The architecture of the system rests on three converging technological pillars: large language models, automated reasoning, and laboratory automation. Large language models provide the flexible, generative capacity to read and synthesize scientific text, propose ideas, and communicate conclusions. Automated reasoning supplies the logical machinery to check that proposed hypotheses and experimental designs are internally consistent and grounded in established knowledge. Laboratory automation, embodied in robotic platforms, translates digital plans into physical experiments. By wiring these components together into a feedback loop, the Chalmers team created a system in which the output of one stage becomes the input of the next, and the results of each experiment feed back into the AI’s evolving understanding of the biological system under study.

Crucially, the AI did not start from a blank slate. The researchers supplied it with a substantial body of scientific knowledge about its experimental subject, including the genome of Saccharomyces cerevisiae, the metabolic pathways that govern its biochemistry, and the accumulated findings of previous studies. Yeast is an ideal proving ground for such a system because its genetics and metabolism have been mapped in extraordinary detail over decades of research, yet the sheer volume of that knowledge now exceeds what any individual scientist can hold in mind, let alone analyze systematically. It is precisely this information overload that the AI scientist was designed to exploit.

Ievgeniia Tiukova, a postdoctoral researcher at the Department of Life Sciences at Chalmers and one of the authors of the new study, explained the scale of the challenge. It is too much information for a human to analyze, she noted, but the AI scientist could identify promising biological questions, recommend experiments to test them, evaluate experimental outcomes, and iteratively refine its understanding based on new evidence. In her view, the most important shift is conceptual: rather than serving solely as decision-support tools, the AI scientist actively generates new scientific knowledge. That distinction separates this system from the many AI assistants that help researchers search literature or analyze data after the fact, because here the machine sits at the center of the discovery process itself.

Integrating the thinking power of AI with genuine experimental capability remains unusual, even within the rapidly expanding field of autonomous research systems. Many projects combine intelligent planning with robotic execution in chemistry or materials science, but validated autonomous discovery in systems biology, where the object of study is a living cell with thousands of interacting components, pushes the approach into far more complex territory. Tiukova draws an analogy to the development of self-driving cars, where AI and machine learning are used to process information, draw conclusions, and take action in the physical world. In the same way, an autonomous laboratory must perceive the state of its experiments, decide what to do next, and act, all while remaining accountable to the standards of scientific evidence.

The physical side of the platform carries a distinguished pedigree. The robot scientist Eve, shown in imagery released with the study, was originally designed specifically for drug discovery and has now been updated with large language models and automated reasoning. Eve’s lineage traces back to Adam, the first general-purpose robot scientist, developed by Ross King, who is now Professor at the Department of Computer Science and Engineering at Chalmers and the University of Gothenburg and the senior author of the new study. Adam was built to autonomously carry out scientific experiments and generate new knowledge, and Eve followed as a second-generation system aimed at accelerating the search for new medicines. The underlying concept of using robotic systems to automate and accelerate scientific discovery has since spread to other areas of research, including chemistry and a range of specialized scientific tasks, making the Chalmers work part of a broader movement that has been gathering momentum for nearly two decades.

King believes the implications for the pace of science are profound. Autonomous laboratories, he argues, will revolutionize research by systematically investigating biological systems much faster than is possible today. AI scientists will collaborate with human scientists to accelerate discoveries across biology, medicine and biotechnology, and such systems have the potential to reduce the time required to explore complex scientific questions while optimizing the use of laboratory resources. The argument is one of throughput: a robotic laboratory does not tire, does not need sleep, and can execute carefully planned experimental cycles around the clock, compressing into weeks a search through experimental conditions that might occupy a human-led team for months or years.

Both Tiukova and King are careful to frame the technology as augmentation rather than replacement. For now, autonomous AI will increasingly undertake the routine cycles of hypothesis generation and experimental testing, the repetitive engine room of experimental science, while human scientists remain essential for defining research priorities, interpreting the broader scientific significance of results, and ensuring ethical oversight. King adds that future generations of autonomous discovery systems will become increasingly capable of collaborating with human scientists, becoming valuable partners in addressing some of the most challenging questions in biology and medicine. The division of labor he envisions is complementary: machines excel at exhaustive, systematic exploration of well-defined hypothesis spaces, while humans supply judgment, context, and the sense of which questions matter.

The study, titled Agentic AI integrated with scientific knowledge: laboratory validation in systems biology, was authored by Daniel Brunnsåker, Alexander H. Gower, Prajakta Naval, Erik Y. Bjurström, Filip Kronström, Ievgeniia A. Tiukova and Ross D. King, with researchers affiliated with Chalmers University of Technology, the University of Gothenburg and the University of Cambridge in the United Kingdom. The work received funding from the Wallenberg AI, Autonomous Systems and Software Program (WASP), the UK Engineering and Physical Sciences Research Council, the Chalmers AI Research Centre (CHAIR), and the Swedish Research Council for Sustainable Development, Formas. The authors declare no competing interests.

What makes the Chalmers result resonate beyond the laboratory is the way it reframes the role of AI in science. For years, the public conversation has centered on AI as a tool for analyzing data or drafting text. A closed-loop AI scientist is something categorically different: an agent that closes the circle between idea, experiment, and evidence, and does so in the messy, unforgiving domain of living cells. If autonomous laboratories can reliably generate and validate biological discoveries with minimal human intervention, the bottleneck in research may shift from the capacity to run experiments to the wisdom of choosing which questions to ask, a bottleneck that will remain, for the foreseeable future, firmly in human hands.

Subject of Research: Autonomous AI-driven hypothesis generation and experimental validation in systems biology using brewer's yeast

Article Title: AI scientist autonomously generates and validates new biological discoveries

Article References: AI scientist autonomously generates and validates new biological discoveries. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: artificial intelligence, autonomous laboratories, robot scientist, systems biology, Saccharomyces cerevisiae, large language models, automated reasoning, laboratory automation, drug discovery, self-driving laboratories, hypothesis generation, Chalmers University of Technology

The commercial case for sodium-ion technology is strengthening rapidly. Recent industrial breakthroughs between 2025 and the present have seen announced manufacturing capacities exceed hundreds of gigawatt-hours, with particularly aggressive deployment in China. Commercial electric-vehicle packs now achieve energy densities above 170 watt-hours per kilogram, while the technology’s inherent safety, thermal stability and strong low-temperature performance make it ideal for grid-scale energy storage, where cost per kilowatt-hour matters more than compact size. High-power sodium chemistries are also attracting interest for drones, portable electronics and robotic platforms that demand rapid charging at low cost.

Yet sodium-ion batteries face stubborn material-level bottlenecks rooted in the element’s physics. The sodium ion is significantly larger than its lithium counterpart, with an ionic radius of 1.02 angstroms compared with 0.76 angstroms, leading to sluggish diffusion kinetics and substantial structural strain during repeated charging cycles. Layered oxide cathodes can undergo phase transitions and transition-metal migration that cause voltage fade, while hard carbon anodes suffer from low initial Coulombic efficiency and unstable interface formation. Alloy-type anodes based on tin, antimony and phosphorus promise theoretical capacities exceeding 1000 milliampere-hours per gram, but swell by more than 300 percent during sodiation, pulverizing particles and draining capacity. At high charging rates, non-uniform sodium deposition, dendrite growth and electrolyte decomposition compound these durability problems.

The review identifies four critical knowledge gaps that have kept traditional research methods from closing the performance gap with lithium. First, the solid electrolyte interphase, the protective film that forms on the anode, remains a molecular black box in sodium systems because sodium-based interphase components dissolve far more readily than their lithium analogues, driving continuous electrolyte consumption. Second, predicting high-voltage degradation and phase transitions with density functional theory calculations is computationally prohibitive. Third, high-quality standardized datasets for sodium materials lag dramatically behind the lithium literature. Fourth, many machine learning models function as inscrutable black boxes, offering predictions without the physical insight researchers need to understand structure-property relationships.

To overcome these obstacles, Hu argues, researchers are deploying an increasingly sophisticated AI toolbox. Graph neural networks represent crystal structures directly as atomic graphs, learning relationships between connectivity and properties without manual feature engineering, and have proven especially powerful for predicting formation energies, phase stability and electrode voltages in layered oxides. Deep neural networks serve as fast surrogate models for electrochemical performance metrics, while generative models such as generative adversarial networks, variational autoencoders and diffusion models can propose entirely new crystal structures that exist in no database. Inverse design flips the traditional workflow on its head: instead of making a material and measuring its properties, researchers specify desired performance targets and let the algorithm identify candidates likely to meet them.

Each cathode family demands a different AI strategy. For layered transition-metal oxides, machine learning models trained on density functional theory data distinguish P2 from O3 structures using descriptors like sodium concentration and mixing entropy, and AI-guided screening has identified phase-stable iron-nickel-manganese-titanium high-entropy oxides with improved cycling retention and air stability. Polyanionic cathodes of the sodium superionic conductor type benefit from active-learning frameworks that slash computational costs while exploring thousands of compositions, with generative models coupled to quantum simulations finding candidates with volume changes below 4 percent. Prussian blue analogues, whose modular open frameworks are plagued by vacancy defects and structural water, are ideally suited to variational autoencoders and Gaussian process regression that optimize composition and suppress degradation mechanisms.

On the anode side, machine learning is equally transformative. Ensemble methods such as XGBoost and random forests trained on biomass-derived datasets now predict reversible capacity and initial Coulombic efficiency from synthesis parameters, revealing how carbonization temperature tunes porosity and graphitic ordering in hard carbon. For alloy anodes, crystal-structure search algorithms combined with machine-learning-accelerated annealing identify thermodynamically stable sodium-alloy phases, while high-entropy alloy concepts redistribute sodiation-induced stress. Anode-free designs, in which sodium plates directly onto a bare current collector for maximum energy density, rely on AI to engineer artificial interphases and electrolyte additives that suppress dendrites, with engineered sodium-metal interfaces reported to reach Coulombic efficiencies approaching 99.96 percent under favorable conditions.

Electrolyte development may be where AI delivers its greatest leverage, because the design space of solvent, salt and additive combinations is astronomically large. Descriptor-based screening using HOMO-LUMO energies, dielectric constants and solvation energies allows algorithms to shortlist solvents that optimize sodium-ion coordination, while deep learning correlates additive chemistry with the formation of inorganic-rich protective layers composed of sodium fluoride, sodium oxide and sodium carbonate. Bayesian optimization combined with quantum calculations rapidly identifies additives that outperform conventional fluoroethylene carbonate, and physics-informed machine learning embeds thermodynamic constraints into models to accelerate prediction of ionic conductivity while preserving physical consistency. The same frameworks are being used to design localized high-concentration and fluorine-free electrolytes that suppress dendrite growth through favorable solvation structures.

The review’s most ambitious vision is the fully autonomous, closed-loop discovery platform, in which AI prediction, robotic synthesis and automated electrochemical testing feed results back into ever-improving models without human intervention. Active learning continuously selects the most informative experiments, Bayesian optimization explores sparse chemical spaces efficiently, and emerging agentic materials-science frameworks can plan, execute and refine entire development workflows. But Hu is candid about the remaining hurdles: sodium-specific datasets remain too small and scattered, dynamic interfacial phenomena span too many length and time scales for current models, and explainable AI methods will be essential to convert black-box predictions into actionable scientific insight. If those challenges can be met, self-driving laboratories could compress the journey from computational concept to validated battery material from years to months, accelerating sodium-ion batteries toward their role as the sustainable backbone of grid storage, electric mobility and beyond.

Subject of Research: AI-driven discovery of cathode, anode and electrolyte materials for sodium-ion batteries

Article Title: Review of AI-driven discovery of cathode anode and electrolyte materials for advancing sodium-ion batteries

Article References: Hu, Z. (2026). Review of AI-driven discovery of cathode anode and electrolyte materials for advancing sodium-ion batteries. Discover Industrial Chemistry and Materials, 1(1), Article 21. https://doi.org/10.1007/s44508-026-00022-x

Image Credits: AI Generated

DOI: 10.1007/s44508-026-00022-x

Keywords: sodium-ion batteries, artificial intelligence, machine learning, graph neural networks, generative AI, cathode materials, anode materials, electrolytes, solid electrolyte interphase, inverse design, high-throughput screening, autonomous laboratories

The central premise is deceptively simple. Single-cell analysis matters because populations of supposedly identical cells are anything but identical. Differences in gene expression, protein secretion, morphology and behaviour drive disease progression, immune responses and therapeutic outcomes in ways that bulk measurements average away. Microwell platforms address this heterogeneity by spatially confining individual cells within miniature compartments while preserving access for imaging, perturbation and downstream molecular measurements. Unlike droplet-based systems, which encapsulate cells in aqueous pockets within oil, microwells keep cells physically addressable and visually observable, allowing researchers to link what a cell looks like, what it does and what it expresses within a single workflow.

The review pays close attention to how engineering choices determine scientific outcomes. Cell-loading strategies—whether gravitational settling, centrifugation-assisted deposition, hydrodynamic focusing or immunomagnetic capture—shape how efficiently wells are occupied and whether the same cell can be tracked across multiple measurement modalities. Well geometry is equally consequential: truncated cone-shaped arrays improve trapping efficiency, L-shaped and paired wells allow controlled encounters between two different cells, and nested nanowell-in-microwell architectures enable motility studies inside confined arenas. Platform architecture and material selection matter too. Polydimethylsiloxane, the workhorse polymer of microfluidics, can absorb small-molecule drugs and distort drug-response assays, while polymer films and hydrogel coatings extend culture lifetimes and support organoid growth. Phototoxicity during live fluorescence imaging and the need to support cell spreading and proliferation impose further constraints on design. The authors argue that these parameters collectively determine the information available for computational analysis—experimental design and analytical power are inseparable.

From this engineering foundation, the review surveys the major application domains where microwells have proven their worth. In cellular behaviour and cell–cell interactions, microwell arrays have enabled researchers to follow lymphocyte cytotoxicity dynamics, serial killing by immune cells, and the activation profiles of paired T cells in real time. Time-resolved cell-pairing arrays have revealed multiple activation states among individual T cells, while single-cell arrays of hematological cancer cells have allowed quantitative assessment of how immune killers engage, destroy and move on to successive targets. Such assays turn immunology from a population statistic into a choreography of individual cellular encounters.

Secretome analysis represents another signature strength. Microengraving methods, first demonstrated for rapidly selecting cells producing antigen-specific antibodies, imprint the proteins secreted by each trapped cell onto an antibody-coated surface, allowing frequencies and rates of cytokine secretion to be measured across thousands of cells simultaneously. Hierarchical loading microwell chips have evaluated single-cell cytokine secretion and cell–cell interactions, while nanoplasmonic microwell arrays now monitor secretion in real time without labels. The field has even extended to single extracellular vesicles and particles, with oil-sealed hydrogel microwell arrays capturing the nanoscale output of individual cells—a frontier the review identifies as rapidly expanding.

Genomic and transcriptomic profiling is arguably where microwells achieved their greatest visibility. Massively parallel polymerase cloning in nanoliter wells enabled genome sequencing of single cells, and Microwell-Seq platforms mapped the mouse cell atlas at unprecedented scale. Seq-Well brought portable, low-cost RNA sequencing to the field, while later generations such as Microwell-seq 2.0 and Microwell-seq3 extended the approach to multiplexed chemical perturbation screens and joint profiling of chromatin accessibility and RNA expression. Addressable microwell arrays support dual-indexed workflows and permit imaging to be integrated with sequencing, so that morphology and transcript identity can be recovered from the same cell. Comparative studies summarized in the review suggest microwell-based methods perform robustly even with cryopreserved clinical samples, an important consideration as single-cell genomics moves toward routine clinical use.

Drug screening and precision medicine complete the application landscape. Microwell chips have measured drug sensitivity in individual prostate cancer cells, supported miniaturized multiplexed high-content screening of drug and immune responses in multichambered formats, and enabled high-throughput generation of patient-derived cancer stem cells for precision medicine. Deep learning frameworks for in silico screening of anticancer drugs at the single-cell level hint at a future where candidate compounds are triaged computationally before ever touching a laboratory plate. Throughout these domains, the authors stress a common advantage: microwells preserve cell identity, allowing spatial, temporal, functional and molecular data to be linked for the same individual cell rather than inferred across disconnected measurements.

That linkage, however, comes at a cost. The datasets generated are enormous, multidimensional and difficult to analyse with conventional approaches, and this is where artificial intelligence enters the story. The review provides a careful, even sceptical accounting, distinguishing AI methods that have been directly demonstrated in microwell-based studies from those developed in the broader single-cell field that remain prospective. Demonstrated applications include automated image analysis, cell tracking, phenotype classification and behavioural analysis: deep learning models have profiled single-cell migration and proliferation on addressable dual-nested microwell arrays, automated the analysis of natural killer cell cytotoxicity in single cancer cell arrays, and enabled stain-free cell viability screening from regularized single-cell imaging. Deep learning has also revealed how cell–cell interactions shape single-cell behaviour in high-throughput coculture systems, and studies of single-cell morphodynamics now predict cell fate decisions during epithelial differentiation—evidence that image-derived features can carry predictive molecular information.

Looking further ahead, the authors outline an ambitious computational roadmap built on multimodal AI, foundation models, large language model agents and autonomous laboratory systems. Foundation models pretrained on cross-species single-cell landscapes have already identified conserved regulatory programs underlying cell types, suggesting transferable biological knowledge that could be fine-tuned for microwell data. Large language model agents have recently been shown to design droplet microfluidic experiments autonomously and to mine the microwell microfluidics literature, and the review extends this vision to microwell platforms themselves: agentic AI systems that could plan experiments, adjust loading and imaging parameters, interpret results in closed loops and iteratively refine hypotheses without human intervention. Autonomous microfluidic labs, combining machine learning with robotic liquid handling and microfluidic control, are identified as the natural endpoint of this trajectory—moving microwell-based single-cell research from measurement toward prediction and, ultimately, self-driving discovery.

The significance of this synthesis lies in its insistence that hardware and software must be co-designed. If the geometry of a well determines which cells are captured, and the imaging regime determines what is visible, then the AI models trained on that data inherit the assumptions and limitations of the platform. By explicitly linking microwell engineering to the information available for computational analysis, the review offers the field a blueprint: researchers who want predictive, AI-ready single-cell biology must build platforms that capture the right cells, preserve their identity, and generate the multimodal, well-annotated data that modern machine learning demands. As microwell arrays and artificial intelligence converge, the humble microscopic well may become the standardized test tube of the autonomous biology era.

Subject of Research: Microwell-based single-cell analysis platforms and their emerging integration with artificial intelligence

Article Title: Microwell platform for single-cell applications and future integration with artificial intelligence (AI)

Article References: Cheung, H. L., Nguyen, D.-N., Ngo, M. T. T., & Dinh, N.-D. (2026). Microwell platform for single-cell applications and future integration with artificial intelligence (AI). Biomedical Microdevices, 28(3), Article 66. https://doi.org/10.1007/s10544-026-00852-8

Image Credits: AI Generated

DOI: 10.1007/s10544-026-00852-8

Keywords: microwell platforms, single-cell analysis, artificial intelligence, microfluidics, cell heterogeneity, secretome analysis, single-cell RNA sequencing, drug screening, precision medicine, deep learning, large language model agents, autonomous laboratories

Chemical vapor deposition, or CVD, is conceptually simple. Precursor gases or vapors are introduced into a heated chamber where they react on a substrate surface, nucleating and growing crystalline films layer by layer. In practice, however, the process is a nightmare of coupled variables. Temperature gradients across the furnace, gas flow rates, precursor partial pressures, substrate surface chemistry, chamber geometry, cooling ramps, and even trace contaminants all influence whether the growth yields a pristine monolayer, a useless patchwork of multilayer islands, or nothing at all. Small changes in one parameter can ripple through the entire system in nonlinear ways, which is why transferring a growth recipe from one laboratory’s furnace to another’s often fails spectacularly.

Traditionally, researchers have navigated this parameter space the way navigators charted unknown seas: iteratively, slowly, and guided by intuition. A doctoral student might spend months running growth after growth, characterizing each sample under a microscope or Raman spectrometer, and adjusting conditions based on experience. The authors of the new analysis contend that this manual, serial approach is fundamentally mismatched to the complexity of 2D material synthesis. The parameter space is too large, the interactions between variables too intricate, and the experimental throughput too low for human trial-and-error to deliver the reproducibility that industrial manufacturing demands.

Enter machine learning. By training models on existing growth data—including published literature and internal laboratory records—researchers can build surrogate models that predict outcomes such as domain size, layer number, crystal orientation, and defect density from synthesis parameters without running an experiment. Bayesian optimization algorithms then use these surrogates to select the next experiment intelligently, balancing exploration of uncertain regions of parameter space against exploitation of promising conditions. Instead of one hundred blind experiments, a well-designed closed-loop campaign may converge on an optimal recipe in a few dozen. The perspective highlights that such approaches have already accelerated optimization of graphene growth on copper foils, uniform monolayer MoS2 deposition, and the control of grain boundaries and twist angles between stacked layers.

But predictive models are only half of the equation. The other half is quality control—a persistent bottleneck in moving 2D materials from laboratory curiosities to certified manufacturing feedstock. Conventional characterization, relying on Raman spectroscopy, photoluminescence mapping, atomic force microscopy, and electron microscopy, is slow, sample-intensive, and often destructive or off-line. The authors argue that AI can transform this step as well. Deep learning models trained on optical microscopy images can classify layer number, identify grain boundaries, and flag defects across wafer-scale samples in minutes. Spectral analysis algorithms can extract quantitative information from Raman and photoluminescence data with greater consistency than human operators. Integrating these inspection tools directly into the synthesis workflow enables real-time feedback: if the model detects that the growing film is drifting out of specification, the process parameters can be adjusted mid-run rather than after the fact.

This convergence of synthesis, sensing, and machine intelligence points toward the article’s central vision: the self-driving laboratory. In an autonomous manufacturing platform, robotic systems handle substrate loading, precursor delivery, and sample transfer; in-line sensors stream data on film growth and quality; and a decision-making algorithm orchestrates the entire loop, proposing experiments, executing them, evaluating results, and refining its strategy. Humans set the goals—the target material, desired thickness, crystal quality, and throughput constraints—while the automated system navigates the path toward them. Early demonstrations of such platforms in materials chemistry, including autonomous flow chemistry rigs and robot-driven thin-film laboratories, suggest the concept is not science fiction but an emerging engineering reality.

The implications for the semiconductor industry are considerable. As device architectures push into the angstrom scale, silicon itself is running out of room to shrink, and 2D materials are among the leading candidates for next-generation transistors, memristors, photodetectors, and quantum devices. The International Roadmap for Devices and Systems has identified monolayer transition metal dichalcogenides as potential channel materials for future logic technology. But fabs operate on tolerances measured in fractions of a percent, and no conventional CVD process today delivers that level of wafer-to-wafer uniformity at scale. AI-driven process control, with closed-loop adjustment based on in-line metrology, offers a credible route to closing that gap—provided the underlying data infrastructure is built to industrial standards.

The perspective is candid about the obstacles. Machine learning in materials synthesis is data-hungry, yet most growth data in the literature is sparse, heterogeneous, and poorly documented. Negative results—failed growths—are rarely published, biasing datasets toward successes and limiting the ability of models to learn failure boundaries. Different laboratories report parameters inconsistently, and furnace-to-furnace variation means that even well-documented recipes are not directly transferable. The authors call for standardized data reporting formats, shared repositories of synthesis records, and the adoption of emerging frameworks such as FAIR data principles—findable, accessible, interoperable, reusable—so that community-scale datasets can be assembled and models can generalize beyond a single instrument.

Safety, reproducibility, and interpretability also demand attention. An autonomous system running hazardous precursors such as metal-organic compounds, hydrogen sulfide, or hydrogen at high temperatures must incorporate robust safeguards, fault detection, and emergency protocols. And for AI recommendations to be trusted by process engineers, the models must not be black boxes; explainable machine learning approaches that reveal which parameters drive a given prediction will be essential for regulatory compliance and for building human confidence in automated decision-making. The authors envision hybrid workflows in which human experts and autonomous systems collaborate, with the AI handling routine optimization while scientists tackle genuinely novel materials discovery.

If those challenges are met, the payoff extends well beyond any single material. A mature, AI-enabled CVD infrastructure would function as a general-purpose platform: the same closed-loop hardware and software stack could be retargeted from graphene to MoS2 to emerging quantum materials simply by swapping precursors and retraining the models. That flexibility, combined with dramatically reduced development timelines, could finally push 2D materials out of the cleanroom demonstration phase and into mass production—powering faster transistors, ultrasensitive sensors, flexible displays, and photonic circuits that today exist only on whiteboards. The message of the new analysis is clear: the bottleneck in 2D materials manufacturing is no longer a lack of ideas but a lack of control, and artificial intelligence is poised to supply it.

Subject of Research: AI-enabled chemical vapor deposition synthesis, quality control, and autonomous manufacturing of two-dimensional materials

Article Title: AI-enabled CVD synthesis, quality control, and autonomous manufacturing of 2D materials

Article References: Mannan, M. I., Mistry, P., Kansal, V., Estrada, V., & Leem, J. (2026). AI-enabled CVD synthesis, quality control, and autonomous manufacturing of 2D materials. npj Advanced Manufacturing. https://doi.org/10.1038/s44334-026-00109-5

Image Credits: AI Generated

DOI: 10.1038/s44334-026-00109-5

Keywords: two-dimensional materials, chemical vapor deposition, machine learning, autonomous laboratories, graphene, MoS2, transition metal dichalcogenides, quality control, semiconductor manufacturing, Bayesian optimization, self-driving laboratory, advanced manufacturing