Roberto finished a new molecular candidate library for our friends at Johns Hopkins University. Great job, Roberto!

BREAKING NEWS: Alan – the big enchilada of the Harvard Clean Energy Project – just got promoted to tenured Associate Professor. Horray and Congrats!

Small correction: Alan actually was promoted to Full Professor :). No idea whether he gets a pay raise.

Congratulations, Alan! It sounds like there's all sorts of good news coming out of the CEP camp.

Thanks to the clean energy team for the frequent updates.

Hello everybody,

It is time again for another progress report on the Clean Energy Project - our friends at IBM-WCG keep reminding us that we are overdue :). Sorry about that, but it’s sometimes hard to make time in our busy research schedule for things that don’t have a hard deadline or threaten to become a bottleneck. Anyways, we now finally get around to giving you an update and hope you’ll enjoy it.

Let’s start with some cool stuff on the team: The most momentous event since our last report in November (aside from blizzard Nemo) is that Alan got tenure and is now a Full Professor in the Department of Chemistry and Chemical Biology here at Harvard University. Amongst other things he got his promotion for the contributions the Clean Energy Project has made in the fields of computational materials science, materials informatics, virtual high-throughput screening, big data, and organic electronics.

In addition, Sule got an Assistant Professorship in Ankara (Turkey), to where she will relocate in the fall (but there is hope that she’ll remain a vital part of the project). A number of awards and fellowships were bestowed on various members of the CEP team (e.g., Aidan won a Rhodes Scholarship, Johannes the IBM-Loewdin and the ACS PHYS Postdoc Award). We wrote several grant proposals for money and cluster time (e.g., from the NSF DMREF “Designing Materials to Revolutionize and Engineer our Future” program), and we partnered up with HGST who sponsor us with hard drive donations. CEP work was presented on several conferences, webcasts, seminars, and talks over the last half year, and we wrote a book chapter on the CEP materials informatics work. This was spearheaded by Carlos, the text is titled “Towards materials informatics for organic photovoltaics”, and it will appear in “Informatics for Materials Science and Engineering”.

Now to the progress on the research front: The WCG returns have recovered nicely after a weak autumn. These days we average between 18 and 20 y/d and we have passed the mark of 15,000 years of harvested CPU time. By now we have performed 150 million density functional theory calculations on 2.3 million compounds with 24 million conformers, accumulating 400TB of data. This is unprecedented – plain and simple! We have upgraded our storage capacity to over 700TB by building and launching Jabba 4, 5, and 6. Thanks again to HGST for their support and Harvard FAS Research Computing for their help. All the Jabbas are also nicely racked now. The project progress is monitored by a new code, which automatically generates statistical analyses and plots, milestones, a timeline, and other information of interest on-the-fly. We have renewed our efforts to revamp our CEP webpage, and the new statistics code will be deployed here once it replaces the old project homepage.

The CEPDB backend was the main focus of our efforts during the last months. We have refactored, improved, and expanded its capability, both for faster queries and to get it ready for its role as an information, analysis, and cheminformatics hub. CEPDB is much more mature and powerful now, and it also contains additional data fields such as INChI, IUPAC, and common names, conversion to canonical SMILES, fingerprinting, a link to matplotlib for graphics, and a number of performance improvements. We gratefully acknowledge the input from our friends at Kitware and ChemAxon. The work on the CEPDB user frontend has also made some really nice progress and we hope to release it soon.

We have further intensified our collaborations with Stanford, Johns Hopkins, and Linz University. We are providing our partners with lists of promising compounds, e.g., for high power conversion efficiency (PCE) according to the Scharber model or for low bandgap semiconductors. According to the Scharber analysis, 35,000 (i.e., 1.5%) of the screened compounds have the potential to achieve a PCE of over 10%, and 1,000 (i.e., 0.04%) of over 11%. This shows how hard it is to design high-PCE materials but at the same time it underlines that a virtual high-throughput screening can still lead to a useful number of candidates.

Our friends at Stanford pointed out to us that one of our top ten molecules (in slightly augmented form) is already known as a high-performance compound. This independent confirmation was extremely encouraging and underscored the value of the CEP results. Our friends at Haverford College have made some progress in synthesizing a compound suggested by us. Our friends at Johns Hopkins University have been using our DFT results to benchmark a new theoretical method (an SSH type model) they’ve been developing. We further provide data dumps to ChemAxon, Kitware, Wolfram Alpha, and NREL. We are currently producing a number of 2nd generation molecular libraries for our different partners.

The development team at Q-Chem Inc was provided with feedback about the performance of some employed algorithms which will hopefully help them to improve the program package. In the meantime, our friends at IBM are working on the review of a fresh revision of the Q-Chem program package for the grid.

We have introduced a new averaging scheme designed to make our results more robust. It takes the different computational results which were independently calibrated and averages over them. A preliminary statistical analysis (still using our old OE11 calibration) suggests that the error bars are surprisingly low: For the HOMO energy we get an average MAD value of 0.07eV and RMSD of 0.09eV. For the LUMO we have an average MAD of 0.09eV and RMSD of 0.12eV. This looks very promising but we need to investigate this more carefully.

In the coming months we will concentrate on expanding our data mining, analysis, and modeling tools, releasing CEPDB as part of the White House Materials Genome Initiative, writing up a number of papers, and launching new generations of workunits. There is still so much more exciting work to be done, and we are looking forward to going this together with you. Thanks so much for all your generous support, hard work, and enthusiasm – you guys and gals rock!!! CEP would not be possible without you. CRUNCH ON!

Best wishes from

Your Harvard Clean Energy Project team