Towards Sustainable Crossbar Artificial Synapses with Zinc-Tin Oxide

Round 1

Reviewer 1 Report

Carlos Silva et al. study crosspoint memristive devices based on zinc-tin oxide active layer. Memristive devices are of high interest due to their effective ability to emulate biological synapses. In this respect a memristive device should demonstrate appropriate characteristics, such as high Roff/Ron ratio, endurance, plasticity, long retention time and low c2c and d2d variation. Several of these characteristics were not presented in the manuscript. Moreover, the novelty of the study is not clear for a second reader and some statements are not supported. Therefore, I recommend a major revision of this manuscript:

1. The introduction part does not cover previous appropriate works on ZTO-based memristors and other memristive structures and applications have already demonstrated crosspoint ZTO memristors. Materials with the other possible resistive switching mechanisms could also be mentioned: photo induced [https://doi.org/10.1002/adfm.201806646], electrolyte gated [https://doi.org/10.1002/aelm.202000511], second- and third-order effects [https://doi.org/10.1038/s41586-020-2735-5].
2. The area of the non-patterned devices was not mentioned.
3. The authors claim that the switching mechanism of their devices is not filamentary. However, there were no experimental demonstration of such behavior. E.g. one may demonstrate it by investigating devices with different areas.
4. The authors are asked to demonstrate retention time and cycle-to-cycle and device-to-device variation characteristics.
5. Why the authors present the Fig. 2a in semi logarithmic scale while the Fig. 2b in linear?

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

This work reports on a zinc-tin oxide (ZTO) based resistive switching device, with fairly good performance and comprehensive characteristics and analysis. Patterned (25 um^2) and non-patterned devices, voltage sweep and current sweep operation, are compared to each other respectively. Synaptic behaviors of potentiation and depression by identical pulse for each trace are also demonstrated. The manuscript is well written.

There are several comments from the reviewer just to further improve the integration of the current manuscript.

1. I would suggest avoiding using the terminology “8-wise” which is poorly defined, to describe the switching behavior. Style of drawing the digits “8” is largely dependent on the culture and personal habit. From my view, the switching behaviors in Figure1 should be “anti 8-wise” other than “8-wise”. The authors also used “C8-wise” terminology, which is ambiguous to me and should be clarified.
2. On page 4, the authors state: “In a biological brain, the relationship between two neurons can either be assisted or inhibited, which leads to an increase (potentiation) or decrease (depression) of the synaptic strength between them.” Actually, the assisted/inhibited connection between neurons and potentiation and depression behaviors are two different concepts. The assisted/inhibited synaptic weight indicates the sign of the weight, which can be implemented with the positive and negative conductance of a differential pair, while the potentiation and depression behaivors relate to the learning or weight update algorithm (see iScience, 23, 101809, 2020, DOI: 10.1016/j.isci.2020.101809).
3. The comparison between voltage sweep and current sweep and related discussion is interesting. However, might the endurance difference also be caused by the lack of current compliance in the voltage sweep?
4. I see an interesting behavior in Figure 4c, where the temperature dependent behavior deviate from Arrhenius law in low temperature region. Does the author have further comment?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The manuscript (electronicmat-1168958), Towards sustainable crossbar artificial synapses with zinc-tin oxide, shows quite interesting results of ZTO memristor characterizations for potential synaptic device applications. Authors present quite comprehensive analysis in this work electrically and physically. Some technical comments would like authors address here before further confirm the outcomes, shown as following -

1. Can authors provide some level of material characterization to further confirm the ZTO layer information, e.g. XRD for crystallinity analysis and XPS for composition analysis.
2. Can authors provide the retention data in this work for the pattern and non-patterned devices? Also, since the sweep types (voltage or current) would impact endurance, did they also potentially impact on the retention performance or not?
3. Although authors presented LTP and LTD characterization by AC pulse response, can authors provide the multilevel states demonstration by DC sweep (current sweep for SET; and voltage sweep for RESET)? This would further support the AC pulse response results in this work.
4. Along with question 3, can authors provide the retention data for DC multi-state or AC LTP/LTD characterizations? Did we find out any relaxation process in states?
5. Please provide a simple benchmark Table to summarize recently semiconductor oxide based memristor device characterizations and reliability performance (e.g. forming, SET/RESET. endurance, retention, and device structure). This would be quite helpful for this referee and potential readers to further insight the research impact in this work as compared to other works. 

Due to the above comments, this referee would like to put the manuscript status as "Major Revision" in the current phase. 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors have addressed all the issues.

Reviewer 3 Report

Authors have replied to this referee' comments in detail. No additional comments from this referee.