My research area at Swarthmore College was the intensity statistics of partially coherent light. Midway in my career at Swarthmore, I spent a post-doctoral semester at Princeton, working in the area of laser photon statistics. A significant result of this research was the publication of a detailed computation of the time dependence of the higher intensity correlation functions of laser light. The conclusions of this work were partially confirmed experimentally in1974 by DeGiorgio, and were confirmed in more detailed experiments by Singh and Qu in 1996.1
From 1973 to 1979 I carried out research on laser fusion and laser isotope separation at the Los Alamos Scientific Laboratory (now the Los Alamos National Laboratory). My contributions to this project included the assignment and interpretation of the infrared spectrum of UF6 and other complex molecules, and the discovery of the existence of new pathways for infrared multiphoton excitation and photodissociation of polyatomic molecules. Because understanding the energy-level structure of UF6 and other molecules was a key step in the development of an industrial method of laser isotope separation, I was elected a Fellow of the IEEE in 1985. A review article2 on multiphoton excitation of atoms and molecules summarized this body of work.
One of the key engineering problems in the Los Alamos approach to laser isotope separation was the development of efficient, high-peak-power laser sources in the 15.9 mm wavelength region. Since there are no efficient lasers that radiate near 15.9 mm, the only possibility was to convert energy from carbon dioxide lasers in the 9—10 mm region. The most efficient 15.9 mm source found to date is the CO2-laser-pumped rotational Raman laser in hydrogen gas, which I and two other Los Alamos staff members invented in 1974. The original patent application was classified, thus leaving a loophole for other inventors. After Los Alamos mounted a successful defense against a challenge from Stanford University, our patent3 was upheld and the claims were significantly broadened when the patent was reissued in 19924.
Laser-pulse propagation effects are important in all uranium laser isotope separation methods, because they affect the efficiency with which laser radiation can be converted to isotopically selective excitation. In 1977, I began to use computational methods to understand the many nonlinear effects that occur when a laser pulse propagates through a medium at or near a resonant absorption frequency. In the course of this research program, which has continued to the present time, I have obtained more than $2,000,000 in external funding and produced ten Ph.D. graduates. The most recent Ph.D. graduate in this area, Dr. Dawn Hollenbeck, is currently employed as a postdoctoral Research Associate at UT-Dallas.
In 1987, I began a research program on nonlinear optics in glass or liquid dielectric spheres, originally with a view towards applications to defense against high-power laser weapons. A highly transparent sphere is a high-Q resonant cavity for electromagnetic waves at certain frequencies. Laser excitation of a high-Q resonance can launch a high-intensity wave that makes of the order of Q round trips around a circumference of the sphere, thus creating an ideal nonlinear optics laboratory. The original proposal5 that an intense hypersonic wave generated through stimulated Brillouin scattering might fracture glass or liquid spheres led to the development of a completely new method for calculating nonlinear couplings of electromagnetic and acoustic waves in spheres. This method uses group theory and the Racah-Wigner angular-momentum calculus (both of which I used in my spectroscopic research at Los Alamos) to take maximum advantage of the exceptionally high symmetry of the sphere for simulating nonlinear interactions of electromagnetic fields. A result of this approach is a computational method that is more accurate, and several orders of magnitude faster, than conventional finite-difference time-domain techniques. My work in this area, summarized in two journal articles6 and one conference paper7, took at least as much time and intellectual effort as my own Ph.D. dissertation.
The research program on nonlinear optics in dielectric spheres culminated in 1997 with the graduation of Dr. Paul G. Quinn, whose Ph.D. dissertation research project was an intensive computational and analytical study of the nonlinear interactions of electromagnetic fields and acoustic waves.8 Several manuscripts are in preparation. In the future, our joint efforts may contribute to an improved understanding of electromagnetic effects in spherical integrated circuits. Dr. Quinn is employed at Raytheon.
In 1990 I began to work with the Solenoidal Detector Collaboration on the problem of data filtering for the Superconducting Super Collider (SSC). The large, general-purpose solenoidal detector was expected to produce one to two proton-proton collisions every 16 nanoseconds. Since the detector was to have been equipped with 2 million sensors, each collision was expected to produce a minimum of 2 megabytes of data. These specifications imply the amazing raw data rate of 1.25´1014 bytes per second. I believed that, if a solution to the problem of filtering data for the signatures of "interesting" events could be achieved, it would probably be the most significant technological accomplishment generated by the SSC. Because Professor Fenyves’ group in Physics was working with fiberoptic detectors, I became aware of the possibility of performing pattern recognition on the outputs of hundreds of thousands of optical fibers. Neural networks appeared to be ideally suited for this purpose. One of Professor Fenyves’ students, Joe Orgeron, completed a dissertation on the design of an electronic neural network for recognition of the specific patterns that were expected to be associated with interesting events. Two of my Ph.D. students, Juvenal Fernandes and Feraydoun Kashefi, designed and simulated various versions of a completely fiberoptic neural network that was intended to provide detection of any desired (fixed) pattern. Dr. Fernandes received his Ph.D. degree in 1995, and Dr. Kashefi received his degree in 1999.
Also in 1998, I began a research effort in optical switching and routing. This effort was supported by a grant from Alcatel Corporate Research Center, which provided a stipend for one Ph.D. student. Collaborative research with Alcatel and support of a graduate student are expected to continue in 2002, with an increase in the level of funding to $175 K annually (pending resolution of intellectual property issues). The topics of collaborative research with Alcatel for 2002 include the following:
All of these research topics require an in-depth understanding of the behavior of optical communication systems from the media and physical layers (layers 0 and 1) to the network (routing) layer (layer 3) and the transport layer (layer 4). Alcatel sought me out as Principal Investigator for this work because my experience covers the physical, hardware and protocol layers of optical communication systems.
In 1999, stimulated by the arrival of an extremely capable doctoral student with a well-formed plan and 25 years of experience at Texas Instruments, I began a research thrust in quantum computing. The student, Doug Matzke, wanted to create some of the tools that will be needed in order to design quantum computers with thousands or millions of gates. Up to this time, the physics community has not considered more than a few gates. Attempting to design, or predict the behavior of, a large quantum computer using a traditional physics approach would be as foolish as attempting to analyze the logical behavior of a Pentium starting from Maxwell¹s equations. Dirac said as much long ago when he remarked that, in the future, progress in physics would depend on talented young individuals¹ finding new approaches to analyze complex systems. Dr. Matzke successfully defended his dissertation, "Quantum computation using geometric algebra", in January 2002.
In 2000, I undertook the supervision of doctoral research in computer architecture, specifically in the area of optimizing digital signal processor architectures for algorithms other than multiply-accumulate and the Fast Fourier Transform. For example, a new algorithm of much practical interest is the Fast Wavelet Transform (FWT). Under my supervision, jointly with Professor Pervin, Jagadeesh Sankaran, the leading assembly-language algorithm designer in the Texas Instruments DSP group, is working on designing architectures and instruction sets for DSP algorithms such as the FWT.
From 1999 through 2002, the dissertation research of Dr. Dawn Hollenbeck on the dynamics of a fiberoptic Raman amplifier has led to a total of $223 K of external funding and a donation of optical networking test equipment worth $2.25 million. Several publications, one of which is in press10, are expected to result from this work, which draws on all of the experience on laser-pulse propagation I have gained over the past twenty-five years. In particular, the numerical methods developed for this research may be valuable intellectual property. In order to protect UTD¹s intellectual property, Dr. Hollenbeck and I filed a patent disclosure titled "Highly parallelizable algorithm for laser-pulse-train propagation in an optical fiber".