Foreword by R. Stanley Williams
I first became aware of the seminal papers on memristors by Leon Chua and the
extension of memristive systems by Chua and Kang in 2004. I was struck by
the similarity of the “pinched hysteresis loop” characteristic of the axiomatically
defined memristor and the experimental data we had been collecting in my group
for nearly 6 years, and I spent a couple of frustrating years trying to figure out if
and how they were connected. After suffering an auto accident in 2006 and being
incapacitated for the month of August, I spent several weeks downloading and
reading about 300 papers. Somehow, my subconscious brain made a connection
among them and I had a genuine eureka! moment, where I understood how and
why our resistance switching devices were memristors. I spent the next year
primarily in writing invention disclosures and patent applications, and in 2008
with my HP colleagues we published our “Missing Memristor Found” paper in
Nature. The next couple of years were spent on expanding our understanding of the
physical mechanisms behind memristance in transition metal oxides, developing
improved compact models that we could use for circuit simulations, improving
our devices, and exploring a wide range of potential applications for the “fourth
fundamental circuit element.” During this period, I read many more papers by
Chua and collaborators, and I realized that the memristor was the literal “tip of the
iceberg” for an enormous and comprehensive body of work on nonlinear dynamical
circuits and networks. I have spent much of the past decade studying and learning
how to utilize nonlinear dynamics and chaos for computation in anticipation of
the saturation of Moore’s scaling for CMOS circuits. I could see that for some
functions, one or two memristors could emulate the properties of a neuron with
better fidelity than a circuit with several hundred to thousands of transistors, thus my
goal was not a one-off replacement for a transistor, but rather dramatic simplification
and efficiency improvements in complete circuits for neuromorphic computing.
However, the existing papers on nonlinear dynamical circuits were necessarily very
formal mathematical developments of unfamiliar (to me) concepts that required
extreme effort on my part to understand. In order to provide guidance for me and
others, I invited Leon Chua in 2015 to present a series of lectures sponsored by
HP that gave him the opportunity to provide an overview of his entire body of
xi
I first became aware of the seminal papers on memristors by Leon Chua and the
extension of memristive systems by Chua and Kang in 2004. I was struck by
the similarity of the “pinched hysteresis loop” characteristic of the axiomatically
defined memristor and the experimental data we had been collecting in my group
for nearly 6 years, and I spent a couple of frustrating years trying to figure out if
and how they were connected. After suffering an auto accident in 2006 and being
incapacitated for the month of August, I spent several weeks downloading and
reading about 300 papers. Somehow, my subconscious brain made a connection
among them and I had a genuine eureka! moment, where I understood how and
why our resistance switching devices were memristors. I spent the next year
primarily in writing invention disclosures and patent applications, and in 2008
with my HP colleagues we published our “Missing Memristor Found” paper in
Nature. The next couple of years were spent on expanding our understanding of the
physical mechanisms behind memristance in transition metal oxides, developing
improved compact models that we could use for circuit simulations, improving
our devices, and exploring a wide range of potential applications for the “fourth
fundamental circuit element.” During this period, I read many more papers by
Chua and collaborators, and I realized that the memristor was the literal “tip of the
iceberg” for an enormous and comprehensive body of work on nonlinear dynamical
circuits and networks. I have spent much of the past decade studying and learning
how to utilize nonlinear dynamics and chaos for computation in anticipation of
the saturation of Moore’s scaling for CMOS circuits. I could see that for some
functions, one or two memristors could emulate the properties of a neuron with
better fidelity than a circuit with several hundred to thousands of transistors, thus my
goal was not a one-off replacement for a transistor, but rather dramatic simplification
and efficiency improvements in complete circuits for neuromorphic computing.
However, the existing papers on nonlinear dynamical circuits were necessarily very
formal mathematical developments of unfamiliar (to me) concepts that required
extreme effort on my part to understand. In order to provide guidance for me and
others, I invited Leon Chua in 2015 to present a series of lectures sponsored by
HP that gave him the opportunity to provide an overview of his entire body of
xi