First results on MgCl2 concentration in all-Fe WiSe flow batteries

Leave a reply

In our previous post we described the research we will be doing on all-Fe flow batteries using water-in-salt (WiSe) electrolytes, thanks to the grant from NLnet. One of the first questions in that list was how the concentration of the WiSe salt affects the efficiency and stability of these batteries. Today I want to share the first set of results on this question, where we look at MgCl2 at different concentrations while keeping everything else in the battery the same.

For these experiments we used 1M FeCl2 (prepared from FeCl2.2H2O) and changed the MgCl2concentration from 3M up to 6M, with a 4.5M experiment in the middle. All the batteries were cycled at 10mA/cm2 using the same pump speed, charged to 20% of their total SOC and then discharged down to 0V. The actual electrolyte volumes ended up being different between experiments (from 9mL down to only 6mL for the 6M solution), so comparing capacities in mAh would be misleading. For this reason all capacities in this post are normalized by volume and reported in Ah/L. With this normalization the charge capacity is 2.68Ah/L in all cases, which makes the comparison a lot easier.

Voltage vs time for 1M FeCl2 with MgCl2 from 3M up to 6M, cycled at 10mA/cm2.

The graph above already shows that the batteries behave quite differently even though they start from very similar places. The first cycle had a low CE in all cases (86-90%), which is something I have seen before with these systems and is basically a conditioning effect of the electrolyte as the Fe chloride is partially oxidized (as it oxidizes slowly in air as a solid) and some electrons during the first cycle are just wasted reducing this Fe. After that, things start to diverge.

Efficiency values vs time for each MgCl2 concentration.

The 3M MgCl2 battery was the most stable one by far. Its CE was a bit erratic during the first 10 hours (one cycle dropped all the way to 84%) but then it settled at around 96% and stayed there for the whole 47 hours of cycling. The EE did start to go down after around 25 hours, ending at 63.7% after peaking close to 69%, but this decline was very slow compared to what we saw in the other two batteries. If we fit a line to the data the EE goes down by only 0.02% per hour.

The 4.5M battery actually gave the highest CE values of the series, between 97% and 98.6% for most of the test. With that said, the EE went from about 67% down to 57% in 42 hours, and since the CE barely moved, pretty much all of that loss comes from the voltage efficiency. There is also a jump at around 20-25 hours that is visible in all the graphs, and the discharge curve of cycle 10 shows some spikes. These were caused by a pump failure we had during this experiment, which interrupted the flow of electrolyte through the cell, so I wouldn’t read too much into that part of the data.

Discharge capacity in Ah/L and capacity retention relative to cycle 2.

The 6M battery is where things went clearly wrong. The CE started lower, at about 92%, and then fell sharply between 17 and 21 hours to around 85%. It never recovered after this. The discharge capacity dropped at exactly the same time, from 2.47Ah/L to 2.27Ah/L, which is a loss of about 8%. The EE ended at 52.6%, the lowest of the whole series, and it was declining at around 0.5% per hour, more than 20 times faster than the 3M battery. This battery also ran for the shortest time, because we had to stop the experiment after around 30 hours when the electrolyte started leaking. The leak was caused by clogging from Fe oxides that built up on the anolyte side, so by the end the battery was in quite bad shape.

Average charge and discharge voltages vs time, along with the voltage gap between them.

The voltage graphs help us understand where the EE losses are coming from. The gap between the average charge and discharge voltages grew in every battery, but in the 3M one it only went from 0.35V to 0.42V over 47 hours. For the 4.5M and 6M batteries the gap reached around 0.5V in 30-40 hours. The end of charge voltage of the 6M battery reached 1.40V after only 30 hours of cycling, and it was still going up when the experiment ended.

Charge and discharge curves early and late in each test, with capacity normalized by electrolyte volume.

Why would more MgCl2 make things worse? Concentrated MgCl2 solutions are quite viscous and their conductivity goes down as you add more salt, so I expected higher overpotentials at 6M from the start. What surprised me a bit is that the initial overpotentials were actually very similar, and the difference comes from how fast they grow. This suggests that something is building up in the cell over time, and in the 6M case we know at least part of the answer. The Fe oxides that clogged the anolyte side would also cover the electrode and make it harder for the electrolyte to flow through it, which fits nicely with the growing overpotentials. I believe the sudden drop in CE and capacity at 17-21 hours is the point where this clogging started to really affect the battery. Why the oxides built up so badly at 6M is something I don’t have a good answer for yet, and it is one of the things we will try to figure out with more cycling experiments.

Long story short, adding more MgCl2 does NOT make for a better all-Fe battery. The 4.5M electrolyte gives a slightly better CE, but the 3M electrolyte keeps its efficiency and capacity for a lot longer, which is what really matters in a flow battery. The next step will be to run the same comparison using CaCl2, and to do some longer tests at 3M to see where the slow EE decline ends up. The raw cycling data for these experiments will be released together with the rest of this first batch of results.

Thanks to NLnet for supporting this research!

Deeper research on all-Fe flow batteries using water-in-salt electrolytes

Leave a reply

During the last year we have been doing some minor research in the world of Fe flow batteries to investigate the use of MgCl2 and CaCl2 water-in-salt (WiSe) electrolytes for all-Fe flow batteries. We have recently obtained a new grant from NLnet to help us better research these chemistries and get to a bigger scale using our larger scale flow cells. The objective of this blog post is to better describe the specific research we will be initially doing and how this will lead to a better understanding of how these flow batteries work.

Our research will use our small scale kits and will be mainly dedicated to the exploration of chemical variations of the all-Fe WiSe chemistries. With this research we would like to answer the following questions:

  1. Which is the highest CE and more stable WiSe electrolyte, CaCl2 or MgCl2?
  2. What is the effect of different WiSe electrolyte concentrations?
  3. What is the highest FeCl2 concentration that can be achieved in these electrolytes?
  4. Can auxiliary salts like NH4Cl improve the batteries?
  5. What is the highest stable energy density that we can achieve?
  6. What are the effects of common minor additives used in Fe flow batteries for improved stability? (citric acid, ascorbic acid, gallium chloride, indium chloride, etc)

Due to the fact that we don’t have the funding or equipment of a large university, we will focus on the more pragmatic aspects of the above questions, that is, we will focus on running the experiments and collecting the battery cycling data (which we will make available for free in our open source website), not so much on answering more complicated “why” questions that would require much more expensive characterizations.

Given the previous experiments we have done and the literature available on Fe plating using WiSe, we will start our experiments with the tests described in the flow chart above. All tests are going to be executed at a current of 10mA/cm2 and all of them will be using 5mm carbon felt that is pretreated with a heating butane torch to improve wetting. All tests will be using a Daramic 900um microporous separator, which has been graciously donated by Daramic for these experiments. All electrolytes will be freshly prepared and filtered using 0.45um PVDF filters before use.

After this we will have a better understanding of how much WiSe can improve Fe flow batteries and whether or not they can achieve stable cycling with limited hydrogen evolution under our testing conditions.

These tests will also build a framework to use for a larger scale flow battery using the all-Fe electrolyte and will inform future experiments we will run. The data for all of these experiments will also become available to anyone who wants to rerun them or analyze the results more closely. The data will be released in chunks as we advance through the different stages of the process.

As always we would like to express our gratitude to NLnet for sponsoring this project!

Why are there no flow batteries with symmetric ferrocyanide electrolytes?

Leave a reply

If you have looked into flow batteries for any length of time, you will have found that the ferrocyanide/ferricyanide redox couple ( [Fe(CN)6] 4−, Fe(CN)6] 3−) is one of the most widely used in the field. This is because this redox couple has very high redox stability, great kinetics, significant solubility (0.7-1.2M depending on the exact salts used) and a redox potential that is lower to that of the Fe2+/Fe3+ redox couple (+0.22V and +0.5V respectively Vs saturated Ag/AgCl), with high stability under high pH conditions.

Potassium ferrocyanide, wikipedia image

However you might have also noticed that there are no published examples of flow battery systems where ferrocyanide salts are used in a common electrolyte, symmetric system. That is, a system where the battery starts with the same exact electrolyte on both the catholyte and anolyte and the redox reactions happen from this mixed state. Common examples are ZnBr2, ZnI2, Vanadium and Fe systems using simple FeCl2 or FeSO4 salts.

Why is this the case? An initial reason is that ferrocyanide forms insoluble substances, Prussian blue or its analogues, with most heavy metal cations, so any battery that wants to do a metal reduction in the anolyte, such as the reduction of Zn2+ to Zn metal or the reduction of Fe2+ to Fe metal, would not work because you would precipitate these solids.

While the above reason makes things more difficult, it is solvable. We have known from the late 1940s that solids of this type can be easily dissolved by using pyrophosphates (see here) and other strategies with strong chelating agents also work. This might tempt you to make a symmetric battery with something like potassium pyrophosphate, zinc chloride and ferrocyanide, where you reduce zinc pyrophosphate at the anode to zinc metal and oxidize ferrocyanide to ferricyanide at the cathode. However this is a bad idea.

Cycling of 0.1M Hexacyanoferrate (taken from this paper). You can see that at 100 cycles it has clearly degraded a lot.

The reason is that anolytes using ferrocyanide, especially when the potential will subject the ferrocyanide side to low potential values, cause ferrocyanide to decompose at the anode to form some Fe metal but also free cyanide. The cyanide will then be free as Zn is already complexed by pyrophosphate but the Ferrocyanide inventory is hard to recover. We know this happens from studies of ferrocyanide solutions under reducing potentials (see here). There are some evident cathodic peaks here that clearly show ferrocyanide is NOT stable at reducing potentials below around -0.5V. When cycled to negative potentials for 100 times, it is clear that it is decomposing. Any free Fe will quickly form Prussian blue or hydroxides at basic pH which will then start destroying the battery. Add to that the fact that we are generating free cyanide and the inventory becomes much harder to deal with.

The above might make you think we can then go the other way and couple ferrocyanide with a higher redox potential couple instead, like these Fe phenanthroline complexes, use a ferricyanide anolyte instead of a ferrocyanide catholyte to create a flow battery in this way. The problem then is that we also have significant anodic current generated at high potentials that correspond to cyanide oxidation and ferrocyanide destruction as well.

The reality is, ferrocyanide is NOT well suited to work in a symmetric system. While it is very stable when working on its own in an asymmetric system, it is not well behaved when exposed to potentials either +1V or -1V away from its redox potential. This basically precludes the creation of a viable flow battery, as batteries with potentials below 1V are not going to be economically viable, especially given the additional costs involved in creating a symmetric system using ferrocyanide.

When using ferrocyanide, also consider that while this salt is relatively safe in its unaltered state, subjecting it to electrochemical abuse WILL generate free cyanide and it’s likely to pose a significant danger to you and others. For this reason, I would recommend to stay away from testing ferrocyanides in symmetric systems entirely, unless you are a trained professional and professionally well equipped to handle both the potential operational hazards and wastes generated from its decomposition products.

Zn-Br sulfamate battery stability

4 Replies

On a previous post I discussed my first attempts at reproducing the Na-sulfamate based Zn-Br battery published by a group of Chinese researchers. My results showed that the chemistry works mostly as they showed, but I was unable to reproduce both the capacity and stability properties of their testing results. This post summarizes some additional research results I obtained with this chemistry and why, I believe, my results have been unable to match theirs.

From the get go, my results showed significant declines in capacity when charging to the Nernst limit. This happened even at lower capacities and even at lower concentrations. Oftentimes with deterioration of the charging potential but sometimes with no changes in charging potential at all. This was irrespective of whether the buffer was prepared with just KAc additions, with HAc+KAc or with different buffer strengths. Additional HAc additions did not recover this capacity, which makes me believe that the losses are due to some permanent loss of the sulfamate inventory. Since these losses often happened with very little or no deterioration of the average charging potential, it also makes me believe these are not due to problems with Zn dissolution. After I opened the batteries I also saw no accumulation of metallic Zn on the anode felt or separator (while when it’s a problem with Zn reversibility you see some clear dead Zn remains).

Charging to the Nernst limit (to 2.1V) shows some clear capacity losses as a function of cycling. The above is for a 0.5M ZnBr2, 0.5M KBr, 0.5M Na-Sulfamate solution in a 5 pH buffer prepared with KAc and 8% Acetic acid. 100% SOC would be close to 6.7Ah/L, as the reaction is limited by sulfamate on the catholyte side. Catholyte and anolyte electrolytes are identical on start. 25mA/cm2 current.

If you read the original Nature paper carefully, you’ll also see that none of their charge curves ever reach the Nernst limit but they are carefully capacity limited to some predetermined value. This initially makes no sense – why would you choose to not use all your capacity? – unless there was a problem with either Zn dendrites or with some other side reaction. Given that Zn dendrites don’t seem to short the battery until much higher capacities, it seems clear that the problem must be elsewhere.

To test this hypothesis I tested the reversibility at 50% of the SOC. It is clear that the deterioration slows down at this point. It is also clear that my cathode – being normal felt – is way less electrochemically active than the carbon nanotube and N-doped felt that is actually used by the Chinese research group. This makes me believe that sulfamate starts degrading at high SOC values, perhaps because N-Br sulfamate starts becoming so concentrated that double bromination becomes possible and then the double brominated N sulfamate is much more likely to decompose with degradation of the sulfamate, possibly into sulfate, ammonium and other brominated side products, like bromate or hypobromous acid. Perhaps the fancy cathode of the Chinese research group has much faster kinetics and is able to handle much faster Br transfers into sulfamate without exposing already brominated sulfamate to double brominations. However, since they don’t charge to the Nernst limit, it makes me believe that they still saw this when they tried charging to higher potentials, hence they didn’t.

Perhaps the most important fact is that capacity recovers if you add more sulfamate, which pretty much confirms that the problem is due to sulfamate degradation.

Same battery as described on the previous image but only charged to 3Ah/L capacity at same current density.

The above implies that sulfamate, while able to support Zn-Br chemistry, is not as stable as it seems on the paper. Careful control over the charged capacity is needed and cathodes that allow very good kinetics for the bromination of the sulfamate are also required. Without significant engineering of the cathode material, it seems that you are limited to around 50% of the SOC – based on the sulfamate – if you want to avoid degradation of the sulfamate as a function of time.

Also, capacities reported by the Chinese group seem to be based only on their catholyte volumes, therefore you have to divide all their values in half if you want to make real comparisons to Ah/L values. They still reach very high capacity values, very close to the actual 100% SOC levels for these systems, although without ever taking the batteries to the Nernst limit. My battery has much higher internal resistance than theirs, which also explains a lot of this difference (as my kinetics are slower, my potential increases much earlier).

Long story short, you cannot just add sulfamate to a Zn-Br electrolyte and expect the battery to work like magic. As it is always the case in batteries, the devil is in the details.

Reproducing Zn-Br flow batteries using Sodium Sulfamate

2 Replies

A recent Chinese Nature paper showed how Sodium sulfamate can be used in Zn-Br batteries to sequester active Br2 into an N-bromosulfamate that is much less aggressive, much more water soluble and even more easily electrochemically reversible than elemental bromine. I also wrote a recent post discussing the potential use of nicotinamide to achieve this (plot twist, it doesn’t work as the nicotinamide Zn complex is not very soluble). In today’s post I want to share with you my attempts at reproducing this chemistry of the Chinese paper using our open source flow battery dev kit.

The paper is very extensive and shares multiple formulations, they share a formulation for normal asymmetric cells as well as formulations to run the batteries using microporous Daramic membranes. Thankfully I have a bunch of 900um Daramic (thanks a lot to Daramic who donated these membranes to us for research). I bought some Sodium sulfamate (NaSA), Zinc bromide (ZnBr2), Potassium bromide (KBr) and potassium acetate (KAc) and proceeded to run some tests.

Tests using 1M ZnBr2, 1M KBr, 2M KAc, 1M NaSA. Charge and discharge were both done at 30mA/cm2.

My tests using the formulations that they disclose exclusively for daramic were not very successful. Formulations using only ZnBr2, KAc and NaSA suffer from either lower capacities because of lower conductivity or issues with hydrogen evolution. This was specially the case when I tried the ZnBr2 1M, KAc 1.5M, NaSA1.5M formulation, which they suggest in the supporting information to reach >50Ah/L. However I think this is a typo and they meant 2M ZnBr2. If you read that paragraph in the supporting information closely you’ll realize why this is the case (they previously refer to a ZnBr2 1M solution and then say this is basically 2x that, but still write it as ZnBr2 1M).

I then proceeded to test using some of the electrolytes they suggest for asymmetric cells, which were much more successful. In particular the 1M ZnBr2, 2M KAc, 1M KBr and 1M NaSA was great, with high CE values and decent EE values (see graph above). I didn’t experience dendrites before reaching the Nernst limit of the cells when using the 900um thick Daramic, which suggests plating is not as aggressively dendritic as with other electrolytes. However dendrites are quite evident when using 300um Daramic, suggesting you need around 300um of Daramic for every 30mAh/cm2. This might explain why the paper restricts most plating to below 90mAh/cm2 when using the 900um Daramic. I have yet to reproduce the Chinese group capacity or cycling stability values, but I believe I have validated the electrochemical principles well.

Clear evidence of dendrites at 78-79mAh (using 8mL of total electrolyte) using a 2M ZnBr2, 1.5M KAc, 1.5M NaSA electrolyte.

It is also worth noting that the Chinese group does some fancy functionalization of their felt with both nitrogen containing groups and carbon nanotubes, which aggressively boosts the conductivity and energy efficiency of the felt for the Br reactions. This is an important different that might justify why they get energy efficiencies closer to 75-80% while mine are just shy of 60%. I also haven’t optimized the compression ratio of my felt, which means that my felt might be under or over-compressed to extract the max EE in this setup. I also lack an oven to properly do air activation of the felt, so my felt is quite suboptimal and just used as-is.

Furthermore, the paper successfully tested a true flow battery setup using Ti-Br. I cannot easily buy TiOSO4 but I decided to try to innovate and test this chemistry in a fully symmetric setup coupled with 0.5M of Fe-DTPA. While Fe-DTPA isn’t expected to be fully resistant to N-bromosulfamate, I figured it might last enough to provide me with some data. Given that the redox potential of the Fe-DTPA redox couple is quite lower than Fe2+/Fe3+, I figured it should give some appreciable voltage in an Fe-DTPA/N-Br-sulfamate battery. Fe-DTPA is also quite soluble and stable at the near neutral pH that favors the N-Br-sulfamate chemistry, so it should work nicely.

Fe-DTPA 0.5M, NaSA 1M, KBr 1M, KAc 1M. Cycled at a current of 30mA/cm2.

The results above, which have never been published before, show that this chemistry works to some extent. The low CE does suggest that a significant portion of the Fe-DTPA is somehow lost, perhaps to oxidation by atmospheric oxygen (I cannot purge my cells with N2 or Argon at the moment), but also likely from just interactions with N-Bromosulfamate across the microporous membrane. With that said, it does show that the new stabilized bromosulfamate chemistry opens up the window to some very interesting options that just didn’t exist before. Perhaps I can test nicotinamide in this setup, where there is no Zn to cause it to precipitate out of solution.

Finally, I wanted to dedicate the above post to Robert Murray-Smith, a fellow chemist in the UK who passed away recently and was a key inspiration for the start of this blog. I know his passing has been very sad for a lot of us in the DIY community, the curiosity and inspiration he instilled in a lot of us will live on. Thank you Robert!