xxii
Preface
to code the 0 and 1 binary bits, instead of the conventional voltage or current, which
collapses to zero when power is interrupted.
News of the nanoscale HP memristor has breathed new life into a device of
antiquity that had been relegated to the dustbin of history. Overnight, it has triggered
a torrent of research and development activities on memristors worldwide in both
industry and academia (see Figure 1b, c in [21]), in anticipation of its disrupting
potential in AI and neuromorphic computing. Companies such as Panasonic and
Fujitsu have sold several hundred million chips with embedded memristor memory
based on tungsten oxide since 2013, and Taiwan Semiconductor Manufacturing
Company (TSMC) has announced that it has developed a 22-nm memristor crossbar
process for ASIC embedded memories that will be available for mass production
in 2019. Aside from its predicted eventual replacement of the over-extended flash
memories, DRAMs, and even hard drives [2] with disrupting nonvolatile memristor
technology, memristors can emulate synapses and ion channels in neurons and
muscle fibers [33], sweat ducts in human skin, and even the primitive amoeba’s
amazing counting ability [29]. Moreover, the memristor’s scalable diminutive
physical size makes it the right stuff for building brain-like intelligent machines.
Furthermore, because even plants can remember events and communicate through
memristors [34], could memristors in fact be the sine qua non for emulating life
itself?
I close my above reminiscence (which is an embellishment on a previous
work [35]) by reproducing the last paragraph of my 1971 article, “Memristor-the
missing circuit element” [4]. In hindsight, this passage foreshadowed the pinched
hysteresis loop fingerprints not only of the ideal memristor predicted in the article
but also of all memristors cited in my later works [29–31, 33, 36, 37]. Although
no physical memristor has yet been discovered in the form of a physical device
without internal power supply, the circuit-theoretic and quasi-static electromagnetic
analyses presented in Sections III and IV make plausible the notion that a memristor
device with a monotonically increasing ϕ − q curve could be invented, if not
discovered accidentally. It is perhaps not unreasonable to suppose that such a device
might already have been fabricated as a laboratory curiosity but was improperly
identified! After all, a memristor with a simple ϕ − q curve will give rise to a rather
peculiar—if not complicated hysteretic—v − i curve when erroneously traced in the
current-versus-voltage plane. (Moreover, such a curve will change with frequency
as well as with the tracing waveform.) Perhaps, our perennial habit of tracing the
v − i curve of any new two-terminal device has already misled some of our deviceoriented colleagues and prevented them from discovering the true essence of some
new device, which could very well be the missing memristor.
Torino, Italy
Fernando Corinto
Siena, Italy
Mauro Forti
Berkeley, CA, USA
Leon O. Chua
Preface
to code the 0 and 1 binary bits, instead of the conventional voltage or current, which
collapses to zero when power is interrupted.
News of the nanoscale HP memristor has breathed new life into a device of
antiquity that had been relegated to the dustbin of history. Overnight, it has triggered
a torrent of research and development activities on memristors worldwide in both
industry and academia (see Figure 1b, c in [21]), in anticipation of its disrupting
potential in AI and neuromorphic computing. Companies such as Panasonic and
Fujitsu have sold several hundred million chips with embedded memristor memory
based on tungsten oxide since 2013, and Taiwan Semiconductor Manufacturing
Company (TSMC) has announced that it has developed a 22-nm memristor crossbar
process for ASIC embedded memories that will be available for mass production
in 2019. Aside from its predicted eventual replacement of the over-extended flash
memories, DRAMs, and even hard drives [2] with disrupting nonvolatile memristor
technology, memristors can emulate synapses and ion channels in neurons and
muscle fibers [33], sweat ducts in human skin, and even the primitive amoeba’s
amazing counting ability [29]. Moreover, the memristor’s scalable diminutive
physical size makes it the right stuff for building brain-like intelligent machines.
Furthermore, because even plants can remember events and communicate through
memristors [34], could memristors in fact be the sine qua non for emulating life
itself?
I close my above reminiscence (which is an embellishment on a previous
work [35]) by reproducing the last paragraph of my 1971 article, “Memristor-the
missing circuit element” [4]. In hindsight, this passage foreshadowed the pinched
hysteresis loop fingerprints not only of the ideal memristor predicted in the article
but also of all memristors cited in my later works [29–31, 33, 36, 37]. Although
no physical memristor has yet been discovered in the form of a physical device
without internal power supply, the circuit-theoretic and quasi-static electromagnetic
analyses presented in Sections III and IV make plausible the notion that a memristor
device with a monotonically increasing ϕ − q curve could be invented, if not
discovered accidentally. It is perhaps not unreasonable to suppose that such a device
might already have been fabricated as a laboratory curiosity but was improperly
identified! After all, a memristor with a simple ϕ − q curve will give rise to a rather
peculiar—if not complicated hysteretic—v − i curve when erroneously traced in the
current-versus-voltage plane. (Moreover, such a curve will change with frequency
as well as with the tracing waveform.) Perhaps, our perennial habit of tracing the
v − i curve of any new two-terminal device has already misled some of our deviceoriented colleagues and prevented them from discovering the true essence of some
new device, which could very well be the missing memristor.
Torino, Italy
Fernando Corinto
Siena, Italy
Mauro Forti
Berkeley, CA, USA
Leon O. Chua