Flow Battery Research Collective

The cost per kWh of this is too high, the use of organic materials and ion exchange membranes puts this closer to what people do in academia and further from the cells that can be reasonably fabricated using DIY approaches. The use of an ion exchange membrane likely reduces the lifetime of this cell a lot. For a static cell, I think the sulfuric acid Cu/Mn battery we have discussed before is far more promising. That chemistry requires no exchange membranes, uses only commodity inorganic materials and cycles to 30-40Wh/L.

For pure Fe batteries I am much more inclined to the Fe flow batteries.

I just wrote a blog post sharing my first results testing the Zn-Br chemistry recently published by a Chinese group. Sodium sulfamate is not very easy to get, but I was able to source it from labdiscounter.nl in the EU. If any of you are in labs where you can also test this chemistry, I would be happy to hear about your thoughts and/or results.

Thanks for sharing your pictures! It always makes us very happy to see anyone reproducing the kit independently.

You can see my current setup below:

I am testing the double reservoir with spill-over communication (that's why the reservoirs look different and are at the center). My pumps are connected in a suction configuration and they enter and exit the cell on their sides. When you have pumps in a push-through configuration having the flow going from bottom to top is important to push air out, but when you suck through the cell the vacuum will push all air out and fill the entire cavity, almost regardless of how you pump. I haven't seen any air being trapped there (no bubbles are evident when shaking or moving the cell so that the flow is in either direction).

As you can see I have changed some connections to luer locks, except the connections for the pumps. I however like how you've made them ALL luer locks, much simpler to disassemble.

As you can see I'm also testing the new clamp-compatible end-plates and clamp. This works SO much easier than the screws, since the compression is centered I am also experiencing fewer problems with the cell and slightly better energy efficiencies since the felt compression seems to be more uniform. Opening and closing this cell is a breeze compared to the screws. I used PLA with 80% infill to print the end plates but I am sure PETG will work great too.

About the PP, it is a bit tricky to find the PP and printer settings that work best for water tight results. I am using the Ivor white PP from smart 3d, which has worked well on the prusa core one. In any case, let me know if there's anything I can help you with.

@rowow

We don't have the equipment or time to prepare these membranes but if you send us samples of 100-400um thick membranes we can test them out.

@rowow I don't know what to tell you. While sulfonated crosslinked polysterene membranes can work great for many flow cell applications (they absolutely have no problem with strongly acidic environments), they will have problems with many flow battery chemistries due to the chemical properties of these membranes and the highly concentrated oxidative environments the membranes are subjected to. You can look at the research on membrane degradation if you want to learn more about this.

These resins have existed for a long time, they have been tested, their problems are well known. You are welcome to try them in flow batteries and share your results here.

As I mentioned the concerns about cost, dendrite formation in deposition chemistries and chemical degradation are mainly why we decided to go with microporous instead of ion exchange membranes. Even if ion exchange membranes were very low cost, for an open source flow battery we believe microporous membranes actually offer better robustness.

This is a polystyrene crosslinked sulfonated resin. It is not going to resist the conditions of flow batteries due to the reactions the aromatic units will go through, which will decompose the resin with time. Bear in mind the problem is NOT solely about the oxidation potential itself, but about the concentration of oxidative species. Potential will control whether some reactions happen or not, but the speed of these reactions will be determined by concentration. A +1.5V potential might be survivable in the long term at mM concentrations but at M concentrations the story can be very different. Sulfonated polystyrenes are not chemically suitable for the application.

Viable low cost alternatives for flow batteries must have more structurally sound backbones, like for example sulfonated poly(ether ether ketone) (SPEEK) membranes (even these won't survive all chemistries). Since the ion exchange membrane is not an easy to remove component and it has to survive for +10 years of very harsh conditions we have therefore decided to use microporous membranes instead. Of course, for demonstration or short term applications, I am sure a polystyrene membrane would be ok with some chemistries.

Thanks for sharing. What is the exact chemistry of the resin you are using? Typically resins used in water softening applications will degrade significantly with the highly concentrated oxidants present in flow battery catholytes. Did you find a fluorinated resin used in water softening that would be low cost?

@gus No, this was both unmodified felt and membranes. This also didn't use any membrane frame, just the photopaper compressed between the silicon gaskets. The charge discharge current was set to 40mA. Before charge/discharge I circulated electrolyte for 15 minutes. I also did 2 cycles to 10mAh at 20mA to make sure I wasn't getting any abnormal behavior.

Also, we're planning to use the suction setup from now on, the safety improvements are worth the potential tradeoffs in my opinion.

This is performance to 1.6V using the "suction" configuration and 3 layers of photopaper with carbon felt on both sides.

@gus Thanks for your reply!

Yes, low temperature is a BIG issue for this system but should be much less so with the trieg present. You can replace 5% of the water with ethanol to see if this reduces the problem.

I am also testing an alternative pump configuration that is working much better, sucking the solution through the cell instead of pushing it through (Kirk's suggestion). So the pumps push solution to the top of the reservoirs and suck it out of the cell, like on this diagram:

This way if the cell gets overpressured it basically stops flowing but tubing never disconnects and spills everywhere. Pumps are also much quieter in this configuration.

I will run some tests with photopaper membrane so that you can get some idea of what I can achieve at this time.

@gus Thanks for writing about your experience and again for reproducing the kit!

One of the issues of the Zn-I chemistry is that I2 can easily generate if the local concentration of iodide is not high enough. This can happen due to several reasons. For example if the electrolyte is not circulated fast enough, if the state of charge is too high or if the current is too high. The triethylene glycol helps avoid this, but it also runs into a limit if the right circumstances happen.

The electrolyte we suggest for testing is 2M I and 1M Zn. The max expected capacity of this electrolyte would be 17.6Ah/L, which our setup (if you use 5mL per side) would be around 176mAh. The 166Wh/L value from the paper you quote has to be interpreted carefully. First, consider that this value is measured in a 5M ZnI2 electrolyte, which is 10M I2, it is 5x as concentrated as the electrolyte we use. Second, this value reported on the paper is accounting for the catholyte volume only, so the actual total Wh/L has to be divided by 2 to compare with the values above, so it would be 83Wh/L. For our electrolyte 5x more concentrated we would expect to get 88Ah/L (105Wh/L at a voltage of 1.2V) which would be higher than the paper you cited, because we indeed expect to extract more capacity because of the use of triethylene glycol.

With that said, Zn-I is a hybrid battery (as you plate a metal) so the mAh/cm2 is also important. If you plate too much Zn not only do you get Zn dendrites but the setup will also clog because of the metal deposition process clogging the felt. The paper you cite uses a cell with an area of 40cm2 but the volume of electrolyte is never disclosed, so we don't know how much they plated in terms of mAh/cm2. However, other Zn-I literature suggests we shouldn't attempt to go above 100mAh/cm2, the lower the safer. For this paper I suspect this value was below the 10mAh/cm2 mark because of the lack of dendrites and the thickness of the membrane. This however means that in our system at 2cm2 we shouldn't charge above 200mAh, so even if we increased electrolyte concentration our max capacity is going to be limited to around 20Ah/L, perhaps we can push this to 26-30Ah/L by reducing volume, but I'm afraid not much more beyond that.

The membrane is another key point. The nafion membranes typically used are much lower resistance and much higher conductivity than a membrane like photopaper. While photopaper works well for demonstrating the cell, its resistance is going to be much higher and therefore our energy efficiencies and current densities will be much lower. If you contact me by chat I can mail you some daramic, which is a commercial microporous membrane, so that you can test it out and get better capacities. Another alternative is to modify photopaper (for example with PVA) to be able to use a single layer.

To answer your questions:

Do you have any advice for me? Where could the root cause of my failures be? - I wouldn't call your results a failure! I think you're doing well. Your charge/discharge curve looks great. You can try the following things:

1. Try decreasing your current to 20mA and see if this way you can charge to the higher SOC, this will show if the problem is just your current density and transfer speeds.
2. Also you can try activating your felt by putting it in commercial bleach for 48 hours, then washing it with distilled water thoroughly before using it. This improves the wetting of the felt a lot and helps the kinetics of the electrochemical reactions.
3. Decrease your volume per side to 4mL, this will reduce the mAh/cm2 which will make clogging less likely as the SOC climbs.

Am I definitely not supposed to use polypropylene felt on the catholyte side? You get best results without polypropylene on either side. Change to felt on both sides to get better SOC values. We also never used it on the catholyte side only on the anolyte side. If you use nonconductive felt on both sides you will reduce the chances of clogging but your energy efficiency will be dramatically lower.

Could there be an issue with the material quality I'm using (even though everything was purchased according to the Bill of Materials sources)? I don't think so, if you got everything from the BOM, then everything should be the exact same I have.

Let me know how things go!

Charging to 6Ah/L at 30mA/cm2 and discharging at 5mA/cm2. At most we only get 2-3 Ah/L of available capacity, same as if we charged to 4Ah/L.

The CE drops a lot when going to higher capacities (even at a current of 40mA/cm2, which means it is not due to crossover as lower currents do not imply a lower CE). This is likely because the stability of Mn3+ species in solution is very limited, so you seemingly cannot exceed a ~2.5Ah/L capacity.

I was doing a literature review of Fe/Mn the other day and happened to find this article on Fe/Mn using MSA (https://www.sciencedirect.com/science/article/pii/S001346862030637X). This article uses an asymmetric setup with FeCl3 on one side (paper says it's FeCl2 but that must be a mistake because the reaction requires Fe reduction on charge) and MnCl2 on the other, both sides using 3M methanesulfonic acid, separated by a Nafion membrane. The Mn3+ is in theory stabilized in the acid media, but given the color of the solution it might be that MnO2 nanoparticles are stabilized instead.

While the paper does not use this in a symmetric setup, I see no reason why this reaction couldn't work symmetrically so I prepared an electrolyte using the following:

• 3g MnCl2.4H2O
• 4mL FeCl3 40% w/w solution
• 4mL 75% methanesulfonic acid (MSA)
• around 1mL of water (final volume was taken to 10mL)

The above creates a solution that is around 1.5M Fe, 1.5M Mn and 3M MSA. This setup has the advantage that both reactions generate no solid products. At a 100% SOC this would give us ~20Ah/L. On charge:

Fe3+ + e- -> Fe2+
Mn2+ -> Mn3+ + e-

The potential difference between these two half reactions is not very high though, so the total expected cell voltage is ~550mV. However this is a "true flow battery" in that power and capacity are fully decoupled as the reaction products are all in solution. Note that Mn3+ is expected to have limited stability, especially at high concentrations, so I would expect capacity to degrade heavily as the Mn3+ gets converted into MnO2, unless this MnO2 is somehow stabilized in solution (which could be as nanoparticles). Interestingly Fe2+ can react with MnO2, so the battery might also self-heal if this happens, just temporarily capacity in the process.

I loaded the electrolyte in a cell with carbon felt on both anode and cathode and used Daramic as a separator (cannot use paper as it reacts with Mn3+). Below are the results of a few cycles at low capacity (0.25Ah/L at 10mA/cm2), just to test the chemistry. It seems to work quite well:

I will continue to run some tests and will let you know what I get.

This test showed some deterioration on cycling:

I took out the catholyte and anolyte when charged (you can see the anolyte (left) and catholyte (right) in the pictures below). There isn't any hydroxide precipitation in either one. However there are some pieces of detached Fe metal on the anolyte, which I think are what causes the slight loss in capacity and increases in ohmic resistance as a function of time.

Running to only 7Ah/L at 10mA/cm2 (1M Fe, 4.5M CaCl2, 1M NH4Cl)

This is the result of charging the 1M Fe, 4.5M CaCl2, 1M NH4Cl cell to the Nernst limit (1.5V)

There is still some decay in capacity due to increases in resistance, although much slower. I will now charge this to 6.5Ah/L, see if it can cycle in a stable manner at that capacity.

I also just made a blog post on this chemistry, with some of the latest results https://chemisting.com/2025/09/15/studying-an-all-fe-chemistry-using-wise-in-our-flow-battery-kit/

@sepi To answer your questions:

1. We are testing concentrated MgCl2 and CaCl2 electrolytes, which have never been published in all-Fe flow batteries. The experiments are therefore innovating in the space.

2. It wouldn't be an issue and in fact this is how Zn-Br commercial flow batteries work, to make them stable they are never charged beyond something like 20-30% of their SOC. The Zn-Br batteries have high enough capacities at 20-30% SOC to make them still quite dense at this much lower SOC rating. For an all-Fe flow battery this would be so low that the cost per kWh would climb a lot, but of course you can always go this route if the compromises are worth it to you.

3. The max voltage the all-Fe chemistry we're studying could give would be around 1.2V given the CV. Ohmic losses because of the solution conductivity and separator thickness are likely a meaningful component of our drop, so are likely the kinetics of plating/stripping and Fe2+/Fe3+ reactions on carbon felt.

4. As far as I know, the main patents on Fe plating flow batteries were issues in the mid 80s, so most of the original patents of this technology have expired. However there are a lot of patents in Fe batteries, particularly dealing with electrolyte modifications and electrode modifications (to reduce H2 evolution).

@sepi Thanks for writing. It is exciting in the sense that Fe systems are great because Fe is low cost, low toxicity, easy to source and very sustainable given how much Fe is present in the earth's crust. However, Fe systems suffer from big problems with hydrogen evolution, as H2 evolution occurs easily at acidic pH (which is needed for the Fe3+ species to be stable in solution). In turn, H2 evolution increases the electrolyte's pH, which then causes problems with Fe hydroxide precipitation. These problems have prevented massive adoption of Fe chemistries in flow batteries, in spite of all the above mentioned advantages.

I had personally never been able to have an Fe system work with our battery system, so the excitement comes from finally having some electrolyte configurations that are sort of working well (at least cycling well at low SOC values with significant CE and EE). The big issue is that there isn't any stability in cycling at high SOC values yet, but at the current state you can run experiments with our kit and help develop Fe battery systems.

The potentials are lower than those expected from the CV experiments, but after recalibrating my potentiostat the losses are actually lower than I thought. So I am now getting potentials near the 0.95V when going to high SOC values.