A Semiconductor Approach To Desalination

By Bernard Murphy and Jim Hogan

We’re not offering breaking news when we observe that the semiconductor industry is in flux. Major consolidations and lack of funding for startups point to an industry that, outside China, is maturing and seems to have lost the recipe for rapid growth. Apologists will argue that analog or MEMS or some other domains are still strong, but this misses the point that the industry as a whole is slowing down. At least part of the problem is arguably that semiconductors are stuck at a fairly low leverage point of a relatively limited number of supply chains (phones, tablets, cars, IoT). One way to break out of this trap would be to enable radically different supply chains.

We believe that one such opportunity is related to water purification. Access to sufficient and safe drinkable water has become a worldwide problem. Here in California after one of our worst droughts on record, we are acutely aware how fragile and finite our water supply can be. Finding ways to alleviate this shortage have become some of the defining problems of our time. If the semiconductor industry could build solutions around this area, we could enable a completely new and rapidly growing supply chain around a much less discretionary need that the next smartphone upgrade.

The current commercial standard for purification is reverse osmosis (RO). This process pushes seawater at fairly high pressure through a filter which blocks hydrated salt ions but allows water through. RO consumes ~3kWh/m³, a sufficiently high energy that if all domestic water usage in the United States were to come from this source, one source estimates American energy consumption would increase by 10%.

A much less energy-intensive approach is based on nanoporous graphene filters. In graphene, a single-atom thick form of crystalline carbon, atoms arrange themselves in a well known hexagonal structure. A key advantage of graphene in this context is its strength. By adding a lot of very small holes or pores to the sheet, you get a nanoporous graphene filter. Pores should be around 0.5nm, which is large enough to let water molecules through but small enough to block salt ions. This filter doesn’t require the high pressures required by RO on the seawater side to achieve high levels of throughput and is therefore potentially much more efficient in both energy and capital cost.

Nanoporous graphene filter allows water to pass through but blocks salt ions

In theory, this approach should allow for up to three orders of magnitude higher throughput than RO for the same energy expenditure. The problem is this has only been demonstrated in theory. The theory is apparently solid and expected to translate well to practice, but you can’t build a desalination plant on theory.

Which brings us to how semiconductor fabrication techniques might be able to help build these novel filters. First we need to build graphene sheets. There are many possible methods, some of which are quite familiar in semiconductor techniques, such as epitaxial growth on copper foil in a low-pressure methane environment. This particular approach naturally leads to a single atomic layer of graphene because the copper catalyzes growth and the initial graphene layer blocks further catalysis. After the graphene has been created, you would need to etch away the copper to have a useful filter, but might leave a support grid to strengthen the sheet.

So far nothing here seems too farfetched for any standard fab. But punching the holes will require some innovation. Remember these have to be ~0.5nm, a tad smaller than the most aggressive feature sizes considered today. Still, pore creation has been demonstrated already in top-down approaches (oxidative etching under ultraviolet light) and bottom-up approaches (pattern copper with non-catalyzing dots of aluminum oxide before growing the graphene layer, then etch away those dots).

You could argue that the chemistry is all very well, but you still need to get down to sub-nm feature sizes. True, but remember this is a single-layer objective—no need to worry about multiple layers, planarization or alignment—and the pattern will be extremely simple, for example a grid of dots. In fact, regularity of spacing is not important either. Only the density of pores and well-controlled pore size is ultimately important. One approach might be interference lithography (IntL). IntL could possibly be maskless, using an X-ray laser, or might employ a conventional X-ray source with a relatively simple mask. And what is normally considered a major drawback for IntL, that it can only pattern arrayed structures, is exactly what we need in this context.

If all of this can be made to work, then fabs could be pumping out (pun intended) wafer-scale filters at an incredible rate to satisfy the needs of desalination plants all over the world. And filters will unavoidably have a finite lifetime so demand for replacement units will continue indefinitely. Of course the market will stabilize eventually, as all markets do but meantime this could be an exciting high-growth opportunity for enterprising foundries.