2D Spintronics for Neuromorphic Computing with Scalability and Energy Efficiency

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

A report of reviewing a manuscript JLPEA-3514013
(2D Spintronics for Neuromorphic Computing with Scalability and Energy Efficiency)

To the authors:
This manuscript is aimed to provide an overview of advancements in 2D spintronics and device architectures designed for neuromorphic applications, with a focus on techniques such as spin-orbit torque, magnetic tunnel junctions, and skyrmions. This project with use of 88 references could be timely and useful for graduate and postgraduate students as well as research workers and teaches. However, due to a number of unsatisfactory issues, which are listed below, I cannot be convinced that readers of JLPEA are satisfied with the current manuscript.

1.    In section 2 (Fundamentals of Neuromorphic Computing): Although the section title is “Fundamentals of Neuromorphic Computing”, a little description was made for fundamental of neuromorphic computing. For example, no description corresponding to Figure 1 was made in this section. In particular, the author should explain how the outputs are obtained from the inputs in neuromorphic architecture, focusing the roles of spiking neural networks and synaptic plasticity.
2.    In section 2, a term“2D spintronics”was used without its definition. The definition is necessary.
3.    In section 2, why can the 2D nature of spintronics materials enable high-density integration. Describe its reasons with use of illustration.
4.    In section 2, we need the definition for a term “real-time computation ”in line 89.
5.    In section 2, references are necessary corresponding to the description made in lines 90-98. 
6.    In section 2, what is “conventional architectures ”used in line 104.
7.    In section 3 (Overview of 2D Spintronic Materials), crystal structural figures of CrI₃, Fe₃GeTe₂, and MnPS₃ should be provided in this section. 
8.    In section, “charge-to-spin (CSI)” is probably “charge-to-spin interconversion (CSI)”in line 127.
9.    In section 4 (Spintronic Device Architectures for Neuromorphic Computing), how are the contents of this section related to those of section 3 (Overview of 2D Spintronic Materials). In particular, no 2D materials are used in subsection 4. 1. (MTJs and Spin Valves). 
10.    In section 4, no description corresponding to Figures 2 and 3 were made.
11.    In section 4, we need explanation describing the relationship between the elements used in Figure 2 and those of neuromorphic architecture shown in Figure 1.
12.    In section 4, what types of materials are considered in lines 167-178, and subsection 4.2.(Skyrmion-Based Devices).
13.    In section 4, LIF in line 224 is used without its definition, though its full name, integrate-and fire, is described later in line 237
14.    In section 5 (Scalability and Energy Efficiency), concerning efficiency of“41.4 %”indicated in Table 1, the definition of the efficiency is necessary.
15.    In section 5, provide reasons for poor error rate of 10^-3 in skyrmion-based device and 10^-2 in VdW heretostructure-device.
16.    In section 6 (Applications in Neuromorphic Computing), there is an incomplete sentence in lines 317-318.  No verb is probably used.
17.    Supposed that 2D spintronic neuromorphic computing system is successfully completed, can it be used for usual functions, which are currently solved by the conventional von-Neumann type computing system? 
18.    Several references are given without volume and page numbers in Refs. 15, 19, 24, 30, 33, 62, 65, 68, and 69.
19.    An author name (Otani Y) is doubled in Ref. 25.

Comments on the Quality of English Language

In caption of Figure 1, a last sentence "Computing paradigm enabled..." could be  "Computing paradigm are enabled..."

In section 6 (Applications in Neuromorphic Computing), there is an incomplete sentence in lines 317-318.  No verb is probably used.

Author Response

Review report 1
This manuscript is aimed to provide an overview of advancements in 2D spintronics and
device architectures designed for neuromorphic applications, with a focus on techniques
such as spin-orbit torque, magnetic tunnel junctions, and skyrmions. This project with
use of 88 references could be timely and useful for graduate and postgraduate students as
well as research workers and teaches. However, due to a number of unsatisfactory issues,
which are listed below, I cannot be convinced that readers of JLPEA are satisfied with
the current manuscript.

We thank the reviewer for the constructive points. In this new version, we made the
following revisions:
• Explained how the outputs are obtained from inputs in the neuromorphic hardware.
• Defined and explained some concepts like “2D spintronics”, “real-time computation”,
“conventional architecture” and so on.
• Clarified why 2D spintronic materials could enable high-density integration.
• Added more references as examples.
• Added a new Figure 2 to illustrate 2D magnets.
• Demonstrated the connection between Section III and Section IV.
• Corrected cited references format.
• Compared in details what tasks neuromorphic computing can solve to von Neumann
computing.

1.  In section 2 (Fundamentals of Neuromorphic Computing): Although the section
   title is “Fundamentals of Neuromorphic Computing”, a little description was made for
   fundamental of neuromorphic computing. For example, no description corresponding to
   Figure 1 was made in this section. In particular, the author should explain how the
   outputs are obtained from the inputs in neuromorphic architecture, focusing the roles of
   spiking neural networks and synaptic plasticity.

Figure 1 was cited in first paragraph where an introduction to von-Neumann and
neuromorphic architecture are given, highlighted in page 1. Figure 1 was moved to Page
1 in the main text.
To explain the outputs in this hardware, we added and highlighted “In SNNs, an
output spike is generated only when the membrane potential exceeds a defined threshold.
This event-driven mechanism allows for synchronous processing, where consumption occur
only when necessary, reducing energy consumption[1]”, “Synpatic weights determine
the strength of the connection between neurons, influencing how input spikes affect the
receiving neuron’s membrane potential.” and “Weight updates in synaptic plasticity are
typically governed by biological learning rules, such as spike-timing-dependent plasticity
(STDP). In STDP, the timing relationship between pre-synaptic and post-synaptic spikes
determines whether synaptic weights are strengthened or weakened. This temporally sensitive
learning rules enables neuromorphic systems to adopt to changing input patterns
and support efficient, continuous learning. ” to Pages 4-5 in the main text.

2. In section 2, a term “2D spintronics” was used without its definition. The definition
is necessary.

Defined 2D spintronics as “2D spintronics refers to the study and application of spinbased
electronic phenomena in 2D materials including atomically thin van der Waals
layers.” in Page 5 of the manuscript.

3. In section 2, why can the 2D nature of spintronics materials enable high-density
integration. Describe its reasons with use of illustration.

Explanation has been added to Pages 5-6 in the main text as “The advantage of using
2D spintronics for high density devices arises from the following key aspects[2]: (1)
monolayer high-quality 2D materials that are free from dangling bonds, unlike thin films
grown via sputtering or molecular epitaxy growth, which often result in non-uniform
isolating islands; (2) seemless heterogeneous integration with most interface, eliminating
lattice mismatch issues that typically occur in as-grown layers; and (3) exceptional scalability,
as 2D materials can be fabricated or exfoliated at the the nanometer scale and
compatible with advanced fabrication techniques. This makes them ideal for producing
ultra-compact devices while maintaining energy efficiency and speed.”

4. In section 2, we need the definition for a term “real-time computation ”in line 89.
“Real-time computation refers to the ability of a system to process and respond to
inputs within a time frame that meets the requirements of the task, typically without
perceptible delay. In the context of 2D spintronics, real-time computation is achieved
through the rapid manipulation of spin states, enabling fast data processing and immediate
responses to incoming information. That is why nanosecond-scale switching time
of spintronic system is preferred for real-time computation. This rapid switching allows
these devices to perform logic operations and memory updates at extremely high speeds,
aligning with the real-time demands of neuromorphic hardware.” This paragraph was
added to Page 5 in the main text.

5. In section 2, references are necessary corresponding to the description made in lines
90-98.

References [3–5] are added and check it on Page 6 in the main text.

6. In section 2, what is “conventional architectures” used in line 104.

“Conventional architectures” refers to traditional computing architectures, particularly
those based on the von Neumann architecture and standard silicon-based hardware,
such as CPUs and GPUs. These architectures are not optimized for neuromorphic computing,
which is designed to mimic the structure and function of the human brain. The
conventional architecture has been changed to “von Neumann architecture” in the main
text.

7. In section 3 (Overview of 2D Spintronic Materials), crystal structural figures of
CrI3, Fe3GeTe2, and MnPS3 should be provided in this section.

We thank the reviewer for pointing this out. Figure 2 on 2D magnets was added to
Page 7 in the main text. It shows not only the crystal structures but also a library of
existing 2D antiferromagnetic and ferromagnetic materials.

8. In section, “charge-to-spin (CSI)” is probably “charge-to-spin interconversion
(CSI)” in line 127.

We thank the reviewer for careful reading. It is charge-to-spin interconversion for CSI
as reviewer points out. This correction has been made in Page 8 in the main text.

9. In section 4 (Spintronic Device Architectures for Neuromorphic Computing), how
are the contents of this section related to those of section 3 (Overview of 2D Spintronic
Materials). In particular, no 2D materials are used in subsection 4. 1. (MTJs and Spin
Valves).

We must admit there is a lack of using 2D spintronics for architecture-level implementation
of neuromorphic hardware. There are some works at the architecture level
reporting the use of 2D materials with their memristive behaviors without magnetic
properties[6] or spintronics for neuromorphic computing[7]. To bridge this gap, future
research should focus on exploring the potential of 2D spintronic materials that combine
both memristive behaviors and magnetic properties, enabling more efficient and scalable
architecture-level implementations of neuromorphic computing. These sentences are
added to the end of Section IV in the revised main text.

10. In section 4, no description corresponding to Figures 2 and 3 were made.

“Fig. 3 shows an example of block-level circuit implementation of multi-level MRAM
synaptic device, where 1T1J configuration is adopted.” was added to the end of Subsection
A in Section IV in the main text.
Fig. 4 now is referred to in “Furthermore, skyrmion-based architectures have been
explored in recent years (Fig. 4).” and “Alternatively, the skyrmion-based reservoir
computing model Fig. 4)” in Pages 12-13 of the main text.

11. In section 4, we need explanation describing the relationship between the elements
used in Figure 2 and those of neuromorphic architecture shown in Figure 1.

The input and output can be one-to-one correspondence between Figure 1 and Figure
3 (was Figure 2). The 1T1J works as the synaptic connection and neuron layers.

12. In section 4, what types of materials are considered in lines 167-178, and subsection
4.2.(Skyrmion-Based Devices).

The materials considered are 2D magnets like Fe3GeTe2 and Fe3GaTe2. This was
added to Page 10 in the main text as “A key feature of multi-state MTJ devices is the
use of domain wall motion, the transition region between two oppositely magnetized
domains, to enable stable memristive or synaptic functionality, using 2D magnets[8].”

13. In section 4, LIF in line 224 is used without its definition, though its full name,
integrate-and fire, is described later in line 237.

Corrected. We move the definition of LIF ahead.

14. In section 5 (Scalability and Energy Efficiency), concerning efficiency of“41.4
%”indicated in Table 1, the definition of the efficiency is necessary.

In this context, energy efficiency is defined as the ability of the heterostructures to
perform neuromorphic computations with minimal energy dissipation, ensuring that a
significant portion of the input electrical energy is effectively used for computation rather
than being lost as heat or other forms of energy waste. To clarify this issue and make the
data consistent between different approaches, we changed the energy efficiency of vdW
case to 2.5 fJ/event and cited the reference in the description. Please check Pages 14 &
15 in the revised manuscript.

15. In section 5, provide reasons for poor error rate of 10−3 in skyrmion-based device
and 10−2 in VdW heretostructure-device.

The error rates are expained in the revised main text, by saying “This error rate could
be from stability of nanoscale skyrmions, thermal fluctuations, edge effect and pining.”
and “The increased error rate may result from the impurities and defects at the interface.”

16. In section 6 (Applications in Neuromorphic Computing), there is an incomplete
sentence in lines 317-318. No verb is probably used.

Corrected to “The combination of high-speed operation, energy efficiency, and scalability
makes 2D spintronic devices key enablers for the advancement of neuromorphic
computing.”

17. Supposed that 2D spintronic neuromorphic computing system is successfully completed,
can it be used for usual functions, which are currently solved by the conventional
von-Neumann type computing system?

If a 2D spintronic neuromorphic computing system is successfully developed, it could
potentially handle many of the tasks currently performed by conventional von Neumanntype
computing systems, but with some important distinctions.

First is task specialization. Neuromorphic systems are optimized for tasks that involve
pattern recognition, learning, and inference, which are common in areas like artificial
intelligence (AI), machine learning, and neural network-based computations. These
systems can process data more efficiently for tasks like image and speech recognition,
decision-making, and sensory processing. In contrast, traditional von Neumann systems
excel at general-purpose computing, where sequential logic and arithmetic operations are
crucial. Second is the energy efficiency. One of the main advantages of neuromorphic
computing systems is their potential for significantly lower energy consumption compared
to traditional von Neumann systems. Third is parallel processing. Neuromorphic systems
are designed to mimic the way the brain works, which involves parallel processing. This
makes them particularly well-suited for tasks like deep learning, which require handling
large amounts of data in parallel. On the other hand, von Neumann systems, while highly
efficient for general-purpose tasks, still rely heavily on sequential processing, which can
become a bottleneck in certain applications. But for general-purpose computing, this
neuromorphic system may be able to replace von Neumann systems for specific tasks.
For example, tasks that require precise, step-by-step, deterministic computation-such as
complex mathematical simulations, data analysis, and software engineering-might still
be better suited for von Neumann systems due to their flexibility, precision, and wellestablished
software ecosystems.

18. Several references are given without volume and page numbers in Refs. 15, 19,
24, 30, 33, 62, 65, 68, and 69.

We thank the reviewer for their careful reading of the manuscript. At the time of
submission, some references were either preprints or accepted manuscripts online, without
full publication details. We have updated the references for the published works, but for
the accepted manuscripts, volume and page numbers are still unavailable.

19. An author name (Otani Y) is doubled in Ref. 25.

Removed doubled names.

Comments on the Quality of English Language:
In caption of Figure 1, a last sentence “Computing paradigm enabled...” could be
”Computing paradigm are enabled...”

Revised accordingly.

In section 6 (Applications in Neuromorphic Computing), there is an incomplete sentence
in lines 317-318. No verb is probably used.

In fact, there is a verb “positions” in the lines. But to make it more obvious, we
changed it to “makes”.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript nicely summarizes the different state-of-the-art Spintronic device concepts that have immense potential for the realization of next generation- neuromorphic computing systems that are aspired to revolutionize the way and extent we understand and utilize computing today. These human brain-inspired computing schemes would surpass the capability of the current digital computing technology in terms of computation parallelism, speed and energy efficiency. I found the manuscript to be well written and matching the standards of a good review article. I just have some minor comments/ suggestions to help strengthen the paper quality and I would strongly recommend the authors to carefully go through the comments below and effectively address the concerns therein.

 

1- The authors need to provide an appropriate citation for their statement on line 91-95, stating: "hardware-accelerated SNNs, SOT-MRAM synapses have been integrated into crossbar arrays, where the conductance of each MTJ emulates synaptic weight adjustments. Recent experimental demonstrations have shown SOT-MRAM arrays achieving high classification accuracy in pattern recognition tasks while consuming up to 85% less power than SRAM-based implementations.

2-  The authors have written that the atomic thinness of 2D magnets (eg. CrI3, Fe3GeTe2, and MnPS3) and antiferromagnets such as MnPS3 enables high-density integration (see lines 121 and 122). In my opinion, atomic thinness is a feature that enables low power operation capabilities but not high-density integration. What a high-density integration capability requires is a small footprint which in turn can be achieved only by scaling down the lateral dimensions of the material, not its thickness. Therefore, the authors need to re-assess their claim and address the issue.

3- Being a review paper and for the sake of completeness in terms of coverage, I would expect the authors to provide some more insight on the limitations of the spintronic device concepts they presented in addition to limitations and challenges on high quality 2D material synthesis.

In this regard, I would ask the authors to discuss variability such as cell-to-cell (spatial) and cycle-to-cycle variability in operations of the 2D spintronic devices as this is a very crucial pre-requisite for any device to make it to the finish line in the race for real applications. In the same context it would also be good to address performance degradation of the devices over time.

4- For any new device technology to move from university research labs to industry fabs is scalability/ mass production capability. Can the authors include their assessment of 2D spintronic devices on this aspect? The scalability concept the authors discussed in their manuscript is in terms of either information storage capacity per square inch of a computing cell or foot print miniaturization enabling high integration density. The scalability concept I am asking the authors to add, however, is 2D material scalability enabling large volume production of the 2D spintronic devices.

 

 

 

Comments for author File: Comments.pdf

Author Response

The manuscript nicely summarizes the different state-of-the-art Spintronic device con- cepts that have immense potential for the realization of next generation-neuromorphic computing systems that are aspired to revolutionize the way and extent we understand and utilize computing today. These human brain-inspired computing schemes would sur- pass the capability of the current digital computing technology in terms of computation parallelism, speed and energy efficiency. I found the manuscript to be well written and matching the standards of a good review article. I just have some minor comments/ suggestions to help strengthen the paper quality and I would strongly recommend the authors to carefully go through the comments below and effectively address the concerns therein.

We thank the reviewer for the constructive suggestions. Major revision have been done in this new version, which include:

• References added to lines 91-95 in the original version and now in page 6 in the revised manuscript.
• Clarified the high-density integration using 2D
• Added large-scale growth of the
• Addressed challenges in

• The authors need to provide an appropriate citation for their statement on line 91- 95, stating: ”hardware-accelerated SNNs, SOT-MRAM synapses have been integrated into crossbar arrays, where the conductance of each MTJ emulates synaptic weight ad- justments. Recent experimental demonstrations have shown SOT-MRAM arrays achiev- ing high classification accuracy in pattern recognition tasks while consuming up to 85% less power than SRAM-based

References [3–5] are added and check it on Page 6 in the main text.

• The authors have written that the atomic thinness of 2D magnets (eg. CrI₃, Fe₃GeTe₂, and MnPS₃) and antiferromagnets such as MnPS₃ enables high-density inte- gration (see lines 121 and 122). In my opinion, atomic thinness is a feature that enables low power operation capabilities but not high-density integration. What a high-density integration capability requires is a small footprint which in turn can be achieved only by scaling down the lateral dimensions of the material, not its thickness. Therefore, the authors need to re-assess their claim and address the

Thank the reviewer for insightful comment regarding the relationship between the atomic thinness of 2D magnets and their potential for high-density integration. This argument raises an important distinction between thickness and lateral dimensions in determining integration density, which warrants further clarification and discussion. We agree that the primary factor enabling high-density integration is the ability to scale down the lateral footprint of a device. Indeed, lateral miniaturization directly impacts how many devices can be packed within a given area, which is a crucial consideration for achieving high-density integration. Atomic thinness, on its own, does not necessarily reduce the lateral dimensions of a material and therefore does not directly translate to higher integration density.

However, we maintain that atomic thinness provides indirect advantages that can fa- cilitate high-density integration. Specifically, the reduced dimensionality of 2D materials allows for the stacking of multiple functional layers in vertical heterostructures without significantly increasing the overall device footprint.  This vertical integration strategy can enhance device density in three-dimensional space, compensating for the limitations imposed by lateral scaling. Furthermore, atomic-scale thickness reduces parasitic capac- itance and enables low-power operation, which is an essential consideration for densely packed circuits where power consumption and heat dissipation become critical challenges. To address the concern, we added the statement to more accurately reflect the nu- anced role of atomic thinness in facilitating high-density integration. We clarified “While atomic thinness alone does not directly achieve lateral miniaturization, it enables ad- vanced vertical stacking and low-power operation, which are complementary strategies for enhancing overall integration density” in the first paragraph of Section III.

• Being a review paper and for the sake of completeness in terms of coverage, I would expect the authors to provide some more insight on the limitations of the spintronic device concepts they presented in addition to limitations and challenges on high quality 2D material synthesis.
• In this regard, I would ask the authors to discuss variability such as cell-to-cell (spa- tial) and cycle-to-cycle variability in operations of the 2D spintronic devices as this is a very crucial pre-requisite for any device to make it to the finish line in the race for real applications. In the same context it would also be good to address performance degradation of the devices over time.

We have added one paragraph to Section VII addressing the challenging in spintronics as “Despite all challenges, some emerging trends in spintronics are shaping the future of energy-efficient and high-performance computing. Next-generation 2D materials, such as magnetic vdW materials and topological insulators, offer exceptional properties like high spin polarization, low-dimensional confinement, and compatibility with existing semicon- ductor technologies. Room-temperature quantum spintronics is another transformative area, enabling the manipulation of spin states without requiring extreme cooling, which is crucial for practical applications in neuromorphic and quantum computing. Additionally, AI-enabled material discovery accelerates the identification and design of novel materials with optimized magnetic, electronic, and thermal properties, facilitating the development of more efficient spintronic devices. The 3D integration of 2D spintronic networks further enhances scalability and performance by enabling dense, multi-layer architectures that combine memory and logic functions, overcoming the limitations of traditional planar designs. Together, these advancements promise to revolutionize the field by improving energy efficiency, enhancing computational capabilities, and enabling new paradigms for information processing.”

• For any new device technology to move from university research labs to industry fabs is scalability/mass production capability. Can the authors include their assessment of 2D spintronic devices on this aspect? The scalability concept the authors discussed in their manuscript is in terms of either information storage capacity per square inch of a computing cell or foot print miniaturization enabling high integration density. The scalability concept I am asking the authors to add, however,is 2D material scalability enabling large volume production of the 2D spintronic devices.

The large-scale growth of 2D magnets is an area of active research. In our view, achieving large-size growth for 2D magnets may be more feasible compared to other 2D non-magnetic materials. Studies have shown that exfoliated non-magnetic materials, such as graphene and transition metal dichalcogenides (TMDs), tend to exhibit higher mobility compared to in-situ growth methods like molecular beam epitaxy (MBE) and chemical vapor deposition (CVD). However, the situation is different for 2D magnets, where the primary focus is on their magnetic properties rather than electronic properties like mobility. Notably, a study has demonstrated an increased Curie temperature in MBE-grown Fe₃GeTe₂[9], which is a promising development. High-quality 2D magnets with lateral sizes ranging from millimeters to centimeters can be achieved through MBE growth, and future research is expected to further advance this field.

The above paragraph has been added to Section VII.

 

References Cited

1. Li, , Zhang, Z., Mao, R., Xiao, J., Chang, L. & Zhou, J. A fast and energy- efficient SNN processor with adaptive clock/event-driven computation scheme and online learning. IEEE Transactions on Circuits and Systems I: Regular Papers 68, 1543–1552 (2021).
2. Zhang, B., Lu, P., Tabrizian, R., Feng, P. X.-L. & Wu, Y. 2D Magnetic heterostruc- tures: spintronics and quantum npj Spintronics 2, 6 (2024).
3. Verma, G., Nisar, A., Dhull, S. & Kaushik, B. K. Neuromorphic accelerator for spiking neural network using SOT-MRAM crossbar array. IEEE Transactions on Electron Devices 70, 6012–6020 (2023).
4. Verma, G., Soni, S., Nisar, A. & Kaushik, B. K. Multi-bit MRAM based high perfor- mance neuromorphic accelerator for image classification. Neuromorphic Computing and Engineering 4, 014008 (2024).
5. Sosa, , Wi, M., Barrera, M., Nasrullah, I. & Wu, Y. Simulating Pattern Recognition Using Non-volatile Synapses: MRAM, Ferroelectrics and Magnetic Skyrmions. arXiv preprint arXiv:2501.03450 (2025).
6. Joksas, D., AlMutairi, A., Lee, O., Cubukcu, M., Lombardo, A., Kurebayashi, H., Kenyon, J. & Mehonic, A. Memristive, Spintronic, and 2D-Materials-Based De- vices to Improve and Complement Computing Hardware. Advanced Intelligent Sys- tems 4, 2200068 (2022).
7. Zhou, J. & Chen, J. Prospect of spintronics in neuromorphic computing. Advanced Electronic Materials 7, 2100465 (2021).
8. Younis, , Abdullah, M., Dai, S., Iqbal, M. A., Tang, W., Sohail, M. T., Atiq, S., Chang, H. & Zeng, Y.-J. Magnetoresistance in 2D Magnetic Materials: From Fundamentals to Applications. Advanced Functional Materials, 2417282 (2025).
9. Wang, H. et al. Above room-temperature ferromagnetism in wafer-scale two-dimensional van der Waals Fe₃GeTe₂ tailored by a topological insulator. ACS Nano 14, 10045– 10053 (2020).

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Reviewer Comments

 

• The introduction section is not satisfactory at all. For a review paper, introduction section should comprehensively provide the targeted topic for the easy understanding of the readers about the topic background, related literature review, and significance of writing this review papers.  A few suggestions to strengthen the introduction section include:
  1. Provide a brief overview of the paper's content while critically addressing scalability and energy efficiency challenges in conventional CMOS systems.
  2. Compare the advantages of 2D materials like graphene, TMDs, and MXenes in neuromorphic systems.
  3. Discuss the principles of spintronic device integration, computing efficiency, and scalability.
  4. Multiple paragraphs discussing about the same point should be merged together to improve the structure of introduction section by ensuring each paragraph focuses on a single point. For instance, paragraphs 2 and 3, both discussing the role of 2D materials in memory devices, should be merged for better coherence.
  5. To conclude the introduction section, provide a breif summary about each of the remaining section of this review paper.
• For readers not familiar with the basics, the review should further include a more explicit and comprehensive description  of spintronic effects such as spin transport, coherence, and injection in 2D materials.  It would also be useful to distinguish between 2D materials (graphene, TMDs,  MXenes), and other materials employed in spintronics, and to explain how 2D materials  differ from them and are superior to them in these applications.
• The subheading of device architecture should be strengthen more by the addition o fmore architectures like spin-FETs and spin logic devices, to provide readers with a clear understanding of the device structures used in neuromorphic computing.
• The area of ‘Scalability and Energy Efficiency’ has been explored to some extent in this  work, but it has not been explored to the fullest. More details with proper reasoning should also be  provided in this section to give more details on this topic as well.
• In this section, it is important to explain what neuromorphic computing is, what material and  device properties are most critical for spintronic neuromorphic systems, and discuss in detail the scalability  and power issues of current CMOS based systems since these are major problems in the field.
• In my opinion, it is crucial to provide a full list of comparisons of spin-based  neuromorphic devices (spintronic memristors, MTJ-based synapses, SOT-driven  neurons, and domain-wall motion devices) with respect to power consumption, non-volatility, speed,  scalability, and operating temperature.
• The review should also pay more attention to the main difficulties in 2D spintronics for  neuromorphic computing such as the efficiency of spin injection, spin relaxation, fabrication variability, the scalability  of the device and the compatibility with CMOS and new architectures. It will be useful to have a  separate section of challenges in this field so that the researchers who work in this area will be able to  concentrate on the solution of certain problems.
• The review should also address the emerging trends such as next generation 2D materials, room temperature  quantum spintronics, AI enabled material discovery for energy efficient devices and 3D integration of  2D spintronic networks as these are important to the advancement of the field.
• The conclusion should restate the key questions, what do we know about AI creativity, why should  we care about it, and what are the implications for the future of AI and work. Finally,  research gaps and future directions are identified, and the need for interdisciplinary collaboration is emphasized.
• Neuromorphic computing is a fast emerging field with ongoing research progress. However, this review  includes only one reference from the year 2025. To make sure it is up to date and  represents the last developments and the present tendencies of the research, I suggest referencing the following extensive studies from  2025.
  10. https://doi.org/10.1021/acs.nanolett.4c06118
  11. https://doi.org/10.1016/j.pmatsci.2025.101471
  12. https://doi.org/10.1021/acs.chemrev.4c00631

Comments on the Quality of English Language

It can be improved.

Author Response

Review report 2

• The introduction section is not satisfactory at all. For a review paper, introduction
section should comprehensively provide the targeted topic for the easy understanding
of the readers about the topic background, related literature review, and significance
of writing this review papers. A few suggestions to strengthen the introduction section
include:

We thank the reviwer for constructive comments and suggestions. We have revised
accordingly, the major changes include:
• Added one more paragraph at the end of Section I to conclude this paper’s content.
• Explained the advantages of using 2D spintronics over 2D materials.
• Clarified the scope of this review paper.
• Discussed principles of spintronics device integration.
• Compared neuromorphic computing capabilities to van Neumann architecture.
• Pointed out the future opportunities to develop the area.

1. Provide a brief overview of the paper’s content while critically addressing scalability
and energy efficiency challenges in conventional CMOS systems.

We have changed the introduction by adding one more paragraph to the end of Section
I, by saying “This review paper summarizes recent progress in the use of spintronics
for neuromorphic computing, with a particular focus on the outlook and future directions
for integrating 2D spintronic materials into neuromorphic hardware. It covers the
fundamentals of neuromorphic computing, an overview of 2D spintronic materials, and
existing spintronic devices and architectures, such as those based on magnetic tunnel
junctions and magnetic skyrmions. Additionally, it discusses the scalability and energy
efficiency of current spintronic neuromorphic systems and explores potential applications
of 2D spintronics in neuromorphic computing.”

2. Compare the advantages of 2D materials like graphene, TMDs, and MXenes in
neuromorphic systems.

2D materials (such as graphene, TMDs, and MXenes) excel in scalability, high-speed
operation, energy efficiency, and flexibility. These materials are particularly advantageous
for low-power, high-speed operations and their integration into compact, flexible
neuromorphic devices. In addition to sharing these benefits, 2D spintronics using 2D
magnets offer unique advantages in memory storage, non-volatility, and stability. This
makes them particularly well-suited for long-term data retention, energy-efficient memory
operations, and high parallelism-key features for advanced neuromorphic computing
systems.
The combined use of 2D materials and 2D magnets leverages their complementary
strengths, offering a promising pathway to achieving the performance demands of nextgeneration
neuromorphic devices.
The first paragraph has been added into Page 4 in the main text.

3. Discuss the principles of spintronic device integration, computing efficiency, and
scalability.

The integration of spintronic devices into computing systems relies on harnessing the
spin degree of freedom of electrons, in addition to their charge, to process and store information.
This is achieved through key components like magnetic tunnel junctions (MTJs)
and spin-transfer torque (STT) or spin-orbit torque (SOT) mechanisms, which enable
data manipulation with lower energy consumption compared to conventional chargebased
electronics. Spintronic devices offer non-volatility, allowing data to be retained
without continuous power, which enhances energy efficiency. Furthermore, they enable
parallel processing and in-memory computation, reducing the need for data transfer between
memory and processors—overcoming the von Neumann bottleneck. In terms of
scalability, spintronic devices can be miniaturized to the nanoscale, making them suitable
for high-density integration in advanced neuromorphic systems. Emerging 2D magnets
further enhance scalability by enabling ultra-thin, flexible device architectures. However,
challenges remain in maintaining consistent switching behavior, reducing error rates,
and achieving high-speed operation at large scales. Successful integration of spintronics
promises to improve computing efficiency by offering faster, more energy-efficient, and
densely packed neuromorphic architectures, paving the way for future high-performance
computing applications.

This paragraph was added to Pages 8-9 in Section IV in the updated main text.

4. Multiple paragraphs discussing about the same point should be merged together
to improve the structure of introduction section by ensuring each paragraph focuses on a
single point. For instance, paragraphs 2 and 3, both discussing the role of 2D materials
in memory devices, should be merged for better coherence.

Paragraphs 2 and 3 discuss different aspects of materials for neuromorphic computing.
Paragraph 2 focuses on 2D materials, primarily those without intrinsic magnetic order.
Toward the end of this paragraph, we provide an example of inducing magnetism in
graphene through the proximity effect with antiferromagnets. In Paragraph 3, we shift
the focus to the use of magnetic materials, highlighting their unique advantages in memory
storage and spintronic applications.

5. To conclude the introduction section, provide a breif summary about each of the
remaining section of this review paper.

A summary of pepar’s content has been added to the end of introduction part.

• For readers not familiar with the basics, the review should further include a more
explicit and comprehensive description of spintronic effects such as spin transport, coherence,
and injection in 2D materials. It would also be useful to distinguish between
2D materials (graphene, TMDs, MXenes), and other materials employed in spintronics,
and to explain how 2D materials differ from them and are superior to them in these
applications.

We have added one more paragraph addressing the difference between using 2D materials
and 2D magnets in Page 4 of the main text.

• The subheading of device architecture should be strengthen more by the addition of
more architectures like spin-FETs and spin logic devices, to provide readers with a clear
understanding of the device structures used in neuromorphic computing.

We expanded Page 3 by providing a comparison between 2D materials and 2D spintronics
using 2D magnets. Including additional architectures would make the review less
focused. While we acknowledge the novelty of spin-FETs and TFETs, these devices fall
outside the scope of our discussion as they have not yet been proposed for neuromorphic
computing.

• The area of ‘Scalability and Energy Efficiency’ has been explored to some extent in
this work, but it has not been explored to the fullest. More details with proper reasoning
should also be provided in this section to give more details on this topic as well.
We added the part explaining why 2D materials can contribute to scalability and
energy efficiency in Pages 4 & 8-9 highlighted in yellow color in the main text.
• In this section, it is important to explain what neuromorphic computing is, what
material and device properties are most critical for spintronic neuromorphic systems, and
discuss in detail the scalability and power issues of current CMOS based systems since
these are major problems in the field.

We changed the first paragraph in the introduction by adding the discussion on neuromorphic
computing basics “Neuromorphic computing represents a burgeoning approach
to computational architecture and device synthesis. Aimed at capturing the human
brain’s efficiency, adaptability, and massive parallelism, neuromorphic computing devices
differ from the conventional von Neumann architecture by integrating memory and computation
in a unified framework (Fig. 1). In spintronic neuromorphic systems, critical
material and device properties include low magnetic damping for energy efficiency, high
spin polarization for improved signal fidelity, and stable magnetic anisotropy to maintain
data integrity. Magnetic materials play a crucial role due to their fast switching
speeds and non-volatility, which facilitate synaptic behavior and memory retention. As
CMOS technology approaches the limits of Moore’s Law, current CMOS-based systems
face substantial challenges in scalability and power consumption. Further miniaturization
is hindered by increased leakage currents and heat dissipation, reducing energy efficiency.
Moreover, the von Neumann bottleneck, which separates computation and memory, imposes
fundamental limits on processing speed and increases latency and power consumption.
Neuromorphic computing addresses these issues by enabling data-intensive tasks
like pattern recognition and sensory processing with significantly lower energy requirements.
Among various material systems explored for neuromorphic devices, 2D materials,
van der Waals (vdW) heterostructures, and spintronic phenomena have emerged as key
enablers, offering unique advantages in scalability, energy efficiency, and novel device
functionalities.”

• In my opinion, it is crucial to provide a full list of comparisons of spin-based neuromorphic
devices (spintronic memristors, MTJ-based synapses, SOT-driven neurons, and
domain-wall motion devices) with respect to power consumption, non-volatility, speed,
scalability, and operating temperature.

We appreciate the reviewer’s suggestion. Our focus is on utilizing 2D magnets for
neuromorphic computing. However, there is still a lack of research on employing 2D
magnets in spintronic memristors or domain-wall motion devices. Therefore, our comparison
primarily considers MTJs, SOT devices, spin valves, skyrmions, and van der Waals
heterostructures, where 2D magnets are studied and fabricated to evaluate single-device
performance.

• The review should also pay more attention to the main difficulties in 2D spintronics
for neuromorphic computing such as the efficiency of spin injection, spin relaxation,
fabrication variability, the scalability of the device and the compatibility with CMOS
and new architectures. It will be useful to have a separate section of challenges in this
field so that the researchers who work in this area will be able to concentrate on the
solution of certain problems.

“Despite all challenges, some emerging trends in spintronics are shaping the future of
energy-efficient and high-performance computing. Next-generation 2D materials, such as
magnetic vdW materials and topological insulators, offer exceptional properties like high
spin polarization, low-dimensional confinement, and compatibility with existing semiconductor
technologies. Room-temperature quantum spintronics is another transformative
area, enabling the manipulation of spin states without requiring extreme cooling, which is
crucial for practical applications in neuromorphic and quantum computing. Additionally,
AI-enabled material discovery accelerates the identification and design of novel materials
with optimized magnetic, electronic, and thermal properties, facilitating the development
of more efficient spintronic devices. The 3D integration of 2D spintronic networks further
enhances scalability and performance by enabling dense, multi-layer architectures that
combine memory and logic functions, overcoming the limitations of traditional planar
designs. Together, these advancements promise to revolutionize the field by improving
energy efficiency, enhancing computational capabilities, and enabling new paradigms for
information processing.” was added to Section VII.

• The review should also address the emerging trends such as next generation 2D materials,
room temperature quantum spintronics, AI enabled material discovery for energy
efficient devices and 3D integration of 2D spintronic networks as these are important to
the advancement of the field.

Refer to previous reply. We have added one paragraph to combine next generation
materials, room temperature quantum spintronics, AI combination and 3D integration.

• The conclusion should restate the key questions, what do we know about AI creativity,
why should we care about it, and what are the implications for the future of AI
and work. Finally, research gaps and future directions are identified, and the need for
interdisciplinary collaboration is emphasized.

We added “Additionally, the key questions surrounding AI creativity-what it truly
means for a machine to be creative, how AI can contribute to creative processes, and
whether AI can genuinely replicate human-like innovation—remain critical. We now know
that AI has the potential to generate novel ideas and even mimic artistic processes, but
whether it can truly possess creativity in the human sense is still debated. This matters
because understanding AI’s role in creativity will shape how it integrates into industries,
disrupts traditional work models, and transforms the creative sectors. The implications
for the future of AI and work are profound, as AI could both complement and challenge
human workers in creative fields. However, significant research gaps remain, particularly
in terms of defining and measuring creativity in machines. Future directions should focus
on advancing AI’s capabilities using neuromorphic architecture, and there is a pressing
need for interdisciplinary collaboration between AI researchers, artists, and philosophers
to address these challenges and explore the full potential of AI creativity. These insights
highlight the transformative potential of 2D spintronics and electronics in shaping the
future of energy-efficient, high-performance computing systems.” to the last paragraph.
• Neuromorphic computing is a fast emerging field with ongoing research progress.

However, this review includes only one reference from the year 2025. To make sure it is up
to date and represents the last developments and the present tendencies of the research,
I suggest referencing the following extensive studies from 2025.
10. https://doi.org/10.1021/acs.nanolett.4c06118
11. https://doi.org/10.1016/j.pmatsci.2025.101471
12. https://doi.org/10.1021/acs.chemrev.4c00631
These three papers were added to Section II.

References Cited
1. Li, S., Zhang, Z., Mao, R., Xiao, J., Chang, L. & Zhou, J. A fast and energyefficient
SNN processor with adaptive clock/event-driven computation scheme and
online learning. IEEE Transactions on Circuits and Systems I: Regular Papers 68,
1543–1552 (2021).
2. Zhang, B., Lu, P., Tabrizian, R., Feng, P. X.-L. & Wu, Y. 2D Magnetic heterostructures:
spintronics and quantum future. npj Spintronics 2, 6 (2024).
3. Verma, G., Nisar, A., Dhull, S. & Kaushik, B. K. Neuromorphic accelerator for spiking
neural network using SOT-MRAM crossbar array. IEEE Transactions on Electron
Devices 70, 6012–6020 (2023).
4. Verma, G., Soni, S., Nisar, A. & Kaushik, B. K. Multi-bit MRAM based high performance
neuromorphic accelerator for image classification. Neuromorphic Computing
and Engineering 4, 014008 (2024).
5. Sosa, L., Wi, M., Barrera, M., Nasrullah, I. & Wu, Y. Simulating Pattern Recognition
Using Non-volatile Synapses: MRAM, Ferroelectrics and Magnetic Skyrmions. arXiv
preprint arXiv:2501.03450 (2025).
6. Joksas, D., AlMutairi, A., Lee, O., Cubukcu, M., Lombardo, A., Kurebayashi, H.,
Kenyon, A. J. & Mehonic, A. Memristive, Spintronic, and 2D-Materials-Based Devices
to Improve and Complement Computing Hardware. Advanced Intelligent Systems
4, 2200068 (2022).
7. Zhou, J. & Chen, J. Prospect of spintronics in neuromorphic computing. Advanced
Electronic Materials 7, 2100465 (2021).
8. Younis, M., Abdullah, M., Dai, S., Iqbal, M. A., Tang, W., Sohail, M. T., Atiq,
S., Chang, H. & Zeng, Y.-J. Magnetoresistance in 2D Magnetic Materials: From
Fundamentals to Applications. Advanced Functional Materials, 2417282 (2025).

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

A report of reviewing a manuscript JLPEA-3514013-R1

(2D Spintronics for Neuromorphic Computing with Scalability and Energy Efficiency)

To the authors:
In the author's response letter, the authors made satisfactory replying to my previous 19 comments. The revised manuscript has improved in accordance with my comments. In addition, several paragraphs are newly added in section IV (Spintronic Device Architectures for Neuromorphic Computing) and section VII (Challenges and future outlooks), so it is recommended for publication.

Reviewer 3 Report

Comments and Suggestions for Authors

The revised version can be accepted for publication in its present form.