VoxSolaris: The Voice of the Sun
Sodium Sulphur Flow Battery

Anatomy of a sodium sulphur cell

The sodium suplhur (NaS) battery was first seriously investigated by the Ford Motor Company back in the the 1960s. It is easy to see why Ford embarked upon the investigation but it is just as easy to see why in the end, they chose not to adopt it. NaS is a form of molten salt battery. The cells have electrodes of liquid sodium and liquid sulphur while the electrolyte is the ceramic, soduim beta alumina with an approximate composition 11 parts aluminum oxide to 1 part sodium oxide. For the cell to work, both electrodes must be liquid and the electrolyte must be at a temperature at which it can conduct ions. Sodium melts at 98°C and sulphur at 113°C but the electrolyte does not conduct sodium ions unless the temperature is around 350°C. This then is the operating temperature of the battery. The electrolyte has melting point is 2100°C and so remains solid. The cells are comonly constructed as a pair of concentric tubes in which the inner tube made of the electrolyte contains the sodium while the outer tube made of a metal such as aluminum, contains the sulphur. During discharge, electrons are striped from the sodium atoms and flow from the sodium anode through the external load to the sulphur cathode. The positively charged sodium ions move through the electrolyte where they react with the sulphur and the electrons to produce sodium polysulphide. Recharging strips electrons from the sodium polysulphide turning it back into sulphur and sodium ions which migrate across the electrolyte into the sodium where they are reunited with their missing electrons to form sodium atoms. The cell is completely enclosed and does not give off any gases.

The Lure of NaS

NaS cells have a columbic efficiency of virtually 100 percent and a charge/discharge voltage ratio of 0.85 so overall electrical efficiency is 85 percent. Because of the chemistry of the alkali metals of which sodium is a member, the energy density can be high, of the order of 200Wh/Kg. Back in the 1960s such energy density, very comparable to what Tesla Motors achieves today, was completely unheard of. NaS also offered a perfectly reasonable power density of 100 to 200W/Kg. Bearing in mind that at the time, although it was a horrible delusion, there was the notion that nuclear energy would one day render electricity 'too cheap to meter'. And while an operating temperature of 350°C might seem insane for a protable wireless as the radio was then known, it isn't insane for a would be replacement to an internal combution engine. And to cap it all the materials are very cheap and abundant. All in all a compelling lure. Can you imagine how modern life might be today had Ford been able to make NaS practical? So what went wrong and could those problems be fixed today?

The Beta Alumina Challenge

The most troublesome aspect of NaS is the electrolyte. Beta alumina is difficult to work with and can be expensive to engineer to the high quality needed in a practical cell. And the electrodes offer little solace. Both are very reactive and therefore corrosive and are all the more so in an elevated temperature. Sodium ignites on contact with air so poses a serious fire hazard so hermetic sealing is critial. That isn't easy to achieve when you consider the thermal expansion the cell will undergo as it is taken up to its operating temperature. The required manufacturing and quality assurance standards are inherently high and so therefore is the price.

Another limitation bestowed by the electrolyte is poor shock resistance. Beta alumina is very brittle. And it suffers from thermal cycling problems. Given the high operating temperature, a practical device on a scale suitable for a car, would experience considerable heat loss. But if the battery were allowed to cool off while not in use, it would have to be reheated before the car could be driven. Not only would that take an innordinate period of time but the electrolyte would fail very quickly. And the consequence of an electroyte breach are not just that the cell becomes electrically useless. If fissures in the electrolyte allow even tiny quantities of the electrodes to mix, an exotherm will result. If the exotherm serves to widen the fissure as is likely, it could run away and lead rapidly to cell explosion. This obviously threatens the integrity of neighboring cells in the battery and ultimately the integrity of the outer casing of the battery. As a minimum the result is a high intensity fire if not an actual explosion.

All this points away from NaS as a viable transport battery but for stationary applications such as grid storage, NaS fares better. Grid storage scales are orders of magnitude larger than transport scales and this cuts thermal losses per unit of energy stored considerably. Furthermore, since neither space or weight are an issue in most stationary applications, heavy duty insulation can cut thermal losses a lot further. But again the beta alumina frustrates, by having a limited lifespan, typically of the order of 2 to 5 years. In a process largely unrelated to use, the electolyte deteriorates over time, slowly becoming a conductor of electrons therby lowering columbic efficiency and increasing self discharge. Ultimately this renders the cells useless. So while the materials might cost less than $10 per KWhr, high manufacturing costs negate this and when you plug the lifepsan into the all important total cost of owership equation, you don't get the number you need to jump out of the bath and shout Eureka!

Across the industry, current research into NaS is rather limited but is not surprisingly, focused on the electrolyte deterioration issue. One big driver for this research is that beta alumina is used as an electrolyte in a wide range of electro-chemical applications and to a greater or lesser extent they all have the same issue. In the particular case of NaS, the deterioration occurs because the aluminum oxide is gradually reduced by the passing sodium ions to aluminum and because sodium metal accumulates in the electrolyte fissures. Where this research might take us is anyone's guess but our sense is that there is essentially nothing that can be done about the reduction of aluminum oxide by passing ions. At best, the process could be slowed down and there are companies who claim they have NaS cells with lifespans of 15 years. One market leader stated "we have found that deterioration of beta alumina is virtually eliminated by high quality of raw materials and high standards of production". It is a plausible claim and certainly there are things that can be done to improve the construction of beta alumina extrusions beyond what the standard method of sintering can achieve. The question is can it be done sufficiently cheaply to serve beyond a few niche markets?

The Zebra Battery

The Zebra battery is a close relative of the NaS battery but is considerably safer. It has a sodium anode and a Nickel chloride cathode. The electrolyte is the same beta alumina used in the NaS cell and accordingly has a similar lifespan. The zebra runs at 270 - 300°C and at these temperatures the nickel chloride electrode is solid. So between it and the beta alumina, an intermeadiate molten salt electrolyte of sodium tetrachloroaluminate is deployed. Like NaS, the Zebra has an electrical efficiency of around 85% with near 100% columbic efficiency and very low self discharge (because of beta alumina). It acheives lower but resonable power densities of around 100Wh/Kg and 150W/Kg respectively and has a cell voltage of 2.58 volts instead of the 2.1 of NaS.

The Zebra having one electrode as a solid does not offer quite as many charge cycles as sodium sulphur due to the need of the solid electrode to regrow to something akin to its original shape. But the usual solid electrode problem of dendrite growth is strongly mitigated by the high temperature. As a material, nickel chloride is much more expensive than sulphur but this is more than outweighed by an inherently safer chemistry and correspondingly lower manufacturing and quality assurance costs. The onset of fissures in the beta alumina will result in much less violent exotherms because the concentration of nickel chloride in the sodium tetrachloroaluminate is low compared to the equivelent situation in the sodium sulphur cell. Although the beta alumina is just as brittle, because of the inherently better safty the Zebra can actually be used for transport and has been successfully deployed in electric buses. Had it been around in the 1960s it could have been a game changer but today its use in transport is seriously challenged by lithium ion which has room temperature operation, comparable or better lifetime and is of comparable cost.

The Wrong Format

The general construction of a NaS or Zebra battery is shown in the diagram below. The cells are placed in a cage supported by spacers and the whole ensemble is sealed in an evacuated box. It makes sense that there are multiple cells since most applications need more than a couple of volts. And it makes sense to put those cells close together from a thermal efficiency standpoint. But it doesn't really make sense from either a reliability or a safety standpoint, particularly for NaS. Put simply, if you have a problem with one cell you have a problem with them all and the more cells you have the more likely it is that you will have a problem. A chain is a strong as its weakest link but a box of fireworks is as dodgy as its dodgiest firework times the number of fireworks!

There is a way of dealing with an exotherm relatively safely. The cells can be constructed such that in the event of a rise in pressure, the outer casing of the cell (the part containing the soodium polysulphide) would drop away from the beta alumina tube and the battery shuts down. This is similar idea to the 'failsafe' mechanisum of nuclear reactors whereby fuel rods fall out of an overheating reactor core. Such arrangements can be very effective at limiting exotherms within cells and stopping them from triggering exotherms in adjacent cells. But things are never that simple in practice. If exotherm detection is too sensitive the outer casings will drop needlessly. Not sensitive enough and you could still have an explosion. And the calibration isn't straightforward at 350°C either.

The Vox-Solaris Recommendation: NaS as a Flow Battery

Of course all of the above does not mean NaS cells are non viable. Far from it. There are numerous grid storage installations based on NaS cells. What all of the above does mean is that basing a NaS battery on NaS cells is not particularly safe and is a very expensive way of exploiting what is the cheapest chemistry on offer. Redesigning NaS as a flow battery, solves ALL of the above issues for stationary applications. And although lithium ion and varients thereof have dramatically reduced the need for a NaS transport battery, the flow design is so much safer it could be used for transport purposes.

The construction would comprise a central reactor and three resevoirs, one of sodium, one of sulphur and one for sodium polysulphide. The central reactor consists of an array of thin circular cells. The cells are essentially the same as standard NaS cells and comprise the same beta alumina electrolyte except that at any given time, only sufficient active ingredients are in contact with the electrolyte to make the cell work. During discharge sodium and sulphur are pumped from their respective resevoirs into the reactor and the resulting sodium polysulphide is pumped to the third resevoir. During recharge the flows are reversed. Sodium polysulphide is pumped to the reactor where it is split by electrolysis back into sodium and sulphur. This arragement avoids the mixing of sulphur and sodium polysulphide on the cathode side of a normal NaS cell and thus avoids any degradation of voltage. The pumping action would be provided more easily by pumping an argon propellant rather than pumping the active materials directly.

Low Thermal Loss, Low Exotherm Risk and High Specific Energy and Power

There are two ways in which the flow design can mitigate thermal losses. Firstly, in any sizable installation, thermal loss can be reduced because there is no need to keep the reseviors at the same temperature as the reactor. Keeping the resevoirs just hot enough to keep their contents liquid rather than at the reactor operating temperature of 350°C, massively cuts thermal losses. Materials entering the reactor need to be heated to the reactor temperature but much of this will be achieved by a heat exchanger cooling materials leaving the reactor. Secondly, thermal losses can be mitigated by shape. The resevoirs would be constructed as spheres as this is the shape that has the highest ratio of volume to surface area. It is also the shape that requires the thinest walls to protect a vacuum, helpful for insulation, from the crushing atmospheric pressure. We gain in three ways. Less material is needed to contain the liguid electrodes, less material is needed for the outer shell and the smaller surface area directly reduces heat loses as these are proportional to surface area. The reactor housing can also be made spherical although this would only be for deployment of a vacumm insulator. Note that while the reactor is the hottest part with the greatest heat loss, it is usually much smaller than the other parts so heat loss is less of an issue. In any event during operation the reactor generates heat that must be gotton rid off. This is done largely by material outflows but clearly has an impact on the extent to which the reactor requires insulation.

The reactor will comprise multiple cells that must allow for liquids to flow under pressure across both sides of the electrolyte whilst maximizing the ratio of surface area to volume. The cells will be individually controlled by shutoff valves which would remain open for the lifetime of the cell, close when cell misbehaviour occurs and then remain closed until the reactor is shut down for maintenence. The mechanism for this has to reliably operate at 350°C and so should be simple. An exotherm will create back pressure against material being pumped into the cell and this could be used to pull a pin that would then allow the valve to fall shut under its own weight. The cells would be surrounded by walls to protect adjacent cells while exotherms would be tolerated by allowing the errant cell to explode downwards into a common sump. The low volume of active material in the cell contributes considerably to the ability to tolerate exotherms. The pumps to drive material around the reactor will be external to the reactor and so not exposed to the temperature.

Although not an important factor for stationary applications, the flow design offers very high specific energy. The energy density of NaS has a theoretical limit of 790Wh/Kg assuming the cell walls, the beta alumina and the outer case can be ignored! With the bulk of the battery being store in spheres that minimizes non active construction material, the flow design can realize more of the energy density offered by the NaS chemistry. We have fleshed out designs that would appear to give better than 300Wh/Kg. And we note that it is easy to give the reactor cells pretty high electrolyte surface areas and thus high power output and fast charge acceptance rates. It all makes one wonder if NaS could not have brought us the viable electric car back in the 1960s after all.