An Open Source DIY Flow battery

Over the past year, I’ve collaborated with my colleagues Kirk Smith, Sanli Faez, and Joshua Hauser on developing an open-source flow battery design and kit. Our aim is to make it feasible for most individuals to construct this flow battery with readily available parts that can be either purchased online or fabricated affordably. We’re targeting a price point below 1000 EUR, inclusive of the potentiostat, to ensure accessibility.

The kit encompasses all necessary components for constructing and utilizing a flow battery for research and development purposes. This includes the battery itself, pumps, electronic components for pump operation, potentiostat, tubing, reservoirs, and a jig for orderly arrangement. Presently, similar setups cost upwards of 9000 EUR, hence our aspiration for significant cost reduction.

A polypropylene FDM printed prototype being tested with Mn/Fe chemistry. This particular test was done without reservoirs, on close circuit circulation to easily detect any leaking.
Image of one of our latest prototypes. This features polypropylene FDM printed reservoirs, a resin printed cell body and a PLA FDM printed jig.

Throughout this endeavor, we’ve explored various fabrication methods for our designs, employing FDM and resin 3D printing techniques alongside traditional CNC fabrication. While all three methods are viable, our experiments indicate that the most optimal results are achieved through traditional milling.

Charge/discharge cycle using photopaper separators and the Zn-I chemistry using the open source flow battery design.

Validation of our design involved utilizing a low-cost photopaper separator and Zn-I chemistry. We’ve achieved successful charge/discharge cycles at capacities ranging from 20-40 Wh/L. However, long-term cycling validation remains ongoing, as we’ve only been testing the final design for approximately a month.

Our design will be presented at the Flow4UBattery Event in Eindhoven, Netherlands, on April 8-9, 2024. You can register here for free, which also includes complimentary lunch (so please make sure you intend to attend if you subscribe). Day 2 of the event will feature a workshop where participants can assemble a flow battery themselves using the design from our kit. Additionally, we’ll be giving away 5 complete kits during the event, each including mystat potentiostats. We’ll also have a fully assembled kit doing cycling so that you can see the fully assembled kit in action!

After this event, we will look into selling these kits online, with all proceeds going towards the development of higher capacity kits with the objective of reaching an open source flow battery stack within the next 2 years. We will also be publishing the full designs and bill of materials online, so that anyone can create their own too!

A Zinc-Iodide flow battery using a matte photo paper membrane

On my last post, I showed the results of charging/discharging a flow battery using a ZnCl2+NH4Cl+KI electrolyte using 4 layers of Daramic as a membrane. However, while Daramic is a low cost material, it is not easily accessible for DIY testing at this moment. For this reason, I wanted to run some tests on materials that are easier to source than Daramic.

I looked for materials around my house that had similar porosity (0.1-5um). I tested several different papers that I had around but none of them worked very well. The porosity of most traditional printer papers is high, with most having 10-20um pore sizes. This means that you need many layers to prevent fast self-discharge from migration of the triiodide across the membrane. Additionally, the papers lost structural integrity quite easily.

Test using a 2m ZnCl2 + 2m NH4Cl + 4m KI electrolyte at a current density of 20mA/cm2. Four layers of matte photo paper were used as separator.

Finally, I stumbled upon matte photo paper as a potential solution. This paper has much lower porosity with <5um pore sizes. Some of these papers might even have pore sizes that are below 1um. This is important for printing photographs, as low pore sizes implies that there is less bleeding of ink when it is applied to the paper, although ink needs to be applied much more slowly to the material (reason why printing with these papers is really slow).

For my initial test, I used 4 layers of matte photo paper. The paper does have a substantially higher ohmic resistance compared to Daramic, so I had to lower the current density to 20mA/cm2. I did 4 cycles of charge/discharge that you can see above (I only did 4 because the lower current meant cycling was quite slow). The CE of 87.54% and EE of 75.72% with a capacity of 33.8 Ah/L shows that photo paper is definitely a good choice for at least the short term cycling of these devices.

On inspection, the photo paper did not show any evident degradation although dendrite penetration happened just as much as it did with the Daramic separator. The separator was also completely black, fully permeated by the catholyte solution which contains triiodide in solution when charged.

Testing a Zinc-Iodide flow battery with a microporous membrane

After concluding my work with the Fe-Mn system, I still wanted to find a system that I could use to build flow batteries. With my new flow battery systems – which I got thanks to my colleague Kirk – I have been able to test different chemistries to come up with the most practical approach for an actual DIY flow battery.

To build such a battery, I first started with a list of requirements:

  • Easy to find chemical supplies.
  • Voltage at least equal to vanadium flow batteries (>1.2V).
  • Energy at least equal to vanadium flow batteries (>25Wh/L).
  • Efficiencies at least equal to vanadium flow batteries (>70% EE).
  • Mildly acidic or basic (no strong acids or bases).
  • Ideally uses a microporous membrane (no need for ion exchange membranes).
  • Long term cycling stability (>1000 charge/discharge cycles).

There aren’t a lot of systems that can comply with all the above requirements. One of the few that seemed plausible was the Zn-I system. In this system, Zn2+ is reduced to Zn metal on the anode and I- is oxidized to I2 (which quickly gets converted to I3) in the cathode. The Zn can be plated with a good density, often at more than 100mAh/cm2 of cathode when using carbon felts.

Results of a 1m ZnBr2 + 1m KCl + 2m KI flow battery using 4 layers of Daramic as the microporous separator, carbon felt anode and cathode with a copper sheet as the anode current collector and a graphite plate as the cathode current collector. The current was 40 mA/cm2

Both the Zn/Zn2+ and I/I3 couples are kinetically very fast – allowing for large currently densities – and have large energy densities. In particular, the solubility of their salts is quite large, so achieving solutions with concentrations >6M is not a problem. At 6M, the theoretical capacity of the battery is around 156Ah/L, rivaling even LiFePO4 batteries.

There are two main problems with this system. The first is the formation of Zn dendrites in the anode, which shorts the battery, and the second, the formation of solid I2 in the cathode after a state of charge of around 66%, which interrupts flow through the battery and causes it to fail. This means that the max State Of Charge (SOC), is often limited to 66% even if no dendrites are present (which is a big if).

Luckily some publications already exist showing that we can sacrifice some efficiency to overcome some of these issues. For example, this paper uses a polyolefin microporous membrane to obtain flow batteries that work at very high current densities and which can “cure” from overcharging due to the constant “leak” of iodine into the anode side. Since the same electrolyte is used as catholyte and anolyte, there is no problem with changes in composition as a function of time.

I decided to use this paper as a template and try reproducing their results. I started by making a 2m solution (this is moles of solute per kg of solvent). You can see my results in the first image above. I was able to obtain a charge of 19.3Ah/L since the 2m solution comes out to approximately 1.5M (because the final volume is larger), this comes out to an SOC of 48%. Trying to charge the battery higher ends up generating I2 solid in my cathode and therefore killing the battery (as no flow becomes possible).

My current was also lower (40mA/cm2 vs 80mA/cm2) because my potentiostat overheated at higher densities (which means I need to build some heatsink to test these current levels). Note that I also ran at higher mL/cm2 since my closed loop is exactly 1mL/cm2 while the paper runs at half of this, meaning that their Zn accumulation per electrode area is half. I also used 4 layers of Daramic to match the thickness of their polyolefin membrane (around 900um).

Results of a 1m ZnCl2 + 1m NH4Cl + 2m KI flow battery using 4 layers of Daramic as the microporous separator, carbon felt anode and cathode with a copper sheet as the anode current collector and a graphite plate as the cathode current collector. The current was 40 mA/cm2

The CE and EE values are pretty decent, but loses from crossover of the microporous membrane are evident. These loses have an advantage though as iodine that permeates the membrane dissolves any Zinc dendrites that might be perforating it, so the battery cycling is very stable.

Since I didn’t observe any gain in SOC due to the presence of the Bromide, I tried to reproduce the results using a lower cost mixture of ZnCl2+NH4Cl+KI. This time I also increased the concentration to 4m to see if I would get a proportional gain in the capacity. The results of this test are showed in the second image in this post.

As expected, there is a drop in the CE and the EE of the device (as more crossover can happen at longer charging times) but the capacity of the battery increases proportionately as well. I was able to charge to double the amount, but the discharge capacity did not increase as much, due to the drop in efficiency. This time I was able to get 35.5Ah/L out of the battery, which is an SOC of around 42%. This is however, already higher than a Vanadium flow battery normally offers (~25Ah/L). Note that higher concentrations are not possible in my system with 4 layers of Daramic as dendrites will start fully crossing the separator at a density of around 56mAh/cm2 (while charge density in the last test is 45.45mAh/cm2).

As you can see, the Zn-I system with a microporous membrane is great. It is stable, has decent energy density, supports large currents and does not require an expensive ion exchange membrane. There are certainly potential improvements to increase the available SOC, but even at the current levels, the system would already be useful in practical applications. Higher separator thickness could also increase the CE and EE of the battery system, without much compromise to the ohmic resistance. Small modifications to the separator – such as adding a carbon coating – can also increase its selectivity and reduce the extent of Iodine migration. Note that you do not want to completely stop it as stopping it leads to dendrites shorting the battery.

I also tested some other microporous membranes that could be easier to obtain than Daramic. Interestingly enough, photographic paper – which has a pore size similar to Daramic – works just as well in this system (really cheap as far as membranes go). I will share the results of these tests in a future post.

New and improved DIY potentiostat/galvanostat for battery charge/discharge cycling

A few months ago, the potentiostat I was using for my battery cycling experiments got damaged with some liquid. The previous potentiostat I used was described in this post. Since a few years have passed since this potentiostat came out, I decided to look at the literature and find any other implementations that might have improved on that previous design. This is how I found the Mystat design, which was published in this paper.

Fully assembled MyStat potentiostat

The Mystat design features many important improvements on the previous design. It has a larger potential range from -11 to 11V and it supports current values up to 200 mA. It also doesn’t lose any of its low range current capabilities, still being able to measure currents in the nA range. The Mystat also allows for easy connection of electrodes using a “headphone jack type” adapter and uses an external 15V power supply instead of relying on the computer’s USB supply for all its energy needs.

I used PCBWay for the fabrication and assembly of the device and decided to create a shared project so that anyone interested can have their own manufactured. You can use that link to buy your own fully assembled potentiostat. The total cost for the potentiostat was around 330 USD, including the power supply and cables that need to be bought separately (links for that are included in the shared project description).

Above you can see the cyclic voltammetry of a 1K ohm resistance using the new MyStat potentiostat.

The best thing is that the potentiostat uses the exact same python software as my old potentiostat, so I was able to use a lot of the coding modifications I had done to improve my battery characterization. You can obtain my updated code here, which allows you to do battery cycling experiments charging to a final capacity or to a final potential and includes an additional display for the uAh measurement in the current cycle.

Problems with the Fe-EDDHA|Mn-EDTA flow battery system

I wrote a blog post about some of my first tests of the electrochemistry of the Fe-EDDHA|Mn-EDTA system and how I planned to build a flow battery using this chemistry. I have since been able to setup a flow battery and perform some initial experiments. So far, there are several fundamental problems with the above chemistry that strongly affects its viability. I will discuss these problems below.

Fe-EDDHA solubility and pH. The Fe-EDDHA is more soluble at more basic pH values and its solubility at neutral pH is limited to perhaps only 0.1-0.2M. Furthermore, there are several less soluble impurities present – different isomers of the EDDHA – which have to be filtered before the cell is operated. Additionally, the pH changes substantially upon charge/discharge, which causes problems with the Mn-EDTA side (you’ll see later on).

Fe-EDDHA anolyte (A2) next to Mn-EDTA catholyte (C2) solutions, both at around 0.2M.

Difficulty in working with Fe-EDDHA solutions. The solutions of Fe-EDDHA are very deeply red, basically black at the concentrations we’re working with here. This means that it is very hard to tell if there are any insoluble substances in the solution. It is also hard to tell the charge state of the solution, as it will only slowly become more transparent as the Fe-EDDHA is reduced to its Fe(II) form.

Fe(II)-EDDHA sensibility to oxygen. Once the Fe-EDDHA is reduced in the anode, the liquid becomes incredibly sensitive to any oxygen dissolved in the solution, which will quickly oxidize the Fe(II)-EDDHA back to Fe(III)-EDDHA. Given that I currently have no means of purging my system of oxygen, this makes it impossible for me to run flow batteries using Fe-EDDHA right now in a reliable way.

Anolyte and catholyte of the system. You can observe the red color of Mn(III)-EDTA appear in the catholyte (C2) as the flow battery is charged.

Mn-EDTA stability at higher pH. The oxidized form of Mn-EDTA, which is Mn(III)-EDTA, is red at a pH below 6.5 and yellow at a pH above 6.5 (you can see that yellow in a previous post). However, the yellow molecule is extremely unstable, therefore it is very important for pH to remain below 6.5. However, at a pH below 6.5 the solubility of the Fe-EDDHA drops substantially. The differences in pH between charge and discharge means that both systems are virtually incompatible, because as the Fe-EDDHA charges it induces changes of pH that very negatively affect the decomposition of the Mn-EDTA catholyte.

For all the above reasons, I decided I no longer will explore the Fe-EDDHA|Mn-EDTA system, as I currently don’t have the technical setup that is necessary to properly study it. Furthermore, the low solubility, difficulty of working with the Fe-EDDHA solutions and the problems with pH, make this system more complicated than I was hoping for, especially in a system I hoped to be simple for DIY.

I will continue to study flow batteries based on Mn chemistries, including Mn-EDTA (as these are very interesting), but I will likely change the anolyte to something that is more aligned with my current setup. Most likely I will have to compromise and use a potentially higher density, low cost anolyte, that will involve plating a metal on the carbon felt anode, like Zn. Hopefully in this manner I’ll be able to find a low cost setup to bring a DIY solution with a capacity at least in the 10-20Ah/L range.