Fig. 4: Confirmation of top-scoring RL-generated radicals with DFT calculations. | Nature Machine Intelligence

Fig. 4: Confirmation of top-scoring RL-generated radicals with DFT calculations.

From: Multi-objective goal-directed optimization of de novo stable organic radicals for aqueous redox flow batteries

Fig. 4

a, Distribution of redox potentials for radicals in the training database, highlighting the target zone for an aqueous, symmetric redox battery. Predicted redox potentials for optimized radicals are shown in blue. b, DFT-computed redox properties for RL-optimized radicals that were predicted to fall within the target zone. EA, electron affinity; IE, ionization energy. c, A comparison of ML- and DFT-derived stability estimates for optimized radicals, compared with the distribution of stability scores for radicals in the training database. A stability threshold of 90 was used as a lower bound for determining whether a radical could be classified as stable.

Back to article page