HOUSTON – Scientists are discovering new organic and inorganic materials at a rapid pace, but progress in discovering hybrid organic-inorganic materials has been very slow due to the complex and oftentimes unknown interactions of these two different classes of materials at their interface. C-Crete Technologies recently discovered a way to create novel hybrid materials quickly by feeding key quantum mechanics calculations on representative material motifs to deep-learning agents, which use layered machine learning based on artificial neural networks.
These deep-learning agents can be used to uncover the complex role of interfacial chemistry. Better yet, they can quickly predict numerous complex motifs to facilitate the design of hybrid materials with the properties desired by a particular industry for a specific use. And, this deep learning offers a significant speed boost over traditional quantum calculations of such properties, allowing calculations that now take days of supercomputer time to run in hours.
One such example is coordination polymers. These polymers adsorbed on quartz or graphene substrates as hybrid materials hold great promise in the design of novel nanostructured materials with applications in gas storage, catalysis, high-density data storage, processing devices, selective ion exchange, the encoding of molecular information to produce biological function, among others. C-Crete’s new method offers a way to create coordination polymers on graphene or quartz with the properties applicable for these uses.
When a polymer and graphene or quartz are brought together as semiconductor heterostructures, a host of phenomena can occur at their interface. The major effect of these phenomena is to break the symmetry at the interface, which leads to a modification of the electronic and structural properties of the substrate. This can create a new class of hybrid materials with specific electronic surface states, something that is unattainable using conventional semiconductors.
C-Crete considered 244 basic material motifs for coordination polymers, comprised of methylene and Group IV halides such as silicon difluoride, silicon dichloride, germanium difluoride, germanium dichloride, stannous fluoride, and stannous chloride. C-Crete screened key properties of more than 120,000 hybrid materials using these material motifs adsorbed on graphene or a quartz layer. The properties included adsorption energy, charge transfer, dipole moment, energy band gap, structural deformation, and interfacial pressure. In the example of methylene, setting all building blocks in the chain to methylene leads to polyethylene, a common insulating polymer.
“One promising application in electronics is in the search for semiconductor heterostructures applicable in perovskite solar cells, p–n junctions and diodes, which require a system with high charge transfers between the adsorbate and substrate,” says Rouzbeh Shahsavari, president of C-Crete Technologies. “Our deep-learning agent rapidly showed that the systems that contain two or more connected germanium difluoride and stannous fluoride on a quartz substrate are excellent candidates for this purpose.”
“Another example is in a search for a chain polymer adsorbed on graphene with large pressure and charge transfer that would lead to systems with one silicon at the starting point of the polymer chain and methylene at the middle of chain,” says Shasavari. “This is nonintuitive, and can have an important impact on utilizing the noncovalent intermolecular interactions observed in many nanoscale electronic devices.”
C-Crete’s deep-learning method for the discovery of hybrid materials could open up a new framework to understand the behavior of different classes of materials, unleash their nonintuitive commonalities and anomalies, and eventually provide for improved design of these materials for tailored applications.
A paper about this work led by Rouzbeh Shahsavari of C-Crete Technologies was published in the July 23, 2021 issue of Nature Scientific Reports
Deep Learning Method to Accelerate Discovery of Hybrid Polymer‑Graphene Composites
Article Publication Date