As an example, in the solar arena, only two metal-chalcogenide-based thin film photovoltaic technologies (cadmium telluride and Cu(In,Ga)(Se,S)2, also known as CIGS) had been able to offer power conversion efficiencies above 9%. We have recently shown world record performance – 9.6% efficiency – in a solution-processed kesterite-based device, in which the relatively rare and/or expensive In and Ga in CIGS have been replaced with plentiful relatively cheap Zn and Sn. Development of these and related high-performance solution-processed metal chalcogenide semiconductors offers outstanding opportunities for meeting price and performance targets in photovoltaics and microelectronics industries.