Big Blue donates cloud to unborn solar cells
Free cycles for plastic sun capture
The World Community Grid, a network of donated processing capacity established by IBM for do-gooder supercomputing projects, said today that it would be donating time to a bunch of Harvard University researchers who are looking for organic materials from which to make flexible and cheaper solar cells. Plastic solar cells, to be colloquial about it.
Because IBM can't say a sentence these days without the word "cloud" in it, it also says it will be encouraging customers who have bought its own cloud computing resources to donate some cycles to the Harvard scientists and hinted it would be donating some of the excess, unsold cycles in its so-called Cloud Computing Centers to the cause.
Most of the solar cells manufactured these days are based on silicon technology, which is great for making relatively expensive yet small, intricate, and sophisticated electronics, such as the chips inside your computer of choice. But when you want to wallpaper a field or a roof with such silicon chips to make solar cells, it gets to be very pricey indeed. Harvard wants to discover the organic molecules that can be mixed to capture sunlight and turn it into electricity without having to resort to the whole chlorophyll thing that plants mastered billions of years ago to store sunlight in a battery we know as sugar. But using a conventional supercomputer (with a few teraflops of computing power, presumably) would take 100 days to virtually test an organic molecule for its suitableness for the flexible solar cell.
Harvard wants to look at tens of thousands of molecules and at the scale of supercomputer that the university can bring to bear, this would require 22 years of computation. That is not going to fix our energy problem.
However, by giving the Harvard researchers access to the World Community Grid, IBM thinks it can help cut that time down by a factor of eleven, helping them isolate promising plastic compounds in around two years. The grid is comprised of some 413,000 volunteers in over 200 countries, who can have multiple machines and who have donated processing cycles for various projects.
IBM says that the grid did five years' worth of computational research in just six months and that a project at the University of Washington to help breed better rice has done 10 million transactions and used 9,000 years of compute time. (This has to be compared to a PC or workstation running the same workload, but IBM is not precise). There are other projects to fight cancer and predict the climate in Africa that are in various stages on the grid.
The goal of the Harvard solar cell research is to not just make cheaper solar cells, but more efficient ones. IBM reckons that a silicon-based solar cell can only capture about 20 per cent of the solar energy hitting its surface and convert it to DC electricity. The department of chemistry and chemical biology thinks it can do better than the physics experts. The best solar cells are pushing 40 per cent under maximal sunlight conditions, so this is a pretty tall order. Even if plastic solar cells were not more efficient, if they were a lot cheaper to make and easier to work with, that would be a huge improvement. Then, we can wallpaper the desert by converting oil into plastic solar cells. What could be sweeter than that? ®
Sponsored: Hyper-scale data management