Research Overview

Transistor switching analysis. Dissertation (Ph.D.), California Institute of Technology (0).

Proposed and demonstrated the first three-terminal solid-state device operating with electron tunneling and hot-electron transport as its operating principles (1).

Initiated a systematic investigation of the energy-momentum relation of electrons in the energy gap of insulators (2, 3, 4) and semiconductors (5, 6, 7). These studies involved exquisite control of nanometer-scale device dimensions, albeit in only one dimension.

First demonstration that hot electrons in Gold retained their energy for distances in the nanometer range (8).

With W. Spitzer (9, 10, 11), established the systematic role of interface states in determining the energy at interfaces of III-V compounds, independent of the detailed nature of the interface. This work anticipated the role of interfaces in band-gap engineering, which is centrally important in modern heterojunction devices.

With W. Spitzer and many other collaborators and students, undertook a systematic study of the physics and commercial importance of Schottky barriers on a wide range of semiconductors. The understanding gained from these studies ramified in several directions over the following years. Much of the nano-scale work accomplished during the period is directly dependent on knowledge of barrier behavior and tunneling.

Built the first working Schottky-barrier-gate field-effect transistor (12). This device (MESFET) has come to be the standard high-frequency transistor used in satellites, cell phones, and other microwave communications systems. Using modern band-gap engineered materials, the device is now known as the HEMT.

Gave the first systematic treatment of ohmic contacts to semiconductor devices. These structures were the first true nanometer-scale devices, and are still the most numerous. Showed that they were tunneling junctions and were critically dependent on the doping density of the semiconductor, but not on the metal used (13, 14).

With S. Kurtin and T.C. McGill, demonstrated several manifestations of a fundamental transition in the nature of solids depending on the relative contribution of covalent and ionic bonding. This paper was highly controversial at the time, and was only published after careful review by, and discussions with John Bardeen (15).

With S.T. Hsu and R. Whittier, reported the first single-electron transistors. These were accidental three-dimensional nanometer-scale devices that resulted in field returns of commercial transistors (16).

Taught first VLSI course at Caltech. First Multi-Project Chip. This course and the multi-project shared-wafer methodology that went with it became the model for an entire generation of courses that contributed greatly to innovation around the world (17).

With M. Delbruck, initiated a program to apply the physical principles learned from electron transport through insulating films to ion transport through membranes of biological interest. At the time the common belief was that the exponential current-voltage characteristics of these systems was due to the individual properties of certain nanometer-scale molecules embedded in the membrane. With a number of collaborators (18, 19, 20) established that the characteristics were due to the population statistics of the molecules, and that the individual molecules had an ohmic current-voltage curve. This result is now taken for granted in the biophysics literature, but was quite controversial at the time.

With B. Hoeneisen, showed that Silicon Carbide Schottky-barrier diodes with nanometer-scale depletion layers were vastly more effective high-power rectifiers than conventional Silicon devices (21). This prediction was based on a deep understanding of the tunneling process and measurements of barrier properties on this remarkable material. The potential of these devices was not realized until twenty years later when single-crystal SiC wafers became available. Today these devices are the workhorses of highpower electronics.

With S. Colley, built and demonstrated the first simple Silicon Compiler. Produced both simulation and layout from higher-level functional description (in this case, finite state machine code). Silicon Compilation was to be the centerpiece of the next 15 years’ work, motivating many of the more detailed contributions. By 1991, every major chip design effort in the world used some variant of this key technology.

As the culmination of many years of work in solid-state device physics, published (with B. Hoeneisen) the first prediction of the nanometer-scale lower limit to the size of transistors (22). These limits were based on fundamental physical laws, and were much smaller than generally expected. These predictions, along with the general notions of scalability that went with them, were instrumental in setting the industry on its path toward nanometer-scale technology. The limits established at that time have held up to the present day, in the face of many years of experimental and theoretical work done at laboratories throughout the world (23). Because these results formed the scientific basis underlying Moore’s Law, they have had enormous economic impact worldwide.

With A. Mohsen, T. McGill, and Y. Daimon, gave the first quantitative treatment of charge transfer efficiency in overlapping-gate CCD structures (24, 25, 26). A new clocking methodology was invented, which allowed a considerable increase in both charge capacity and transfer efficiency (27). This work formed the basis for the high transfer-efficiency CCD devices now used for imaging applications.

Described two unique concurrent computing structures: a serial compare-under mask chip (class project from the first VLSI course, actually finished in 1971), and (with E. Cheng and R. Lyon), a serial pipelined multiplier. Articulated the concept that pipelined structures passing data to nearest neighbors formed an optimal VLSI structure. This principle paved the way for much more general work on systolic algorithms.

Distilled many of these thoughts into an article, written jointly with Ivan Sutherland, in Scientific American (28). This popular account was the first that received any attention from the Computer Science community.

With M. Rem, proposed the foundations of a Complexity Theory for VLSI, in which time, area, and energy were the dimensions of a cost vector (29, 30). This work has led to a fundamental expansion in the notion of computational complexity.

Gave the first public discussion of the role of silicon foundries in promoting technological innovation (31). The first Caltech conference on VLSI, at which that talk was given, was also the occasion of the first description of Dave Johannsen’s graduate research on silicon compilation, and was the first time the term had been coined. The coincidence of silicon compilation (now called Synthesis) and Silicon Foundries (32) led to an entire new business model for the semiconductor industry, now called Fabless Semiconductor. This segment is currently responsible for over half the economic value of the entire semiconductor industry.

Articulated the impact that the VLSI technology would have on Computer Science education, a set of predictions that have now appeared in the standard curricula (33).

The book, Introduction to VLSI Systems, written with Lynn Conway as co-author, appeared (34). This book captured many of the insights of the previous 10 years’ work in a form that could be taught to students with a wide variety of backgrounds.

With M. Chen, presented the first formal semantics for general VLSI systems (35). This work led to a completely general hierarchical approach to system specification and simulation (36, 37, 38, 39).

With T. Lin, extended the hierarchical semantics work to include a physically based treatment of time delay (40, 41, 42).

With J. Wawrzynek, described a very general concurrent computational approach to problems requiring the solution of finite-difference equations in time. Used the approach to produce high-quality musical instruments in real time. The structure used for this application was a programmable interconnect technology that became the basis for a large class of commercial Field-Programmable Gate Arrays (43, 44, 45).

With M. A. Mahowald, described the first analog silicon retina (46). The approach to silicon models of certain neural computations expressed in this chip, and its successors, foreshadowed a totally new class of physically based computations inspired by the neural paradigm. More recent results demonstrated that a wide range of visual and auditory computations of enormous complexity can be carried out in minimal area and with minute energy dissipation compared with digital implementations.

The book Analog VLSI and Neural Systems was published (47). This book condensed the insights gained during the previous eight years of work into a single volume, accessible to students with a wide range of backgrounds. Several recent reviews have spelled out in some detail the compelling advantages of realizing adaptive systems directly in analog VLSI. Reduction of system power dissipation by a factor of 10,000, and of silicon area by a factor of 100 are being demonstrated.

Experience gained in using photo-response of semiconductor structures for barrier-energy and band-gap studies led to system-level structures that sensed and processed images in various ways. With numerous collaborators, a large variety of imaging structures were developed. One branch of this effort resulted in CMOS imagers, now the most prevalent of all image sensors. A particular subset of these, the X3 sensors, have produced some of the finest images ever captured by any photographic technology.

Throughout the entire period, worked to bring about a general awareness of Computation as a physical process, rather than purely a mathematical one. Strongly advocated the importance of unifying technology and architecture into a single discipline, and emphasized the importance of this unity for the future of the field at large.

The book Collective Electrodynamics: Quantum Foundations of Electromagnetism, published by MIT Press, unifies electromagnetic phenomena with the quantum nature of matter (48).

Recent work on Collective Electrodynamics is evolving an entire introductory level physics course based on macroscopic quantum systems. This approach allows students to develop a deep intuition for fundamental physical processes by way of simple laboratory experiments.

G4v - an Engineering approach to Gravitation

G4v is an internally-consistent, quantum coupled treatment of gravitation and electromagnetism. This theory is a direct extension of Einstein's 1911/12 approach. It diff ers from previous attempts in a number of important ways:

1. The theory is based on Mach's Principle and provides a conceptual base for the Equivalence Principle.
2. It is not a metric theory; it is formulated in flat space-time.
3. Neither the electromagnetic nor gravitational fields are quantized. The wave functions and four-potentials are continuous functions of space and time. Quantization results from the interaction of matter and field wave functions.
4. The speed of light c is equal to the gravitational potential. It is not constant, but varies with position and time.
5. The quantity of matter coupled gravitationally is not the mass m; it is the Compton wave number.
6. The theory is based on four-vector coupling. It is thus locally Lorentz-invariant in regions where the speed of light can be considered constant.
7. The source of the electrical four-potential is the charge/current density four-vector, and that for the gravitational four potential is the energy-momentum four-vector. Both quantities are de fined for the wave function of the source matter, and appear as terms in the a ffected matter wave function.

In addition to obtaining the correct value for the light-deflection problem, G4v also obtains exactly the same expressions as GR, to the fi rst order beyond Newton, for perihelion precession, Gravity Probe B, gravitational redshift and Shapiro delay. The total gravitational radiation from binary systems is identical for circular orbits, but has a very slightly di fferent eccentricity dependence, which may be testable as more high-eccentricity binary-pulsar systems are discovered.

The G4v predictions for gravitational-wave radiation patterns from binary systems, and for the antenna patterns of observatories like LIGO are markedly di fferent from those of GR. (50, 51) These predictions should be testable within the next few years.