MIT’s associate professor of mechanical engineering Tonio Buonassisi, former MIT professor Daniel Nocera (now at Harvard University), MIT postdoc Mark Winkler (now at IBM) and former MIT graduate student Casandra Cox (now at Harvard) have published a paper that is a new analysis laying out a roadmap for a research program to improve the efficiency of the “artificial leaf” systems, and could quickly lead to the production of a practical, inexpensive and commercially viable prototype.
Artificial leaf systems would use sunlight to produce a storable fuel, such as hydrogen, instead of immediate use electricity. This fuel could then be used on demand to generate electricity through a fuel cell or other device. The artificial leaf process would liberate solar energy for use when the sun isn’t shining, and open up a huge range of potential new applications.
Nocera and his team produced a proof of concept artificial leaf in 2011. The small device when placed in a container of water and exposed to sunlight would produce bubbles of hydrogen and oxygen. The device combines two technologies: a standard silicon solar cell, which converts sunlight into electricity, and chemical catalysts applied to each side of the cell. Together, these would create an electrochemical device that uses an electric current to split atoms of hydrogen and oxygen from the water molecules surrounding them.
The goal is simply to produce an inexpensive, self-contained system that could be built from abundant materials.
Cox explains the purpose of the paper, “What’s significant is that this paper really describes all this technology that is known, and what to expect if we put it all together. It points out all the challenges, and then you can experimentally address each challenge separately.”
The Nocera team’s original demonstration leaf from 2011 had low efficiencies, converting less than 4.7 percent of sunlight into fuel, Buonassisi says. But the team’s new analysis shows that efficiencies of 16 percent or more should now be possible using single-bandgap semiconductors, such as crystalline silicon.
Winkler describes the effect of the modeling and the discipline employed to set the model up with, “We were surprised, actually. You’ve just got to question the conventional wisdom sometimes.” Conventional wisdom held that the characteristics of silicon solar cells would severely limit their effectiveness in splitting water, but that turned out not to be the case.
The key to obtaining high solar-to-fuel efficiencies is to combine the right solar cells and catalyst – a matchmaking activity best guided by making a roadmap. The approach presented by the team allows for each component of the artificial leaf to be tested individually, then combined.
Modeling broke the artificial leaf concept into parts. Buonassisi explained the voltage produced by a standard silicon solar cell, about 0.7 volts, is insufficient to power the water-splitting reaction, which needs more than 1.2 volts. One solution is to pair multiple solar cells in series. While this leads to some losses at the interface between the cells, it is a promising direction for the research.
An additional source of inefficiency is the water itself as it is part of the pathway that the electrons must traverse to complete the electrical circuit and has resistance to the electrons, Buonassisi said. So another way to improve efficiency would be to lower that resistance, perhaps by reducing the distance that ions must travel through the liquid.
Cox explains the water circumstances with, “The (water) solution resistance is challenging. But, there are “some tricks” that might help to reduce that resistance, such as reducing the distance between the two sides of the reaction by using interleaved plates.”
“In our simulations, we have a framework to determine the limits of efficiency” that are possible with such a system, Buonassisi says. For a system based on conventional silicon solar cells, he says, that limit is about 16 percent; for gallium arsenide cells, a widely touted alternative, the limit rises to 18 percent.
Models often set observers on edge with grand projections and questionable assumptions. This MIT work is far superior to those. Here the assumptions come from a proved up concept device and the knowledge gained from the experiments. The model is more of a mapping of working through the assorted technologies and sciences needed to optimize what is already known, and to arrive at the projected value of such work.
Buonassisi explains models to determine the theoretical limits of a given system often lead researchers to pursue the development of new systems that approach those limits. “It’s usually from these kinds of models that someone gets the courage to go ahead and make the improvements,” he said.
“Some of the most impactful papers are ones that identify a performance limit,” Buonassisi says. But, he adds, there’s a “dose of humility” in looking back at some earlier projections for the limits of solar-cell efficiency: Some of those predicted “limits” have already been exceeded, he says. “We don’t always get it right, but such an analysis “lays a roadmap for development and identifies a few ‘levers’ that can be worked on.”
Sixteen percent is a very different prospect from 4.7%. A 300% improvement is well worth the chase. And on the way this kind of thinking may well show there is even more potential in other ideas.
It’s a technique others would do well to notice and emulate.