PaulC61
commented
1 year ago Hey, here's a quick into to dyes for solar cells:
In this article they present two automatically generated databases for both dye sensitized solar cells (DSCs) and perovskite solar cells (PSCs), with a total data entry of 660,881 representing 57,678 photovoltaic devices. I've been focused on modelling the DSCs which covers 41,680 devices.
The paper above details how these databases are available in MongoDB, JSON formats and downloadable via figshare.
Below is a paper describing the dye sensitized solar cell database (DSSCDB) which is currently not available at the URL given but I am planning to clean it up a bit and re-release it (with permission) somehow . This database was created manually and standardized by an expert in the field, so it's unsurprising that it only contains around 4,000 device entries. But look it was the first of it's kind and was used as a benchmark for the previous database.
The DSSCDB is composed of metal and metal-free dyes. The current SMILES notation across the different dyes is standard, not annotated with anything but people who are trying to model these dyes, for either the macro properties of a solar cell device (such as Photo Conversion Efficiency (PCE)) or the molecule spectrum absorption/emission maxima, take these SMILES and either fragment them into fingerprints or calculate physico-chemical properties directly from the SMILES.
Finally, here is a study from our lab where we used pharmacophore models to screen for new dye organic, metal-free compounds using the DSSCDB. The rationale was that there is an accepted organic dye architecture, D-pi-A, which has been shown to facilitate photo-induced charge separation and a pharmacophore model should encapsulate the elements of this generalized structure. We're looking now to expand this methodology to the larger dataset I first introduced.
https://www.nature.com/articles/s41524-022-00823-6
In this push–pull structure the highest occupied molecular orbitals (HOMO) are mainly localized on the donor part of the dye, while the lowest unoccupied molecular orbitals (LUMO) are mainly localized on the acceptor part of the molecule. This spatial orientation not only favors the electron injection but also slows down recombination between electrons in the conduction band of the semiconductor and oxidized molecules. This architecture also facilitates wide absorption spectrum.
The tricky part about modelling the macro property of PCE in solar cell devices is that the dye molecule is only the beginning. These things are multi-component, messy sandwiches. Consider just the TiO2 (or other metal oxide) semiconductor layer, where it can come in various nanoparticle sizes, film thicknesses and scattering layers (if even correctly reported). This is only the beginning as I haven't even touched on co-sensitizers.. What's more, I've found out recently that even the experimental conditions under which PCE is measured has some influence on the final value reported.
So as far as GlobalChem is concerned, a good place to start may be navigating the dye chemical space alone, where what would deem impactful in terms of only the dye itself are the broadest possible UV/vis absorption maxima range.
In regards to why people care about DSCs: Their efficiency isn't very high as of now but they can be used for transparent solar cells (solar windows), they can be fabricated on flexible substrates or fibers (wearables), and are cheap to produce at scale. There's a lot more to talk about here so please ask questions but this serves as an intro.
Sulstice
Let's understand some stuff about solar cells. Send a link to the paper here.