Superior Prediction of Graphene Nanoflake Properties with Unbiased Graph Embedding

In order to apply machine learning to study the structure-property relationships of particular nanomaterials domain knowledge is typically required for feature extraction. However, this process may introduce bias if there is a focus on known aspects of structure, impeding the discovery of new science. Here, we develop an approach that uses only atomic Cartesian coordinates to predict the electron affinities, band gap energies, Fermi energies and ionization potentials of simulated graphene nanoflakes from a pubically available data set [1]. The workflow developed represents nanoflakes with graphs that are more representative than the ball-stick atom-bond representation that is intuitive to humans and generates fixed-size embeddings of these graphs using the neural embedding framework graph2vec [2]. Pairing the graph embeddings with a convolutional neural network produced highly accurate predictive models with hold out test set R2 from 0.9 − 0.96 for nanoflakes with a very challenging variation in size from tens to thousands of atoms. These predictions were benchmarked against results for optimised predictive models with geometric domain-driven features [3] exceeded their model accuracy for predictions of Fermi energy, electron affinity and ionisation potential and met their model accuracy for band gap energy.