TL;DR
This paper introduces the VERDE materials DB, a novel, open-access database containing ground and excited-state properties of organic molecules, crucial for designing next-generation electronic materials. By integrating computational data with machine learning, VERDE aims to accelerate the discovery of new materials for applications like organic solar cells and redox flow batteries, as demonstrated by a case study on photoaffinity labels.
What Problem Does This Paper Address?
The discovery of new organic materials for applications like organic photovoltaics (OPVs) and energy storage is hampered by the vast chemical space of possible molecules and the high cost of experimental characterization. While existing databases offer ground-state properties, they often lack crucial excited-state information (like S1, T1 states) that dictates photophysical behavior and photochemical reactivity. This gap makes it difficult to rationally design materials that function effectively under light or in non-equilibrium states, which are common in organic electronic devices and chemical reactions. The authors aim to bridge this gap by providing a comprehensive, accessible database of excited-state properties.
Methods
The VERDE materials DB is built upon a high-throughput virtual screening (HTVS) workflow. This workflow begins with generating molecular structures using RDKit and performing a low-energy conformational search. Subsequent computational steps involve semi-empirical optimizations (PM7 and RM1-D) to obtain geometries that closely approximate DFT results. Finally, M06/6-31+G(d,p) calculations are used for single-point energy calculations on the lowest-energy conformer, followed by DFT optimizations and frequency calculations for ground (S0), excited singlet (S1), and triplet (T1) states. Redox potentials are derived from radical cation calculations. A key aspect is the calculation of 0-0 transition energies (E0-0), which are crucial for understanding photophysical properties. The data is then integrated with the Materials Data Facility (MDF) for discoverability and accessibility.
Datasets
The VERDE materials DB itself is a newly introduced dataset. It currently contains data for over 1,500 organic molecules, with more than 13,000 density functional theory (DFT) calculations performed. This dataset focuses on π-conjugated organic molecules relevant to green chemistry, organic solar cells, and organic redox flow batteries. The paper also references existing datasets like GDB-13, QM7, QM7b, QM9, and the Harvard Clean Energy Project (CEP) dataset, highlighting how VERDE materials DB complements them by including excited-state properties.
Key Insights and Ideas
- Excited-state data is critical: The paper strongly emphasizes that understanding excited-state properties is not just beneficial but essential for designing functional organic electronic materials and for comprehending photochemical reactions.
- Bridging the computational-experimental gap: By providing readily accessible, computationally derived excited-state data, VERDE materials DB democratizes access to information that was previously difficult and expensive to obtain experimentally.
- Machine learning enablement: The database is designed to serve as a rich resource for training machine learning models, enabling the discovery of new materials and structure-property relationships that might be missed by traditional screening methods.
- Focus on light-responsive materials: The selection of molecules and the inclusion of photophysical properties directly address the need for materials in solar energy conversion, photocatalysis, and other light-driven applications.
Key Results
- The VERDE materials DB has been established, containing optimized geometries and properties for 1,500 organic molecules, including ground (S0) and excited states (S1, T1), redox potentials, dipole moments, excitation energies, and 0-0 transition energies.
- A case study on diazirines identified four promising candidates for photoaffinity labels with predicted high photostability, based on their calculated E0-0 values. These candidates are anticipated to be at least as stable as a known benchmark compound.
- The database is integrated with the Materials Data Facility (MDF), making it searchable and accessible via advanced search capabilities.
Limitations and Future Work
- Computational cost: While DFT calculations are more accessible than experiments, they still represent a significant computational cost, limiting the current size of the database.
- Functional dependence: The authors acknowledge that vertical excitation energies can be functional-dependent and that their choice of reporting E0-0 values is a compromise.
- Experimental validation: While the diazirine case study is promising, experimental verification of the predicted properties and photostabilities is noted as ongoing work.
- Future expansion: The authors plan to continue growing the database through internal projects and collaborations. They also aim to make VERDE materials DB more interactive by allowing users to upload structures for computation and by integrating with other data services for machine learning model repositories and benchmarking.
Summary updated