Working on a large scale open source flow battery design and kit

During the past couple of years we have been working on the design of a small flow battery kit for the study of flow batteries (you can read a previous post about it here). With the help of an NLNet grant, we have fully developed the first two versions of our small scale kit – with the second one achieving even longer scale tests – and are now working on the development of a larger surface area kit that can be used to study flow batteries or build flow battery applications at a relevant scale for practical power storage.

A flowframe being printed on our Prusa Core One on Polypropylene. Such a large flat area piece is very hard to print without warping in non-heated chambers.

This new large area setup – which we have discussed in our forum – is now moving out of the design stage thanks to the acquisition of new systems for both 3D printing (Prusa Core One) and vinyl cutting (Cricut Maker4) that are allowing us to move to their in-house fabrication and testing. The new cell is designed to be compatible with the area of 3D printing beds and features an active area of 13.5×13.5cm, which is 182.5cm2. In practice this means a single cell will be able to handle a capacity of nearly 22Wh when using the Zn-I chemistry used in our small scale design.

We borrow some of the features that worked great for our small scale kit – like using a polypropylene enclosed flow frame – and add features that are needed for a scalable design (like all entry/exit points on the same side, stacking compatible design, etc).

Finished flow frame held against light so that you can see the flow field design. This design is not final and will change as we do fluid testing.

The fabrication of these pieces is especially challenging due to their size and the complexity of the flow fields in them. The flow fields have to follow complex paths to prevent the creation of large shunt currents within the device. A really good 3D printer with a heated enclosure is required to be able to print this in polypropylene with no warping and good enough resolution. Even with such a printer, a lot of fine tuning is still required to achieve the low level of surface roughness and high level of water tightness that is needed for this particular application.

While we are thinking about the potential of using heat welding or even adhesives to be able to put the cell together, the easiest first approach will be to use the same silicon gaskets as we used for the fabrication of our initial cells. To cut our gaskets in house we are using a Cricut Maker4 vynil cutter, which is able to make this high precision cuts without any problems.

One of our 3D printed polypropylene flow frames next to a silicone gasket

With these materials now fabricated, we are now close to putting together our first tests of a larger scale flow battery cell. The idea of these initial tests will be to test the cell for leaks and make sure we can circulate water without problems before we try to run any active materials. We will be using thick wood as endplates for this initial test, as this is the easiest to source, hard material, that can be used. Thanks to our use of fully enclosed gaskets, the end plates will also have zero contact with any water or active materials.

This flow battery kit work is being funded by the Financed by Nlnet’s NGI0 Entrust Fund. We are also collaborating with the FAIR Battery project.

A low cost, open source, Cu/Mn rechargeable static battery

Building a rechargeable battery is not an easy task. Although many great technologies are available (like LiFePO4 or even lead acid batteries), building these batteries isn’t trivial because of the technological hurdles, manufacturing requirements, chemical substances, knowledge and safety requirements. It would be ideal if we had access to an open source rechargeable battery technology that was easy to construct in practice with readily available materials, robust and at low cost.

This is a sample 0.5mL cell using a copper anode and a carbon felt cathode (0.3mm). A polypropylene felt separator is used between both. The cell is 1cmx1cmx0.5cm in volume.

In a previous post I talked about Cu/Mn batteries and how several different papers describe batteries using Cu sulfate and Mn sulfate along with sulfuric acid to create robust batteries with significant capacities, even above 40Ah/L. Such batteries would be close to ideal as they are easy to build, use earth abundant materials and – in theory – are very robust. However, my efforts to reproduce these batteries were plagued with failure as I was unable to reproduce both their reversibility and their capacities.

Furthermore, this Cu/Mn technology using sulfuric acid has been patented by a French company (years before the Chinese articles started sharing the chemistry in 2017). This means that this chemistry is not open source and significant battles could arise from the use of this technology at a wide scale. This also explains why the patent applications mentioned by some of the Chinese researchers in their papers cannot be found (probably the patents were denied because of the preceding French patent).

Experimental results of the Cu/Mn cell using methylsulfonic acid. The electrolyte was prepared with 0.05g of FeSO4.7H2O, 2g CuSO4.5H2O, 1.8g of MnSO4.4H2O, 2mL of 70% CH3SO3H and 8mL of RO water. Cycling was done at 5mA/cm2

To tackle my problems reproducing this chemistry, the hurdles with intellectual property and some issues dealing with the solubility limit of copper/manganese sulfate mixes, I have modified this technology to use methansulfonic acid (CH3SO3H) instead of sulfuric acid. Methanesulfonic acid is easier to get than sulfuric acid – because it has no regulatory restrictions – and the solubility of both copper and manganese mesylates is higher than that of their sulfates, meaning that even higher capacities than with sulfuric acid should be possible.

The above experimental results show cycling of the cell shown in the first picture. This chemistry achieves a CE above 90% with an EE above 65%, the cycling is also very stable with very reversible MnO2 formation in the highly acidic media. The electrolyte tested is roughly 0.8m Cu, 0.8m Mn and 1m CH3SO3H. I haven’t tried changing the acid concentration or preparing more highly concentrated electrolytes yet, as I am still fine tuning the cell fabrication process to enhance reproducibility. The cells right now can be charged to 20Ah/L, which is already an interesting level of capacity, although 40Ah/L should be possible.

Image of dendrites due to electric field abnormalities around the edges of the Cu anode.

A very important issue I’ve noticed is that dendrites tend to appear around the edges of my Cu anodes due to electric field instabilities, as the Cu prefers to grow in the free solution rather than through the polypropylene separator. This can cause the battery to short when charging to very high capacities). Cells without separators – with just the electrodes hanging parallel as in the case of some of the Chinese papers – could help alleviate the issue. I am also trying passivating the edges using nail polish, to see if this fully solves the issue.

While the Cu/Mn battery chemistry using H2SO4 is clearly patented, the innovation using CH3SO3H is not protected, neither covered by the scope of the current patent nor previously published anywhere else (my innovation to the best of my knowledge). The publication of this blog post should ensure that this technology will remain patent-free.

Update:

The results above show the cell being charged to 1.45V (This is likely the Nernst limit of the cell). I got a discharge capacity of 33.6Ah/L at 10mA/cm2. Electrolyte is identical to the one mentioned before. The CE and EE are still the same at higher capacity. Dendrites do not seem to get worse provided enough space is left between the edge of the copper electrode and the edge of the separator.

Long term cycling of our Flow Battery kit using a Zn-I chemistry

Our goal at the Flow Battery Research Collective (FBRC) during the past year has been to develop and manufacture a flow battery kit that can be used to study flow batteries at a small scale in a low cost yet reproducible manner. Today I want to discuss all the problems we have found when attempting long term cycling in both versions of our kit and how we have addressed them so far.

Long term cycling results with the first version of our kit. Around 5 days of cycling.

Our first kit design – which you can see above – was able to do long term cycling of an electrolyte composed of Zinc Chloride (1m) and Potassium Iodide (2m) in a potassium acetate/acetic acid buffer (pH=5.2). We were able to obtain capacities higher than 10Ah/L (on total volume of electrolyte — note previous values on this blog were based on catholyte only) without much capacity fade at the 25 cycle range. I didn’t try longer term cycling because of issues with pumps getting damaged by the tubing I was using leaking iodide (see how orange the right pump above looks, that’s not supposed to happen). At this point we were also testing several types of tubing, reason why you can see different colors of tubing in the setup.

However, this design had two big issues. The first is that the polypropylene bodies were getting over compressed at the top and under compressed at the center, so after a single long term cycling event (cell cycling for 5 days), the cell would never seal well again after being opened due to warping of the body. This made this design completely unfeasible and forced us to accelerate the development of v2.

Long term cycling results of second version of our kit. Cycling was also around 5 days.

The second cell design – showed in the second image above – has a square design with even compression and complete isolation of the cell body from the electrolyte by using a sealed polypropylene flow frame (you can read more about it in the FBRC website). However, despite using the exact same electrolyte and membrane, this design has been showing some increasing decay of the electrolyte as a function of time. The cycled capacity is a bit lower and after 28 cycles the testing had to be stopped due to obvious deterioration of the charging potential.

Flow batteries using microporous membranes can suffer from significant issues related to hydraulic pressure differences between anolyte and catholyte. As the battery is cycled, there is net water transfer between both sides and this can lead to the accumulation of polyiodide species in a Zn-I flow battery. This deteriorates the charging potential and reduces the SOC of the battery. You can ready more about this decay mechanism in this paper. When my testing was done, there was indeed an almost 50% less fluid in the anolyte vs the catholyte side, interestingly the completely opposite effect when compared to the paper. Changing the supporting electrolyte concentration or Zn salt used – to make the water changes less extreme – or adjusting electrolyte flow rates, can help eliminate these volume balance issues.

New flow frame design for the second version of our kit. You can see the flow channel has been colored red by the catholyte. The flow frame was 3d printed. Since 3d printing leaves some porosity (even at 100% infill) you can see some coloring even away from the flow channel.

There are however other problems with the current tests. The first is that the chemistry is not stable to oxygen – both due to iodide and zinc reactivity – so working under atmospheric conditions without any inert gas purging of the electrolyte has likely made my tests more unstable, due to Zinc passivation and pH instabilities caused by the reactions of iodide with oxygen. Since the new flow frames are 3d printed – therefore not 100% free of pores – they might be letting more oxygen in, which might also be why the system is more unstable than v1 in this regard. Getting flow frames manufactured through polypropylene sintering or injection molding could get rid of this issue.

Another issue – shown below – is the use of graphite foil (grafoil) as our electrode material. Although grafoil does the job fine, it does seem to be porous to iodine, which crosses it and reacts with the underlying copper, increasing series resistance and creating a protuberance on the copper plate (which is readily visible in the image after long term cycling). I scratched the grafoil to then reveal a white powder, which is copper iodide, leading to a further loss of capacity. This problem could be solved by using a proper bipolar plate material or by sealing the grafoil in some manner. Thicker grafoil could also be enough to ameliorate the problem.

Catholyte side graphite foil, 0.5mm, after 5 days of cycling. You can see a bump in the foil caused by buildup of copper iodide material below the electrode.
A high capacity short cycle result using a Zn-I electrolyte. This is using total volume of electrolyte, so a capacity that would be in line with current Vanadium batteries.

I also want to note that we have achieved much higher capacities – see above – although not in long term cycling, as these higher capacities generate much more concentrated electrolytes that have significant problems (such as the precipitation of insoluble solid elemental iodine and the corrosion of the pumps and copper current collectors).

As you can see, getting long term cycling results is not easy. There are many hurdles to overcome to be able to provide a kit that is able to cycle a chemistry like Zn-I long term, without any problems, in a reproducible manner. All these hurdles can be overcome though – as there are plenty of examples of people cycling these batteries long term – so it is a matter of arriving at a proper combination of materials and chemistry. Our work continues! Thanks a lot for all your support.

This flow battery kit work is being funded by the Financed by Nlnet’s NGI0 Entrust Fund. We are also collaborating with the FAIR Battery project.

The Cu/Mn battery mystery

Since 2019, several groups with Chinese authors have published papers describing batteries using a Cu/Mn chemistry (1, 2, 3). This chemistry is very interesting as it has very cheap chemicals (just copper sulfate, manganese sulfate and sulfuric acid) and doesn’t seem to require any significant electrode preparation. The papers use either carbon cloths, carbon felts or copper plates, all with similar results. However, this chemistry is not as simple or as easy to reproduce as they make it seem. This blog post covers my attempts at reproducing these results.

Testing setup

To reproduce these findings I used 3mm thick carbon felt, celgard 2500 as a separator (as tested in reference (2)) and graphoil as current collector material. I also chose an area of 1cm2 in order to minimize material use and simplify calculations. I also treated the felt with a blow torch to improve its wetting abilities, by holding it in front of the torch for 10 seconds per side. I bought copper sulfate heptahydrate, manganese sulfate monohydrate and 15% sulfuric acid from laboratoriumdiscounter.nl. For the electrolyte I prepared a solution containing Cu 0.8M, Mn 0.8M and 0.8M H2SO4. The cell was immersed in 10mL of electrolyte. Given that this is a static battery that deposits MnO2 on the cathode and Cu on the anode, capacity should be limited by electrode volume and not by the volume of solution.

Charge/discharge cycles. Charge was done at 10mA/cm2 to 1mAh, discharge was done to a voltage of 0.4V. Total volume of the electrodes is 0.6mL. Highest discharge density is therefore around 1.2Ah/L.

At low capacities, the battery behaves as shown in the figure above. The CE of the battery is significantly below 100% (~84%) and the energy efficiency is also quite low (~68%). This contrasts with the published literature which often shows CE efficiencies above 90% and energy efficiencies above 70%. I significantly increased the charge to 2.5mAh (4.16Ah/L), which showed a significant decrease in CE, EE and capacity with cycling. Specifically the discharge voltage started decreasing substantially with cycling.

Charge/discharge cycles. Charge was done at 10mA/cm2 to 2.5mAh, discharge was done to a voltage of 0.4V. Total volume of the electrodes is 0.6mL. Highest discharge density is therefore around 3.3Ah/L.

Trying to go to even higher capacities (10mAh), as exemplified in paper (1) which shows values of up to 50mAh/cm2, I got the results showed below. There are very fast decreases in both CE and the EE, with the starting CE being slightly above 85% but dropping aggressively from that point going forward. In contrast with the lower discharge rate experiments, in this case the charging voltage did deteriorate aggressively as well.


Charge/discharge cycles. Charge was done at 10mA/cm2 to 10mAh, discharge was done to a voltage of 0.4V. Total volume of the electrodes is 0.6mL. Highest discharge charge density is therefore around 13 Ah/L.

The electrolyte also shows significant signs of decomposition. The image below shows you a comparison of a pristine vs a cycled electrolyte. You can see how the cycled electrolyte becomes extremely dark, due to the presence of MnO2. This is confirmed by addition of ferrous sulfate, which immediately makes the liquid clear up (as Fe2+ is able to reduce MnO2 to Mn2+). The MnO2 is formed away from the electrode because of the formation of Mn3+ which migrates away and then disproportionates into Mn2+ and MnO2. This explains why there are significant loses in the CE as a function of charging, both due to Mn3+ disproportionation and self-discharge caused by Mn3+ migration into the anode.

Comparison of pristine (left) and cycled (right) electrolyte.

The publishes papers make it seem as though this chemistry is extremely straightforward and reversible, but the facts of Mn3+ formation and disproportionation heavily complicate this approach. It is therefore puzzling to me how the results of these researchers were produced, especially the ones in (1) as their setup uses flooded cells us well, even in the complete absence of any separator. I made similar attempts using copper plates as anodes, 0.4M, 0.5M and 0.6M sulfuric acid and 0.5M, 0.8M and 1.2M Manganese sulfate solutions but couldn’t find any differences in the basic results, the only difference being that current densities needed to be much lower when a copper plate was used, likely due to the much lower surface area.

Let me know if you have any ideas about what I might be missing in the construction and testing of this Mn/Cu chemistry.

Revisiting the idea of using chelates for the Fe/Mn flow battery

On my last post I wrote about the potential of using Fe/Mn in acidic solution to create an Fe/Mn flow battery. I cited a paper published a few years ago which shows that you can achieve reversible Mn3+ chemistry in a solution of sulfuric acid and hydrochloric acid, I then proceeded to confirm this reversibility using cyclic voltammetry of Mn2+ solutions in hydrochloric acid.

However, it quickly became clear from analysis of the paper that this was only at very low capacities. This is because Mn3+ becomes unstable as its concentration increases in solutions, turning into MnO2 and Mn2+.

A 0.5M Fe-DTPA + 0.5M Mn-EDTA solution in an acetate buffer (prepared with 100mL of 8% acetic acid + 10g of potassium acetate)

Given the very low volumetric densities that can be achieved with the acid setup, there’s no option but to revisit the use of more stable and reversible forms of manganese. The best candidate seems to be Mn-EDTA. This complex has already been shown to work in flow batteries at the 0.5M-1.0M range (see here).

I had already thought about using this complex and wrote several posts about its potential use in combination with Fe-EDTA or Fe-EDDHA (see here). However, there is a big problem with the pH compatibility of the Mn-EDTA with the Fe-EDTA or Fe-EDDHA. The issue being that Mn3+-EDTA is only stable under acidic pH conditions, where the solubility of both Fe-EDTA and Fe-EDDHA is limited to around 0.1M. These chelates are only highly soluble under basic pH conditions, which are fully incompatible with Mn-EDTA.

CV of the solution shown in the first image. The half-wave potentials for both reactions are -0.11V and 0.61V, both Vs Ag/AgCl. The above CV was done with a scan rate of 10mV/s.

The question is whether there is any easily accessible Fe chelate that is both compatible with Mn-EDTA in solution (so that we can create a symmetric electrolyte) and that can create soluble solutions at >0.5M concentrations in a pH ~5-6 buffer. Note that I need both chelates to be dissolved at >0.5M at the same time since I want the electrolyte to be symmetric so that it can work using a microporous membrane.

The answer is Fe-DTPA. This chelate is highly soluble at acidic pH values and – best of all – it is soluble enough to actually be in >0.5M solution in the presence of Mn-EDTA at this high concentration. Above you can see a picture of the Fe-DTPA+Mn-EDTA solution. The solution also contains an acetate buffer, which should ensure pH stability on charge/discharge, which should prevent degradation of the Mn-EDTA.

The second image shows a CV of the Fe-EDTA/Mn-EDTA buffered solution, showing that both the Fe and Mn electrochemical reactions are reversible. The half wave potentials are -0.11V and 0.61V, giving us an expected potential for the flow battery of +720mV. This is close to what I had measured before for Fe-EDTA/Mn-EDTA. This proves that the DTPA does not change the electrochemical characteristics of the system very much. The above test also confirms there acetate buffer is stable to the generated Mn3+-EDTA.

The next step is to build a flow battery using the above solution and see what performance characteristics we can get. With the current solutions this system will be limited to around 8-9Wh/L. However I haven’t tested the solubility limits of the chelates in this buffer.