In this interview, Robert Mitchell from CPI discusses how battery materials engineering is driving improvements in battery performance, manufacturing efficiency, and cost. He explores materials optimisation, the challenges of scaling from lab to production, automation and AI in R&D, and where he expects battery chemistries to be in five years.
Could you briefly introduce yourself, your role at CPI, and what led you to focus on battery materials engineering?
I’m Rob, a Principal Scientist at CPI working to support companies in developing battery materials from powder to cell. I first became interested in materials during my Master's by working on the synthesis of powders for solid oxide fuel cells, which led me to specialise in inorganic materials science, a focus I developed alongside electrochemistry through my PhD.
I worked on a number of material types, including photocatalysts, porous structures to harvest light, quantum dots, and then, through further postdoctoral roles, catalysts. I then moved to CPI as a Senior Scientist working on graphene materials.
All these materials shared similar challenges: a need to work together across disciplines, solving fundamental material challenges as well as device-level understanding; a need for quality control; an understanding of how to formulate materials to deposit in the right format; and the understanding of how features at the nano or micro scale can affect performance at the macro scale.
In 2019, we started to move into battery materials. These lessons were valuable in building the capabilities required to support the growing battery industry, and, for me personally, in growing from a scientist working in the labs to a Principal Scientist leading teams, and ensuring we grow with the industry in a more externally facing capacity.
Image Credit: IM Imagery/Shutterstock.com
What changes in recent years have made materials optimization so critical to competitiveness?
Materials have always been a critical element to the competitiveness of a battery, with fundamental chemistry defining many aspects of capacity, safety, and performance. LFP was known as a structure for decades, but it was the Goodenough group in the late 90's that first demonstrated its use as a battery cathode material. Commercial relevance followed in the mid-2000s, once uniform carbon coatings and nano-sizing overcame its intrinsically low electronic conductivity.
In recent years, the costs of battery cells have plummeted, leading to further considerations of how money can be saved in the process. Cathode materials in particular represent around 30-40% of the cost of a cell, varying considerably by chemistry and with the raw material prices of the day, and falling as a share at pack level.
There is increasing desire to develop synthesis routes for improved materials, specifically to gain more control over the morphology of particles produced, thus improving cell performance. This can include producing multimodal particle distributions, improving coating methodologies, or changing the growth process to produce single-crystalline materials.
Could you walk us through the battery material synthesis process and the main ways engineers tune material properties?
The fundamental process to produce battery materials targets the material's crystal structure at the atomic and micron levels simultaneously, as well as the optimal size of materials for electrochemical storage and release of metal ions.
Typically, early steps involve mixing precursors, which can be in solid form for a classical solid-state reaction depending on the precursor selection, or through mixing liquids.
For example, layered oxide materials such as NMC or sodium ion equivalents often start with combining precursors in the liquid phase and reacting to form a mixed metal hydroxide, which is grown to the correct size through pH control and complexing additive addition (NH3).
Some alternative routes forego the mixed metal hydroxide precursor in favour of direct solid-state reaction, which can avoid formation of the sodium sulfate byproducts generated by conventional co-precipitation, a significant waste challenge in battery material production. Mixing is then limited by solid-state diffusion, so higher calcination temperatures are needed and control over particle size and morphology is largely lost.
LFP materials generally start with either solid-phase materials or spray-dried precursors that agglomerate to the right size. After this initial step, calcination is often used to heat them to a high temperature to form the right crystal phase. This is generally performed in an atmosphere to control the reduction/oxidation behaviour of the metals involved in the structure.
Coating processes are also important. These can be incorporated during the synthetic process, the preferred route for LFP, e.g. carbothermal reduction methods which produce a material with a thin carbon shell, or for NMC, often during post-processing through methods such as liquid phase addition, spray coating, or ALD (atomic layer deposition).
For anode materials such as graphite, production routes differ depending on whether natural or synthetic approaches are taken.
Natural graphite is mined from the ground with the required ordering of the hexagonal carbon layers already in place, whereas synthetic graphite produces this ordering through treatment at high temperatures up to 3000 °C. Alternative methods are under investigation to lower the energy of this process.
Both natural and synthetic materials are sized for the application, and typically spheronised, whereby the graphite flakes are folded into a spherical shape, improving tap density and packing while giving more isotropic Li-ion access by exposing sheet edges at the particle surface. The trade-off is yield, which is typically only 50-60%.
Coatings are also important for performance, with one option mixing the material with pitch and decomposing it into a thin carbon layer, which aids the interfaces when incorporated into a battery electrode.
What most commonly goes wrong when a material moves from lab scale to pilot or production scale?
The most common pattern we see is that performance attributes which can be achieved at the small scale are not replicable at larger scales.
At gram scale, precursors can be mixed to high uniformity beyond what a large-scale blending process can produce. Calcination happens in a small crucible where the entire bed sees effectively the same temperature and gas composition, which needs engineering optimization to be achieved when moving to larger saggars or a rotary furnace, in which material at the center of the bed experiences a different atmosphere to material at the edge.
The expensive part, in both time and money, is that the consequences are only visible electrochemically. Confirming whether a scaled process has preserved performance means building and cycling cells, so every iteration of the production step carries a long and costly feedback loop.
Routes to improve this can target indicative performance factors to shorten the iteration loop. For example, we supported scale-up of a material from a gram-scale planetary ball mill process to a multi-kg process using a scalable low-energy mixer. The ball mill was producing variable feeds in the precursor which on calcination led to secondary phases as observed by XRD. The optimized process used particle size distribution (PSD) of the resultant mixed precursors as a proxy for success of the process. A defined PSD led to single phase product following calcination.
Scale-up constraints need considering during material design, not afterward. This means identifying early, cheap measurements that predict downstream performance, so a process can be controlled and monitored at interim stages. Examples include precursor PSD as an indicator of likely final phase purity, ‘soft sensors’ like mixer torque to indicate rheology, and in-process monitoring such as ultrasound or X-ray CT.
Which microstructural features or particle-packing metrics are most predictive of battery performance, and how do you measure them?
One of the key challenges in battery material production is that there are few simple ways to estimate electrochemical performance without direct testing in a battery, so most specification sheets for battery materials also include metrics for capacity measured in a test cell. Parameters that can indicate battery performance are:
- Impurity content. This is measured by dissolving the material in acid and analyzing it by ICP-MS, with high levels of metals a particular concern during long-term cycling.
- Moisture level in a material. High moisture levels lead to early failure. This can be measured, for example, through coulometric Karl Fischer titration.
- Crystal phase. Crystal ordering defines the ability to store metal ions in the structure, and so to function in an electrode. Phase is studied by X-ray diffraction, coupled with pattern matching and refinement through e.g. the Rietveld method, to understand the exact environment and diagnose factors such as Li/Ni mixing in NMC type materials.
- Particle-size distribution and shape. This governs a material’s packing behavior and can be tested either through particle imaging (electron microscopy) or by laser-diffraction-type measurements.
- Tap density. This indicates how efficiently the powder packs. It is measured by tapping a known mass of powder in a cylinder until the volume stops changing, using a standard method. A higher tap density supports a higher electrode density after calendering, and so a higher energy per unit volume in the cell.
- Surface area. This affects material formulation and can influence SEI formation and first-cycle capacity loss. Surface area is typically measured as BET surface area by nitrogen physisorption.
When we work with customers to diagnose material and cell challenges, the most common issues come from moisture, impurities, and crystal phase, so these are typically confirmed first. Particle size, morphology, and surface area define how the material will form into an electrode and can be adjusted by post-processing. Tap density is largely set during the synthesis process.
Overall, one metric in isolation cannot guarantee battery performance: battery performance results from a combination of many factors in the material, how it was made into an electrode, and how it is tested electrochemically. Two specifications can look similar on paper, yet lead to differences in cell performance. This is why validation in a battery cell is important for the specification sheet.
Can you share examples of impurities that have particularly significant impacts on battery performance or lifetime?
Stray metal ions are a common cause of battery failure. During cycling, these can be deposited elsewhere in the battery and lead to dendrite formation. In particular, stray particles of Cu, Fe, Ni, Cr, and Zn are problematic. This is part of the reason for high bulk purity requirements for battery materials (>99.9%).
Moisture can cause serious issues with batteries, with limits typically set in the tens to hundreds of ppm depending on the material, and below roughly 20 ppm for the electrolyte. Moisture reacts with the electrolyte to produce HF, which can attack various parts of the internal structure, such as dissolving the current collector or causing dissolution of the transition metals in the cathode.
Typically, battery manufacturing avoids these issues through use of high-purity input materials and through processing in a dust- and moisture-controlled environment.
This is also where a challenge comes for battery circularity. The impurity specification is finely controlled for first life applications. Depending on how the materials are extracted from the battery, impurities can be retained, which affects confidence in the material. There are regulatory drivers: the EU battery regulations require increasing recycled content in future cells, with key increments in 2031 and 2036. That means material suppliers must understand the impurities present in recycled feedstock and be able to validate their impact on cell performance. Without that, an OEM cannot incorporate recycled material into an electric vehicle and stand behind the warranty. In practice, battery producers will push recyclers toward processes that control impurity content tightly at minimum cost.
How do you balance trade-offs between energy density, fast charging, safety, cycle life, and cost when optimizing electrode structures, particularly for LFP?
To achieve high electrode densities, a material with high tap density in powder form is required. Strategies to achieve this include alternate precursors, multiple cycles of milling/grinding and furnace treatment, or production of carbon layers through multiple steps.
Processes to produce high-density LFP in as few steps as possible are desirable to keep the costs low. Compared to the traditional carbothermal reduction process, adding extra steps can add cost and complexity. In some instances this is compensated for by improvements at the electrode level, achieving a lower cost per kWh.
Another key parameter in optimizing electrode performance is porosity within the layer: one way to achieve higher electrode compaction levels is to reduce overall electrode porosity. This must be balanced with the ability to infiltrate the electrode with electrolyte for effective cycling.
An additional factor is the carbon network within the structure. Typically, approaches to maximise electrode performance will employ lower-dimensional carbons such as carbon nanotubes, which, combined with the carbon layer on the LFP, lead to effective percolation of the conductive network and can reduce the mass fraction of carbon used.
Electrode engineering combines the structural design requirements of the electrode with the capability of the manufacturing process to deliver quality at scale. The balance between these factors is set as much by what the coating and calendering process can reliably produce as by what the material can theoretically achieve.
What development strategies or workflows have you found most effective for accelerating materials and electrode optimization while reducing cost and time?
Optimizing formulation or electrode properties can involve extensive trial and error, which can be costly in both time and performance. At CPI, we use design of experiment approaches to learn faster, linked to automation of the R&D process through use of robotic systems to evaluate and reach the optimal specification faster.
The increased throughput enables deeper probing of a formulation space. A manual process might take half a day to produce one slurry and coating; an automated system can prepare 20-30 compositions and 50 electrode coatings in two days, and turn those into cells with further automation. This maps a wider parameter space than testing a small number of pre-selected formulations, and identifies the interactions between variables instead of assuming them.
For example, a project we worked on looked at conductive additives, with screening through electrical conductivity of the finished electrode. The parameter space included materials with multiple morphologies, including spherical carbon black, plate-like graphitic additives, and rod-like carbon nanotubes. Through use of the automation systems we could tune a multicomponent blended system in which the packing and conductive network were balanced to maximise conductivity whilst working at a total carbon content of 1–2%. Doing this manually would have required extensive time and thus opportunity cost, and would likely have missed the combinatorial effects.
Alongside throughput, the reliability of automation is key to producing high quality data. Automated dispensing gives exact, traceable quantities for every addition, removing transcription error and allowing final compositions to be checked against target. Automation also means each sample sees the same mixing energy, coating speed and drying profile, so differences in the data come from the formulation, not from handling. That comparability makes the data usable for modelling, unlike noisy manual data.
This mirrors the approach taken in industry, where automation drives system productivity. The same logic applies further down the line: automation makes R&D data structured enough to model and to support AI-driven approaches to materials development.
How should collaboration between material suppliers, cell manufacturers, and end users evolve to better use materials engineering as a competitive advantage?
Having a close technical relationship between material suppliers and cell manufacturers is critical for a number of reasons:
- Supply chain resilience, securing offtake
- Technical validation and roadmap alignment
- Awareness and derisking of product line changes which may result in changes to processes
Cell manufacturers will typically operate a smaller pilot line to evaluate material differences and future specifications separately to the main production line. This should be closely matched to the end user's requirements to enable offtake of produced cells.
Material improvements do not always translate into adoption. One example is LMFP, which currently has not been adopted as quickly as predicted, partly due to challenges in the material (lower conductivity, manganese dissolution, a two-plateau voltage profile that is more pronounced at intermediate Mn content) and partly due to its higher-than-LFP costs.
Good collaboration and technical understanding across the supply chain are important to enable translation of material improvements into battery performance. Vertical integration is another approach exemplified by BYD, operating from mine to application. The BYD Blade 2.0 is a useful illustration of improved cathode materials translating directly into cell-level performance. Most launch coverage described it as a move to LMFP. The type approval filing for the first large pack lists LFP, though, and the reported cell energy density points to a Gen 4 high-density LFP rather than an LMFP cathode.
For the end user, the benefit of an improved material has to be seen in terms of performance, typically at the same or reduced cost.
Which emerging characterization tools, processing technologies, or AI/data-driven methods are you most excited about for improving battery materials and manufacturing?
Some particularly interesting angles include:
- Polaron, an Imperial College London spin-out, applying generative AI to microstructural image data to characterise electrodes and design improved microstructures, including automated analysis of particle size and cracking
- Illumion uses charge photometry to image particles during charging and discharging processes, giving greater insights into the effect of material improvements on performance
- Automating the R&D process and integrating AI to give faster insights.
AI cannot physically make materials, and is best used when large, high-quality data sets can be fed into it, enabling evaluation of trends and iteration. Therefore, developing physical hardware that can effectively interact with AI is important. In addition, the cell assembly process can lead to marked differences in performance. In one report identical commercial electrodes gave anywhere from 170 to 2500 cycles in coin cells, depending on stack height and the steel grade of the casing, against a pouch cell benchmark of 3500 cycles. (A. Smith et al., Batteries & Supercaps, 2023, 6, e202300080) Standardised processing and transparency in the production process are critical for comparing materials whilst building large data sets.
This is why CPI focuses on automating R&D from powder to cell, allowing production of large quantities of data on the processes of slurry mixing, electrode coating, and battery cell evaluation, which would otherwise be neither attainable nor consistent by manual production methods.
We are developing closed-loop routines for battery material evaluation from powder to cell as part of the FULL-MAP Horizon Europe project, which is building an acceleration platform for materials design and route to commercial implementation.
Other examples of physical hardware that can support this large data set production for the synthetic process through access to wider compositional spaces include the A-Lab at Lawrence Berkeley National Laboratory (California) and DIGIBAT at Imperial College London, itself one of the FULL-MAP partner facilities.
Where do you expect battery materials engineering to be in five years?
Most gains of the last five years came from material morphology, electrode and pack engineering applied to chemistries that were already commercial, rather than from new chemistries arriving. On that basis I would expect LFP and graphite to remain the mainstays of mass market batteries for mobility and energy storage, with NMC continuing as the chemistry of choice for high energy applications.
Silicon incorporation will keep rising as first cycle losses and volume expansion are brought under control, mainly deployed in high energy systems where cost tolerance is higher and the cathode is not the limiting factor. Calendar ageing is an underappreciated challenge. Silicon is seen as a key western strength, shown recently by the US Department of Defense issuing a conditional loan commitment of up to $1.4 billion to Sila Nanotechnologies for silicon-carbon anode capacity and a cell plant for defence and specialty applications.
LMR is another chemistry of interest, having been known for thirty years and never commercialised because voltage fade from oxygen loss and manganese migration limited cycle life potential. Recent gains from a material engineering standpoint have focused on dopants, surface coatings and particle-level morphology control. GM and LG Energy Solution are targeting US production by 2028. LFP followed a similar path, sitting dormant for decades until coating and sizing made it viable, and LMR may prove another case of a long known material coming good through materials engineering rather than new discovery.
Sodium ion has real potential against LFP, but is currently limited by two points: volumetric energy density and supply chain. For energy storage, due to the lower density of sodium ion materials, a 20-foot container reaches just over 2 MWh against around 6 MWh for LFP (CRU), which favours LFP until the expected cost benefits of sodium materialise. Anode-free and metal anode designs would address this energy density gap, with the recent Mana Battery and Saft agreement one example, though aimed at defence and industrial UPS rather than grid storage. On the supply chain side, promising approaches include that of companies such as Batri, producing hard carbon from consistent anthracite feedstock rather than variable biomass. Sodium ion’s key current advantage is in low temperature performance, and adoption will continue to grow as the above points are tackled.
Solid state will continue to develop, but I would expect volumes to remain low, with the key challenges being the interfaces, the modifications needed to existing production lines, and cost compared to mass market chemistries.
For all of these chemistries the key constraint is manufacturability, and the organisations that last are the ones that establish reproducibility, quality and a route to cost competitiveness early in the scale up process, not late.
About the Speaker
Robert Mitchell has a PhD in Electrochemistry and 15 years experience in developing materials for energy applications. Rob leads CPI research programs in batteries and critical materials, delivering over £4.9 million in projects to address commercial challenges in these industries. He supports industrial R&D and technology scale-up across the supply chain, supporting customers to derisk technologies.
This information has been sourced, reviewed, and adapted from materials provided by the author in a private capacity and do not reflect the views of CPI. CPI support battery process development and scale up, from powder to cell – see links for more detail: AMBIC centre: https://www.uk-cpi.com/about/national-centres/advanced-materials-battery-industrialisation-centre
For more information on this source, please visit CPI - Centre for Process Innovation.
Disclaimer: The views expressed here are those of the interviewee and do not necessarily represent the views of AZoM.com Limited (T/A) AZoNetwork, the owner and operator of this website. This disclaimer forms part of the Terms and Conditions of use of this website.