IEEE Circuits and Systems Magazine - Q4 2019 - 9

Memristive technologies are young and on their way to maturity.
Hence, the path to maturity is no less long for the
models describing them.
of these top 30 papers are concerned with modeling, even though the majority of them are device
level works.
Considering the overall search combining the findings using all of above keywords, visualized in Fig. 3,
the share of circuits and systems verified by physical
implementation and measurements is as low as 12% (48
out of a total of 383 works)1. We contend that this disproportion between practical works and simulated designs
is alarming and has negative effects. Our search shows
some indications that in communities using PCM and
ReRAM keywords, they are more aware of (and hence
pay more attention to) the importance of practical implementation. However, the ratio of circuit and system
works done in those communities seems to be lower.
We note that such disproportion between simulation
and implementation exists in certain other communities such as digital CMOS circuit design too. However,
it is important to note that in that case, such disproportion is justified by the maturity of the models and
Computer-Aided Design (CAD) tools developed for
CMOS technologies thanks to the heavy investment of
major companies and consequently comprehensive and
continuous effort of engineers and scientists in device
and circuit development and modeling. Therefore, a
successful simulation is, to a large extent, a good guarantee of a successful implementation for the majority
of digital CMOS circuits. Despite the maturity of CMOS
technology and respective models and CAD tools, a
successful simulation is much less of a guarantee for a
successful implementation in the case of analog CMOS
circuits. Therefore, in the respective community, new
designs, and concepts are often well-received and commonly spread, if and only if they are backed up by physical implementation and verification. In consequence,
the proportion of high-quality publications verified by
physical implementation is much higher compared to
their digital counterpart.
III. The Pitfalls of Circuit & System Simulation
We contend, that in the memristor community the shortage of physical implementation and verification is an important problem due to three main reasons, which we
discuss here.
1
The share of circuit and system simulations-only is 15% (57 out of 383
non-repetitive works).

FOURTH QUARTER 2019

A. Immaturity of Technology
In the case of CMOS transistors, the material and fabrication process is to a very large extent determined and
clear. The exact technology and fabricating company
determine further details. When it comes to memristors, in contrast, this can vary significantly in terms of
material, fabrication, and operation dynamics. Some of
the circuits and systems are designed for "memristors"
without any considerations for specifics of operations
of different device types. Even among a single type, for
example, ReRAMs, although Titanium dioxide (Ti O 2) is
one of the more well-known types, the literature contains many other materials and fabrication processes
such as Tantalum Oxide (Ta O x) [70], [71], Hafnium
Oxide (H f O x) [72]-[74], amorphous silicon [75], carbon
nanotubes [76], ferroelectric [77], SiNW [56], and silverbased ReRAMs [78], [79]. Given the sparse and ad-hoc
approaches towards developing these memristors, they
often remain in a pre-mature or maturing phase. Significantly less mature compared to any CMOS technology
of the day. This negatively affects their characterization
as well as their reproducibility. Some of the main challenges to be addressed at device level include device
variability, cyclic variability, OFF/ON ratio, endurance,
retention, and device speed.
Fig. 4 shows our measurement results performed on
8 ReRAMs of the same technology, fabrication round,
and die packed in a single package2. In our measurements,
2

These memristors use metalization of chalcogenide material as the
switching mechanism. We do not have permission to publish more specific details.

12%
Other Works
Circuit/System Implementation
88%

Figure 3. Distribution of "Top 30" literature works with all keywords (derived from memristor, ReRAM, and PCM) showing
the share of circuit/system implementation/measurements
(13%) from all articles found in our searches. Other works
include theoretical papers, device level, or circuit and system
designs which were not implemented (measured).

IEEE CIRCUITS AND SYSTEMS MAGAZINE

IEEE Circuits and Systems Magazine - Q4 2019

Table of Contents for the Digital Edition of IEEE Circuits and Systems Magazine - Q4 2019