How To Optimize Materials and Devices via Design of Experiments and Machine Learning: Demonstration Using Organic Photovoltaics

In silico analysis of design of experiment methods for metabolic pathway optimization

Microbial cell factories allow the production of chemicals presenting an alternative to traditional fossil fuel-dependent production. However, finding the optimal expression of production pathway genes is crucial for the development of efficient production strains. Unlike sequential experimentation, combinatorial optimization captures the relationships between pathway genes and production, albeit at the cost of conducting multiple experiments. Fractional factorial designs followed by linear modeling and statistical analysis reduce the experimental workload while maximizing the information gained during experimentation. Although tools to perform and analyze these designs are available, guidelines for selecting appropriate factorial designs for pathway optimization are missing. In this study, we leverage a kinetic model of a seven-genes pathway to simulate the performance of a full factorial strain library. We compare this approach to resolution V, IV, III, and Plackett Burman (PB) designs. Additionally, we evaluate the performance of these designs as training sets for a random forest algorithm aimed at identifying best-producing strains. Evaluating the robustness of these designs to noise and missing data, traits inherent to biological datasets, we find that while resolution V designs capture most information present in full factorial data, they necessitate the construction of a large number of strains. On the other hand, resolution III and PB designs fall short in identifying optimal strains and miss relevant information. Besides, given the small number of experiments required for the optimization of a pathway with seven genes, linear models outperform random forest. Consequently, we propose the use of resolution IV designs followed by linear modeling in Design-Build-Test-Learn (DBTL) cycles targeting the screening of multiple factors. These designs enable the identification of optimal strains and provide valuable guidance for subsequent optimization cycles.

Mapping the correlations between bandgap, HOMO, and LUMO trends for meta substituted Zn-MOFs

Bandgap is a key property that determines electrical and optical properties in materials. Modulating the bandgap thus is critical in developing novel materials particularly semiconductors with improved features. This study examines the bandgap, highest occupied molecular orbital (HOMO), and lowest unoccupied molecular orbital (LUMO) energy level trends in a metal organic framework, metal-organic framework 5 (MOF-5), as a function of Hammett substituent effect (with the constant σm in the meta-position of the benzene ring) and solvent dielectric effect (with the constant ε). Specifically, experimental design and response surface methodologies helped to assess the significance of trends and correlations between these molecular properties with σm and ε. While the HOMO and LUMO decrease with increasing σm, the LUMO exhibits greater sensitivity to the substituent's electron withdrawing capability. The relative difference in these trends helps to explain why the bandgap tends to decrease with increasing σm.

Self-Driving Laboratories for Chemistry and Materials Science

Self-driving laboratories (SDLs) promise an accelerated application of the scientific method. Through the automation of experimental workflows, along with autonomous experimental planning, SDLs hold the potential to greatly accelerate research in chemistry and materials discovery. This review provides an in-depth analysis of the state-of-the-art in SDL technology, its applications across various scientific disciplines, and the potential implications for research and industry. This review additionally provides an overview of the enabling technologies for SDLs, including their hardware, software, and integration with laboratory infrastructure. Most importantly, this review explores the diverse range of scientific domains where SDLs have made significant contributions, from drug discovery and materials science to genomics and chemistry. We provide a comprehensive review of existing real-world examples of SDLs, their different levels of automation, and the challenges and limitations associated with each domain.

Interpretable AI and Machine Learning Classification for Identifying High-Efficiency Donor-Acceptor Pairs in Organic Solar Cells

To enhance the efficiency of organic solar cells, accurately predicting the efficiency of new pairs of donor and acceptor materials is crucial. Presently, most machine learning studies rely on regression models, which often struggle to establish clear rules for distinguishing between high- and low-performing donor-acceptor pairs. This study proposes a novel approach by integrating interpretable AI, specifically using Shapely values, with four supervised machine learning classification models, namely, support vector machines, decision trees, random forest, and gradient boosting. These models aim to identify high-efficiency donor-acceptor pairs based solely on chemical structures and to extract important features that establish general design principles for distinguishing between high- and low-efficiency pairs. For validation purposes, an unsupervised machine learning algorithm utilizing loading vectors obtained from the principal component analysis is employed to identify crucial features associated with high-efficiency donor-acceptor pairs. Interestingly, the features identified by the supervised machine learning approach were found to be a subset of those identified by the unsupervised method. Noteworthy features include the van der Waals surface area, partial equalization of orbital electronegativity, Moreau-Broto autocorrelation, and molecular substructures. Leveraging these features, a backward-working model can be developed, facilitating exploration across a wide array of materials used in organic solar cells. This innovative approach will help navigate the vast chemical compound space of donor and acceptor materials essential in creating high-efficiency organic solar cells.

Multivariate Bayesian Optimization of CoO Nanoparticles for CO2 Hydrogenation Catalysis

The hydrogenation of CO2 holds promise for transforming the production of renewable fuels and chemicals. However, the challenge lies in developing robust and selective catalysts for this process. Transition metal oxide catalysts, particularly cobalt oxide, have shown potential for CO2 hydrogenation, with performance heavily reliant on crystal phase and morphology. Achieving precise control over these catalyst attributes through colloidal nanoparticle synthesis could pave the way for catalyst and process advancement. Yet, navigating the complexities of colloidal nanoparticle syntheses, governed by numerous input variables, poses a significant challenge in systematically controlling resultant catalyst features. We present a multivariate Bayesian optimization, coupled with a data-driven classifier, to map the synthetic design space for colloidal CoO nanoparticles and simultaneously optimize them for multiple catalytically relevant features within a target crystalline phase. The optimized experimental conditions yielded small, phase-pure rock salt CoO nanoparticles of uniform size and shape. These optimized nanoparticles were then supported on SiO2 and assessed for thermocatalytic CO2 hydrogenation against larger, polydisperse CoO nanoparticles on SiO2 and a conventionally prepared catalyst. The optimized CoO/SiO2 catalyst consistently exhibited higher activity and CH4 selectivity (ca. 98%) across various pretreatment reduction temperatures as compared to the other catalysts. This remarkable performance was attributed to particle stability and consistent H* surface coverage, even after undergoing the highest temperature reduction, achieving a more stable catalytic species that resists sintering and carbon occlusion.

De-risking Pretreatment of Microalgae To Produce Fuels and Chemical Co-products

Conversion of microalgae to renewable fuels and chemical co-products by pretreating and fractionation holds promise as an algal biorefinery concept, but a better understanding of the pretreatment performance as a function of algae strain and composition is necessary to de-risk algae conversion operations. Similarly, there are few examples of algae pretreatment at scales larger than the bench scale. This work aims to de-risk algal biorefinery operations by evaluating the pretreatment performance across nine different microalgae samples and five different pretreatment methods at small (5 mL) scale and further de-risk the operation by scaling pretreatment for one species to the 80 L scale. The pretreatment performance was evaluated by solubilization of feedstock carbon and nitrogen [as total organic carbon (TOC) and total nitrogen (TN)] into the aqueous hydrolysate and extractability of lipids [as fatty acid methyl esters (FAMEs)] from the pretreated solids. A range of responses was noted among the algae samples across pretreatments, with the current dilute Brønsted acid pretreatment using H2SO4 being the most consistent and robust. This pretreatment produced TOC yields to the hydrolysate ranging from 27.7 to 51.1%, TN yields ranging from 12.3 to 76.2%, and FAME yields ranging from 57.9 to 89.9%. In contrast, the other explored pretreatments (other dilute acid pretreatments, dilute alkali pretreatment with NaOH, enzymatic pretreatment, and flash hydrolysis) produced lower or more variable yields across the three metrics. In light of the greater consistency across samples for dilute acid pretreatment, this method was scaled to 80 L to demonstrate scalability with microalgae feedstocks.

Inside the Borate Anomaly: Leveraging a Predictive Modelling Approach to Navigate Complex Composition-Structure-Property Relationships in Oxyhalide Borate Glasses

This study employs a systematic and predictive modelling approach to investigate the structure and properties of multi-component borate glasses. In particular, this work is focused on understanding the individual and interaction effects of multiple constituents on several material properties. By leveraging advanced modeling techniques, this work examines how the inclusion and variation of B2O3, CaF2, TiO2, ZnO, and Na2CO3 influence the glass network, with particular attention to modifier fractions ≥ 30 mol%. This research addresses the gap in knowledge regarding the complex behavior of borate glasses in this high modifier fraction range, known as the borate anomaly, where prediction of glass structure and properties becomes particularly challenging. The use of a design of mixtures (DoM) approach facilitated the generation of polynomial equations indicating the influence of mixture components on various responses, enabling the prediction and optimization of glass properties over broad compositional ranges despite being within the anomalous region. This methodical approach not only advances our understanding of borate glass systems but also underscores the importance of predictive modelling in the accelerated design and development of novel glass materials for diverse applications.

Digitally Controlled Printing of Bioink Barriers for Paper-Based Analytical Devices: An Environmentally Friendly One-Step Approach

The patterning of hydrophilic paper with hydrophobic materials has emerged as an interesting method for the fabrication of paper-based devices (PADs). Herein, we demonstrate a digitally automated, easy, low-cost, eco-friendly, and readily available method to create highly hydrophobic barriers on paper that can be promptly employed with PADs by simply using a bioink made with rosin, a commercially available natural resin obtained from conifer trees. The bioink can be easily delivered with the use of a ballpoint pen to produce water- and organic solvent-resistant barriers, showing superior properties when compared to other methods such as wax-printing or permanent markers. The approach enables the pen to be attached to a commercially available cutting printer to perform the semiautomated fabrication of hydrophobic barriers for PADs. With the aid of digitally controlled optimization, together with features of machine learning and design of experiments, we show a thorough investigation on the barrier strength that can be further adjusted to the desired application's needs. Then, we explored the barrier sturdiness across various uses, such as wide range aqueous pH sensing and the harsh acidic/organic conditions needed for the colorimetric detection of cholecalciferol.

Design of Molecularly Imprinted Polymers Using Supercritical Carbon Dioxide Technology

The design and development of affinity polymeric materials through the use of green technology, such as supercritical carbon dioxide (scCO2), is a rapidly evolving field of research with vast applications across diverse areas, including analytical chemistry, pharmaceuticals, biomedicine, energy, food, and environmental remediation. These affinity polymeric materials are specifically engineered to interact with target molecules, demonstrating high affinity and selectivity. The unique properties of scCO2, which present both liquid- and gas-like properties and an accessible critical point, offer an environmentally-friendly and highly efficient technology for the synthesis and processing of polymers. The design and the synthesis of affinity polymeric materials in scCO2 involve several strategies. Commonly, the incorporation of functional groups or ligands into the polymer matrix allows for selective interactions with target compounds. The choice of monomer type, ligands, and synthesis conditions are key parameters of material performance in terms of both affinity and selectivity. In addition, molecular imprinting allied with co-polymerization and surface modification are commonly used in these strategies, enhancing the materials' performance and versatility. This review aims to provide an overview of the key strategies and recent advancements in the design of affinity polymeric materials using scCO2.

A Data-Driven Approach to Molten Salt Synthesis of N-Rich Carbon Adsorbents for Selective CO2 Capture

Applying a design of experiments methodology to the molten salt synthesis of nanoporous carbons enables inverse design and optimization of nitrogen (N)-rich carbon adsorbents with excellent CO2 /N2 selectivity and appreciable CO2 capacity for carbon capture via swing adsorption from dilute gas mixtures such as natural gas combined cycle flue gas. This data-driven study reveals fundamental structure-function relationships between the synthesis conditions, physicochemical properties, and achievable selective adsorption performance of N-rich nanoporous carbons derived from molten salt synthesis for CO2 capture. Taking advantage of size-sieving separation of CO2 (3.30 Å) from N2 (3.64 Å) within the turbostratic nanostructure of these N-rich carbons, while limiting deleterious N2 adsorption in a weaker adsorption site that harms selectivity, enables a large CO2 capacity (0.73 mmol g-1 at 30.4 Torr and 30 °C) with noteworthy concurrent CO2 /N2 selectivity as predicted by the ideal adsorbed solution theory (SIAST = 246) with an adsorbed phase purity of 91% from a simulated gas stream containing only 4% CO2 . Optimized N-rich porous carbons, with good physicochemical stability, low cost, and moderate regeneration energy, can achieve performance for selective CO2 adsorption that competes with other classes of advanced porous materials such as chemisorbing zeolites and functionalized metal-organic frameworks.

Top-Emitting Quantum Dot Light-Emitting Diodes: Theory, Optimization, and Application

The superior optical properties of colloidal quantum dots (QDs) have garnered significant broad interest from academia and industry owing to their successful application in self-emitting QD-based light-emitting diodes (QLEDs). In particular, active research is being conducted on QLEDs with top-emission device architectures (TQLEDs) owing to their advantages such as easy integration with conventional backplanes, high color purity, and excellent light extraction. However, due to the complicated optical phenomena and their highly sensitive optoelectrical properties to experimental variations, TQLEDs cannot be optimized easily for practical use. This review summarizes previous studies that have investigated top-emitting device structures and discusses ways to advance the performance of TQLEDs. First, theories relevant to the optoelectrical properties of TQLEDs are introduced. Second, advancements in device optimization are presented, where the underlying theories for each are considered. Finally, multilateral strategies for TQLEDs to enable their wider application to advanced industries are discussed. This work believes that this review can provide valuable insights for realizing commercial TQLEDs applicable to a broad range of applications.

Monolithically Integrated Complementary Ferroelectric FET XNOR Synapse for the Binary Neural Network

Neuromorphic computing, which mimics the structure and principles of the human brain, has the potential to facilitate the hardware implementation of next-generation artificial intelligence systems and process large amounts of data with very low power consumption. Among them, the XNOR synapse-based Binary Neural Network (BNN) has been attracting attention due to its compact neural network parameter size and low hardware cost. The previous XNOR synapse has drawbacks, such as a trade-off between cell density and accuracy. In this work, we show nonvolatile XNOR synapses with high density and accuracy using a monolithically stacked complementary ferroelectric field-effect transistor (C-FeFET) composed of a p-type Si MFMIS-FeFET at the bottom and a 3D stackable n-type Al:IZTO MFS-FeTFT, achieving 60F2 per cell (2C-FeFET). For adjusting the threshold voltage and improving the switching speed (100 ns) of n-type ferroelectric TFT, we employed a dual-gate configuration and a unique operation scheme, making it comparable to those of Si-based FeFETs. We performed array-level simulation with a 512 × 512 subarray size and a 3-bit flash ADC, demonstrating that the image recognition accuracies using the MNIST and CIFAR-10 data sets were increased by 3.17 and 14.07%, respectively, in comparison to other nonvolatile XNOR synapses. In addition, we performed system-level analysis on a 512 × 512 XNOR C-FeFET, exhibiting an outstanding throughput of 717.37 GOPS and an energy efficiency of 196.7 TOPS/W. We expect that our approach would contribute to the high-density memory systems, logic-in-memory technology, and hardware implementation of neural networks.

Machine Learning Driven Channel Thickness Optimization in Dual-Layer Oxide Thin-Film Transistors for Advanced Electrical Performance

Machine learning (ML) provides temporal advantage and performance improvement in practical electronic device design by adaptive learning. Herein, Bayesian optimization (BO) is successfully applied to the design of optimal dual-layer oxide semiconductor thin film transistors (OS TFTs). This approach effectively manages the complex correlation and interdependency between two oxide semiconductor layers, resulting in the efficient design of experiment (DoE) and reducing the trial-and-error. Considering field effect mobility (𝜇) and threshold voltage (Vth ) simultaneously, the dual-layer structure designed by the BO model allows to produce OS TFTs with remarkable electrical performance while significantly saving an amount of experimental trial (only 15 data sets are required). The optimized dual-layer OS TFTs achieve the enhanced field effect mobility of 36.1 cm2  V-1  s-1 and show good stability under bias stress with negligible difference in its threshold voltage compared to conventional IGZO TFTs. Moreover, the BO algorithm is successfully customized to the individual preferences by applying the weight factors assigned to both field effect mobility (𝜇) and threshold voltage (Vth ).

Autonomous and dynamic precursor selection for solid-state materials synthesis

Solid-state synthesis plays an important role in the development of new materials and technologies. While in situ characterization and ab-initio computations have advanced our understanding of materials synthesis, experiments targeting new compounds often still require many different precursors and conditions to be tested. Here we introduce an algorithm (ARROWS3) designed to automate the selection of optimal precursors for solid-state materials synthesis. This algorithm actively learns from experimental outcomes to determine which precursors lead to unfavorable reactions that form highly stable intermediates, preventing the target material's formation. Based on this information, ARROWS3 proposes new experiments using precursors it predicts to avoid such intermediates, thereby retaining a larger thermodynamic driving force to form the target. We validate this approach on three experimental datasets, containing results from over 200 synthesis procedures. In comparison to black-box optimization, ARROWS3 identifies effective precursor sets for each target while requiring substantially fewer experimental iterations. These findings highlight the importance of domain knowledge in optimization algorithms for materials synthesis, which are critical for the development of fully autonomous research platforms.

Using Data-Driven Learning to Predict and Control the Outcomes of Inorganic Materials Synthesis

The design of inorganic materials for various applications critically depends on our ability to manipulate their synthesis in a rational, robust, and controllable fashion. Different from the conventional trial-and-error approach, data-driven techniques such as the design of experiments (DoE) and machine learning are an effective and more efficient way to predictably control materials synthesis. Here, we present a Viewpoint on recent progress in leveraging such techniques for predicting and controlling the outcomes of inorganic materials synthesis. We first compare how the design choice (statistical DoE vs machine learning) affects the type of control it can offer over the resulting product attributes, information elucidated, and experimental cost. These attributes are supported by discussing select case studies from the recent literature that highlight the power of these techniques for materials synthesis. The influence of experimental bias is next discussed, followed finally by our perspectives on the major challenges in the widespread implementation of predictable and controllable materials synthesis using data-driven techniques.

Predictive Synthesis of Copper Selenides Using a Multidimensional Phase Map Constructed with a Data-Driven Classifier

Copper selenides are an important family of materials with applications in catalysis, plasmonics, photovoltaics, and thermoelectrics. Despite being a binary material system, the Cu-Se phase diagram is complex and contains multiple crystal structures in addition to several metastable structures that are not found on the thermodynamic phase diagram. Consequently, the ability to synthetically navigate this complex phase space poses a significant challenge. We demonstrate that data-driven learning can successfully map this phase space in a minimal number of experiments. We combine soft chemistry (chimie douce) synthetic methods with multivariate analyses via classification techniques to enable predictive phase determination. A surrogate model was constructed with experimental data derived from a design matrix of four experimental variables: C-Se bond strength of the selenium precursor, time, temperature, and solvent composition. The reactions in the surrogate model resulted in 11 distinct phase combinations of copper selenide. These data were used to train a classification model that predicts the phase with 95.7% accuracy. The resulting decision tree enabled conclusions to be drawn about how the experimental variables affect the phase and provided prescriptive synthetic conditions for specific phase isolation. This guided the accelerated phase targeting in a minimum number of experiments of klockmannite CuSe, which could not be isolated in any of the reactions used to construct the surrogate model. The reaction conditions that the model predicted to synthesize klockmannite CuSe were experimentally validated, highlighting the utility of this approach.

Screening for electrically conductive defects in thin functional films using electrochemiluminescence

Multifunctional thin films in energy-related devices often must be electrically insulating where a single nanoscale defect can result in complete device-scale failure. Locating and characterizing such defects presents a fundamental problem where high-resolution imaging methods are needed to find defects, but imaging with high spatial resolution limits the field of view and thus the measurement throughput. Here, we present a novel high-throughput method for detecting sub-micron defects in insulating thin films by leveraging the electrochemiluminescence (ECL) of luminol. Through a systematic study of reagent concentrations, buffers, voltage, and excitation time, we identify optimized conditions under which it is possible to detect sub-micron defects at high-throughput. Extrapolating from the signal to background observed for detecting 440 nm wide lines and 620 nm diameter circles, we estimate the minimum detectable features to be lines as narrow as 2.5 nm in width and pinholes as small as 70 nm in radius. We further explore this method by using it to characterize a nominally insulating poly(phenylene oxide) film and find conductive defects that are cross-correlated with high-resolution atomic force microscopy to provide feedback to synthesis. Given this assay's inherent parallelizability and scalability, it is expected to have a major impact on the automated discovery of multifunctional films.

CaTiO3/g-C3N4 heterojunction-based composite photocatalyst: Part I: Experimental design, kinetics, and scavenging agents' effects in photocatalytic degradation of gemifloxacin

A critical, challenging environmental issue is explored pollution of water supplies by discharging industrial/pharmaceutical/hospital/urban wastewaters into the aquatic environment. These needs introducing/developing novel photocatalysts/adsorbents/procedures for removing or mineralizing various pollutants in wastewater before discharging them into marine environments. Further, optimizing conditions to achieve the highest removal efficiency is an important issue. In this study, CaTiO₃/g-C₃N₄ (CTCN) heterostructure was synthesized and characterized by some identification techniques. The simultaneous interaction effects of the experimental variables on the boosted photocatalytic activity of CTCN in the degradation of gemifloxcacin (GMF) were studied in RSM design. The optimal values for four parameters were: catalyst dosage: 0.63 g L^-1, pH: 6.7, C_GMF: 1 mg L^-1, and irradiation time: 27.5 min, with approximately 78.2% of degradation efficiency. The quenching effects of the scavenging agents were studied to show the reactive species' relative importance in GMF photodegradation. The results illustrate that the reactive ^•OH plays a significant role, and the electron plays a minor role in the degradation process. The direct Z-scheme mechanism better described the photodegradation mechanism due to the great oxidative and reductive abilities of prepared composite photocatalysts. This mechanism is an approach to efficiently separating photogenerated charge carriers and improving the CaTiO₃/g-C₃N₄ composite photocatalyst activity. The COD has been performed to study the details of the mineralization of GMF. The pseudo-first-order rat (from the Hinshelwood model) constants of 0.046 min^-1 (t_1/2 = 15.1 min) and 0.048 min^-1 (t_1/2 = 14.4 min) were respectively obtained from the GMF photodegradation data and COD results. The prepared photocatalyst retained its activity after five reusing runs.

Active learning applied to automated physical systems increases the rate of discovery

Active machine learning is widely used in computational studies where repeated numerical simulations can be conducted on high performance computers without human intervention. But translation of these active learning methods to physical systems has proven more difficult and the accelerated pace of discoveries aided by these methods remains as yet unrealized. Through the presentation of a general active learning framework and its application to large-scale boundary layer wind tunnel experiments, we demonstrate that the active learning framework used so successfully in computational studies is directly applicable to the investigation of physical experimental systems and the corresponding improvements in the rate of discovery can be transformative. We specifically show that, for our wind tunnel experiments, we are able to achieve in approximately 300 experiments a learning objective that would be impossible using traditional methods.

Machine learning integrated photocatalysis: progress and challenges

Discovering efficient photocatalysts has long been the goal of photocatalysis, which has traditionally been driven by serendipitous or try-and-error strategies. Recent developments in photocatalysis integrated with machine learning techniques promise to accelerate the discovery of photocatalysts, but are also facing significant challenges. In this review, advances in machine learning integrated photocatalysis are first presented from the perspective of three main photocatalytic processes: light harvesting, charge generation and separation, and surface redox reactions. Next, progress in using machine learning to understand complex photoactivity-structure relationships and identify the factors governing activity follows. A future photocatalysis paradigm is then provided with the integration of artificial intelligence, robots and automation. Lastly, we discuss the current challenges in machine learning integrated photocatalysis. This review aims to provide a systematic overview and guidelines to the broad scientific community interested in photocatalysis and artificial intelligence for solar fuel synthesis.

A Brief Introduction to Chemical Reaction Optimization

From the start of a synthetic chemist's training, experiments are conducted based on recipes from textbooks and manuscripts that achieve clean reaction outcomes, allowing the scientist to develop practical skills and some chemical intuition. This procedure is often kept long into a researcher's career, as new recipes are developed based on similar reaction protocols, and intuition-guided deviations are conducted through learning from failed experiments. However, when attempting to understand chemical systems of interest, it has been shown that model-based, algorithm-based, and miniaturized high-throughput techniques outperform human chemical intuition and achieve reaction optimization in a much more time- and material-efficient manner; this is covered in detail in this paper. As many synthetic chemists are not exposed to these techniques in undergraduate teaching, this leads to a disproportionate number of scientists that wish to optimize their reactions but are unable to use these methodologies or are simply unaware of their existence. This review highlights the basics, and the cutting-edge, of modern chemical reaction optimization as well as its relation to process scale-up and can thereby serve as a reference for inspired scientists for each of these techniques, detailing several of their respective applications.

Tier-grown Expansion of Design-of-Experiments Parameter Spaces for Synthesis of a Nanometer-scale Macrocycle

A method to find optimum synthetic conditions was devised by combining a data-driven empirical model with a traditional mechanistic model. In this method, an experimental parameter space was empirically obtained by Design-of-Experiments optimizations with machine-learning supplements and was strategically expanded by examination of the mechanistic model of the reaction paths. An extra tier grown on the original 3×3×3 parameter space succeeded in allocating an optimum reaction condition in the expanded 3×3×4 parameter space. The method was specifically devised for the synthesis of a macrocycle, [n]cyclo-meta-phenylenes ([n]CMP), and the largest congener with n=12 was synthesized and fully characterized for the first time. Crystallographic and photophysical analyses revealed favorable features of [12]CMP for the material applications.

Computational Design of Antimicrobial Active Surfaces via Automated Bayesian Optimization

Biofilms pose significant problems for engineers in diverse fields, such as marine science, bioenergy, and biomedicine, where effective biofilm control is a long-term goal. The adhesion and surface mechanics of biofilms play crucial roles in generating and removing biofilm. Designing customized nanosurfaces with different surface topologies can alter the adhesive properties to remove biofilms more easily and greatly improve long-term biofilm control. To rapidly design such topologies, we employ individual-based modeling and Bayesian optimization to automate the design process and generate different active surfaces for effective biofilm removal. Our framework successfully generated optimized functional nanosurfaces for improved biofilm removal through applied shear and vibration. Densely distributed short pillar topography is the optimal geometry to prevent biofilm formation. Under fluidic shearing, the optimal topography is to sparsely distribute tall, slim, pillar-like structures. When subjected to either vertical or lateral vibrations, thick trapezoidal cones are found to be optimal. Optimizing the vibrational loading indicates a small vibration magnitude with relatively low frequencies is more efficient in removing biofilm. Our results provide insights into various engineering fields that require surface-mediated biofilm control. Our framework can also be applied to more general materials design and optimization.

Activated recovery of PVC from contaminated waste extension cord-cable using a weak acid

Waste electronic and electrical equipment are complex mixtures of valuable and/or toxic materials, which pose serious challenges in their recycling or disposal, for example, electrical transmission wires insulated in polyvinyl chloride materials. These materials are frequently found contaminated with toxic chemical elements, such as Pb, Hg, Cr, or Cd, and are discarded without decontamination. To resolve this problem, we developed a microwave-assisted extraction process to remove toxic metals from plastic e-waste. We processed diluted (30 wt%) citric acid at 210 °C for 1 h inside a pressurized vessel heated by microwave, and found it was suitable not only for the extraction of the toxic metals (∼100%) but also for a significant plastic recovery (>50 wt%). To predict an optimized process window, the support vector regression machine learning algorithm was applied, which reduced the amount of experimentation required while still giving accurate results. Conditions optimized for the reference sample also led to maximum extraction of toxic metals from real-life extension cord waste. We also report that the recovered plastic's properties remained intact after the extraction.

Data-Driven Analysis of Hole-Transporting Materials for Perovskite Solar Cells Performance

We have created a dataset of 269 perovskite solar cells, containing information about their perovskite family, cell architecture, and multiple hole-transporting materials features, including fingerprints, additives, and structural and electronic features. We propose a predictive machine learning model that is trained on these data and can be used to screen possible candidate hole-transporting materials. Our approach allows us to predict the performance of perovskite solar cells with reasonable accuracy and is able to successfully identify most of the top-performing and lowest-performing hole-transporting materials in the dataset. We discuss the effect of data biases on the distribution of perovskite families/architectures on the model's accuracy and offer an analysis with a subset of the data to accurately study the effect of the hole-transporting material on the solar cell performance. Finally, we discuss some chemical fragments, like arylamine and aryloxy groups, which present a relatively large positive correlation with the efficiency of the cell, whereas other groups, like thiophene groups, display a negative correlation with power conversion efficiency (PCE).

Synthesis of a Negatively Curved Nanocarbon Molecule with an Octagonal Omphalos via Design-of-Experiments Optimizations Supplemented by Machine Learning

A saddle-shaped nanocarbon molecule was synthesized, which revealed the existence of negative Gauss curvatures on a >3-nm molecular structure possessing 192 π-electrons. The synthesis was facilitated by a protocol developed with Design-of-Experiments optimizations and machine-learning predictions, and spectroscopy and crystallography were used to reveal the saddle-shaped structure of the molecule. Solution-phase analyses showed the presence of dimeric assembly, and crystallographic analyses revealed the stacked dimeric structures. The stacked crystal structure was scrutinized by various methods, including Gauss curvatures derived from the discrete surface theory of geometry, to reveal the important role of the molecular Gauss curvature in dimeric assembly.

Automation and Microfluidics for the Efficient, Fast, and Focused Reaction Development of Asymmetric Hydrogenation Catalysis

Automation and microfluidic tools potentially enable efficient, fast, and focused reaction development of complex chemistries, while minimizing resource- and material consumption. The introduction of automation-assisted workflows will contribute to the more sustainable development and scale-up of new and improved catalytic technologies. Herein, the application of automation and microfluidics to the development of a complex asymmetric hydrogenation reaction is described. Screening and optimization experiments were performed using an automated microfluidic platform, which enabled a drastic reduction in the material consumption compared to conventional laboratory practices. A suitable catalytic system was identified from a library of Ru^II -diamino precatalysts. In situ precatalyst activation was studied with ¹ H/³¹ P nuclear magnetic resonance (NMR), and the reaction was scaled up to multigram quantities in a batch autoclave. These reactions were monitored using an automated liquid-phase sampling system. Ultimately, in less than a week of total experimental time, multigram quantities of the target enantiopure alcohol product were provided by this automation-assisted approach.

Extrapolating missing antibody-virus measurements across serological studies

The development of new vaccines, as well as our understanding of key processes that shape viral evolution and host antibody repertoires, relies on measuring multiple antibody responses against large panels of viruses. Given the enormous diversity of circulating virus strains and antibody responses, comprehensively testing all antibody-virus interactions is infeasible. Even within individual studies with limited panels, exhaustive testing is not always performed, and there is no common framework for combining information across studies with partially overlapping panels, especially when the assay type or host species differ. Prior studies have demonstrated that antibody-virus interactions can be characterized in a vastly simpler and lower dimensional space, suggesting that relatively few measurements could predict unmeasured antibody-virus interactions. Here, we apply matrix completion to several large-scale influenza and HIV-1 studies. We explore how prediction accuracy evolves as the number of measurements changes and approximates the number of additional measurements necessary in several highly incomplete datasets (suggesting ∼250,000 measurements could be saved). In addition, we show how the method can combine disparate datasets, even when the number of available measurements is below the theoretical limit that guarantees successful prediction. This approach can be readily generalized to other viruses or more broadly to other low-dimensional biological datasets.

Machine Learning-Assisted Polymer Design for Improving the Performance of Non-Fullerene Organic Solar Cells

Despite the progress in machine learning (ML) in terms of prediction of power conversion efficiency (PCE) in organic photovoltaics (OPV), the effectiveness of ML in practical applications is still lacking owing to the complex structure-property relationship. Therefore, verifying the potential of ML through experiments can amplify the use of ML models. Herein, we developed a new series of π-conjugated polymers comprising benzodithiophene and thiazolothiazole with fluorination and alkylthio chains (PBDTTzBO, PFSBDTTzBO, and PFBDTTzBO) for non-fullerene (NF) acceptors based on our random-forest ML model for OPVs. Notably, the order of the ML-predicted PCEs of these polymers with IT-4F (9.93, 11.35, and 11.47%) was in good agreement with their experimental PCEs (5.24, 7.35, and 10.30%). In contrast, an inverse correlation was observed between the predicted (9.20, 12.29, and 12.20%) and experimental (11.98, 1.57, and 6.53%) PCEs with Y6. Both the findings are interpreted in terms of surface morphology, transient photoconductivity, charge carrier mobility, polymer orientation, and miscibility, quantified by the Flory-Huggins parameters. Herein, we present an ML-assisted polymer design for high-performance non-fullerene organic photovoltaics (NFOPVs) and elucidate the importance of the subtle alterations in the morphology of NFOPVs.

Limitations of machine learning models when predicting compounds with completely new chemistries: possible improvements applied to the discovery of new non-fullerene acceptors

We try to determine if machine learning (ML) methods, applied to the discovery of new materials on the basis of existing data sets, have the power to predict completely new classes of compounds (extrapolating) or perform well only when interpolating between known materials. We introduce the leave-one-group-out cross-validation, in which the ML model is trained to explicitly perform extrapolations of unseen chemical families. This approach can be used across materials science and chemistry problems to improve the added value of ML predictions, instead of using extrapolative ML models that were trained with a regular cross-validation. We consider as a case study the problem of the discovery of non-fullerene acceptors because novel classes of acceptors are naturally classified into distinct chemical families. We show that conventional ML methods are not useful in practice when attempting to predict the efficiency of a completely novel class of materials. The approach proposed in this work increases the accuracy of the predictions to enable at least the categorization of materials with a performance above and below the median value.

Geometrically engineered rigid island array for stretchable electronics capable of withstanding various deformation modes

Integration of rigid components in soft polymer matrix is considered as the most feasible architecture to enable stretchable electronics. However, a method of suppressing cracks at the interface between soft and rigid materials due to excessive and repetitive deformations of various types remains a formidable challenge. Here, we geometrically engineered Ferris wheel-shaped islands (FWIs) capable of effectively suppressing crack propagation at the interface under various deformation modes (stretching, twisting, poking, and crumpling). The optimized FWIs have notable increased strain at failure and fatigue life compared with conventional circle- and square-shaped islands. Stretchable electronics composed of various rigid components (LED and coin cell) were demonstrated using intrinsically stretchable printed electrodes. Furthermore, electronic skin capable of differentiating various tactile stimuli without interference was demonstrated. Our method enables stretchable electronics that can be used under various geometrical forms with notable enhanced durability, enabling stretchable electronics to withstand potentially harsh conditions of everyday usage.

Development of a fibrin-mediated gene delivery system for the treatment of cystinosis via design of experiment

Cystinosis is a rare disease, caused by a mutation in the gene cystinosin and characterised by the accumulation of cystine crystals. Advantages of biomaterial-mediated gene delivery include reduced safety concerns and the possibility to cure organs that are difficult to treat using systemic gene transfer methods. This study developed novel fibrin hydrogels for controlled, localised gene delivery, for the treatment of cystinosis. In the first part, fabrication parameters (i.e., DNA, thrombin, and aprotinin concentrations) were optimised, using a Design of Experiment (DOE) methodology. DOE is a statistical engineering approach to process optimisation, which increases experimental efficiency, reduces the number of experiments, takes into consideration interactions between different parameters, and allows the creation of predictive models. This study demonstrated the utility of DOE to the development of gene delivery constructs. In the second part of the study, primary fibroblasts from a patient with cystinosis were seeded on the biomaterials. Seeded cells expressed the recombinant CTNS and showed a decrease in cystine content. Furthermore, conditioned media contained functional copies of the recombinant CTNS. These were taken up by monolayer cultures of non-transfected cells. This study described a methodology to develop gene delivery constructs by using a DOE approach and ultimately provided new insights into the treatment of cystinosis.

Machine Learning and Optoelectronic Materials Discovery: A Growing Synergy

Novel optoelectronic materials have the potential to revolutionize the ongoing green transition by both providing more efficient photovoltaic (PV) devices and lowering energy consumption of devices like LEDs and sensors. The lead candidate materials for these applications are both organic semiconductors and more recently perovskites. This Perspective illustrates how novel machine learning techniques can help explore these materials, from speeding up ab initio calculations toward experimental guidance. Furthermore, based on existing work, perspectives around machine-learned molecular dynamics potentials, physically informed neural networks, and generative methods are outlined.

Machine Learning-Assisted Design of Material Properties

Designing functional materials requires a deep search through multidimensional spaces for system parameters that yield desirable material properties. For cases where conventional parameter sweeps or trial-and-error sampling are impractical, inverse methods that frame design as a constrained optimization problem present an attractive alternative. However, even efficient algorithms require time- and resource-intensive characterization of material properties many times during optimization, imposing a design bottleneck. Approaches that incorporate machine learning can help address this limitation and accelerate the discovery of materials with targeted properties. In this article, we review how to leverage machine learning to reduce dimensionality in order to effectively explore design space, accelerate property evaluation, and generate unconventional material structures with optimal properties. We also discuss promising future directions, including integration of machine learning into multiple stages of a design algorithm and interpretation of machine learning models to understand how design parameters relate to material properties. Expected final online publication date for the Annual Review of Chemical and Biomolecular Engineering, Volume 13 is October 2022. Please see http://www.annualreviews.org/page/journal/pubdates for revised estimates.

Instantaneous Property Prediction and Inverse Design of Plasmonic Nanostructures Using Machine Learning: Current Applications and Future Directions

Advances in plasmonic materials and devices have given rise to a variety of applications in photocatalysis, microscopy, nanophotonics, and metastructures. With the advent of computing power and artificial neural networks, the characterization and design process of plasmonic nanostructures can be significantly accelerated using machine learning as opposed to conventional FDTD simulations. The machine learning (ML) based methods can not only perform with high accuracy and return optical spectra and optimal design parameters, but also maintain a stable high computing efficiency without being affected by the structural complexity. This work reviews the prominent ML methods involved in forward simulation and inverse design of plasmonic nanomaterials, such as Convolutional Neural Networks, Generative Adversarial Networks, Genetic Algorithms and Encoder-Decoder Networks. Moreover, we acknowledge the current limitations of ML methods in the context of plasmonics and provide perspectives on future research directions.

Computational Method-Based Optimization of Carbon Nanotube Thin-Film Immunosensor for Rapid Detection of SARS-CoV-2 Virus

The recent global spread of COVID-19 stresses the importance of developing diagnostic testing that is rapid and does not require specialized laboratories. In this regard, nanomaterial thin-film-based immunosensors fabricated via solution processing are promising, potentially due to their mass manufacturability, on-site detection, and high sensitivity that enable direct detection of virus without the need for molecular amplification. However, thus far, thin-film-based biosensors have been fabricated without properly analyzing how the thin-film properties are correlated with the biosensor performance, limiting the understanding of property-performance relationships and the optimization process. Herein, the correlations between various thin-film properties and the sensitivity of carbon nanotube thin-film-based immunosensors are systematically analyzed, through which optimal sensitivity is attained. Sensitivities toward SARS-CoV-2 nucleocapsid protein in buffer solution and in the lysed virus are 0.024 [fg/mL]-1 and 0.048 [copies/mL]-1, respectively, which are sufficient for diagnosing patients in the early stages of COVID-19. The technique, therefore, can potentially elucidate complex relationships between properties and performance of biosensors, thereby enabling systematic optimization to further advance the applicability of biosensors for accurate and rapid point-of-care (POC) diagnosis.

Microfluidic Screening-Assisted Machine Learning to Investigate Vertical Phase Separation of Small Molecule:Polymer Blend

Solution-based thin-film processing is a widely utilized technique for the fabrication of various devices. In particular, the tunability of the ink composition and coating condition allows precise control of thin-film properties and device performance. Despite the advantage of having such tunability, the sheer number of possible combinations of experimental parameters render it infeasible to efficiently optimize device performance and analyze the correlation between experimental parameters and device performance. In this work, a microfluidic screening-embedded thin-film processing technique is developed, through which thin-films of varying ratios of small molecule semiconductor:polymer blend are simultaneously generated and screened in a time- and resource-efficient manner. Moreover, utilizing the thin-films of varying combinations of experimental parameters, machine learning models are trained to predict the transistor performance. Gaussian Process Regression (GPR) algorithms tuned by Bayesian optimization shows the best predictive accuracy amongst the trained models, which enables narrowing down of the combinations of experimental parameters and investigation of the degree of vertical phase separation under the predicted parameter space. The technique can serve as a guideline for elucidating the underlying complex parameter-property-performance correlations in solution-based thin-film processing, thereby accelerating the optimization of various thin-film devices in the future.

High-temperature ionic logic gates composed of an ionic rectifying solid–electrolyte interface

Direct data collection from extremely high temperature environments is vitally important for the progress of industrial technologies such as combustion-engines, turbines and furnaces for various purposes. However, present semiconductor-based information devices are not suitable for such high-temperature applications due to thermal excitation of electronic carriers. Herein, we demonstrate high-temperature ionic AND and OR logic gates composed of the oxide-ion-conducting yttria stabilized zirconia (YSZ) and the mixed oxide-ion and electron conducting La₂NiO_4+δ as an ultra-high temperature information device. The ionic AND and OR gates developed in this work exhibited proper and stable electrical responses at 1073 K. The ionic logic gates shown in this work are promising demonstrations for robust information devices in extreme environments.

In this work, high-temperature ionic logic gates composed of ion rectifying YSZ/La₂NiO_4+δ junctions are demonstrated.

Machine learning-assisted optimization of TBBPA-bis-(2,3-dibromopropyl ether) extraction process from ABS polymer

The increasing amount of e-waste plastics needs to be disposed of properly, and removing the brominated flame retardants contained in them can effectively reduce their negative impact on the environment. In the present work, TBBPA-bis-(2,3-dibromopropyl ether) (TBBPA-DBP), a novel brominated flame retardant, was extracted by ultrasonic-assisted solvothermal extraction process. Response Surface Methodology (RSM) achieved by machine learning (support vector regression, SVR) was employed to estimate the optimum extraction conditions (extraction time, extraction temperature, liquid to solid ratio) in methanol or ethanol solvent. The predicted optimum conditions of TBBPA-DBP were 96 min, 131 mL g^-1, 65 °C, in MeOH, and 120 min, 152 mL g^-1, 67 °C in EtOH. And the validity of predicted conditions was verified.

Machine Learning-Assisted Development of Organic Solar Cell Materials: Issues, Analyses, and Outlooks

Nonfullerene, a small molecular electron acceptor, has substantially improved the power conversion efficiency of organic photovoltaics (OPVs). However, the large structural freedom of π-conjugated polymers and molecules makes it difficult to explore with limited resources. Machine learning, which is based on rapidly growing artificial intelligence technology, is a high-throughput method to accelerate the speed of material design and process optimization; however, it suffers from limitations in terms of prediction accuracy, interpretability, data collection, and available data (particularly, experimental data). This recognition motivates the present Perspective, which focuses on utilizing the experimental data set for ML to efficiently aid OPV research. This Perspective discusses the trends in ML-OPV publications, the NFA category, and the effects of data size and explanatory variables (fingerprints or Mordred descriptors) on the prediction accuracy and explainability, which broadens the scope of ML and would be useful for the development of next-generation solar cell materials.

Comparison of the Surface Properties of Hydrothermally Synthesised Fe3O4@C Nanocomposites at Variable Reaction Times

The influence of variable reaction time (tr) on surface/textural properties (surface area, total pore volume, and pore diameter) of carbon-encapsulated magnetite (Fe3O4@C) nanocomposites fabricated by a hydrothermal process at 190 °C for 3, 4, and 5 h was studied. The properties were calculated using the Brunauer-Emmett-Teller (BET) isotherms data. The nanocomposites were characterised using Fourier transform infrared spectroscopy, X-ray diffraction analysis, thermogravimetry, and scanning and transmission electron microscopies. Analysis of variance shows tr has the largest effect on pore volume (F value = 1117.6, p value < 0.0001), followed by the surface area (F value = 54.8, p value < 0.0001) and pore diameter (F value = 10.4, p value < 0.001) with R2-adjusted values of 99.5%, 88.5% and 63.1%, respectively. Tukey and Fisher tests confirmed tr rise to have caused increased variations in mean particle sizes (11-91 nm), crystallite sizes (5-21 nm), pore diameters (9-16 nm), pore volume (0.017-0.089 cm3 g-1) and surface area (7.6-22.4 m2 g-1) of the nanocomposites with individual and simultaneous confidence limits of 97.9 and 84.4 (p-adj < 0.05). The nanocomposites' retained Fe-O vibrations at octahedral (436 cm-1) and tetrahedral (570 cm-1) cubic ferrite sites, modest thermal stability (37-60 % weight loss), and large volume-specific surface area with potential for catalytic application in advanced oxidation processes.