A Machine Learning Model on Simple Features for CO₂ Reduction Electrocatalysts

Abstract

Electroreduction of CO2 is one of the most potential ways to realize CO2 recycle and energy regeneration. The key to promoting this technology is the development of high-performance electrocatalysts. Generally, high-throughput computational screening contributes a lot to materials innovation, but still consumes much time and resource. To achieve efficient exploration of electrocatalysts for CO2 reduction, we created a machine learning model based on an extreme gradient boosting regression (XGBR) algorithm and simple features. Our screening model successfully and rapidly predicted the Gibbs free energy change of CO adsorption (ΔG CO) of 1060 atomically dispersed metal–nonmetal codoped graphene systems, and greatly reduced the research cost. The competitive reaction, the hydrogen evolution reaction (HER), is also discussed with respect to such a screening model. This work demonstrates the potential of machine learning methods and provides a convenient approach for the effective theoretical design of electrocatalysts for CO2 reduction.

Origin of the N-coordinated single-atom Ni sites in heterogeneous electrocatalysts for CO2 reduction reaction

“…33,34 It has been established that the local coordination environment of a single-atom metal site is essential for determining its catalytic performance. 35,36 We began by constructing the potential congurations of single-atom Ni sites and evaluating their formation energy values. For the local structure of the Ni sites of Ni-N-C SACs, the central Ni atom adopts a (quasi-)planar tetracoordination conguration, and as mentioned earlier, the ligand species can be either pyridinic N, pyrrolic N, or C. Interestingly, the prototypical a-NiN x C 4Àx can be conveniently constructed by creating a double-vacancy in the basal plane of graphene and replacing certain C atoms with N atoms (Fig.…”

Applying Machine Learning to Rechargeable Batteries: From the Microscale to the Macroscale

“…It is,t herefore,q uite convenient to construct al arge dataset containing the energetic information of ac lass of functional materials by DFT calculations,w here ML models can then be adopted to construct as caling relationship between the structural and energetic properties of the materials.T his scheme is widely applied in designing electrochemical catalysts [43] for the hydrogen evolution reaction, [44] oxygen evolution/reduction reaction, [45] nitrogen reduction reaction, [46] and carbon dioxide reduction reaction. [47] As ar esult, ML is effective in promoting the development of metal-air batteries.Asimilar paradigm has been established with Li-S batteries for predicting the binding energies between lithium polysulfides and cathode hosts. [48] It should be noted that the choice of descriptors for the materials and the construction of ML models or algorithms are equally important in such ML studies.…”

“…It is, therefore, quite convenient to construct a large dataset containing the energetic information of a class of functional materials by DFT calculations, where ML models can then be adopted to construct a scaling relationship between the structural and energetic properties of the materials. This scheme is widely applied in designing electrochemical catalysts [43] for the hydrogen evolution reaction, [44] oxygen evolution/reduction reaction, [45] nitrogen reduction reaction, [46] and carbon dioxide reduction reaction [47] . As a result, ML is effective in promoting the development of metal–air batteries.…”

Harnessing the power of machine learning for carbon capture, utilisation, and storage (CCUS) – a state-of-the-art review

“…Wan et al 268 also proved that GBR model exhibited the best prediction performance to select the superior electrocatalysts for CO 2 reduction. Moreover, Chen et al 269 ) for new MOFs catalysts for carbon dioxide fixation. In addition, biological fixation is also an attractive method to convert CO 2 into organic compounds by using organisms such as microalgae.…”