Electrolyte Solutions to Battery Problems (Part 2)
Battery chemistry is a complex beast, an intricate combination of materials science, organic and inorganic chemistry, electrochemistry, and surface science. The puzzle is part of the fun, but there’s a lot to learn. Recently, electrolyte chemist Dr. Jennifer Allen gathered a selection of helpful tips, nuanced points, and hot takes based on misconceptions and oversights she’s learned from throughout her decade as a battery researcher. In this second of a two-part series, she discusses cell design & experimental considerations.
These kids and their newfangled operando cells
After reading last week’s piece on electrolyte theory, you might be tearing your hair out about how insanely complicated electrolytes and interfaces are. We can never know anything, this is a disaster, etc. Luckily, we have operando cells to clear all of that up: cells that can be characterised while cycling, as opposed to the usual ex situ (postmortem) analysis. Operando cells can provide unique insights into mechanisms underpinning battery function, without some of the confounding issues associated with postmortem samples. This can be incredibly useful for studying delicate structures like SEIs. Operando cells are also really cool proofs-of-concept that get us feeling inspired and excited about the scientific method. However – and I hate to rain on the parade – the value of these experiments depends heavily on the compromises made in the cell design.
An operando experiment with an industrial cell is the dream. In reality, accommodations are generally required to permit operando measurement; of course, these compromises come with consequences. For instance, a flooded cell designed to probe the electrolyte volume may not be representative of the electrolyte reactions occurring in a cell with a much smaller electrolyte volume, particularly with respect to solubility. Some operando cell designs don’t use a separator, while many others use glass fibre separators (more on that later). Electrodes made for operando cells can be fragile, with low active masses, low densities, and sometimes strange recipes (binder-free, carbon-free). Operando cells are often prone to air ingress over time, due to unique cell casing materials, which limits the length of time the cell can be operated. The cell operating time is also limited by instrument availability constraints, resulting in many formation studies with few longer-term studies or even studies past e.g. cycle 10. Operando cells are also often limited to slow cycling rates, which may not be representative of how real-world cells are cycled. Another issue with operando cells is that they are inherently very low throughput and tend to suffer from poor reproducibility. It’s hard to get insights into the long-term degradation of practical cells if you can only study the first few cycles of a single, really weird cell.
By contrast, an ex situ experiment may allow for study of an industrial cell, or at least a more conventional cell that employs more standard materials, C-rate, cycle life, pressure, air impermeability, etc. Of course, ex situ studies introduce all sorts of new potential pitfalls, primarily contamination concerns. Still, if exposure to potential contaminants can be managed, and the potential decomposition of chemically and mechanically sensitive species can be prevented, I love a good ex situ experiment – and I say this despite (or perhaps because of) having worked with ten different operando cell designs over the years.
Although academia has a bias towards operando studies, perhaps for their novelty and ingenuity, I think sometimes (sometimes!) an ex situ study performed on more representative cells can be more meaningful.
Glass separators and glass-eating electrolytes
In small-scale work, separators are often ignored and assumed to be inert or irrelevant compared to the chemical complexity of the electrodes and electrolyte. But think of it this way: the separator is in contact with the anode, cathode, and electrolyte, all at the same time, and is responsible for preventing catastrophic short circuits. Separator engineering is its own separate art. I know many researchers using glass fibre separators are doing so due to practical resource constraints, but I am nevertheless obligated to say it: we know battery electrolytes make HF, and we know HF reacts with borosilicate glass.
The exact reactions between HF and glass fibre separators are surprisingly hard to pin down. Ex situ studies at 60 °C and 80 °C indicate HF reacts with the B2O3 and SiO2 in borosilicate glass to produce water, which in turn reacts with PF6– to produce more HF, etc. These continuous reactions are catastrophic for the electrolyte solution, leading to substantial PF6– decomposition.
By contrast, experiments in cells have shown that glass fibre separators have HF-scavenging qualities that lead to higher cell capacities and longer lifetimes. It’s been suggested that silica-containing separators degrade the electrolyte when stored but improve performance in cells because the H2O produced from reaction of HF with SiO2 is removed in cells via reduction to LiOH and H2, thereby stopping the cycle of PF6– degradation. In principle, glass fibre separators degrade the electrolyte; in practice, they artificially improve cell performance (compared to the performance that would be expected with a more conventional industrially-relevant separator).
Regardless of the reaction mechanism, we may conclude that a glass separator is not the ideal representative material to study HF-contaminated electrolytes. Due to their thickness, glass fibre separators also result in relatively large electrolyte volumes being used, when a smaller volume may be more representative of the final use case. Even in cases where a fluffy and absorbent separator is specifically needed, there are other, better (though admittedly less accessible) options, like thick nonwoven polypropylene separators.
There is no such thing as a half cell
I’m kidding. But, while I’m griping about cell formats, I do think the term half cell allows us to forget that both electrodes are interacting with the electrolyte and affecting the cell.
Imagine two battery researchers. Both have just prepared identical NMC-Li coin cells, to be cycled in a two-electrode configuration. One researcher says, “I’ve made a half cell. I’m isolating the NMC cathode to separate out all the messy crosstalk effects of a graphite anode.” The other researcher says, “I’ve made a lithium metal full cell. I’m getting insight into the function of the lithium metal anode and how it interacts with NMC.”
Given the extent of feedback loops and crosstalk reactions in lithium-ion cells, I tend to treat half cell results with a degree of caution when interpreting full cell results. The continuous interactions between the Li metal electrode and the electrolyte solution limit the value that can be gained from half cells, particularly when looking to do long-term cycling tests or when looking to screen electrolyte candidates.
A three-electrode cell is an alternative that can provide valuable insights; if a three-electrode cell isn’t possible, some full cell configurations can be equally or more useful than half cells. If, for example, you wanted to isolate the effect of an NMC cathode, it might be preferable to cycle NMC vs. LTO or delithiated LFP (rather than Li), because they operate at a much higher potential. There is even evidence that the surface film formed on NMC is more similar with graphite and LTO anodes as compared to Li anodes.
Even then, crosstalk is annoyingly intertwined with cell performance, and it’s incredibly hard to evaluate full cell behaviour as two isolated cell halves. There are always going to be necessary compromises made in model experiments, but it’s at least good to know the limitations of the method and what the results can – and cannot – tell you about full cells.
What’s a good electrolyte baseline?
A good experiment needs, of course, a control condition. But the best baseline or control cell may not always be the most obvious one. In the electrolyte research space, it’s surprisingly common to find studies where the control case is a cell that employs zero electrolyte additives; often simply 1 M LiPF6 in a blend of cyclic and linear carbonates. For cells with low-voltage anodes (graphite, Si, Li), this is a poor choice of baseline that artificially inflates the value of the experimental modification. If there are no electrolyte additives used at all, then it’s easy for any new additive to be somewhat better than nothing. However, this is not representative of the research environment around cells with graphite anodes, where many beneficial SEI-forming additives are known and frequently employed. Comparing a cell employing a new electrolyte additive to a cell with no electrolyte additives feels akin to comparing a cell with a new functionalised carbon to a cell with no conductive carbon at all. Why would we ever make a cell containing no conductive carbon at all when we know it would be bad? The same goes for electrolytes without additives (in the context of low-voltage anodes).
Also, on the topic of control cells: electrolytes are already so temperamental. It’s helpful to always run new control cells, preferably made at the same time as the experimental cells. Even two cells made with the same electrodes, separator, electrolyte, and cell parts in the same environment by the same person can still vary significantly in performance, for a million reasons. Batteries are finicky: maybe electrolyte expired or was contaminated; maybe materials, even if from the same batch, weren’t homogeneous; maybe components weren’t dried exactly the same; etc. ad nauseam. This is also the case for purchased supplies, where batch quality may vary, but doubly so for anything prepared in-house in a research environment. It’s vital to have a reliable, and meaningful, control to accurately portray how amazing your experimental innovation is.
The exact right additive concentration
Speaking of the amazing experimental innovation – with how tightly electrolyte results are tied to the exact system under study, how much can we assume when trying to adopt an innovation from the literature into our system? This often comes up when deciding how to go about testing a promising new electrolyte additive.
Unlike salts and solvents, electrolyte additives are essentially by definition employed in small amounts for targeted outcomes. This means the ‘perfect’ concentration of an electrolyte additive can vary wildly between different cells. Even when using the same active materials and the same electrolyte composition, some factors to bear in mind are the cell format and casing material; the electrolyte volume; the separator material; the electrode surface area, thickness, porosity, and tortuosity; and cycling factors including the upper cutoff potential, C-rate, and number of cycles.
If we think about a film-forming electrolyte additive, its concentration might be better expressed relative to the electrode surface area rather than the electrolyte weight. Even factors like the necessity of SEI self-repair will be important – Are lots of acids being generated in the electrolyte? Is lithium plating occurring? Is the electrode expanding significantly? – as the optimal concentration of a film-forming additive can be slightly higher than what is actually needed to form the initial film. If an article reports the ‘optimal’ concentration of an additive, it’s something worth taking note of, but it’s also not something written in stone. The trusty ‘buy it and try it’ method will probably give you the best idea of what concentration is optimal for your specific system.
Karl Fischer titration of electrolytes, and when it fails
Now that we’ve worried about the electrolyte composition, we can also worry about the electrolyte purity (yay!). Karl Fischer titration is widely used to quantify water in a solvent, electrolyte solution, or other cell components. The water content is significant because LiPF6 (and some other fluorinated salts, like LiBF4) will hydrolyse to form HF, which is harmful to the cell. We can restate this as: water is an electrolyte contaminant that will participate in unwanted reactions to produce harmful products. Or: water is bad because it will react away into other bad stuff.
Researchers can use Karl Fischer titration of a solvent or a freshly mixed electrolyte to check if it’s wet. In this scenario, the water content is used as a proxy for the levels of HF that will form over time. But what happens if we perform KF titration of a premixed or old electrolyte solution?
Once a nonaqueous electrolyte is prepared, trace water begins to react with LiPF6 to form HF, and will continue reacting until it is almost entirely consumed. After a few days, the water content of the electrolyte solution will be fairly low. At this point, KF titration will only show the small amount of remaining water, potentially leading to the erroneous conclusion that since there’s no water in solution, there’s no HF in solution.
If planning KF titration of an electrolyte solution, it’s important to know when it was prepared. Ideally, one would do KF titration on the solvent, or on the electrolyte solution as soon as it’s mixed. The latter method also provides insight into the water content of non-solvent components (i.e., salts and additives, or the bottle used for mixing the electrolyte).
This is not to say that researchers must only work with freshly prepared electrolytes; in many instances, this isn’t practical. But if KF-determined water content is no longer a proxy for the HF content when an electrolyte solution isn’t fresh, then how can researchers directly measure the HF content of older electrolytes? Here’s a few options to choose from, which vary in accessibility and specificity. The most chemically specific, NMR spectroscopy of the electrolyte solution will reveal HF both in the 19F and 1H spectra (though perhaps easier to detect in the 19F spectra due to the labile nature of acidic protons). Much more accessible, acid-base titration will quantify all H+; this can be used as a proxy for HF, though it will also capture other acids. Lastly, ion exchange or ion exclusion chromatography, capillary electrophoresis, or fluoride ion-selective electrodes can be used to quantify F– as a proxy for HF. In short, if HF is the real baddie, then sometimes you need to look for HF and not for H2O.
Closing thoughts
Electrolytes and interfaces are hard work. My intention here is to help researchers design the most useful experiments and perform the most accurate analysis of our results, while also understanding the limits of our methods and data. Once an interpretation makes its way into a scientific publication, there’s a chance it will be accepted and repeated by readers, especially those who aren’t experts in that particular field. By treating our analyses with rigor, caution, and nuance, then we can perhaps put a bit more faith in our conclusions. I can only hope I’ve also convinced you along the way that electrolyte chemistry is complex, fascinating, and not to be underestimated.
🌞 Thanks for reading!
📧 For tips, feedback, or inquiries - reach out
📣 For newsletter sponsorships - click here