Plausible Physical Mechanisms for Unusual Volatile/Non- Volatile Resistive Switching in HfO₂-Based Stacks

Round 1

Reviewer 1 Report

The manuscript by Quinteros, Antoja-Lleonart, and Noheda discusses plausible physical mechanisms for unusual volatile/non-volatile resistive switching in
HfO2-based stacks. The discussion in the manuscript is on possible reasons why or why not the switching in current voltage characteristics of Ti/HfO_2-x/Co_Pd can be explained and what are potential consequences based on the observations and reasoning. The value of this manuscript is not so  much on the preparation/description of the devices and conducted experiments, but more of what the implications are of experiments who didn't quite work out as the authors would have expected. From this point of view, the manuscript is somewhat of an oddity because most of the time, one would hesitate to show that experiments didn't work. Here, the authors turn around and try to reason what are potential reasons and if this can be used to show potential avenues to improve and make use of the observed effect after all. I don't see a reason why such a manuscript should not be published since it contains valid and well reasoned points to improve our understanding and maybe how to think around the corner to get to a workable solution.

I have two small points the authors may consider.

You talk at some point about "an unintentional ion migration". Usually this would be detrimental to a device because it indicates a certain operation instability. That point you don't really discuss if I am not mistaken. The way you use it would mean you introduce strain to get to the symmetry lowering and hence FE. I think you would have both effects.

One point in your discussion about FE and AFE you didn't mention, but I think plays a role in oxides, are oxygen defect. In films we usually have them most of the time wheather we like it or not. Your discussion reminded me of the work done by Bednorz on Cr-doped STO single crystals, which was compared to Cr-doped films because there can be optically active centres in the crystal leading to a switiching behaviour if an electric field is applied if I remember correctly. This work at IBM was one of the three possible routes for memory devices. In films defects where detrimental to this kind of switching and oxygen defects where the probably cause of it. My question would be, how would oxygen defects play in your explanations?

Author Response

Please see attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

The manuscript presents an unusual type of resistive switching in a HfO2-based stack, compares it to switching in other HfO2-based stacks and suggest two different explanations for the experimental findings. The paper contains interesting experimental findings and novel ideas to explain them. However, the paper is not well structured and contains many unjustified or contradictory conclusions. It resembles a collection of data and ideas rather than a systematic scientific investigation and will not be very useful for the respective community in the present form. I therefore cannot recommend this paper for publication. Please find my detailed comments below:

• The IV-curve depicted in Figure 1 is definitely interesting. The authors claim that this type of switching is much more reliable than the bipolar switching observed in other HfO2-based stacks but no reliability data is shown. The authors should provide Weilbull plots of C2C and D2D data as it is usual in the community to prove small variability of memristive devices.
• The bullet points of the pros and cons of this type of switching on page 3 is very unusual for a journal publication. It should be written in full text form and well explained instead of only summed up in keywords. This aspect might also fit better in a summary at the end of the paper instead of in the middle of the description of the data.
• The comparison of the results of the different stacks, prepared in different labs is very confusing and no real conclusions can be extracted from it. The different temperature might be a reason, but in order to confirm this, systematic temperature studies should have been performed on one of the same system. Comparisons of different growth systems and labs are generally not very useful as long as the exact differences cannot be identified. The fact that HfO2 can show bipolar switching is well known. It is one of the most well investigated memristive systems. The authors could have cited many papers instead of inserting data from samples provided by the Juelich group. Also the comparison to sample batches that have been fabricated later in Groningen only shows that the observed type of switching is difficult to reproduce. The comparison of the different stacks is very lengthy and confusing and does not provide the reader with any useful information since the explanation why they exhibit different switching behaviour is missing. It only dilutes the information that might be interesting for the readership. I strongly recommend to stick to one sample stack in this paper and discuss in detail the observation of the unusual type of switching.
• The suggestion of the threshold switching as origin of the unusual switching sounds reasonable. Since similar curves have been reported together with finite element simulations previously (see Funck at al. Adv. Electr. Mater. 2016). The short-coming of this suggested mechanism is that it is not obvious which of the layers in the complex stack exhibit threshold switching. Usually threshold switching is observed in low band-gap oxides or Mott systems like VOx, NbOx or charge transfer insulators like NiO (references to previous work are missing here. The authors claim that threshold switching is only observed in phase change materials which is not true). For wide band gap semiconductors such as HfO2 it is less obvious. The authors have also mentioned in the beginning of this manuscript that the HfOx might be highly conductive and not contribute to the switching and that the TiO2 interface layer might play the dominant role. Based on this suggestion, threshold switching might take place within the TiO2 layer. That might be definetely be more likely because of the lower band-gap.
• The suggestion of anti-ferroelectricity as origin of the observed hysteresis curve is pure speculation. It should have been at least proved that the HfO2 in this stack is antiferroelectric which is not the case here. If the authors could better prove their idea it could be published in a high ranking journal. However, as long as no further justification is provided, I recommend skipping this aspect from the manuscript.

Author Response

Please see attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

This work reported an unusual switching behavior of multiple oxide-layered structure. Through comparison with the behaviors of devices with various stacking layers, the mechanism was analyzed. Possible mechanisms were also discussed by comparing the device behavior with threshold switching devices and ferroelectric devices. The device performance is interesting, but actually more electrical measurement data need to be elaborate. The manuscript was written in an unusual way.

There are also several specific issues need to be addressed:

1. The Ti or TiO2 layer: this layer was deposited as Ti. However, subsequently, the authors refer to it as TiO2. Although the authors discussed later about this layer might not be fully oxidized, it is still not convincing why it should be fully oxidized? A Ti layer thinner than 5nm might be fully oxidized by not 20 nm. Could the author provide more evidence about the oxidation status of this layer?
2. If we assume the Ti layer is as deposited, or only slightly oxidized. The device shown in Fig. 1 is actually two resistive devices, i.e., Si/SiOx/Ti and Ti/HfOx/Pt, connected back-to-back. This is the case of complementary RRAM switching (See D.J. Wouters, IEEE EDL, 33, 1186, 2012; Federico Nardi, IEEE, TED, 60, 70, 2013), which shows similar behavior as the authors showed about threshold switching. It is recommended that the authors look into this possibility.
3. For the device behavior in Fig. 1, how long does it wait between successive voltage sweep? Will the device behave as in pristine state if continuously apply positive voltage after the 4^th cycling in Fig. 1b?
4. In the advantage claims, the author mentioned good C2C and D2D stabilities. Any data to support these claims?
5. The device name method is interesting, but not academically way. Where are the device fabricated should not be relevant to be performance of the device or of interesting to the readers.
6. Reviews on the mechanism of threshold switching and ferroelectric device are not necessary and actually make the manuscript too long to follow the essential findings.

Author Response

Please see attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The authors did not agree with most of my points of criticism and conducted only minor changes of the manuscript. I therefore still disagree that this paper will be usefull for the community in the present form.

Author Response

Please see attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

No further comments. 

Round 3

Reviewer 2 Report

The manuscript has strongly improved due to the restructuring. I recommend puublishing the paper in the present form.