Materials Horizons Blog

Patients with diabetes frequently develop chronic ulcers on their lower extremities that are extremely challenging to treat. To accelerate healing, numerous wound dressings have been developed to protect damaged tissue and possibly deliver therapeutics. However, few dressings can provide real-time monitoring of wound healing and patient health. Since diabetic ulcers are prone to persistent hyperglycemia, ischemia, prolonged inflammation, and bacterial infections, frequently assessing biomarkers is essential to optimize treatment and maximize healing outcomes.

In a recent work by Hou et. al., a multifunctional wound dressing is developed to not only shield diabetic foot wounds from insult and bacterial infections, but record wound site temperature, pH, and glucose levels. Additionally, the flexible electronic also monitors physiological signals such as patient heart rhythm and brain activity. To create the diagnostic dressing, a poly(ethylene glycol)-based polymer possessing cationic and anionic moieties, as well as the self-complementary hydrogen-bonding group, ureidopyrimidinone (UPy) is developed. The polymer abbreviated PADU after the first letter of each constituting monomer, forms flexible substrates due to intermolecular UPy interactions as well as ionic bonding between cationic and ionic segments.

 

 

When applied to metals, rubber, plastic, glass, and biological tissues, PADU substrates display remarkable adhesion, notably exceeding an adhesive strength of 13 kPa for both diabetic and healthy mouse skin. Following damage to the patch, dynamic hydrogen bonding also endows PADU materials with self-healing capabilities, restoring 96% of their original mechanical strength after 24 hours. In practical use this gives PADU substrates a huge advantage because may be able to sustain electronic function despite physical damage from bodily movement.

In addition to its superb mechanical properties, PADU was also found to exhibit antibacterial activity. Killing of bacteria is achieved through disruption of the cell membrane by cationic groups within the polymer, making it active against even drug-resistant pathogens. Co-cultures with mammalian cells and blood cells, however, showed no cytotoxic or hemolytic effects, and biocompatibility was further confirmed by subcutaneous implantation into mice.

When glucose, pH, and temperature sensors were printed on PADU substrates, real-time monitoring of environmental conditions were reported with high accuracy. Gold foil electrodes could also be easily integrated into the patch to produce dressings capable of collecting electrocardiograms, electromyographic signals, and electroencephalograms. Due to their strong adhesion to skin, PADU-based sensors could even collect electrophysiological readings while patients were in motion, exceeding the accuracy of commercial sensors.

The dressing developed in the work by Hou, et. al. surpasses the capabilities of most diabetic wound dressings due its multifunctional performance as an adhesive and versatile electronic. Although initially designed for diabetic ulcers, the technology will likely find broad use as a diagnostic dressing for any type of slow-healing wound. In future studies, electronic patches may even be studied as implantable materials to monitor surgical site healing or internal injuries. Overall, the work demonstrates the vast utility of flexible bioelectronics and highlights the importance of developing new devices that not only treat but monitor chronic wounds.

To find out more, please read:

Skin-adhesive and self-healing diagnostic wound dressings for diabetic wound healing recording and electrophysiological signal monitoring
Zishuo Hou, Tengjiao Wang, Lei Wang, Junjie Wang, Yong Zhang, Qian Zhou,   Zhengheng Zhang, Peng Li and Wei Huang
Mater. Horiz., 2024, Advance Article, DOI: 10.1039/D3MH02064A

---

About the blogger

---

Kelsey DeFrates is a Materials Horizons Community Board member. She received her Bachelor of Science in Bioengineering at Rowan University (Glassboro, New Jersey) in 2018 and completed her Ph.D. in Bioengineering through the University of California, Berkeley – UCSF joint graduate program in 2023. For her thesis research, Kelsey worked with Professor Phillip Messersmith at UC Berkeley on the development of supramolecular materials for drug delivery and tissue regeneration. In 2023, Kelsey joined Professor Christopher Hernandez at UCSF as a Chancellor’s Postdoctoral Fellow, where she is working to understand how bacteria sense and respond to physical stimuli. In future work, Kelsey hopes to use this information to develop engineered living materials for healthcare and sustainable building.

Neuromorphic computing, inspired by the structure of the human brain, aims to overcome the limitations of traditional computing architectures by more closely integrating processing and memory functions. It is believed that this approach is a step towards dramatically improving the efficiency of artificial neural networks by in-memory computing. Specifically, compared to conventional graphics processing units, the memristor crossbars connected by synapses would markedly enhance the training and inference of the artificial neural networks in terms of speed and power for natural language processing, image classification, and so forth.

Due to their tantalizing properties, including unique control of ionic processes, mechanical flexibility, and low cost processability, organic materials, such as small molecules, polymers, graphene oxides, and halide perovskites, have sparked considerable research interests for crossbar memristive materials for artificial neural networks. Nevertheless, some weaknesses, such as dissatisfactory environmental stability, irreproducible switching behavior, and lack of understanding of the switching mechanisms, inevitably limit their applications for such application. Thus, it is of vital importance to rationally design new organic materials for crossbar memristors and synapses, to thoroughly understand their mechanisms, and to characterize their performance.

Recently, Vishwanath et al. designed, fabricated, and evaluated one-dimensional halide perovskites for crossbar memristive materials for application in artificial neural networks. They synthesized two kinds of one-dimensional halide perovskites, one with the organic cation of propylpyridinium and the other with the organic cation of benzylpyridinium. The substituted pyridines were used as templating agents to construct the one-dimensional structures. Compared to the three-dimensional perovskites, these exhibit better resistive switching performance due to their larger band gaps.

In order to maximize the improvement of their reliability, endurance, and retention, a device configuration of Ag/PMMA/HP/PEDOT:PSS/ITO was adopted, in which the halide perovskites switching matrix is sandwiched between the poly(methyl methacrylate) (PMMA) isolated layers and the poly(3,4-ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS). By employing the widely-used solution-processable technique, the team elaborately fabricated the largest dot point and crossbar halide perovskites memristive arrays so far (50000 devices across an area of 100 cm² and 16×16 crossbar).

Furthermore, they comprehensively analyzed the analog programming window for the halide perovskites. Concurrently, a spiking neural network with halide perovskite synapses was trained to classify the handwritten digits from the Modified National Institute of Standards and Technology database, which corroborates the applicability of the spike-timing-dependent plasticity learning of one-dimensional halide perovskite memristive synapses.

In summary, this novel study pioneers a new path for high-performance neuromorphic computing with innovative halide perovskites as the active material. These insightful results not only offer a solid foundation for the future explorations of halide perovskites in state-of-the-art neuromorphic computing, but also highlight the significance of materials innovation in unlocking the potential of next-generation computing technologies.

To find out more, please read:

High-performance one-dimensional halide perovskite crossbar memristors and synapses for neuromorphic computing
Sujaya Kumar Vishwanath, Benny Febriansyah, Si En Ng, Tisita Das, Jyotibdha Acharya, Rohit Abraham John, Divyam Sharma, Putu Andhita Dananjaya, Metikoti Jagadeeswararao, Naveen Tiwari, Mohit Ramesh Chandra Kulkarni, Wen Siang Lew, Sudip Chakraborty, Arindam Basuf and Nripan Mathews
Mater. Horiz., 2024, Advance Article, DOI: 10.1039/d3mh02055j

---

About the blogger

---

Wen Shi is currently an Associate Professor at School of Chemistry, Sun Yat-sen University and a Materials Horizons Community Board member. He received his Ph.D. in physical chemistry from Tsinghua University in 2017. From 2017 to 2021, he worked as a scientist at Institute of High Performance Computing (IHPC), Agency for Science, Technology and Research (A*STAR) in Singapore. Dr. Shi’s current research interests are in theoretical computations and simulations of functional materials.

 

In the ever-evolving landscape of materials engineering, researchers are pushing the boundaries of sustainability and functionality. One groundbreaking avenue of exploration is the integration of mycelium, the root-like structure of fungi, into biocomposites. In this blog post, we delve into a recent study that harnesses the unique properties of mycelium through 3D printing and indirect inoculation, resulting in a material with enhanced mechanical strength and diverse applications.

Shen et al. formulated a biocomposite designed to offer mechanical robustness and compatibility with mycelium. To achieve this, the researchers ingeniously selected chitosan and cellulose and introduced leftover coffee grounds as a sustainable source of nutrients. This careful combination created a material that not only exhibited shear thinning behavior, ideal for 3D printing but also laid the foundation for mycelium colonization (Figure 1).

One distinguishing feature of this study is the use of indirect inoculation for mycelium colonization. Traditionally, direct inoculation involves mixing the inoculum with the composite material before printing. However, the researchers chose a different route, incubating printed samples on a bed of live mycelium. This indirect approach, although taking longer for full colonization, turned out to be more effective.

The study of the mechanical properties of biocomposites revealed a strong influence on the orientation of the 3D printing tool path and the alignment of the cellulose fibers. The authors printed parts of different shapes, and the mechanical properties were dependent on the printing design. However, the fully colonized material showed a notable increase in mechanical strength, surpassing previously reported mycelium composites.

 

The influence of mycelium extended beyond mechanical properties to wettability and absorption characteristics. The fully colonized composite developed a smooth hydrophobic “skin,” demonstrating improved water contact angles. Under submerged conditions, the colonized compound demonstrated lower water absorption and volume swelling, attributed to the presence of hydrophobic mycelial hyphae.

The team explored the capabilities of mycelium to develop biosealed mycelium containers as self-sealing living boxes and the creation of flexible textile-like materials through precise 3D printing and mycelium colonization. By printing biocomposite panels with consistent gaps and allowing mycelium to cover them, flexible hinges were formed, enabling the creation of a material capable of bending and stretching in multiple directions.

In conclusion, the fusion of 3D printing, indirect inoculation, and mycelium colonization represents a leap forward in the field of sustainable biocomposites. The mechanical properties, wetting characteristics, and adaptability of the biocomposite open avenues for green alternatives in packaging, textiles, and more.

To find out more, please read:

Robust myco-composites: a biocomposite platform for versatile hybrid-living materials
Sabrina C. Shen, Nicolas A. Lee, William J. Lockett, Aliai D. Acuil, Hannah B. Gazdus, Branden N. Spitzerab  and Markus J. Buehler
Mater. Horiz., 2024, Advance Article, DOI: 10.1039/D3MH01277H

---

About the blogger

---

Dr. Danila Merino is the PI of the SusBioComp group at POLYMAT and a Materials Horizons Community Board member. Her research focuses on the development of new sustainable biocomposite materials derived from biomass specifically designed to protect and enhance agricultural and food systems with minimal environmental impact.

To address the need for large-scale electrochemical energy storage (EES), much research attention has moved beyond Li-ion batteries due to safety and security-of-supply issues. Sodium, an alkali-metal which is much more abundant and well-distributed globally, is of keen interest as its mining process is cleaner and freer from ethical concern. Moreover, to avoid the high flammability of organic electrolytes, some researchers are looking towards aqueous sodium-ion batteries as a potential contender for future EES systems. This has the added benefit of increasing the ionic conductivity by as much as two orders of magnitude versus organic equivalents, potentially enabling higher rate capability. However, moving from traditional carbonate-based electrolytes to water means a narrowing of the electrochemical stability window and the need for electrodes to facilitate the intercalation and de-intercalation of hydrated cations. Given the smaller accessible voltage and the larger charge carrier, aqueous sodium-ion batteries are still plagued by low specific energy and limited lifespans.

Therefore, the development of new electrode materials to maximise specific capacity is an important research direction. For the anode material, much attention has been paid to the development of polyanionic materials, such as sodium superionic conductors (NASICON), but also carbon-based materials such as polypyrrole and polyimide systems. In recent work by Maria K. Ramos et al., however, the synthesis of a graphene-based composite thin-film was presented, incorporating two compounds that had been shown to have high capacities but suffered from low conductivity and significant volume changes.

Specifically, the researchers highlighted the difficulty of producing ternary thin-films by traditional fabrication routes (e.g., spin-coating, vapour deposition etc.), spurring the development of a liquid-liquid interfacial route (LLIR) for the self-assembly of materials at the interface of immiscible liquids to give a continuous network that can be deposited on any solid substrate. MoS₂, known to facilitate the intercalation and de-intercalation of hydrated sodium ions, and non-toxic copper oxide nanoparticles with high theoretical specific capacity, were combined with graphene in this way to produce ternary films that were electrochemically characterised.

Interestingly, the researchers detailed three different thin-film preparation approaches using their LLIR method (Figure 1). The in-situ approach, whereby graphene oxide and Cu²⁺ were simultaneously reduced in a dispersion of MoS₂, yielded a thin-film anode material that demonstrated a very high specific capacity of 1377 mA h g^-1 (c.f. specific capacity of typical lithium-ion batteries is < 200 mA h g^-1).

In summary, the successful implementation of this in-situ liquid-liquid interfacial method for thin-film preparation provides encouragement for its use to produce other composite electrode materials, and a greater understanding of its scalability. The demonstration of such a high-capacity aqueous sodium-ion battery electrode should encourage greater exploration of this more sustainable, beyond-lithium EES technology.

To find out more, please read:

Nanoarchitected graphene/copper oxide nanoparticles/MoS₂ ternary thin films as highly efficient electrodes for aqueous sodium-ion batteriesMaria K. Ramos, Gustavo Martins, Luiz H. Marcolino-Junior, Márcio F. Bergamini, Marcela M. Oliveira and Aldo J. G. Zarbin
Mater. Horiz., 2023, 10, 5521-5537, DOI: 10.1039/d3mh00982c

---

About the blogger

---

Dr Josh J Bailey is an Illuminate Fellow at Queen’s University Belfast, focused on the implementation and optimisation of ionic liquids used in polymer electrolyte fuel cells and is a Materials Horizons Community Board member. He received his doctoral degree from University College London, UK, as part of the Centre for Doctoral Training in Advanced Characterisation of Materials, investigating electrode degradation in solid oxide fuel cells. His research interests span fuel cells, lithium-ion batteries, solid-state batteries, and flow batteries, both in terms of the design of novel electrodes, electrolytes, and membrane materials, as well as the study of materials degradation, with a view to improving performance and durability.

For a considerable period, the engine of progress was fuelled primarily by economic incentives. However, this paradigm has shifted due to increased awareness of the environmental consequences of society. Focus has turned towards embracing sustainability as a precursor to assimilating the fruits of progress into industry. This trend leverages the conversion of different waste feedstock like plastics, metals, etc, into new added value matter. The added value matter could be fuels, solvents, organic substrates, new polymers, and functional materials.

Embracing the notion that “the new is often the well-forgotten old,” the use of biomass as a feedstock for materials with practicable qualities has been revisited and revitalized. The biomass is feedstock mainly derived from agricultural and forestry resources, and animal resources. Compared to other waste feedstock, such as plastic, electronic, and construction waste, biomass already fits more within the sustainable economic strategy due to its natural origin. The other side of this coin is the insufficient mechanical properties of biomass-derived materials which leads to poor durability and recyclability of functional materials from biomass.

Recent work from Leixiao Yu, Lingyan Gao, Shengyi Dong and team suggests a supramolecular strategy to overcome these limitations. They reported the conversion of 6 types of biomass (cellulose, guar gum, sericin protein, chitin, corn protein and potato starch) to functional materials via copolymerization with thioctic acid (TA) to afford poly[TA-biomass]. The material formation is driven by hydrogen bonding between TA and the polar functional groups in the biomass. Despite such non-covalent forces being reversible and inherently weaker than covalent bonds, prepared materials are proven to be highly impact resistant. The prepared poly[TA-biomass] is highly adhesive and water-resistant, however, it could be fully depolymerized by simple ethanol treatment and involved in the next cycle of polymerization-utilization without any obvious decay in mechanical strength. This anticipates potential applications of poly[TA-biomass as anti-water, impact resistant materials. The team expanded the potential application directions to the biomedical field by demonstrating high biocompatibility, nontoxicity, and antimicrobial effects towards both gram-positive and negative bacteria, attributed to TA. For instance, the newly prepared poly[TA-biomass] may hold promise for smart packaging or wound healing materials.

This recent work is a perfect example of the “waste to wealth” approach, where materials chemistry assisted in transforming common feedstock into functional materials. Combining waste feedstock with a supramolecular strategy is a promising concept that can be broadened to the use of other types of feedstock (plastic, metal) and a broad family of non-covalent interactions (hydrogen bonding, π- π stacking, hydrophobic effects). At the moment, however, this research directs the attention of the community to biomass as a promising feedstock for functional materials design.

To find out more, please read:

A supramolecular approach for converting renewable biomass into functional materials
Yunfei Zhang, Changyong Cai, Ke Xu, Xiao Yang, Leixiao Yu, Lingyan Gao and Shengyi Dong
Mater. Horiz., 2024, Advance Article, DOI: 10.1039/D3MH01692G

---

About the blogger

---

Dr Olga Guselnikova is a member of the Materials Horizons Community Board. She recently joined the Center for Electrochemistry and Surface Technology (Austria) to work on functional materials as a group leader. Dr. Guselnikova received her PhD degree in chemistry from the University of Chemistry and Technology Prague (Czech Republic) and Tomsk Polytechnic University (Russia) in 2019. Her research interests are related to surface chemistry for functional materials. This means that she is applying her background in organic chemistry to materials science: plasmonic and polymer surfaces are hybridized with organic molecules to create high-performance elements and devices.

As we progress towards sustainable sources of energy, achieving efficient hydrogen generation through water splitting becomes increasingly vital. However, the challenges associated with electrode and membrane degradation due to corrosion in seawater hinder large-scale applications. While indirect seawater splitting through pre-desalination can circumvent corrosion issues, it introduces a drawback demanding additional energy input, rendering it economically less attractive. Therefore, the development of cost-effective water splitting electrocatalysts is essential for making the overall electrolysis process economically viable for widespread adoption and industrialization. Metal oxide electrocatalysts with active site engineering is a cutting-edge strategy to obtain high activity and durable catalysts for long-lasting performance under harsh saline conditions.

Recent work by Meng and team presents heterogeneous spin state molybdenum dioxide (MoO₂) as a promising electrocatalyst to address the above critical challenges. Their novel approach involves the incorporation of Fe into MoO₂ nanosheets on Ni foam (Fe-MoO₂/NF) through a rapid carbothermal shocking method. This synthetic process facilitates the lattice dislocations, effectively exposing rich O vacancies and inducing a low-oxidation state in Mo sites, especially during the rapid Joule heating process. This results in the manipulation of spin states between Fe and Mo atoms. The resulting catalyst demonstrates remarkable performance for hydrogen evolution reaction (HER) and oxygen evolution reaction (OER) in alkaline seawater.

The integration of heterogeneous spin states into MoO₂ disrupts the d–d orbital coupling resulting in modified electronic configuration. This significantly affects the binding energy between the active sites and reaction intermediates, thereby enhancing the electrocatalytic activity. The catalytic activity of Fe–MoO₂/NF is demonstrated by ultralow overpotentials recorded for both HER (17 mV@10 mA cm⁻²) and OER (310 mV@50 mA cm⁻²). Furthermore, the catalyst exhibits high selectivity in alkaline seawater splitting, showcasing its potential for efficient hydrogen production in challenging environmental conditions. To demonstrate the practical applicability of this newly developed electrocatalyst, the Fe–MoO₂/NF is assembled into an anion exchange membrane seawater electrolyser that achieves a low energy consumption of 5.5 kW h m⁻³, emphasizing its practical application in renewable energy systems.

An important aspect of this research is the successful coupling of Fe–MoO₂/NF with a solar-driven electrolytic system, yielding a solar-to-hydrogen efficiency of 13.5%. This demonstrates the catalyst’s compatibility with solar energy, opening avenues for sustainable and clean hydrogen production. Theoretical insights into the electronic structure of Fe-incorporated MoO₂, along with the abundance of oxygen vacancies, provides a deeper understanding of the catalytic mechanisms involved. Distortion of the Mo–O bonds, optimized through this method, plays a crucial role in enhancing the binding energy of adsorbed species during the electrochemical processes.

Looking forward, these findings hold significant promise for practical water splitting at a scalable level, especially in the context of solar-to-hydrogen production. The successful integration of Fe–MoO₂/NF into a solar-driven electrolytic system is a critical step toward sustainable and eco-friendly widespread adoption and integration into large-scale hydrogen production systems. In the pursuit of practical water splitting for solar-to-hydrogen production on a larger scale, the key challenge lies in making the process economically viable and technologically feasible. Thus, the prospects for this work include optimizing the synthesis method and extending for other heterogeneous spin such as Ni and Co for seawater splitting, which might extend the versatility of the proposed strategy, offering a simple and efficient approach for efficient electrocatalysts. The simplicity and efficiency of the proposed strategy make it an attractive option for large-scale implementation. As the demand for green energy solutions and the reduction of carbon footprints continue to grow, the significance of advancements in water-splitting, especially saline water electrolysis driven by solar energy, cannot be overstated. This study adds to the growing body of evidence that renewable energy sources can be used to produce hydrogen, which will pave the ways towards a more greener and sustainable energy future.

To find out more, please read:

Rapid carbothermal shocking fabrication of iron-incorporated molybdenum oxide with heterogeneous spin states for enhanced overall water/seawater splitting
Jianpeng Sun, Shiyu Qin, Zhan Zhao, Zisheng Zhang and Xiangchao Meng
Mater. Horiz., 2024, Advance Article, DOI: 10.1039/D3MH01757E

---

About the blogger

---

Shahid Zaman is currently a postdoctoral fellow at Hydrogen Research Institute, University of Quebec Trois-Rivières, Canada, and he is a Materials Horizons Community Board member. He received his Ph.D. in Material Physics and Chemistry from Huazhong University of Science and Technology in 2021. From 2021 to 2023, he worked as a postdoctoral fellow at the Southern University of Science and Technology, China. Dr. Shahid’s current research interests are nanomaterials for electrocatalysis in proton exchange membrane fuel cells and water electrolyzers.

Peripheral nerve injury is a common complication of surgical procedures and traumatic insults that can result in severe discomfort including chronic pain and sensory defects. Ensuring that the injured nerve regenerates is therefore critical for the long-term well-being of affected individuals. A promising therapeutic strategy to enhance regeneration is to target both the initial inflammation and subsequently promote axonal regrowth. How to deliver drugs sequentially and on time for each patient is a complex engineering challenge requiring innovative solutions.

Toward this end, Shan and the team developed a stimuli-responsive drug delivery scaffold for a dual drug delivery that makes it possible to target both inflammation and axonal sprouting sequentially. The team created a composite material that consists of a Poly-L-lactic acid (PLLA) shell that provides mechanical support to the regenerating axons and that can be loaded with the novel bioactive and stimuli-responsive scaffold.

To achieve a staggered drug release profile, the team encapsulated vitamin B12 (vB12) with anti-inflammatory properties directly inside calcium crosslinked alginate hydrogel and used an additional multilevel encapsulation approach consisting of microspheres loaded with nerve growth factor (NGF)-adsorbed nanoparticles enabling their delayed release compared to vB12. The team then leveraged ultrasound as a stimulus to hierarchically open polymer chains in ultrasound-responsive calcium cross-linked alginate hydrogels.  The rate of release of both vB12 and NGF was not predetermined and could be tuned by changing ultrasound intensity thus enabling the adjustment of the delivery profiles based on the individual patient’s healing progress. Interestingly, the ultrasound treatment alone improved peripheral nerve regeneration by promoting the secretion of neurotrophins in rodent models. Taken together, Shan and colleagues developed a novel strategy to promote nerve repair which utilizes ultrasound as an easily administrable stimulus that makes it possible to adjust the treatment of injured peripheral nerves based on patient’s needs.

To find out more, please read:

A responsive cascade drug delivery scaffold adapted to the therapeutic time window for peripheral nerve injury repair
Yizhu Shan, Lingling Xu, Xi Cui, Engui Wang, Fengying Jiang, Jiaxuan Li, Han Ouyang, Tailang Yin, Hongqing Feng, Dan Luo, Yan Zhang and Zhou Li
Mater. Horiz., 2024, Advance Article, DOI: 10.1039/D3MH01511D

 

---

About the blogger

---

Anna Stejskalova is currently a postdoctoral fellow at the Wyss Institute at Harvard Medical School and a member of the Materials Horizons Community Board. Dr. Stejskalova’s research focuses on female reproductive health and preterm birth utilizing organs on a chips and advanced biomaterials.

Quantitative prediction of material function based on multi-scale modeling is of vital importance for not only systematic performance optimization, but also precise materials design. Due to the nonintuitive and nontrivial structure-property relationship across different length scales, a comprehensive characterization of the hierarchical behavior of functional materials still remains a formidable challenge to date. To deal with this burning issue, recently Hernandez et al. made full use of data science techniques and developed an interpretable neural network architecture, viz. NestedAE, to link and quantify material properties across various length scales.

In NestedAE, an autoencoder is used to represent each physical scale of the materials, and a series of autoencoders are connected. Thus, the successive transfer of ‘‘important’’ information from one scale to another can be realized by the latent space of each autoencoder. In contrast to the previous approaches, in NestedAE each autoencoder possesses a different architecture and it is trained upon its own data set. Moreover, both the latents from the previous autoencoder and the features from the data set are reconstructed by the autoencoder.

To demonstrate the applicability of this newly developed machine learning architecture, Hernandez et al. employed NestedAE to compute the density-functional-theory bandgaps of metal halide perovskites based on their atomic and ionic properties. Furthermore, their power conversion efficiencies were also predicted. It was proven that the predicted results agreed well with the previous experimental observations, and its application on the metal halide perovskites established the correlation between the fundamental atomistic-level structural properties and the macroscopic device performance.

In summary, this computational study developed an interpretable machine learning architecture, NestedAE, to quantitatively predict the material properties at many length scales and to correlate the basic chemical structure and the macroscopic performance. These insightful results pioneer a new way for hierarchically optimizing and designing new functional materials.

To find out more, please read:

NestedAE: interpretable nested autoencoders for multi-scale materials characterization
Nikhil Thota, Maitreyee Sharma Priyadarshini, and Rigoberto Hernandez
Mater. Horiz., 2024, Advance article, DOI: 10.1039/D3MH01484C

 

---

About the blogger

---

Wen Shi is currently an Associate Professor at the School of Chemistry, Sun Yat-sen University and he is a Materials Horizons Community Board member. He received his Ph.D. in physical chemistry from Tsinghua University in 2017. From 2017 to 2021, he worked as a scientist at Institute of High Performance Computing (IHPC), Agency for Science, Technology and Research (A*STAR) in Singapore. Dr. Shi’s current research interests are in theoretical computations and simulations of functional materials.