Skip to main content
Research Highlight

Prediction of Carbon Dioxide Adsorption via Deep Learning

Topic: Materials

Scientific Achievement

A new method was developed for the discovery of fundamental descriptors for gas adsorption through deep learning neural network (DNN) approach. This approach has great potential to identify structural parameters for gas adsorption.

Significance and Impact

This work demonstrated that mesopores play a more important role on CO2 adsorption under high pressure and surface area is an independent textural parameter that can be synergistically coupled with other textural parameters in determining gas-solid interactions and thus gas-uptake capacities.

Research Details

  • Porous carbons with different textural properties exhibit great differences in CO2 adsorption capacity but it is unclear what role each textural variable plays in CO2 adsorption.
  • We use a random 1000 samples as the training data to train a DNN with two hidden layers, and 20 data samples for cross-validation (prediction). Additionally, we use the leave-k-out method to check our model accuracy with the random 1000 training samples.
  • The surface area, micropore volume, mesopore volume, adsorption temperature and pressure were considered as five neurons in the input layer.
Z. Zhang, J. Schott, M. Liu, H. Chen, X. Lu, B. Sumpter, J. Fu and S. Dai Prediction of Carbon Dioxide Adsorption via Deep Learning. Angew. Chem. Int. Ed. 2019, 58, 259-263. DOI: 10.1002/anie.201812363