To appreciate why this matters, it helps to understand what formation actually is. When a lithium-ion cell is assembled, it is not yet a functional battery. It must first be charged and discharged in carefully controlled cycles, a process called formation, during which the solid electrolyte interphase, a critical protective layer on the anode, is established. This stage is among the most expensive and time-consuming steps in battery manufacturing, and manufacturers around the world use different formation protocols, varying the charging currents, voltages, and cycle patterns, in pursuit of faster production or better performance. The trouble is that these protocol differences leave distinct fingerprints on the cell’s long-term degradation behavior. A prediction model trained on batteries formed one way may fail badly when applied to batteries formed another way, which is precisely the problem the research team set out to solve.

Traditional approaches to remaining useful life prediction, the researchers note, often suffer from poor accuracy and weak robustness when batteries have undergone different formation protocols. The core difficulty is that formation-induced differences create systematic biases in the early-cycle data that prediction models rely on. Features extracted from charge-discharge curves may look deceptively similar across cells while hiding fundamentally different aging trajectories. The new method confronts this head-on with a two-pronged strategy: an unscented transformation tailored for variable formation protocols, and a hybrid framework combining an Unscented Particle Filter with a Gated Recurrent Unit neural network, complete with a protocol-sensitivity compensation module designed to mitigate the bias that protocol differences inject into predictions.

The first component of the framework is an exponential empirical degradation model, fitted using an Unscented Particle Filter, that captures the underlying aging mechanism of the battery. Empirical degradation models describe how capacity fades over cycles using mathematical functions, and the particle filter is a Bayesian inference technique that continuously updates the model’s parameters as new measurements arrive. The unscented variant is particularly well suited to this task because it propagates uncertainty through nonlinear transformations without requiring the model to be linearized, a step that can introduce serious errors in battery systems where degradation is anything but linear. By anchoring the prediction in a physically motivated degradation model, the framework avoids the worst failure mode of purely data-driven approaches: extrapolating confidently into regions where the training data offers no guidance.

The second component is where the deep learning enters. The researchers constructed a Gated Recurrent Unit neural network, a streamlined relative of the long short-term memory architecture that has become a workhorse of time-series prediction. The network takes key features from constant-current and constant-voltage charge-discharge cycles as its inputs, treats the battery’s state of health as a hidden state that evolves over time, and outputs the remaining useful life under varying formation protocols. Gated recurrent units are prized for their ability to capture temporal dependencies in sequential data while remaining computationally lighter than full LSTM networks, making them attractive for applications where predictions must be updated continuously as new cycling data streams in.

What makes the architecture genuinely novel is the way these two components collaborate. Rather than letting the neural network operate alone, the framework integrates empirical modeling with data-driven prediction, allowing the particle-filter-based degradation model to refine the network’s RUL estimates while the protocol-sensitivity compensation module adapts dynamically to variations in formation conditions. In effect, the empirical model provides a physically grounded skeleton, the neural network fleshes out the complex, protocol-dependent details, and the compensation module corrects for the systematic distortions that different formation histories introduce. This model-data fusion philosophy reflects a broader trend in battery prognostics, where researchers increasingly recognize that neither pure physics nor pure machine learning is sufficient on its own.

Validation was carried out on public data from the University of Michigan Battery Lab, specifically the fast formation study dataset published through the university’s Deep Blue Data repository. This dataset is a valuable testing ground precisely because it contains batteries cycled under different formation protocols, allowing the researchers to systematically quantify and model the impact of protocol differences on long-term degradation and remaining useful life prediction. The results were striking: the proposed method achieved an R-squared value of 99.41 percent, meaning it explained nearly all of the variance in the observed degradation trajectories, and a root mean square error of 2.23 percent, indicating tight agreement between predicted and actual values. The team reports that the method significantly outperformed comparison methods in both state of health and remaining useful life prediction accuracy, demonstrating superior cross-protocol generalization and robustness compared with existing hybrid approaches.

The practical implications of this work extend well beyond the laboratory. Accurate remaining useful life prediction is essential for determining appropriate battery replacement times, a decision with enormous consequences for safety, cost, and sustainability. In electric vehicles, overestimating a battery’s life risks stranding drivers with a depleted pack, while underestimating it wastes money by replacing cells that still have serviceable capacity. In aerospace and space applications, where some of the co-authors’ institutions operate, the stakes are even higher, since battery failure can compromise entire missions. A prediction framework that remains accurate regardless of how the cells were formed could allow manufacturers to mix batteries from different production lines or formation batches into the same fleet management system without sacrificing prognostic reliability, something current methods struggle to guarantee.

There is also a manufacturing angle that could prove commercially significant. Formation is one of the most energy-intensive and capital-heavy stages of cell production, and the industry has been actively exploring faster formation protocols to cut costs and increase throughput. One reason manufacturers hesitate to adopt aggressive fast-formation schemes is uncertainty about how those choices will affect long-term life and whether existing quality-control and prognostic tools will still work. A prediction method that explicitly compensates for protocol sensitivity could lower that barrier, giving manufacturers the confidence to experiment with formation strategies that shorten production times without losing the ability to forecast field performance. The research team, which includes authors affiliated with CATL, the world’s largest battery manufacturer, is well positioned to understand these industrial pressures.

The study, led by Gaige Chen and Haitao Meng, who contributed equally, with Yahong Yang, Xiaoshen Zhang, Wei Deng, and Jinsong Liu, was supported by the National Natural Science Foundation of China and builds on the group’s earlier work on transfer learning for state of health prediction under cross-formation protocols. As batteries proliferate through every corner of the energy economy, tools like this hybrid UPF-GRU framework point toward a future in which the lifespan of every cell can be forecast accurately from its earliest cycles, regardless of the manufacturing choices that shaped it. That kind of predictive certainty could reshape how batteries are designed, produced, warranted, and retired, turning one of the industry’s most persistent blind spots into a solved problem.

Subject of Research: Machine learning-based remaining useful life prediction of lithium-ion batteries across different formation protocols

Article Title: Remaining useful life prediction of lithium-ion batteries across formation protocols

Article References: Chen, G., Meng, H., Yang, Y., Zhang, X., Deng, W., & Liu, J. (2026). Remaining useful life prediction of lithium-ion batteries across formation protocols. Ionics. https://doi.org/10.1007/s11581-026-07472-5

Image Credits: AI Generated

DOI: 10.1007/s11581-026-07472-5

Keywords: lithium-ion batteries, remaining useful life, formation protocols, unscented particle filter, gated recurrent unit, state of health, battery degradation, machine learning, battery manufacturing, prognostics, CATL, University of Michigan Battery Lab

The appeal of the anode-free design is easy to grasp. In a conventional sodium metal battery, a foil of metallic sodium is preloaded into the cell, adding weight, volume, cost, and hazard. In the anode-free version, the anode is simply a bare current collector, usually copper or aluminum, and the sodium that plates onto it during charging is extracted in situ from the cathode. Strip the sodium back during discharge and the cycle repeats. In principle this maximizes both gravimetric and volumetric energy density, because the cell carries no excess anode mass whatsoever. Current pouch-cell prototypes deliver roughly 200 to 210 watt-hours per kilogram at the cell level, still 20 to 30 percent below commercial lithium-ion cells, but the headroom is real if the chemistry can be tamed.

The problem is that sodium is a difficult metal to plate reversibly. Compared with lithium, sodium ions bind more weakly to substrates, carry a larger ionic radius, and behave differently in solution, all of which conspire to produce higher nucleation overpotentials and three-dimensional, mossy deposits rather than compact films. Where lithium nucleates with an overpotential of roughly 10 millivolts, sodium often demands more than 50. The solid electrolyte interphase that forms on sodium is typically soft and organic-rich, fracturing repeatedly as the metal expands and contracts, exposing fresh sodium to the electrolyte each time. Each fracture consumes both sodium and electrolyte irreversibly. In a cell with a sodium reservoir these losses are buffered; in an anode-free cell they are fatal.

The review’s authors formalize this with an inventory balance that deserves to become standard practice in the field. They divide the sodium in a cycled cell into three categories: reversible sodium that remains electrically connected and can be stripped again; dead sodium, which is metallic but electronically isolated or entombed in insulating interphase; and chemically bound sodium, consumed into interphase phases and electrolyte decomposition products. Coulombic efficiency, the metric nearly every paper reports, captures only the sum of the last two. Because dead sodium is in principle partially recoverable while chemically bound sodium is not, the two loss channels respond to entirely different design levers. The authors propose that reporting the three terms separately, using techniques such as sodium-23 nuclear magnetic resonance, titration gas chromatography, and X-ray photoelectron spectroscopy, would finally let researchers identify which failure pathway dominates in a given system rather than inferring it.

The arithmetic of efficiency is brutal. Working from the retention relation that multiplies initial Coulombic efficiency by the average efficiency raised to the power of cycle number, the authors derive minimum efficiency targets for practical operation. To retain 80 percent of capacity over 500 cycles with a first-cycle efficiency of 90 percent, a cell must sustain an average Coulombic efficiency of at least 99.98 percent. Even at a perfect first cycle, 500 cycles demand 99.96 percent. A cell averaging 99.7 percent, which sounds excellent, reaches the same retention threshold after only about 74 cycles. Measured against these criteria, fewer than 13 percent of reported studies achieve efficiencies above 99.7 percent under lean-electrolyte conditions, and the gap between the state of the art and commercial viability amounts to nearly a sevenfold reduction in sodium loss per cycle.

What makes this gap worse is that most laboratory numbers are collected under conditions that flatter the chemistry. More than 80 percent of reported systems operate below 2 milliamp-hours per square centimeter of areal capacity, while the industrial benchmark is at least 3. Coin cells are typically flooded with 30 to 50 microliters of electrolyte, corresponding to more than 10 grams per ampere-hour, whereas industrial cells require 3 or less. When chemistries that shine in coin cells are translated into multilayer pouch cells, a consistent penalty emerges: efficiencies above 99.9 percent often fall to 99.0 to 99.5 percent, driven by edge plating at cell perimeters, uneven stack pressure, tab corrosion, and electrolyte starvation. Of twenty-four representative systems surveyed, only one simultaneously satisfies both the lean-electrolyte and high-areal-capacity thresholds. The review’s blunt conclusion is that anode-free sodium batteries are integration-limited, not materials-limited.

That said, the toolbox of interventions is expanding rapidly, and the review organizes it by function rather than by material class. Electrolyte engineering has moved from empirical salt-solvent mixing toward mechanism-guided control: weakly solvating dual-salt systems, high-entropy sole-solvent formulations, zeolite molecular-sieve films that impose geometric confinement on solvation clusters, and nano-silica suspensions that immobilize reactive anions all steer interphase chemistry toward inorganic-rich, mechanically robust layers. On the current collector side, fluorinated carbon scaffolds act as built-in fluorine donors that seed sodium fluoride-rich interphases, with one fluorinated carbon nanotube collector sustaining over 99.9 percent efficiency across 800 cycles at 3 milliamp-hours per square centimeter. Alloy coatings of indium, tin-copper intermetallics, and high-entropy alloy nanolayers lower nucleation barriers, while three-dimensional porous hosts redistribute ion flux and buffer the mechanical stress of plating. Sodium sits at roughly 80 percent of its melting point at room temperature, so it creeps plastically under modest pressure, which means controlled stack compression between 0.1 and 0.4 megapascals can close voids and restore contact to otherwise disconnected deposits, a lever lithium does not offer so readily.

The cathode side, often neglected in anode-free discussions, turns out to be a coupled problem rather than an isolated one. At operating voltages above roughly 4.0 volts, layered oxide cathodes accelerate electrolyte oxidation and dissolve transition metals, and X-ray photoelectron spectroscopy has directly detected nickel and iron species on the cycled sodium anode, proof that cathode degradation migrates across the cell and poisons the interphase where sodium plates. Electrolytes that form thin, inorganic-rich cathode-electrolyte interphases suppress this crossover, and several pouch cells have now operated up to 4.3 volts across temperatures from minus 40 to 60 degrees Celsius with cell-level energies exceeding 200 watt-hours per kilogram. The design lesson is that the two interphases must be chemically aligned: stabilizing one while ignoring the other simply relocates the failure.

Manufacturing economics add a final layer of constraint. Sodium itself is cheap, but the supporting ecosystem is not. Fluorinated imide salts cost five to six times more than the incumbent sodium hexafluorophosphate, and projected cell manufacturing costs of 80 to 120 dollars per kilowatt-hour erode the cost advantage that motivates sodium chemistry in the first place. Sodium’s extreme moisture sensitivity demands dry rooms with dew points near minus 60 degrees Celsius, capital costs 50 to 100 percent higher than conventional lithium-ion lines, yet fewer than 5 percent of studies even report humidity tolerance. A technology-readiness assessment across 36 strategies finds that only three reach prototype demonstration level, and those tend to be the ones using commercially compatible materials and processes rather than exotic syntheses.

The review closes with a roadmap that reads less like a wish list and more like an audit. Short-term priorities through 2026 center on standardized pouch-cell protocols with controlled pressure, lean electrolyte dosing, and humidity-controlled filling. The mid-term focuses on electrolyte-interphase co-design validated at the ampere-hour scale, and the long term points toward solid-state anode-free cells for grid and mobility applications. The authors also propose a coordinated Battery300 initiative, modeled on the U.S. Department of Energy’s Battery500 program, to unify evaluation standards around targets of at least 300 watt-hours per kilogram, efficiency above 99.95 percent in pouch cells, and open benchmarking databases. Their underlying message is clear and slightly uncomfortable for the field: the path to practical anode-free sodium batteries runs not through another record-breaking coin cell, but through measurement conventions, inventory accounting, and the patient, unglamorous work of making every component survive contact with every other one.

Subject of Research: Anode-free sodium metal battery design, failure mechanisms, and system-level integration for practical energy storage

Article Title: Anode‐Free Sodium Metal Batteries: From Materials Design to System‐Level Integration

Article References: Hussain, H., Ali, S., Ali, M., Aman, S., Ali, M., Naseer, U., Farooq, S., Mahmood, A., He, S., Jiang, Y., & Yousaf, M. (2026). Anode‐Free Sodium Metal Batteries: From Materials Design to System‐Level Integration. Advanced Science, Article e77874. https://doi.org/10.1002/advs.77874

Image Credits: AI Generated

DOI: 10.1002/advs.77874

Keywords: sodium batteries, anode-free, solid electrolyte interphase, Coulombic efficiency, current collector, electrolyte engineering, dead sodium, pouch cells, energy density, battery manufacturing, solid-state electrolytes, machine learning

The carbon footprint concept emerged in the early 2000s as a practical way to link products, processes, and organizations to their greenhouse gas emissions, expressed in carbon dioxide equivalents. It was quickly institutionalized through two landmark frameworks. The Greenhouse Gas Protocol, published in 2004 by the World Resources Institute and the World Business Council for Sustainable Development, introduced the now-ubiquitous classification of emissions into Scope 1, Scope 2, and Scope 3 – direct emissions, purchased energy, and value-chain emissions respectively. Meanwhile, the ISO 14040 and 14044 standards formalized Life Cycle Assessment, or LCA, the rigorous methodology that tallies environmental burdens from raw material extraction through manufacturing, use, and disposal. Together these frameworks gave companies a common language for identifying emission hotspots, benchmarking performance, and setting reduction targets, and they now feed into national climate commitments, ESG reporting, and consumer carbon labels.

But the review identifies a persistent structural problem: the landscape is methodologically fragmented. Analysts must choose among process-based LCA, which offers detailed product-level accounting but suffers from truncation errors that omit indirect upstream emissions; input-output analysis, which captures economy-wide supply chains but relies on coarse sector averages; and hybrid models that combine both but demand intensive data and modeling effort. Newer variants push further – dynamic LCA incorporates time-dependent emission factors that track decarbonizing electricity grids, while spatially resolved methods capture regional differences in energy mixes. Each approach embodies a trade-off between specificity, completeness, data availability, and computational effort, and no single method dominates across all manufacturing contexts.

The consequences of this fragmentation are quantifiable. Differences in system boundaries, allocation rules for co-products, functional units, and emission-factor databases such as Ecoinvent, GaBi, and GREET can produce substantially different results for similar products. Sensitivity studies compiled in the review illustrate the scale of the problem: variability in the energy mix has driven differences of up to 50 percent in food-product footprints, while choices of allocation method alone shifted industrial results by as much as 40 percent. Uncertainty treatment compounds the issue. Simple one-at-a-time sensitivity analysis, widely used because it is cheap and intuitive, cannot capture interactions between variables. Monte Carlo simulation propagates uncertainty across thousands of scenarios but depends heavily on the quality of input distributions. Global sensitivity methods such as Sobol indices and the Morris method reveal which parameters and interactions actually drive variability, yet their computational demands limit routine industrial use. Critically, the review notes, uncertainty is usually assessed only after a footprint model has been built, rather than being woven into data selection and boundary definition from the start.

Scope 3 emissions present perhaps the thorniest challenge. For many organizations, value-chain emissions dominate the total footprint, yet they are the hardest to quantify because supply chains are complex, geographically dispersed, and data-poor. The review highlights a particularly troubling blind spot in battery manufacturing: conventional factory-level assessments frequently account for on-site cell assembly and pack integration while excluding the embodied carbon of cathode and anode active materials sourced from other countries – precisely the components that dominate a battery’s total emissions. The result is a systematic underestimation of the true carbon burden, obscured further by static, historical data that fail to reflect evolving energy systems and technological change.

The authors’ proposed remedy comes from an unexpected direction: the world of digital engineering. Model-Based Definition, or MBD, transforms the 3D CAD model into the authoritative digital representation of a product, embedding geometry, tolerances, materials, and manufacturing process metadata directly within the model as machine-readable information. Unlike traditional 2D drawings, an MBD model can be automatically interpreted by downstream applications – and, crucially, by LCA tools. When a designer changes a material, a tolerance, or a manufacturing feature, the associated metadata can instantly update the corresponding life cycle inventory entries and recalculate the footprint. In this vision, the product model becomes a digital carrier of sustainability data, bridging the long-standing disconnect between engineering design and environmental analysis.

The review situates MBD within a broader model-based ecosystem. Model-Based Systems Engineering formalizes system requirements and architecture, allowing environmental constraints such as emission targets and recyclability to be treated as design requirements subject to early trade-off analysis. The Model-Based Enterprise extends product data across the organization through standards like STEP AP242 and the Quality Information Framework, enabling a continuous digital thread from design intent to factory execution. Commercial tools such as SolidWorks Sustainability and Autodesk Insight already perform on-the-fly carbon estimates from model metadata, and academic workflows have predicted emissions for welded assemblies and machined parts with minimal manual intervention. In construction, Building Information Modeling systems similarly derive embodied-carbon inventories automatically, demonstrating the transferability of the approach.

To illustrate the potential, the authors sketch a hierarchical, bottom-up carbon-accounting framework for battery production. Each component – cathode, anode, electrolyte, separator, casing – is defined through MBD with embedded material and process metadata linked to life cycle inventory databases. Emissions roll up through assembly stages, then to the factory, the industrial zone, the city, and the national sector, preserving traceability at every level. In a worked example, the MBD-driven approach yields a cradle-to-grave footprint of roughly 88 kilograms of CO2-equivalent per kilowatt-hour of cell capacity for an NMC battery line – about 80 kilograms from materials plus 18 kilograms from process electricity – corresponding to approximately 17,600 tonnes of CO2-equivalent per year for a facility producing 200 megawatt-hours of cells. Conventional top-down estimates for comparable facilities, built from aggregated electricity, fuel, and waste statistics, typically report substantially lower values because they truncate upstream material production. The authors stress the MBD figure is not an upper bound but a physically grounded reconstruction of the true carbon burden.

The authors are careful to position the framework as a future research direction rather than a mature industrial system. MBD cannot resolve methodological choices such as system boundaries, allocation rules, or emission-factor selection – those remain matters of expert judgment. Significant obstacles remain, including semantic misalignment between CAD and LCA ontologies, boundary inconsistencies between design models and life cycle scopes, and the need for versioned, traceable emission-factor metadata. Fully automated inventory generation, AI-driven emission prediction, and continuously updated digital carbon twins all require further validation. The review calls for minimum machine-readable data requirements, standardized mapping mechanisms between product attributes and inventory flows, improved environmental-data traceability, and industrial case studies comparing MBD-enabled and conventional workflows.

If the vision matures, the implications extend well beyond compliance dashboards. A manufacturing world in which every design decision – a cathode chemistry swap, a more efficient drying oven, an optimized formation protocol – propagates instantly through a traceable digital hierarchy would transform carbon accounting from retrospective reporting into proactive, model-driven management. Designers could visualize trade-offs among performance, cost, and carbon intensity during the earliest phases of product development, when decisions carry the greatest leverage. Regulators would gain auditable, consistent data pipelines across supply chains. And the persistent, credibility-eroding discrepancies that have plagued carbon footprinting for two decades could finally give way to numbers that engineers, policymakers, and consumers can all trust – because they were built from the same digital truth as the product itself.

Subject of Research: Methodological limitations of carbon footprint assessment in manufacturing and a proposed model-based digital framework for design-integrated carbon accounting

Article Title: Carbon footprint assessment in manufacturing: Methodologies, limitations, and a future MBD-enabled digital framework

Article References: Yassine Sibai, D., Kassab, A., Pannier, C., Ayoub, G. Y., & Mohanty, P. (2026). Carbon footprint assessment in manufacturing: Methodologies, limitations, and a future MBD-enabled digital framework. Cleaner Engineering and Technology, 34, Article 101325. https://doi.org/10.1016/j.clet.2026.101325

Image Credits: AI Generated

DOI: 10.1016/j.clet.2026.101325

Keywords: carbon footprint, manufacturing, life cycle assessment, GHG Protocol, Model-Based Definition, digital thread, Scope 3 emissions, uncertainty analysis, sensitivity analysis, battery manufacturing, sustainable design, digital twins